Physics of The Electric Guitar - Manfred Zollner

0.
Getting into the mood (Satire) 0-01

0.1 Barking up the wrong tree 0-02
0.2 Signature guitars 0-11
0.3 Strings and the orcus (cf. Schiller) 0-15
0.4 Unamplified E-guitars 0-17
0.5 Noehle-Gluehstrumpf 0-23
1. The fundamentals of string oscillations 1-01

1.1 Transversal waves 1-01
1.2 Wound strings 1-05
1.3 Inharmonic partials 1-10
1.3.1 Dispersion in the time domain 1-10
1.3.2 Dispersion in the frequency domain 1-18
1.4 Longitudinal waves / dilatational waves 1-20
1.5 Plucking a string 1-27
1.5.1 Deconvolution of dispersion 1-27
1.5.2 Plectrum influence 1-31
1.5.3 String bounce 1-35
1.5.4 String rattle 1-41
1.6 The decay of string oscillations 1-42
1.6.1 Plane string oscillations 1-42
1.6.2 Spatial string oscillations 1-46
1.6.3 Partial level and summation level 1-51
1.6.4 Worn out strings 1-53
1.7 String lifetime 1-54
2. The string as a waveguide 2-01

2.2 Image waves as a model for reflection 2-11
2.3 Standing waves 2-12
2.4 Transient phenomena 2-18
2.5 Reflection 2-20
2.5.1 Reflection factor 2-20
2.5.2 A resonator as string bearing 2-23
2.6 Internal dissipations 2-24
2.7 Dispersive bending waves 2-26
2.8 The generalized waveguide model 2-36
2.8.1 Ideal string, bridge pickup 2-36
2.8.2 String with singlecoil pickup 2-41
2.8.3 String with humbucking pickup 2-47
2.8.4 Dispersive waveguide components 2-50
2.9 Magnetic pickup with dilatational waves 2-52
3. String magnetics 3-01

3.1 Steel, nickel, bronze 3-01
3.2 String loudness 3-04
3.3 Magnetic string parameter 3-08
3.3.1 Measurements with a string loop 3-08
3.3.2 The magnetic skin effect 3-10
3.3.3 Measurements with a yoke 3-12
4. The electromagnetic field 4-01
4.1 Fundamentals of magnetostatics 4-02
4.2 The magnetic potentials 4-07
4.3 Materials in the magnetic field 4-10
4.3.1 Soft magnetic materials 4-14
4.3.2 Hard magnetic materials 4-14
4.3.3 Nonmagnetic materials 4-14
4.4 Pickup magnets 4-15
4.4.1 Alnico magnets 4-16
4.4.1.1 Alnico-III and Alnico-I 4-22
4.4.1.2 Alnico-II 4-23
4.4.1.5 Alnico-V 4-24
4.4.1.6 Additional Alnico materials 4-25
4.4.1.7 Comparison of Alnico materials 4-25
4.4.2 Cunife magnets 4-30
4.4.3 Ferrite magnets 4-31
4.5 Magnetic aging 4-32
4.6 The magnetic circuit 4-35
4.7 Depiction of magnetic fields 4-40
4.7.1 Magnetic field strength and flux density 4-41
4.7.2 Magnetic potentials 4-44
4.7.3 Spatial fields 4-45
4.8 Field geometry inside materials 4-46
4.9 Mathematic field theory 4-50
4.10 Magnetodynamics 4-57
4.10.1 Magnetic voltage induction 4-57
4.10.2 Self induction, inductivity 4-58
4.10.3 Permeability 4-61
4.10.4 Magnetic losses, magnetic skin effect 4-64
4.11 Magnetic field forces 4-72
4.11.1 Maxwell force 4-72
4.11.2 Field related pitch modulations 4-74
4.11.3 Field related level modulations 4-77
4.11.4 Field related dissipations 4-81
4.11.5 Indirect sound effects 4-85
4.11.6 Coulomb force 4-87
4.11.7 Lorentz force 4-87
4.12 Magnetic figures of merit (table) 4-88
5. Magnetic pickups 5-01

5.1 Singlecoil pickups 5-01
5.2 Humbucking pickups 5-09
5.3 Coaxial singlecoil pickups 5-13
5.4 The pickup's magnetic field 5-17
5.4.1 The static field without string 5-17
5.4.2 The static field with string 5-25
5.4.3 The alternating magnetic field 5-29
5.4.4 Magnetic window (aperture) 5-34
5.4.5 Absolute pickup sensitivity 5-43
5.4.6 Staggered and beveled polepieces 5-45
5.4.7 Fender Jaguar and Lace 5-48
5,4,8 DeArmond Pickups 5-52
5.5 Basic pickup parameters 5-56
5.5.1 DC resistance 5-56
5.5.2 Coil inductance 5-58
5.5.3 Coil capacity 5-61
5.5.4 Resonance quality 5-62
5.5.5 Polarity 5-64
5.5.6 Time variance 5-69
5.5.7 Wire coating, wax 5-70
5.5.8 Flatware 5-72
5.5.9 Absolute sensitivity, loudness 5-73
5.6 Pickup measurement devices 5-75
5.7 Hum sensitivity 5-79
5.8 Nonlinear distortions 5-86
5.9 Equivalent networks 5-91
5.9.1 Models and analogies 5-91
5.9.2 Impedance models 5-92
5.9.2.1 Singlecoils with low eddy current losses 5-94
5.9.2.2 Eddy currents in the nonmagnetic conductor 5-97
5.9.2.3 Equivalent two-pole networks 5-101
5.9.2.4 Eddy currents in the magnetic conductor 5-103
5.9.2.5 Singlecoils with high eddy current losses 5-106
5.9.2.6 Gibson-Humbucker: screw-coil 5-109
5.9.2.7 Gibson-Humbucker: plug-coil 5-118
5.9.2.8 Gibson-Humbucker: coupling of the coils 5-120
5.9.3 Equivalent transmission networks 5-122
5.9.4 Connected pickups 5-126
5.10 Analysis of the transfer behavior 5-129
5.10.1 Measurements with a shaker 5-129
5.10.2 Measurements with a Helmholtz coil 5-131
5.10.3 Measurements with a coaxial coil 5-133
5.10.4 Measurements with a tripole coil 5-134
5.10.5 Measurements with a laser-vibrometer 5-135
5.10.6 Measurement accuracy 5-145
5.10.7 Finite Element Modeling 5-148
5.11 Directional characteristic of pickups 5-150
5.11.1 String polarization 5-150
5.11.2 Wave polarization 5-153
5.12 Pickup noise 5-155
5.13 Pickup microphonics 5-157
5.14 Pickups with shorted turns 5-163
5.15 Database 5-167
5.16 Patents and Inventions 5-206
6. Piezoelectric pickups 6-01

6.1 The piezoelectric process 6-01
6.2 Electric loading 6-03
6.3 The piezo pickup as a sensor 6-04
6.4 Reciprocity 6-08
6.5 The piezo pickup as an actuator 6-11
6.6 The disassembled pickup 6-14
6.8 Piezo pickup vs. microphone 6-17
6.9 Microphonics 6-19
6.10 Differences compared to magnetic pickups 6-21
6.A Supplement: Piezoelectric state equations 6-23
7. Neck and body 7-01
7.1 The guitar neck 7-01
7.2 The frets 7-04
7.2.1 Fret positions 7-04
7.2.3 Fret materials 7-10
7.2.3 The Buzz-Feiten-system 7-11
7.3 Neck and string geometry 7-15
7.3.1 Head and neck angle 7-15
7.3.2 String trees 7-17
7.4 String dynamics 7-18
7.4.1 Playing forces 7-18
7.4.2 Bearing forces 7-21
7.5 Reflection and absorption at the bridge/nut 7-25
7.5.1 Reflection and absorption parameter 7-26
7.5.2 Reflection analysis 7-27
7.5.3 The mechanical bridge impedance 7-39
7.5.4 Measurement results 7-45
7.6 Vibration measurement techniques 7-51
7.6.1 Impedance / admittance measurements 7-51
7.6.2 The spectrum of decaying tones (Volagramm) 7-56
7.6.3 The decay time T30 7-64
7.7 Absorption of string oscillations 7-66
7.7.1 Radiation absorption 7-66
7.7.2 Internal dissipation 7-67
7.7.3 Winding attenuation of wound strings 7-69
7.7.4 Bearing absorption 7-71
7.7.4.1 Coupling of transversal waves 7-71
7.7.4.2 Absorption of longitudinal waves 7-75
7.7.4.3 Residual absorption 7-76
7.7.4.4 Bearing conductance 7-77
7.7.5 Finger-, hand- and capodaster-attenuation 7-84
7.7.6 String aging 7-85
7.7.7 Flatwound strings 7-86
7.8 "Specialized" literature 7-87
7.8.1 The fairytales of the primary tone 7-88
7.8.2 "Stratone" 7-92
7.8.3 BS-Journalists 7-100
7.9 Does the body wood affect the tone? 7-102
7.10 Special bridge constructions 7-117
7.10.1 Simple models 7-118
7.10.2 Bridges without vibrato (Gibson / Fender) 7-122
7.10.3 Bridges with vibrato (Fender / Bigsby / Rickenbacker) 7-132
7.11 Solid vs. semisolid body 7-139
7.12 Vibration – soundwaves – sound 7-142
7.12.1 Linear string oscillations 7-142
7.12.2 Nonlinear string oscillations 7-152
7.12.3 The causes of timbre 7-161
7.12.4 So what? 7-164
7.13 Neck curvature and fret/string distance ("action") 7-165
7.14 Damping reduction 7-170
8. Psychoacoustics 8-01
8.1 Tone systems 8-02
8.1.1 Das Pythagorean tone system 8-03
8.1.2 Just intonation 8-07
8.1.3 Tempered intonation 8-10
8.1.4 Equal tempered intervals 8-13
8.1.5 Typical guitar mistuning 8-16
8.1.6 The stretched intonation 8-16
8.2 Frequency and pitch 8-17
8.2.1 Frequency measurement 8-17
8.2.2 Frequency and pitch accuracy 8-19
8.2.3 Pitch detection 8-23
8.2.4 Grouping of partials 8-25
8.2.5 Inharmonicity of partials 8-28
8.3 The character of keys 8-37
8.4 Consonance and dissonance 8-40
8.5 Timing and rhythm 8-47
8.6 Loudness and timbre 8-54
8.7 Listening tests 8-71
8.7.1 Psychometry 8-71
8.7.2 The unamplified E-Guitar 8-76
8.7.3 Tactile sensations 8-82
9. Guitar Circuits 9-01

9.1 Potentiometers 9-01
9.2 Tone-Caps 9-05
9.3 Pickup cables 9-09
9.4 Guitar cables 9-10
9.5 Metal sheets 9-15
10. Guitar amplifiers 10-01

10.1 The input stage 10-01
10.1.1 The input tube 10-02
10.1.2 The tube's input resistance 10-04
10.1.3 Triode charts 10-07
10.1.4 Nonlinearities, distortion 10-11
10.1.5 Cutoff frequencies 10-24
10.1.6 Time variances 10-28
10.1.7 Noise, hum, microphonics 10-30
10.1.8 Noise processes 10-32
10.1.9 Pentodes in the input stage 10-35
10.2 The second stage 10-36
10.2.1 Cathode-basis-circuit 10-37
10.2.2 Cathode follower 10-38
10.2.3 The mixing stage 10-46
10.3 The filter section (Tone Stack) 10-49
10.3.1 Bass-Middle-Treble 10-49
10.3.2 Equalizer 10-58
10.3.3 Presence-Control 10-61
10.4 Phase reversal (Phase Splitter) 10-62
10.4.1 Paraphase 10-62
10.4.2 Kathodyn 10-65
10.4.3 Difference amplifier 10-67
10.4.4 Halfewave antimetry 10-69
10.5 The power stage 10-75
10.5.1 Class-A, tetrode, pentode 10-76
10.5.2 Class-A push pull 10-85
10.5.3 Class-B 10-87
10.5.4 Class-AB, class-D 10-90
10.5.5 The impedance paradox 10-91
10.5.6 Negative feedback 10-92
10.5.7 The source resistance of the power stage 10-94
10.5.8 Biasing the power stage 10-98
10.5.9 Stress and aging 10-107
10.5.10 The magic sound of a 6L6 10-115
10.5.11 Match Point 10-119
10.5.11.1 Selecting, matching (and leg pulling) 10-119
10.5.11.2 Tube testing 10-122
10.5.12 Selected tube circuits VOX, Marshall, Fender 10-124
10.5.13 Comparing analysis: Power tubes 10-144
10.5.14 Pentode / triode / ultralinear 10-176
10.6 The output transformer 10-157
10.6.1 The linear model 10-157
10.6.2 Impedance matching 10-162
10.6.3 Winding capacitance 10-164
10.6.4 The nonlinear model 10-166
10.6.5 Comparing analyses 10-176
10.7 Power supply 10-188
10.7.1 Heating circuit 10-188
10.7.2 Filter capacitor 10-189
10.7.3 The internal resistance 10-193
10.7.4 Rectifier tubes 10-194
10.7.5 The smoothing filter 10-195
10.7.6 The mains transformer 10-196
10.8 Effects 10-204
10.8.1 Reverb 10-204
10.8.2 Vibrato / Tremolo 10-212
10.8.3 Phaser / Flanger / Chorus 10-218
10.8.4 Wah-wah-pedal 10-220
10.8.5 Fuzz-box 10-221
10.8.5.1 Diodes 10-227
10.8.5.2 Transistors 10-230
10.8.5.3 Range Master (Dallas Arbiter) 10-232
10.8.5.4 Tube-Screamer (Ibanez) 10-235
10.8.5.5 Fuzz-Face (Dallas Arbiter) 10-237
10.8.5.6 Roaring semiconductors 10-238
10.9 Operational behavior 10-239
10.8.1 Tube-sound vs. transistor-sound 10-239
10.8.2 Tube-Watt vs. transistor-Watt 10-244
10.8.3 Coupling capacitors 10-249
10.8.4 Sound event vs. listening event 10-267
10.10 Comparing analyses 10-271
10.10.1 Right you are if you think you are 10-271
10.10.2 Stage topology 10-275
10.10.3 Headroom charts 10-278
10.10.4 Comparison of nonlinear distortions 10-285
10.10.5 Audibility of nonlinear distortion 10-290
10.10.6 Comparison of Frequency responses 10-296
10.10.7 Comparison of VOX, Fender, Marshall 10-299
10.10.8 Modeling amps 10-316
10.11 Tube data 10-325
10.11.1 Nomenclature 10-325
10.11.2 Triodes 10-326
10.11.3 Power tubes 10-330
10.11.4 Tube parameters 10-338
11. Guitar loudspeakers 11-01

11.1 Construction and function 11-01
11.2 Electrical two-pole characteristic 11-07
11.3 Frequency response 11-11
11.4 Directional characteristic 11-29
11.5 Efficiency and maximum sound pressure 11-38
11.6 Nonlinear distortions 11-52
11.7 Alnico- vs. ferrite magnet 11-64
11.8 Loudspeaker cabinets 11-71
11.8.1 Basics 11-71
11.8.2 Comparison of cabinet materials 11-80
11.9 Beamblockers and Diffusors 11-86
11.10 Horn loudspeaker 11-91
11.11 Studio monitors 11-96
11.12 Loudspeaker cables 11-100
11.A Supplement: Measurement techniques 11-101
11.A.1 Measuring microphones 11-101
11.A.2 Reverberation time 11-101
Supplement: Vibration and waves A-01

A.1 Oscillations vs. waves A-01
A.1.1 Forced oscillations A-02
A.1.2 Free oscillations A-02
A.1.3 Forced waves A-03
A.1.4 Free waves A-05
A.1.5 Standing waves A-05
A.2 Longitudinal waves A-07
A.2.1 Pure longitudinal waves A-07
A.2.2 Dilatational waves in strings A-07
A.3 Transversal waves A-10
A.3.1 Pure transversal waves A-10
A.3.2 Transversal waves in strings A-10
A.4 Bending waves A-12
A.4.1 Bars under zero tension, pure bending waves A-12
A.4.2 Vibrations of a stiff string A-16
A.4.3 Eigenmodes of bending waves A-17
A.5 Wave resistance A-21
A.6 Stiffness A-25
A.7 Impulses A-27
A.8 Ultimate end: cryo… A-28
References
Glossary
Animations (see www.gitec-forum.de, no download possible)

Some Words from the (main) Translator
Phewwww ...!!! What a labor of love this has been! Four years in part dominated by trying to
find the right words and expressions ... not just technical jargon, not just precise scientific
equivalences, but also doing justice to the sometimes highly personal style of the author ...
Here's how this translation of the German "Physik der Elektrogitarre" into the English
"Physics of the Electric Guitar" came about. There's quite a bit of back-story:
From a very early age my life has been connected to science, electrical engineering, and
electro- and psychoacoustics - and to music. My grandfather was a violinist with a permanent
position in the orchestra of the opera in Stuttgart, Germany. My dad, Eberhard Zwicker, was a
high-calibre scientist: an electrical engineer and electro-acoustician by trade who had
specialized in Psychoacoustics. As one element in a group of scientists around the globe who
investigated the fundamentals of the hearing system, his findings – together with those of
others – laid the groundwork for many practical applications. Examples are diagnostic
systems to check the hearing system of newborn and very young children, or the development
of systems that could measure the actual loudness of sounds as it is perceived by us. Not least,
the development of the data-compression algorithms behind the now ubiquitous mp3-music-
format is another result of that psychoacoustical research.
I had been playing the guitar for a couple of years when I first met Manfred Zollner in the mid
1970's at a family-party of the members of my dad's team of scientists who worked with him
at his lab at the Technical University in Munich. Manfred had just gotten his diploma in
electrical engineering and was starting in a position as an assistant professor at the lab, also
working on his PhD – but what fascinated me much more was the Gibson ES-335 he had
brought along to the party ... an instrument that I had been dreaming of! What Manfred had
with him also was a little battery-powered amp that he had modded to achieve controllable,
smooth distortion. With his playing being light-years ahead of mine at the time, the sound he
got that night at the campfire was terrific, and I of course clobbered him completely with lots
of questions about guitars and amps. He seemed to know everything, and that he worked on
the side as a techie at the music shop my friends and I frequented only added to my
fascination. Clearly, the man was brilliant in a number of ways - and he was also quite a
character.
Over the years, I could stay in touch with Manfred. As can be seen in his introduction on the
"Board of GITEC"-page (https://www.gitec-forum-eng.de/landing-page-news/vorstand/) his
path led him (after obtaining his PhD) first to the music equipment industry, and then to his
own company developing and manufacturing precision instrumentation of acoustical
measurements. Subsequently, he held a university professors chair for many years until his
retirement. I retired myself in early 2016, and around that time the contact to Manfred
intensified again - because he had founded GITEC, and I was of course EXTREMELY
interested. I had the privilege of being invited to join the GITEC board in late 2015 (where I
served until late 2019).
Manfred had published his book "Physik der Elektrogitarre" in 2014, and it was clear to me
that this was – in its approach, depth, and style – a most unique, unparalleled oeuvre ... that
merited the largest possible readership. The latter, however, would remain rather limited
because of the use of the German language. In a number of board meetings, I urgently pushed
the idea that it was necessary to offer global access to the book by translating it into English.
Some Words from the (main) Translator © Tilmann Zwicker, 2020 1
The board completely agreed. We were acutely aware, though, that there were serious
challenges. Manfred had in fact already earlier investigated possibilities - and while there
seemed to be some people around that could do the job, the cost seemed prohibitive in view of
the finances available to GITEC. Because we still wanted to push the issue, we decided to
jointly put some private money on the table - and started negotiations with translators.
At the same time I started the project of introducing an English version of the GITEC website.
In the corresponding framework, I started to translate some of the articles that had been
published on the German-language GITEC-site. After all: despite not being a native English
(or American) speaker, I had lived in the US for a number of years, I had worked at the
European Patent Office (where English is one of the three official languages) for the better
part of my professional life, and I had good technical insight – so I felt competent to do that
kind of work. However, from the book translation I did shy away ... simply because of the
sheer amount of word involved in that task.
Over the course of half a year we checked out some translators who seemed able to work on
the book but found the results of some test translations not satisfactorily at all. Moreover,
most of the translators we contacted showed merely a lackluster interest ... simply because a
volume of around 1300 pages of material that combines extremely serious science,
pronounced specialization in music-equipment, quite specific satire and not infrequent acidic
comments is not what's included in the business model of your regular translator. In mid
2016, we were quite frustrated ... and I decided to do what I had sought to avoid before: offer
myself as a translator. Putting the articles into English had proven to be a very enjoyable
experience that also helped me keep my command of the language at a good level. So maybe
a VERY long-term project of translating the book would not be impossible for me. On top of
that, I had some insights into the kind of character the author was, and I felt I could find the
right kind of style that would represent him well.
My good pal and (at the time) co-board-member Wolfgang Hönlein was sufficiently
encouraged by my willingness to start actual work on the project to indicate that, with the
assistance of a British friend of his, he would contribute. We also considered that we could
invite other contributors to possibly form a kind of "cloud-translation" project that I could
"oversee".
There was indeed much enthusiasm among a number of people who heard about the
translation project - however, none of them had really much (or any) experience with
translation work. Some of the folks tried out what it was like ... but realized very quickly that
this required too much of an effort, or that their skills were not at the level required. On the
other hand, it also became clear to me that coordinating and checking the results of a "crowd-
translation" project would not be much less work than doing the translation myself. In the
context of the crowd-translation experiment, I am indebted to Franz Wolter for his
considerable efforts trying to make a contribution.
In the end, Wolfgang (with the help of his friend Andrew Graham) did supply the translation
of the whole Chapter 4, and GITEC-friend Volker Eichhorst (with help of his friend Gabriel
Mallory) saw to it that sub-Chapter 7.7 and the Supplement got translated. As a "collateral
benefit" from trying out translating, Ralf Jamer helped with translating the article
"Overdrive, Fuzz & Distortion" in the GITEC website's knowledge-base.

The rest ... that was up to me to take care of, and although I am almost ecstatic to be able to
state now (in autumn 2020) that the translation work is completed, I am quite sure that I will
miss the work in a way. I have learned – in more ways than one! – a lot in the framework of
doing the translation, and it certainly was much fun. For better or worse, GITEC has
generated a lot of material on top of the book (and keeps doing so), so that I have an ample
supply of text in case there are any severe withdrawal symptoms ... after all this time of
dedication to pushing "Physik der Elektrogitarre" into the English-language realm.
Here, I would like to thank all those who have supported me in this endeavor
- first and foremost the Professor M.Z. himself: he read all the translated material with great
patience, and with the precision that distinguishes all his work, and gave important feedback.
- Wolfgang Hönlein for continued encouragement, discussions, and input, and his wife
Brigitte for a never-ending supply of "tea and sympathy" (or, rather, coffee and biscuits, and
lunches!) during these discussions,
- the board and the consulting friends of GITEC,
- the special support, interest, feedback, and encouragement of
- Andrew Flanders in the US,
- Tim Wrigley in Australia
- Doug and Robbie Laughlen in the UK
- Bertrand Dauvergne for much encouragement and helping me with the "rather different"
Chapter 0 (on top of being a great musical collaborator!).
- Elizabeth Corcos for constant encouragement and support, and for helping me to maintain a
good level of (spoken) English,
- last but absolutely not least my lovely Sabine (who must have anticipated my retirement
with rather different expectations about my activities) for the always patient and loving
support.
I hope many readers get something out of this translation.
Be well and take care - live long and prosper,

Tilmann

Foreword by Paco Beslmeisl (translated by Dr. T)
Hi there … or rather: “Grüß Gott” – as we say in the south of Germany!
Yeah, well – ‘course I’ll happily write up a few words regarding Physics of the Electric
Guitar … I own a few of those, after all, and I’ve got a load of experiences with reviews, be it
rave or otherwise (you gotta endure a lotta tuff stuff as a guitar player). I don’t know much
about the technical side (beats me why an inductee plays a part of that whole R’n’R-Hall-of-
Fame shebang while an inductor can’t), but if Billy Gibbons can write the foreword to the
Marshall book … hey, I know three chord shapes, as well. Some time back I always thought:
Gibbons, ah, that’s them monkeys – but then he’s a guitarist, too. And one who – unless his
beard gets entangled in the strings – will play like the devil. In terms of Lone-Star-State
standards, anyway – y’all know what I mean. So what does one write in a foreword, actually?
Billy writes in English … that really pushes the envelope for me: I’m a Bavarian, born and
raised … and I can scrounge up some bits and pieces of Saxon speak (that’s always a hit for
in-between-tunes babble when we play the beer tents in Upper Bavaria … even the trashed
guys under the tables who can’t hold their liquor will at least hold their sides cracking up). I
trust that my friend Dr. T. will do a reasonable job with the translation into English, or
American, rather – I always left that language to the singer in our band. When I drive him
home after the gig (the man had to hand in his license three years ago … DUI), he always
sings FERNANDO by the ABBAs to me, and I try to hold my own throwing in some Kraftwerk
tunes … stuff one sings when driving home (on the AUTOBAHN). I won’t go into detail
regarding his tales regarding breaks between sets ‘cause this here should somehow be about
the guitar. In fact, they built me a signature model. I like the French ladies, so I wanted to call
the color “vert prinantier” but they rudely changed that. Allegedly English sounds better, so:
faded vomit green. OK by me, then. I don’t really care what I play as long as its old ‘cause:
the new stuff doesn’t have that kind of vib-e-ration. One time, I put a body (of a guitar,
silly!!) in the freezer for two weeks, to do that cryo-tuning – heard about it from my pals at
the “Six ‘n’ Four String Slinger” mag. That really sucked, though, can’t recommend it at all.
Not with my freezer, anyway. The body (of the guitar – how often do I need to emphasize
that!) was too wide by a few 16ths of an inch so the door didn’t fully close … which I didn’t
register. When I returned (six gigs in Switzerland; played everything slower by 20 BPM –
was cool!), an abominable stink met me already at the door. I thought: well, no ventilation for
two weeks (or longer, I’m not always on top of that game), but when I got to the kitchen I
nearly dropped dead. In front of the freezer, on the floor, everything covered in this green
glop with the tiny white hairs on top. Yuck! The stench was undescribably atrocious –
everything had thawed, some stuff had oozed out, mildew everywhere. The guitar body stunk
so much that I rushed to sell it to our rhythm guitarist Ernie. He said he’s not bothered, he’s
used to the smell from home – thank God he’s on the other side of the stage. I now have a real
aversion to cryo-tuning … you get me, don’t you?
So, this has turned out to be a really nice foreword, hasn’t it? I’m not getting any dough for it;
doin’ it for free. But maybe someone gets the idea to gift me with another signature guitar?
Billy Gibbons’ Pearly Gates, for example? To refer, at the end, back to the beginning –
learned that in essay-writing in school. Cool, ain’t it?!
Cordially and guitar-istically yours,

Paco Beslmeisl
Preface-2
Preface – by the author
The present book is the result of the lifelong practical and theoretical dealings I have had with
the electric guitar. The associated practical experiments started already in the 1960’s when
guitar-dominated, so-called “Beat-Music” got its global breakthrough. Looking back,
“Memphis Tennessee” was the initial spark that – barely ignited – found a propellant charge
in the tunes of the Beatles and the Rolling Stones that still continues to burn intensely to this
day. Subsequently, Eric Clapton became the big hero in terms of sound and style – and in fact
he still is, at least as far as his early years are concerned. My funding situation back in the day
required that amplifiers had to by built in DIY-fashion on a budget, with the ensuing insight
that teachings in school were not nearly comprehensive enough. This automatically led to
enrolling in the course for electrical engineering at the Technical University in Munich, with a
focus on electro-acoustics. It was here where the theoretical part started.
Particularly formative were the lectures by Hans Marko (systems theory), Rudolf Saal
(network theory), Eberhard Zwicker (acoustics), and Hans Meinke (radio frequency
technology). It may be surprising that RF-technology plays a decisive role in acoustics – but
Meinke’s theory of transmission lines, in combination with the theory of electro-acoustical
analogous networks, would prove to be ideal for the description of string vibrations. Founding
a company that designed and manufactured instrumentation equipment for measurement of
sound did temporarily lead to a banning of all guitars to the attic, but it did also make for the
emergence of precision instrumentation that formed the basis for the hardware in the lab later.
From 1990, more and more guitars succeeded to wander back to the basement and then up
into the beletage: the newly commenced work as a university professor (acoustics, signal- &
systems-theory) generated free time and aroused the curiosity how exactly these devices
operated. Uh-oh … that may sound a bit unsettling … let’s rephrase: the newly commenced
task as university lecturer led to such an intense workload that a balance was urgently
required, and that turned up in the form of various guitars. There was a curious inherent
proliferation process among the latter over the years – and they all wanted to be played and
analyzed. After several years of more sporadic experiments, systematic research on the
electric guitar set in from 1999, including written documentation.
Initially I had hoped that a few equations on vibrations, and some formula on the magnetic
field could adequately do the job of describing the topic. About 100 pages titled “How does
the electric guitar function?” emerged from that assumption. I then realized my very limited
understanding of just that functionality. This in fact is quite a good situation for any scientist.
As a model, the simple hypothesis of a transversally oscillating string with complex-valued
bearing impedances was of merely limited suitability. Considering the available literature, the
engineer in me tried to make do with it, but as a guitarist I found grave deficiencies. There
was, after all, an upside to having been able to study musical performance practice in the
clubs in Munich’s hip Schwabing quarter. That was back in the day as a student, when
conservative educators warned, with a wagging finger, about the amalgamation of Apollonian
and Dionysiac goings-on, and now it bore fruit. Still, the already mentioned transversal wave
remained in the foreground, but the revelation that a string on an electric guitar played by a
virtuoso will act out in a way entirely different from the teachings of all text books I knew –
that revelation slowly worked its ways from cortical depths into consciousness.
Unfortunately, 100 pages had already been written up.
Translation by Tilmann Zwicker © M. Zollner & T. Zwicker, 2014 & 2019

Preface-3
From now on, the musician prevailed over the engineer, and unmistakably articulated the
question why calculations were always done on the freely vibrating string, when that string
constantly hits the frets (as required by other regions in the brain during daily guitar playing).
Put another way: it is not entirely wrong to model the electric guitar as an LTI-system, but the
“sound” cannot be described that way. About at the same time of that recognition, irregular
occurrences in the spectrum of the strings showed that a non-negligible longitudinal wave
(dilatational wave) needed to be preset on top of the transversal movement of the string. This
longitudinal wave could at first be assumed only hypothetically but (from 2005) found
comprehensive confirmation via the use of a laser-vibrometer. By that, the book that at that
point had been considered to be very much on course, suddenly lost its whole structure.
Chapters and figures had to be repositioned, pagination did not fit anymore, whole passages
had to be completely re-written.
Today, more than 16 years have passed, and more than 1200 pages written for the book have
resulted. That extent had never been envisaged – but then, how far can research be planned?
Reading the first pages again after 16 years, ideas for a redesign pop up on impulse – but
giving in to them would preclude any finalization. Thus: over and done – that’s it!
The first two chapters deal with string vibrations. Much space is dedicated to dispersion,
because its effects on the harmonicity are considerable already in the middle frequency range
(see e.g. Fig. 1.11). The combination of the theory of electrical transmission lines and electro-
mechanical network-analogies facilitates a presentation of wave-propagation and -reflection
(Fig. 1.20) that is easy to grasp for the telecommunications engineer. Effects of the strings
hitting the frets or buzzing are discussed in Chapters 1.5.3 and 1.5.4 only in short – the
supplement is delivered in Chapter 7.12. The purpose of Chapter 2 is predominantly to
elucidate the real shape of the string vibration – sine-shaped partial-modes are of little help
here. By using a special presentation of the differential string-stiffness (Fig. 2.13), a way
could be found to visualized transversal movements without involving too much math.
Chapter 3 discusses the magnetic string parameters. It turned out shorter than originally
planned because the investigated strings barely differed in their magnetic parameters.
Chapters 4 and 5 describe magnetic pickups. Extensive and varied measurements

demonstrated a relatively simple correspondence between string velocity and pickup voltage.
This is a result that is well supported by theoretical considerations. The aperture-window
(Chapter 5.4) is merely 1 cm long and independent of the coil geometry; the latter, however,
influences the frequency response and the absolute sensitivity.
Relatively short, Chapter 6 is dedicated to piezo pickups. Combining quadripole-theory,

digital signal processing, and electro-acoustics (reciprocity, Chapters 6.4 and 6.5) transpired
to be particularly interesting.
The investigations into vibrations of guitar neck and guitar body (Chapter 7) started from the
premise that the resistive part of the string bearing (the so-called conductance) would deliver
the main contribution to the string damping, and thus the wood of the body would be essential
to the sound of the electric guitar (an assumption flogged to death by trivial “specialist”
literature month by month). Extensive investigations regarding the decay behavior of the
plucked string show, however, an entirely different result: for the solid-body guitar, the bridge
is yielding only at few frequencies to such an extent that the bridge-absorption gains
importance relative to the string-internal absorption. Moreover, this absorption can be mostly
traced to the bridge-design itself, and practically not at all to the wood used for the body.
© M. Zollner & T. Zwicker, 2014 & 2019 Translation by Tilmann Zwicker

Preface-4
Neck resonances are of a bit more significance; of particular relevance, however, is the upper
surface of the frets, because it determines whether and where the string bounces off the frets
(attack, snap).
Chapter 8 outlines the physical basics of prevailing tonal systems (tunings) and explains
some music-relevant essentials of psychoacoustics (e.g. spectral and virtual pitch, grouping of
partials, consonance/dissonance, timing and rhythm, timbre and loudness).
Chapter 9 is dedicated to the electric circuitry within the electric guitar. It was not planned to
be very extensive, because sufficient literature is already available on this topic [H. Lemme].
Chapter 10 (guitar amplifiers) and Chapter 11 (guitar loudspeakers) give information on the
electro-acoustic equipment. That in fact is a never-ending story – a topic that cannot be
comprehensively presented even with 440 pages. Even so, there is now an extensive
metrological analysis – supplementary listening experiments are desirable.
To make things somewhat less prosaic, Chapter 0 rises a bit beyond the world of physics.
Corresponding feedback has generally been very positive, it’s only Ms. Growse-Glowsock
who causes some agro.
As a last point, thanks should be given to the government of the State of Bavaria – with its
steady donations, it has provided a though small yet still important contribution to this
research project. My little acoustics lab and I were always full of joy when the dean declared
that, despite unavoidable funding cuts, another € 1200.- would be at our disposal. Not per
month, no, of course not. We do not see ourselves as an elite-cluster, my lab and I. We do
understand that € 1200.- per year can only be mustered because an elite-cabinet in the state of
Bavaria’s capital Munich spares neither trouble nor expenditures. And we understand that for
16 years we have been practically alone because permanent staff for the lab cannot be
financed. Having said that: my lab did have some trouble comprehending, and repeatedly
asked back, why another vice-president for the university had to be added to the first one if
there is as good as no money available. The additional VP requires (and gets!) staff, rooms,
and – again as my lab has found out – a new computer. To be honest, the thing with the
computer was difficult to communicate (to the lab) because we still have old NT-computers
hanging around. Some empathy could be created, after all, by the wisdom passed down from
generation to generation: that only that can grow that is present. What is not present (e.g.
sufficient permanent staff posts) does not grow. And so, dear acoustics lab, it is only too
reasonable that our university has by now received its third vice-president. L'enfer, c'est les
autres (Sartre)…
Regensburg, in autumn 2014
Manfred Zollner

0. Tuning-in (… not dropping out …) and getting into the groove
No - let’s not (yet) talk about the rhythm-“groove”. This short pre-chapter is not supposed to
correct the tuning of the guitar but to tune the reader to the called-for reading-groove …
tuning-in to electromechanical systems theory, to science – but also to fantastical blurbs.
Physics of the guitar – that is a wide subject (Too Far Afield, even?). There are non-linear
differential equations, time-variant systems, non-homogeneous anisotropic materials, spinodal
decompositions, diverging magnetic fields, and dispersive continuum-waves. Cast under a
proliferating cloud of wafting catch phrases that could be not more dopey, bogus, fallacious
and plain wrong. Undermined by self-proclaimed gurus who spam their unproven
assumptions with steady regularity into magazine columns. Outshone by the infallible
splendor of science that however prefers to bestow its affection onto those more noble
instruments, preferring to ponder the violin, the pianoforte, and the church organ – rather than
the armamentarium of Mr. James Marshall H. A science that will fastidiously check the
spelling of the name of famous Lord John Rayleigh in order to at all cost avoid any mix-up
with Sir Walter Raleigh, but is unable to distinguish between Jimi and Jimmy, just as it fails
to get right the difference between rock and pop♣.
So … yet another book about the electric guitar! That thing that the genius Segovia sought to
deny the designation “musical instrument”. That guitar “wired for sound”, somehow
operating with electrical current but still allegedly needing to “resonate” down into the very
last wood-fiber after each plucking of a string. This here is not going to be easy – not for the
author, not for the reader. Well then: if one does make an assertion regarding the effect of a
shielding pickup cover, then supporting it with good reasons should be mandatory. Three
purposeful reasons are: the physical/mathematical model, the results of measurements, and a
correspondence of the two. However, a physical/mathematical model requires a certain basic
knowledge in physics and mathematics – in fact that’s an enormous understatement because
in order to comprehend a coupling of modes, a good deal of specialist knowledge needs to be
present. Therefore this book “Physics of the Electric Guitar” has not turned out as a book that
will advise the musician which guitar to buy, but it is a documentation of years of research
work. Still, since the author is not your regular theory-dweeb, either, but a practicing guitarist,
the odd thought has made it directly from the left-hand part of the brain onto the paper, and
remains comprehensible without any grand education in math or physics. Or so the author
hopes, anyway! At least, these thoughts should not be any more cloudy than the allegation
that alder would result in both fat and subtle bass, and in both accentuated and mushy
articulation [guitar literature].
So: if you are not that much (or not at all) interested in formal-analytical description: do turn
the page(s) … more practically oriented passages and simplifying summaries always lie
ahead. It is the guitar that remains the topic of this book, and not theory for its own sake. For
the following pages, a few paragraphs from Chapters 7 and 8 shall be pulled ahead, to tune-in
without a lot of math. After that, the (science-) band begins to play … we’re gonna get down
to business.
♣
Memory hook for the gig: rock first, pop later!
0-2 0. Tuning-In & Getting into the Groove
0.1 Barking up some (wrong?) tree
Woodrow W. Worm, PhD, “Woody” to his friends, director of research at the guitar
manufacturer Tawdro, has kindly invited us (my photographer and me) to join him on a hike
as he inspects “his” woods; questions regarding wood in general, and regarding its sound
specifically, may be asked. So: "Dr. Worm, Tawdro is a well-known..."
"The globally operating guitar manufacturing enterprise Tawdro Inc. sells its world-
renowned guitars across all continents♣. We are a long-standing, tradition-minded business
that has remained under company ownership for 150 years. Uh … under family ownership …
I mean it’s owned … it belongs to the Tawdrant family. They originally hail from the eastern
parts of Germany and carried the name Drantow at the time. Carpenters by trade, they came
to the Home of the Brave on the early 1800’s. Their original name was misspelled so often
that it was changed to Tawdrant in the end.” “Aha! That’s the origin of the company name?”
"Precisely. From Roland Tawdrant, venerable founder of the company. However, Rotawd
would have sounded strange somehow. Hence: Tawdro.”
"Understood! Still, Dr. Worm, for a guitar, doesn’t Tawdro somehow sound … well …
there’s the association towards ‘tawdry’ …”
"I have no idea what you mean. In my dissertation about the third indo-germanic phonetic
change, I have established clear proof that ….” "There were no less than three of those?”
"Of course not! That’s exactly what I provided proof for! About the name: in the 17th century,
in the geographic East-German/Slawic context, 'dran’ incidentally had a very different
meaning that today would relate nicely to good guitars. The middle-high-German ‘trannck’ –
mutating via the early-Franconian ‘trann’ into the later “tranig” – originated from the
northeastern German ‘schtyrannckhaft’, as it was already shown in Mai 1956 by Nana
Tucketti Slay-Ryde and Johans Begoud Toonite in their reference book: De Thri-Teimes Fone
Tshanshe off Tschermanske-Indish ..." "Please, Dr. Worm – we wanted to discuss wood..."
"Oh yeah – right. These etymologic details will indeed concern only true specialists. In short:
they often changed name like that back in the day. Just think about Son Gibbo, Martinius
Frido Christophon, Peef Ehartla, or Fend Erleo, or Smitty ‘Rushes’ Paolo. Many company-
and brand-names came about that today globally command respect. In fact, my work with …”
"The wood, Dr. Worm, the wood …."
"Of course. Wood is the fundamental ingredient of the guitar vibration. That is why it is
THAT important, isn’t it? Without wood there is no vibration, no tone, no nothing, is there?
Wood – that’s the heart to the guitar. Not just the heart – it’s the soul. But that’s impossible to
convey to a technician. If a merchant offers me a batch of Honduras mahogany, I first smell
into every chink and grasp the olfactory overall composition. That’s like it is with music, or –
better – with wine! Your tongue has to shape up – you know what I mean? Oenophile?”
"I’m more into beer … so the wood determines the sound for the electric guitar, as well?”
"Certainly! Without wood there is no sound, no guitar! I shall demonstrate this with …”
Abruptly, Dr. Worm’s elaborations are interrupted: a specific tree absorbs his attention
completely and stops the lecture. Dr. Worm circles the tree, approaches it, walks away and
back, extends his hands, raises them, lowers them. No, that is no sudden attack of Qi-Gong –
we are privileged to witness a tree-claiming. Dr. Worm intones a slowly swelling vowel
similar to an “ommmmm” but breaking off after a few seconds with a loud "aikkk".
"Ommmm-aikkk, ommmm-aikkk!" Fascinating!
"Dr. Worm, sir, could you explain to us what …” "Silence – not now!"
Obviously, a tree-claiming must not be disturbed. Quietly, we wait in the background so as
not to again disrupt the events in such an unqualified fashion.
After several minutes, Dr. Worm disengages from the tree, approaches us and elucidates:
♣
Seminar for execs on marketing: “the first sentence is the most important one“.
Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

0.1 Barking up the (wrong?) Tree 0-3
"These force-fields – did you feel them, too? This will be true premium wood! Check the piles
over here – that’s it already drying.” He picked up a few of the piled-up branches, smelled
them, tapped his finger against them, and seemed to sense vibrations inaudible to us. “In
about 40 to 50 years, when it is well seasoned and dried, use that to build an electric guitar –
you will get a strong bass, loud low-mids, assertive high-mids, and a dominant treble.”
"That is a most interesting and obviously typical example. Couldn’t we also describe that kind
of sound by “more of everything?”
"That would be highly unprofessional – no, the expert evaluates the bass, the low-mids, the
high-mids, and the treble. In more detail: the lower bass range, the upper bass-range, the
lower mids, the higher mids, the presence, the absence, the dominance, the brilliance and the
articulation. ‘More of everything’ does not make for precision discrimination, does it?”
"But then, where is the distinction between ‘strong bass, loud low-mids, assertive high-mids,
and dominant treble’? If everything is loud, where is there something specific?”
"That is amateurish thinking. For my master thesis “About the Wood in general and the
Sound in particular”, I have done a literature search and worked through a multitude of
books and magazines on electric guitars. Let us just take ash as it is deployed in Fender
guitars, for example. Specialist literature describes its sound as:
Ash⊕: mellow, rocking, soft, bass-y, brilliant, emphasis on the mids, no pronounced share of
mids, balanced, lively, powerful, tight, warm bass, long sustain, dry, airy, hard-wood-y, rich
in attack, strong assertiveness (because ash is of stiff structure), responds considerably faster
than alder.
Look, you have to be aware of all this if you want to build a guitar. Indeed, that is not a rush-
job, no simple saw-&-glue-together, but its fine craftsmanship. Artisan craftwork, crafty
artwork. Otherwise we wouldn’t require those years of training and formation, those
advanced olfactory and gustatory seminars of ongoing education …”
"Even gustatory??"
"Yes, sure – the lay-person is not aware of all that. Good guitar-wood needs to be grasped
with all sensory channels. I do not only smell the wood – I taste it, as well.”
"Really quite fascinating. But let’s go back to your literature search, where you said: ash
sounds both mellow and rocking. Isn’t that a contradiction?”
"By no means! These are citations from different reference books! Of course only the expert is
familiar with this so-called semantic differential. Von Bismarck is said to already …”
"The battleship? The one that sunk the Hood?"
" … and was sunk itself shortly after … so many lives lost on both sides … tragedy … where
were we? No NOT THE SHIP! Von Bismarck was the chancellor of Prussia and then of
Germany in the late 1800’s, but I don’t mean him … a later Von Bismarck, there was a keen
thinker in the family… The name slipped my mind. Gandalf maybe … no, that would be
Tolkien … or Gottfried, or Gilbert … or Sullivan … no, that would be the composers. Maybe
Gottfried, after all … “ "Dr. Worm, Sir, please … the rocking ash ..."
"Sure, ash. Rock, that indeed does not always equate to just rock – there’s hard-rock, soft-
rock, prog-rock, under-the-rock, metal, death metal, beyond-death metal, grinch, grunge,
grump, pump, hunk, and hulk!” "What – him, too???" "What do you mean: him too?"
"Does the Hulk have a special sound? I thought he’s just green?”
"I don’t understand what you mean. A “green” sound – our area of trade is not aware of that.
But this is not uncommon at all in science! Especially in the interdisciplinary realm, close to
the fringes, pushing the limits – you will find a lot of ignorance there. That’s just why we have
specialist literature that exactly specifies the sound of the wood.”
⊕
Literature sources are given at the end of the chapter
© M. Zollner & Tilmann Zwicker 2009 & 2019 Translation by Tilmann Zwicker
"So: mellow-ly rocking?" "That’s for ash, of course."

"But how do the assessments of “emphasis on the mids” and “no pronounced share of mids”
fit together? Would it be possible that one of the expert authors is not that competent, after
all? Or that the wood is not that decisive to the sound, anyway?”
"No, of course not – wood is always decisive. One expert will write “emphasis on the mids”,
because he perceives the sound as such: with emphasized mids. The next expert will write “no
pronounced share of mids” because he will perceive the mids as not pronounced. That is not
a contradiction at all!
"And that was discovered by that Bismarck person?"
"Von Bismarck! … Um, no … well, yes, I think so. Or rather in parts, I think. The semantic
differential differentiates the semantics. You yourself have asked about the differentiating
aspect when I first elaborated! The differences in the semantics, in the teachings about the
meaning of words. That’s Von Bismarck – it seems he is even acknowledged by some
psychological psycho-acousticians. And that is quite something! I’ll only mention Berkeley.
Have you already seen those guys?”
"In Boston?"
"Why in Boston? In Berkeley!"
"Oh … not Berklee but Berkeley!"
"I see, you had those other people in mind. Here we have more of a phonetic differential. Did
you know that already in the Middle Ages …”
"!!!"
"Okay, right – the wood. Well then: if one type of wood sounds bassy, mid-emphasized, and
trebly, then that’s balanced, isn’t it? And a long sustain may well sound dry. The opposite
would be … well, one would have to say … opposed to dry sustain … but ash does actually
not show this kind of contrast. To the contrary, the mellowly-rocking, airy-balanced dry
sustain is indeed a characteristic for ash. Contrary to alder, that is.”
"Oh – that’s interesting. What characterizes alder, then? Does alder sound different compared
to ash?”
Dr. Worm jerks to a halt, raises his right index finger and utters, almost in a whisper: "alder is
the perfect material for the electric guitar. Alder is the master builder’s wood. If I had to
build an electric guitar right now, alder would be my one choice. About alder, my literature
search indicates:
Alder⊕: silky, mellow, warm, tender, many harmonics, restrained share of treble, fat bass,
rather subdued share of bass, strong mids, round share of mids, much sustain, accentuated,
squishy, good presence, undifferentiated, balanced, full sound, a sound thinner than that of
basswood, faster response than basswood.
That’s how experts judge in specialist books. Now doesn’t that sound very different compared
to ash, after all! Knowing this, we can build a custom guitar for every customer as requested.
Of course, only the expert knows this – wood is not understood by just anybody.”
"Indeed, Dr. Worm – we, too, have some difficulties to get it all straight in our heads. Fat,
tender, subdued share of bass, and with squishy-ly accentuated presence yet being
undifferentiatedly-balanced … that Von Bismarck fellow is again behind this?”
"Right, that later one. Yes. A most differentiating description, indeed.” "Really?? Excuse me,
that is outlandish! How can one and the same wood sound squishy and accentuated? With a
bass that at the same time is both tender and fat, and rather subdued on top of that! The reader
will discount that as pure hokey-pokey!”
⊕

"Now, that does hurt me a bit! This criticism of centuries-old knowledge – that is not justified!
These insights have helped for hundreds of years to build violins that to this day …… and it
will assist in all likelihood for hundreds more years!”
For a moment, Woodrow W. Worm, PhD, is almost angered, abruptly turns around, takes a
few steps … but then stops again and elaborates in conciliatory tone: "I understand now that
you cannot understand this. Look, consider we have some luthier writing a book about the
electric guitar. He will just be compelled to put in a chapter about wood, won’t he? And since
he will – like probably every luthier – at some point have heard an alder-Strat with its fat
bass, he may well write that in his book. Don’t you think so?”
"We start to understand. Another author will own an alder-Strat with more subdued bass …”
"Presumably so. After all, we have – on a global scale – a large number of alder-Strats.
Thousands. Millions, even! Still, not everyone having a Strat at hand should be allowed to
write a book about it – isn’t that true? Only the expert may do that, right? Because in books,
pure opinions mutate into dogmas, into axioms, don’t they? Specialist books are
objectifications of subjective assessments.”
"But if we now impute … sorry: attribute … such different – even opposite – characteristics
to the wood, wouldn’t it be better to say: the wood has practically no effect on the sound of an
electric guitar? You will not want to publish contradictory doctrines in textbooks, will you?”
"Science does live on dispute, it subsists on the dialectic contention of diverging spheres.”
"Wow! Whhhaaahht?"
"Wood is, after all, an object embracing objective characteristics …”
"The soul …"
"That is something you will never comprehend: it is exactly the soul that is not the objective
but the transcendent, holistic mystical. No, I allude to the objective criteria that exist far
beyond any validation. In the terms set out by Plato, I say: wood, as a spiritual universal
essence, has an existence outside of human thought. Seen from that angle, the textbook author
delivers his personal subjectivization of the objective. Do you follow?”
"We’re trying: the textbook as coexistence of objectified subjectivity and subjectified
objectivity. In a way: as platonic coexistence?”
"That’s about right. Aristotle looked at it in a different manner, as did Hilbert, by the way –
Fuchs elaborated on that already back in 1972: an accentuation of axiomatic contemplation
implies that we keep – of the factual material of notion from which the basic concepts of a
theory are formed – in the axiomatic design of the theory only that which is formulated as
extract in the axioms, while abstracting from all other content. That’s Knaur, in 1972.”
A clearing had come in sight, and Dr. Worm picks up the pace as he purposefully approaches
a young basswood tree. His flow of speech had stalled – but only for a moment. "I can
exemplify that with this young lime, or basswood, tree. It represents a wood highly suitable
for electric guitars – although it is underestimated by many. This lime tree here” – he
competently kicks the trunk, such that the whole universal essence is shaken by unbridled
vibrations – “has a very good response, as you can clearly recognize, but will give a squishy
sound. That does, however, not imply that basswood will – in the sense of Plato – necessarily
sound squishy always and everywhere. It does not even need to be called basswood at all: in
Hilbert’s terms it could also be designated table, chair, or beer mug. But let us by all means
leave the name, let’s continue to simply designate it ‘basswood’ – it is called that, after all. In
my literature search, I have compiled everything at our disposal regarding basswood:
Basswood⊕: mellow, low mids, squishy, good response, undifferentiated, somewhat mid-
laden, similar to alder, relatively little sustain, warm sound that lacks zappy-ness, unobtrusive,
forceful, rather dull-sounding.
I believe that the above three examples of ash, alde,r and basswood quite clearly show the
effects of the wood, and what specialist literature is capable of.”
"You are correct, our opinion on the matter starts to solidify. Alder with its accentuated-
squishy, mid-emphasized, mellow-full tone is thinner in sound than the well-squishy-ly
responding basswood with its soft-powerful low mids?”
"In a very compressed fashion, yes. According to the textbooks: yes. Yes, by all means. To
summarize even more succinctly: basswood sounds similar to alder; however, alder sounds
thinner than basswood. More like poplar – which by the way sounds like basswood. I shall
right away reveal the sound characteristics of other woods that are excellently suitable for
electric guitars:
Poplar⊕: the tonal characteristics correspond to those of basswood, clear treble, more airy
than basswood, unobtrusive, round sound, like basswood but thinner, the tonal characteristics
correspond to those of alder but lack warmth and brilliance, more crisp than basswood, round
tone, rather short sustain.
Maple⊕: rich in attack, singing tone, hard sound, much sustain, rich in harmonics, lively, not
warm, warm bass, lacking warmth, mid-emphasizing sound, brilliant.
Mahogany⊕: mellow, very bass-y, delicate brilliance, warm mids, good sustain, silky, warm
sound.
Rosewood⊕: powerful and harmonic sound, airy basic character, loose and full bass range,
sparkling treble.
Let’s hang on to this fact: the wood defines the sound of the electric guitar. The – I am
tempted to say: new-fangled – electronics can only add nuances! The basic tone is generated
by the wood.”
"Indeed, we have also already seen this opinion. A well-respected author writes in 1977 A.D.:
‘every piece of wood has its intrinsic sound’. A few pages on, the same author opines (in the
same book): ‘the sound of an electric guitar depends mainly on the pickup’, and in 1994, he
proclaims in a new edition: ‘for solid-body guitars, as well, the body has a decisive influence
on the sound’. In the same new edition, we again read a few pages later: to a large part, the
difference in sound between electric guitars is due to the pickups’. So there we have it again,
what the (original) elders already knew: all things are connected … everything depends on
everything else. What is more important, though: pickup or wood?”
"In my literature search I have looked into this issue, as well. The thing is: for the luthier who
knows everything about wood but has had no course on electro-acoustics, the sound of the
electric guitar is in the wood. However, those who have graduated in physics or electrical
engineering but cannot tell a board of beach wood from swamp ash, nor from birch – to those
the sound is exclusively due to the pickup. See the following literature collection:“
⊕

♦ Wood does not influence the sound (Pearson/Webster, in: May p.144).
♦ Wood must have an influence, differences in pricing between guitars are due to the wood
(May, S.144).
♦ Using high-grade wood is futile (Zills, in: May, p.86).
♦ Wood has an influence on the sound (Evans/Evans, in: May, p.145).
♦ The influence of the wood on the sound must not be underestimated (Gitarre & Bass, 3/97).
♦ Experts agree that the sound of a solid-body is mainly determined by the electronics
(Carlos Juan, Fachblatt Musikmagazin, 1996).
♦ The sound of an electric guitar depends relatively strongly on the wood (Meinel, p.47).
♦ The sound is not mainly determined by the pickup; rather, the wood provides the
foundation (Jimmy Koerting, Fachblatt Musikmagazin).
♦ Pickups convert the vibrations they are subjected to and do not form the sound themselves
(G&B 5/06).
♦ The tonal characteristic of the electric guitar is substantially determined by the choice in the
wood. Pickups and amplifiers support the sound of the guitar but rarely change, influence, or
mould it fundamentally (Day et al., p.205).
♦ Solid guitars can, however, be manufactured in almost any shape and size; no considerable
effects on the sound should be expected by this. (Day et al., p.140. That’s the same Day as in
the previous citation).
♦ The wood does not only determine the sound color but in particular the information of the
string vibration (Gitarre & Bass, 02/00).
♦ The electrified plank-guitar is predominantly an acoustic instrument. The wood determines
the sonic character; the pickups only to a very small extent. Hence a humbucker is nowhere
near to be able to exorcise the characteristic sound- and attack-evolvement from a Strat with
alder- or ash-body (Udo Klinkhammer, Gitarre & Bass, 2/00).
♦ Looking at the process of the sound generation of the electric guitar, we quickly grasp that
the quality and type of the wood used will influence the sound of an instrument just as
massively as the construction (Day et al., p.206).
"Now that is a clear vote: the majority sees the wood of an electric guitar as determining the
sound. If that were not the case, we could build great-sounding guitars just as well from
inexpensive materials. Which is not what the specialist trade can be interested in. Or at the
most there is a supplementary interest. That’s why every brand manufacturer points out that
they have only the most expensive tone-woods underneath their sunburst finishes. And that,
my friend, easily necessitates to a price of one or two grand. Dr. Worm again kicks against
the trunk of the basswood, as if to underline his words: the products issued by his company
were indeed also looking for recognition and intrinsic value – and therefore for high retail
prices. From the tree, a butterfly that had been disturbed due to the rather massive tremor in
“its” bass wood took off, zappily got off the starting blocks, resonated all the way to the
wingtips but then landed again with an undifferentiated, squishy decay in its wing motions.
Relatively little sustain – the thought flashed through us.
"But, Dr. Worm, Sir – may we call you Woody? – if now the professional circles report so
inconsistently about the wood: hasn’t anybody compared guitars made of different types of
wood? If ash and poplar sound so differently: couldn’t we just compare an ash-Strat with a
poplar-Strat”
"Woody it is, then … indeed that has been done, as e.g. the report in the Fender-issue of G &
B shows. However, this listening comparison yielded only ‘minute differences’. Could be both
an individual opinion and verified expert knowledge. But there are more comparison tests …”
Dr. Worm – Woody – had stopped because from afar a buzzing engine noise had become
audible. “They’re sawing away again”, he said with disgusted air. “For building-timber.”
The direction from which the noises could be heard seemed to unsettle him. It was the
direction of where we had started our educational forest-walk. With a short “I gotta see that”
he turned and started back on our path, almost running. His facial expression vetoed any
further question. Time dragged on, minutes passed – only now we became aware that we had
walked downhill for some distance, and now it was an uphill rush towards the buzzing noises
of the saw (increasing from a perceived 0.2 Asper to now 0.4 Asper). A smokey-tar-y
component suddenly attacked our olfactory afference, still undifferenciated but quickly
gaining in dominance. While the information of the n. opticus by itself could have been
interpreted as a kind of fog, the cooperation of first and second brain nerves clearly indicated:
something’s burning! Forest workers became outlined against the smoke, force fields
diverged … we had been here before? In the center of the scene: the ash tree, the wood for the
master craftsman (at least in 40 to 50 years). Right in the midst of it, but less upright and less
proud than it had been only an hour ago, rather cut up into sub-sets now, still unsorted, lying
around on the ground in bundles. The thinner ones of the master-woods branches, previously
piled up to dry, they had been thrown together forming a heap, flames flickering already,
affording warmth to those hands that only minutes before had callously decapitated the
wooden bretheren. Ash to ashes … Benef’cent is the might of the flame, when o’er it man
doth watch, doth tame. Woody lost it completely, enraged, beserk, his balanced round bass
gone with the wind, rich in attack he went up against one of the lumberjacks, with his treble
content having lost any moderation: “You can’t do that! That was wood for the masters!”
“That’s how us here’ve always been doin’ it” – strong mids came back from one of the
workmen. “We cold, we light ’em up”, his neighbor contributed with resonant bass, and a
more trebly but still squishy voice added: “la leña seca bien arde, amigo!”
We decided to better not get involved in this final dispute, as much as it might have been of
scientifically fundamental and typical character. We pondered the rising smoke. Lively-
powerful, the grey curled out of the glow, converted into white, pulling a Fibonacci-sequence-
like bifurcation right behind it, just before it dissolved itself, rapidly ascending to a higher
plane. The warm fundament grabbed us with its tight bass, while it dive-attacked from above
with distinct hard-woodiness. No doubt at all: it had to be ash – that much we had learned
from the elaborations of Woodrow ‘Woody’ Worm, PhD. Ash through the ashes …
And some supplementary opinions1:
G&B (Gitarre & Bass), 9/02, p.80: “Bob Benedetto, whom many (practically all) take to be the best
luthier alive, states: popular opinion demands wood that has slowly grown (slow growth shows in
narrow tree rings). According to my knowledge, that is a myth. … some of my best guitars are made
from spruce that some would take as substandard. Check out the old masterpieces from Stradivari or
Guaneri – they are made from wood with wide tree rings, as well. Maybe we have, for years, fallen for
the advertisement in the brochures of a few companies that promote wood with narrow grain. … Once
I went to a wood supplier in Pennsylvania and bought the worst wood I could find. I built a guitar
from it that sounds excellent – after all, Scott Chinery bought it.”
Tom Lockwood, Guild-Guitars, in: U. May, p.145: "Manufacturers like ourselves only use the
highest-grade material, that’s only about 5% of the yield. We therefore ask a mill producing 100.000
board feet to let us select about 5000 feet. The remainder we have no use for, and that has a
tremendous impact on the price."
1
Translator's note: the citations were in German and I could not trace the originals in English (where
appropriate). I therefore re-translated them into English. This will without doubt have led to a different wording
compared to the original. The same generally applies to citations thoughout this book.

“Taylor builds good guitars because we now how to do it. To prove that, we have built an acoustic
guitar from an old, rotten pallet we found in the garbage. The top was from a scrapped plank of which
we could not really determine the wood. We so elaborately glued together the top from 6 slats that it is
hard to even detect that, and the holes from the nails … were highlighted with small aluminum discs.
This pallet-guitar was one of the most noticed guitars at the winter-NAMM-show (Bob Taylor, ISBN
3-932275-80-2).
"Besides, I actually think that the component wood is, in general, overrated“ Ulrich Teuffel, Teuffel-
Gitarren, in G&B, 5/04, p.85.
D. Holz: Holztechnologie 25/1, 1984, p. 31-36: about some correspondences between forestal-
biological and acoustical characteristics of tone-wood (resonance wood): “A connection between the
year rings and the acoustically important properties of resonance woods cannot be specified.” G.
Ziegenhals on the topic: "Recent investigations at the Inst. for Musical Instrument Making” generally
support this.” FAMA-Seminar, DEGA 2001.
♦ The Les Paul Custom sports an ebony fingerboard. An ebony fingerboard gives a slightly more
mid-rangy sound (Luthier Thomas Kortmann, gitarrist.net).
♦ An ebony fingerboard results in a brighter and more brilliant sound (Gerken).
♦ A fingerboard made of Rio-rosewood will render the sound more brilliant. (Kortmann,
gitarrist.net).
♦ The maple fingerboard makes for the clearer sound; the rosewood fingerboard will sound meatier.
[Duchossoir, Stratocaster-Book].
♦ Rio-rosewood produces a ‘full octave of additional harmonics’ (Day et al.)
♦ For me, maple fingerboards work much better than the ones made from rosewood because they
have a tigher, stronger tone (Eric Johnson, G&B, special Fender issue).
♦ The "Slab-Board" (rosewood fingerboard) is one of the secrets of the renowned old crystal clear
vintage-sound especially in Fender guitars (Day/Rebellius, p.72).
♦ Electric guitars with a neck-through construction behave much more favorably compared to a bolt-
on neck. The gain in sustain is striking. (Meinel, 1987, p.63).
♦ Set neck and bolt-on neck have equivalent decay times. (G&B, 3/97).
♦ The bolt-on neck diminishes the sustain of the guitar (Lemme 1982, p.59).
♦ The bolt-on neck can generate a long sustain, as well. (Lemme 1994, p.50).
♦ Overall, maple necks are known for giving the instrument a percussive touch (G&B 4/06).
♦ One-piece maple necks sound just like necks with glued-on fretboard (Lemme 1982, p.62).
♦ (There are) practically no differences between three special guitars that are distinct only in the way
the neck is attached (glued-on, bolt-on, neck-through) (A. Paté, Nantes 2012).
♦ The maple top contributes a lot to the sound character of the Les Paul (Gibson-CEO
Henry Juskiewicz, in: Bacon/Day, Les Paul Book, p.61).
♦ The Les Paul Customs had a body completely made from mahogany, just like Les Paul preferred it
to the mix of maple and mahogany. (Bacon/Day, Les Paul Book, p.20).
♦ G&B, 9/05: Les Paul: back then my idea was to manufacture the whole guitar, i.e headstock, neck,
and body, from one and the same piece of wood. They didn’t do it. When I asked the president of
Gibson why not, he replied: “because now it’s less expensive.”
♦ G&B, 7/02, comparison test: "The Fame LP-IV indeed sound most authentic. Its sound is very
similar to that of the original (Gibson Les Paul).”
Fame LP-IV: maple neck, oak fingerboard, alder body, mahogany top.
Gibson Les Paul: mahogany neck, rosewood fingerboard, mahogany body, maple top.
♦ G&B Fender special issue S.76: ash-Strat vs. poplar-Strat: only 'minute differences'.
♦ G&B 10/04: alder-Strat vs. poplar-Strat: differ only in 'finest nuances'.
♦ Of course, the body wood decisively shapes the Fender sound. … A true connoisseur hears totally
different characteristics in a 61 Strat compared to a late 64.
A few paragraphs on, we then read in the same (!) comparison test: as one will imagine, the sound
results are very close to each other (G&B 3/06).
♦ G&B 5/06: Squier-mahogany-Strat vs. Squier-basswood-Strat: using the neck- or middle-pickup
the two guitars sound all but identical.
♦ G&B 9/05: Still, the PRS EG surprises with authentic Strat-sounds (mahogany neck, rosewood
fingerboard, mahogany body).
♦ G&B 2/00: Despite the humbucker, a Strat can never become a Les Paul.
♦ G&B 7/06: Gary Moore: some people believe that you hear a Stratocaster on 'Ain't nobody', but in
reality it’s my own signature Les Paul.
♦ Jimmy Page recorded the complete first Led Zeppelin album using a Telecaster. The guitar sound
on that album is exactly that of a Les Paul. (G&B Fender special issue).
♦ G&B 9/05: and so despite identical basis (mahogany neck, rosewood fingerboard, mahogany body)
the three PRS-SE guitars each deliver typical sound characteristics à la Strat, SG/LP-Special, and
Standard Paula, respectively, and this on a high sonic level.
♦ E. van Halen: "Die Strat had too little sustain. Hence mahogany" (G&B 7/04).
♦ Larry Carlton: "The Tele doesn’t kick butt sufficiently. Hence Gibson" (G&B 5/01).
♦ Cavities (in the solid body) have no influence on the sound (Lemme 1982, p.54).
♦ "To improve the body's resonance, the core body is drilled with eleven 1,5"∅ cavities."
(Duchossoir, Tele-Book, p.31).
♦ "The cavities in the Les Paul have no influence on the sound characteristic of the model; we tested
it. (Henry Juskiewicz, Gibson CEO, Les Paul Book, p.61).
♦ "Cavities increase the ability to resonate." (Day et al., p.140).
♦ Resonance chambers: "It is difficult to avoid the impression that the router was called in often, and
wood was taken away until the manufacturer was of the opinion that now the guitar is light enough"
(Day et al., p.143).
Eric Johnson: "More than 75% of the sound is in the fingers". (G&B 5/01).
E. van Halen: "It’s not really the equipment, it’s in the fingers". (G&B 7/04).
Jimmy Page: "You know, I’m getting a lot of sounds out of that guitar that you will normally not get
from it." (G&B Fender special issue).
Richie Sambora: "But you also hear that Hendrix went through only through the amp. It’s his fingers.
The same with Jeff Beck: you may use his rig and his guitar but you will never sound like him. It’s in
the fingers." (G&B 11/02).
Jan Akkerman: " It all comes down to your hands." (G&B, 1/07).
Jaco Pastorius: "Piss off the amp and the instrument. It's all in your hands." (G&B 1/06).
Jeff Beck: "no shenanigans, no mumbo-jumbo – just the fingers." The man does get it right ...
v. Bismarck G.: Psychometrische Untersuchungen zur Klangfarbe... Akustik und Schw.-Technik, VDI 1971.
Wheeler T.: The Guitar Book. Harper & Row, New York, 1978.
Wheeler T.: American Guitars. Harper & Row, New York, 1982.
May U.: Elektrische Saiteninstrumente in der populären Musik. Dissertation, Münster, 1983.
Bacon T., Day P.: The Gibson Les Paul Book. Outline Press Ltd., London, 1994.
Day P., Rebellius H., Waldenmaier A.: E-Gitarren, GC Carstensen, 2001.
Lemme H.: Elektrogitarren, 1977, 1995, 2003.
Meinel E.: Elektrogitarren. VEB Verlag Technik Berlin 1987
Gitarre & Bass, Musik-Media-Verlag, Cologne.
Fachblatt MusikMagazin, Spezial-Zeitschriftengesellschaft, Unterschleißheim (previously: Munich).

0.2 Signature Guitars 0-11
0.2 Signature guitars
To have a signature guitar built by a well-known company – for a guitarist, that is like being
knighted. There’s fame, the records are selling in the millions, and now the maker of that be-
all-and-end-all go-to guitar asks whether one’s name can be put on the headstock. And so
Tawdro releases a Paco-Beslmeisl-guitar, and presents it to him with maximum ballyhoo at
the SchlockMuCom. Such special edition models are usually developed in close cooperation
with the correspondingly honored guitar player, and address all those who seek to sound just
like their revered idol. A genuine Paco-Beslmeisl-Signature – only the dyed-in-the-wool fan
owns that. Complete with yellow pick and a bandana.
The corresponding actual genesis, however, may be rather sobering: “Oh man! Look, Henry,
the October numbers are not really up to plan – got an idea?" "Let’s do another signature,
then.” "Yeah, that’s what I thought of, too. Whom could we choose?” “Whom … well … that
so-called manager of that WARLORD guy just mailed about when at last he’ll get one. His
present axe is a shambles.” “WARLORD – that douche should first bring back the rig we loaned
him and that allegedly is now in an ebay-auction the third time already. That guy – we will at
the most supply him with a lid on his coffin – in company colors.” “Jeez – you ARE having a
bad day; it was just an idea! I kinda dig the WARLORD – he’s not as grotty as people always
say. What about ol’ BLIND BOOBY BROONZY?" "You crazy? Nobody gives a sh.. about him
anymore. I was rather thinking of SLOWHEAD." "Right back atcha: that jackass these days
resells his guitars right away; he’s a no-go. What about BIERMEISTER?" "Not notorious
enough here in the States. We should stick to one of ours. Or a Brit. No, rather not – they’re
tough to understand. Texas, that would be good – ‘Don’t mess with Texas’ ‘n’ all’. Yeah – I
wanted to visit Austin again, anyway.” “Speak about Austin: what are HEALEY AND THE
DIFFERENCE up to these day?" "Told ya: no Brits, and certainly not that underpowered
HEALEY! Power we need, so maybe JOE ROCKER. His Strings from Hell sells like hot cakes at
the moment. Right: JOE ROCKER, that would be it.” "What? That guy is gonna kick the bucket
any minute now. He was constantly high on heavy stuff and wouldn’t come down … now
he’s in intensive care. In Axxes, Lix ‘n’ Trix they already published a kind of orbituary.”
“Awright!! That’s it, then! Think about it: if he makes it, we present him his signature at the
next WAMM. If he doesn’t we issue it at the time of the funeral. Posthumously, sort of, in a
black gigbag adorned with a silver cross.” "We could also bring it to the IC room right now.
That would make for an epic pic: ROCKER with his eyes half closed, mouth hanging open –
and our signature axe right across the bed. I could hold it in place … we’ll want to avoid a
disaster like the one at Ronnie’s rehab last year when the guitar slid off the covers and
crashed.” "Done! Go call the head physician for a permission of the shoot and such – we’ll
pay him a flight to Vegas with two weeks in a suite … they always have some kind of
conference there, anyway. I’ll inform the custom shop. At last they can use up the birch slats
from Patagonia; those were going to be woodworm fodder soon.” "Should I offer the chief
physician some sweetening if he gets difficult – maybe a complimentary ticket for his wife, as
well?” “Get real, man – why would he want to take his wife?! He’s looking at a voucher for
the all-inclusive package, and I mean FULL inclusive.” “Okey dokey; well just need the text
for the official statement, then. Something like: in every clear minute … well: in every free
minute, JOE has contributed to the design and development because he insisted that his sound
comes across at 100% in this signature. He brought us his original axe to measure it, and by
his own hand wound another 25 feet of wire onto the pickups. Even the barf-green – he
designed it himself. What a hoot!! That makes signature model number … ???” “Must be the
twenty-fifth or so, I think.” “ Very well: LIMITED EDITION!"
Signature models carry the names of famous guitarists, and are designed in close cooperation
with these. Say the ads, and say the test reports. And many an epigone buys such a special
model, hoping that now he/she will be able to get the same sound as the hero. The latter will
present just that same model to the camera, and will play the instrument exclusively, live and
in the studio. Will he, really? Now, many of the top players own not one but 10 or 50 or even
200 (or more!) guitars. Will they suddenly play only that signature model?? The specialist
literature knows more:
Jeff Beck’s Fender-Signature-model is already marketed in a second edition. He himself (as

stated by G&B) “uses almost exclusively just a regular Fender Stratocaster (only the tailpiece
and the nut are taken from his Signature model)”. Regarding his album “Blow by Blow”, on
the cover of which he is shown with a Gibson Les Paul, he says: “because of that cover, many
people believe that they hear a Gibson guitar on that album. It was Strats and Teles, though.”
(G&B, 2/01). Conversely, the Gibson book states: “For the recordings of this LP (meant is
“Blow by Blow”), Jeff Beck used this guitar (meant is a brown Les Paul) almost exclusively –
even though a Fender Telecaster with humbuckers can be heard here and there, as well. On
some of the tracks, Beck started to use a Fender Stratocaster, and since then has been as good
as married to that guitar and that manufacturer”.
Jimmy Page "is known predominantly as a Les-Paul-player. However, he recorded all of the
first Led Zeppelin album using a Telecaster (!) that Jeff Beck had given him. Replying to a
remark that the guitar sound on that album was exactly that of a Les Paul, Page once told the
interviewer from Guitar Player: “You know, I can get many sounds out of the guitar that you
would normally not get from it. That confusion goes back to the early sessions that I played a
Les Paul on. Those recordings may not sound like a Les Paul but I did use one.” G&B Fender
special, p.37.
Moreover, Messrs. guitarists the will be happy to switch the supplier. Here’s Richie Sambora
in an interview by G&B (10/02): "Also, I am lucky to have a few 59’s and a ‘60s sunburst Les
Paul. Those are my favorites right now. As such, Fender has been marginalized a bit.” G&B:
"But didn’t they recently make a signature model for you?” Sambora: "True! But what can I
do (laughs). ... Actually, I play everything that I get my hands on and that sounds halfway
decent.” Right above the headline ‘Richie Sambora Standard Stratocaster’, we find in the
Fender brochure: "Designed under the direct supervision of some of the world's most
influential players, these models have been painstakingly crafted to accommodate each artist's
unique specifications and playing style" (Fender-Frontline).
Duchossoir’s book on the Strat, preface by E. Clapton: "The Stratocaster is about as close to
being perfect as any electric guitar can be". Clapton-ad: "The one and only electric guitar♣."
On the other hand, we read in Bacon/Day: "I have never found a guitar quite as good as that
one” – with Clapton referring his lost Gibson Les Paul. Why should I care about what I said
yesterday?! ‘The Gibson’ cites Clapton using an ES-5, the ‘Cream sound’ is due to Clapton’s
SG, or to his Firebird, or to his 335, respectively, and he famously used a Telecaster, as well.
The acoustic-sound of Cream, however, stems from the Epiphone and Guild guitars of Mr.
Eric Patrick Clapp. It seems many more signature models will be in order. There is already
one issued by Martin … that apparently was scorned by E.C. for his UNPLUGGED oeuvre.
♣
Stratocaster, G&B 4/06

0.2 Signature Guitars 0-13
Mark Knopfler: "If I want a fatter sound, I’ll use my Les Paul – it is simply more dynamic.
That does not mean, however, that I could not do the same thing with a Stratocaster1." That
might not entirely reassure the gentlemen at Fender, since they write about the Mark-
Knopfler-Stratocaster in the brochure: "His unmistakable tone comes from three Texas
Special single-coil pickups and a five-way switch." And Mark even goes one better:
"Sometime I use the Les Paul to get a particularly clean sound1."
Gibson’s Lucille is B. B. King’s signature guitar. Charles Dennis, guitarist in B.B.‘s band,
comments: "We were out there somewhere and Lucille couldn’t make it – she was still on the
plane. We had to play the job, though, and he player a Fender. What can I say: in his hands it
still sounded just like Lucille2."
Yngwie Malmsteen has been given a signature model by Fender, but remarks: "But the only
ones that I actually play, are Strats from the late 60’s and early 70’s”, (G&B 11/02, S.63)."
"Our desire with the whole Signature Series was to build the guitars exactly the way the artists
play them. We didn't just want to build something that everybody was going to buy and then
the artist had to have his different". Fender-exec Dan Smith in Duchossoir’s Telecaster book.
Lenny Kravitz got a signature Flying V from Gibson. However: "I can hardly remember the
details. I don’t now anymore what I changed on it – just that I shortened the neck some.”
True, as the test report in G&B discloses: the scale is 625 mm, compared to the 626 mm on
the original Flying V. Also: “It is much lighter.” True, as well: 3.2 kg compared to 3.3 kg.
"And it sounds better. That makes a big difference.” The tester does indeed state that there s a
difference: "To my surprise, it delivered more sustain that first expected, but it cannot match
the regular V. In terms of attack and the evolution of the tone, it lags behind.” What! The
tester does some straight talking? That’s is rather unusual … he even detects a constructional
flaw: “Due to the shallow neck angle, the strings can exert next to no pressure on the bridge;
they run across it almost without any bend angle. One consequence is an unintended and
annoying sitar-effect on the e-string”. That, on the other hand, the vibrato is a weak point, that
is typical: "the Maestro-Vibrato has always had the deficiency that it is not adjustable –
neither in terms of the spring-tension nor regarding the height of string retainer. On the guitar
under test, the lever hovers so closely above the pickguard that it is just about possible to get
the finger under the handle.” Still, Mr. Kravitz heartily condones the fact that the signature
model sets the customer back a cool € 6990.-, in contrast to the regular model at € 2190.
"Such things are always more expensive … (grin),” Does he actually play it? "In the studio I
always use a selection of Les Pauls. Mostly goldtops or vintage flames from ’58, ’59 or ’60”.
Typical stuff you will just simply use if you “store about 140 guitars in a storage area
specially rented for the purpose.” ... getting a signature guitar as no. 1413. Having said that:
the Flying V seems not to be a total loss, after all: Mr. Kravitz poses in the bathtub with such
a piece. Watch out, though, dear fans: that is the white V. The black V “I would have never
subjected to the paint”. Because (aren’t our artists so precious?!): the tub is not filled with
water but with red paint. The situation became rather dangerous for Kurt Cobain: "He played
Jazzmaster- and Mustang-guitars – until he received a signature model. He committed suicide
in 19941." Come to think of … maybe … had he stayed with the regular stuff ….
1
G&B Fender special issue.
2 3
G&B 9/06 G&B 06/2004 p.72, G&B Gibson special issue p.126.
Mike Einziger (Incubus): "For a long time already I had no fun playing the PRS. I just
wanted something new. I wanted to change, without any sanction. And so I decided that I
never again would legally tie myself in such a way to a guitar company.” Is that the reason
why there is so far no signature model for Mike Einziger? “Correct. I have no interest in that.
To be honest, I find that simply silly … (laughs). I mean, what should I change in an
instrument that in its own ways is already perfect?” G&B, 1/07, p.48.
... and many more ...
The specialist magazines further fuel the signature- and custom-market by detailed reports on
how the great guitar-wizard has had his (or her) axe and amp modified. Frequently with the
hint attached: “if you want to do the same, be prepared to shell out € 8000.-.“ Many an
epigone will save over the years to reach this (or a smaller) number to come closer to his/her
hero. And if the original Blackie is out of reach, then at least let’s go for a set of 3 new pots
for € 600.-. Or – for the Marshall – let’s get that more authentic (!) output transformer from
the US. The old lag over there is not even able to send to Old Europe an offer that would
correspond to mercantile convention, but he does have, no less, several transformer variants in
his self-wound assortment. Better sound? Only if you believe (Chapter 10.6).
For the sake of fairness, we do need to cover another variant: there’s the well-off forty- or
fifty-something who gets onstage with his mates on the weekend purely for fun. He really
enjoys that they all envy him for his original 1963-Rickenbacker. He doesn’t mind that it was
expensive; to the contrary: that’s why he bought it. And of course because the old Beatles
songs are a pleasure to play on it. Actually, if such a Ricky could be had for € 100.- at every
yard sale, he would have rather chosen the old Epiphone Casino. Or some other pricey
‘unique feature’. Just like his wheels, a tuned up Helby-S Corba – that cannot be found on
every street corner, either. Without any malice now: making music has got a lot to do with
emotions. Including the audience (“incredible, a ‘Richenbaken’”), and the artist (“how can
that bloke next to me coax such awesome tones from his el-cheapo?”). Therefore it is not
uncommon – actually it is even imperative – that many musicians attribute a power of
inspiration to their instruments that cannot be verified scientifically. Looking at that
translucent-blue stained maple top … oh man! On the rear, a tiny sticker becomes visible with
a 4-four-digit figure starting with a 9 … that’s how impulsive comfort shopping happens
(especially if GAS – gear acquisition syndrome – plays a role, as well). Finally: a 12-string
that not everybody has. Didn’t that one player back in the day – what was his name … he
must have played one like that or something similar … man, these rare stringed bodies can get
to you … it’s so … oh … where were we? Which chapter was this supposed to be? Ah, yes,
right, special models! Custom-Shop, Artist-Gallery, Signature-Model ... of course! And why
not? Not due to any logic and rationale! Not because of any alleged extra-fidgeting and some
supplementary wisdom of some trendy idol, but out of pure lust and passion. Right - that had
to be said! Sure, the sales guys are perfectly aware of this, and every year they provide ample
ordnance for the passionate buyer with the bursting wallet: model of the year, limited edition,
custom colors, custom woods, with the original signature by Mr. X (surcharge is $ 4000.-
with no less than about $ 5.- going to endangered jungles), and of course the original 2nd-hand
gear used by the big stars. That will be seven digits, then, for the particularly well-endowed
money-bag. No joke at all. Seven digits – that’s $ and €, not Yen.
"Any lively joy is, too, a fallacy, a vapor, because no fulfilled desire can yield persisting
satisfaction. Because, too, any possession and any happiness is merely on loan from chance
for an undetermined time.” That would be Schopenhauer. Probably wasn’t a guitar player.

0.3 How the Vibrations of the Strings arrived in the Underworld 0-15
0.3 How the vibrations of the strings arrived in the Orkus
There’s scarcely any test-report on guitars that does not praise the exorbitant vibration-
propensity of the investigated electric guitar: “the design shows considerable resonance
properties; after each string attack it vibrates intensely and clearly noticeably ” [G&B 9/06].
Or: "From a vibration point-of-view, the MTM1 ranks at the highest level, since the whole
structure resonates intensely into the last wood fiber after each string attack, resulting in a
slow and continuously decaying sustain” [G&B 8/06]. Or: "Combined with the given open
freedom of vibration, we arrive at a brilliant sonic image” [G&B 8/06]. Or: "Less mass can be
more easily be made to vibrate” [Luthier Thomas Kortmann, Gitarrist.net]. Or: "At Fender they
even proceeded to build the bodies from several pieces of wood. … Of course, the ability of
the wood to resonate will be reduced by such a number of pieces varying in size, as well".
Und o.a.: "At the time, the fact that ash also has almost optimal resonance characteristics was
noted with appreciation. It doesn’t even bear contemplating what had happened if, back in the
day, Leo Fender had opted for mahogany” [Day et al]. Or: "Clearly noticeable, both Strat and
Tele show very good resonance properties right to the outermost wood fibers” [G&B 4/06].
Take note: this is about solid body electric guitars, and not about acoustic guitars. The clearly
noticeable vibration of the guitar is taken as a criterion for quality. Let’s have the father of
the solid body, Lester William Polfuss, have a word here: "I figured out that when you've got
the top vibrating and a string vibrating, you've got a conflict. One of them has got to stop and
it can't be the string, because that's making the sound." Mr. Polfuss wanted only the string to
vibrate, and not the top of the guitar. Well, one could object that the man was a musician and
not an engineer. Still, he was a musician who, answering the question about who in fact had
designed the Gibson Les Paul, said: "I designed it all by myself". The string is supposed to
vibrate, and the body is supposed to keep quiet. Only the overly pedantic will interject here
that in fact it is the relative movement that counts, i.e. if the string remains in rest, the body
could instead … no, enough about the theories of relativity; it works better the other way
‘round. But then, what does “better” mean? What characterizes the better sounding guitar? In
his dissertation [16], Ulrich May cites D. Brosnac who realizes that a guitar made of rubber
would absorb all string energy within a short amount of time, i.e. it would not sound right.
This is easily understood but does not prove whether ash, or maple, etc. are better suited.
Obviously there are unsuitable body materials that withdraw quite a lot of vibration energy
from the string. Rubber would be one of them. But who would want to build a guitar from
rubber? Presumably, dough for steamed bread would be unsuitable, as well♣. For another
approach, fresh from the sleep clinic: a bed of a length of 1.45 m (4.75 ft) is uncomfortable
for most grown-ups, therefore a bed with 2.12 m (6.95 ft) must be more comfortable than a
bed of 2.05 m (6.72 ft). To be more specific to our field: what the luthier has learned with
respect to the acoustic guitar cannot be wrong for the electric guitar. A guitar needs to vibrate.
Right into the outermost wood fiber. Intensely and clearly noticeable.
So, what in fact is noticeable, or perceivable, for the human in general and for the guitar tester
in particular? That of course will depend on the stimulus and the receptor – but regarding
vibration, the subcutaneous Pacini-corpuscles are most sensitive at stimulus frequencies of
200 – 300 Hz, and can sense vibration amplitudes of as little as 0.1 µm. However, that also
implies that for frequencies above 250 Hz, the sensitivity increasingly drops rapidly. Sound-
shaping harmonics therefore remain mostly outside of the reach of the sense of touch, the
feeling of vibration..
♣
due to the strong „damp-ing“.
Fig. 0.1 shows the frequency dependency of the vibration threshold, i.e. the vibration
amplitude that needs to be reached such that any vibration sensation can emerge in the first
place. The exact shape of the curve depends not only on the frequency and the amplitude but
also in the area of the vibrating surface, and on the stimulated location. The shown
dependency can be seen as typical for the thenar (area below the thumb). If a guitarist, upon
plucking the strings, feels a vibration in the body or the neck of the guitar, these will be
predominantly in the low-frequency domain. If, as a calculation to check the assumptions,
we take a force at the bridge of 10 N, a mass of 4 kg, and 250 Hz as excitation frequency, we
get a displacement of 1 µm. Hence it is no wonder that noticeable vibrations may be
generated, even without any resonance-amplification.
Fig. 0.1: Vibration threshold

(“Vibrationsschwelle”).
Only values that lie above the threshold lead to a
vibration perception. According to this curve, a
vibration with an amplitude of 0.4 µm is
noticeable at 300 Hz, but not anymore at 800 Hz.
“Schwingungsamplitude” = vibration-amplitude
“Frequenz” = frequency
Therefore, the question is not so much whether perceivable vibrations can occur but how
these should be interpreted. Taking up Les Paul’s idea again, any noteworthy vibration of the
guitar body would be counterproductive. With a lot of mass (a ten-pounder Paula), we would
approach his ideal at the cost of comfortably carrying the instrument – and we would still
disregard vibration-amplifying natural oscillations (Eigenmodes). The neck of the guitar in
particular cannot be arbitrarily made heavier; it will vibrate noticeably in every guitar.
However, what would happen if we could manufacture a guitar to be vibration-free? For
comparable plucking, comparable strings would vibrate identically on every guitar of that
kind! Individuality is imperfection, and it would fall by the wayside in this scenario. For the
acoustic guitar, the luthier seeks to form the transmission factor in a frequency-dependent
fashion, and therefore makes some frequency ranges radiate better, but others weaker instead.
This way an individual sound results. The same principle could be applied for the electric
guitar, and neck and body could be made to vibrate more strongly at certain frequencies i.e. to
dissipate the vibration energy more quickly. Whether this is indeed desirable – that can only
be assessed in an overall consideration of all sound-forming elements. It would however be a
particular coincidence if it were exactly those frequency ranges that would require the
strongest damping, in which the vibration perception is especially sensitive. One thing is clear
beyond doubt: the source for the sensed vibration energy is the string. The more intense “the
whole structure resonates”, the less the string vibrates because it looses its energy to “the
whole structure” very quickly. One may disagree or agree with Les Paul’s ideas – going
against the law of energy conservation is not advisable.
Disagreeing with Day et al., however, is at everybody’s liberty: “The vibrato system itself
received a knife-edge arrangement at the six corresponding holes, such that the whole system
had a very low-friction bearing but could still conduct the string vibrations optimally into the
body. Yep, that’s a well-known path: For the ignoble goes down to the (c)orc/pus in
silence. Schiller, Nänie (Nania). Or something like that.

0.4 The Sound of the unamplified Guitar 0-17
0.4 The sound of the unamplified guitar
How does the expert test an electric guitar? By first listening to it without amplification (dry).
"It is certain, that – contrary to common belief – the desired sound in electric guitars and
basses does not predominantly depend on the pickups. Rather, the wood creates the basis. A
guitar made from plywood will not sound good even with the best of pickups. When a
customer approaches me here in the ‘Guitar Garage’ in Bremen and wants to discuss pickups,
I first listen to the instrument without an amp” [Jimmy Koerting, Fachblatt Musikmagazin].
Or: "For the first assessment of the sound quality we need neither towering amps nor
distortion devices, a small combo suffices. Of course, it would be even better to test the tonal
behavior in a quiet corner playing ‘dry’, purely acoustically, and check with regard to attack,
balance and sustain” [G&B 3/97]. How then can two guitars that differ in their ‘dry’ sound be
unable to make this difference heard via the amplifier? "Surprisingly, the differences in sound
show up to a much lesser degree when played through the amp, compared to the ‘dry’ test”
[G&B 7/06]. Compared were: Gibson New Century X-Plorer and V-Factor. From another
comparison test: "The Platinum Beast sounds (dry) powerful, warm and balanced, with a
velvety brilliance and delicate harmonics. The Evil Edge Mockingbird somehow comes
across as feeble, poor in the mids, with somewhat more pronounced bass, but instead is more
brilliant and richer in harmonics. Thanks to the hot humbuckers, everything sounds very
different when connected to the amp because – hard to believe – both instruments now sound
all but identical” [G&B 8/06].
Extreme examples seem not to be of any help here. Plywood (or even rubber!) is called into
action to serve as body-wood in order to justify the significance of, and necessity for, high-
grade woods for the guitar body. That’s the one extreme: with a totally unsuitable (highly
absorbing) body, you cannot build a good guitar. Ergo-1: the wood is more important than the
pickups. The other extreme: you switch a trebly (“underwound”) Strat pickup for a bassy,
treble-devouring Tele-neck-pickup boasting a thick brass cover, and postulate Ergo-2: the
pickup is more important than the wood. Both considerations are too lopsided.
From the point of view of systems theory, the vibrating string is a generator that on the one
hand excites the body and the neck to vibrate, both of which themselves radiate airborne
sound. On the other hand, the relative motion between string and pickup generates the
induced voltage. Airborne sound and voltage are therefore correlated – they result from one
and the same source. If the string vibration dies off already after a few seconds, the pickup
cannot make for a gigantic sustain. Or maybe it can, after all? Within certain limits it could
indeed – in combination with a suitable amplifier (+ loudspeaker). If the amplifier limits the
signal (overdrive, crunch), it actually changes the decay behavior. That’s the decay behavior
that is audible via the loudspeaker, because the decay of the string vibration is not changed,
anyway. Or is it?? Now, the situation begins to become multitudinous … and exactly for this
reason we find so many contradictory opinions in guitar literature. If guitar and loudspeaker
are located close together, feedback can certainly influence the string vibration, too. Which
may be the reason for the expert-advice to first listen without an amplifier. Still: no guitarist
will buy an electric guitar to always play in unamplified fashion. At some point, plugging-in
will happen, and now the predictions from the ‘dry’ test are supposed to by vindicated. The
probability of a favorable ending of the experiment is not entirely at zero – electric and
acoustic sounds are somehow related (cor-related!), but how exactly cannot be seen at first
glance.
Let us imagine a simple experiment: the pickups of a Stratocaster are screwed directly into
the wood – this is to fully secure them in place. Oh, you reason that this step alone already
changes the sound? Hm. Anyway, this special sound is taken as the reference. We have guitar,
pickups – and now we get to the exceptional: once we play with pickguard, and once without.
It’s a plastic pickguard so that no metal layer can cause any eddy-current damping. Is a
difference in sound audible if the guitar is played with pickguard, and then without? In the
acoustic sound: definitely yes, in the electric sound: definitely no. If the pickguard is present,
it is caused to vibrate via the guitar body. Having weakly damped natural frequencies
(Eigenmodes), it can radiate audible sound in several frequency ranges. Do these vibrations of
the pickguard act retroactively onto the strings. In theory: yes, because “All things are bound
together. All things connect.” (causality statement by Chief Seattle, sometime in the mid
1800’s). Practically: no, since between pickguard and strings we have the guitar body which
weighs in at a serious multiple of the mass of the pickguard. The string vibrations are
influenced by the pickguard to such an insignificant extent that the electrical sound is not
audibly changed. The radiated airborne sound, however, does of course change. Or another
example: a singer performs in a concert hall. Listener A listens in the concert hall while
listener B listens from the neighboring room via an open door. Now we close the door – what
does change? A lot for listener B, practically nothing for listener A. Very theoretically we can
again call for Chief Seattle’s lemma, and demand a correction value for the wall absorption,
but in this case there is no practical effect, as much as we might agree the Chief in general.
What’s the singer got to do with the electric guitar? In both cases there are different
transmission paths which change the sound conducted by them in a different manner.
Knowledge about one transmission path does in general not allow for any conclusion about
the other. The listener in the concert hall cannot be certain whether the other one (The Man
Outside…) can hear anything at all. For the guitar, that implies: what can the nice acoustic
sound do for me, if the pickup coil is ruptured? Careful though, were getting again into the
domain of extreme positions. So let’s assume an incomplete sound-insulation for listener B.
He/she will then be able to give some statements: when is there singing, or a pause. Maybe
he/she even recognizes which one of three singers is in the process to try to get to the high C
at the given moment: the little one, the handsome one, or Fat Lucy. Issues with intonation will
be audible even through the closed door, as long as the insulation is not complete. And even
more so, if these issues are present in the expectation of the listener in the first place.
The thing with the expectations needs to be considered for the guitar, as well: it is astonishing
how some guitar testers fall victim to their own conviction. Irrevocable credo: "of course, the
original Les-Paul-mix consisting of mahogany neck with rosewood fretboard and mahogany
body with thick maple top will result in the one-and-only Les Paul sound”. That’s exactly
how this needs to be written – in this case in a comparison test for guitars♣. And then a copy
with an alder body (stigmatized with "!" in the test report) dares to sound good – even
commands the tester’s respect. "... come alder … come mahogany, it is anyway able to
convince us with a first-class sound”. Well, well, don’t you exaggerate! Don’t forget: we are
talking about alder here! And lo and behold: "...all in all a bit subdued and a little bit shy.”
There you go – typically alder! However, oh great Polfuss, what happens only one column
later, with the Fame LP-IV that’s also in the test group? "Those who dig a typical powerful,
no-frills Les Paul sound, you should check out the Fame LP-VI. It indeed sounds the most
authentic. Its sound is very close to the original in every range.” Question: according to the
test-info, the Fame LP-IV has a maple neck, an oak fretboard, an alder body and a mahogany
top. Did I get something wrong here?
♣
G&B 7/02

However, why don’t we postpone the discussion on materials to later and return to the
question: how far is the conclusion from the ‘dry’ test to the electric sound legitimate?
Apparently there are “robust” signal parameters that win out in any signal path, and “fragile”
parameters that change as they pass through a transmission medium. Pitch is fairly robust:
whether the guitar is amplified or not, you will hear if it is in tune. Maybe not to the last cent
(of pitch!), as the psycho-acousticians know, but with adequate accuracy for these first
considerations. The balance between treble and bass, however, depends on the tone control
settings on the amp – that is a trivial as it is uncontroversial. As hard as the sound from the
guitar body may try – it will loose out to the fully dimed bass knob. “That’s not what we
mean”, the expert may object, “in the ‘dry’-test I listed to the foundations of the sound – to
the soul of the wood.” Now, please: dear physi-cists and psycho-cists, don’t you get malicious
here! A guitar tester does not have to have too much of a grasp of either physics or
psychology, and he may present such a statement. The soul of the wood does not present
itself prima facie, though. Many séances are required during which the spirit can permeate the
matter. A lot of knocking on wood will be necessary, a tuning fork will have to be pressed
against the solid body of a Stratocaster (at least according to Fender advertising), and ear-
training over many years will be mandatory. We should be able to reach a consensus at least
when it comes to this latter point, shouldn’t we? The discussion is, after all, not supposed to
be about the guitar-o-phobe agnostic suffering from chronically progredient dysacusis. It is
about the more or less pronounced aficionado of the instrument – who, with a more or less
extensive auditive experience, may indeed hear details in the sound that are not accessible to
the layperson.
Enter the following problem: how do we describe such details in the sound? That is a classic
task of pyschophysics and psychometrics, and it often leads to a misunderstanding just as
classic: a verbal description (dead, boxy sound) will be rejected at the scientific docking site
as too ambiguous and imprecise, just as the exact description (degree of amplitude modulation
of 8.43% at 944 Hz and with fmod = 6,33 Hz) is rejected by the musical/mystical faction as
figment-y and way too abstract. Any proposals of compromise trying to connect the two
worlds are consistently dismissed by both factions. Well then: rather than talking about the
soul’o’wood, quite often a dead, or lively, sound is cited. How are dead matter and alive
matter different? Alive matter will move! Ah … you object already now because the pencil
dropping from the table would then be alive? O.k. let’s then turn to the basic philosophical
consideration of life in particular and of existence in general … NOT! Alive means
movement – done and dusted! To translate that to the guitar: an artificial tone with strictly
harmonic partials that all decay with the same time-constant – that will sound dead.
Conversely, if the partials decay with different speed and with various beats, a sensation of
movement and lively-ness will result. Here, the term “movement” may certainly be looked at
in its original meaning as change of location: as a sound source changes its location in a
(sound-reflecting) room, time-variant comb-filters make for differences in the signal
spectrum, and the movement in space causes the “movement” in sound. In primeval days it
presumably was conducive to survival to prioritize moving sound emitters over ones fixed in
place, and at the same time early linguists discovered that speech sounds can carry
information only if they include change. Without entering too far into foreign territory: there
would be sufficient reasons why human hearing is constantly on the hunt for spectral changes.
Even if the electric guitar is somewhat younger than roaring tigers and vandals going
“Arrrghh!”, the hearing possesses this ability to analyzes and it will use it. A lively tone rich
in beats sounds more interesting than a dead one – at least as long as instrument-typical
parameters are being kept.
Similar to the string pitch, beatings of partials can be rather robust towards the transmission
parameters, and therefore it certainly is imaginable, that the expert can derive criteria of the
electric sound already from the ‘dry’ test. Now, what does this robustness of the signal
parameters depend on? Frequency-dependent signal parameters, such as the spectrum, loose
their individuality if the corresponding frequency-dependent system parameter (the
transmission function) has a similar shape. Three examples:
1) psycho-acoustics [12] describe the balance of trebly and bass-y spectral contingents with
the perceptional characteristic “sharpness”: sounds with an emphasis on treble have a strong
sharpness. Turning down the treble control decreases the sharpness. Significant for the
calculation of the sharpness is not so much the spectral detail, but the (smoothed) shape of the
spectral envelope. To be more precise: sharpness is derived from the weighted loudness/pitch-
diagram which will capture the frequency range relevant for the electric guitar at merely
around 20 sampling points. Using the same spectral resolution, transmission frequency
responses of guitar amplifiers may also be represented (Fig. 0.2). Looking at the relationship
between the two datasets we can conclude that the sharpness of the ‘dry’ guitar sound will in
general not correspond to the sharpness of the amplified sound. Put another way: changing the
controls of the amplifier, we can change the sharpness of the sound – from this angle,
sharpness is not a robust signal parameter.
Fig. 0.2: Tone control of a Fender amplifier (transmission factor). The points at the upper picture frame mark the
critical-band grid (discretization of the abscissa for calculation of sharpness).
2) Beats between partials may be described as amplitude fluctuations in the time domain,
while they can be seen as sum of closely adjacent partials in the frequency domain. For
example, two same-level partials of slightly different frequency (e.g. 997 Hz and 1003 Hz)
lead to the perception of a 1000-Hz-tone fluctuating in loudness with 6 Hz [3]. To change this
beating, a highly frequency-selective operation needs to be carried out that would be untypical
for tone controls on amps. As such, beats between partials are therefore robust relative simple
tone-control networks.
3) The spectrum of a quickly decaying sine tone (Fig. 0.3) is predominantly limited to a
narrow frequency range. Changes in the decay characteristic will therefore need to be carried
out also via highly frequency-selective changes. In other words: a linearly operating, guitar-
amp-typical tone-control network will leave the decay behavior of single partial practically
unaffected; the decay behavior is robust in this respect.

Fig. 0.3: Decaying sine-oscillation, f = 1000 Hz, time constant τ = 0,3 s.
These simplified presentations do need to be supplemented by a few points: it’s not just the
transmission factor of the guitar amp that changes the spectrum given by the stings – the
loudspeaker (incl. its enclosure), too, acts as a filter, and in the detail its transmission curve is
more frequency dependent than that of a tone-control network. The speaker membrane does
not reach the high resonance Q-factors of decaying guitar partials; for that it would have to
itself produce clearly perceivable tones – which it exactly does not. The last filter in the
transmission path is the room with it reflecting surface. Its effect is not entirely negligible
even for the ‘dry’ test, and when playing through amp and speaker, the speaker distance
weighs in as another variable. However, as long as we stay within close range of the
loudspeaker, the effects of the room may be regarded as being equivalent for both playing
situations.
Special consideration needs to be given to those effects that result in more than what a simple
tone control does. The addition of artificial reverb can extend decay processes and feign
liveliness that is not included in the original signal in such a form. Chorus/phaser/flanger are
time-variant filters of a high Q-factor, and their use always targets changes in the fine-
structure of the partials. Compressors (in particular the multi-channel variant) change the
decay constants of individual groups of partials. Overdrive has similar effects but adds extra
partials. It is thus certainly possible to influence the signal parameters that have been
categorized as ‘robust’ above. Still, without radical effects we can be successful within certain
limits to infer the sound of the amplified electric guitar from the unamplified guitar. Which
of the many beat- and decay-parameters, however, would be important for that ’good’ sound
… that is only appraisable implicitly, in the best case. Moreover, we then get into the wide-
open field of temporal and spectral masking [12], and therefore we can only draw the
fundamental conclusion that the sound of the unamplified guitar should in principle not
be evaluated. In particular in view of the expert’s special knowledge (that has been
accumulated over decades), and his/her specially trained ear, this rule does allow for
exceptions … in individual cases, and for that expert, the ‘dry’ test may reveal “everything”,
after all. The group of such experts who may take advantage of that exception comprises:
guitars testers of all guitar magazines, all guitar sales personnel, all guitarist who have had, or
have wanted to have a guitar for more than a year, an all listeners (both CD and vinyl) who
have the exact sound of Jeff Beck’s signature guitar still ringing in their ears (see Chapter 7).
And please, dear experts that now have received such extensive legitimization for your
obviously indispensable ‘dry’ tests: we now should have consensus that the assessment of
tactile vibrations is nonsense, shouldn’t we?!
Concluding the topic of guitar testing, here are a few further citations:
Yamaha Pacifica guitars (maple neck, alder body) in a comparison test: "Acoustically, the basic
characteristics of the Pacificas are readily comparable. Plugged in, however, they differentiate
themselves rather clearly corresponding to the pickup complement” [G&B 6/04].
Gibson Les Paul Faded Double-Cutaway: "Right from the first plucking of the strings, it is clear that
there is less damping of the resonance characteristics of the wood due to the low-key varnishing, The
guitar resonates from head (machine heads) to toe (strap-pin) so intensely that I could even sense it in
my own body” [G&B 6/04].
Ibanez IC400BK: "The slight underexposure of the E6-string found in the ‘dry’ test is suddenly gone
as the pickups provide support.” [G&B 6/04].
Squier-Stratocaster, comparison: mahogany body vs. basswood body: Using the middle and neck
pickup, respectively, both guitars sound nearly identical.” [G&B 5/06].
"Grabbing the Pensa-Suhr guitar and playing it unamplified, any reasonably trained ear immediate
hears what it’s at. … Both seated and standing up, you feel the fantastic vibration behavior of the
excellently tuned woods in your belly” [Fachblatt, 6/88].
"Despite the humbucker, a Strat can (sonically) never become a Les Paul” [G&B 2/00]. Ozzy
Osbourne about Joe Holmes: "I don’t actually like Fender guitars. But Joe gets this fulminant Gibson
sound with them” [G&B 2/02]. "Jimmy Page recorded the complete first Led Zeppelin album using a
Telecaster. The guitar sound on that album is exactly that of a Les Paul.” (G&B Fender special issue).
Mark Knopfler: "If I want a fatter sound, I’ll use my Les Paul – it is simply more dynamic. That does
not mean, however, that I could not do the same thing with a Stratocaster." [G&B Fender-Heft]. Gary
Moore: “some people believe that you hear a Stratocaster on 'Ain't nobody', but in reality it’s my own
signature Les Paul.” [G&B 7/06 p.91].
High mass of wood (3,9 kg): Due to the big mass of wood, the response seems to be a bit ponderous,
and the notes do not get off the starting blocks as quickly. [G&B 7/06].
Still heavier (4,15 kg): The guitar resonates intensely, has a direct and dynamic response; every chord
and tone unfolds crisply and with great liveliness [G&B 8/06].
Despite the enormous mass of wood (3,85 kg) almost every note responds crisply and dynamically,
unfolding very swiftly [G&B 7/06].
"Less mass can be made to vibrate more easily” [Thomas Kortmann, gitarrist.net].
A slender guitar body makes for a slender tone [G&B 7/02].
Thinner body = less bass [G&B 4/04].
Fat neck = sonically advantageous [G&B 8/02]. Thin neck = round, fat sound [G&B 10/05]. Thin
neck: The less mass that needs to be moved, the more direct and quickly response and unfolding of the
tone get off the starting blocks. [G&B 3/05]. Crisp and direct in the response, every tone gets off the
starting blocks quickly and with great livelihood, despite the immense mass of wood (that indeed
needs to be set in motion to begin with!) [G&B 9/05]. A thin neck has no acceptable vibration-
characteristic whatsoever [G&B 3/97). Of sonic advantage is that the neck weighs in with a lot of
mass [G&B Fender special issue). The Ibanez JEM 777 features an extremely thin neck-construction:
the sound character is powerful and earthy [Fachblatt, 6/88]. Of course the neck shape also
contributes to the sonic character [G&B, 12/06]. What is not true at all is that fat necks sound better
than thin ones. I have built the same guitar with a fat neck and a thin neck, and could not detect any
difference [luthier Thomas Kortmann, Gitarrist.net]
Nay, that's past praying for [Shakespeare].

0.5 Growse-Glowsock 0-23
0.5 The Growse-Glowsock Affair
She already expected us. Second table on the right, as arranged. We, that was yours truly,
specialist editor at Guitar Licks & Tricks, and Dick Johnson, our photographer (mind you,
this is famous D.J., not the mute dweeb from Chapter 0.1). We had scheduled the date for
15:00, and she was on time. We as well, of course. It’s not an every-day occurrence that you
get the opportunity to meet the marketing CO of a famous tube distributor. Ms. Ann-Cathrin
Growse-Glowsock, Psy.D., gave a most professional impression already from a purely visual
point of view. “Whewww … she lights me up like an AC30-deployed EL-84”, Dick
whispered under his breath as we approached. His thoughts must have already been on-topic,
because today it was going to be about amplifier tubes. Meeting in a cafe seemed strange at
first, but Growsock (as she was dubbed in the editorial office) had already apologized for any
inconvenience: “we have such a bedlam in the test area right now, I don’t dare let anybody in
there.” Of course we suspected that they had some exquisite new hyper-tubes – still secret.
We hoped that we could elicit the odd detail out of her.
Having done the introductory prelude (here’s my card – can I please have yours?) and a secret
look to the bounty (indeed: “CO Marketing”) we quickly got down to business: “Dr. Growse-
Glowsock …” “ Please, that’s Ann-Cathrin for you guys!” “Thank you! Ann-Cathrin, how do
you manage again and again to find these great NOS-tubes?” “Well – that’s a most difficult
question – and right at the start!” she smiled flirtatiously, “that would be what we shall ask
Ed, our director of purchasing – he’ll join us later.” Wow, this is gonna be a blast! “So you
produce all the tubes here in Valleymoon?” “Oh no, of course not, we have a global network
of suppliers. The US, Russia, China, Cambodia, Algiers, Laos, and many others.” Of course
… stupid question … wherever you can buy quality products. By now Dick had set up his
camera and butted in, in his inimitable fashion: “your super-bulbs are really so GRAND, I‘m
over the moon with them.” "Well, right now the KT-88 is indeed a top seller,” she stand-
offish-ly replied. My God, Dick – she’s a manager with a doctorate … could you find an any
more dopey come-on? Another try: “ Ms. Growse-Glowsock, with a doctorate under your
belt, do you fare better in this man’s world? You, as a woman …” Oh sh.., that’s not it, either.
“I mean, not all of your competitors have staffed their exec-positions with university
graduates, have they?” Phew …in the nick of time … “Would you pose this question to a
man, too?” Her green eyes were painted every so lightly with this glittery stuff (well, not the
eyes, but just above) … it looked really good, even though she squinnied now and then. Green
glitter-eyes with that ginger mane … oh, man …well … thank God she was not looking for an
answer but continued: “actually, I first took courses in geography. Economic geography, to be
precise. But during the 10th semester I realized that I was not going to get hired anywhere. So
I broke off my studies and worked some casual jobs for a while.” "Was that already in the
electronics sector?” “No, that came but later. I worked at the university’s copy-shop. That’s
where I took notice of a psychology professor. Or rather, he took notice of me.” There was a
bit of a mischievous smile on her face. The old story: I once had a girl, or should I say, she
once had me (L/McC). Psych-Prof … ‘course, as a specialist journalist, you can’t compete.
“So you got your doctorate with the psychologists?” “Yes! At the Institute for Speculative
Psychology, with Professor van Bonner. You know him?” “Sorry, no. Speculative
psychology?” “Right: what might Schopenhauer have said to Nietzsche? That was the subject
of my thesis.” “Very interesting. So what would he have said?” “Not much! Which is why I
didn’t have to write all that much – tee-hee!” Now the green ones smiled again. “Right,
always economize,” Dick barged in again. Before he could add a ‘typical female!’, I kicked
him under the table. That must have hurt because he already hauled off for a counterstrike –
but at that moment an immense behemoth approached our table, and Dick was distracted.
"Ah, here’s Ed now, our managing director of production,” she exclaimed with a honey-tone
voice,” Eddy-darling, sweety – we’re here!” Must have been easily some 250 lb of sweetness
that came crawling towards us. Designer specs, grey braid at the back of the neck: Eddy.
“Director of purchasing or of production?” I quietly asked because my memory couldn’t fit it
all together. He heard it and introduced himself right away: “Edward Growse, purchasing and
production.” Understood – it does happen in big corporations that a board member takes care
of two divisions for some time ... or maybe cost-saving measures? Whatever. “I tidied up, we
can go in” Ed reckoned, saving the day for Dick, because a few pics had to be taken (location,
location ... or genius loci, as Growsock probably would have put it). We paid up and piled
into the SUV. “Fasten seatbelts, please – Eddy really hits the breaks when a speed camera is
indicated!” Sure, we’d do the same. And off we went: on the road to Tubilic, Ltd.
"Ann-Cathrin, you mastered in psychology but now work in tube marketing? …” “Indeed.
You know, when after 16 semesters I checked out the employment market for speculative
psychologists …” “…you realized that …” “exactly! And then Hans, my professor, had less
and less time for me because of his wife, so it was a lucky coincidence that Eddy and his band
played a gig at the university.” Such is life – hence the double name. Spontaneous idea: back
home they would have a vacancy for the chief district executive … no, maybe not. The west
of the city almost silently rolled past the V12, the streets became narrower and more
contorted. As Ed pulled into the driveway with the triple garage, a giant Great Dane yelpingly
jumped up to the fence. They wouldn’t have a little nosh for us before we … “Jeez, you got a
lovely place here – that tube business brings in some heavy dough, doesn’t it?” Oh no - who
had made this retard my photographer! Luckily, Growsock had already gone ahead to the
door, and Ed pretended not to have heard anything: “Let’s go downstairs to the test-field right
away.” Yikes! So THIS WAS Tubilic! He boxes and she types up the invoices! That’s almost
like we had seen it in Tonopah at the pickup guru’s … Never mind, we’ll see it through now.
Ed already opened one of the many basement doors. Neon tubes flickered to life, bathing
meticulously stacked-up small cardboard boxes in cold light. Tidy it was – gotta give the guy
that. Gold Lions, old GECs, new Tungsols, everything accurately piled up. “Ed. You have …”
the remainder of the sentence was drowned out by infernal bellowing that all of a sudden
burst forth from the other side of the door. “Bonzo would like to say hi”, Eddy remarked with
a malicious smile, and opened the door. This was the Scottish version of the Great Dane.
They stand about a yard high at the shoulder. In their younger years. Fully grown that may
increase to 4 ½ ft. The Giant Scottish Great Dane will measure yet another foot on top of that,
at the very least. As long as they do not bob up … it won’t, will it? … Noooo … of course it
will. The dog was completely overjoyed, woof-woof, pant-pant … if at least not that 2-ft-
tongue … and that deafening roar in the reverberant basement … the things you gotta endure
as journalist … “Has he happened to waggle down an expensive tube sometime?” Well then,
Dick could indeed also shoot good questions, although the present situation necessitated a lot
of accompanying gesturing – multi-medial communication, in a way. But as quickly as it had
come, the episode was over: upstairs Growsock clanked a pot, there was one last "Ch-ch-ch",
and the dog was gone. I’ll have to get at least 20% hardship allowance for all this … or else
the editor in chief will have to do the job himself next time.
Where were we? “Ed, which is the better tube, the 6L6-GC, or the 5881?" Again, Ed
displayed that malicious smile: “That would be the 5881 – we get a better markup on that.”
Big laughs. “But don’t you write that in the article. On the other hand: most people know that
anyway, don’t they?” He added: "6L6-GC for your average moron, 6L6-WGC for the one
seeking to spend a bit more, 5881 as premium-merc, and 5881-WXXS for the snobs. The
insides are always the same”. Ed’s laughter was suddenly interrupted by very enraged green
eyes that must have finished taking care of guzzle-guzzle and wanted to attend to the visitors.

"His humor takes some getting-used-to, but with tubes he’s really on top of the game, like no
other” she sought to distract. “Of course, the insides are NOT always the same: this 5881-
WXXS here e.g. is a heavy-duty-version with brown base, while the regular 5881 over there
has the black base. That is something entirely different.” Dick shot a querying glance at
Growsock and released the shutter: once brown, once black. No, he had not forgotten to put in
a color film – we are a specialist publisher, after all, and have been working with digital for
years. Digital in color, even.
For confirmation, Growsock now brought on the big guns: instrumentation! "With the 5881,
we get about 10% more compared to the 6L6" she remarked. “Occasionally – when a special
order comes in, we can even measure power. We have bought a gadget specifically for that.”
She pointed to a contraption that probably was a tube-testing device: “on the left side, the
5881 has about 10% more than the 6L6." Dick industriously kept shooting, and I decided to
enter the professional discourse: “Left, that would be …” “She means the instrument on the
left side”, Ed seconded, ”we operate in a highly targeted, concise manner. No superfluous
pleasantries. 10% more on the left, and everybody knows what is meant.” "So what does the
left device register?” “Well, the tube that has been inserted. Plug in, there we go.” Quickly,
another question – before Dick comes up with the next mischievous idea: “That will then be
the plate-current that is shown on the left instrument?” Ed was not in the mood to get a lot
into theory, though, and preferred to remain very practical: “we first do a selection process on
all tubes: those with the straight glass container get loaded onto the blue trolly, the bulbous
ones are placed on the yellow one. I think every musician has the right to selected tubes. In
fact, the guys in China should already do the selecting, but since Sinh Ter has left the export
division, we occasionally receive the tubes in a rather colorful mixture. Logn San, the new
guy, is just too …” “He is in training and will be certified soon. We ale vely ‘appy wit te tube
man’factulel.” Growsock’s humor certainly was of a different caliber. "The individual
numbers printed onto the boxes, this 34/-52, for example, that ... " "… that’s already on the
boxes. Although I think you can order the cartons without those numbers. Nobody does that,
the market demands the numbers, and we serve the market.”
This had turned out to be an interesting meeting, after all. “You imply that you have not laser-
printed these labels individually but … isn’t that deceit, somehow? Or even …” Now our
psychologist sooo got going … there was a job market for speculative psychology, after all:
“you have no clue, do you! You’re absolutely 404! A musician on stage, opening up to the
world, in a way baring his or her soul – will he or she not need the maximum in gear
performance that the market can offer? Feed selected premium ware to the combo amp,
maybe even remakes of the legendary black-plate powerhouses with the larger and longer
base – that will give him or her that vibe … no: the FORCE to be truly inspired!” “Jack, you
got shots of the meter? Maybe Annie … sorry: Ann-Cathrin … could sit beside it giving you a
smile? I’ll get into some more technical stuff with Ed, meanwhile.” She didn’t give up that
easily, though. “You sell an 6L6 for $5 to the players – they surely feel they're getting pure
rubbish. Can’t go in front of an audience with that. A 5881-JKAS at $49.90” … (Ed’s ‘with
the same junk inside’ got drowned out by the dramatically mounting fortissimo) … “that’ll do
the magic and make them play like gods. Like Clapton at 22, like Morse and Moore combined
in one person!” O.K. – she knew her stuff. “So it’s all psychology?” “Nonsense – of course an
el-cheapo tube can’t cut it like a premium tube will.” Now they seemed to switch roles;
apparently Eddy-darling wanted to remain at the wheel, too: “That needs to be clarified from
the ground up. Tubes: not just anybody can do that. We are the champions here.”
He elaborated: “a JKAS for example will give you those satin highs, with a well defined
share of bass and lots of headroom. The JRK, on the other hand, delivers particularly delicate,
mild treble, well-defined bass and powerful mids, and the BLL has those tight basses, strong
mids and satin treble, and headroom in spades. A 5881 puts out delicate treble, creamy
sustain, particularly mild mids, with great headroom. The 5881-TLT brings creamy treble,
fine sustain, powerful bass …“ he faltered “… no, that’s strong bass. Powerful bass, that’s the
5581-WNK. And, of course, the 5881-WNK/JRK-STR-highgrade. The latter combined with
particularly silky and super-clean treble.” Growsock applauded ecstatically. “That’s why I put
him in charge of production”, she remarked smugly, and immediately added “Without Eddy,
this joint wouldn’t do so wonderfully – he knows tube specs like no other.”
"And the numbers, those on the boxes?” Dick tried to dig deeper. “Numbs are for dumbs," Ed
laughed. "It’s only on the balance sheet, where numbers you have got to read.” Not a man
modeled after Leo F., then. We needed something tangible, though, to keep the head-editor-
boss out of our hair. Next try: “The JKAS features gold-plated grids, doesn’t it? That’s in
order to …” “Gold is a precious metal”, Growsock embarked. “The more precious the metal,
the more classy the sound – pure logic. You wouldn’t want to wear an aluminum ring on your
finger, either, now would you?! Gold grids, and a black-anodized glass bulb.” “…and the
longest cylinder possible,” Dick barged in, only to sulkily shut up again with an “Ouch!” "Are
NOS-tubes indeed as good as they are said to be? They’ve been lying around for a number of
decades, after all?” “In most cases, it is not possible to exactly date NOS-tubes,” Growsock
submitted sibyllinically. “We are always happy when again somebody finds a case in some
attic, and we hope that such tubes continue to be found for a long time. Myself, I just a few
months ago discovered a huge supply back in the old country, in the basement of the house
my grannie was in the process to sell. More than 1000 pieces! One has to wonder about all the
stuff that people hoard.” “And these are truly old?” ”Of course! My granny’s house was in the
area where the GDR used to be, actually very close to the SOG-Tube-combine. She always
said ‘vee haf nossing ofer herrr’, but what the little they had, sey haf nott srone avay. I was
just surprised that Ed didn’t find those tubes. He rummaged around in that basement for days
before I arrived. Wonder what he was looking at and for, my darling blind-shell!” “Main
thing is you keep turning up those antique precious tubos – I’ll sell ‘em.”
Those two truly had found their perfect work-sharing arrangement. Ann-Cathrin and Edward:
enterprising, slaving away serving the discerning guitarist, supplying premier tubes. Their
business was indeed going well, although … “The competition does not concern you at all?”
“Well, the guys at TOD, The Other Distributor, they do niggle us. But the grapevine says they
are not getting a grip on their personnel expenses. We have a different scenario here.” The
green ones were gleaming again. “Plus, we do have some big names under contract, our party
really rocks! What’s-his-name – no, musn’t tell you who – buys three new quartets after every
gig … and he’s gigging almost daily. What wicked endorsement!” “Huh? Doesn’t
endorsement imply something like sponsoring? The guy pays for his tubes?” “Sure, his roady
was a bro’ in the old commune – convinced our man that this brown, way-cool – no: way-hot
sizzle only is on when he's burning our prewar-MOV’s. We call it an endorsement because, in
a way, he’s endorsing our V12. And that’s just him alone! That is so cool. That endorsement,
that is so … so …” “Von Holen?” “Right!”
Dear Ann-Cathrin Growse-Glowsock, Psy.D., dear Growse, Edward – thank you for having
us. And special regards to Stronzo, or whatever its name was.

Counter statement, on behalf of Ms. Glowstock, Phy.D.
In the so-called “pre-release of Physics of the Electric Guitar”, a series of untrue allegations
about me are included. In this respect, I state:
1) My name is not at all Growse-Glowsock but Grous-Glowstock.

2) I have done studies neither in economic geography, nor in speculative psychology. Correct
is rather that my doctorate had the subject: “The difference between being in itself and being
per se from the point of view of solipsism – and the corresponding criticism by
Schopenhauer”.
3) My assigned doctorate supervisor was neither an alcoholic nor was he “Prof. Hans van
Bonner”. It seems there was confusion with Edward Grous’s student band “Van Bonjovous”.
4) The “pre-release” creates the impression that I had red hair. Correct is that I am blond; a
natural blonde, all over.
5) The “pre-release” creates the impression that I had relations with my doctorate supervisor
that ended at the intervention of his wife. Correct is that his wife did not even know about me
at the point in time.
6) The “pre-release” creates the impression that our company would select tubes merely
according to color and/or shape. Correct is that we certainly select according to other aspects.
For this, we deploy expensive special equipment.
7) The “pre-release” creates the impression that I would not know what is indicated on the
“left instrument”. Correct is that I know very well that “mA” is indicated there.
8) My grandmother did not live in the GDR, but in Poland; she hailed from Upper Silesia.
Never were any tubes found in the basement of her house. She passed away already 11 years
ago, not “a few months ago”.
9) The “pre-release” creates the impression that 50% of our company’s tubes would be
rejects. I state: this is untrue. 50% of our tubes are not rejects.
10) Edward Grous and I do not drive a V-12 but an S-63 that, according to the manufacturer’s
specifications, has not 12 but 8 cylinders (source: WWW.Mercedes-AMG.com).
11) The “pre-release” creates the impression that we would gain economic advantages from
“Von Holen”. Correct is that we do not know “Von Holen” at all.
August 24, 2010, Anna-Katerczyna Grous-Glowstock
Statement by the author:

Applicable law requires the publication of a counter statement without appraisal of its content.
I wish Ms. Glowstock that she may recognize with Schopenhauer that her being in itself and
per se is not so terrible, after all.
August 25, 2010, Prof. Dr.-Ing. Manfred Zollner
1 Basics of the Vibration of Strings
As a stringed instrument, the guitar belongs to the subgroup of composite chordophones/lute

instruments/crossbar instruments. The strings form frequency-determining oscillators; they
radiate their vibration either directly as airborne sound or – after conversion into an electrical
signal by the pickups – via the guitar amplifier. Being a mechanical oscillator, the string is
briefly fed energy by a plucking action … not a lot of energy but enough to entertain an
auditorium even without an amplifier. It would actually be possible to heat up one liter of
water to boiling temperature using this plucking energy: to achieve this objective, the guitarist
would have to pick the string about 60.000.000 times. That sounds worse than it is – picking
the string 5 times per second it would take about 2 years if we assume that no break is taken,
and that the heat-insulation is perfect. Old Sisiphus would be happy to “enjoy” such working
conditions. Admittedly, approaching the topic of producing art from a mechanistical/
operationistical angle receives ambivalent assessment from the involved research disciplines.
Elementartistic schools of though have to put up with being by the gestalt-psychologists that
the whole is more than the sum of its parts, after all. It doesn’t really help to counter the
insight “Hendrixian genius is more than pure superposition of vibrations” with the
existentialistic appearing question “yeah well – and where is he now?” … all too different are
the doctrines. The following considerations therefore target exclusively vibration mechanics –
as a part of the whole … as an essential part of the whole.
Translator’s remark: in this chapter, often the bridge and the nut of the guitar are taken as the points between
which the guitar string vibrates. Of course, all basic considerations apply to the fretted string in the same way –
the string then vibrates between bridge and fret. This is not always explicitly indicated, and therefore the term
“nut” should be considered to appropriately include the term “or fret”, as well.
1.1 Transversal waves
The strings of an electric guitar are made of steel, with its density ρ just below 8 ⋅103 kg / m3.
A steel string with a diameter D is stretched to the length L by applying the tension force Ψ.
Fretting the string on the fretboard shortens the length. Typical lengths are just shy of 65cm
for the open (unfretted) string (= scale M). Plucking the string (with a finger or a pick)
displaces the string in the transversal direction; subsequently there is a free, damped vibration.
After the plucking-release, a transversal motion (transversal wave) propagates from the
plucking position in both directions of the string. The propagation speed c of this wave
running along the string is (with ρ = 8 ⋅103 kg / m3):
Propagation speed
Given a string of a diameter of 0,35 mm and a tension force of 50 N, c calculates to 255 m/s.
However, this propagation speed (in the direction of the string) must not be confused with the
velocity that the string oscillates back and forth with in the transverse direction. To avoid any
confusion, the transverse velocity is termed particle velocity v. More detailed investigations
reveal that c is not constant but depends on the frequency (dispersion); more about this in
Chapter 1.3.
© M. Zollner & T. Zwicker 2002, 2020 Translation by Tilmann Zwicker

1-2 1. Basics of the vibrations of strings
Moving with the propagation speed, the transversal wave runs off in both directions and is
reflected at both ends (nut and bridge, respectively). As a reflection, it then returns to the
point of origin. We may imagine and model the process of reflection as a superimposed signal
originating from a mirror source positioned behind the end of the string (Fig. 1.1). In this
model, the primary wave excited by the plucking runs beyond the end of the string (i.e. it is
not reflected), but an additional superimposed (added) mirror wave runs opposed to the
primary wave. At the fixed end of the string, both waves meet. It is obvious that the
displacement of the mirror wave needs to be in opposite phase to the primary wave such that
the end of the string indeed remains (ideally) at rest and immobile. This phase reversal is
valid at both nut and bridge in the same way.
Fig. 1.1: Propagation of a transversal wave on a clamped string.
As the reflections arriving from nut and bridge reach the origin-point of the plucking, they
continue further, are then reflected again at the respective other end of the string, and run back
to the plucking point with the original phase. Arriving there after having covered 2L, one full
period of the fundamental oscillation T has passed. The reciprocal of T is the fundamental
frequency fG of the string. A steel string of a length of 0,65 m and a diameter of 0,35 mm
oscillates – at a tension force of 50 N – with a fundamental frequency of 196 Hz (note G3).
The frequencies of the open strings (in regular tuning) are E = E2 = 82.4Hz, A = A2 = 110Hz,
d = D3 = 146.8Hz, g = G3 = 196Hz, b = B3 = 246.9Hz, and e' = E4 = 329.6Hz.
The fundamental frequency of the string depends on the tension force Ψ, the density ρ, the
diameter D, and the length L. Quadrupling the force, or halving the length, or halving the
diameter, respectively, doubles the fundamental frequency:
Fundamental frequency
The tension force Ψ required to obtain a certain fundamental frequency calculates based on
the length L of the string, and on the material data of the density ρ and the diameter D.
Fundamental frequency and string-length appear as a product; given a string tensioned with a
constant force, fundamental frequency and string length are therefore reciprocal to each other:
Tension force
Because the actual oscillation processes are rather complicated, idealizing models are
employed. In the simplest case, planar polarization, frequency-independent propagation
speed, absence of losses, and ideal reflections are assumed. The string is described as a linear,
time-independent LTI-system.
Translation by Tilmann Zwicker © M. Zollner & Tilmann Zwicker, 2002 & 2020
The periodic repetition caused by the reflections can be seen as a (temporal) convolution of
the excitation impulse with a causal Dirac-pulse. Causal means that the signal is zero for the
negative time axis. A causal Dirac-pulse contains equidistant Dirac-impulses for . A
temporal convolution corresponds, in the spectral domain, to a multiplication of the excitation
spectrum with the spectrum of the causal Dirac-pulse. This latter spectrum necessarily is
complex, since the time-function (causal Dirac-pulse) is neither odd nor even (mapping
theorem). Using partial fraction decomposition, it can be shown that a co-tangent-shaped
spectrum of the imaginary part is linked to the causal Dirac-pulse; the spectrum of the real
part is a spectral Dirac-comb. This complex spectrum would have to be multiplied by the
excitation spectrum – however this is still too complicated for most considerations.
For this reason, further idealization is in order. The (un-damped) oscillation is not induced at
t = 0 but continues from the infinite past to the infinite future. The period of the oscillation
may be developed into a Fourier series since it is in a steady state with regard to its
periodicity. A line spectrum results as the spectrum of the oscillation, with the frequency
lines at the integer multiples of the fundamental frequency.
This way, the overall oscillation can be seen as the sum of superimposed (added) single tones
– they are called partials or (because of the integer frequency relations) harmonics. The
fundamental is the 1st harmonic, with the 2nd harmonic located at double the frequency of the
fundamental. In music, the 2nd harmonic is called the 1st overtone. This terminology extends
to the higher harmonics correspondingly (3rd harmonic = 2nd overtone, etc.).
Reality differs considerably from these idealizations. A line spectrum requires a periodic
signal of infinitely long duration. In signal theory, the term ‘periodic’ implies that a certain
section of the signal is infinitely repeated in identical shape. However, as it oscillates back
and forth, the string looses energy, and therefore an identical repetition of any signal section
is not possible. The oscillation of the string therefore is a non-periodic signal that has no
actual line spectrum affiliated to it; rather, the spectral lines are broadened into funnels due to
the damping. The reasons for the energy loss are dissipation and radiation: the motion energy
in the string is partially converted directly into heat, and partly radiated as sound-energy. The
frequency dependent propagation speed (dispersion) – discussed more extensively in Chapter
1.3 – constitutes an additional effect that must not be ignored for more detailed investigations.
Even though the string oscillation is in fact of dispersive and dissipative character, it is still
purposeful for the understanding of the motion processes to use a simplified, idealized view.
This holds in particular as long as we only regard short sections of the time signal.
An idealized plucking will displace the string triangularly (Fig. 1.2). After the pick (or the
finger) has lost contact to the string, the latter will ideally oscillate freely and without
damping. The shape of the lateral displacement can be seen as superposition of two partial
waves running in opposite directions. Both partial waves are identical at the moment of
plucking but run away from each other in opposite directions for t > 0; the magnitudes of both
propagation velocities are equal. For t = 0, the displacement of each partial wave at the nut
and the bridge is zero; it is at its maximum value at the plucking location. The triangular
shape continues in a point-symmetrical (odd) manner at the nut and the bridge as mirror wave.
The displacements of both partial waves are superimposed to yield the displacement of the
string. The same holds correspondingly for all derivatives, e.g. for the propagation speed.

Fig. 1.2: Propagation of a triangular wave after plucking the string. The phase shift is .
The string (indicated as the bold line) is modeled as superposition of two partial waves running away from each
other. The abscissa is the coordinate along the string (length of the string); the ordinate is the lateral
displacement. A parallelogram yields the delimitation line for the string displacement (lower right). These
diagrams are not time functions!
The actual string vibration is the sum of two partial waves running in opposite directions.
Both triangularly displaced partial waves run at constant speed. The particle velocity of each
point on the string is constant per section; however, the movement in one direction happens
with a different particle velocity compared to the movement in the other direction.
Superimposing both waves yields an unexpected result: each location on the string is either at
rest, or it vibrates with the constant (!) particle velocity ±v. String locations close to the nut
or to the bridge do not vibrate more slowly but during a shorter time compared to locations at
the middle of the string (Fig. 1.3).
Fig. 1.3: Time function of the particle velocity of the

string at three different points a, b, c (see Fig. 1.2 for
comparison). From this, the time function of the
displacement of these points can be derived via
integration. For the v-spectrum, the superposition of two
line spectra with phase-shifted si-envelope results. A
temporal integration corresponds to a division by jω in
the frequency domain.
In this model-consideration it is important to distinguish the actual string oscillation

(measurable in reality), and the components from which it is put together in the model. The
partial waves may not be considered in isolation; they are “artificially generated” to support
the visualization of the concept.
1.2. Wound strings 1-5
1.2 Wound strings
The thinner strings of the electric guitar (E4, B3) consist of solid steel. If the thicker (bass-)
strings (E2, A2, D3, sometimes G3 as well) were manufactured the same way, unavoidable
flexural stiffness would result in considerable inharmonicities (Chapter 1.3). For this reason, a
thin core made of steel is wound with a helically abutting winding (Fig. 1.4). For electric
guitars, the winding consists of steel or nickel, while for acoustic guitars it is made of bronze.
Using this construction, the flexural stiffness is determined mainly by the core. The winding
merely contributes the required additional mass.
Several criteria are relevant for the relationship κ = DK/DA between core diameter DK and the
outer diameter DA: in order to reduce the flexural stiffness, κ should be made as small as
possible. However, the normal stress now very quickly approaches the limit of tensile strength
even for high-strength steel. Simple machinery steel, for example, has a minimum tensile
strength of around 430 N/mm2 (St 44). For strings, this would be not adequate at all since –
for regular tuning and in rest condition – up to 2000 N/mm2 is required here. During playing,
additional strain occurs that (in the interest of long durability) still needs to remain well below
the breaking point. Moreover, high endurance towards changing strain is demanded as well.
In addition, the string must not corrode too fast, it should not be too brittle (in order to agree
with string bending), and it moreover needs to have certain magnetic properties. Overall,
these are very challenging demands – not easily fulfilled by just any manufacturer of wires.
Fig. 1.4: Wound string.

The string-core is either round or polygonal
(e.g. hexagonal).
For most wound strings, the core-diameter measures 1/3rd to 2/3rd of the outer diameter. In
particular for the higher-frequency strings, a smaller κ-value leads to breakage, and moreover
the winding-wire would have to be bent very strongly. Higher κ-values relieve the core but
bring stronger inharmonicities, and also result in too small a diameter of the winding wire
(this also calling for issues with durability). Besides the ratio of core diameter to overall
diameter, the absolute values are significant, too. To generate a certain pitch (e.g. E2), the
heavier strings need to be (and may be) stretched more than the light strings. Doubling the
diameter quadruples the mass; if the pitch is supposed to remain constant, the tension force
also needs to be quadrupled – with the normal tension (pulling force / cross-sectional area)
remaining unaffected by this.
The winding of a string often employs round wire; flat wire is used more rarely. Due to the
oblique grooves, strings wound with round wire feel somewhat rough; strings wound with flat
wire (flatwound strings) give a feel similar to the plain strings but they sound differently.
Somewhere midway we find sanded-down strings: here the core is first wound with round
wire, and subsequently the outer sections of the winding are slightly sanded in order to reduce
the surface roughness.

On acoustic guitars, heavy strings facilitate a louder sound but require to be pressed down
onto the fretboard with more force. The signals generated by electric guitars can be amplified
to almost any degree, and therefore we frequently find, on these instruments, lighter strings
than on acoustic guitars. In fact, it was only the reduction of the tension- and thus playing-
forces by up to 50% that enabled the development of new techniques (bending strings, finger
vibrato) on the electric guitar.
Every string manufacturer offers sets of strings with different diameters – designations are
usually "heavy", "medium", "light", or "super light". For a more precise characterization, all
string diameters are in addition given in mil (1 mil = 1/1000 inch = 25.4 mm /1000). On
electric guitars, the so-called 009-set is found quite often, consisting e.g. of strings with the
diameters 9-11-15-24-32-42. However, there are 009-sets also with different gradation, for
example 9-11-16-26-36-46. In string sets with thinner strings (“light gauge strings”), the three
treble strings are solid (“plain”) while the heavier strings are wound; in heavier gauge string
sets, the G-string is would, as well.
Fig. 1.5: Tension-force of a string dependent on the outer diameter. ρcore = 7900 kg/m3, ρwinding. = 8800 kg/m3.
For the string length, 25.5" = 64.8cm (e.g. Stratocaster) was taken; shorter lengths decrease the tension force Ψ.
The effect of κ on Ψ is small. 013- and 014-string-sets are mainly found on steel-string acoustic guitars.
Fig. 1.5 shows how tension force Ψ and string diameter are related. The strings are depicted
as steeply inclined lines, with the G-string shown both with and without winding. Frequently
used diameter combinations are shown as a shallow curved line. The calculations are based on
rigid (unyielding) string-bearings. Spring-loaded bearing (e.g. a vibrato system) necessitates
higher tension forces. For frequency dependent spring effect see Chapter 2.5.2.
For solid strings, the tension force of the string Ψ is calculated from the density ρ, the
fundamental frequency fG, the (outer) diameter D, and the string length (scale) M:
Tension force of the string
Due to the air enclosed in the winding, the density of wound strings is about 10% less
compared to solid strings (given the same outer diameter):
In this formula, is the density of the winding, is the density of the core material.
indicates the density of a solid string of the same outer diameter (used for comparison), is
the average density of the wound string. κ = DK/DA = core- / outer-diameter. A more precise
consideration requires minor corrections in case the core is not round but features a square or
a hexagonal cross-section, and if the winding comprises sanded down round wire, or flat wire.
Fig. 1.6: Normal stress of the string dependent on κ. Customary values are shown in bold. The values for the
(solid) e-, b- und g-strings are marked as a circle at the right border of the graph. M = 25.5" = 64.8 cm . The
representations are valid for stiff (unyielding) string bearings; spring-loaded bearings (vibrato) yield increased
normal tension.

For solid strings, the normal tension σ (tension force / cross-sectional) calculates as:
Normal tension (solid string)
Given equal fundamental frequency and length, the normal tension does not depend on the
string diameter. If light strings seem to break more easily than heavy ones, this is due to the
additionally acting plucking force – light strings offer little overhead here. For wound strings,
σ calculates as:
Normal tension (wound string)
An average density reduced by 10% needs to be applied as density for wound strings. A
particular influence is due to the ratio of the diameters κ. Fig. 1.6 shows, for all 6 strings, the
normal tensions; towards the top, the risk of breaking the string increases; towards the right,
there is more inharmonicity (Fig. 1.7). Contrary to fracture of the string (which of course must
be avoided), inharmonicity is not inherently a bad thing – it even may impart a special
“liveliness” to the sound of the string (Chapter 8.2.5).
The inharmonicity that appears in particular for heavy strings in their higher partials is due to
the flexural stiffness. According to [1], the frequency of the n-th partial calculates as:
Spreading of partials
This formula (dating back to Lord Rayleigh) holds for solid strings with D as the string
diameter. For wound strings we rearrange the math as follows:
Here, B is the flexural stiffness that depends on the core diameter DK and on Young’s
modulus E, and m' is the length-specific mass depending on the outer diameter DA. We
obtain as the parameter of inharmonicity b:
Inharmonicity-parameter
In Fig. 1.7, several ranges are marked for b. These encompass, for wound strings, the range of
customary outer diameters, and of customary values of κ (compare Fig. 1.6). For solid strings
(lower-case letters), κ = 1 holds.
Fig. 1.7: Parameter b for the inharmonicity of partials of typical guitar strings; E-A-D-G = wound, g-b-e = plain.
For a wound D-string with an outer diameter of 36 mil, we obtain: b = 12e-5 for κ = 0.6;
Core-/outer-diameter: κE = 0,33 – 0,42 κA = 0,33 – 0,50 κD = 0,40 – 0,60 κG = 0,48 – 0,60.
Scale = 65 cm. For a scale of 63 cm, all values for b need to be increased by 13%.
Material Density ρ in 103 kg / m3 Young’s modulus E in 109 N / m2
Steel 7,8 - 8,1 200 - 220

Nickel (Ni) 8,90 199
Copper (Cu) 8,92 120
Brass (Cu, Zn) 8,1 - 8,6 ≈ 100
Bronze (Cu, Sn) 8,2 - 8,9 ≈ 110
German silver (Cu, Zn, Ni) ≈ 8,6 ≈ 130
Nylon (Polyamid) ≈ 1,2 ≈ 3,5
Table: Material-data. Steel, Brass, Bronze and German silver are available in various compositions.

1.3 Inharmonic partials
The simple ideal string has a length-specific mass m', and a tension-stiffness π2Ψ /L created
by the tension force Ψ. Conversely, the real string also includes a flexural stiffness that
impedes bending the string – this is an undesirable effect that causes dispersive wave
propagation. The heavier the string is, and the less it is tensioned, the more the flexural
stiffness manifests itself (i.e. especially in the bass strings of the guitar). To achieve an
improvement, heavy strings are wound with thin wire of one or more layers. The flexural
stiffness is then predominantly determined by the thinner core, while a high mass loading is
still possible. However, since the core cannot be made arbitrarily thin, the impact of
dispersion may only be reduced but cannot be removed. Precise analyses indicate a
propagation speed c( f ) that increases towards higher frequencies. It causes the partials to
“spread out” and loose their harmonicity to a certain degree. Therefore, the term “harmonics”
is incorrect in the strict sense of the word and may be replaced by the term “partial”.
1.3.1 Dispersion in the frequency domain
In a linear (or at least linearized) system, any oscillation shape may be represented as a
superposition of single mono-frequent oscillations. The propagation of a transversal wave is
described by the wave equation. A position- and time-dependent transverse displacement
ξ(z,t) is created along the propagation direction z, with the temporal derivative being the
particle velocity.
Wave equation
In this equation, represents the oscillation amplitude, ϕ0 indicates the phase angle at the
position z = 0 and at the point in time of t = 0, ω is the angular frequency, and k is the wave
number. The angular frequency yields the periodicity in time ; the wave number
yields the periodicity in space . For a fixed position z, the phase grows linearly with
the time t, for a fixed point in time, the phase decreases linearly with the position z:
Phase function
The periodicity in space (wave length λ) and the periodicity in time (oscillation period T) are
linked via the propagation speed (= phase speed) c:
Propagation speed
A steady free oscillation can only originate if all reflections running in a z-direction
superimpose with the same phase, i.e. if the phase shift across the length 2L amounts to an
integer multiple of 2π:
Frequencies of partials
In this equation, the propagation speed c is assumed to be frequency-independent; the partials

are then situated at integer multiples of the fundamental frequency.
However, in reality the string features dispersive wave propagation (i.e. the propagation
speed is frequency dependent): high-frequency signal run at higher speeds than low-frequency
signals, and therefore frequencies of the partials grow progressively (i.e. are spread out) with
increasing frequency. The underlying mechanism is the already mentioned flexural stiffness
that manifests itself in particular in oscillation shapes with strong curvature (i.e. at small
wave-lengths = at high frequencies). It should be noted that this is a linear effect. The
frequencies of the inharmonically spread out partials can be calculated with the following
formula [appendix]:
with Spreading of partials
Herein, the symbols mean: fi = frequency of inharmonic partial, fG = fundamental frequency

without dispersion, n = order of the respective partial, b = parameter of inharmonicity, E =
Young’s modulus (approx. ), DA = outer diameter, κ = core- / outer-diameter,
L = length of the string, ρ = density.
With a solid string of a diameter of 1,2 mm tensioned such that a fundamental frequency of
82,4 Hz results, the dispersion would detune the 20th partial from 1648 Hz to 2774 Hz – quite
a considerable effect. Using, instead of a solid string, a wound string of the same length-
specific mass, the flexural stiffness is reduced – and so is the inharmonicity. In wound guitar
strings, the core diameter is smaller than the outer diameter by a factor of about 0.3 to 0.6.
The density ρ will be about 7900 kg/m3 for solid strings, while for wound strings the effective
density is about 10% less than the core density (Chapter 1.2). Given a wound E2-string
(with an outer diameter of 1,3 mm), the calculation yields (using b = 1/8141) a spreading of
the 20th partial from 1648 Hz to 1688 Hz, i.e. by 2,5%.
In the formula of the spreading-parameter b, the length of the string L occurs with the power
of 4, while the fundamental frequency only occurs with the power of 2. If, for example,
fretting the octave halves the string-length, the percentile in detuning of the 20th partial
increases from 2,5% to 9,5% − that is from just shy of a half-step to three half-steps.
However: the 20th partial of the fretted octave lies in a different frequency range, and the
direct comparison between the 20th partial of the open string and the 10th partial of the octave
shows the same detuning of (2,5%). In other words: for a given string and the same absolute
frequency, the inharmonicity is always of the same strength irrespective of the fretted note.
The down-tuning of a guitar also increases the inharmonicity: if – in the above example –
the low E-string is down-tuned by a whole step (82.4 → 73.4 Hz) and the regular-tuned open
E-string is compared with the down-tuned E-string fretted at the 2nd fret (i.e. in both cases we
have the note E2), the inharmonicity of the 20th partial is at 2.5% for the regular tuning, and
3.9% for the down-tuning.
At this point we shall not investigate how far these inharmonicities of the partials are actually
audible; details about the topic are included in Chapter 8.2.5. [10] reports about hearing
experiments, and in [2] a computation method for piano strings is developed.

Fig. 1.8 shows the relationship between the order n of the partial and the spread frequency fi
as it can be observed for a wound low E-string of a diameter of 1,3 mm. The fundamental
frequency is 82,4 Hz, the spreading parameter is b = 1/8000.
Fig. 1.8: Inharmonic spreading of the partials for a low E-string. The thin line marks a harmonic relation.
“Teiltonspreizung” = spreading of the partials; “Teilton Nr.” = partial no.
Fig. 1.8 attributes to a given partial its spread-out frequency. For the following considerations,
however, the reverse relationship is required, as well: we have a partial at a given frequency fi,
and want to know how much was it spread, or what its frequency fn is. Fig. 19 provides the
answer. The abscissa fi shown corresponds to the ordinate in Fig. 18.
Fig. 1.9: Spreading (on a percentage basis) of partials as a function of the (spread-out) frequency (low E-string),
b = 1/8000. “Prozentuale Teiltonspreizung” = spreading of partials on a percentage basis; “gespreizte Frequenz”
= spread frequency.
Only at the discrete frequencies with n = integer we find an in-phase superposition of

all waves running in the same direction. For a full revolution (z = 2L), the phase shift amounts
to , and the travel time required corresponds to the n-fold period T of the oscillation, i.e.
. Because here the travel time for a specific phase is referred to (e.g. for the zero
crossing), the term used is the phase delay τp, with the corresponding propagation speed
being the phase speed cp.
for z = 2L;
When using the formulas giving phase delay and phase speed, we need to bear in mind that
the spread-out frequency is used. It is for this reason that the right-hand side of the equation
should contain fi but not fn:
Fig. 1.10 depicts the frequency dependency of the phase delay and the phase speed. On the
abscissa we find the spread frequency fi, i.e. the frequency where the oscillation actually
occurs. The calculation here is done for the low E-string (E2) with b = 1/8000.
Fig. 1.10: Phase delay (“Phasenlaufzeit”) (z = 2L) and phase speed (“Phasengeschwindigkeit”),
low E-string, b = 1/8000. “gespreizte Frequenz” = spread-out frequency
For the following considerations, theoretical calculations are compared to measurements. An

Ovation guitar (Viper EA-68) constitutes the measuring object – it includes a piezo-pickup
mounted in the bridge. The Viper is not a typical Ovation: its body has a thickness of 5 cm,
and being largely solid it can be counted as a solid-body guitar. The built-in amplifier was not
used; rather, the pickup was directly connected to an external measuring amplifier featuring
very high input impedance. For the majority of the measurements, D'Addario Phosphor-
Bronze strings EJ26 were deployed (.011 − .052). If not specified otherwise, the guitar was in
standard tuning E-A-D-G-B-E.
Fig. 1.11 juxtaposes calculation and measurement. There is a problem in principle with the
(or any) spectral analysis: to obtain a high frequency resolution, a measurement with a long
time duration is necessary – analysis-bandwidth and -duration are reciprocal, after all.
However, with long measurement duration, dissipation makes itself felt at high frequencies –
the signal is not in steady-state anymore. Any measurement will therefore represent a
compromise. In Fig. 1.11, the duration of the analysis amounts to 85 ms, and instead of
narrow spectral lines the result are funnel-shaped extensions (DFT-leakage). Pointing
upwards, the tips of the funnels indicate the frequency of the respective partial; the minima of
the curves are of no significance. To compare, Fig. 1.11a holds (as dots) the calculation
results for harmonic partials: the correspondence is weak – at 2,3 kHz, the frequency-
discrepancy is already as big as the distance between two partials. Fig. 1.11.b shows the
spread partial frequencies with a significantly better correspondence. Any remaining
differences will be discussed later – as will be the frequency-dependence of the level.

Fig. 1.11.a: Measured spectrum (lines), calculated harmonic partials (dots).
Fig. 1.11.b: Measured spectrum (lines), calculated spread partials (dots, b = 1/8500).
The problem mentioned above regarding the selectivity occurs particularly in spectrograms.
To generate them, many single spectra are superimposed as color- or grey-scale-coded lines
(Fig. 1.12). Herein, the level (dB-value) is entered as a function of time (ordinate) and
frequency (abscissa). However, a spectrum can never be determined at a point in time but
only for a time-interval. If we shorten the duration corresponding to this interval in order to
obtain a good time–selectivity, the spectral selectivity deteriorates. In Fig. 1.12, the time
window has an effective length of 1,9 ms, with the effective bandwidth being 526 Hz. In the
low-frequency range, red/yellow bars follow each other with an interval of 12 ms; these are
the reflections of the plucking process. The reciprocal of this periodicity corresponds to the
fundamental frequency. Towards higher frequencies, the intervals become shorter –
corresponding to the spreading of the frequencies of the partials. The quantitative evaluation
is not (yet) a good match for Fig. 1.10: as is evident, the inharmonicity occurring towards
higher frequencies is much more pronounced on Fig. 1.12.
The reason for these apparent discrepancies is found in the way the analysis is done: a
spectrogram shows the envelope shapes corresponding to given frequency ranges, and not the
propagation of a certain oscillation phase. For this reason, it is the group delay that needs to
be considered for the comparison, and not the phase delay. The phase delay is the negative
quotient of phase and angular frequency, while the group delay is the negative differential
quotient.
Phase delay
Group delay
Fig. 1.12: Spectrogram of the plucking process of a low E-string (top); computer simulation (bottom). The
resonances occurring at multiples of 1,4 kHz are due to expansion waves (Chapter 1.4). “Frequenz” = frequency.

Fig. 1.13 illustrates the differences. The

uppermost graph is the time function of
a signal resulting from 5 neighboring
tones. As this signal runs through a
system with frequency-proportional
phase, the envelope and the carrier
below it are shifted by the same delay
(middle graph). Phase delay and group
delay are equal in this case.
If there is a linear, but offset

relationship, phase- and group-delay
differ (lower graph). The envelope is
not shifted as much as a certain carrier-
phase (marked here with a dot). The
time function is not only shifted but has
also changed its shape. Due to the
frequency-independent group delay, the
shape of the envelope has, however, not
changed.
Fig. 1.13: Explaining the difference between

phase-delay and group-delay.
For a dispersive string, the group delay for a full oscillation period is calculated as:
for z = 2L; Group delay
Inserting into this equation a low value for fi (e.g. fG) yields a group delay that is – with good
approximation – the reciprocal of the fundamental frequency, i.e. about 12 ms. For higher
frequencies this value drops to about 7,8 ms which is a good match to the high-frequency
impulse distances observed in Fig. 1.12.
The lower section of Fig. 1.12, shows a computer simulation for the spectrogram depicted
above it. While the differences are not to be ignored (multiple decay processes of the excited
resonances and superimposed expansion waves make for an early unraveling of the original
line structure), we can still see already in this simple analysis a good correspondence of the
dispersive effects,
From the point of view of systems theory, the dispersive propagation may be described as an
all-pass: a linear, loss-free filter with a frequency dependent delay-time. Compared to an
ideal all-pass, the vibration energy of a real string decays – but let’s postpone dealing with
this effect a bit. Linear filters are described by their complex transfer function in the
frequency domain, and in the time domain by their impulse response. The magnitude of the
transfer function of an all-pass is equal to one for all frequencies (loss-free transmission). If
the phase of the all-pass transfer function were zero, input and output signal would
correspond (trivial case). If the phase were proportional to the frequency, all frequency
components would be delayed by the same delay time, and the system would not be termed
all-pass, but delay line. In a non-trivial all-pass, the phase is not proportional to the
frequency. The phase delay thus is frequency-dependent – for a string this occurs in such a
way that high frequencies appear at the output of the all-pass after a shorter delay than low
frequencies.
Of course, the delay time also depends on the distance traveled. Assuming precise
manufacture with place-independent mass and stiffness along the string, the string represents
a homogenous transmission line: the propagation speed is frequency-dependent but place-
independent. The phase shift thus shows proportionality to the traveled distance – at any
frequency (with a frequency-dependent proportionality factor). This assumption corresponds
well with the real string; we find somewhat more serious problems with the places of
reflection at the nut and bridge … we will have to look into this more specifically later.
It already has been explained with respect to Fig. 1.12 that for the propagation of envelopes it
is not the phase delay that is important, but the group delay. In non-dispersive systems, phase
delay and group delay are identical, but in the dispersive string the group delay is smaller than
the phase delay. As a description of the transfer characteristics of an all-pass, we typically
find the frequency response of the group delay in the frequency domain, and the impulse
response in the time domain; both characteristics are equivalent and can be converted one
into the other.
The frequency responses of phase delay and group delay are shown in Fig. 1.14. The abscissa
is the spread-out frequency fi, rather than the n-fold fundamental frequency.
Fig. 1.14: Phase delay and group delay across half the string length (from bridge to mid-string), E2, b = 1/8000.
“Phasenlaufzeit” = phase delay; “Gruppenlaufzeit” = group delay; “gespreizte Frequenz” = spread frequency.

1.3.2 Dispersion in the time domain
Guitar strings are plucked with the finger or a plectrum (pick). A slowly increasing force pulls
the string from its rest position, then this force suddenly stops, and the string executes a free
damped oscillation. The idealized time function of this excitation is a force-step: at the point
in time t = 0 the force jumps from an initial value to zero. Starting from the plucking point, a
step-wave travels in both directions. However, this wave will now change its shape due to the
dispersion: the high-frequency components of the step travel faster than the low-frequency
ones. The step is being pulled apart in both the frequency- and time-domains. From the
viewpoint of systems theory, the dispersive propagation may be modeled by an all-pass. The
latter is a linear, loss-free filter with frequency-independent transfer coefficient and
frequency-dependent delay time (Chapter 1.3.1). Transfer function and impulse response
represent the transmission-relevant quantities of an all-pass.
The impulse response of a linear system is formed by the inverse Fourier-transform of its
transfer function. Convolution of any arbitrary input signal with the impulse response yields
the output signal. According to this definition, if the system is stimulated at its input e.g. with
a step, the output signal is the result of a convolution of step and impulse response. For this
special case, a simplification is possible: the step is the (particular) temporal integral of the
impulse. Like differentiation, integration is a linear operation, and therefore the sequence of
impulse/integrator/system may be exchanged for impulse/system/integrator (commutative
law). The step-response of a linear system therefore corresponds to the integrated impulse
response, just like the impulse response corresponds to the derivative of the step response.
The model system used in the following to emulate the plucked string is an all-pass with a
step-function being fed to its input.
In Fig. 1.15.a we see on the left the measurement result from an E2-string plucked halfway
between nut and bridge (z = L/2). On the right, the step-response of an all-pass is shown for
comparison. There are clear differences but also some commonalities: the step response
permanently switches its polarity after 3 ms; this delay time corresponds to the low-frequency
group delay for half the string length. From about 1 ms – corresponding to the shorter high-
frequency group delay – we see fast oscillations. In the output signal of the piezo, the high-
frequency oscillations have more damping (treble cut). Moreover, there is a dip at 0 – 2 ms
caused by the plectrum. After 3 ms, decay processes of the longitudinal resonances appear
(Chapter 1.4) – these are not present in the simulation of the all-pass.
Fig. 1.15.a: Piezo-signal (left) and simple simulation of an all-pass (right); excitation by a step at mid-string
and t = 0. For the piezo-signal, sign and offset were chosen for best fit.
Fig. 1.15b: Piezo signal (left) and all-pass/low-pass simulation; step-excitation at mid-string and t = 0.
”Tiefpass” = low-pass.
For Fig. 1.15.b, the same all-pass as in Fig. 1.15a was used but supplemented by a simple
low-pass in order to model the treble-cut (dissipation). The amplitude of the early oscillations
can effectively be damped this way.
Two remarks regarding the bandwidth: the piezo-signal was sampled with 48 kHz. It
received a band-limitation at 20 kHz by a low-pass filter, just like the all-pass simulation. The
lower frequency limit of the measuring amplifier is 2 Hz. DC-coupling is not purposeful and
would only crate offset-problems. As a consequence, the zero-point of the ordinate is
arbitrary. Moreover, the sign was reversed such that the step happens from zero to positive
values as is customary in systems theory.
Fig. 1.16 depicts a longer section taken from the piezo signal. With increasing time, the step
is pulled more and more apart, and therefore no “period” is equal to another. Assuming, for
one revolution (z = 2L), 12 ms at low frequencies and 4 ms for higher frequencies, the step is
spread out already across several “periods” after 5 revolutions with 60 ms | 20 ms. A short-
term spectrum measured over a short time duration therefore captures signal components that
have been reflected differing numbers of times, depending on the frequency range.
Fig. 1.16: The first 60 ms of the piezo signal; E2-string, plucked at mid-string with a plectrum.

1.4 Longitudinal waves
For the guitar string, the most important wave type is the flexural wave running along the
string with a relatively slow phase speed (Chapter 1.1). However, additional waves may be
generated that all have a significantly higher propagation speed but contain – relatively – little
energy. Due to the high propagation speed, already their fundamental frequency is relatively
high. Still, theses additional waves are worth a look.
In analyses relating to Fig. 1.11, an anomaly at multiples of about 1,4 kHz showed up time
and again. At first, this was interpreted as a pickup resonance, until it transpired from
supplementary measurements that this irregularity depended on the length of the string.
Consequently, not the pickup but the string had to be the source. For bodies with dimensions
that are large compared to the structure-borne wave-length, it is known that both transversal
and longitudinal waves can appear, and combination-type waves, as well [11]. In long, thin
rods we find, on top of the tension-force-dependent flexural waves, mainly dilatational
waves (extentional waves) manifesting themselves. Their propagation speed is constant and
non-dispersive:
Dilatational wave speed
For solid steel strings the math yields cD ≈ 5100 m/s; with 64 cm as string length we calculate
a (tension-force-dependent) fundamental frequency of about 4 kHz for this dilatational wave.
In wound strings, the longitudinal stiffness depends mainly on the diameter DK of the core,
while the mass depends on the outer diameter DA. Given a length-specific compliance n' and a
length-specific mass m', the propagation speed calculates as:
Dilatational wave speed with winding
Compared to the former formula, the correction factor core-diameter / outer-diameter needs to
be considered, as well: for customary strings this ratio is about 0,32 ... 0,42. With the latter
number, the fundamental frequency f the dilatational wave decreases to about 1,3 ... 1,6 kHz,
– a good fit to the measurements. Even more precise results may be achieved by including
both the filling-factor and the stiffness of the winding in the considerations.
The resonances of the dilatational waves can be clearly seen both in Fig. 1.11 (at multiples of
1,4 kHz) and in Fig. 1.15 (after 3 ms). The following model describes the effects on the
transmission: when plucking the string, two transversal waves running in opposite directions
are generated (Chapter 1.1). The place- and time-dependent field quantities force and particle
velocity are connected via the transmission-line equations (Chapter 2), and the wave
impedance of the transversal wave calculates as about 1 Ns/m. The bridge (with its piezo
pickup) represents the line termination, it may be seen as a very stiff spring (operation below
resonance). The output voltage of the unloaded piezo pickup is proportional to the
displacement of the bridge. The latter causes a mode coupling, i.e. a small portion of the
transversal wave is converted into a dilatational wave. The input impedance of the
dilatational-wave line forms a loading of the transversal-wave line and thus influences the
transfer coefficient of the piezo pickup.
1.4 Longitudinal waves 1-21
The termination impedances of the string are seen, as a first-order approximation, as large
compared to the wave impedances (for more detailed considerations, neck- and body-
resonances would need to be looked into). The input impedance of an open-circuit
dilatational-wave line shows a co-tangent-shaped frequency dependency, including maxima at
the multiples of the fundamental frequency of the dilatational wave. At these maxima, the
possibility of the bridge acting like a spring is impeded, and its displacement (and thus the
sensitivity of the piezo pickup) is reduced.
Fig. 1.17: Measurement (left) and dilatational- wave simulation (“Dehnwellensimulation”, right); step-excitation
at mid-string at t = 0, E2-Saite. “Tiefpass” = low-pass.
In Fig. 1.17, the all-pass simulation was supplemented by a dilatational wave, yielding
significant improvement. Any remaining differences are due to the plectrum (low-frequency,
left-hand section of the figure) and to reflections at the nut (high-frequency, right-hand
section of the figure). Both these effects were not included in the simulation.
The principle effect of the dilatational-wave line on the piezo pickup may be described via
discrete elements: at very low frequencies, only the longitudinal stiffness acts, and the model
system consists of a spring. To emulate the lowest Eigen-oscillation, the mass is thought to be
concentrated in the middle of the string with a spring each left and right of it. Above this
resonance, the movement of the mass decreases due to the inertia, and half the spring forms
the input impedance. To model the higher Eigen-resonances, the string is subdivided into
more and more partial springs with interjacent partial masses. A shortening of the spring
corresponds to an increase of the stiffness such that the piezo is loaded by a spring with
continuously increasing stiffness as the frequency increases. With this, the piezo-sensitivity
decreases towards high frequencies in a staircase-shaped manner, with the steps located at
multiples of the dilatational-wave resonances.
In the upper section of Fig. 1.18, the spectral analysis of Fig. 1.11 is repeated. The low E-
string (E2) was plucked with a plectrum at a distance of about 5 mm from the bridge. The
lower section of the figure shows the result of the simulation calculation, with the dispersion-
caused inharmonicity, the dilatational-wave loading, and a simple treble damping (1st-order
low-pass) being considered. Both sections of the figure show similar irregularities at integer
multiples of 1,4 kHz – these can be explained as dilatational-wave resonances. The spectral
envelope has a similar shape in both graphs, but differences remain in the details. The most
important reason for these differences is in the frequency of the partials, the calculation of
which was based on an ideal tensioning of the string in the formulas discussed up to now. The
real nut and bridge impedances are, however, not infinite: neck, body, neighboring strings,
and many small parts all vibrate as coupled parts of a complicated system. This results in a
multitude of structural resonances.
Fig. 1.18: Magnitude spectrum: measurement (top) and model calculation (bottom).
All vibrations may not only appear in one but in three directions – and torsional vibrations are
possible, in addition. Not all resonances will substantially influence the bridge impedance but
the dilatational waves obviously have a non-negligible effect. In Fig. 1.18, the resonances of
the dilatational waves are exclusively considered relative to the frequency response of the
piezo (global envelope) – they are not considered regarding their influences on the exact
frequencies of the partials (see additional info about this in Chapter 2.5). Because of the high
Q-values of the resonances and the connected steep cutoff slopes (dB/Hz), already a
resonance-shift of a mere few permille (!) causes a clear change in the levels of the lines.
Moreover, additional spectral lines result (clearly visible at 2,8 kHz). The mechanical
parameters of a guitar cannot be established with an accuracy of in the permille-range, and
thus the limitations of the modeling come into view.
During the investigations, the model based on dilatational waves originated early on as a
working hypothesis to explain the step-shaped envelope. Three years later, an experimental
setup deploying a laser vibrometer became operational – it delivered further supporting
findings:
The laser-based setup consists of stone table weighing in at 250 kg, with a Polytec laser-head
mounted to it. A steel wire of 0,7 mm diameter is stretched in parallel to the table surface; one
end of the wire finds its support in a knife-edge bearing located on a U-brace bolted onto the
table surface. The other end of the wire is mounted to an impedance head (Brüel&Kjaer 8001)
located on a wall across the hall at a distance of 13,3 m; it measures the longitudinal force.
The wire is tensioned such that its fundamental frequency is 5 Hz; given a length of 0,65 m
for the string, the equivalent would be a fundamental frequency of 102 Hz. A laser vibrometer
sampled the vertical vibration of the wire; the same vibration was also sensed by a pickup
mounted under the string on the stone table. This “long string” was excited via a pick made of
Pertinax moving downwards in a hammer-like fashion and thus having the effect of a short
transversal displacement impulse (Fig. 1.19).
With the location of the excitation being close to the bearing of the string, the short section of
the string acts like a stiff spring; the long section of the string – with the input impedance
being the wave impedance – may be disregarded in comparison. In conjunction with this
string stiffness, the mass of the Pertinax pick forms a 2nd-order oscillation system … at least
as long as force is being transmitted. Consequently, the string displacement is in the shape of
a half-sine in the transversal direction. Fig. 1.19 shows this idealized transversal movement,
and also the results of laser-measurements for comparison. Increasing in width due to the
dispersion, this half-wave impulse runs along the string as a flexural wave; its group speed
(1.3.1) amounts to 133 m/s at low frequencies, and to about three times as much at high
frequencies. The first reflection can therefore be expected to be back at the laser vibrometer
not earlier than after 66 ms. However, as early as after T = 5,15 ms, the laser beam measures a
reflection that is repeated with decreasing amplitude in equidistant intervals. Given an overall
running path of 26,6 m, this yields a propagation speed of cD = 5165 m/s – the typical value
for (dispersion free) dilatational waves in steel wires.
Fig. 1.19: Laser measuring setup (left).

“Wirbel” = machine head; “Saitenreststück” = remaining section
of the string; “Fallhammer” = drop hammer; “Schneidenlager” =
knife-edge bearing; “Laserstrahl” = laser beam; “Saite” = string;
The graphs shown below depict measuring results of
the transversal displacement of the string with
different time-axis scaling. Below right: the
idealized shape of the curve is indicated as a dashed
line and with a horizontal shift. The excitation
happens at about 1 mm distance from the knife-edge
bearing; the measuring point of the laser is very
close to it at 5 mm from the knife-edge bearing.
“Auslenkung” = displacement.

The dilatational wave remains almost invisible to the laser-vibrometer because the laser
beam can react only to transversal but not to axial movements♣. Periodicities of T = 5,15 ms
are nevertheless measured – this is due to a coupling of the two wave types: the string is bent
at its bearing, and here the dilatational wave returning after 5,15 ms triggers a secondary
flexural wave that is visible to the laser beam.
The measurement results from the laser setup are shown in Fig. 1.20, the longitudinal force
measured at the end of the string being subject to integration. Without support-bearing, the
dilatational wave of the string (having been triggered at the left-hand bearing) reaches the
right-hand bearing after 2,6 ms. The excitation impulse is comparable but not identical to the
one shown in Fig. 1.19. With the support-bearing, the force sensor receives its first
excitation after 2,6 ms, as well – there is, however, some attenuation. Without the support-
bearing, the second impulse arrives 5,2 ms after the first one, with support-bearing this
happens already after 2 ms. The reflection of the longitudinal-force-wave is in phase at both
clamps (rigid clamping); at the support-bearing we obtain complex factors for both reflection
and transmission. The small ripples visible in the left section of Fig. 1.20 can be traced to
unavoidable resonances in the left-hand bearing; they have no special significance.
Fig. 1.20: Laser setup with/without support-bearing. The support-bearing separates the string length into two
parts 816 cm : 511 cm. The diagrams show the temporal integral of the longitudinal string-force; the unit is
Newton ⋅ millisecond (Nms). The positive sign indicates that first compression and then strain reach the sensor.
“Laserstrahl” = laser beam; “Kraftsensor” = force sensor; “Kraftintegral” = force integral; “Stützlager” = support-bearing.
In the right-hand section of Fig. 1.20, the reflections differ (in their shape) from the primary
impulse starting at 2,6 ms. Between 4,6 and 7,5 ms, three bipolar impulses can be observed:
on its path from the source (at the left-hand bearing) to the force sensor, each of them has
traversed the support-bearing once and has additionally received several reflections at the
support-bearing. In the case of a uni-polar impulse changing to a bi-polar one, we can assume
high-pass filtering. The change of shape of the impulse allowed only for the conclusion that
the reflection acts as a high-pass, and the transmission as a low-pass.
♣
Effects of lateral contraction are too weak.
Fig. 1.21 shows results of calculations using a dilatation-wave model. A 1st-order low-pass
(cutoff frequency at 1,5 kHz) emulates the transmission across the support bearing, while a
1st-order high-pass (cutoff frequency at 1,5 kHz) models the reflection. The cutoff frequency
was determined via “curve-fitting” (vulgo: we tried until we got a match). The agreement is
remarkable.
Fig. 1.21: Measurement (left) and model calculation (right); string with support-bearing, as in Fig. 1.20. The
time “zero” is shifted by 2,6 ms to the start of the first impulse. The lateral string displacement determined via
the laser (close to the left-hand bearing) was used as the input signal for the model calculation.
“Kraftintegral” = force integral.
A movable brass-cylinder (∅ 4 mm) served as support-bearing (Fig. 1.20), with the string
forming a bend angle of 5° around it. Using the parallel-axes theorem, the axial moment of
inertia of a cylinder (mD2/8) can be recalculated into the generatrix moment of inertia
(3mD2/8), with m = mass and D = diameter. Longitudinal movements of the string roll the
cylinder back and forth on its base; propelling force is the torque F⋅D, with F = longitudinal
force in the string. With respect to the longitudinal movement of the string, the inertia of the
rolling movement of the support-bearing can be recalculated into an equivalent translation
using the equivalent mass mä = 3m/8. Here, m is the actual mass of the cylinder (volume x
density), and mä is the equivalent mass to be shifted from the point of view of the string. The
source impedance of the dilatational wave arriving at the support bearing is the impedance of
the dilatational wave. Given a steel wire of a diameter of 0,7 mm, ZW is about 15,8 Ns/m (see
appendix). The wave transmitted across the support bearing also forms a loading of the latter
with ZW. The support-bearing itself is described via the equivalent mass (Fig. 1.22). Using
this, the cutoff frequency of the low-pass results as: fx = 1/(πCRW) = ZW/(πmä), and the
equivalent mass may be calculated as 3,4 g. From the latter, the mass of the cylinder follows:
m = 8,9 g. The cylinder used in the experimental setup indeed had a mass of 8,5 g – the results
of the model are nicely confirmed. Whether the cutoff frequency is set to 1500 Hz or 1578 Hz
will change the curves in Fig. 1.21 by merely by the width of a stroke.
Fig. 1.22: Electrical analogous circuit [3] of the support-

bearing. The mechanical wave impedance is transformed into
an electrical conductance; the equivalent mass is transformed
into a capacitance (FI-analogy).
The reflection- and transmission-processes may also be calculated using the equations for the
transversal wave given in Chapter 2.5; in this case the parallel connection of RW and C needs
to be taken for the bearing impedance: . This corresponds to a
high-pass HP1.
In order to localize the origin of the dilatational wave, the string was plucked at a distance of
51 cm from the left-hand string bearing (Fig. 1.23). If already the impact of the drop hammer
onto the string would trigger a dilatational wave, then the measured force integral would have
to be a dispersion-free image of the string displacement at the location of the origin. However,
the result is in fact a better match to the displacement measured closely to the bearing – the
only conclusion being that the main portion of the dilatational wave is generated only at the
time when the (dispersively broadened) flexural wave has reached the left-hand bearing. This
hypothesis is supported by the delay times depicted ion Fig. 1.23, as well.
Fig. 1.23: (top to bottom)

- Transversal displacement at the left-hand plucking
point,
- Transversal displacement at the left-hand string
bearing,
- Integral of the longitudinal force at the right-hand
string bearing
Conclusion: dilatational waves merely make for 2nd-order effects on a guitar, but their
influence may not be entirely neglected, either. The plucking action mainly generates a
flexural wave – but as soon as this hits a bearing (nut, bridge, fret), part of the flexural wave-
energy will be transformed into a dilatational wave. Dilatational waves propagate without
dispersion and create resonances in the frequency range above 1 kHz. A bearing with a small
surface towards the string will only partially reflect a dilatational wave; part of the dilatational
wave-energy will be transmitted across the bearing into the other part of the string. The
reflected portion manifests itself partially as a dilatational wave and partially as flexural wave.
Fig. 1.24 shows the significance of this mode-coupling: a string of 13,3 m length was
plucked close to its left-hand bearing, with the laser measuring-point right next to it. At 20 cm
from the plucking position, a Telecaster pickup (electrically loaded with 110 kΩ // 330 pF)
was mounted below the string. The integral of the pickup voltage is shown in Fig. 1.24 in
normalized fashion. The flexural wave passes the pickup 1 ms after its generation and induces
a voltage there. The dilatational wave that
is also generated runs along the string, is
reflected, and arrives back at the bearing
after 5,2 ms. Here, a secondary flexural
wave is generated (among other waves)
that passes the pickup after another
millisecond. In Fig. 1.24, the maximum of
this secondary impulse reaches almost 40%
of the magnitude of the primary impulse.
At least for this experimental setup, this is
an impressive testimony for the
significance of the dilatational wave.
Fig. 1.24: Measurments with a magnetic pickup.
“Auslenkung” = displacement; “Spannungsintegral” =
voltage integral
1.5 The plucking process 1-27
1.5 The plucking (or picking) process
The guitar string is plucked (or picked) with the finger (-nail) or a plectrum (pick, fingerpick).
The following calculations and measurements describe the excitation with a pick because this
represents the dominating approach for electric guitars.
1.5.1 Dispersion-deconvolution
Compared to the (particle) velocity of the string, the speed of the pick is relatively slow; in
fact, the displacing of the string can be regarded as quasi-stationary. For low-frequency
movements, the string acts as a spring with a lateral stiffness sQ (depending on the scale M),
the tension force Ψ, and the distance x between location of picking and bridge:
Lateral stiffness
Usually, the location of picking is about 6 – 10 cm from the bridge, with a lateral stiffness of
about 1000 – 2000 N/m resulting. Given a typical displacement of 2 mm, the potential
excitation energy will be around 2 – 4 mWs. No significantly higher energy levels will be
obtainable due to the distance of string to fretboard, but lower energy levels may certainly
occur with light plucking. Because the lateral stiffness is similar for all 6 strings, the
excitation energy of all strings is comparable, as well.
First, the string converts the excitation energy into vibration energy that is on the one hand
radiated as airborne sound, and that on the other hand will directly be converted into heat
energy. If all of the vibration energy would remain within the string, the latter would heat up
by about 1/1000th of a degree – no really much at all. A well-built acoustic guitar will convert
a considerable portion of the vibration energy into airborne sound: in an anechoic chamber,
peak sound pressure levels of just shy of 90 dB may be reached at 1 m distance.
Measurements with a Martin D45V yielded an airborne sound energy of about 1 mWs. This,
however, represents merely an orientation because beaming and plucking strength were not
determined precisely – indeed the investigation of acoustic guitars is not the actual aim here.
When analyzing the string oscillation from an instrumentation-point-of-view, several systems

need to be distinguished: generator, string, and pickup. The generator describes the string
excitation. Idealized, the plucking delivers a force-step, but in reality differences to the ideal
step are found depending on the movement of the pick. For the first few milliseconds, the
string may be described quite well as a loss-free, dispersive, homogeneous transmission line;
for more extended observations, damping increasing towards high frequencies needs to be
considered. The pickup converts mechanical vibrations into electrical signals. Its sensitivity
depends on the oscillation plane of the waves, and moreover we encounter strong frequency
dependence. The term “pickup” shall here be used rather broadly at first; it includes all
frequency dependencies that are not directly due to the plucking process or to the flexural
wave. A distinction into further subsystems may be necessary – depending on the
circumstances.

The objective of the present investigations was to describe the transmission behavior of the
above systems. Since all three subsystems interact (the plucking process cannot be analyzed
without the string, the pickup will re-act towards the string), an isolated system analysis was
not possible. In some respects, the vibration instrumentation also provided limitations, in
particular if measurements up to 10 kHz or even 20 kHz are targeted.
The below measurements were taken with the Ovation Viper already mentioned. The string
was plucked with a plastic plectrum given realistic conditions (in situ). This provided, as a
first approximation, a step-shaped imprinted force; however, more precise investigations
show significant deviations from this. The problem is not so much the actual step itself
(which of course may not be of infinitely fast speed: natura non facit saltus), but much more
the way the force develops ahead of the actual step. First, the plectrum relatively slowly
presses the string to the side. Just before the step, a relative movement between string and
plectrum commences which may in turn include both sliding friction and static friction (slip-
stick). In this, the force fluctuates quickly. After the plectrum separates from the string, it
moves according to a damped Eigen-oscillation (natural vibration) that may include another
short contact to the string. It is almost impossible to directly measure the forces occurring at
the tip of the plectrum – especially not up to 20 kHz. However, the piezo-signal allows for
conclusions regarding the excitation signal.
To describe it, the overall transmission line is divided into three subsystems: the plectrum-
filter that forms the real force transmission from the ideal step, the string-filter modeling the
dispersive flexural-wave propagation, and the piezo-filter emulating the transfer
characteristic of the pickup (incl. connected resonators). If on top of the step-transmission, the
reflections are of interest too, a recursive structure is required (Chapter 2.8).
The individual filters are taken to be linear – this should be a correct assumption at least for
light plucking of the string. Moreover, the piezo-filter is of time-invariant character. The
string definitely does not have that quality: an old string features a much stronger treble-
damping than a new one. Within a single series of experiments, however, the string may be
seen as time-invariant as long as no detuning occurs. The plucking process is difficult to
repeat the exact same way; it is time-variant, as well. Using suitable mechanical contraptions,
an acceptable (albeit not ideal) reproducibility is possible.
The overall system between step-excitation and piezo-signal is described via an overall
transfer function and a step response (or impulse response). Without supplementary
knowledge, a division into the individual subsystems is not possible. Assuming restricted
conditions, it is, however, possible to determine approximated transfer characteristics.
First considerations are directed towards the wave propagation. The frequency dependence of
the group delay could already be shown using short-term spectroscopy, with good agreement
between physical explanation (cantilever) and measurement. The measurements of the
evolution of the levels of the partials during the first milliseconds indicates only very little
damping; therefore assuming a loss-free all-pass is justified.
The following considerations relate to the low E-string plucked in its middle with a plectrum.
While the step runs from the middle of the string, the levels of the partials do not change, but
the phases are shifted such that the step is spread out (Fig. 1.16). If we shift the phases back
using an inverse filter, the step reappears. It is changed by the piezo-filter, though, and after a
short time, the saddle reflections superimpose themselves (Fig. 1.25).
Shifting back the phases corresponds to a de-convolution using the impulse response of the
all-pass, or a multiplication with the inverse transfer function of the all-pass. We need to
consider here that a de-convolution is only possible for one single line-length (e.g. L/2), and
for this reason the steps following later on the time-axis in Fig. 1.25 still show all-pass
distortion. Due to the de-convolution, the step spread out across the time range from 1 – 3 ms
is concentrated to the zero point on the time axis. The signal occurring ahead of that is the
excitation by the plectrum, convolved with the impulse response of the piezo-filter. Now, this
is where things get complicated: the plectrum-filter and the piezo-filter cannot be separated
without any further assumptions. There are an infinite number of possibilities to separate a
product into two factors.
Fig. 1.25: Original piezo-signal (left), de-convolved piezo-signal (right); low E-string plucked in the middle.
“1. Sprung mit/ohne Disperion” = 1st step with/without dispersion.
However, in order to fundamentally understand the plucking process, an exact system-

separation is not necessary in the first place. We already obtain a good approximation from
defining the signal shape ahead of the first step as the plectrum-excitation. For a more exact
analysis, measurements with the laser vibrometer are being prepared.
Already a simple evaluation of many plucking processes reveals various mechanisms

influencing the vibration:
The distance between plucking location and bridge is responsible for characteristic comb-
filters; this will be discussed in-depth later.
Shape and hardness of the plectrum influence the treble response.
The attack angle of the plectrum influences the bass response.
Bouncing and “slip-stick” processes lead to comb-filtering.
Fig. 1.26: String movement fro friction-free plectrum excitation; guitar top horizontal (sectional image).
“Plektrum” = plectrum.
In Fig. 1.26 we see (from left to right) four consecutive points in time of an excitation
process. The guitar top is horizontal and the plectrum is steered in parallel to it. On the left,
the plectrum touches the string without transmission of any force. In the second figure, the
string is displaced along a line perpendicular to the plectrum and running through the zero
position of the string. In the third figure, the displacement progresses, and in the fourth figure
the string just starts to leave the plectrum and vibrate along the dashed path. The whole
process is taken to be free of friction.
Given constant horizontal plectrum-speed a sawtooth-shaped string displacement results. A

piezo-pickup built into the bridge will react mainly to movements normal to the guitar top (as
will your usual magnetic pickup with coils), and therefore only the vertical vibration is of any
significance. With slow plectrum movement, the string acts as a spring. The vertical force is
proportional to the vertical displacement, and both increase time-proportionally up to a
maximum value. The excitation force then instantly breaks down to zero.
In reality, the plectrum will not move precisely in horizontal fashion. Rather, contact forces
will deflect it upwards. Moreover, its angle of attack will change, and for thin plectra bending
will occur in addition. The sliding friction between string and plectrum also allows for small
deviations from the dashed line, and there might be stochastic slip-stick movements. The
latter stem from the difference between sliding friction and static friction: if the plectrum-
parallel string force becomes greater than the static friction force, a relative movement
between string and plectrum sets in along the plectrum. Since the smaller retention force is
now substantially surpassed, the string can slip over a small distance – until it is stopped again
via the (higher) static friction force.
For Fig. 1.26, the plectrum is angled at 63° relative to the guitar top, but remains parallel to
the longitudinal axis of the string. The smaller this angle of attack becomes, the easier it is for
the string to continuously slip towards the bottom. Increasing this angle to 90° (i.e. the
plectrum is perpendicular to the guitar top), the string is displaced only horizontally at first –
there is no vertical movement. It some point the plectrum has to yield, though – either it
boggles towards the top, or it bends or changes its angle such that the string can move
downwards. The associated excitation impulse has a shorter duration compared to the angled
plectrum: the “boggling” can happen only during the very last millisecond, so to say.
If the plectrum is not held exactly in parallel to the longitudinal axis of the string but at a
slight angle, the friction changes. This is because the string does not slide along the surface of
the plectrum anymore but skips along the edge of the plectrum. In most cases, the edge is
rough – which increases the stochastic component in the excitation. The latter effect is further
increased for wound strings.
Therefore, the guitar player has many possibilities to influence the excitation impulse – and
thus the sound of the guitar. This begins with the choice of the pick, its free length, and its
angle relative to the guitar top and relative to the longitudinal axis of the string. In addition to
the plectrum, the fingertip may contact the string during the plucking process (teeth have also
been know to get used here …), and on top of it all the location of the plucking may be varied,
and the strength of the plucking, of course.
A simple, step-shaped excitation is conducive to the system-theoretical description of the

string. Since moreover the evaluation of its reproducibility is done with relative ease, this
excitation was the basis for many measurements. However, that does not mean that the ideal
step-excitation represents the desirable objective for the guitarist.
1.5.2 Influence of the plectrum
It is most purposeful to discuss the effects of the plucking process on the sound in the
frequency domain (Fig. 1.27). The force impulse shown in the figure has an arbitrary duration
of T = 80 ms; is the maximum value (negative in the present case). FS describes the
spectrum corresponding to this sawtooth impulse, and Fδ pertains to the time-derivative of the
sawtooth impulse. Within the frequency range pertinent to the guitar it makes no big
difference whether the impulse starts at –80 ms (as it does in the figure) or much earlier … it
is only important that the actual step occurs at t = 0. For this reason, we use the term step
excitation despite the fact that strictly speaking we have an impulse. We obtain the
mathematically correct limiting case as T moves towards ; the first fraction in the spectral
function vanishes in this case and – with 1/jω – a pure (rectangular) step-function remains.
The time-derivative of this ideal step is the Dirac impulse that corresponds to a constant
(white) spectrum Fδ. In systems theory, (Dirac-) impulse excitation and impulse response are
most commonly used; step excitation and step response are somewhat closer to the practical
application. Disregarding the frequency f = 0 that does not actually exist, both descriptions are
equivalent and may be converted from one to the other.
Fig. 1.27: Sawtooth impulse:

time- and spectral-function
Because in reality the force process occurring upon plucking does not correspond to the
depiction in Fig. 1.27, we define a plectrum-filter that shapes the actual force process from
the theoretical rectangular step. The magnitude of the frequency response this plectrum-filter
has describes the impact of the plucking process onto the sound.
The following figures show the analyses for the already mentioned Ovation guitar. The low
E-string was plucked with a thin nylon-pick (Meazzi 19), while the piezo-signal was fed
directly into a high-impedance measuring amplifier – and cleared of the dispersion via de-
convolution with an inverse all-pass (Chapter 1.3.2) Fig. 1.28 shows two time functions
obtained that way. Compared to Fig. 1.27, there are several striking differences: the force
increase (in terms of its amount) is not linear but progressive; during the last few milliseconds
several peaks appear (slip-stick); after the step, reflections are visible that presumably are
caused by longitudinal resonances.
Fig. 1.28: De-convolved piezo-signal; two different plucking processes.

“Sprung ohne Dispersion” = step without dispersion.
In Fig. 1.29 we see different plucking processes in comparison. The left-hand column shows
the dispersion-free, de-convolved piezo-signal while the right-hand column shows the
magnitude spectrum belonging to the differentiated piezo-signal. The derivative makes for an
easier evaluation: the ideal rectangular step is linked to a constant (white) spectral function.
The first line a) depicts an almost perfect step. Only from about 3 kHz, a treble loss occurs; it
is connected to the rounding off of the step. There may be several reasons for this: the tip of
the plectrum is rounded off, and therefore the string is not displaced in an exactly triangular
manner. This effect is probably further increased by the bending stiffness of the string. The
high frequencies are consequently attenuated already in the excitation signal. In addition,
dispersion effects in the string need to be considered that also manifest themselves in the high
frequency range.
In the case of b), the force rises to its magnitude maximum only during the very last
milliseconds. This will occur if the plectrum has a high angle of attack and moves in parallel
to the guitar top. The shape is more impulse-like, and in the spectrum the bass is attenuated.
The analyses c) to e) indicate a progressive treble damping as it is typical for a round, hard
plectrum.
For the remaining analyses, the force increases first (in its magnitude) and then moves
through a magnitude minimum (the force acts in the negative direction). Presumably, this
includes a sliding along the string of the plectrum, the latter getting stuck on the string for a
short time and then finally separating from the string.
a)
b)
c)
Fig. 1.29: Excitation step, and spectrum of the differentiated step for various plectrum movements.
d)
e)
f)
g)
h)
i)
j)
Fig. 1.29: Continuation from the previous page.

Fig. 1.30: De-convolved piezo-signal for the string excited in its longitudinal direction (scratched string).
“Entfaltet” = de-convolved.
Fig. 1.30 documents an interesting detail: here, the low E-string was excited using a sharp-
edged metal plectrum at mid-string in the longitudinal direction, i.e. the plectrum scratches
along the string, jumping from one winding to the next. The signal transmitted by the piezo
was again de-convolved i.e. cleared of the dispersion. As the plectrum jumps across the
winding, a flexural wave is generated. The first (de-convolved) impulse of this wave is shown
at 0 ms (the second impulse appears at 3,7 ms). However, in addition a dilatational wave of
about 1,4 kHz occurs (Chapter 1.4). This (non-dispersive) dilatational wave propagates with a
considerably higher speed than the transversal wave; its start is shifted by 3 ms towards the
past due to the de-convolution. In fact, the de-convolution algorithm does separate according
to wave-type but it corrects the phase delay of any 1,4-kHz-signal by -3 ms. Further details of
the dilatational wave (in particular regarding its coupling to the transversal wave) have
already been described in Chapter 1.4.
The plucking processes shown in Figs. 1.29 and 1.30 are typical for guitars but represent
merely a relatively arbitrary selection. There is also a multitude of other possibilities to excite
the string – and we need to particularly consider that the tip of the thumb or the first finger
may also come into contact with the string. It is therefore not necessarily an indication of
excessive vanity if the well-known professional guitarist, after an extensive narrative
highlighting his wonderful custom-built paraphernalia, concludes the interview about his
equipment with a confident: “90% of the sound is in the fingers, though”.
1.5.3 String-bouncing
If a string is plucked with little force, it will approximately react as a linear system. This
means that doubling the initial displacement will also double the displacement at any instant
of the subsequent vibration process. Of course, any displacement is limited – at some point
the string will hit the frets on the fretboard. In doing so, it generates a somewhat rattling,
buzzing sound. To some degree, this is in fact a means of musical expression and thus not
something generally undesired.
In the book “E-Gitarren” by Day/Waldenmaier we find the recommendation: "A slight tilt of
the bridge makes it possible to adjust the action of the high E-string a little lower than that of
the low E-string. The latter has a more pronounced vibration amplitude and requires more
space that the high strings ". However, the transverse stiffness for all customary string sets is
higher for the low E-string (E2) than it is for the high E-string (E4) – why then would the
stiffer string require more space for its vibration? It is o.k. to concede this space to it; that
decision is, however, just as individual as the choice of the string diameter and cannot be
justified with a generally larger amplitude.
Fig. 1.31: String displaced at A (bold line), intermediate positions of the vibration (thin lines). In the left-hand
figure, the string was pressed to the guitar body and then released, on the right it was pulled up and released.
“Sattel” = nut; “Steg” = bridge.
The string is displaced in a triangular fashion by the plectrum (or the finger-tip, or –nail, or
teeth …). After the plucking process, the string moves in a parallelogram-like fashion – given
that we take a dispersion-free model as a basis (Fig. 1.31). However, this movement in the
shape of a parallelogram can only manifest itself if the string does not encounter any
obstacles. Frets are potential obstacles; their immediate vicinity has the effect that the string
does not only occasionally establish contact but hits them on a regular basis … with the
parallelogram-shaped movement being correspondingly changed. Fig. 1.32 shows (seen from
the side) a neck with the typical concave curvature. The axis-relations of this figure hold for
the following figures, as well.
Fig. 1.32: Fretboard geometry (strongly distorted due to the scale); lower surface of the resting string (dashed).
The frets are distorted into lines due to the strong magnification of the vertical dimension.
“Sattel” = nut, “Steg” = bridge; “Griffbrett” = fretboard.
If the string pressed down at point A (Fig. 1.33) has no contact to the frets, it can freely decay
in the dispersion-free model case. The string that has been lifted up, however, hits the 10th fret
already after less than half the vibration period – its vibration-shape is completely destroyed.
Fig. 1.33: String-parallelogram. On the left, the string was pressed down and then released (uninhibited
vibration); on the right it was lifted up and then released (fret-bounce at the 10th fret). “Griffbrett” = fretboard.
The well-versed guitarist will vary his/her “attack” as required and shape the sound of the
respective picked note via change of the picking-strength and –direction: both pressing-down
and lifting-up of a string happen. However, in particular when using light string sets, a further
vibration pattern occurs. It is generated as the string contacts the last fret (towards the bridge)
when being pressed down during plucking (Fig. 1.34). As soon as the string is released, a
transverse wave propagates in both directions and is first reflected at the last fret and then at
the bridge. Consequently, a peak running towards the nut is generated – it is reflected there
and bounces onto the first fret (right-hand part of the figure).
Fig. 1.34: String displacement at different points in time. On the left, the first half-period is shown, on the right
we see the subsequent process including bouncing off the first fret. Plucking happens at point A with contact to
the fretboard. The time-intervals are chosen such that the resolution is improved at first and after t = T/2.
Without dispersion. “Griffbrett” = fretboard.
Immediately the question pops up: how often does this case happen? Contact-measurement at
the last fret tells us: a lot. For better understanding, Fig. 1.35 depicts the connection between
plucking force (transverse force) and initial string displacement (at A). Since the transverse
forces often reach 5 N (or even 10 N occasionally), contact to the last fret often occurs.
Fig. 1.35: Connection between transverse force and string displacement, open string (left), string fretted at the
14th fret (right), plucking point 14 cm (–––) and 6 cm (---) from the bridge. 2,1 mm clearance between the string
and the last fret (= 22nd fret). B-string, 13 mil, calculations.
“Saitenauslenkung” = string displacement; “Querkraft” = transverse force.
We can see from Fig. 1.35 that the string operates as a linear system only for soft plucking.
As soon as the string gets into contact with the last fret, the force/displacement characteristic
experiences a knee – a jump in the stiffness of the string occurs. This degressive characteristic
tends to correspond to the behavior of a compressor: despite stronger plucking force, the
string-displacement grows only moderately. However, here we also find a source of potential
misunderstanding, for displacement does not equal loudness! With the string establishing
contact to the last fret, the shape of the vibration deviates from the mentioned parallelogram,
and changes result in the spectrum, and thus in the sound.
For the following graphs, the E4-string of an Ovation guitar (EA-68) was plucked using a
plectrum; the electrical voltage of the piezo pickup built into the bridge was analyzed (i.e. the
force at the bridge). The location of plucking was at a distance of 125 mm from the bridge,
and the plectrum was pressed towards the guitar body such that a fretboard-normal vibration
was generated. Fig. 1.36 shows time function and spectrum for the linear case (no contract
between string and last fret). The voltage of the piezo jumps back and forth between 0 V and
0,4 V, with a duty cycle resulting from the division of the string (517:125, scale = 642 mm).
Given the transfer coefficient of 0,2 V/N (Chapter 6), the corresponding force at the bridge
calculates as 2 N, this representing good correspondence to Fig. 1.35. In this example, 2 N
forms the limit of linear operation – using a larger force makes the string bounce off the frets.
Fig. 1.36: Time-function and spectrum of the piezo-signal. The upper half of the left-hand graph shows the
measured time function, below is the result of the calculation. On the right is the measured spectrum and the
(idealized) envelope. Open E4-string, fretboard-normal vibration. “Frequenz” = frequency.
The analyses shown in the following graphs (Fig. 1.37) correspond to Fig. 1.36 but are based
on (fretboard-normal) string excitations of different strengths. For the upper two pairs of
graphs we can see proportionality in the time domain and in the spectral domain: the level
spectrum is simply shifted upwards for stronger plucking. As soon as the plucking force
exceeds 2 N (in the lower two pairs of graphs), the string touches the last fret and bounces off
it. Time function and spectrum become irregular. The strong peak in the time function finds
its counterpart in the location function (Fig. 1.34); it may be interpreted as the interaction
between two excitations:
a) string displacement, force step at t = 0 (idealized), and
b) opposite-phase force step at the last fret; occurring at the instant as the string leaves the last
fret (t ≈ 0,2 ms).

Fig. 1.37: Time-function and spectrum of the piezo-voltage. String plucked with different force. See text.
A spectral analysis encompassing the whole of the auditory range is conducive for the
acoustic guitar, and the same holds for a piezo-pickup (Chapter 6). In Fig. 1.38, three of the
sounds from Fig. 1.37 are shown as third-octave spectra. On the left, we see the spectra of
strings plucked lightly and with medium strength, respectively – the system is still linear and
the spectra merely experience a parallel shift. Strong plucking (right figure) leads to a level-
increase merely in the middle and upper frequency range; below 1 kHz, there is even a
decrease in level. As other strings are played, or as the E4-string is fretted at other frets, this
effect tends to remain, but the spectral differences are specific to the individual case.
Fig. 1.38: Third-octave spectra, open E4-string, overlapping analysis of main- and auxiliary third-octave.
On the left, and for the dashed curve on the right, there is not yet any bouncing off the frets. Strong plucking
(solid line of the right) causes the string to touch the last fret and bounce off it. 1st and 2nd harmonic actually
decrease in this process, while there is a strong increase in level at middle and high frequencies.
From this, we can deduce a compressor-like behavior in any guitar: for light plucking, the
string operates as a linear system, and slight changes in the picking strength lead (with good
approximation) to similar level changes in the whole frequency range. However, already at
medium picking strength, the string bounces off the frets – the lower the action and the lighter
the strings, the lower is the threshold to this occurring. Now, if filtering (due to magnetic
pickups) accentuates a specific frequency range, this compression is perceived with different
strength. Fender-typical single-coil pickups emphasize the range around 3 – 5 kHz. This will
lead to less perception of compression compared to humbuckers sporting resonance
frequencies around 2,5 kHz. This may not happen for all played notes, but it does happen in
the example shown in Fig. 1.38. So does a humbucker compress more strongly than a single-
coil? “Somehow”, yes – but not causally. The source of the compression is the string (in
conjunction wit the frets) that compresses in different ways in various frequency ranges.
Pickups and amplifiers make this different compression audible in different ways.
Here’s an opinion voiced in the Gitarre & Bass magazine (02/2000): "What happens when I,
for example, pick the low E-string first softly and then more and more strongly via a slightly
distorted amp? The Strat behaves much more dynamically and you can open the throttle ever
more until, purely theoretically, the string throws in the towel and breaks. The Les Paul shows
an entirely different character: first, the increasingly harder picking also generates more
loudness, but then the whole thing topples over: the notes don’t get louder anymore but more
dense – almost as if there were a compressor/limiter switched in. Say what?! Indeed, the
information of the string vibrations resulting from the behavior of the wood determines the
tonal characteristic of the Les Paul, but not the fatter sounding humbuckers.”

The G&B-author was careful (?) enough not to throw in something like “and that shows that
mahogany compresses more strongly than alder”. Still, he infers: “now we understand, why a
Strat even with Humbuckers can never turn into a Les Paul. You can at most make the tone
warmer and fatter, but the typical compression is out of reach.” Unfortunately, the author
does not report which experiments or models were the basis for his last conjecture.
Fig. 1.39: Third-octave spectrum, Stratocaster, neck-

pickup, E2-string (42mil) fretted at the 5th fret. Plucked
from lightly to strongly. Distance between plectrum and
bridge: 13 cm. Clearance of the open E2-string to the
last fret: 2,3 mm.
As we can see from Fig. 1.39, a Stratocaster, too, compresses in the range of the low
partials. While the level-difference between light and very strong plucking is no less than 39
dB at 4 kHz, the fundamental changes only by 7 dB. Your typical Gibson Humbucker will
only transmit the spectrum of the low E-string up to about 2 kHz and therefore misses the
dynamic happening in the 4-kHz-range that a Fender pickup will still capture. However, in the
experiment reported in G&B, it is likely that behavior of the amplifiers was almost more
important: “via a slightly distorted amp”. There you go! The Gibson Humbucker will have
generated approximately double the voltage of the Fender single-coil. That makes the
amplifier participate in the signal compression: it will compress (or limit) the louder signal
(that of the Les Paul).
However, that does not mean that the compression is determined merely by the action on the
guitar, and by the amplifier. As the string bounces off the fret, a metal hits metal (at least on
the electric guitar). The result is a broad-band bouncing noise that extends to the upper limit
of the audible frequency range. String- and fret-materials are of particular significance in this
bouncing noise: pure-steel wound strings generate a more aggressive, treble-laden noise
compared to pure-nickel wound strings. Old string with their winding filled up by rust, grease,
etc, will sound duller than fresh strings. And the fret-wire that the string hits (that may in fact
be any fret in the course of the vibration) contributes, with its mechanical impedance, to the
bouncing noise, as well. A detailed analysis of the mechanical neck- and body- impedances
follows in Chapter 7; string/fret-contacts are analyzed in detail in Chapter 7.12.2.
1.5.4 String-buzz
If the string is plucked with little force, it reacts approximately as a linear system. This
implies that double the initial displacement also leads to double the displacement at every
moment during the subsequent vibration process. Of course, the displacement cannot become
indefinitely large – at some point the string will hit the frets on the neck (Chapter 1.5.3,
Chapter 7.12.2). If this contact to the fretboard happens right after the plucking itself, it
becomes part of the attack process of the respective tone. Later occurring contacts to the frets
(with the limit at later than about 50 ms) will become audible as single events – given they are
strong enough. Weak or short string/fret contacts are, to some degree, a means of expression
and therefore not generally undesirable.
Fig. 1.40: Time-function and spectrogram of the piezo voltage resulting from a strongly
plucked low E-string (E2).
In Fig. 1.40 we see the piezo voltage taken from an OVATION Adamas SMT (open E2-string),
with the string so strongly plucked with a plectrum that a clear buzz became audible. The
spectrogram reveals – after the broadband first plucking impulse has passed – further string-
to-fret hits around 200 and 350 ms; these act like high-frequency echoes. The string hits the
frets repeatedly and strongly, and generates a clearly audible buzz.
Besides the impulses occurring with a separation of 12 ms, very low-frequency vibrations are
visible in the time-function. These point to the reason why the string bounces off the fret not
only at the very beginning of the vibration. However, an exact analysis of the low-frequency
vibration cannot be derived from the time-function. This is because the cutoff-frequencies
found in the piezo pickup, the amplifier and the analyzer at around 2 Hz result in strong phase
shifts. The cause of the low-frequency signal components is a rotation of the plane of
vibration (Chapter 7.7.4, Chapter 7.12.1).

1.6 The decay process
After being plucked, the sting vibrates in a free, damped oscillation process. “Free” implies
that no further energy is injected; “damped” indicates that vibration energy is converted into
sound and caloric energy (radiation, dissipation). Any further string damping (e.g. via the
fingertips of the palm of the hand) shall not be considered here at this time.
1.6.1 One single degree of freedom (plane polarization)
The simplest oscillation system consists of a mass, a spring, and a damper. The mass force is
proportional to the acceleration (inertia, NEWTON), the spring force is proportional to the
displacement (stiffness, HOOKE), and the damper force is proportional to the (particle)
velocity (friction, STOKES). The time derivative of the displacement yields the velocity; the
time derivative of the velocity yields the acceleration [3].
After the excitation a “periodic” oscillation of the frequency fd results. Instants of equal phase
(e.g. maxima, zeroes, and minima) occur at equal distances in time – which led to the us of
the term period T = 1/ fd. However, signal theory does not actually see this decay process as a
periodic signal: due to the exponential decay, the individual periods fail to be identical.
Mechanics, on the other had, do use the term periodic vibration here because the duration of
the periods in time-invariant ( ... non est disputandum).
The resulting envelope has three parameters: the frequency fd, the initial phase ϕ, and the time
constant of the envelope . In this general form, the equation for the oscillation is:
Oscillation equation
For t = 0, the e-function yields 1; with increasing time, it decreases towards 0. The phase shift
ϕ may be taken to be zero for the first considerations. The time constant determines how
fast the oscillation decays: the smaller is, the faster the decay. Instead of , literature
offers a multitude of other parameters, as well – they can easily be converted into each other.
The letter τ is frequently used for the time-constant; in the present context we will rely on this
letter only when we get to the calculation of levels. What needs to be avoided in particular is
confusion between the degree of damping and the decay-coefficient, since the latter is
sometimes also designated with !
It may be the displacement, the (particle) velocity, or the acceleration that represents the
physical oscillation. A sensor converts these quantities into a voltage u(t) that subsequently is
analyzed.
Fig. 1.41: Damped oscillation of 100 Hz; exponential decay; time-constant .
1.6 The decay process 1-43
Given mass m, spring-stiffness s, and friction W, we calculate frequency and time-constant:
Parameters of the oscillation
If the friction W is set to zero, the un-damped system results. It has an infinite time-constant:
the e-function now has the constant value 1, and the vibration does not decay anymore. A
weakly damped vibration with a frequency fd = 100 is shown in Fig. 1.41. The shape of the e-
function is indicated as a dashed line with its tangent crossing zero at . At the point in time
of , the envelope has decreased from 1 to 1/e ≈ 0,37.
In instrumentation, the decay process is often depicted as level-curve. Level is a logarithmic

measure that may be determined in various ways. It always constitutes a time-average over a
weighted measurement interval; the averaging is done using the squared signal quantity. We
often see an exponential averaging where the weighting is of exponential form, and is done
such that the signal components lying further back in the past contribute less prominently to
the measurement. The averaging time constant τ is specified as parameter of the exponential
averaging; the value is used frequently, with the corresponding standardized way
of averaging being labeled FAST. The decay constant of the dampened oscillation must
not be confused with the averaging time constant τ of the level measurement.
The level measurement comprises three consecutive operations: squaring, averaging, and
logarithmizing. Squaring and logarithmizing are non-linear operations; the order of sequence
must therefore not be interchanged. It is only the averaging that is a linear filter operation: a
1st-order low-pass in the case of the level measurement. In the time domain, the averaging is
described by a convolution [6]: the result of the averaging corresponds to the convolution of
squared signal and impulse response h(t) of the averager. For damped oscillations we get:
(for causal signals)
Here, h(t) is the impulse response of the averager, u(t) is the damped oscillation, the star
symbol stands for the convolution. The average m(t) is calculated for the point in time t with
the time-variable ψ integrated from 0 to t. Therefore, the average value m(t) does in this case
not indicate the average over the whole decaying oscillation but the average from the
excitation to the (variable) point in time t. The averaging time constant τ is large compared to
the oscillation period T; the contribution of the sine function can thus be disregarded in good
approximation. Using this, the time-variant average is:
for

When calculating levels we need to consider that we are working with a squared signal, which
is why we need to opt for the formula for power levels. The reference value needs to be
chosen such that the correct absolute value results for the steady case ( ). Using, on the
other hand, , we get the relative level that decays starting from 0 dB.
dB = decibel = reference value
Fig. 1.42 shows the course of the level of a damped oscillation determined via exponential
averaging. The time-constant of the damping is . Having an understanding of the
equation of the oscillation, we could also give the exact course of the level. To do that, it is
merely necessary to logarithmize the e-function (shown as a dashed line). The level
determined via measurements deviates significantly from this calculation. In the figure, we
see two graphs with the averaging time-constants 0,125 s and 0,5 s, as well as the theoretical
behavior (dashed).
Fig. 1.42: Level of an exponentially

damped oscillation. Damping time-
constant = 4 s, averaging time
constanr, τ = 125 ms and 500 ms. For
500 ms, the asymptote is too high by
1,2 dB, and for 125 ms, it is too high
by 0,3 dB.
“Pegelverlauf” = course of the level;
“Zeit in Sekunden” = time in seconds
After a short attack phase (mainly determined by τ), the level drops off with approximately
the time constant . As is evident, the measurement curves run in parallel to the exact values
after a short time, but remain too high. Therefore the slope – and thus the system damping –
can be determined with good accuracy; for measurements of absolute values, however,
considerable errors may arise. Using L(t), the level difference is calculated as:
The shorter the averaging time-constant gets relative to the damping time-constant, the more
exact the tracing of levels via measurements becomes. Still, the averaging time-constant must
not be chosen too short, either, because then the (squared) oscillation may not be fully
averaged anymore, and ripples in the level-graphs would result.
Moreover, Fig. 1.42 indicates that the measured level maximum is lower than expected. The
position of the maximum is determined via differentiating and zeroing:
The larger the averaging time-constant is chosen, the lower the maximum.
From a signal-theory point-of-view, a damped oscillation belongs with energy signals. The
signal energy is derived as integral over the squared signal value; it differs from the physical
energy:
Z = impedance
The signal energy of the damped oscillation may be calculated from the equation of the
oscillation using integration:
The average value across m(t) yields the same signal energy irrespective of τ. If the energy is
derived via mmax, however, a correction is required due to .
Besides the exponential averaging there are also other ways to average: block-averaging is
done with constant weighting across a fixed time interval, Hanning-averaging uses a sine-
shaped weighting. Block averaging is also called linear averaging, a rather confusing term
that is common in the area of spectral analysis, though. While the exponential averaging is
always run from the start of the signal to the point in time of the measurement (marked with a
star on Fig. 1.43), linear averaging is done from the start of the signal over an interval of fixed
duration (1 s in the figure). In exponential averaging, only the end of the interval is shifted, in
linear averaging, however, this is done to both start and end. The Hanning-averaging uses a
fixed duration of the averaging (2 s in the figure), as well, but weighs the signal with a sin2.
Hanning-averaging is often deployed in DFT-analyzers – as are many other DFT-windows
(Blackman Kaiser, Bessel Gauß, Flat-Top, etc.).
Fig. 1.43: Different ways of averaging:

exponential averaging (upper left),
linear averaging (upper right),
Hanning-averaging (lower left).
“Zeit in Sekunden” = time in seconds”
All ways of averaging are calibrated such that for steady signals (constant level), equal results
are obtained. With levels varying over time, differences occur. In frequency-selective
analyses (DFT, 1/3rd-octave, etc.), also further system-immanent errors contribute: a filter will
react more sluggishly to the input signal as the filter band becomes narrower. In broadband
level-measurements (e.g. 10 Hz – 20 kHz), no significant errors will occur, but in selective
measurements of partials (e.g. 2500 Hz – 2519 Hz), they might creep in, depending on
circumstances.
1.6.2 Spatial string vibrations
After a guitar string is plucked, spatial vibrations will propagate on it. The transversal waves
introduced in Chapter 1.1 are of particular significance. Given that the axis along the string is
taken as z-coordinate, transversal waves can propagate both in the xz-plane and the yz-plane;
superpositions are possible, as well. For electric guitars, the vibration plane perpendicular to
the guitar top is especially important, while for acoustic guitars the vibration parallel to the
guitar top also has effects.
The wave equation includes a dependency both on place and time. However, investigations
into the vibrations of guitar strings are mostly based on a fixed location (the place of e.g.
pickup, or bridge) so that merely the time remains as variable. As a simplification, the string
vibration occurring at a given location tends to be seen as superposition of many
exponentially decaying partials (Chapter 1.6.3). In this scenario we need to consider, though,
that for each partial, vibrations may appear in two planes. Sometimes one of the two
vibrations has next to no effect and may be disregarded, but in some cases both need to be
taken into consideration.
The following approaches first start from the assumption that plucking the string will result in
two same-frequency vibrations orthogonal in space. The time constants of the damping are
still different for the two vibrations, the effect on the output is different, and they may be
phase-shifted relative to each other. At the output, both are superimposed:
d = top-parallel part
Particularly in acoustic guitars, the top-normal vibration is tightly coupled to the resulting
sound field, and therefore vibration energy is relatively quickly withdrawn, and the damping
time-constant is short. The top-parallel vibration does not lead to as efficient a radiation (d is
smaller); it thus has a longer time-constant. In the level-analysis, the decay shows up with a
characteristic kink (Fig. 1.44).
Fig. 1.44: Open E2-string, FAST-level of the 2nd partial; left: calculation; right: measurement (Martin D45V).
To confirm our hypotheses about the vibrations, two experiments were carried out. In order to
adjust the neck, the OVATION Adamas SMT allows for the removal of a cover plate (of
∅13cm) in the guitar body. This detunes the Helmholtz resonance and thus changes the low-
frequency coupling to the sound field. With the cover taken off, the low frequencies receive
weaker radiation; the time constant should therefore be longer.
Fig. 1.45: Left: Ovation Adamas SMT, level of fundamental (F#2), with closed (“Deckel geschlossen”) and
removed (“Deckel geöffnet”) cover plate.
Right: Ovation Viper EA-68, level of fundamental (F#2), with (“mit Magnetfeld”) and without influence of the
magnetic field.
Fig. 1.45 (left) depicts the decay curves for the fundamental of the tone F# fretted at the 2nd
fret on the low E-string. The measurements confirm the assumption. In a second experiment,
a permanent magnet was brought close to the low E-string on an OVATION Viper. Due to the
attraction force the stiffness of the string is reduced in one plane of vibration – the vibration
frequency is thus reduced in that plane. This leads to a beating of the orthogonal
fundamentals now slightly detuned relative to each other (Fig. 1.45, right).
However, even without any magnetic field, the top-normal vibration of a particular partial
does not necessarily occur at the exact same frequency as that of the top-parallel vibration of
the same partial. This is due to the reflection factors of the string clamping (nut, bridge) – the
former are dependent on the vibration direction. The spring-stiffnesses at the edges may be
different for the two directions of the vibration, resulting in slight differences in the vibration
frequencies. The decay process will then include beatings that render the sound more “lively”.
Fig. 1.46 shows results of calculations and, for comparison, sound pressure levels measured
with an acoustic guitar (MARTIN D45V, anechoic room, microphone at 1 m distance ahead of
the guitar). Various patterns emerge:
The level differences between the two sub-vibrations determine the strength of the
interference. At a difference of 20 dB, the amplitude fluctuates merely by 10%, while at 6 dB
difference the fluctuations grow to 50%. Differences in the damping determine for which
period the beating persists. If both sub-vibrations decay with the same damping, the level-
difference does not change, and neither does the beat-intensity. Conversely, if the decay is
different, the beats are strongest at the instant when both levels are equal. The frequency
difference determines the periodicity in the envelope: the larger this difference, the faster the
fluctuations. Moreover, the phase of the sub-vibrations is of significance – in particular if
different damping occurs i.e. if the beats are limited to a short time-interval. The interference-
caused cancellation will only present itself if both sub-vibrations are in opposite phase during
said time-interval.
Another degree of freedom comes into play if we allow for non-linearities. For example, the
friction may depend on a higher order of particle velocity, or the spring-stiffness may depend
on the displacement. This may cause, for example, that the level of a mono-frequent vibration
does not decay linearly with time but shows a curvature. Addressing such aspects requires
considerable effort – no corresponding investigations were carried out in the present
framework.
Fig. 1.46 Top: Decay processes given phase-differences. Left: both vibrations with the same frequency; right:
beats due to a frequency differencde of 1.2 Hz. The damping cannot be determined precisely anymore from the
initial slope of the curve. Bottom: measurements with a MARTIN D45V
An interesting set of curves emerges if the excitation energy remains constant while the string
damping varies. First, however, we need to define more precisely the term “damping”: any
real string executes a damped vibration. In this case, damping means that vibration energy is
continuously withdrawn from the string, with displacement amplitude (potential energy) and
velocity amplitude (kinetic energy) decreasing over the course of time. Springs and masses
store energy while resistances “remove” energy. Sure, energy cannot actually be removed –
rather its mechanic incarnations are converted into caloric energy (heat); but in any case the
“removed” energy is not available anymore to the vibration of the string.
In the acoustic guitar, we need to distinguish between the ‘good’ and the ‘bad’ losses. If all
of the energy in the string is converted to sound-energy with an efficiency of 100%, we do
have damping (a loss), but the objective of generating sound has been achieved with the
utmost efficiency. If, conversely, 90% of the energy in the strings is converted directly into
heat due to inner friction, and only 10% are radiated, we have an undesirable loss. To
illustrate this with an EXAMPLE: a watering can supplies water to a flowerpot. If the water
flows through a small cross-section, it will take a long time until the can is empty. With a
larger cross-section, the process will be quicker – but it’s always the whole of the water that
arrived in the flowerpot. This situation changes if there is a hole in the bottom of the can – an
additional degree of freedom is now present that influences the efficiency ◊. Applying this to
the string: via tight coupling between string and sound field, the energy flows from the string
quickly – the string is damped strongly but all energy reaches the sound field (100%
efficiency). The efficiency drops only as friction-resistance is included in the guitar.
In electric guitars, the objective is entirely different. They do not need to radiate sound
energy – that’s taken care of by the loudspeaker. Due to the lack of radiation loss, the string
damping is lower, the decay is longer – the guitar has longer/better sustain.
Several quantities are disposable in order to describe damping: one is the time constant of the
damping (or time constant of the envelope) of the individual partials. During the length of
a time constant, the level of the respective partial drops by 8,686 dB. A vibration with a level
dropping off by 60 db within 10 s has a time constant of 1,45 s. The duration of time that it
takes a level to drop by 60 dB is – in room acoustics – also called the reverberation time TN.
The latter is suitable to describe a damping, as well: the formula holds. Fig. 1.47
shows the course of the levels of the fundamentals (G#) measured via the piezo pickup.
During the initial second, the time constants differ by a factor of 18.
Fig. 1.47: Measurements with Ovation guitars: SMT (acoustic guitar, left); Viper (electric guitar, right).
The following considerations are based on the law of conservation of energy. In the plucking
process, the string is given a certain potential energy that is in part dissipated and in part
radiated. As an EXAMPLE, a string is to be plucked with 5 mWs; it then decays in different
ways. Which sound pressure level is generated at a distance of 1 m if we assume – to begin
with – that 100% of the vibration energy is radiated as sound wave?
For any exact calculation we would have to know about the beaming – as a simplification let
us assume an omni-directional characteristic here. In fact, this assumption is a good
approximation for the (quite level-strong) 2nd partial of the E-string [1]. The energy E of the
spherical wave [3] is calculated as:
with Z0 = 414 Ns/m3
Herein, p(t) is the sound pressure at the distance R = 1m; the integral over the damped
vibration was already calculated at the end of Chapter 1.6.1. The equation can be solved for
the sound pressure amplitude:
in the example for und .
From the (now known) sound pressure, the level can be calculated e.g. for exponential FAST-
averaging (Fig. 1.48, left section, different ). ◊
The time constant of the damping influences both the maximum value and the speed of
decay. The luthier can increase the peak sound pressure level via high mechano-acoustical
coupling – the loudness will then decrease more quickly, though. Lower coupling will enable
him (or her) to achieve longer sustain, but then the guitars is not as loud. The plucking energy
is present only once, after all. Now, if we allow the string to vibrate in two planes, the
seemingly impossible is in reach: a loud guitar with long sustain. The top-normal vibration
generates a loud attack. The quick decay of this loud attack is “drowned out” after a short
time by the more slowly decaying top-parallel vibration.
Fig. 1.48: Left: FAST SPL for different degrees of coupling between string and sound field (η = 100%).
Right: FAST-SPL for two superimposed orthogonal vibrations (η = 100%). Equal energy.
Fig. 1.48 (right hand section) shows an example with both vibrations being excited with 5
mWs. The quicker decay happens at a time constant of the damping of 0,5 s, the longer decay
has a time constant of 5 s. The dashed lines indicated the levels of the individual vibrations.
An efficiency of 100% is assumed again for both vibrations.
Of course, in practice an efficiency of 100% is not achievable; part of the vibration is

converted into caloric energy already within the string, and in the guitar body, as well.
Reducing the efficiency to 50% will also reduce the time constant of the decay by half (this
may be deduced via the transmission-line equation). The course of the level will then be
determined by two parameters: the mechano-acoustical matching, and the dissipation in the
guitar (Abb. 1.49).
Fig. 1.49: Calculated SPL for an excitation energy of 5 mWs (left) and 2.5 mWs (right). The solid line indicates
an efficiency of 100 %, the dashed one an efficiency of 50%.
1.6.3 Partial and summation-levels
The real guitar string does not consist of a single concentrated mass and a single concentrated
stiffness – rather, these quantities are continuously distributed along the length of the string.
As a consequence of this spatial distribution, a multitude of Eigen-vibrations (natural
vibrations) manifest themselves (Chapters 1.1. and 1.3), all of which decay with their
individual frequency , initial phase and damping . The actual overall vibration is a
superposition (addition) of the individual vibrations that also appear in two planes each –
again with different parameters. This already rather complex description is, however, still a
simplification because we would have to consider non-linear behavior in addition, especially
for strong plucking.
Typically, low-frequency partials show long sustain while high-frequency partials decay
quickly – especially with old strings. The course of the levels of individual partials needs to
be determined frequency-selectively, e.g. using a narrow band-pass filter with its center-
frequency tuned to the frequency of the given partial. Choosing a filter bandwidth that is too
wide will make the neighboring partial influence the measuring result; with too narrow a
bandwidth, fast changes in level will not be captured correctly. From a systems-theory point-
of-view, two filters are connected in series: the string and the band-pass. The output signal
results from the filter input-signal (string vibration) convolved with the impulse response of
the band-pass filter. The narrower the band of the filter, the slower its impulse response
decays, and the less the course of the level of the partial is correctly captured.
This is an inherent problem existing irrespective of how the narrow-band filtering is achieved.
A DFT (Direct Fourier Transform) can be interpreted as a filter-band: for this the DFT-
window (e.g. Hanning) is moved along the time axis, and the now time-variant voltage of
each discrete frequency point is interpreted as time-discrete output voltage of the filter (STFT
= short-time Fourier Transform).
In the STFT, the time signal u(t) to be analyzed is first multiplied with a weighing window;
this weighing function is different from zero only for a short time. The DFT is calculated
across the signal weighted this way, resulting in a complex instantaneous value at the
individual frequency f. Then, the window is shifted by one sample period, und again a DFT is
calculated … and so on.
STFT
Convolution
Formally, the integration for the STFT happens across the infinitely lasting time t. De facto,
however, this is done merely across the window-section that is shifted by t'; the e-function is
due to the Fourier transform. The convolution integral has the same structure – its first factor
is seen as time function to be filtered. Its second factor results – as impulse response – in a
vibration of the circular frequency ω that is weighted with g(t). This shows that the STFT
works like a (digital) filter – including all associated system-typical selectivity-problems.

Fig. 1.50 Course of the level of the fundamental (G#): 40-dB-Kaiser-Bessel-window (left), 60-dB-Kaiser-
Bessel-window (right).
While merely one single (theoretical) long-term spectrum exists, there are any numbers of
short-term spectra that in some cases differ substantially. In Fig. 1.50, the same decay process
is investigated using two different DFT-windows. The beats visible in the left-hand section of
the figure are leakage effects of the DFT-window, as they would appear similarly also with
the Hamming window and the 40-dB-Gauss-window♣. Although this analysis could not be
actually termed ‘wrong’, it is more purposeful to use a window with stronger side-lobe
attenuation (e.g. 60 dB; right hand section of the figure).
A 512-point DFT at 48 kHz sampling rate will have a frequency-line distance of 94 Hz. This
frequency grid is too coarse to obtain a good resolution of an E2-spectrum (fundamental
frequency 82,4 Hz). Using an 8k-DFT reduces the line distance to 5.9 Hz; however, at the
same time the block length rises to 171 ms. Basis of the selective level measurement is now
an averaging time of 171 ms (due to the filter, with a weighting corresponding to g(t)), and
this smoothes out all quick changes in level. A compromise needs to be found between these
two extremes.
The overall level can be calculated via summation of the temporal course of the partial-levels.
However, this does not work by simply adding the dB-values; rather, it is necessary to add the
individual power data (addition of incoherent sources). Since power is always positive, the
overall level can never be smaller than the individual levels – if the latter are all measured
using the same type of averaging, that is! Given different averaging, the value of the sum can
indeed have a short-term value smaller than the individual values.
In summary, the following picture emerges: the power of the partials decays (in
approximation) exponentially while the level of the partials decreases linearly. If the
fretboard-normal and the fretboard-parallel components of the vibration show different
damping, a kink can appear in the course of the level. If moreover the frequencies are also
different, beats can result. Averaging techniques that are unavoidable when taking
measurements will smoothen-out the course of the level. Directly after the plucking attack,
the overall level is influenced strongly by the level of the high-frequency partials but these
decay rather rapidly. After a short time, a few low-frequency partials dominate: they decay
slowly. Therefore, the overall level often decays non-linearly – quickly at first, and then more
and more slowly. Because many partials are involved, there is no sharp kink but a rounded off
shape of the decay.
♣
More extensively elaborated in: M. Zollner, Signalverarbeitung, Hochschule Regensburg, 2010.
1.6.4 Old strings
For wound strings, the energy share converted into heat depends strongly on the age of the
strings. Dirt and remains of skin are deposited in the grooves of the winding; this causes
additional damping. Corrosion may also contribute. The mass introduced into the winding has
the effect of a detuning; however, the strongest impact is perceivable in the damping of high
frequency partials: an old string sounds dull. With electrically amplified guitars it does not
help to turn up the treble control, because the decay constant cannot be extended that way.
Fig. 1.51: The decay of an open E2-string: left for a low-frequency partial, right for a high-frequency partial.
“Alte Saite” = old string; “neue Saite” = new string.
In Fig. 1.51 we see the course of the partial-levels of a decaying E2-string. For the 2nd partial
(164,8 Hz), the differences between old and new string are within the limit of reproducibility:
the vibrations decay with practically the same speed. This is very different at high
frequencies: the decay duration for the old string is reduced to 1/7th. The time constant for the
decay of the old string is merely 0,1 s; under no circumstance must any measurement of the
decay therefore be taken with the FAST setting.
For the E4-string, no ageing could be found: neither with the fundamental, nor for the higher
harmonics. The string had been wiped with a cloth before the measurement, and apparently
any residue lets itself readily enough be removed from the solid strings. In contrast, simple
wiping does bring only very mild relief for the wound strings. Better results are said to be
obtained by ultrasonic baths, or boiling the strings in suitable solvents; we did not carry out
any analysis to that end.
Besides corrosion and residue, a further ageing process is to be considered: over time, the
frets grind small transverse grooves into the strings – action and homogeneity consequently
change. Mass and stiffness are not distributed uniformly along the string anymore but depend
on the location. For the model of the string, an inhomogeneous transmission line with
location-dependent wave-impedance results. Each groove makes for a small mismatch and
thus triggers minor reflections. This effect was not analyzed in the scope of this present work.
In conclusion, Chapters 1.5.3 , 7.7.6, and 7.12.2 should be mentioned: for old strings, it is not
only the decay process that is different but also the excitation. New strings sound more
brilliant because every bounce off a fret generates a broadband impulse. In old strings, the
deposits act as treble-attenuating buffer.

1.7 Lifetime of strings
How long do guitars strings last? Depends … The collector may be most enthusiastic about
that original No-Caster still carrying its original strings that after almost 70 years. The
professional may change strings after every gig – or only as a string breaks because the sound
of new strings may not be what is liked: “James, the Paula sounds so piercing.” “ Which one,
Milord?” “The one with the E.C. carved into the headstock … should be No. 8.” “Pardon me,
Sir, No. 8 is the one with the foot-long whisker pinched in the bridge; the one with E.C. is No.
38. I have just put on fresh strings, and they are not played-in yet”.
Strings almost always break at places where they are strongly bent. This is because here the
mechanical load is even higher than along the free section of the string. Thus, it would be in
the interest of longevity to round off all sharp-edged support points. At a sharp edge, the
nickel coating (that in fact provides protection against corrosion) can mutate into an electro-
chemical string murderer: if the nickel coating is damaged, humidity and sweat combine with
the two metals (steel, nickel) to form a local electric cell. The resulting electrical current leads
to subsurface corrosion and, in the end, to string breakage. Fender recommends to add a drop
of machine oil or Vaseline to the support points of the string in order to keep humidity and
sweat away from the string. That’s good advice – that needs to be supplemented with the
following: on the Stratocaster, the treble strings experience a sharp bend on an edge on the
vibrato block. It is worth the trouble to deburr that edge with a high-grade round file (similar
problem areas can be found on other guitars). Why doesn’t Fender deburr that edge in the first
place? Well … Fender does sell strings, too …
Nicely supported and guided, strings can last for months even when played frequently – but
they do sound increasingly dull (Chapters 1.6.4, 7.7.6, and 7.12.2). They will therefore be
changed before their final “snap”. Whether this happens after a few days or after a few
months – that depends to such an extent on the individual approach and taste that it is
impossible to give any guide values here. Frequent, heavy handed playing will shorten the
lifetime, wiping the strings now and then, and using care products may extend it. In any case,
when using the latter, care needs to be taken that such products are compatible with the
material of the fretboard!
2. The string as a transmission line
Within the terminology of systems theory, a special transmission channel that transmits
signals from the source to the receiver constitutes a transmission line. In the framework of
the electric guitar, our thinking in terms of a transmission line will in the first place probably
be target the guitar cable. However, while the latter does transmit electrical signals from the
guitar to the amplifier in the sense as given above, we do not need the general line theory in
order to describe its function. This is because for short lines, a simplification to concentrated
line elements is adequate. The guitar cable indeed is a short line – short relative to the
electrical wavelength that is in excess of 30 km. Transmission line theory is supposed to
describe predominantly long lines with dimensions in the order of the wavelength or a length
longer than that. In this sense, the guitar string does represent a long mechanical
transmission line. The source of the propagating mechanical wave is the place where the
string is plucked. Receiver of the signal transmitted via the string is the bridge that decouples
part of the incoming signal energy and feeds it to the guitar body. The remaining part of the
signal energy is fed back to the string as reflection. The nut (or the “active” fret) reflects, as
well, leading to the manifestation of a standing wave on the string.
String vibrations are the basis for all musical signals generated in the pickup; the following
section is dedicated to these vibrations. A pickup may also generate interference, but this will
be investigated elsewhere (Chapter 5.7). The guitar string is a mechanical system that, strictly
speaking, reacts non-linearly in a complicated manner; we will assume it to be linear and
time-invariant in order to simplify things. Given such boundary conditions, we can define – as
system quantities – masses, stiffnesses and resistances, and acting on these we have the
signal quantities of force, and of vibration velocity = particle velocity. The local distributions
of the signal quantities run along the string as a wave – the propagation speed being c. On
electrical lines, we find very similar relationships: here the system quantities are capacitance,
inductance and resistance, and the signal quantities are current and voltage. Using the
analogous mathematical description, mechanical and electrical lines will be juxtaposed in the
following. The mechanical line is the guitar string; the analogous electrical line is supposed to
serve as model for illustration it does not actually exist, and it certainly is not the guitar cable!
Translator’s remark: in this chapter, again often the bridge and the nut of the guitar are taken as the points
between which the guitar string vibrate i.e. as the string bearings. Of course, all basic considerations apply to
the fretted string in the same way – the bearings are then bridge and fret. This is not always explicitly indicated,
and therefore the term “nut” should be considered to appropriately include the term “or fret”, as well.
2.1 Transversal waves
On a mechanical transmission line, mechanical waves propagate. These waves may be

longitudinal or transversal waves, or a combination hereof. In a pure transversal wave, the
differentially small line particles oscillate laterally relative to the direction of propagation,
either in a planar movement, or in rotating fashion. In a pure longitudinal wave, the particles
oscillate in the direction of propagation; for a guitar string this would be along the string axis,
having rather minor significance compared to the transversal wave. In a simple electrical line,
an electrical field is generated between two parallel conductors. Within the conductors,
currents are flowing, and differences in electrical potential (i.e. voltages) result between the
conductors.
© M. Zollner & Tilmann Zwicker, 2002 & 2020 Translation by Tilmann Zwicker
2-2 2. The string as a transmission line
Electrical line theory distinguishes between various conductor geometries – this will not be
required for the fundamental considerations to be discussed here.
It is the local distribution of the signal quantities that propagates along the transmission line
with the speed c. Defined as a function of place and time, the force F is a signal quantity on
the mechanical line: F(z,t). Herein, z is the place coordinate in the direction along the string,
and t is the time. A first reason for misunderstandings pops up: it is not the tension force Ψ of
the string that is meant here but the wave force F. Tuning the string, a tension force Ψ is
exerted onto the string; after conclusion of the tuning process this will (ideally) remain
constant. In addition, plucking the string will introduce a lateral transverse force; this force is
meant with F. On top of the force distribution across place and time we require also a
movement quantity to describe the changing geometry. For this we basically look at the
distribution of the lateral velocity that may be converted into acceleration via differentiation
and into displacement via integration. To avoid confusion with the propagation speed c
(which is signal-independent constant), this signal speed is termed (particle) velocity v(z,t).
The signal-carrying wave quantities are thus the force F(z,t) and the velocity v(z,t). In the
important transversal wave, the direction of the latter is transverse to the string axis, for the
longitudinal wave, it is in parallel.
Either wave quantity may not be directly observed. Even as we see that a string indeed
vibrates, it is impossible to say whether the particle velocity is 1 m/s or 5 m/s. Conversely, the
displacement can be estimated – at least if it is sufficiently strong. Easiest to interpret are
therefore graphical representations of the displacement which is often designated with x or ξ.
However, ξ, is dependent on place and time: ξ(z,t). This function could be represented in
space via a z,t,ξ-coordiate-system, with ξ being the elevation above the z,t-plane. Sections
along t = t0 = const result in a place-function ξ(z,t0); sections along z = z0 = const yield a time-
function ξ(z0,t). The place-function is a snapshot showing the location-distribution of the
displacement at one point in time. The time-function is a snapshot indicating the course of
the displacement of one special point on the string. Spacial representations above a z,t-plane
do, however, have the big disadvantage that the time t is in fact not a space-coordinate. This is
not a problem for the general definition of the term “space” but it is not very descriptive for
fundamental considerations. A real problem, though, is simplifying ξ(z,t0) to ξ(z) = position
function, and simplifying ξ(z0,t) to ξ(t) = time-function. Indeed, t0 and z0 are both constant
quantities, but ξ(z) and ξ(t) remain two distinct, different functions that should not be
designated with one and the same letter ξ. We will write ξZF(t) = ξ(z0,t) for the displacement-
time-function in order to facilitate that distinction, and ξOF(t) = ξ(z,t0) for the displacement-
place-function.
For three different wave-shapes, Fig. 2.1 shows the place-function of the displacement at
seven different points in time. In each of the three graphs, a transversal wave runs from right
to left. As a contrast to the real string, the wave propagation depicted in Fig. 2.1 is not
dispersive i.e. the wave maintains its shape. In the real string, the propagation happens with a
frequency-dependent speed (dispersion), and the wave changes its shape during the
propagation, because higher frequencies propagate with higher speed. For introductory
considerations, we may neglect dispersion, but for more exact analyses it will have to be
taken into account, with c being not a constant but dependent on frequency.

Fig. 2.1: Transversal waves. Each line shows the lateral displacement of the string at one point in time. The
wave propagates (in each of the three columns) from right to left; the lower lines show later points in time. At
the far right, a short-duration lateral excitation happens, causing a wave running to the left with constant
propagation speed. The three graphs depict three different excitation functions.
In the following, ξ = ξOF(z) is to be interpreted as the analytical representation of a function,

with Fig. 2.2 showing the corresponding graphical representation (for a specific example). A
function is a rule that unambiguously allocates to each argument z a function value ξ.
Rather than the term ”allocate”, we often use “map”, and thus a function is also a mapping:
the set of z-points is mapped onto the set of ξ-points.
A transformation also is a mapping, because again sets are mapped onto each other. In the
following, the term “transformation” is – as a specialization – defined as describing the
shifting of the zξ-plane. Each point on this plane is described as a pair of values; the origin
e.g. by z = 0, ξ = 0. Shifting every point on the zξ-plane by the same distance in the same
direction results in a special transformation that in this case is termed shift or translation.
Analytical geometry of the plane calls this a parallel shift of the plane in itself – the shift
belongs to the class of concordant congruent mappings.
Fig. 2.2: Graphical representation

of the function ξ = ξOF(z). Applying
the transformation shifts the
function graph in the positive z-
direction.
Right: ct = 2.
Functions, mappings and transformations are allocation rules. For the following
considerations we will use these specializations of the terms: the z-ξ-allocation is termed
function, while the shift of all z,ξ-points that leads to a shifting of the function graph (the
function curve) is designated a transformation. The shift of the function graph in the direction
of the z-coordinate is of particular importance since this is the axis of the string (i.e. the
direction along the string), with elastic waves running along the string in that direction. The
place-function of the displacement describes the connection between the place z and the
displacement ξ. For the string, each z is tied to a distinct ξ for any special point in time.
Analytically described by ξ = ξOF(z), the function graph is a depiction of the string
displacement.
Depending on the changing time t, the function graph changes its position; it shifts along z.
Mathematically seen this shift is a time-dependent transformation (specifically: a translation).
It is either termed a coordinate transformation or an argument transformation, because the
transformation rule changes merely the function-argument z.
All values (ξ) of the function are retained; they are, however, mapped to new z-values via the
transformation.
Time-dependent translation
The transformation changes the argument of the function: z becomes z – ct0. Herein, c is the
propagation speed♣ of the wave, and ct0 is the distance covered during the time t0. We may
interpret ξOF(z) as place-function at the time t = 0, and ξOF(z – ct0) as place-function at (a
different) time t0. The function graph defined by ξOF(z) is shifted in the z-direction by the
transformation: if c is positive, the shift is towards the right, and for negative c it is towards
the left. Besides the place-function that describes the displacement for a fixed point in time t0
as a function of place, we may also consider the time-function giving the displacement for a
fixed location z0 as a function of time. If one of the two functions is known, the other can be
calculated from it.
Fig. 2.3 exemplarily depicts a triangular place function ξOF(z). The location (z) is newly
defined relative to the specific location z0 = 8 on the string: z = z0 – ct. Basis for this
substitution is the consideration that it does not make any difference for the calculation
whether the wave runs towards the location z0 or whether the observer moves towards the
wave starting from location z0. ξOF(z0 – ct) becomes the new function ξZF(t) that originates
from ξOF(z) via argument-transformation: ξOF(z) ⇔ ξZF(t). More generally: the place
function becomes the time function via argument transformation, and vice versa. ξOF and
ξZF show a similar behavior but they are not identical.
For a positive c (with the wave running towards the right), one function originates from the
other via horizontal stretching, via mirroring relative to the ordinate, and via horizontal
shifting. Although other mapping steps would also be definable, these three partial mappings
are to be considered. The horizontal stretching (performed in the direction of the abscissa)
allocates a new scaling to the abscissa: the place becomes the time, and vice versa (z = ct).
The mirroring results in a reversal of the direction of the abscissa. Both partial mappings
could also be called “stretching with negative coefficient”. As a last step, the curve – mirrored
and stretched in the direction of the abscissa – is subsequently also shifted in the direction of
the abscissa; the place function becomes a time function (or vice versa). For the wave running
towards the left (negative c), the mirroring is omitted, i.e. the direction of the abscissa is not
inverted. Both graphs in Fig. 2.3 are displacement functions; the functional connection
between abscissa and ordinate is, however, different.
Fig. 2.3: Place und time function.

The wave runs towards the right to
the point z0 = 8; the displacement of
that point is shown in the time
function. For the physical units see
the text.
“Ortsfunktion” = place function;
“Zeitfunktion” = time function
♣
In literature equations are also found that fundamentally start with a positive c and use a plus- or a minus-sign
depending on the propagation direction.

In Figs. 2.2 and 2.3, the variables do not possess any physical units – this is not unusual for
mathematical representations. We could add units, or interpret the location coordinate z in a
normalized manner … e.g. normalized to 1 m. That would make z0 = 8 in fact mean z0 = 8 m.
If, in addition, we assume that the time t is normalized to 1 s, the propagation speed for this
example would be c = z/t = 2 m/s.
In Fig 2.23 it does not make any difference whether the wave (on the left) runs towards the
observer located at the fixed point z0 = 8 at a speed of 2 m/s, or whether the observer runs
(starting at z0 = 8) towards the motionless (!) wave at a speed of 2 m/s. In both cases the
observer sees the same time function. Also (please do remain calm now, dear physicists):
waves on guitar strings do not run at light speed. Not even approximately.
The graphs shown so far have represented place- and time-functions of the displacement
because the latter is easily observed on vibrating strings. From the point of view of systems
theory, however, the (particle) velocity v is of greater importance because power and
impedance result from it (along with the force). The velocity v (at the place z0) is the partial
temporal derivative of the displacement ξ (at the same place):
Time function: displacement → velocity
With both v und ξ depending on two variables in the general representation, a partial
derivative for t is required. In it, the differentiation is done merely for t with the condition that
z = z0 remains constant:
Both Functions for the same place z0
However, place and time are interdependent via the propagation speed: z = z0 – ct. It
therefore is possible to reshape the time-differentiation d/dt into a place-differentiation d/dz,
and with this to move from the place-function of the displacement ξOF(z) directly to the place-
function of the velocity vOF(z) (chain rule of differential calculus):
Place-function: displacement → velocity
In all these equations, the sign of the velocity v is oriented relative to the direction of ξ:
movement in the direction of ξ yields a positive v. The conversion of the velocity-place-
function is done, just as for the displacement, via substitution: z = z0 – ct.
Place-function → time-function
For known place-function and known propagation speed, the time-function is unambiguously
defined – and vice versa. For known displacement and known propagation speed, the velocity
is unambiguously defined, and vice versa.
Fig. 2.4 shows the place-function of triangular-shaped displacement waves; the corresponding
velocity waves have a square shape. In the figures, z is the abscissa; seven subsequent points
in time are shown in the figures. Despite the shape of the displacement-place-function being
the same, the velocity-place-function differs in the sign. In the formula used so far, this
change of the sign has been covered by c: for waves running towards the right, c has been
defined as being positive; for waves running to the left it was negative.
Fig. 2.4a: Place-functions of the wave

running towards the left. The marked
point is first moved upwards: the
velocity of this point starts in the
positive.
“Auslenkung” = displacement
“Schnelle” = (particle) velocity
Fig. 2.4b: Place-functions of the wave

running towards the right. The marked
point is first moved downwards: the
velocity of this point starts in the
negative.
“Auslenkung” = displacement
“Schnelle” = (particle) velocity
N.B.: At the left and right border, the wave disappears from the picture frame; there is no reflection.
Displacement ξ and velocity v describe the deformation of the string; the force F may be
interpreted as their cause. As was already mentioned, it is not the tensioning force that is
meant here, but the transverse force. It is purposeful at this point to look at the electrical
transmission line rather than at the mechanical one. At the root of both lines we have the same
type of differential equation (it is merely the system parameters that are designated
differently). Considerations of analogy enable us to extrapolate from the behavior of one
line to the behavior of the other [3]. It is particularly obvious to transfer the insights gained
from the electrical line theory [5] to the mechanical line using the force-current-analogy.
Doing this, the following correspondences result: capacitance ↔ mass, inductance ↔ spring,
electrical admittance ↔ mechanical impedance, electrical voltage ↔ (particle) velocity,
current ↔ force. For reasons of simplification, we are exclusively looking at loss-free lines
with negligible short-term signal damping. Dispersion is not included in the considerations.
As a wave propagates along an electrical line, voltage and current are linked at every position
on this line by the wave impedance ZWel: . For loss-free lines, the wave
impedance is of purely resistive character (i.e. it is real). There is no contradiction here: the
line indeed accepts energy – however, this energy will not be dissipated as heat but will be
transmitted. In order to avoid reflections, we usually assume an infinitely long line. This is not
mandatory, though: as long as the wave is not facing any ‘obstacles’, we can do calculations
using the wave impedance. Applying the F-I-analogy to the electrical line yields:
Mechanical line quantities
Distinguishing it from the electrical wave impedance ZWel, we term the mechanical wave
impedance ZW.

With the length-specific mass m', and the length-specific compliance n', the mechanical wave
impedance ZW is calculated as:
Mechanical wave impedance
In this formula, Ψ represents the tensioning force of the string, ρ the density, and D the
diameter. For a 009-gage string set, 0,68 Ns/m (E2♣) and 0,14 Ns/m (E4) result – see also
Chapter A.5.
In the wave propagating without perturbation, this real quantity connects the force F and the
velocity v at every location. As example: the E4-string vibrates with an amplitude of 1 mm;
its velocity amounts to 2π⋅330s-1⋅0,001m = 2,07 m/s (330Hz, sine-shape, peak value). Given
ZW = 0,14 Ns/m we obtain for the peak value of the force-wave: F = 0,29 N. Because ZW is
real, force and velocity are in phase at every location. However, this holds only for the wave
propagating without perturbation. As soon as reflected waves are superimposed, there are
other dependencies. The table below indicates the connections between the wave quantities:
Place-function → time-function: . Time-function → place-function:
Place-function Time-function
Displacement
Velocity
Force
Applying the formulas introduced so far, place- and time-functions can be converted into each
other, and relationships between displacement, velocity and force can be set up. We have,
however, not paid sufficient attention to the sign – its definition is not as trivial as it first may
seem. For the displacement, we obtain still relatively simple relationships: the displacements
in the ξ-direction are defined positively. For a wave progressing in the +z-direction, a positive
displacement therefore implies: seen in the direction of the propagation, the displacement is
‘to the left”, while for the wave running in the –z-direction, positive displacement means ‘to
the right”, seen in the direction of the propagation.
Evidently, there are two different possibilities for the definition of the sign: either referring
to the absolute coordinates, or referring to the direction of propagation. If waves propagating
in different directions are to be superimposed, absolute coordinates are more purposeful;
with them, the superposition can be done – independently of the propagation direction – as a
simple addition. For displacement, this definition is obvious: displacements in the ξ-direction
are positive. For the velocity and the acceleration this approach is recommended, as well.
Positive acceleration therefore implies that the string moves in the ξ-direction with
increasing velocity. For the force, the following holds: a positive force generates a state of
pressure in the spring. In an upright-standing coil spring, a pressure state can be generated as
the upper end is pressed downward, or the lower end upward – both cases have the effect of a
positive force.
♣
For wound strings, the calculation needs to consider a density reduced by 10% due to the encased air.
Fig. 2.5: Line-element. The circles are transversely movable

masses; the springs model the transverse stiffness.
F2 > F1 indicates that there is a greater pressure force within
the spring on the right. Consequently, a downward-directed
acceleration (indicated by the arrow) acts onto the mass.
Given the sign convention explained above, this acceleration
is negative (negative ξ-direction).
In order to illustrate the transversal forces, the spring-mass-model according to Fig. 2.5 may
serve. The displacement ξ of the points of mass is to be seen directly as distance to the zero-
line, and the transversal force F acting within the springs can be taken from the deformation
of the springs. The acceleration forces relevant for the masses result as the difference of the
two adjacent spring-forces. The force-difference F = F2 – F1 has the effect of an acceleration
directed downwards; the inertia-formula therefore requires a minus sign. Dividing the
equation by the differential length dz of the line element, the force becomes the length-
specific force, and the mass m becomes the length-specific mass m'.
Law of inertia
The transversal force F acting in a spring depends, via the compliance n, on the change of the
length Δξ F = Δξ/n. The change of the length is the difference between two adjacent
displacements; by relating it to dz, the compliance n becomes the specific compliance n'.
Hooke’s law
The specific compliance (compliance per length) is the inverse of the tension force Ψ of the
string (to be discussed later). A further differentiation of the spring force yields two terms that
can be put into an equation:
this yields:
The differential equation derived this way is called the wave equation. It interconnects the
second place-derivative (curvature) with the second time-derivative (acceleration). The
general solution consists of the superposition of an arbitrary number of waves that each may
run towards the left or towards the right. However, the magnitude of the propagation needs to
be equal for all waves because it depends – as a constant – on the transmission line
parameters (string parameters). For waves running towards the right, we defined c (arbitrarily)
as positive, and for waves running towards the left as negative. The wave impedance ZW = F/v
is also carrying a sign; given the sign-convention used previously here, a positive wave
impedance is for the wave running towards the right, and a negative wave impedance is for
the wave running towards the left.

Fig. 2.6a: Transversal wave running towards the right.

The propagation speed c is positive, as is the wave
impedance.
Fig. 2.6b: Transversal wave running towards the left.

The propagation speed c is negative, as is the wave
impedance.
Fig. 2.6 depicts a progressing wave at the two points in time of t1 and t2; the difference in
displacement allows for deduction of the momentary velocity. For example, for the wave
running towards the right, the mass tagged with * moves downwards, and its velocity
therefore is negative. However, the force F shown here is not the inertia force but the force
transmitted in the springs. Via the place-function, the displacement F is unambiguously
determined; in order to determine v, though, we need to additionally know c.
It is not obligatory to connect the springs as shown in Fig. 2.6. Alternatively, the upper end of
the spring could be connected to the mass positioned adjacent, and the lower end could be
connected to the mass on the right. However, this connection would require reversal of the
sign of the force! As a consequence, the wave running towards the left would have a positive
wave impedance, and the wave running to the right a negative one. Both changes do not
represent a contradiction: the spring-mass-model is a direct visualization of a mechanical
tension state. To start with, the sign in this model may be arbitrarily defined – subsequently,
however, all following calculations are committed to this definition. Instead of the spring-
mass-model, it would also be possible to define place-discrete shear stresses, but again this
would entail freedom in setting the sign.
The following graph (Fig. 2.7) gives an overview for different triangular displacement waves.
Seven function graphs – positioned one above the other – indicate seven consecutive points in
time; the start is in the uppermost line each. All depictions show the place-functions along the
z-coordinate. For all examples it is assumed that a transversal wave only moves along the
string. As soon as we allow for a superposition of waves running in different directions, a new
degree of freedom is introduced for the velocity (Fig. 2.8). The force, however, is always
connected unambiguously with the displacement.
Fig. 2.7: Place-function of the displacement (= “Auslenkung”), the (particle) velocity (= “Schnelle”), and the
transverse force (= “Kraft”) for three different waves.
In Fig. 2.8 we see the superposition of two waves running in different directions. At the fifth
point in time, the velocity is zero for all points in the string. This special condition cannot be
realized with one single wave; for c ≠ 0 the displacement would otherwise have to be always
zero for the whole of the string.
Fig. 2.8: Place-function of the displacement (= “Auslenkung”), the (particle) velocity (= “Schnelle”), and the
transverse force (= “Kraft”). The sum of the force cannot be calculated from the sum of the velocity anymore.

2.2 Mirror waves as a reflection model 2-11
2.2 Modeling of reflections with mirror waves
As long as the wave impedance of a string remains the same at all locations, there is an
unperturbed wave propagation. Conversely, any local change in the wave impedance has the
effect of part of the wave reversing its direction and running back to the source (i.e. it is
reflected). Particularly strong changes happen at the bearing points of the string: the bearing
impedance F/v is very high, while v is almost zero due to the small bearing compliance. In an
acoustic guitar, the bridge needs to feature a certain compliance in order to feed part of the
energy of the string vibration to the guitar body (and have it radiated as airborne sound from
there). For the electric guitar, however, the radiation of sound via the body is not a priority;
the impedance of the bearing points is very high, and the velocity of the bearing points is
approximately zero.
A reflection at a bearing may be described in two ways: either we consider the perturbation of
the wave impedance and formulate laws for the reflection, or we ignore the change in the
wave impedance and force the bearing condition v = 0 via two waves running against each
other. Let us apply the latter approach here: the wave propagating in the direction of the
bearing is supplemented by a mirror wave that runs towards the bearing from the other side.
Both waves can run across the bearing in an undisturbed (!) fashion – just as if the bearing
points would not exist at all. The parameters of the mirror wave need to be chosen such that at
every point in time the bearing condition of v = 0 at the bearing persists. The wave and the
corresponding mirror wave add up; the sum emulates the reflection process.
Fig. 2.9 shows a triangular displacement wave running to the right towards the bearing
indicated by a vertical line. In the right-hand section of the figure, a mirror wave runs towards
the first wave; the two displacement waves are point-symmetric (for this bearing that is
defined as being un-yielding). Correspondingly, the velocity is shown in the middle graph.
Due to the point-symmetric character, displacement and velocity are always zero at the
bearing. Via the wave impedance (carrying a sign), we arrive, starting from the velocity, at the
axisymmetric force (graph on the right). However, this v-F-transformation only holds for the
individual waves but not for their sum. The actual bearing force is double the force that would
exist for the individual unperturbed wave running across the bearing. Using the above sign
definition we get: displacement and velocity are reflected with opposite phase, the force is
reflected with the same phase. It does not make any difference in the function graphs
whether we interpret the wave running towards the right in Fig. 2.9 as the cause that has as
effect a reflection running towards the left, or whether we see it the other way round (i.e. the
wave running towards the left is reflected towards the right). Identical graphs result from both
cases.
Fig. 2.9: Model of the reflection via a mirror wave running in the opposite direction. The bearing is in the middle
of each graph. We see 7 consecutive points in time from top to bottom.
“Auslenkung” = displacement, “Schnelle” = particle velocity, “Kraft” = force.
2.3 Standing waves
The waves considered so far all featured a direction of propagation: either towards the right
i.e. in the direction of increasing z-coordinate (positive propagation speed c), or towards the
left (negative c). Such waves are called propagating waves or travelling waves. They
transport active (wattful) energy: . Two superimposed, equal-energy waves
running towards each other yield zero energy flux, though. There is reactive energy in the
potential spring-energy or in the kinetic mass-energy; however, the mean power value across
full periods of the vibration is still zero.
For a transmission line terminated at its end with an infinite bearing impedance Z, it is not
possible to feed any energy to the bearing. This is because the velocity at the bearing is
always zero: . Therefore all of the wave energy arriving at the bearing is
reflected – with the amplitudes of the waves running to and from being necessarily equal. The
superposition resulting from this is designated standing wave. This term holds for every
waveshape but is particularly descriptive for sinusoidal waves (Fig. 2.10). In the propagating
wave, the amplitude is constant and the phase changes as a function of time, while in the
standing wave, the phase (as a function of place) remains constant but the amplitude changes
over time.
Fig. 2.10: Propagating sinusoidal wave (left); standing wave (right). Along the place-coordinate (z) , the
displacement is shown at three consecutive points in time. “Auslenkung” = displacement.
Literature often describes waves running on transmission lines as sinusoidal. For guitar
strings, however, we find (at least during the plucking process) a triangular shape. At the
plucking point, the string is deflected by a transverse force, and for a moment there is
(approximately) a triangular string deflection. As soon as the contact between pick (or finger)
and string breaks off, two triangular waves run from each other in opposite directions. They
are reflected at the string bearings and form – as a sum of all reflections – a standing wave.
Instead of reflections we could also define mirror waves (see the previous chapter) that run
in an unimpeded manner across the bearing points (without reflection). In that model, the
boundary conditions of the triangular excitation shape, and of the idealized bearing
condition of ξ ≡ 0 need to be respected. Given the simplifying assumption of lossless
propagation and reflection, every wave is reflected an infinite number of times. Therefore an
infinite number of mirror waves is required that all run along the string with equal magnitude
of the propagation speed. All waves running with a positive c can be combined
(superimposed) into one summation wave running to the right; the same way all waves
running to the left can be combined. The standing wave thus may be described by two
summation waves running in different directions.

Since in the present model we assume dispersion-free propagation, the propagation speed c is
not dependent on frequency. The distance in time between two reflections (of the same event)
occurring in the same direction (!) therefore is T = 2l/c for all spectral components. Herein l
corresponds to the length of the string – it needs to be run through twice until the subsequent
reflection occurs e.g. at the right-hand bearing. Given knowledge of the place-function of the
excitation, the sum of the waves is easily described with this: its place-periodicity is double
the length of the string, and the displacement- and velocity-place-functions are point-
symmetric relative to the bearing. At the point in time t = 0 both summation waves are
identical but run away from each other in opposite directions for t > 0. The term summation
wave indicates the sum of all waves travelling in the same direction. The summation wave
running towards the left needs to be added to the summation wave running towards the right
in order to obtain the actual wave on the string. [Animations can be found at: https::/www.gitec-forum-
eng.de/knowledge-base-2/collection-of-animations/].
Fig. 2.11 shows a string deflected in triangular fashion between its bearing points. The top
row starts on the left with the initial state. To the right, the two summation waves are depicted
– the displacement may be thought as both being combined. At the point in time t = 0 the two
summation waves are identical, and therefore only one single curve can be seen. In the right-
hand section of the figure we see a later point in time, with the summation waves having
already diverged a bit. The superposition of the two summation waves (second row in the
figure) gives the actual course of the displacement – which at the bearing points needs to be
zero always (unyielding bearing). In the right-hand graph of the lower row of the figure,
several subsequent points in time of the wave propagation are depicted.
Fig. 2.11: Propagation of a triangular displacement wave. The unyielding string bearings are marked as dots.
The wave runs back and forth in a zigzag shape between the dashed end positions.
“Summenwellen” = summation waves.
Books frequently depict string vibrations in a sine-shape – similar to the graphs in Fig. 2.10.
These are, however, mono-frequent special cases. The shape of the displacement is – at the
moment of excitation – triangular, as shown in Fig. 2.11. The string oscillates back and forth
in a zigzag shape; the wave-shape changes over time, though. The damping increases with
frequency and blunts the shape, and in addition dispersion occurs (the high frequencies run
with a higher propagation speed). These changes in shape are not embraced here; Fig. 2.11
shows a simplification of the basic behavior. The plane of vibration is not considered, either:
the vibration of the string is a movement in space, with rotation of the plane of polarization
occurring at the bearings. Even with the string plucked e.g. precisely perpendicular to the
fretboard, a fretboard-parallel component will emerge over time.
From the place-function of the displacement shown in Fig. 2.11, the place-functions of the
velocity and the force may be deduced – and their time-functions, as well. The velocity is the
excitation quantity for the magnetic pickup, and the force is the quantity affecting the bearing
(as it is processed e.g. by piezoelectric pickups). It was already shown that the velocity results
from the place-derivative of the displacement – however this holds only for propagating
waves and not for standing waves. In the following equation, the propagation speed c needs to
be inserted including its sign: for waves running towards the right this is by definition
negative, and thus –c becomes positive.
Place-function: displacement → velocity
The standing wave therefore is to be dissected – as shown in Fig. 2.12 – into two summation
waves. The place-derivative of each of these waves is then multiplied by –c. For triangular
excitation, the result is depicted in Fig. 2.12: the triangular displacement does not oscillate up
and down – rather, a zigzag-wave runs back and forth between the triangular border-positions.
The velocity has the shape of a rectangular impulse that is reflected in opposite phase at the
bearings. The force-wave has a rectangular shape, as well, but the reflection happens with the
same phase here. All three place-functions are standing waves aggregated from two
summation waves each. Between the summation waves, a simple conversion (ξ ⇔ v ⇔ F) is
possible, while for the actual aggregated functions (standing waves) a simple correspondence
can only by found between the displacement and the force: .
In order to be able to attribute the place-function of the force unequivocally to the

displacement, Fig. 2.13 again shows the spring-mass-model. For two conditions, it very nicely
demonstrates the triangular displacement, and the rectangular distribution of the (spring-)
force.
With the place-functions known, we can now determine the time-functions. Again this holds:
the summation-place-functions can be converted into the summation-time-functions with little
effort, while for standing waves this is not directly possible. First we will look at the bearing
force that results from two summation waves running towards each other. The starting
condition (v ≡ 0) forces both summation waves F(z) to have an identical shape at the starting
point in time (t = 0); the bearing condition (ξ = 0) forces an odd (point-symmetric) course of
the displacement ξ(z) relative to the bearing, and an even (axisymmetric) course of the force
F(z), due to the differentiation. Because both axisymmetric summation waves run towards the
bearing with equal-amount propagation speed c, the bearing force amounts – at any given
point in time – to twice the force acting on the bearing due to a single summation wave.
Therefore, the time-function of the bearing force can be determined from the place-function
of the force-summation-wave via a simple argument-transformation (z = ct), see Fig. 2.14.
The periodic time-function of the bearing force is linked to a spectrum of discrete lines with
the fundamental frequency of f0 = 1/T; T = 2l/c is the time-periodicity here. The spectral
envelope is an si-function [si(x) = sin(x)/x]; its zeroes result from the partitioning of the
place-related displacement: a string partitioning of 4:1 cancels the 5th harmonic.
Animations at: https::/www.gitec-forum-eng.de/knowledge-base-2/collection-of-animations/

Auslenkung Schnelle Kraft
Fig. 2.12: Triangle wave: place-function for 9 different points in time (T = periodicity). The velocity triangle is
reflected with opposite phase, the force triangle with the same phase. Direction of propagation: ----→, ←⋅⋅⋅⋅⋅⋅.
“Auslenkung” = displacement, “Schnelle” = (particle) velocity, “Kraft” = force.
Fig. 2.13: Triangle wave: spring-mass-model (compare to Fig. 2.6). The place-function of the force may be
deduced from the deformation of the springs. In the graph on the left, the force left of the salient point is
negative, the force to the right is positive (sign convention: compression stress = positive sign). In the right hand
graph, the force left of the salient point is positive; to the right it is negative.
Fig. 2.14: Time-function and magnitude spectrum of the bearing force; triangular displacement similar to Fig.
2.12. The average of the progression of the force over time is zero, and the DC-component in the spectrum thus
is zero, as well. Integer multiples of the quintuple of the fundamental frequency are cancelled if the distance
between bridge and plucking point is 1/5th of the length of the string (graph on the left).
Fig. 2.15 presents the result of a voltage measurement. The E4-string of an Ovation (EA-86)
was plucked using a plectrum, with the built-in piezo pickup serving as sensor. The shape of
the voltage is basically rectangular (i.e. a pulse) – the superimposed vibrations are effects of
the dispersive wave propagation. We can interpret the piezo pickup in a simplified fashion as
a force-voltage converter transforming the wave forces acting in the bridge into a
correspondingly proportional electrical voltage. The duty factor of the pulse corresponds to
the division-ratio of the plucking point on the string (32:32, 51:13).
Fig. 2.15: Electrical voltage measured in the piezo pickup built into the bridge. Basically, the shape is
rectangular – cause of the vibrations are resonances and dispersive wave propagation. Distance from plucking
point to the bridge: 32cm (left), 13cm (right); String length = 64 cm.

At the point in time of the plucking action (assumed to be at t = 0), the actual velocity of the
string is zero for all points on the string – the string starts off displaced but at rest. For the
two summation waves, opposite signs follow: vR(z, t=0) = -vL(z, t=0). Here, the index marks
the direction of the propagation: R = running right, L = running left. Moreover, the actual
velocity is always zero at the bearings (assumed to be immobile), and therefore the
summation waves need to be point-symmetric relative to each other for t = 0. Consequently,
vR(z, t) = -vL(-z, t) is valid for all points in time. From these two conditions it follows that both
summation waves are even functions for t = 0: vR(z, t=0) = vR(-z, t=0), vL(z, t=0) = vL(-z, t=0).
For the electric guitar, the string velocity is the input quantity for the magnetic pickup.
Determining the spectrum is more complicated than for the piezo pickup because velocity
sensors cannot be operated at the bridge. Typically, the magnetic pickup is located below the
string at 3 – 15 cm away from the bridge, this distance being designated zTS while the
corresponding delay time is termed τTS. To determine the string velocity above the pickup, we
start from the triangular string displacement, and require several transformations. The actual
displacement is dissected into two summation waves, and the local derivative yields the place-
function of the velocity. Then, an argument-transformation (z = z0 – ct) yields (from the place-
function) the time-function, with the time-delay τTS corresponding to a phase-shift in the
frequency domain. At the bridge, the actual velocity is the result of two components that
always sum up to zero due to the above mentioned symmetries: vΣ(t) = v(t) – v(t). At the
position of the pickup, the delay time needs to be considered with different signs: vΣ(t) = v(t
+τTS) – v(t –τTS). With the displacement law of the Fourier-transform, this results in:
Herein, VΣ(jω) is the velocity spectrum of the string at the location of the pickup; this
spectrum results from the velocity spectrum V(jω) of a summation wave via multiplication
with a sine-function. The summation wave of the velocity features a harmonic spectrum of
discrete lines with the zeroes in the si-shaped spectral envelope determined by the plucking
location – as it was for the bearing force. This spectrum is to be multiplied with the above-
mentioned sine-function, the zeroes of which are determined by the position of the pickup.
Fig. 2.17 shows the velocity spectra for an E2-string. Depending on the pickup placement and
the place of plucking, a characteristic, sound-determining envelope results (shown in red in
the figure).
Fig. 2.17: Level-spectrum of the actual string velocity at the place of the pickup. The place of plucking is 11 cm
from the bridge; the pickup is located at a distance of 15 cm (left) and 5cm (right) from the bridge. Scale length
(length of the string) = 66 cm. f0 = 82,4 Hz.
2.4 Transitory processes
Systems theory describes linear, time-invariant (LTI-) systems via their impulse response. In
fact, the impulse response h(t) is a system-quantity; it may however also be seen as a signal-
quantity found at the output of the system that in turn is excited at its input with a (Dirac-)
impulse δ(t). Using, instead of the Dirac-impulse, its particulate integral over time, the output
of the system yields the particulate integral of the impulse response: this is the step response.
Dirac-impulse and step are idealizations that, in reality, occur merely approximately. As a
pre-consideration, let us excite the string with a force step: a transverse force acting
externally on the string changes its value from 0 to F at the point in time of t = 0, with the
string at rest (not deflected) for any time t < 0. It is unimportant for the model consideration
how such a force-step can be realized, but it is important that F remains constant – and in
particular that it does not depend on the displacement. The string bearing is immobile at one
position (z = 0), and the other (right-hand) bearing is very far away. At the distance d from the
bearing at z = 0, the external force F acts on the string (Fig. 2.18).
Fig. 2.18: Place-function of the displacement. Shown

from top to bottom are 7 subsequent states. The
immobile bearing is given by a dot; a constant external
force acts at the place marked by a star. For the first 5
graphs, the mirror wave arriving from the left is
indicated by the dashed line; for the last two graphs it
is not shown. The further course of the wave is
represented as a dotted line in the bottom graph.
The wave impedance ZW is defined via the (mechanical) string data. As long as no reflection
has arrived at the excitation point (star), ZW describes the quotient between force F and
velocity v. Since the excitation point is, however, loaded by two transmission lines (to the left
and to the right), the input impedance is doubled i.e. it is 2 ZW (seen from an external point of
view). In considerations of analogy with an electrical line, we need take into account that the
F-I-analogy results in reciprocal impedances: impedance ↔ admittance. Imprinting a constant
force at the location of the star will generate a transverse movement with the constant
velocity: v = F/(2 ZW). The reflection is considered via a mirror wave arriving from the left; it
reaches the location of the star after the time τ = 2d/c (c = propagation speed). For t > τ, the
quotient between F and v is not determined by ZW anymore, because there are now two waves
superimposed at the location of the star. The two counteracting velocity-waves interact such
that the point of the string marked by a star changes its velocity from v to zero at t = τ. This
point remains at a fixed displacement for t > τ. The displacement at this location may be
calculated:
Maximum displacement at the location of the star, t ≥ τ
The parallelogram of forces yields the same value if the tension force of the string Ψ, and the
transverse force are F, are formulated orthogonally: .

2.4 Transient processes 2-19
The point in time of t = τ separates two different processes: during t < τ the transient process
(an aperiodic movement) takes place. For t > τ, the stationary final state between the bearing
(symbolized by the dot) and the point where the force is applied (symbolized by the star) is
reached.
In the case that (as is shown in Fig. 2.18) the right hand bearing is very far away, a slope
(indicated as dotted line in the figure) runs to the right without perturbation. The section of
the string between the (right-hand) bearing and the position given by the star stops moving at
the point in time of τ; it then remains at rest. However, if reflections can happen on the right-
hand side, as well, a continuously vibrating standing wave results. Still, this model does not
simulate the plucking process because in the latter the force does not jump from 0 to F but
from F to 0. Given LTI-conditions, though, an F→ 0 jump may be seen as the sum of a
negative force-step and a force constant at all times:
; ; .
The boundary conditions now are: for negative time a constant force acts on a point of the
string – the string is displaced but at rest. At the point in time t = 0, the force jumps from
to 0 with an oscillation starting that is superimposed onto the triangular displacement. The
initial situation (t < 0) is shown in Fig. 2.19. The external force (constant over time) finds
its counter-forces in the bearing forces FL and FR. While the signs of the string-internal forces
and the external forces require some getting-used-to, they are consistent. For positive time t >
0, the external force vanishes – from this point in time the two bearing forces thus need to
be void of any mean value (Fig. 2.20).
Fig. 2.19: Spring-mass-model for t < 0 (left), and corresponding string-internal force-place-function.
Fig. 2.20: Time-function for the bearing forces; at t = 0, the excitation jumps to zero. The signs of the bearing
forces are defined comparably among each other: if the string is displaced in the ξ-direction, a counter-force
needs to act at both bearings; their direction is indicated with an arrow. The string-length is M – it is divided into
a left-hand (L) and a right-hand (R) section. T1 = TR/M, T2 = TL/M, T = 1/fG.
2.5 Calculation of reflections
In Chapter 2.2, we had introduced a model of mirror-waves in order to describe reflections of

waves. In it, the wave under consideration runs across the bearing (and disappears), and at the
same time a mirror wave running in the opposite direction emerges from the bearing. As an
alternative to these two waves propagating in an unperturbed fashion, it may be expedient to
look at only one single wave that is reflected at the bearing according to certain criteria. This
reflection model gives advantages in particular for the type of modeling of the string that uses
delay units.
2.5.1 The reflection factor
Every propagating wave (travelling wave) transports energy: in the electrical transmission
line this is the energy of the magnetic and electric field, while in the mechanical line it is
kinetic and potential energy. The mean values of the mechanical energy calculate as:
The two transported (mean) energies are equal at each place of the transmission line. As the
wave arrives at the end of the string, this energy cannot disappear into nothingness; it is either
coupled into the bearing (and transported further there, or dissipated) or it is (fully or
partially) reflected.
All bearings show complicated bearing impedances. The bearing impedance is anisotropic
i.e. depends on the plane of vibration, and it is dependent on frequency. The compliance is the
inverse of the complex bearing impedance and is defined as a complex admittance:
Admittance = conductance + j ⋅ susceptance
Impedance = resistance + j ⋅ reactance = 1/admittance
An unyielding, rigid bearing (small admittance, high impedance) can absorb forces but does
not allow for movement; the compliant bearing behaves conversely. Strings are anchored in
relatively unyielding bearings. For the electric guitar only, the bearings may be totally rigid –
in the real world, such an ideal is of course not possible. If the bearings on an acoustic guitar
were fully unyielding, they could (due to v = 0) not receive any energy from the string, and
could not transmit it further to eventually radiate sound.
The bearing impedance (or admittance) connects the two field quantities of force and
(particle) velocity; their product is the power P. The requirement for continuity demands
Fstring = Fbearing and vstring = vbearing. On the string, the quotient F/v is equal to ZW for the
propagating wave, but at the bearing, this quotient may take on any value. At first, this
appears to be a contradiction. If a 2-N-force-wave runs through a transmission line of a wave-
impedance of 1 Ns/m, the velocity is 2 m/s. As this wave now encounters a bearing of a
bearing-impedance of 10 Ns/m, the bearing cannot fully absorb the wave energy. The bearing
“extracts” from the arriving wave that part of the energy that matches the bearing impedance
in terms of the F- and v-components. The remainder of the energy is “sent back”.

2.5 Calculation of reflections 2-21
Therefore, two waves (incoming and reflected) running in opposite directions are
superimposed at the bearing and at every point of the string. Force and velocity thus result
from the sum of two values. The wave running in the opposite direction at the plucking
location has to be considered including its reflection – and the subsequently generated
reflections, as well. All waves are reflected after having run the length of the string, i.e. more
and more waves superimpose. The sum of all superimposed waves results in the steady-state
condition that may be calculated via the tools offered by network analysis. Calculating the
line impedance Z(z) for this steady-state condition at any arbitrary point z will not yield the
wave impedance (at least not for the general case).
At this point, the wave “sent back” in the above example is unknown. In order to calculate it,
we formulate the force F(z) and the velocity v(z) acting at each point of the line as a sum of
two waves♣. The waves Fh(z) and vh(z) running towards the bearing are given; knowing one of
the two is sufficient; the other can be calculated from it. The reflected waves Fr(z) and vr(z),
while also linked via ZW, are yet unknown. The bearing impedance delivers the missing
condition, because at the bearing point (e.g. at z = 0) the quotient of F(z = 0) and v(z = 0) is
equal to the bearing impedance ZL. As has already been the case a number of times, the sign
springs a surprise: at the right-hand bearing, ZL = F/v holds, and at the left-hand bearing, we
have ZL = – F/v. This reversal of the sign is easiest seen in Fig. 2.5: a left-hand bearing can be
generated by making F1 = 0; the left-hand mass is now removed, and the formula indicated as
“law of inertia” carries a minus-sign. Similarly, F2 = 0 yields a plus-sign. The wave
impedance includes its peculiarity in terms of the sign, too: for waves running towards the
left, ZW is negative, for those running towards the right, it is positive (Chapter 2.1).
Superimposing the waves running back and forth we would have to do the math with two
different wave impedances. However, for the following calculations ZW is strictly positive –
for the waves running to the left we insert a minus-sign. In the below calculation we consider
a wave running (“hither”) towards the left onto the left-hand bearing (z = 0), and a wave
reflected towards the right:
The ratio of the complex amplitudes within the back-and-forth-running wave is the complex
reflection coefficient r. It is dependent on the wave impedance ZW and on the bearing
impedance ZL:
Reflection coefficient
There are three interesting special cases: for (matching condition), the reflection
coefficient becomes zero: the wave is not reflected and the bearing absorbs the whole of the
wave energy without reflection. For , the reflection coefficient of the velocity
becomes +1: the velocity wave is completely reflected with the same phase, and the force
wave is completely reflected with opposite phase.
♣
F, v, and Z are complex; we make do without the underscoring here.
For , the reflection coefficient of the velocity becomes −1: the velocity wave is
completely reflected with opposite phase, and the force wave is completely reflected with the
same phase. This is the case of the unyielding bearing where the velocity of the string is
always zero. Of course, a guitar string must not be operated with r = 0 – otherwise a
“periodic” vibration would never come into being. With r = ±1, the vibration would never
decay – at least within the idealizations underlying here.
In Chapter 1.6, investigations regarding the decay process of the string vibration were
introduced. It the vibration of an E2-string decreases (strictly exponentially) e.g. by 60 dB
within 12 s, it decays by 0,06 dB per 12 ms (1 period), corresponding to 0,7%. The reflection
coefficient therefore is 0,993 per period. Since the wave on the string is reflected twice per
period, this absorption of 0,7% needs to be divided up between bridge and nut (or fret), e.g.
0,3% at the nut/fret and 0,4% at the bridge. Typically, a refection coefficient of close to 1 is
found.
Given strictly real bearing impedance, the reflection coefficient is real because ZW is real, as
well. For a real r, the phase shift between the original and the reflected wave is either 0° or
180°. In contrast to the reflection at an imaginary bearing impedance, the amplitude of the
reflected wave is now smaller than that of the original wave. For a guitar string, the bearing
impedance ZL is large compared to ZW, yielding the following as an approximation:
A negative-real reflection coefficient indicates that the velocity-reflection happens with the
opposite phase. If the real part of the reflection coefficient is not zero, active energy flows
into the bearing points (dissipation, string damping). It makes no difference for the string
whether this energy is radiated from the guitar body, or is converted directly into heat within
the bearing – the drained energy is not available anymore as vibration energy.
The other extreme would be a purely imaginary bearing impedance as it is formed by a mass
or a spring. Even if the bearing is composed of several masses and springs, at any one single
frequency there will be either one inert or one stiff bearing impedance. For a purely
imaginary bearing impedance, numerator and denominator of the reflection coefficient are
complex conjugate; the absolute value of r therefore is 1. That is exactly 1! The waves
running back and forth are phase-shifted relative to each other, but the absolute value is
conserved: the vibration energy does not decrease. However, since the phase of propagating
waves changes as a function of the place (wave equation), a phase-shifted reflection may be
seen as non-phase-shifted reflection from another place. We can imagine that the wave is
reflected without phase shift but at a small distance behind the bridge, with the phase shift
resulting from this detour corresponding to the actual reflection. Depending on the sign it may
be necessary to shift this imagined reflection place ahead of the bridge. The same holds for
the nut (or fret). The effective string length may therefore differ from the geometric one:
depending on the bearing impedance, and on the frequency, the length may be longer or
shorter. This influences the frequency of the partials:
A springy bearing extends the effective length of the string, and it decreases the vibration
frequency; the softer the spring, the lower f is. A mass-loaded bearing shortens the string and
decreases the frequency; the lighter the mass the higher f is.

2.5 Calculation of reflections 2-23
The reflection coefficient of the real string has both a real and an imaginary component, with
both depending on the frequency. The real part causes the damping of the string, while the
imaginary part has a detuning effect. In addition, string-internal mechanisms need to be
considered – the present chapter is dedicated to the loss-free transmission line.
EXAMPLE: a tensioned string (L = 64 cm, ρ = 8⋅103 kg/m3, S = 0,5 mm2, Ψ = 100 N) is

suspended immovably on one side and springily on the other with s = 10.000 N/m, s ≠ s(f).
From this follows: ZW = 0,633 Ns/m, c = 158,1 m/s, fG = 123,5 Hz (without influence of the
spring). Considering the elastic edge suspension, the fundamental frequency fG decreases:
The absolute value of the reflection coefficient is 1, the angle is smaller than 180° by 5,6°.
Running through a full length of the string, the phase of the wave is changed by 180°; a phase
delay of 5,6° corresponds to a path-length of 2 cm. The one-sided elastic suspension
effectively lengthens the string by 2 cm, decreasing the fundamental frequency to 119,8 Hz♣.
The relative detuning is identical for all harmonics (disregarding the dispersion).◊
2.5.2 A resonator serving as bearing for the string
Any real bearing of a string needs to feature not only components behaving like springs, but
also masses – and that makes bearing resonances unavoidable. At the resonance frequencies,
the reactances (or conductances) compensate each other. Impedance and admittance are
exclusively real. At all other frequencies, impedance and admittance remain complex [3].
As an example, a loss-free spring/mass-system will be investigated in the following. The

impedance of its bearing computes to:
For the impedance of the bearing becomes zero (no force despite movement), while
for the bearing acts like a spring (spring-controlled). For it acts inert (mass-
controlled). Below resonance, a string coupled to the bearing is in effect elongated. Above
resonance, it will in effect be shortened. Even assuming the string to be dispersion-free, the
frequencies of the partials are not laid out harmonically anymore: below the resonance
frequency of the bearing, the frequency of the partials decreases, and above the resonance
frequency of the bearing, it increases. The reflection coefficient for the velocity is:
The frequency dependence of the reflection coefficient leads to a 2nd-order rational

function. The even numerator- and denominator-potencies are identical, while the odd ones
have an inverted sign. Numerator and denominator thus are complex conjugate relative to
each other. This kind of frequency dependence is termed all-pass function.
♣
Real bearings are much stiffer; with them the detuning is smaller.
The magnitude of an all-pass function is 1, and the phase shifts by for , n being
the order of the all-pass function. For f = 0, rv = –1 holds: the velocity wave is reflected with
opposite sign. For we obtain rv = +1; for we again get rv = –1.
Therefore, having a resonator terminating the transmission line has the effect of an additional
phase shift. Natural vibrations (partials) occur at those frequencies where the phase shift for a
full travel-path on the string (2L; back and forth) is an integer multiple of 2π. Assuming
dispersion-free wave propagation on a fully clamped-down string, partials at integer multiples
of the fundamental frequency result. However, if a bearing acts as a resonator, an additional
phase shift is introduced that generates (in our example) an additional partial. For resonators
of higher order, several additional partials occur.
Fig. 2.21: Phase shift along a full travel path

along the string. One string bearing is configured
as a resonator resonating at 1,415 kHz. An
additional natural frequency is the result of the
narrow-band additional phase shift.
Fig. 2.21 shows the phase shift occurring for a string vibrating at a fundamental frequency of
100 Hz and a full travel path (double the string length). The phase is negative as it is
customary for delays in recent literature. One string bearing is configured as a resonator with
a resonance frequency of 1,415 kHz (dot on the abscissa). At the bottom of the graph, the
frequencies of the partials are indicated with bars. The partial at 1,4 kHz is substantially
detuned downwards by the bearing resonance, and an additional partial is generated at 1,42
kHz. All other de-tunings are too small to be recognizable in the figure.
The spectral derivative yields the group delay (Chapter 1.3.1). The slope of the
phase function is virtually constant with the exception of the range around the bearing
resonance. Thus, the group delay is also generally constant – only in the range of the bearing
resonance it becomes longer. This leads to a warping in the spectrogram (Fig. 1.8).
2.6 Line losses
Ideal masses and springs store energy but do not dissipate them as heat. These elements are
therefore termed “loss-less”. In contrast, any real string also features friction-resistances that
irreversibly convert the vibration energy into caloric energy. Line theory considers these
energy losses via distributed, differentially small resistances. It is insignificant for the model
whether the losses are due to mechanical friction in the string (inner damping), or result from
the string directly radiating sound energy (i.e. without detour through the guitar body).

2.6 Line losses 2-25
With a level decreasing by 10 dB within 10s in the decay process, a low E-string looses about
2,8% of its vibration energy per period of the fundamental. The string therefore is included in
the groups of weakly damped systems (Q-factor Q = 2249) – it may be seen as a transmission
line with low losses in good approximation. Considering moreover that the main share of the
measured losses is not from the string itself but from the bearings shows that this
approximation is justified to a high degree.
For transmission lines with low losses, the assumption is that phase- and group-velocity are
practically not affected by the damping. It is only the vibration amplitude that decreases
slightly as the signal passes through the line. For line lengths in the range of the length of a
string, the amplitude damping is so small that it may be disregarded altogether in many cases.
However, if the signals are subjected to numerous reflections, and if the objective of the
investigations is a decay process lasting several seconds, then the amplitude damping must
not be ignored anymore. It is not always necessary, though, to formulate a differentially
distributed damping: insofar as merely discrete points on the string are of interest, ladder
networks consisting of loss-free delay elements and delay-free damping elements provide a
useful model (Chapter 2.8).
Trying to calculate the internal losses in the string brings curious issues to light: the loss
factors for steel given by different books differ by a factor of 14. Even in one and the same
book we may find differences of 600%. That may be because microphysical loss effects
depend on manufacturing processes, or because there is not the one steel. It is more likely
though, that ‘internal’ losses also include radiation losses. A loss factor of d = 0,0001 (Gahlau
et al., Geräuschminderung durch Werkstoffe und Systeme, Expert Sindelfingen 1986) appears
plausible; it yields a level decay of 0,22 dB/s for 82,4 Hz – significantly less than that of
typical measurement results (0,6 dB/s), and leaving room for further damping mechanisms.
The d = 0,0006 specified only 14 pages on in the same book, however, is too high (1,3 dB/s).
We probably better abandon hope for any consistent terminology – all too entrenched are the
habits. Terms like damping factor, damping coefficient, degree of damping, loss factor, etc.
may certainly (?) be applied in a consistent manner within one and the same publication, but
interindividual differences are the rule. It is therefore not surprising that an author specifies
the aperiodic boundary case (called critically damped oscillation elsewhere) with d = 1, while
another (equally renowned) colleague specifies d = 2 for the same case. You can live with
such a scenario  – but you gotta be aware (sapienti sat).
The situation is more conducive for the calculation of direct radiation losses. Under the
heading “air damping”, we find in [9] formulas for the radiation of active energy, and
evaluations for bass strings. The losses mount with increasing frequency, and decrease as the
string-diameter grows. The calculations in [9] relate to the damping of the fundamentals –
higher harmonics tend to be radiated less well implying lower string damping♣. For guitar
strings, calculations yield radiation-induced time-constants of the amplitude in a range of 20 s
(open E2-string) to 2 s (open E4-string). We can therefore disregard radiation losses for the
low guitar strings, while for the high strings these losses are at the borderline (measurement
values are e.g. 1,7 s).
♣
In addition, we can consider that fretboard and guitar body are located in close vicinity of the string and act as
reflectors. This compounds the calculation of the radiation impedance.
As a bottom line, we may state: inner damping and radiation losses may be disregarded as
long as merely the wave propagation along short sections of the string is discussed. When
analyzing vibrations of longer duration, we find – in electric guitars – damping mechanisms
having a greater effect towards the higher frequencies (Chapter 7.7), and additional
frequency-selective absorptions (e.g. resonances of the bridge). For acoustic guitars, we need
to expect substantial absorptions in the low-frequency range, as well, since a non-negligible
share of the vibration energy is fed to the bearings (bridge, frets).
2.7 Dispersive bending waves
The simple transmission line theory assumes place-independent wave impedance and
frequency-independent propagation speed. However, the transversal waves of the guitar string
propagate in a dispersive fashion, i.e. with frequency-dependent speed. The high frequencies
run faster than the low ones (Chapter 1.3.1). The reason is the bending stiffness that increases
the transverse stiffness, the latter in turn depending on the tensioning force.
Modeling the string as a dispersive transmission line takes much effort and is not always
necessary. In most cases, only two or three points on the string are of interest (nut/fret, bridge,
and point of plucking). Possibly, the position of the pickup also needs to be added in. It is
easy to model the parts of the line between the discrete points via all-passes (Chapter 2.8).
However, if precise description of the reflection conditions is required, we need a more
detailed model. The simplest solution is found for steady-state (mono-frequent) partials:
propagation speed and wave impedance are only weakly dependent on the frequency. For
narrow-band considerations they may in fact be assumed to be constant. Transient processes
extend across a frequency range, though; in such cases we need to apply frequency-dependent
quantities.
We had introduced a simple element for modeling the dispersion-free string in Abb. 2.5. As
characterizing quantities, force and velocity were sufficient (both quantities being signal-,
place- and time-dependent). However, the rigidity of the real string requires that in addition to
the (transverse) force F, a place- and time-dependent bending moment M is specified, and
also that we introduce an angular speed w. This gives us a frequency-dependent phase delay
(Fig. 1.6). The dispersive line element cannot be described as a quadripole (two-port
network); rather, we need to specify a four-port network (octapole) [11]. The input
quantities of the latter are F1, M1, v1, w1; its output quantities are F2, M2, v2, w2. Because the
transverse dimensions of the string are small relative to the wavelength, we may disregard
shear deformations and rotational inertia moments (Euler-Bernoulli theory for beams). Thus,
the length-specific mass m', the length-specific compliance n', and the bending stiffness B
remain as the system quantities (inside the four-port network).
The rigid string features two wave impedances ZF = F/v and ZM = M/w, and two wave powers
PF = Fv and PM = Mw. Two bearing impedances each are active at both string bearings
(nut/fret, bridge), and in addition the four signal quantities may be intercoupled in each
bearing. For example, the edge-force may generate an edge-moment, or a displacement will
necessarily lead to torsion. Since all these relationships appear depending both on frequency
and direction, simplifications and approximations are indispensable.

2.7 Dispersive flexural waves 2-27
Waves of lower order (fundamental and low-frequency harmonics) are not influenced much
by the rigidity. The effective overall rigidity is practically only determined by the tensioning
force Ψ, with the dispersion remaining insignificant (Fig. 1.4). However, for higher-order
partials the influence of the rigidity may not be ignored anymore – especially for the low
strings. The (overall-) rigidity as it is significant for the higher partials consists of two
components: a frequency-independent portion caused by the tensioning force, and a
frequency-dependent portion caused by the rigidity. The differential equation for the bending
wave is of 4th order; we therefore require four boundary conditions, and four independent
fundamental solutions are possible. As had been the case for the rigidity-free string, a wave
running forward and a wave running backward appear in the longitudinal direction, but in
addition, an exponential fringe field is superimposed close to the bearings. Fortunately, this
fringe field decays already at a short distance, and further away from the bearings we may
therefore do the math with only one wave type. Without the fringe field, we obtain a simple
coupling between F, v, M, w: knowing one of these four quantities suffices to describe the
other three. One single wave equation is good enough to describe the string vibration (in one
plane); we need a frequency-dependent wave number k(ω) for it, though.
This simplification is not valid for the description of reflections, though, because the latter
indeed occur especially within the fringe zone. In this context, “fringe” refers to the
beginning and the end of the string, and not the mantle-surface of the cylinder. Within the
fringe zone, we need to formulate – in addition to the wave equation – a fringe field with its
own wave number k', designated fringe-field number. Although in the fringe field the signal
quantities F and v are still linked via ZF (as M and w are linked via ZM), F may take on any
value independently of M (and the other way round) due to the fringe field. While in the
dispersion-free string the reflection coefficient depends only on the ratio of wave impedance /
bearing impedance, two wave impedances and two bearing impedances (per bearing each!)
define the reflection coefficients in the stiff string. Thus, it is (at least theoretically) possible
to reflect the Fv-wave entirely at the bearing, and to entirely absorb the Mw-wave. This does,
however, not mean that there is no Mw-wave running in the reverse direction: the fringe field
will take care of the existence of an Mw-wave already at a short distance – the energy
necessary for this is “withdrawn” from the Fv-wave.
Within the abundance of all the reflection conditions possible in every vibration plane, there
are some special cases that may be easily analyzed:
• Open end of the string: the string ‘dangles in the air’; its end cannot absorb any
transverse force F, nor any moment M. While this seems rather lacking in practical
relevance, it may appear at resonance.
• Clamped string: transverse velocity v and angular speed are zero.
• Guided end: angular speed w and transverse force F are zero.
• Supported string: transverse velocity v and moment M are zero.
The real string bearing is not represented in any of the above special cases. This is because the
string does normally not end at the bearing but is guided across it. Often the string rests in a
small notch that permits for line-shaped contact only. This inhibits any transverse movement
but allows for forces, angular movements and moments. If we interpret this bearing as a large
blocking-mass, it will reflect Fv-waves but not Mw-waves! For the extreme case of a string
featuring a stiffness that is only determined by the bending stiffness (beam), a barrier-mass
reflects 50% of the incident wave energy – the other 50% are coupled as a bending wave into
the section of the string beyond the bearing. In the other extreme case (B = 0), though, 100%
of the energy is reflected.
In order to assess the significance of the bending stiffness, let’s look at the following model
case: the string is supported by a knife-edge bearing not allowing for any lateral movement.
The string also continues indefinitely beyond this first bearing, the other bearing has ideal
reflecting characteristics. The percentile energy-portion transmitted beyond the bearing is
shown for an A2-string in Fig. 2.22. At low frequencies, the bending stiffness is negligible;
the energy is almost completely reflected. However, already from middle frequencies a
significant percentile is coupled across the (immobile!) bearing. On the other side of the
bearing, we do not see a pure Mw-wave; rather, the fringe field again takes care of generating
a combination of Fv- and Mw-waves.
Of course, a real string cannot extend indefinitely; it ends after a few centimeters at the tuner
(“machine head”), in the string retainer, in the body, or wherever else there is space to attach
it. Fig. 2.22 clearly indicates that it does make a difference where and how the string is
fastened, though. The string-part beyond the bearing may indeed tap considerable vibration
energy if it has corresponding length, forming a coupled resonator. Still, the power-percentile
shown in Fig. 2.22 is not necessarily lost at each and every reflection. The share of energy
coupled across the bearing may itself be reflected e.g. at the tailpiece, run back to the bearing,
and then is once more coupled across the bearing into the main part of the string. Also, the
real string does not have a line-shaped contact to the bearing: via a contact area (groove), not
just a pure transverse force may be received but a moment as well. Some bridge/nut-
combinations are deliberately (?) designed with larger contact surface, or directly as
clamping-devices. For the latter, a further model-case will be discussed at the end of the
chapter.
Fig. 2.22: Degree of power transmission (“Leistungs-Transmissionsgrad”) for an A2-string (40 mil); one end is
clamped, the other borne on a knife-edge. On the right, the decay time (“Abklingzeit”) of the partials purely due
to the transmissions is given for a level-decrease of 30 dB via with three calculated lines (decay time T30,
Chapter 7.6.3). The ratio of core diameter to outer diameter is κ = 50% (––––), or 33% (-----), or 100% for the
solid string (”Massivsaite”).
The grey areas show results of measurements (A2, 40 mil, κ = 50%) taken on the stone table. For a, the string
was clamped at both ends, for b one end was clamped and the other supported: remaining string length is 30 m,
weakly damped.
Case a fits well to the “orientation line” presented in Chapter 7 (Fig. 7.66); in addition to the bearing, string-
internal damping mechanisms are at work, as well.
Case b should be compared to the 50%-line above it. This (calculated) line considers only the absorption
occurring at the support-type bearing. In contrast, the measurements (grey area) also include string-internal
damping mechanisms, and the absorption at the other bearing (clamp).
In an E2-string, the losses due to the transmission are even larger.

To calculate the conditions for vibration and reflection, the string is divided into small
cylindric sections of the length dz. At rest, the circular separation planes (cross-sections) are
perpendicular to the z-axis. As the string is excited, the cross-sectional surfaces remain flat
but are not in parallel anymore due to the bending moments: they form an angle of curvature.
The laws of motion, inertia, and strength result in a partial differential equation for the rigid
string (for detail see the supplement):
Differential equation for the string
The differential equation (DEQ) is a partial one because it includes the derivatives for both
place z and time t; it is linear because the variables of transverse displacement ξ, place z, and
time t are present in the first power only; it includes constant coefficients because the system
quantities of tension force Ψ, bending stiffness B, and length-specific mass m' are not
dependent on z and t (idealized); and it is homogenous because it does not comprise an
external excitation.
B and m' are determined from the material data and the geometry of the string; the tension
force Ψ results from the required fundamental frequency fG. Any function ξ(z,t) that will
satisfy the DEQ is a solution for it. According to DANIEL BERNOULLI, the solution for
sinusoidal movement is formulated as a product including a purely time-dependent and a
purely place-dependent factor:
Solution approach
The first factor ξ includes the angular frequency ω and the initial phase ϕ; a partial
differentiation regarding the time t becomes a multiplication with jω. The second factor holds
the wave number k; a partial differentiation regarding the place z becomes a multiplication
with –jk. Introducing the corresponding derivatives into the DEQ yields:
Characteristic equation
The characteristic equation may be cancelled by ξ (the case ξ ≡ 0 being trivial). This yields a
conditional equation for k that includes only a dependency on the system quantities. Because
this equation is of 4th order, there are four independent solutions for which four independent
boundary conditions need to be specified. In terms of the solution approach, two k-values are
real, and the exponent therefore is imaginary (-jkz). This describes a sinusoidal wave running
to the left or to the right, respectively. The other two values for k are imaginary, and the
exponent thus is real – describing an exponentially increasing/decaying fringe field
originating from the string bearing. Only the decaying fringe field is of practical importance.
The general equation of motion is a superposition of the two wave equations and the equation
of the decaying fringe field:
general solution
The time-dependency is found in the three independent complex amplitudes , the frequency
is identical for all three components.
In the following, we will consider a string ( ); the (left-hand) bearing is located at z = 0.

The expressions
, ; Excitation
describe a sinusoidal wave running left towards the bearing (ξ = displacement). A part of its
energy is reflected at z = 0, the remainder is transmitted:
, ; Reflection, transmission
In the general case, reflection coefficient ζ and transmission coefficients ψ are complex. On
the considered section of the string ( ), three displacements are superimposed:
γ represents the complex fringe-field coefficient. Beyond the bearing (i.e. in the range of the
transmission) two displacements are superimposed:
Here, too, a fringe field with a different wave number k' and fringe-field coefficient δ is
generated (in addition to the transmitted share). Fringe fields and waves are functions of the
place (z) and the time (t). The dependency on time is described by ξ with ω as circular
frequency; the place-dependency is described via the fringe-field numbers k and k'. For the
propagating waves, is reciprocal to the wavelength λ. The fundamental frequency
fG is the lowest eigenfrequency (natural frequency) of a string; λG corresponds to double the
length of the string – in the E2-string this is about 1,3 m for 82,4 Hz. Partials in the range
around 10 kHz therefore have a wavelength around 1 cm (λn =λG/n). This still much exceeds
the string diameter – we thus may do the math using approximations. For the high strings,
these conditions are met to an even higher degree.
From the fringe-field number k', a limit distance zg = 1/k' may be estimated; it indicates at
which distance the fringe field has decayed to 1/e. Since the characteristics are those of a
flexural wave, the calculations require somewhat more effort (in particular for the wound
strings). Fig. 2.23 shows typical values of zg.
Fig. 2.23: Limit value zg of the fringe field (in mm)

E2-string: 53 mil, κ = 0.4
E2-string: 42 mil, κ = 0.4 ----
G3-string: 24 mil, κ = 0.5
g3-string 20 mil, plain
e4-string: 12 mil, plain
“Randfeld – Grenzdistanz” = limit value zg of the

fringe field

Running towards the bearing, the excitation wave is specified by its amplitude and its
frequency ω. For this same wave, F, M and w are defined via the system quantities B, m' and
Ψ, and so are the wave impedances ZF and ZM, as well as the velocity v with v = ∂ξ/∂t. The
bearing (at z = 0) is – to begin with – defined by its two bearing impedances ZFL = F(0) / v(0)
and ZML = M(0) / w(0). Considering the string to be a linear system, there is the superposition of
three oscillations in the range of : the given excitation wave ( ), the reflected wave
(ζ), and the fringe field (γ). At first, ζ and γ are two unknown quantities; however, they may
be calculated via the two bearing impedances.
The system quantities of the string are tension force Ψ, length-specific mass m' = ρ S, and
bending stiffness B = ES2 / 4π. Herein defined are ρ = density, S = cross-sectional surface, and
E = Young’s modulus. For wound strings, it is predominantly the core that defines the
bending stiffness; the densities of core and winding may differ [appendix]. From these
quantities, the wave number k and the fringe field number k' may be calculated:
Both k and k’ are system quantities, as well, i.e. they are signal independent. The rigid,
tensioned string can be transformed into two borderline cases by varying B and Ψ: for B = 0
we obtain the dispersion-free string (fully flexible), and for e Ψ = 0 we get the cantilever
(without any tensioning force). The wave numbers are calculated as:
String without bending stiffness
Beam without tensioning force
The phase velocity c is frequency-dependent for , and for it is constant. The

wave-reflection coefficient ζ is calculated as:
The formulas now do start be become rather lengthy – but they still do not fully describe the
bearing. In fact, the simplification based on two bearing impedances ZFL and ZML (as it is
sometimes found in literature) is not always sufficient. In the general case, a coupling
between the transversal quantities F or v, and the bending quantities M or w, respectively,
may occur; the bearing impedance in that case receives the form of a matrix, and moreover an
additional coupling term. Using this, a formal description is still explicitly possible, but the
practical use of the formulas is increasingly limited because the individual bearing quantities
cannot be measured with sufficient accuracy anymore. The vast diversity of bearing
parameters forces to simplify – and it calls for the question how well these simplifications fit
in the individual case.
The string vibration may be approximated in different ways:
a) The simplest approximation describes the string without its bending stiffness. The partials
are positioned harmonically, and the propagation velocity is frequency-independent. To
describe the bearing and the reflection, a single bearing impedance ZFL is sufficient – it may
be determined e.g. with an impedance head. For the fundamental fG and the lowest partials,
this approximation is adequate in many cases, but already in the middle frequency range we
recognize clear deviations between calculation and measurement (Figs. 1.5, 1.7).
b) The calculation of the partials with consideration of the bending stiffness represents an
easily obtainable improvement. On average, the actual spreading of the partials is quite well
met. Considering moreover also the dilatational waves (Fig. 1.17) yields a useful
approximation for the level spectrum.
c) In order to calculate the decay processes, the bearing impedances need to be known. For
very light strings, we may disregard the bending stiffness, but for heavier strings knowledge
of the bearing impedance ZFL is required besides knowledge of the bearing impedance ZML.
d) The supposed “fully comprehensive” description of the bearing quickly degenerates into a
confusingly extended system of equations: in two orthogonal vibration planes, we need to
define three bearing impedances each – not to forget additional coupling impedances between
the two planes. In addition, the impedance of the longitudinal wave should be borne in mind,
again including mode coupling to the two orthogonal transversal waves. Presumably, a torsion
wave on the string may be ignored – but this assumption is still under scrutiny: for the bowed
string, the torsion wave is significant. Since all bearing- and coupling impedances depend (in
some cases strongly) on the frequency, a confusing multitude of parameters results.
The next example shows that the bending stiffness of the string can make for problems even
at low frequencies although the tensioning stiffness should in fact be predominant in this
frequency range. For the calculation, we assume an idealized support bearing that is immobile
in the transverse direction. The transversal velocity therefore is zero at this bearing. However,
for bending processes that are coupled to the angular speed, this bearing is supposed to feature
a moment of inertia (blocking mass). Due to the (material- and geometry-dependent) bending
stiffness of the string, and due to the inertia of the bearing, a resonance may arise that
(depending on the circumstances) may absorb a significant part of the vibration energy, or
may couple this energy into the section of the string beyond the bearing (total transition).
For a very small or a very large blocking mass, the resonance frequency will appear at very
high or very low frequencies – it will then not cause any disturbance. However, given a
corresponding dimensioning, resonances can appear in the middle frequency range, as well.
Such resonances are not generally undesired – possibly, the luthier seeks to obtain a
somewhat stronger absorption exactly in that frequency range. However, to use this in a
targeted manner, the (frequency-dependent) moment of inertia at the bearing would have to be
known – this poses problems for the instrumentation.
The following calculation circumvents the instrumentation issue and defines idealized bearing
parameters; the approach does not orient itself on a special realization. In the discussion of the
cone-parameters following later, we will again look into this subject matter, and we shall dive
some more into the details.

As EXAMPLE, we look at a wound E-string of a diameter of 46 mil. It rests on a bearing such

that its lateral movement is zero; ZFL thus becomes infinite. The reflection coefficient:
therefore is simplified. It reads:
The bending impedance ZML of the bearing is negative because the excitation wave runs
towards the left:
Θ = moment of inertia of the bearing, W = bearing resistance
Using this, the complex reflection coefficient can be calculated:
Together, the bending stiffness B and the moment of inertia Θ form a resonance that can be
located e.g. in the range of the middle frequencies (Fig. 2.24). With a suitable choice of W,
total absorption is possible within a narrow frequency range. Such extreme cases may not be
expected in typical string bearings, but it is still clearly evident that the bending stiffness can
have effects at middle and low frequencies, too. ◊
Fig. 2.24: Magnitude of the reflection coefficient of an

E2-Saite, 46 mil, core/outer diameter 50% (κ = 0.5).
The bearing is unyielding in the transversal direction,
but has a moment of inertia Θ towards bending stress.
Θ = 4,2*10-8 kgm2 corresponds to the rotation moment
of inertia of a steel ball of a diameter of 10 mm.
W = 1,07*10-5 Nsm.
“Reflektions-Faktor” = reflection coefficient
Finally, let us look again at the calculation of the transmission coefficient ψ. Given known
excitation and known bearing impedance, ψ can be calculated; conversely, an unknown
bearing impedance (preferable ZML) may be calculated if ψ is known. For the bearing that is
immobile in the transverse direction, and given a fully flexible string, the transmission is zero
– the vibration energy is entirely reflected. Strings that are not ideally flexible can, however,
transmit part of their vibration energy across such a bearing (Fig 2.22).
The calculation of the transmitted part assumes that a flexural wave with the transversal
velocity v propagates in the main section of the string. The transverse velocity v(z = 0) at the
bearing is supposed to be zero by definition (ideal knife-edge bearing). We see the angular
speed w(z = 0) as the coupling quantity; it is identical on both sides of the bearing. The
following equation describes v as it occurs in the main section of the string:
From this, the place-derivative ( ) yields the angular speed:
Angular speed at the bearing
For the knife-edge bearing (ZFL = 0), the reflection coefficient ζ contained herein results from:
The ideal knife-edge bearing does not have any bending impedance ZML. However, the
flexural wave arriving at the bearing still does meet a bending impedance: the one of the
string extending beyond the bearing (z < 0). This impedance is:
Input impedance of the remaining section of the string
The bearing impedance ZML is negative, because the excitation wave runs towards the left to
the bearing (z > 0). Using ZML, the (complex) reflection coefficient is simplified:
Reflection coefficient of the knife-edge bearing (z>0)
With this result, the fringe-field coefficient γ is also defined for the knife-edge bearing:
Fringe-field coefficient of the knife-edge bearing (z>0)
Using the above, we can now calculate the angular speed present at the bearing: w(0) = j k v.
However, the progressive wave does not simply travel across the bearing being unimpressed:
directly at the far side of the bearing we have w(-0) consisting of the ψ-part of the progressive
wave, and the δ-part of the fringe field. The fringe field has decayed at a small distance (z < 0),
though, and only the ψ-part of the (transmitted) wave running away from the bearing remains.
Given w(0), the remaining section of the string is now excited in the range z < 0 (transmitted
part); the wave running away and the fringe field superimpose here:
At z = -0, the transversal velocity needs to be zero, too (knife-edge bearing), thus δ = -ψ
holds.

The absolute scaling is calculated using the angular speed w(0) of the bearing:
ψ represents the complex transmission coefficient – it states which part of the excitation
wave runs across the bearing. Given a cantilever without any tensile strain (Ψ = 0), k' / k = 1
holds. The transmitted amplitude portion amounts to 70%, and the transmitted energy portion
is 50%. The other 50% of the energy are reflected. In a guitar string, the tension force Ψ
dominates, and thus the reflected portion is larger (Fig. 2.22).
We may not entirely ignore the coupling across the bridge, though, as shown by the
following experiment. For a semi-solid guitar (Gibson ES-335 TD with trapeze tailpiece), the
strings of which continue for 10 cm beyond the bridge to the tailpiece, the E2-string was set in
motion by tapping it between the bridge and the tailpiece, and then immediately damped
again. By this, the section of the string between bridge and nut was set in motion, as well, and
sustained audibly. However, tapping the string directly above the bridge there is practically no
excitation – the transverse impedance of the bridge is indeed very high.
The coupling across the bridge is also pointed to by this experiment: the decay time (sustain)
of the ES-335 TD was established for the E2-string, using third-octave bandwidth.
Subsequently, the palm of the hand was placed on the section of the string between bridge and
tailpiece, damping it. Again, sustain was measured (for the section of the string between
bridge and nut), and it indeed was shorter across the whole frequency range.
Neither experiment provides absolute proof: the sting is supported at the nut, as well, and
excitation or damping could have been present here, too. Therefore we carried out a
supplementary experiment on the vibration test rig: a solid steel wire of 0.7 mm diameter
and 13,3 m length was stretched between two bearings each with the shape of mono-pitched
roof. A laser beam samples the transversal velocity of this “string” at 4 mm ahead of one of
the bearings. Beyond the bearing the string continues for 65 mm to where it is fastened (i.e.
this is the remaining section of the string). Between the bearing and the measuring point of
the laser, the string is hit with a small drop hammer providing an impulse-shaped excitation.
The transverse displacement over time is shown in Fig. 2.25: once with un-damped remaining
section of the string, and next to that with damped remaining section of the string. The
bending-coupling is not pronounced very much but it is still clearly visible.
Fig. 2.25: Transverse displacement (“Transversal-Auslenkung”) without (left) and with (right) damping of the
remaining section of the string (Chapter 1.4).
2.8 The generalized transmission-line model
The guitar is part of a signal-processing system generating sound from the movement of a
plectrum (pick). With the input quantities of plectrum-force and plectrum-velocity, and the
output quantities of bearing force and bearing velocity (in an acoustic guitar), or pickup
voltage and pickup current (in the electric guitar), respectively, the string is a subsystem of
the guitar. In Chapter 1.5 we had defined the plucking process as imprinting a force step with
the effect that a special square wave runs back and forth on the string. This (more of less)
periodic repetition of the excitation signal may be very nicely described with signal-flow
diagrams, as they are also used in the context of digital FIR-/IIR-filters. It is not a problem
that the signals in digital filters are usually time-discrete and discrete-valued, while the signals
on the string are time- and value-continuous. In the simple transmission-line model, only the
delay times occurring between the string bearings are emulated via delay lines. Conversely,
plucking point and pickup position may be arbitrarily chosen.
2.8.1 Ideal string, bridge pickup
The following signal flow diagrams SFD (block diagrams) represent the signal processing via
arithmetical operations. The basic operations are delay, summation, subtraction, and
multiplication with a constant. The graphs do not give any indication of the source- and load-
impedances and must not be confused with a circuit diagram.
A transverse force jumping to zero at the time t = 0 is defined as the excitation signal for the
string. This force step runs in both directions from the plucking point; its phase velocity is c.
The delay time necessary to reach bridge or nut, respectively, depends on c and the distance
that needs to be covered. At the end of the string, each force step is reflected – here, we need
to distinguish between rbridge = R and rnut = r. Thereafter, both force steps circle in a recursive
loop with an overall delay time of T = 2L/c. Fig. 2.26 shows the corresponding SFD:
Fig. 2.26: Signal flow diagram (SFD) for non-dispersive string vibration. T1 and T2 are delay times from the
plucking point to the bridge (“Steg”) and the nut (“Sattel”), respectively; R is the reflection coefficient at the
bridge, r is the reflection coefficient at the nut, T/2 is the delay time between bridge and nut, or nut and bridge,
respectively. E = input, A = output (bridge).
The SFD shown in Fig. 2.26 differs from the ideal string in one significant aspect: the impulse
created by the plucking runs back and forth on one and the same string, while in the SFD, the
paths in the two directions manifest themselves in two separate, serially connected signal
branches. Still, the signal processing is identical, and in both cases one cycle includes two
reflections.

2.8 The generalized transmission-line model 2-37
By repositioning of single delays, the SFD can be reshaped to result in a ladder network of
three systems (Fig. 2.27):
• A basic delay T1, modeling the delay time from point of plucking to the bridge.
• A recursive system with the delay time T, modeling the string vibration maintained via the
reflections (IIR- and AR-filter respectively)
• Am interference filter with a delay difference of 2T2, modeling the shaping of the sound
color via the point of plucking (FIR- and MA-Filter, respectively). For any one reflection
at the nut /(or fret), r ≈ –1 holds.
This representation has the main advantage that the “plucking”-filter (FIR-filter) and the
section of the generator (IIR-filter) are considered independently from each other in separate
stages. Assuming un-damped, loss-free vibrations (R⋅r = 1), the IIR-filter (operating just shy
of self-oscillation) generates – after impulse excitation – a periodic signal. Obligatorily, there
is a matching harmonic line spectrum with the frequency distance of the lines equal to the
fundamental frequency of the string.
Fig. 2.27: Rearranged signal flow diagram (only a single signal path string → bridge). The sequence of the
FIR-filter (2T2) and the IIR-filter (T) is permutable (commutative mapping in the linear system).
“Sattel” = nut, “Steg” = bridge.
Rearranging the FIR-delay time is done with , resulting in:
Using simple methods known from signal processing [e.g. 5], we can now derive from the
SFD shown in Fig. 2.27 the behavior regarding frequency. If we take, as excitation signal, a
short impulse (idealized a Dirac) periodically repeated in the IIR-filter, a spectrum with
equidistant lines of constant height results. This spectrum is filtered as it runs through the
subsequent systems, i.e. it is modified.
A pure signal delay by a constant delay time (e.g. T1) only changes the phase spectrum but not
the magnitude spectrum. We will ignore this basic delay since it is immaterial for the
following considerations whether or not the output signal arrives a few milliseconds later.
However, the delay time in the FIR-filter must not be ignored since here two signals are
superimposed that are delayed with respect to each other – with the resulting frequency-
selective amplifications and cancellations (comb-filter). The sequence of FIR/IIR, or IIR/FIR,
respectively, must not be interchanged.
The filter effect of the comb-filter is extensively described in literature; we will only cover it
in short here. The temporal input signal of a delay line arrives at the output after a delay
(generally: Tx), the spectrum of the input signal is to be multiplied with the transfer function
to yield the spectrum of the output signal. The transfer function H of a (pure) delay line with
the delay Tx is:
Transfer function of a delay line
In a comb-filter, delayed signal and un-delayed signal are added or subtracted, respectively;
this yields the transfer function of the comb-filter:
FIR-filter
The designation FIR-filter (Finite Impulse Response) is due the impulse response being of
finite duration. The magnitude of the frequency response is the magnitude of a sine-function
with zeroes at 0 Hz and integer multiples of the reciprocal of the delay time Tx. This
calculation is formally correct but inconvenient for illustrations, as Fig. 2.28 shows. Similar
problems are known from time-discrete signals if the sampling theorem is not adhered to: too
low a sampling rate results in (usually undesirable) reverse convolution. In the present
special case, however, the ambiguity due to the sampling is helping. Via the identity
and a few intermediate steps, the FIR-transfer function may be converted into:
FIR-filter, reformulated
Herein, d represents the distance between the plucking point and the bridge, and M is the
length of the open string (scale). For the fretted string, the scale needs to be applied here, as
well, because it is included in the formula for the propagation speed of the wave. If the open
string is plucked precisely in the middle, the long-term spectrum holds only odd harmonics –
the zeroes of the sine-function are located at the even harmonics. The closer the plucking
point is to the bridge, the wider the minima of the envelope are spaced. The conversion only
holds in the steady-state part (discrete line-spectrum) but not for the transitory process. This is
a basic condition for every transfer function, though: it always holds for the steady state only.
Furthermore, we need to consider that the delays in the above model are frequency-
independent – dispersion is not (yet) emulated. Spread-out spectra require, instead of simple
delay lines, all-passes that approximate the string dispersion in the frequency response of
their delay (Chapter 2.8.4).

Fig. 2.28: FIR-filter frequency response (magnitude, ---) and filtered line spectrum for d = M/5. The lines shown
are identical in both graphs; the graph on the right shows the transformed FIR-transfer-function.
In Fig. 2.29 we see the measurements for a plucked E2-string. The distance between plucking
location and bridge amounted to d = 4,7 cm and 1,5 cm, respectively. From the results, the
first minimum of the comb-filter calculates as 1,1 kHz and 3,5 kHz, respectively. In the low
frequency region, the comb-filter structure is clearly visible in the spectral envelope – it is
however perturbed by strain-wave resonances (Chapter 1.4, marked via dots). In anticipation
of Chapter 2.8.4, Fig. 2.29 already includes the dispersive spreading of the spectral envelope.
In addition, further selective damping mechanisms have an effect, especially in the high
frequency domain. The associated causes will be elaborated on in Chapter 7.
Fig. 2.29: Measured spectra; E2-string (impulse excitation), d = 4,7 cm (top) and d = 1,5 cm (bottom).
The shown envelope was spread out (dispersion) and slightly attenuated towards the high frequencies.
The two measurements were taken with two different E2-strings (OVATION Viper EA-68).
While the FIR-filter determines the spectral envelope, the recursive filter defines the
frequency of the individual spectral lines. The impulse response of a recursive filter is of
infinite length, which is why the term IIR-filter (Infinite Impulse Response) is common for
this filter type. With both reflection coefficients being equal to 1, a short excitation impulse
would circulate in the loop indefinitely without attenuation; such a filter is called borderline
stable. Real strings have reflection coefficients of <1; the impulse-shaped excitation therefore
decays over time. For a run through the full loop, both reflection coefficients act in
multiplicative manner (R⋅r).
Given R⋅r = 0,9, for example, the height of the impulse decreases e.g. from 1 to 0,9 for a
single loop, to 0,81 for a double look, and to 0,9n for an n-fold loop. The amplitudes of the
impulses following each other with a distance in time of T represent a geometric progression;
for R⋅r < 1, the term used is exponential decay. Chapter 1.6 had already included quantitative
statements regarding the decay process; for the guitar string, the loop coefficients are very
close to 1 (e.g. 0,993). In the FIR-filter only a single reflection occurs, and therefore r = −1
may be used with very good approximation. However, in the IIR-filter, the loop is run through
an infinite number of times, and consequently this approximation is not allowable.
Chapter 2.5 had shown that the reflection coefficient is not constant but frequency-dependent.
The reason are resonances in the bridge and the nut (or fret) formed from a combination of
springs and masses. These springs and masses are not necessarily all found within the bridge
(or the nut or fret) but may be located e.g. in the neck of the guitar and act on the nut [8].
When integrating a frequency-dependent reflection coefficient into the SFD (Fig. 2.17), we
need to pay attention to the fact that the system shown as circle (R⋅r) becomes a filter that
way: is the frequency dependent transfer function of this reflection filter. The
decay time-constant for each partial results from the loop-delay-time T (frequency dependent
if the dispersion is considered), and from . The SFD (Fig. 2.27) does not consider the
reason for the damping: it is the overall damping that is modeled via . If required,
several individual filters may be connected in series, for example to be able to model the
internal string damping in a separate subsystem.
Ahead of the input designated with E in Fig. 2.27 we need to position the plectrum filter that
shapes the real excitation force from the ideal step (or from an impulse). The piezo-filter, or –
for acoustic guitars – the body- and radiation-filter follows the output A. The structure-borne
sound path is not modeled herein. If we think of the nut merely as a vibration absorber, this is
not necessary, either: the damping caused by the nut is considered in R⋅r, after all. However,
part of the vibration energy flowing into the nut might be radiated, or fed back to the string
via the bridge – something that necessarily would have to work in reverse, as well. The
dilatational waves discussed in Chapter 1.4 use a similar bypass (albeit directly via the string).
Additional recursive loops enable a simple emulation of such parallel paths. It should be
emphasized again, though, that this does not automatically make for a correct representation
of the energy flows. In the SFD, a summation point adds two signals (e.g. two forces), but it
does not model the impedances – these would a have to be considered separately depending
on the circumstances.

2.8.2 String with single-coil pickup
The SFD presented in Fig. 2.26 is now extended by the output of a magnetic pickup,
assuming that the pickup will not influence the vibration of the string. This assumption is not
fundamentally justified, because the attraction force of the permanent magnet does change the
string vibration, and moreover the law of energy conservation demands that the string delivers
the electrical energy generated. While the latter effect may be neglected when high-
impedance pickups are deployed, strong magnets are indeed known for their interference
when adjusted too close to the strings (Chapter 4.11). However, on order to explain the
transfer characteristic in principle, the attraction does not need to be modeled.
Fig. 2.30 depicts the simplified model for the ideal string and a single–coil pickup. T1 and T2
designate the delay from the plucking point to the bridge and the nut, respectively. ,
respectively, is the delay from the location of the pickup to the bridge and the nut. Multiple
rearranging of the drawing yields a ladder network consisting of four different filters:
• A basic delay from plucking point to pickup

• An FIR-filter with the long delay 2T2 (or 2τ2, respectively)
• A recursive IIR-filter to model the string vibration
• An FIR-filter with the short delay 2τ1 (or 2T1, respectively)
The sequence of these four subsystems may be changed arbitrarily. The pitch depends on the
IIR-filter, and the sound color depends on the FIR-filters with their interference effect
retraceable to the delay times T1 and τ1. There are three cases for the position of the pickup
and the plucking point: T1 < τ1, T1 > τ1, and T1 = τ1. It is immaterial whether pickup or
plucking point is located closer to the bridge. For example, the pickup may be mounted 10 cm
off the bridge, and the string is plucked 4 cm from the bridge, or the pickup may be mounted
4 cm from the bridge and the plucking may happen 10 cm from the bridge – in a linear model,
the result will be the same (Fig. 2.35). What is not modeled: the string hitting and bouncing
off the frets.
Fig. 2.30a: Ideal string with single-coil

magnetic pickup, T1 ≥ τ1.
Reflections at bridge and nut are taken to
be loss-free (R = r = -1).
“Grundlaufzeit” = basic delay time
Fig. 2.30b: Ideal string with single-coil

magnetic pickup, T1 ≤ τ1.
Reflections at bridge and nut are taken to
be loss-free (R = r = -1).
“Grundlaufzeit” = basic delay time
The step response associated with the step excitation is indicated in Fig. 2.31. Like Fig. 2.30,
Fig. 2.31 shows that when changing from T1 < τ1 to T1 > τ1, merely the delay times T1 and τ1
need to be interchanged. The periodicity of this dispersion-free filter is T = 2(T1 + T2) = 2(τ1 +
τ2). Two square impulses are located within that period, centered around the point in time t0,
and T – t0, respectively. For T1 < τ1 we get t0 = τ1, while t0 = T1 results for T1 > τ1. The
impulse width amounts to Δt = | T1 – τ1 |.
The impulse width corresponds to the delay time of the transversal wave running from
plucking point to pickup. If this distance is e.g. 4 cm, the impulse width calculates as 4⋅T/2⋅64
= T/32. Herein, the scale is assumed to be 64 cm. If the string is plucked exactly over the
pickup, the two square impulses are perfectly contiguous.
Fig. 2.31: Step response of the filter from Fig. 2.30. Left: T1 ≠ τ1; right T1 = τ1. Input quantity for the filter is a
force step at the plucking point. Output quantity is the string velocity over the magnet of the pickup – the source
voltage of the pickup is proportional to this velocity. The terminal voltage results from low-pass filtering of the
source voltage (Chapter 5.9). In particular for the low strings, the frequency-dependent propagation velocity
(dispersion, Chapter 2.8.4) takes care of reshaping the rectangular waveform. In order to model this effect, the
delays in Fig. 2.30 need to be realized as all-passes (Fig. 2.39).

The calculation of the overall transfer function of the 4 serially connected individual filters
requires a multiplication of the individual transfer functions, resulting in somewhat more
complicated frequency responses (Fig. 2.32).
Fig. 2.32a: Transfer frequency response, E2-string plucked 12 cm away from the bridge. Scale = 64 cm.
Left: bridge-pickup (4 cm distance from the bridge); right: neck-pickup (16 cm distance from the bridge).
Fig. 2.32b: Transfer frequency response, string plucked 12 cm away from the bridge; bridge pickup (5cm
distance from the bridge, scale = 64 cm). Left: E2-string, right: A-string.
It should be noted as particularly important that the two FIR-filters act string-specifically and
do not have a global filter effect (as the magnetic pickup discussed in Chapter 5 would show
it). The winding of the pickup coil is permeated by field-alterations of all 6 strings, and thus
the resonance peak of the pickup will affect all 6 strings in the same way. The cancellations of
the FIR-filter, however, are based on the propagation speeds of the waves, and these are string
specific. As already elaborated, these propagation speeds do not depend on the (fretted) pitch,
but on the pitch of the open string. The latter determines the propagation speed cP, after all. It
is therefore not possible to generate the FIR-characteristic electronically with an effects
device … not with your regular pickups, anyway.
Fig. 2.33 shops the FIR frequency responses of a Stratocaster dependent on the pickup
position. The effect of the second FIR-filter (plucking location) was not included in the
calculations. To ensure a clear representation, the minima are only shown to a depth of 18 dB;
according to the theory, the graph should extend to – ∞ dB in the minima.
Fig. 2.33: Calculated FIR frequency responses for the Stratocaster; without dispersion. The dynamic is limited to
18 dB. In the lowermost graph, the effect of dispersive propagation is shown as a dotted line (compare to
Chapter 1.8.4).

In Fig. 2.34, we see a comparison between measurement and calculation. A Stratocaster is

connected to an instrumentation amplifier (input impedance: 100 kΩ) via a cable of a
capacitance of 200 pF. The E2-string is plucked directly at the bridge with a plectrum, and the
signal of the bridge-pickup was evaluated.
The comparative calculation of the line spectrum includes both FIR-filters, the IIR-filter, and
the equivalent circuit diagram of the pickup (Chapter 5.9.3). In addition, a small treble-
attenuation was included to emulate the window of the magnetic field (Chapter 5.4.4). There
is a clear correspondence. The measured spectrum nicely depicts the spreading of the
frequency of the partials; it was emulated for the calculation using a simple model. The
simulation easily reproduces the comb-filter structure, as well – at high frequencies, however,
differences between measurement and calculation become evident. For a further
improvement, e.g. the reflection coefficients would have to be adapted.
Fig. 2.34: Spectrum of an E2-string plucked directly at the bridge (Stratocaster, middle pickup).
Top: measurement (with DFT-leakage). Bottom: calculation (with dispersion, compare to Chapter 2.8.4).
The inharmonic spreading is considerable; the 70th “harmonic” is at 7,37 kHz rather than at 5,84 kHz.
It was already noted with respect to Fig. 2.30 that the plucking point on the string and the
location of the pickup result in an FIR-filter each (with different delay times). Two serially
connected filters represent two commutatively connected mappings the sequence of which
may be interchanged. It therefore should not make any difference whether the string is
plucked at point A with the pickup being located at point B, or the string is plucked at point B
with the pickup located at point A.
To check this hypothesis, the E2-string of a Stratocaster was plucked over the neck-pickup
while the signal from the bridge-pickup was recorded. Subsequently, the E2-string was
plucked over the bridge-pickup and the signal of the neck-pickup was recorded. Fig. 2.35
shows the DFT-spectra of both signals. The agreement is uncanny – especially considering
that the reproducibility of the plucking process is not particularly good.
Measured signal of the bridge-pickup; string plucked over the neck-pickup.
Measured signal of the neck-pickup; string plucked over the bridge-pickup.
Fig. 2.35: Spectrum of the E2-string of a Stratocaster; pickup and plucking position interchanged.

2.8.3 String with humbucking pickup
In the hum-cancelling humbucking pickup, two coils are connected in opposite phase. In
order for the electrical output signals to interact constructively, the magnetic permanent flux
is reversed in one of the coils. Many pickups (e.g. Gibson) generate the permanent field using
a bar magnet located under the coils; the field is conducted through the coils using so-called
pole-pieces. Other designs (e.g. Fender) use 6 individual magnets in each coil; in one of the
coils, the north-pole is directed upwards, in the other it is the south-pole. The two coils are
usually connected serially in opposite phase; opposite-phase parallel connection is less
common.
The humbucker samples a wave running along the string at two adjacent areas. The distance
between the two pole-pieces is 18 mm for the Gibson Humbucker – there are, however, also
very narrow humbuckers that fit into the housing of a regular single-coil pickup.
Fig. 2.36: Signal flow diagram for a humbucking pickup with two equivalent coils.
In Fig. 2.36, τ1 represents the (single) delay time between the coil located closer to the bridge
and the bridge, while Δτ is the delay between the two coils. Using suitable conversion, we
arrive at a simple ladder-network of two FIR-filters. The first filter models – with same-phase
superposition – the delay time Δτ between the coils; the other filter emulates – using opposite-
phase superposition – double the delay time between the middle of the humbucker and the
bridge. The humbucker positioned at a location x differs from a single-coil pickup located at
the same position only in the Δτ-filter. The modeling as ladder network offers the
considerable advantage that the overall transfer function can be represented as the product of
the individual transfer functions. Given a humbucker with a distance between pole-pieces of
18 mm, we get an additional signal cancellation for the E2-string in the range around 3 kHz;
for the higher strings, the humbucker-minimum is located at correspondingly higher
frequencies. The exact frequency of the minimum depends not only on the pole-piece
distance, but also on the dispersion (Chapter 1.3)
As is shown in Fig. 2.37, the differences between single-coil pickup and humbucker are
string-specific: for the E4-string, only small variations in the treble range will be recognizable,
while for the E2-string, the humbucker will absorb the 3-kHz-range that is important to obtain
a brilliant sound. Reducing the distance of the two humbucker coils to 13 mm (as it was done
e.g. in the Mini-Humbucker fitted to the Les Paul Deluxe) will shift all interference-minima
toward higher frequencies. A particularly small distance of the coils (7 – 9 mm) is realized in
the single-coil format; still, a treble loss remains for the low strings.
Fig. 2.37: Comb filtering caused by the humbucker

delay time Δτ. Distance of the poles typical for the
Gibson Humbucker: 18mm; for the Mini-
Humbucker: 13mm. 7,6mm can be found in
humbuckers in single-coil format (e.g. DP-184).
The minima of the comb-filter graphs were cut to
obtain a clearer representation. Dispersion was
considered (Chapter 1.3).
If the two humbucker coils do not feature the same sensitivity in both coils, we get differences
in particular in the range of the humbucker-minimum (Fig. 2.38). Such imbalances have their
roots in different numbers of the turns of the coils (deliberately produced for the Burstbucker)
and/or in the field guides: the pole pieces in the shape of slugs have a different magnetic
resistance compared to the threaded pole-screws. For differing coils, the SFD may not be
separated into two FIR-filters, and thus Fig. 2.38 shows the frequency responses of the overall
signal flow diagram.
Fig. 2.38: Magnitude frequency responses for unmatched humbucker coils. Left: bridge humbucker (distance to
bridge 45 mm); right: neck humbucker (distance to bridge 147 mm). The sensitivity of the coil with threaded
pole pieces (screws) is better by 1 dB compared to the “slug”-coil (–––), or smaller by 1 dB (---). Dispersion was
considered (Chapter 1.3).

For a Gibson ES-335 TD (E2-string), Fig. 2.39 considers the transfer function of the
equivalent circuit established in Fig. 2.36. In Fig. 2.40, the RLC-transfer-function (Chapter 5-
9) is added in. Via Fig. 2.41, we can compare a measurement. For all graphs, dispersive wave
propagation was included.
Fig. 2.39: Gibson ES335, E2-string, model without RLC-filter. Left: bridge pickup. Right: neck pickup.
Fig. 2.40: ES335, E2-string, model with RLC-filter and 707-pF-cable. Left: bridge pickup. Right: neck pickup.
Fig. 2.41: ES335, E2-string plucked directly at the bridge; bridge pickup. Left: model calculation; right:
measurement. The differences do not refute the basic model assumptions; rather, they indicate how important the
modeling of both strain-wave and bearing impedances is – this was not included here.
2.8.4 Dispersive line elements
In Chapter 1.2, we had discussed that the propagation speed of the transversal waves is
frequency dependent (dispersion), leading to a “spreading out” of the frequencies of the
partials. This effect may be modeled in the SFD using frequency-dependent delay times. If we
first assume the string to be loss-free, the magnitude spectrum will not change during the
wave propagation. The phase spectrum does change – but not with a linear-phase
characteristic like it would in a delay line. Rather, it assumes the characteristic of an all-pass
function due to the frequency-dependent delay time. From the spreading of the partials, we
can deduce the all-pass transfer-function (Chapter 1.3.1), and from this the all-pass impulse
response (Chapter 1.3.2) via inverse Fourier transform. The simulations shown in Chapter 1
were calculated using such an SFD.
All-pass: linear system with a frequency-independent magnitude transmission coefficient and

frequency-dependent phase shift.
Minimal-phase system: linear all-pass-free system.
Linear-phase system: linear system with frequency-proportional phase shift.
System order: number of the independent storage elements in the system.
For a wound E2-string (b = 1/8000), Fig. 2.42 shows the phase shift ϕ as it appears in a
transversal wave running the distance of 8,65 mm (Chapter 1.3.1). Cascading 74 of the digital
filters indicated in the figure yields a good approximation of the overall phase-shift of an E2-
string of 64 cm length (single travel path). The relatively high number of filters is due to the
chosen sample frequency: a 2nd-order all-pass can turn shift the phase by no more than 2π.
Fig. 2.42: Block-diagram and frequency response of the phase of a 2nd-oder canonic digital all-pass filter.
Sample frequency: fa = 48 kHz, a = 0,5378, b = -0,03668. The frequency response of the filter phase is indicated
as a dashed line; the differences to the phase of the string (––––) are insignificant. “Frequenz” = frequency.
Given fa = sample frequency, the transfer function of the digital all-pass is:
; ;
If the sample frequency is changed, the parameters a and b need to be adapted, as well.

The phase delay of the all-pass filter shown in Fig. 2.42 features the same tendency as it is
found in dispersive waves on strings: high frequencies get to the output of the filter faster than
low ones. Given a step excitation, we will therefore see a reaction in the high frequency range
first; the low frequency components follow with a delay (Fig. 2.43).
Fig. 2.43: Step response of a cascade of 14 (left) and 74 (right) all-pass filters. Data as in Fig. 2.42. In addition to
the all-passes, a slight treble attenuation was included (one 1st-order low-pass at 10 kHz).
On the one hand, dispersion has the effect of a progressive spreading of the frequency of the
partials. For the perceived sound, it is more important, though, that the FIR-filters depicted in
Figs. 2.30 and 2.36 are subject to the same mechanism, as well: their interference effect
happens progressively spread out towards higher frequencies. Given a dispersion-free E2-
string, the bridge pickup of a Stratocaster would feature an interference cancellation at 3 ⋅ fG ⋅
65cm / 5cm = 3214 Hz. However, your commercially available string is not free of dispersion,
and therefore the interference cancellation mentioned above will happen somewhere in the
range of 3330 – 3520 Hz, depending on the specific design of the string. In case the
loudspeaker contributes narrow-band resonances in that same frequency range, a change of
the make of strings may indeed bring audible differences. In this context, it should not be left
unmentioned, though, that moving the guitar loudspeaker may well lead to changes in the
sound: the room represents an FIR-filter, as well – due to the various occurring sound paths.
2.9 Magnetic pickup with excitation by dilatational waves
Does an axial shift in the string induce an electrical voltage in the magnetic pickup? The
distance between string and pole-piece of the pickup does remain constant, after all – which is
why we would not expect any voltage. Measurements do not support this hypothesis, though.
Apparently, the distance between string and pole-piece is not the only criterion for the
generation of a voltage: due to hysteresis and associated memory processes, a dilatational
wave running along the string may indeed induce a voltage in the pickup, as well. The
following model considerations discuss the basic context:
Fig. 2.44: Dilatational wave (left), string with elementary magnets and pickup coil (right). Both figures show
considerably simplified, discretisizing models.
In the left-hand section of Fig. 2.44, we see a model of a string depicted at 11 different times;
the bold points are masses, and in between them there are springs♣. On the top left, a
compression impulse is generated that propagates along the string with progressing time
(dilatational waves are generally free of dispersion). A pickup mounted beneath the string
generates a permanent magnetization within the string – this is shown in the right-hand graph
with a few elementary magnets. The dilatational-wave impulse sequentially shifts each of the
elementary magnets: first a little to the right, then back to the original position. This shift
varies the magnetic flux axially penetrating the coil. Looking at the right-hand graph seen in
Fig. 2.44: for the elementary magnets shown towards the left, the variation of the location
(resulting from the impulse) causes an increase in the magnetic flux penetrating the coil, and a
decrease for the elementary magnets shown on the right.
The efficiency of the voltage induction caused by this effect depends on many factors: the
magnet, the turns-number of the coil, or the material of the string. Of particular importance
for the above model consideration are two parameters: the distance between the elementary
magnets and the coil, and the angle between the axis of the elementary magnets and that of
the coil. The compression impulse running along the string from left to right generates – in
the coil – first an increase in the flux, and then a decrease. These variations in the flux induce
an electrical voltage in the coil (law of induction: the voltage induced per turn of coil-winding
corresponds to the temporal derivative of the magnetic flux penetrating this turn).
♣
The shown change in diameter is strongly exaggerated.

2.9 Magnetic pickups with excitation via dilatational waves 2-53
For a half-wave-shaped displacement impulse, Fig. 2.45 presents the time functions of the
flux change ΔΦ, and the corresponding temporal derivative. The graphs – put together from
simple functions – are meant to merely familiarize us with the shape in principle; an exact
calculation would require considerable effort. Given a geometric distance of the ranges of
maximum sensitivity of about 1 – 2 cm (the typical dimensions of pickups), we obtain the
distance in time of the extrema of about Δt = 2 – 4 µs (with a propagation speed of dilatational
waves of about 5 km/s).
Fig. 2.45: Variation of flux (left) and its temporal derivative, caused by a compression impulse.
The signal shown in the right-hand part of Fig. 2.45 may be interpreted as impulse response
hUξ. The first index (U) points to the pickup voltage U being seen as the output value that
results from the differentiation of the magnetic flux. The second index ξ relates to the source
signal: a displacement impulse. From the impulse response h(t) of an LTI-system [6, 7], and
using the help of a Fourier-transformation, we arrive at the transfer function H(jω) of this
system♣. Herein, input and output signals remain the same; they are merely represented in
different “domains”: the impulse response connects (via the convolution) the input- and the
output-time-function, and the transfer function connects (as a multiplication) the input- and
the output-spectrum. The Fourier transform of the impulse response hUξ is therefore the
transfer function HUξ. Model-considerations for equivalent circuits have, however, shown that
HUv represents the more easily interpretable transfer function (Chapter 5.9.3), rather than HUξ.
Instead of the displacement impulse, a (particle-) velocity impulse is applied as trigger of the
dilatational wave (the corresponding displacement function is the step). Instead of exciting a
dilatational wave within the string with an excitation impulse, the temporal integral (the
displacement step = velocity impulse) of the dilatational wave is impressed optionally. This
additional integration is taken into consideration in Fig. 2.45 by requiring that the induced
voltage shown in the right part of the figure is subject to an integration (commutativity of
LTI-systems). Since the right-hand part of the figure was derived from the left-hand part via
differentiation, we can use the left-hand graph to establish the time-course of hUv – merely the
units are different. The following summary results:
A dilatational wave resulting from a displacement impulse induces the pickup voltage shown
in the right-hand graph of Fig. 2.45. A dilatational wave resulting from velocity impulse
induces the voltage shown in the left-hand graph of Fig. 2.45.
♣
The additional low-pass filtering occurring in pickup and cable is ignored here to begin with.
The Fourier-transform of hUv (i.e. the transfer function |HUv|) is depicted in Fig. 2.46. We can
see a frequency-proportional rise in the frequency range particularly relevant for the magnet
pickup (< 10 kHz) – corresponding to the magnitude frequency-response of a differentiator. It
becomes clear that the exact shape of the hUv-curve is of minor importance: any (odd
numbered) origin-symmetric impulse response will exhibit the characteristic of a
differentiator in the low-frequency range. Due to the high propagation speed of the
dilatational waves, the maximum of the transfer characteristic is located at such high
frequencies that its specific range does not need to be determined.
Fig. 2.46: Magnitude of the transfer function of (particle) velocity → voltage, without LC low-pass (left).
Voltage induced in the pickup: string-excitation by a drop hammer, w/out dispersion, w/out LC low-pass (right).
Excitation of the string via a drop hammer generates two subsequent impulses in the pickup
winding: first, we get the impulse induced by the dilatational wave, and then the impulse
induced by the (slower) flexural wave. If at first the dispersion (that occurs only for the
flexural wave) is disregarded, a voltage shaped similarly to the one shown in Fig. 2.46 would
be expected. A displacement impulse shaped similar to a sinusoidal half-wave (small figure)
runs along the string both as a (simplified) dispersion-free, slow transversal wave, and as fast
dilatational wave. The first temporal derivative of this impulse corresponds to the voltage
induced by the transversal wave; the second derivative corresponds to the voltage generated
by the dilatational wave. However, the dispersive propagation of the flexural wave leads to a
considerable reshaping of the impulse. Therefore, the shape of the voltage shown in Fig. 2.46
on the right will not occur during measurements in reality. Rather, all-pass-induced impulse
deformations appear (Chapter 1.3.2, Chapter 2.8.4). In order to be able to compare the above
theoretical model-calculations with measurements, the transversal-wave impulse (that looks
similar to a full sine oscillation) needs to be first sent through an all-pass filter.
The measurements used for comparisons in the following were done using a 30 m long string
of 0,7 mm diameter mounted below a Jazzmaster pickup. Due to its very low winding
capacitance, and given suitable electrical loading, this pickup allows for broadband
measurements up to about 20 kHz. While not exactly typical for use in electric guitars, the
corresponding circuitry is highly qualified for measurements. At 3 mm from its mounting
point (clamp), the string was excited by a short displacement impulse, leading to the
propagation of a dilatational and a flexural wave along the string. At a distance of 68 cm from
the clamp, the transversal velocity was sampled both with a laser vibrometer and with the
Jazzmaster pickup, and the resulting signal was digitally stored.

2.9 Magnetic pickups with excitation via dilatational waves 2-55
In Fig. 2.47, we see on the left the transversal velocity measured by the laser vibrometer. The
dilatational-wave impulse reaches the measuring point 0,13 ms after the drop hammer has
struck the string – this instant represents the origin of the time-scale. The laser vibrometer
practically ignores the dilatational-wave impulse; the pickup, however, shows an impulse that
resembles a twice-differentiated sinusoidal half-wave impulse (Fig. 2.46). After about 1 ms,
the high-frequency components of the flexural wave reach the measuring point, and the low-
frequency components follow after about 6 ms (dispersive propagation); these waves are
received by both sensors in a similar way.
Fig. 2.47: Time function measured after impulse excitation of the string; laser (left), pickup (right).
From the point of view of systems theory, the tensioned string represents – with good
approximation – an LTI-system that maps input quantities onto output quantities. A
separation according to the two wave types yields two sub-systems: a dispersion-free delay
line (dilatational wave), and a dispersive delay line (flexural wave). De-convoluting the
output quantity of the system measured at the pickup gives the input quantity of the system.
The effect of this de-convolution is shown in Fig. 2.48: of the pickup voltage indicated on the
right in Fig. 2.47, the time-snippet between 1 ms and 7 ms was de-convoluted with the
impulse response of the all-pass (Chapter 1.3.2). The result was drawn into the right-hand half
of the left-hand section of Fig. 2.48; for comparison, the original dilatational-wave impulse is
presented on the left. The part of the figure on the right shows the twice-integrated functions
corresponding to the displacement. While the curves juxtaposed in the figure are not identical,
they still are very similar – this could not be expected given the original functions that are,
after all, of an entirely different character.
Fig. 2.48: Comparison between measured dilatational-wave impulse and de-convoluted flexural-wave impulse.
“Beschleunigung” = acceleration; “Auslenkung” = displacement; “Dehnwelle” = dilatational wave;
“entfaltete Transversalwelle” = de-convoluted transversal wave.
The pronounced similarity of the shape of the curves presented in Fig. 2.48 leads to the
following conclusion: dilatational wave and flexural wave have approximately the same
time-function at the moment of their formation. This hypothesis may be further
corroborated via mapping the dilatational-wave impulse onto the flexural-wave impulse. For
this, the section from 0 ms to 1 ms of the impulse shown in Fig. 2.47 on the right is integrated
and convoluted with the impulse response of the all-pass: the signal depicted on Fig. 2.49 on
the right results. This latter signal corresponds with good approximation to the signal of the
flexural wave (on the right in Fig. 2.47; repeated in Fig. 2.49 on the left). An example for
which measurement and calculation correspond even better still is given in Fig. 2.50.
Fig. 2.49: Pickup voltages: flexural wave (left); impulse derived from the dilatational wave (right).
Fig. 2.50: As in Fig. 2.49, but established at a different pickup-position (55 cm instead of 68 cm).
“Messung” = measurement, “Modell” = model.
There is no absolute scaling of the ordinate in the above figures – for that a transfer
coefficient for the individual pickup would be required. To get an impression of the wave-
parameters, the following table lists typical (rounded-off!) values. The relationship of the two
wave energies depends on the respective string bearing.
Flexural wave Dilatational wave

Maximum displacement 30 µm 5,7 µm
Maximum (particle) velocity 0,4 m/s 0,07 m/s
Maximum force 0,2 N 1,2 N
Wave impedance 0,5 Ns/m 17 Ns/m
Maximum power 88 mW 88 mW
Impulse-energy 8,0 µWs 8,5 µWs

3. Magnetics of the string
In order to be able to change the magnetic resistance in the magnetic circuit, the vibrating
string needs to consist from ferromagnetic material. Ferromagnetics come in great variety – if
they are to be suitable as basic material for guitar strings, one feature is a predominant
requirement: they have to withstand the extremely high tensile stress. Just about every
guitarist will have broken a string during play at least once; that clearly shows how close to
the limit we are operating! Typical tensioning forces of strings fall into the range of between
50 N and 140 N. Given the rather small cross-sectional areas this implies tensile stresses of
up to 2000 N/mm2. Given to such high stress, only high-strength ferromagnetic special steel
qualifies as material for strings. As a protection against corrosion, the surface of the string is
usually coated with a thin layer of nickel or gold; this layer has no magnetic effect due to its
small thickness. Wound strings behave differently: their core diameter is about 30 – 60% of
the overall diameter, with the winding consequently giving a substantial contribution to the
cross-sectional area (the latter growing with the square of the diameter). Testimony to this
issue is the effect we get when trying to use – on an electric guitar – strings with steel core
and bronze winding as manufactured for acoustic guitars. Compared to the solid treble strings,
such wound strings are picked up with too little volume – because bronze is not magnetic.
The three bass strings of the electric guitar (E-A-D) are therefore wound with a magnetically
conductive material: usually with nickel, nickel-plated steel, or special non-corrosive steel. In
the following paragraphs, the magnetic properties of typical steel strings are discussed.
Subsequently, Chapter 4 will contribute a detailed description of magnetic fields.
3.1 Steel, nickel, bronze
High tensile strength requires a smooth surface because cracks and pores would increase the
risk of breakage. As a protection against corrosion, the string surface may be coated (TINNED
MANDOLIN WIRE); there are also uncoated strings, though. “Tinned” does not compulsorily
imply that the surface is coated with tin: in fact the coating of typical guitar strings is formed
of nickel (NICKEL PLATED STEEL). The two highest treble strings (E4, H3) are always solid
(PLAIN), and the three bass strings (E2, A2, D3) always sport a winding (WOUND); the G-string
is solid (plain) in light string sets, and wound in heavy ones. The winding does not absorb
tensile forces but merely serves to increase the mass. Other than steel, less stress-resilient
materials may be used for the winding, as well.
Without doubt, the material of the strings does influence the sound of the guitar. The reason
for this is, however, not that self-evident. Obviously, we will think of the inner damping of
the material. When bending steel, nickel, copper, or other metals, different amounts of energy
are converted into heat (dissipated). The decay of vibration therefore is material-dependent.
The differences between the customary metals are, however, not pronounced to the extent that
an audible difference in sound will result in tones of short duration.
3-2 3. Magnetics of the string
The main effect results from the string bouncing off the frets. Even with regular strength of
plucking/picking, the string will hit and bounce off the frets many times (Chapter 7). In this
process, the winding (or coating) acts as elastic and therefore sound-determining buffer
between fret and core of the string. An exact description of the string-bounce process is only
possibly with a very high effort: each individual string/fret contact is a non-linear occurrence
that will rule out the otherwise so helpful principle of superposition. The great number of
these non-linear contacts can only be described in a non-linear, stochastic model – which
would include a frightful variety of parameters.
Every string/fret-contact implies a mechanical impact. Mechanics know two kinds of impacts:
the elastic one, and the inelastic one. For the elastic impact, there is no generation of thermal
energy during the contact phase – it is termed the loss-free condition. However, this does not
mean that the string is not loosing any energy, but only indicates that the sum of the energy in
both partners involved in the impact is constant! The vibration energy transferred to the fret is
lost to the string at first: the string experiences a damping from the elastic impact. Also, we
may not expect that the vibration energy stored in the fret is re-transferred to the string later –
in fact, a substantial portion of the energy is lost in the fretboard and the neck of the guitar.
Given an inelastic impact, energy is dissipated irreversibly already in the deformation of the
material during the impact phase, i.e. it is irretrievably converted into caloric energy.
Each contact between string and fret is also a source of two fresh secondary waves running in
opposite directions. The energy contained in these secondary waves is not introduced to the
system from the outside but withdrawn from the original wave-energy. After each contact, the
system is again a linear one, and all waves may be superimposed. The contact phase itself,
however, is a non-linear, drive-level-dependent process that cannot be described via
superposition. The multitude of contacts renders the system non-linear during the first 10ths of
a second; only the subsequent decay process is linear.
A string/fret contact (other than where the string is actually fretted) may only be avoided with
very slight plucking of a (normally adjusted) string; in this case every analysis shows that the
levels of the higher-frequency partials decay substantially faster than the low-frequency ones.
The short impact of the string on the fret during the string-bounce represents a broad-band
excitation that “refreshes” the treble, in a manner of speaking. Instead of being plucked one
single time, countless “pickings” rain down on the string and make for a treble-rich, brilliant
sound.
Auditory experiments with a E2-string confirm this hypothesis: between a string wound with
nickel-plated steel (Fender 250) and pure-nickel-wound string (Fender 150), there is a just-
about significant, noticeable difference. However, raising the height of the bridge to the extent
that any post-plucking string/fret-contact is avoided makes the two string-types sound the
same. It needs to be emphasized here that the string/fret contacts are not generally perceived
as string-buzz or clatter. Rather, these contacts merge, as auditory events, to a single
homogenous plucking sound (ATTACK), as long as the contact noises do not dominate too
strongly, or are audibly modulated by low-frequency components. Each string/fret contact
transforms part of the low-frequency vibration energy into high-frequency vibration energy;
therefore the attack of “bouncing” string sounds more trebly. Nickel – as a material that is the
softer compared to steel – at the same time absorbs more of this add-on treble, and therefore
nickel-wound strings have a sound not quite as brilliant as steel-wound strings.

3.1 Steel, nickel, bronze 3-3
On guitars having a piezo pickup mounted rather than a magnetic pickup, the magnetic
conductance of the string winding does not play any role. Strings for these guitars therefore
typically sport a winding made from brass or bronze. What again holds: harder, low-loss
winding materials result in a more brilliant sound; softer winding materials also sound
“softer”, i.e. not as brilliant.
The “Zebra”-strings made by DR with their double-start winding represent a peculiarity: they
are manufactured with two different winding threads positioned next to each other. The
bronze-wire is supposed to generate the sound typical for flattop steel-string guitars, the steel-
wire is supposed to score with the magnetic pickup (see Chapter 3.2).
"Every other coil is nickel-plated steel, every other coil rare phosphor bronze, wound on hex
cores", it says in the Internet ad. Only on the packaging we then read: "...by winding
phosphor-bronze plated steel wire side-by-side with 8% nickel plated steel wire. Phosphor-
bronze brings out the acoustic tones of your guitar. 8% nickel plated steel is designed to
increase the response of a Piezo pickup in the bridge, or a magnetic pickup mounted in the
soundhole, as well as the pickups in the archtop guitars." Nickel for the piezo? Be that as it
may … However: a bronze wire, as it is customary for an acoustic guitar, turns into a bronze-
coated steel wire. To meet the cosmetic expectations, the flimsiest of coating is sufficient …
there’s that reddish gleam. It musn’t be much more, either, because bronze is a magnetic
insulator! Just imagine that across the winding, an electric current would have to flow (along
the string) … and then the guys wind around the core once bare copper wire, and alternately a
combination of copper wire and enameled copper wire. This example speaks for itself. While
bronze is not a perfect magnetic insulator, it still is less efficient than steel or nickel by several
orders on magnitude. Fig. 3.1 shows the approximate shape of the magnetic flux – strongly
simplified in order to keep the calculation effort at bay. Finding: the magnetic resistance of
the winding is determined predominantly by the surface touching the winding (Hertzian
stress). In this range, the flux density is high, the material is magnetically saturated, and the
exact calculation proves time-consuming.
Fig. 3.1: Magnetic flux in a

wound string. Single-layer
winding (left), double layer
winding with a bronze-coated
winding wire (right). The lines
of flux are not calculated
precisely; in a real string, core
and winding influence each
other mutually.
Measurement on a 0,042"-Zebra-string showed that it is less sensitive by 2 dB compared to

a steel-wire-wound 0,042"-Fender-string (Type 350). The core wires of both string have the
same diameter and the same magnetic properties – the difference results from the winding
exclusively. If one to the two winding wires were indeed made from solid bronze, the
magnetic efficiency of the remaining other winding wire would practically disappear.
Whether bronze-coated steel wire actually has a significant influence on the acoustic sound
… that would be a topic for more extensive experiments. The issue was not looked into,
though.
Unfortunately, not all manufacturers of strings give information regarding the actual build of
their strings. Tom Wheeler uses the heading "Welcome to Fantasyland" for the chapter on
strings in his reference oeuvre "Guitar Book". And he continues: “Advertisements for string
often bristle with misleading information; one almost forgets that the only serious path to a
good sound is paved with auditory experiments”. Indeed – it ain’t easy. Gerken at. al opine:
“phosphor-bronze strings sound a little more mellow that 80/20 bronze or brass strings”; in
Day et al., it conversely reads: “Phosphor-bronze sounds more brilliant than bronze”. Both
books were issued (in Germany) by the same GC-Carstensens publishers within only 2 years.
Often, the declarations about materials used flounder on the marketing primacy: brass (which
is a copper-zink-alloy), for example, turns into “bronze”. The reason might simply be that
brass is also the term for horn instruments … as played in that other kind of “band” … the one
in the football stadium. Do guitarists seek association with that scenario? Probably not, the
contrary may actually be true. (The translator recalls Pat Metheny’s “Forward March” here …) So:
“bronze” rather than “brass”. This ab-use has even migrated in German guitar-“literature”.
Now, how do you call the winding made of “real” bronze (a copper-tin-alloy), then? Right:
name it “bronze”, as well! Or maybe “phosphor bronze”, to distinguish it from the (boring)
other “bronze”. Come to think of: the mentioning of phosphor is not necessarily off, because
bronze tends to become porous … indeed, phosphor is added: has a cleaning effect and
reduces the porosity, and the high hardness of Cu3P brings more brilliance to the sound. How
much P the manufacturers add – that remains shrouded in the mystery that is string marketing.
Similar vagueness is found in “pure nickel strings”. Strings made from pure nickel could
never, ever withstand the high tensile load – you have to use steel. Only the surface (nickel
plated) or the winding (nickel wound) may consist of nickel. The winding may be made from
pure nickel or form nickel-coated steel. The manufacturers are reluctant to hand out the
specifics, though. Only the advertisement for most recent development is clear about which
side one’s bread is buttered on: “special strings for lefties” …
3.2 The loudness of the strings
If you exchange on your guitar the 009-string-set for an 011 one, will it sound louder?
Practical experience says: yes – theoretical considerations advise caution, though. First, we
should look at a meaningful intermediate quantity rather than the loudness that is difficult to
establish. Using the AC-component of the force at the bridge (acoustic guitar, pickup built
into the bridge) come to mind, or the induced AC-voltage (magnetic pickup). Keeping the
boundary conditions constant (!), there is no way around realizing that neither the bridge-
force, nor the pickup voltage includes any dependency on the string diameter.
The force at the bridge first: the excitation force transferred to the string as it is plucked may
be modeled as sum of two sub-components of equal value causing transversal waves running
in opposite directions (Chapter 2). These two waves superimpose at the bridge with equal
phase: the force at the bridge (only the AC component is of interest here) thus corresponds to
the plucking force – that’s independent of the string diameter. Still, the diameter of the string
has an effect on the sound because it affects the transverse stiffness (see appendix), and thus
the displacement of the string The heavier the string, the larger the plucking force for a given
displacement can be, and the louder the guitar will sound – if the guitarist takes advantage of
this. With equal plucking force, heavier strings bounce less (Chapter 1.5.3) and sound fuller.
We could have analyzed the dependency of internal damping mechanisms and radiation losses
on string diameter – but that had less priority and was put on the backburner.

3.2 Loudness of the string 3-5
In the magnetic pickup, the vibrating string induces an electrical voltage that is proportional
to the velocity of the string. Redoubling the amplitude of the string displacement leads to
double the velocity and thus to double the induced voltage – at least as long as we take the
linear model as a basis. However, a number of other factors enter into the transfer coefficient
of a pickup, as well: winding- and magnet-parameters, the distance between string and
magnet, the direction of the string vibration, and the string diameter … to name but the most
important ones. Our first considerations are directed to the induced voltage and its level.
The dependency of the pickup voltage on the diameter of the string was experimentally
determined using a test bench fitted with a shaker. For all measurements, a Stratocaster
pickup was deployed, with a string being sinusoidally moved up and down over its D-magnet
at a frequency of 85 Hz. The direction of the vibration was along the axis of the magnet, with
a displacement amplitude of 0,22 mm. Varying the amplitude between 0,15 and 0,50 mm
gave no indications of any substantial non-linearities: the voltage remained proportional to the
displacement in this range. The clear width between magnet and string was 2 – 5 mm; no
abnormities could be detected for these distances. The pickup-voltage level changed with
about 2,1 dB/mm for light strings, and with about 2,7 dB/mm for heavy strings. Solid strings
with diameters between 0,23 and 0,66 mm yielded proportionality between pickup voltage
and cross-sectional area of the string. Redoubling the string diameter quadruples the output
voltage (all other parameters remaining constant).
The proportionality between voltage and cross-sectional area only holds for solid strings,
though. In wound strings, the winding is magnetically not fully effective. In the experiment,
the core wires of Fender strings type 150 (pure nickel wrap), type 250 (nickel plated steel
wrap), and type 350 (stainless steel wrap) were compared. The core wires are hexagonal with
a diameter of about 0,4 mm. In terms of figures, the winding increases the cross-sectional area
by a factor of seven – the measurement shows merely double the voltage, though, for the core
with winding compared to the core without winding.
Fig. 3.2 explains why the winding is so inefficient magnetically: the individual layers only
touch at narrow fringe areas, and this is what predominantly determines the magnetic
resistance (Hertzian stress). While a part of the magnetic flux will find its way without air gap
via the helix-shaped path along the winding, this path is much longer and shows, relative to
the core, a magnetic resistance larger by a factor of 10. The magnetic effectiveness of the
winding depends, other than on the permeability, also on the mechanical tension in the
winding. If all windings are densely and tautly placed next to each other, larger areas of
contact result, with the string representing a smaller magnetic resistance. The annular area
marked grey in Fig. 3.2 is to be seen as an equivalent: a corresponding hollow cylinder would
have the same magnetic properties as the winding (measurement results from the Fender
strings).
Fig. 3.2: Wound string: the areas indicated in grey on the right are magnetically effective (compare to Fig. 3.1)
In contrast to the figure, the core of the Fender strings is hexagonal.
The winding of a string contributes to the sound in more ways than one: the mass of the
winding increases the mass of the string, but it does so without substantially increasing the
string stiffness. The hardness of the winding determines the harmonic content generated as
the string bounces off the frets. The magnetic characteristics of the winding determine the
(electric) loudness of the string. Now, loudness is a quantity that is not easily described and
that depends on many parameters, e.g. on the levels of the partials that in turn may be traced
back to the electrical partial-voltages generated by the pickup. Assuming a fretboard-normal
string vibration, the voltage of the fundamental depends on the cross-section of the string, on
the string velocity, and on the string-to-magnet distance. In the frequency range of the
fundamentals, the transfer coefficient of the pickup is still substantially independent of the
frequency and may be seen as constant (although it could well be modeled as frequency-
dependent, see Chapter 5). The clear width between string and magnetic pole of the pickup is
– for the time being – also seen as constant, so that merely string velocity and string cross-
section remain as parameters to be considered.
The voltage of the fundamental is proportional to the particle velocity of the string (law of
induction) and to the string cross-section (measurement results): U ∼ v⋅S. The string velocity
depends on the fundamental frequency and the string displacement, the latter being traceable
back to plucking force and transverse stiffness sQ. For a constant distance to the bridge, the
transverse stiffness is directly proportional to the tensioning force of the string. This force has
similar values for all 6 strings.
Assuming a constant plucking force, we obtain for the string displacement ξ:
; sQ ∼ Ψ; Ψ∼ ; } ξ ∼
The string velocity is proportional to the product of displacement and frequency. What
therefore remains for the tension is a simple frequency dependency that is independent of the
cross-section:
v∼ ; v ⋅ S ∼ 1 / fG ; U ∼ v⋅S } U ∼ 1 / fG
If all 6 strings on the guitar were solid, and given the above conditions, the E2-string would
generate the quadruple voltage relative to the E4-string. Because in each string the second
harmonic is of double the frequency of the fundamental, the same relationship would be
found here, as well. This simple consideration may not readily be transferred to all partials,
but we can already say without diving into the depths of loudness-calculation that the bass-
strings would be too loud in comparison to the treble strings. However, the wound strings are
magnetically less efficient than the solid treble strings, and therefore all strings generate (via
pickup, amplifier, and loudspeaker) a similar loudness as a first approximation.
Fig. 3.3 presents the dependency of the level of the fundamental on the frequency. This graph
may serve as rough orientation regarding the loudness of the strings (although of course
loudness and level are two different quantities). If all strings were solid, the dashed 1/f-line
would result. The measurement values (gathered with a Fender 150 string set: 042-032-024-
016-011-009) are indicated as the bold line. All measurements were performed over one and
the same magnet of a 1972-Stratocaster-pickup. The figure on the right shows the results
taken from a typical bronze-wound string set (again measured with the Stratocaster pickup).
3.2 Loudness of the string 3-7
Fig. 3.3: Level of the fundamental of the 6 guitar strings (magn. pickup). Winding: nickel (left), bronze (right).
The ratio of core diameter to overall diameter presents a significant parameter of the wound
string. For E2-strings this ratio is in the order of 0,33, for G-strings rises to about 0,6 – with
manufacturer-typical variations (Chapter 1.2). The winding-inefficiency is predominantly due
to the geometry and can therefore not be influenced much via ferromagnetic parameters.
Comparative measurements on Fender E2-strings of the types 150 (nickel-wrap), 250 (nickel
plated steel-wrap), and 350 (stainless steel-wrap) yielded comparable voltage levels for the
150 and 350 types, with the type-250-string generating 1 dB more relative to this. About half
of this efficiency increase could be attributed to the slightly larger core diameter. An
unobtainably high precision would have been required to exactly research the underlying
reasons: for a measurement accuracy of 0,1 dB, the core diameter would have to be
determined (and maintained) with a precision of 0,6% – for a core diameter of 0,4 mm this
implies a tolerance of 2,4 µm! Furthermore, the distance between string and magnet would
have to be adjusted with a precision of 40 µm. While the latter requirement appears doable, it
is certainly not trivial given a test bench made entirely of plastic components. Therefore,
tolerances of some 10ths of a dB have to be expected for all statements regarding levels.
The pickup-industry has already early on attended to the variations on string gauges;
adjustable or different-length magnets were included in the pickups (staggered Magnets,
Chapter. 5.4.6). However, apparently the differences are judged to be more on the
insignificant side, because in many magnetic pickups the 6 magnets protrude to the same
extent from the pickup housing. Be warned about unauthorized modifications, though: it is
not advisable to move the magnets in old Fender pickups – the fine-as-a-hair winding wire is
in direct contact with the magnet and can be damaged very easily. In modern pickups with a
plastic bobbin, shifting the magnets should be possible but even in this case a consultation call
with the manufacturer might be a wise idea.
Supplementing the measurement with the shaker, the levels of the strings were also subject to
an auditory evaluation. A well-versed guitarist played a Stratocaster (flush pole-pieces) fitted
with Fender 150 strings and did his best to pick the individual strings with equal force. With
much effort, it was possible to detect any significant difference in the overall level between
the D- and the G-string: the level of the G-string was about 4 dB higher relative to the D-
string. Due to a lack of reproducibility, the level differences of the remaining strings could not
be determined with sufficient accuracy. When playing regular lead and rhythm, differences
were not noticeable. The D/G-difference was just about detectable – if one really concentrated
on the task. However, as soon as the player directed his attention to the music to be played
(this would be have to be seen as the normal approach), the differences between the strings
did not stand out anymore. We did not further investigate whether there was any
compensatory action in a senso-motoric control circuit, of whether the perceptional threshold
had shifted.
3.3 Magnetic parameters of the strings
When it comes to strings, manufacturers swiftly turn into poets: "Gleaming nickel squiggles
around Swedish hex-steel and guarantees brilliant (sic!) tone with never-ending sustain.
These are your weapons of choice to deal with any degree of overdrive and get assertive solo-
sounds with bite at absolutely unbelievable killer distortion. Hotter’n Hell!", opines Gibson
sales. Which one of those not-so-few-anymore and probably not-quite-resting-in-peace
deceased 6-string-slingers will have signaled this under-worldly temperature assessment to
the ground floor?
It would appear that the required high breaking stress cannot leave a lot of latitude for
differences in the magnetic parameters. The solid strings and the core wires of the wound
strings differ only little when it comes to magnetics. Even the effects of different winding
wires remain unspectacular: measurements with nickel-wound string (Fender 150) and steel-
wound strings (Fender 350) show no difference when subjected to the shaker-equipped test
bench. The string wound with nickel-coated steel wire yielded a level higher by 1 dB … but
half of that effect is due to the somewhat thicker core wire. That does not mean that these
strings must sound the same: the mechanical vibration-behavior may well differ – but the
magnetic properties are still very similar, even if nickel and steel show different hysteresis
curves. The core-characteristics are equal in all three string-types, and together with pre-
magnetization- and saturation-effects this leads to similar magnetic parameters.
To measure these magnetic characteristics is not easy but still just about doable with sufficient
precision – and with justifiable effort. Since every measurement process includes
inadequacies inherent in the system, we will present – in the following paragraphs – several
methods of analysis to gather the magnetic data of strings. An extensive presentation of
electromagnetic fields follows in Chapter 4.
3.3.1 Measurements with the string-ring
Measuring magnetic parameters is complicated: the magnetic field in not homogenous, and
there is a non-linear relationship between the field strength H and the flux density B. A
substantial simplification can be obtained if the field-geometry can be shaped in such a way
that it can approximately be seen as homogenous. An annulus-shaped (torus-like) examination
piece that is completely wound with copper wire on its lateral surface will generate an
azimuthal circulatory magnetic field. When described using cylinder coordinates, this field
may be seen – in the space within the examination piece – as location-independent … at least
as long as DC-current flows though the copper wire. Two challenges need to be mastered in
this scenario: manufacturing a ring made of steel as it is used for strings, and the measurement
of the magnetic flux density.
For the following measurements, guitar strings were wound to form a ring. Winding a string
of a length of 85 cm into 6 turns yields a “string-ring” with a diameter of 4,5 cm. Start and
end of the string should join up as much as at all possible to minimize the effects of the
unavoidable air gap. The magnetically effective cross-sectional area of this ring is the 6-fold
of the cross-sectional area of the individual string – in the case of a 17-mil-string this will
give us an overall area of 0,9 mm2. The ring as a whole is wound – along its 14-cm-long
“core” – with a single layer of enameled copper wire (∅ = 0,5 mm); in the present
experiment, 239 turns were required.

3.3 Magnetic parameters of the string 3-9
The azimutal magnetic field strength H in the interior of this annular coil amounts to:
Field strength in the annular coil
In this formula, N1 is the number of turns of the primary coil (in our example 239), I is the
excitation current, and D represents the diameter of the ring (45,8 mm). Given I = 5 A, we
calculate H = 8,3 kA/m – this is a value sufficiently high for string-steel. In order to measure
the magnetic flux density, a second winding is wound – as a secondary coil – onto the first
one. In our example this has N2 = 100 turns. Using AC-operation, an AC-voltage is induced
into the secondary coil. This voltage depends – among other factors – on the change of the
flux density B (law of induction, Chapter 4.10).
The voltage induced into the N2 windings is U = . The flux Φ is calculated from
the product of flux density and surface area. Because the string is – compared to air – the
much better conductor for magnetic fields, we need to use (in this example) not the cross-
sectional area of the coil but six-fold the cross-sectional area of the string used. For the sake
of completeness is should be mentioned that this simplification reaches its limits as the
magnetization approaches saturation. Fig. 3.4 presents measuring results from a 17-mil-string.
On the left we see the sinusoidal current (f = 10 Hz) and the impulse-shaped induction
voltage. Since this voltage is the time-derivative of the flux density, it may be integrated to
obtain B (right-hand graph). Clearly visible is the almost square-shaped B-curve that points to
a pronounced saturation.
Fig. 3.4: Excitation current I and induction voltage U (left); Fields strength H and flux density B (right).
As we vary the frequency of the excitation current, shape and phase of the B-curve change, as
well: evidently there are delays in the build-up of the magnetic field that could not be really
expected given the low frequencies at work here. The reason for the delays is the skin effect:
eddy currents weaken the H-field, and only as they decrease, the field can be built up to
strength. The H-field reacts to changes in the current in a delayed fashion, and therefore the
magnetic flux also reacts with a delay to such current changes (Chapters 3.3.2 and 4.10.4). To
minimize the effect, all string-rings used were fashioned using lacquered strings – that way,
eddy currents can circle only within the individual string (Figs. 3.5 and 5.9.17). To measure
the hysteresis, eddy currents do not need to be determined quantitatively: it is sufficient to
decrease the frequency in successive measurements until the differences become smaller than
the envisaged measuring error. For this, imprinted voltage is more purposeful than imprinted
current.
3.3.2 Skin effect in steel strings
As a string moves within the magnetic field of a pickup, its position relative to the pickup
magnet changes. As a consequence, field strength and flux density within the string also
change. A variation in the flux density will induce, in the electrically conductive string, an
eddy current (Fig. 3.5), that itself generates its own magnetic field in opposite direction to the
primary field. Because the strength of the eddy current depends on the change of the primary
field, the primary field is more and more squeezed out of the string as the frequency increases.
At high frequencies, a substantial magnetic flux is left merely in a thin outer layer (i.e. the
skin) of the string. Therefore, the magnetic conductivity decreases with increasing frequency.
This so-called skin effect is dependent on the basic magnetic conductivity of the material (a
large µ results in a large B), and on the electrical conductivity (a large σ results in a large I).
An extensive discussion of the skin effect will follow in Chapter 4.10.4.
Fig. 3.5: Metal cylinder permeated axially by the magnetic field H, with eddy current I (left); radial distribution
of the magnetic flux density in a 17-mil-string (middle). For µr = 100 (right), there is almost no field distribution:
the magnetic flux density is practically independent of the location. Approximation: µr is constant.
Given a sinusoidal vibration, the temporal change of the flux density is particularly strong at
the zero-crossing. At these instants, the magnetic field will therefore not be able to permeate
the complete string material – there will be delay in the build up of the field. In the left-hand-
section of Fig. 3.6, the hysteresis loop measured at 1 Hz is depicted; on the right we see the
broadening at increased frequency. The skin effect is relevant if the whole hysteresis reaching
into saturation is measured. Given the string vibrating over a pickup magnet, we find other
conditions, though: within the string there is a strong DC-field with a rather small change
superimposed. In this case it is not the differential permeability that is important, but the
reversible permeability, the latter being much smaller in steel string than the differential
permeability ( µ rev < 70, Chapter. 3.3.3).
Fig. 3.6: Hysteresis loop, maximum inclination (left); frequency dependent broadening (right).

3.3.3 Reversible permeability
The connection between magnetic field strength and magnetic flux density is a non-linear one,
and it is also dependent on the previous history: the hysteresis loop shown in Fig. 3.6 can only
be run through clockwise (see also Chapter 4.3). However, for small variations around an
operating point (DC-field and superimposed AC-field), the changes do not happen along a
small section of the hysteresis curve, but on much more shallow curves. Their much less
pronounced inclination (dB / dH) is the reversible permeability µ rev. Fig. 3.7 shows
measurement results determined with a string ring. A low-frequency sinusoid (1 Hz) forms
the large drive signal, with a weak 266-Hz tone superimposed. The B-field does not follow
the reversals in the drive signal on the large hysteresis but on the flat small lines (that in fact
are lance-leave-shaped loops, as magnification would reveal). The gradient of these flat lines
is highest for the flux density approaching zero and decreases as the magnitude of the flux
density increases.
Fig. 3.7: Hysteresis curve, determined with a two-tone signal (1 Hz @ 0 dB; 266 Hz @ -32 dB). Right: slope of
the flat lines shown dependent on the flux density, i.e. this is the reversible (relative) permeability. The dashed
curve holds for the ‘reversal’ of the hysteresis i.e. for the upper branch of the hysteresis.
Already early on, R. Gans published a formula connecting B and µrev♣. It turns out, however,
that this “Gans-sian curve” may only be regarded as a rough orientation; even the supposed
independence of H is not present⊗. Fig. 3.8 shows corresponding measurements taken with 5
solid strings in comparison to the “Gans-sian curve”.
'Gans-sian curve'; compare to Chapter 4.10.3
Fig. 3.8: Measured relative permeability, calculated “Gans-sian curve” (= “Ganssche Kurve”).
♣
R. Gans, Annalen der Physik, 23, p. 399; 1907.
⊗
H. Jordan, Annalen der Physik, 21, S. 405; 1934.
3.3.4 Measurements with the yoke
Putting together a string-ring wound with two coils is highly time-consuming. For any
investigations into the market, simpler measurement approaches would thus be desirable. In
the following, the test-bench utilizing a yoke will be introduced: it employs a ring-shaped
electromagnet with an air gap. The string to be measured is inserted into the latter.
To measure magnet parameters, advantage is taken of the continuity conditions that appear
at boundaries as the field permeates them [e.g. 7]. At the string/air-boundary, the tangential
component of the field strength H is continuous. Therefore, if the axis of the string is directed
in parallel with the field lines within the air gab, the field strength Hi internal to the string
corresponds to the field strength HL within the adjacent air layer. The field strength in the
interior of the string can therefore be determined without having to actually enter the string.
To measure HL, two coils of different diameter are wound around the string: a tightly fitting
inner coil with the diameter D1, and coaxially a second coil with the diameter D2 > D1. As a
sinusoidal AC-flux Φ flows through the string, induction voltages are generated in both coils.
These voltages depend on Φ, on the frequency f, and on the turns-numbers. If both coils
feature the same number of turns N, opposite-phase connection makes it possible to
compensate for and cancel out the part of the voltage that results from the magnetic flux
flowing through the inner coil. As a consequence, the combination of the two coils measures
only the magnetic flux in the ring-shaped range between the two coil surfaces. Using this
approach, the field strength HL in air can be determined via µ0 (the known permeability of
air). HL corresponds to the axial field strength in the string (provided there is homogeneity).
Fig. 3.9: Coaxial annular coil.

Left: two windings with 1 turn each.
Right: ring-winding for measuring H.
Fig. 3.9 presents a cross-section of the measurement setup. The magnetic field generated by
an electromagnet (not shown in the figure) is directed perpendicularly to the viewing-plane. It
runs in parallel to the string axis and permeates two coil-windings concentrically surrounding
the string. The overall cross-sectional area is designated SS, the cross-sectional area of the
inner winding is S1, and that of the outer winding is S2. For reasons of clarity, each winding
consist of merely one turn in the figure; in practice about 100 turns each yields a good
compromise between sensitivity and (small) size. The number of turns of the two coils should
be exactly the same♣; they are connected in opposite phase. Given this setup, only the
magnetic field flowing between the two windings into the ring-surface forms a contribution to
the induced voltage. In the right-hand part of Fig. 3.9, two ends of the windings are connected
such that a winding WH encompassing the ring surface (S2 – S1) results. The voltage induced
in WH depends, according to the law of induction, on the turns number N, and on the temporal
change of the magnetic flux ΦRing permeating the ring surface. This flux is again a product of
ring surface, magnetic field strength H, and the permeability of air µ0.
♣
If both coil-voltages are recorded separately, correction can also be achieved via post-processing.

From the ring induction voltage UH, the field strength at the ring surface H can be
calculated:
; ;
S2 – S1 = ring surface, µ0 = 4π⋅10-7 Vs/Am
Prerequisite for exact measurements is a homogenous H-field; with pole shoes of high
magnetic conductance this can be generated with sufficient accuracy. The measuring coils can
be wound with very thin wire, making small dimensions possible. About 100 turns will
generate induction voltages in the range of 10 – 100 µV, which is comfortably measurable
with a low-noise amplifier. Highly significant is avoiding measurement of external interfering
fields (connecting lines, shielding, grounding!). In case not the voltage of the ring winding is
recorded, but rather the individual voltages of the two coaxial coils, particularly high
precision is required: the ring-voltage UH results from the difference of two voltages that may
potentially differ by a factor of 100. Any imbalance between the measurement channels (even
as small as in the ‰-range) may lead to unacceptable errors.
On top of the field strength at the ring surface H (that approximately corresponds to the
axial field strength of the string), the axial flux density of the string needs to be measured as
the second field quantity. Magnetic flux in the string Φ and flux density in the string can be
determined via the voltage U1 induced in the inner coil. However, this involves a systematic
error because the inner coil will not directly touch the string in a test-bench suitable for
various string diameters. Instead of measuring only the part of the flux that flows through the
string, a part of the flux that flows through the surrounding air is measured in addition. Given
high permeability of the string, this error would possibly be negligible – but in the saturation
range the string-permeability is precisely NOT high anymore, and the error would be
inacceptable. Still, there is an elegant way to directly measure the magnetic polarization J of
the string. J may be imagined as “material-bound part of the flux density”. Given imprinted
field strength H, the flux density results in air. Introducing ferromagnetic
material into this H-field will increase the flux density to . This is transformed to:
J = magnetic polarization
Thus J is the share by which the flux density increases (from B0 to B), depending on the given
material. Now the voltages induced into the windings W1 and W2 can be rearranged into:
is the part of the voltage that would be induced into the inner coil if there were no
string present. The part of the voltage delivered by the string is added as the second summand
. In both voltage equations, may be eliminated, and J can be calculated♣:
Given known geometry of the coils, the field strength and the polarization in the string can
now be determined from the two coil-voltages U1 und U2.
♣
The letter J is – loco citato – also used for the electrical current density!
The accuracy for the H-measurement is determined by the ring voltage UH and the area of the
ring. The potential problems with forming the difference have already been noted. The ring
area should be very small in order to capture exclusively the field in the air directly next to the
string; this makes the precise determination of area difficult, though. A solution is the use of
Helmholtz-coil enabling us to generate a highly accurate magnetic field and to calibrate the
H-measurement that way. For establishing the value of J, especially the area-ratio k = S2 / S1
needs to be precisely known. Calibration is done without string: the value of k is corrected as
necessary until J reaches zero. For the integration (which advantageously is performed with
digitized signals in a simple manner), attention needs to be paid to extremely precise offset-
compensation. If errors occur here, the hysteresis curve fail to close in the case of multiple
revolutions; it will rather diverge, or be represented with the wrong width.
Fig. 3.10 shows measurement results of a “no-name” string that were gathered with the
measurement setup as described above. The H/J-relation is typical for metals that are
magnetically hard to a lesser degree. We obtain J HC = 1,6 kA/m for the coercitivity, and we
get JR = 1,4 Tesla for the remanence. A comparison with “name products” (Fig. 3.11)
indicates small differences regarding the magnetic parameters. The sources of these
differences cannot be clarified unequivocally – it may be assumed, though, that the tolerances
due to the test bench are in a similar order of magnitude. Let us therefore remind ourselves
that measuring magnetic parameters requires much effort, and despite this effort they can only
achieve a modest accuracy.
Fig. 3.10: Hysteresis-loops measured on the yoke-test-bench for a “no-name” string (∅ = 0,43mm, plain).
The measurement frequency (2 Hz) is sufficiently low for individual strings.
Fig. 3.11: Hysteresis-loops measured on the yoke-test-bench for “name” strings (∅ = 0,43mm, plain).

4. The Magnetic Field
Macroscopic magnetic effects were already known in ancient times: Magnetite attracts iron
particles. This force effect can be described by a vector field in which a defined field intensity
is assigned to every point in space, characterized by its strength and direction. Every magnet
produces a magnetic field in its vicinity which decreases rapidly with distance. Energy
conservation is of course valid: No energy needs to be added for retention of the field (!). If
there is a displacement of a piece of iron by the magnetic force, mechanic energy is “gained”;
at the same time the magnetic field is weakened. If, conversely, the piece of iron is detached
from the magnet, the same amount of energy has to be added which will increase the energy
of the field accordingly.
Materials which generate and sustain a permanent magnetic field are called Permanent
Magnets. This characteristic is predominant for Magnetite (Fe3O4) and some other metals.
The root cause of the magnetic field is electrons moving around the atomic nucleus and their
own intrinsic spin. According to the Bohr-Rutherford atomic model, electrons move in
stationary orbits without energy dissipation but produce a magnetic field. A more or less
intense magnetic field effect evolves in macroscopic space according to the direction and
strength of these fields and the coherence effects of neighboring atoms.
In the same way an electric current flowing through a wire will produce a magnetic field. This
field will further increase if the wire is wound to form a coil. However, contrary to the
permanent magnet, the electromagnetic field disappears if the current is switched off. The
name of this kind of magnet is derived from its operational principle: Electromagnet.
Permanent magnets and electromagnets have the same effect. Both produce magnetic fields
and forces on iron and similar metals. Energetically there seems to be a difference. A
permanent current flow is necessary in order to sustain the magnetic field for the
electromagnet, which means that energy needs to be supplied. However, one has to
distinguish between the one-off portion of energy which is needed to build up the field and
the continuous supply of energy which heats up the wire (current × voltage × time = energy).
In an ideal conductor (superconductor), the magnetic field could be sustained permanently
without the continuous addition of energy.
In addition to the force effect of magnetic fields, there is also the effect of Magnetic
Induction. A change of the magnetic field over time produces (induces) a voltage in a wire
coil. This effect is exploited in a magnetic Guitar Pickup, in which a vibrating steel string
changes the magnetic field of a permanent magnet, inducing a voltage in the coil of the
pickup. Knowledge from several areas is helpful to understand the principles of the pickup, in
particular Magnetostatics, which describe the stationary magnetic circuit (magnetization of
the string), Magnetodynamics, which describe time-variant changes in magnetic fields
(induction effects), and the Two-port and Systems Theories which are needed to describe the
transfer behavior as a function of frequency. The following chapters will introduce these three
disciplines in detail.
© M. Zollner 2002
4-2 4. The Magnetic Field
4.1 The Basics of Magnetostatics

We will start the following considerations with an electromagnet because the causal relation
between field-generating current and resulting magnetic field are clearly visible.
Electromagnets do not play any role for pickups, but the results that are obtained can be easily
transferred to the permanent magnets which are used in pickups.
Fig. 4.1: The magnetic field surrounding a current carrying wire; iron filings (left), field lines (right); [18, 19].
When an electric direct current flows through a very long, straight wire, a circular magnetic
field is generated around it. The effect of the magnetic field can be visualized by elongated
iron filings which are introduced into the space surrounding the wire. The filings line up into
circular lines, concentrically wrapped around the wire (Fig. 4.1). In this visualization method,
the circular lines are not perfectly aligned but easily recognizable by eye. Using the iron
filings, a method to visualize the invisible magnetic field had been discovered. The lines
marked by the iron filings (circular in this example) were designated field lines. The magnetic
field does, of course, not only exist within the field lines, rather it fills the entire space. The
line-like description is a discrete visualization of a (continuous) vector quantity equally
distributed in space.
The evolution of circular structures has two origins. The elongated filings are oriented in a
tangential direction by the magnetic field (normal to the position vector) and they arrange
themselves into groups connected together at their ends. Iron filings are a good medium to
visualize the effects of the otherwise invisible magnetic field. However, an exact quantitative
description of the field is not possible with this method. Nevertheless, the empirically
deduced circular form is the basis of an abstract analytical description of the field, called the
r
magnetic field strength H . The word “magnetic” is sometimes omitted if no confusion with
the electrical field strength is possible.
In the example of a long wire with current flow in one direction, the vector of field strength
points in the direction of the field lines, tangential to the circles or normal to the position
vector. The value of the field strength vector decreases proportional to 1/r with increasing
distance. However, before we start with the exact calculation we must first define the
reference systems.
© M. Zollner 2002, translated by W. Hönlein

4.1 Basics of Magnetostatics 4-3
r
The magnetic field is a vector field and the descriptive field-parameter H has a value and a
direction. Not every field has a vector character. For example, a spatial temperature
distribution is described by a scalar field with every point having a value but no direction.
The direction of a vector is given by an angle deviation with respect to a reference system.
In a two-dimensional scheme polar coordinates are particularly suited for the description of
the direction. The direction of a vector originating from zero is defined relative to the
abscissa, with angle deviations being counted positive in the counter-clockwise (CCW)
direction. The spherical coordinates are defined in a similar way in three dimensions. The
definition of the positive CCW direction fits into other coordinate systems (Cartesian,
complex e-function and Euler) but, ultimately, it is arbitrary: coordinates based on the
clockwise direction would also be possible. However, once the sense of direction is defined,
it has to be maintained throughout the following considerations. The direction of a magnetic
field, i.e. the direction of the magnetic field vector, is defined by the tangent to the magnetic
field line at every point in space. A tangent, however, is a straight line and not a ray.
Consequently, there are two possible reference directions 180° to each other.
The directional reference system for magnetic fields valid today has an historical foundation.
It is derived from the needle of a compass. The Earth is a huge permanent magnet, producing
a weak magnetic field between the North and South Poles. If a compass needle (a little bar-
shaped permanent magnet) is suspended so that it can move without restriction, the magnetic
force will turn it to be parallel to the field lines. The part of the compass needle that points to
the geographic North Pole was defined as magnetic north pole of the compass needle. At the
same time it was deliberately defined that the field lines emerge from this Magnetic North
Pole. This definition, however, yields that the geographic North Pole§ must be a magnetic
South Pole! In the following the North Pole is always the Magnetic North Pole. As for the
relationship between current and magnetic field direction we also have to define direction
conventions. In metallic conductors the term current flow designates the flow of free electrons
(electrical current = charge displacement over time). The direction opposite to the electron
flow is called the technical current direction (plus to minus within the electrical load). In
graphical representations this technical current flow direction is often depicted by an arrow.
The relationship between the above mentioned current and magnetic field direction can
readily be visualized with the right hand rule: if the thumb points into the direction of the
current flow the other (fisted) fingers will point in the direction of the circular magnetic field.
The field lines of an infinitely long straight conductor are concentric circles centered on the
axis of the conductor. This field is called parallel-plane, because the same circular field line
schemes will evolve on all planes which are in parallel to each other. The computation of this
simple scheme is easy but has one disadvantage in that it does not exist in reality because an
infinitely long conductor does not exist. Real magnetic fields can have considerably more
complicated structures, which can usually be described by, mostly rough, approximations.
Finite element modeling (FEM) programs, that divide the fields into small sections, may
provide solutions, but will soon reach their limits in the case of fields relevant in, and around,
pickups. In the following chapters we will first describe the basic relationships in an idealized
manner. The particularities of pickups will be addressed at the end.
§
Between the Geographic North Pole and the Magnetic South Pole there is a distance of about 1400 km. In
central Europe the magnetic field lines have an inclination angle of approx. 65° with reference to the surface.
The magnetic flux density is approx. 45µT.

The magnetic field originating from a single wire carrying a current is relatively weak. A
strong magnetic field emerges, by superposition (addition) of the individual fields, only if the
wire is wound to a coil. The superposition principle is depicted in Fig. 4.2. In this case, we
have two parallel wires with an equal amount of current flowing in directions that are
opposite to each other. Usually the technical current flow is defined from positive to negative.
In the cross section, a current flow into the picture plane is represented by a cross ⊗, and the
opposite direction out of the picture plane towards the viewer’s position is marked by a circle
with a point .
Fig. 4.2: Magnetic field of two parallel wires; anti-parallel current direction. Individual fields (left), superposition
(right)
Every wire produces a circular magnetic field propagating with the speed of light. The delays
related to the propagation speed are virtually negligible for the small dimensions of a pickup
(< 10 cm) and the low frequencies (< 20 kHz), Thus, a quasi-stationary magnetic field can be
assumed and no magnetic wave equations are necessary. The magnetic fields of both wires
have to be vector-added at every point in space resulting in the eccentrically circular lines.
Instead of the “superposition of magnetic fields,” we can also speak of the “vector
summation” of the magnetic field strength originating from both wires. However, this
superposition is only valid for a linear system. Air permeated by a magnetic field has linear
characteristics, iron has not. First, we will address the linear systems.
The magnetic field around a current carrying conductor has some characteristics which can be
immediately and intuitively understood. It is proportional to the current, decreases with
increasing distance and has rotational symmetry with respect to the conductor. Formally the
scalar value of the vector of the magnetic field at the point of measurement can be determined
by
I
H= , The magnetic field strength outside a straight conductor
2π r
in which H is the scalar value of the field strength, I is the current strength and r is the
shortest distance of the measuring point to the conductor axis. The formula is only valid for
the space outside an infinitely long straight conductor. Again, it should be stressed that, in the
case of two wires (Fig. 4.2), the scalar values cannot simply be added. Rather the field
strength has to be a vector sum. If, for example, two equally large field vectors are normal to
each other, the scalar value of the total field strength is not doubled, but only increases by a
factor of 2 .

4.1 Basics of Magnetostatics 4-5
The scalar value of the magnetic field strength can be increased if the current is increased or if
several wires are acting together. Figure 4.3 depicts several parallel current carrying wires. It
is clearly visible that the field lines in between the wires are focused into a channel. A similar,
but not identical picture can be obtained, if one single wire is wound up into three screw-like
coils.
Fig. 4.3: Cross section through the spatial magnetic field of 6 current-carrying parallel wires (left). The spatial
magnetic field of a current carrying coil (right) [19].
In Fig. 4.2 and 4.3 the field lines are used as visualization of the invisible magnetic field. The
tangent to the 3-dimensional field line marks the direction of the field strength and the
distance between the drawn field lines marks the magnitude of the field strength. The shorter
the distance between neighboring field lines, the higher the magnetic field strength. The
scaling factor can be chosen deliberately: Whether, the lines are drawn, e.g. with a distance of
1 mm or 5 mm for a magnetic field strength of 500 A/m, only depends on the clarity of the
description of the total field distribution. The real magnetic field is of course not restricted to
the drawn field lines but is continuously distributed in space.
Thus, the field lines do not represent points of equal field strength, so they should not be
confused with the isobars of a weather chart or the lines of a rcontour map. Rather, a curve
becomes a field line because the vector of the field strength H is a tangent vector on every
point of the curve. The direction of the field is defined for every point in space by the
differential quotient of the vector of field strength. From a geometrical point-of-view, the
integration of this spatial differential equation represents the connection of differentially small
direction arrows into integral curves, i.e. field lines.
Field lines are curves of equal field strength only in very simple cases such as in Fig 4.1. In
general, the value of the field strength changes rwhen moving along a field line. Thus, it stands
to reason to examine the line integral over H because the field strength is a line-specific
quantity. Calculation of the line integral means following a field line and integrating the
product of the field strength and the differential (small) line-length
r ds. The field strength is
the tangent vector to the field line along the line and, therefore, H is always parallel to ds.
The quantity which is calculated by the line integral is called the magnetomotive force V in
analogy to the case of the electric field. If one chooses to evaluate the line integral not ralong
the
r field line, but on a general space curve, one has to calculate the scalar product of H and
ds .

In contrast to electric field lines, magnetic field lines do not have an origin and an end. In
most cases they form closed loops, but infinitely long complex space curves are also possible.
The integral along a closed loop field line, called the contour integral, yields the
magnetomotive force. This force corresponds to the electric current confined by the field
line, in other words the source of the magnetic field. This relationship can be easily seen in
Fig. 4.1: In an infinitely long wire in which a current I is flowing the field strength at a
distance r from the wire is H = I / (2πr) and the contour integral along the circumference of a
circle with radius r yields I.
Even in the case that the contour integral does not run along a field line, but along an arbitrary
closed path in space, its value represents the enclosed current. In this case the scalar product
has to be applied, since the field strength vector is not necessarily pointing in the direction of
the closed loop. The current passing through rthe area defined by the contour path is given by
the surface integral of the current density J . This surface integral is called the magnetic
r
flux Θ. With it, it is possible to
r establish a relationship between the electrical origin J of the
field and the magnetic effect H :
r r r r
∫
Θ = J ⋅ dS = H ⋅ ds ∫ Magnetic flux law (Laplace Law)
S s
r r
In this equation J is the vector of current density (Amperes/Square Meter)§, H is the vector
of magnetic field strength
r (Amperes/Meter). The flux passes through a surface S defined
r by
the contour line s. ds is an infinitely small linear element of this contour line; dS is an
infinitely small surface element of the entire surface delimited by s. The surface element is
defined as a vector: The scalar magnitude of this vector is the surface area; its direction is
perpendicular to the surface element. The product of the vectors is the scalar product, which
is defined as the product of the vector magnitudes multiplied by the cosine of the angle
between the vectors. The circle on the integral symbol indicates that integration has to be
carried out along the closed curve s, i.e. a contour integration needs to be applied. If the
contour integral does not run along the entire (closed) circumference, but rather along a curve
between two points A and B, one obtains the magnetomotive force V:
VAB J
B
r r B
V AB ∫
= H ⋅ ds A
H
Magnetomotive force
A
VAB >0
The magnetomotive force is derived from a scalar product and is, consequently, a scalar.
Scalars do not have a direction but they do have an orientation (also called direction
character). VAB = -VBA is valid. Most often the orientation is depicted by an arrow: The sign of
the magnetomotive force is positive if the potential (4.2) decreases with the direction of the
arrow; in this case the direction is identical with ther magnetic-field-strength direction. If one
points with the thumb of the right hand into the J - direction (technical current direction),
then the bent fingers point into the V-direction.
§
Sometimes J is also used for the polarization or the magnetic dipole moment – this is not meant here. In
addition the current density is sometimes called j; j is used for − 1 here.

4.2 Magnetic Potentials 4-7
4.2 Magnetic Potentials
The magnetic field strength, as defined in chapter 4.1, is a differential length-specific quantity
whose line integral yields the magnetomotive force. This figure can be interpreted as an
integrated quantity (along the line) as well as the difference between the scalar potentials
associated with the start and end points of the line. The potential defines the “magnetic
power” of every point in space, whereas the magnetic field strength describes the spatial
change of this “power”. The word potential is derived from the Latin word “potentia” which
means “ability, force, power, influence”. The definition of a potential is also common in other
areas, e.g. the gravitational field can be derived from the “potential energy”. However,
assigning an absolute power to every point in space within the framework of a relative scale
immediately leads to the question about the zero point of this scale. In the case of
temperature, there is an absolute zero deduced from energetic considerations. However, for
the magnetic field this scaling is arbitrary. Strictly speaking the magnetic potential is not
defined by a relative scale but rather by an interval scale, with zero being defined by a
constant deliberately chosen for convenience. If one computes the magnetic field or the
magnetomotive force as a potential difference, this constant will disappear. This leads to the
legitimate question why a pseudo-absolute quantity (potential) is defined, if one continues to
work with differences (intervals). The explanation can be found less in the area of physics but
more in mathematics. The field and potential theory, which is based on complex function
theory, offers a universal tool for the description of all fields, independent of their individual
scaling.
In the scalar potential and the vector potential, mathematics provides us with two abstract
quantities whose physical interpretation is somewhat arduous. First of all an obvious
misinterpretation has to be ruled out: Even though the magnetic field is a vector field, it has
both a vector potential and a scalar potential. The vector potential is a vector quantity
associated to every point in space, the scalar product is a scalar quantity associated to every
point in space. The scalar potential is, however, not the scalar value of the vector potential.
The scalar potential ψ is the quantity which leads to the magnetomotive force V through the
formation of differences. If the distance between two points approaches
r zero, the respective
potential difference converges towards the magnetic field strength H . Hence, the differential
quotient to be determined is the gradient:
⎛ ∂ψ ∂x ⎞
r ⎜ ⎟ Magnetomotive force as a function of the scalar potential.
H = − gradψ = − ⎜ ∂ψ ∂y ⎟ The unit of the scalar potential is the Ampere.
⎜ ∂ψ ∂z ⎟
⎝ ⎠
The scalar potential ψ is, as suggested by its name, a scalar, the gradient isr a vector. It points
along the direction of the highest field growth. The field strength vector H points along the
direction of the highest field decrease (H-decrease) since the equation contains a minus sign.
The gradient of a constant is zero. As the gradient formation is a linear operation, an offset
does not change the gradient. As a consequence, the (arbitrary) definition of the potential zero
has no influence on the field strength: grad(ψ) = grad(ψ + const).

It is easy to deduce the field strength from the scalar potential by calculating the gradient
(spatial differentiation). Conversely, one has to calculate the line integral in order to deduce
the scalar potential from the field strength. As always, integration needs an additive constant –
the latter defines the absolute potential-zero. In the following equation this potential-zero is
assigned to the point in space P0. A line integral has to be formed between P and P0:
P P Scalar potential as function of the magnetic field

r r 0 r r
∫ ∫
ψ ( P ) = − H ⋅ ds = H ⋅ ds strength.ψ(P) is the scalar potential at point P. At point P0
the scalar potential is arbitrarily set to zero.
P0 P
The magnetic scalar potential exhibits a specific feature: It is not defined universally and,
where it is defined, it is either discontinuous or ambiguous.
r No scalar potential is allowed in
sections of space where an electric current density J ≠ 0 is present. Inside an electric
conductor no scalar potential exists. Not that it is zero, rather it is not defined. Outside the
conductor a scalar potential can be defined, e.g. in the air, which is considered to be an
insulator. If one defines the potential reference point [ψ(P0) = 0] at a point P0 outside of a
straight current-carrying conductor and circles the conductor on a circular line, the potential
will assume positive values. After a full circle one again arrives at P0. The potential at this
point equals the magnetomotive force. After two revolutions (arriving at the same point!) it
amounts to twice the magnetomotive force. The scalar potential defined this way is
continuous but ambiguous. Alternatively, one could restrict the definition range to one single
full revolution. Then the scalar potential would become unique but would be discontinuous,
because it changes its value abruptly at the borderline.
The second method is used frequently, i.e. a unique but (spatially) discontinuous scalar
potential. For rthis, a sector or domain is defined in which an electrical current flow is not
allowed (here J = 0 is valid), and boundary lines are introduced so that this area will become
“simply connected”. In a simply connected area every closed path may be reduced to a point.
In Fig. 4.4 the area outside the conductor is such a sector if the border line is introduced as a
section boundary. It prevents a multiple circulation around the conductor, but at the same time
produces a discontinuity (at the direct transition from C to A).
A
C
Fig. 4.4: Simply-connected area around a current carrying
conductor. The line to the right is a sector boundary. The scalar
potential will grow from A over B to C. The arrow indicates the
direction of the H-vector.
It might be seen as disadvantage that the scalar potential is only defined outside the
conductor. However, it does have the advantage that one (univariate) scalar is sufficient to
describe of the field instead of the three field strength components (Hx, Hy, Hz) that would be
otherwise necessary.

4.2 Magnetic Potentials 4-9
The magnetic vector potential is defined in addition to the magnetic scalar potential. It
enables field descriptions inside as well as outside the conductor. However, the magnetic
vector potential is not a very clear and accessible quantity. In fact, its existence is derived
from formal mathematical considerations and subsequent numerical (FEM) calculations of the
field (Potential and Field Theory, 4.9). The calculation of two-dimensional fields with the
FEM-software “ANSYS” is onlyr feasible with the vector potential and r not with the scalar
potential. The vector potential A is dependent on the field strength H via a special spatial
differentiation, the rotation or curl:
r r r r
µ ⋅ H = ∇ × A = rot A Vector potential§ A
Here µ is a material constant, the so-called permeability (chapter 4.3). In Cartesian

coordinates the rotation is calculated as the difference of partial differentials and can be
depicted with the nabla operator ∇:
⎛ ∂Az ∂y − ∂Ay ∂z ⎞
v ⎜ ⎟ Curl in Cartesian coordinates.
rot A = ⎜ ∂Ax ∂z − ∂Az ∂x ⎟ The unit of the vector potential§ is Vs/m.
⎜⎜ ⎟⎟
⎝ ∂Ay ∂x − ∂Ax ∂y ⎠
For magnetic fields that can be represented by a two-dimensional scheme, e.g. parallel-plane
fields, the vector potential has only one component. Both of the other components are zero.
For example, the Hz-component is zero for an H-field only defined in the xy-plane. In the
associated vector potential only Az is non-zero. This is the component of the potential which is
perpendicular to the xy-plane.
Fig. 4.4 represents such a parallel-plane field. The current flows into the plane of projection
and an H-field emerges in the xy-plane. The vector potential has only an Az-component
parallel to the current flow. The equation specified simplifies to:
r r r ⎛ + ∂Az ∂y ⎞
µ ⋅ H = ∇ × A = rot A = ⎜⎜ ⎟⎟ 2D-vector potential
⎝ − ∂Az ∂x ⎠
The vector potential presents an elegant method to define boundary conditions. This is
necessary, for example, to reduce the complexity of computations or to set boundaries to
infinite domains in FEM calculations. In addition, it is relatively simple to define field lines
with the vector potential (chapter 4.7). Figure 4.5 depicts the spatial vector relationship
between the current density and field strength, and the flux density and vector potential,
respectively.
J r r B
rot H = J
r r
rot A = B
H A
r r r r
Fig. 4.5: Spatial relationship between J and H (left) and B and A (right).
§
r
The symbol A must not be mixed up with the area vector!

4.3 Matter in Magnetic Fields
It has been found to be useful to describe fields, in analogy to fluxes of matter (water circuit),
by potential and flux quantities. The expressions flow and flowing are to be used in a
figurative sense. There is no real flow in the magnetic circuit, contrary to the water circuit.
The scalar pressure is the quantity of drive in the water circuit. If the pressure is not equal in
the entire fluid, but varies as a function of position, there are vector pressure differences or
gradients and forces acting on the fluid particles which, as a consequence, move or flow in the
opposite direction to the gradient. Hence, the pressure can be interpreted as a scalar potential
in which its gradient would be comparable with the field strength. The velocity of the fluid,
however, cannot be deduced directly within this scheme. Other characteristics of the fluid,
like viscosity and inertia as well as the boundary conditions, have to be considered.
An electric circuit is quite similar. The gradient of the electric scalar potential is the electrical
field strength and its line integral is the electric voltage. The material quantity “impedance,”
or the admittance, has to be known in order to deduce the current flow from the voltage or the
electrical field. This is not different for the magnetic circuit. The magnetomotive force or,
alternatively, the magnetic field strength, is the quantity of drive and the magnetic resistance
determines the amount of magnetic flux. As already mentioned this flux is immaterial and,
like all the other magnetic quantities, is not visible. As long as the entire magnetic field is
confined in a single material, the introduction of a magnetic flux could be dispensed with. The
introduction of the flux quantity is advantageous if several materials have to be considered.
The continuity condition is especially useful. It means that the entire incoming node flux is
zero for an incompressible liquid. If, for instance, a node is formed by three tubes and in the
first tube the incoming flux is 5 m³/s, in the second tube the incoming flux is 4 m³/s, then,
consequently, the incoming flux in the third tube has to be -9m³/s, i.e. the outgoing flux at the
node is +9 m³/s. This law is also known as Kirchhoff’s (nodal) rule or Kirchhoff’s first law.
The electric current divides itself at a conductor node and the magnetic flux at a material node
based on the same principle.
The flux quantity already yields clear descriptions without the presence of nodes. If the cross-
section of a tube with impermeable walls (!) varies, or the flow resistance depends on
location, there is still the same flux through every cross-section, given that the fluid is
incompressible. This is equivalent to electrical engineering: The same current flows through
serial resistances even though their ohmic values might be different.
The potential quantity is defined in integral and differential form. The integral quantities in
the water circuit are the pressure and the local pressure difference. The differential value is
the pressure gradient. The flux quantity is also defined as integral and differential, as the total
flux in the water circuit, e.g. m³/s, and as the flux density, i.e. flux per transverse section
(m/s). The relationship between the differential potential and the flux quantity is established
via the specific conductivity or the reciprocal specific resistance. In this case “specific”
means material specific as well as volume specific.
The quotient of pressure gradient and flux density is the specific flow resistance in the water
circuit. If the flow resistance is large, the water flow is low. A higher fluid viscosity leads to a
higher specific resistance and a slower current. For the electrical current, the quotient of the
electrical field strength (in V/m) and the current density (in A/m²) yields the specific
resistance (in Ωm). Poor conductors have a high specific resistance, i.e. they are “highly
resistive”. The specific conductance is defined reciprocally to the specific resistance. In an
electric circuit it is the quotient of current density and electric field strength.

4.3 Matter in Magnetic Fields 4-11
The specific conductivity in the magnetic circuit is called the permeability µ, which is the
quotient of the magnetic flux density B and magnetic field strength H:
Vs
B H = µ = µr ⋅ µ0 µ 0 = 4π ⋅10 − 7 = 1.257 µH/m Permeability µ
Am
In many cases the permeability µ is divided into two factors, the absolute permeability µ 0 and
into the dimensionless relative permeability µ r. The absolute permeability, which is also
called the magnetic field constant, has the unit Vs/Am, or Henry / Meter (H/m). Care has to
be taken here. The italic H is the equation symbol for the field strength, the non-cursive H
stands for the unit Henry (1H = 1Vs/A). Sometimes the unit Henry is also abbreviated by Hy
to avoid confusion. µH means one microhenry, which is 10-6 H. Again one has to
differentiate: The italic µ is the quantity of permeability, the upright µ is a prefix which means
“one millionth”.
The relative magnetic permeability of the vacuum is 1. Thus µ 0 can be interpreted as the
permeability of the vacuum. The absolute permeability µ 0 can also be applied to air with a
high accuracy. For many materials the relative permeability µ r shows only a minor deviation
from 1. These are called non-magnetic materials. In physics one further distinguishes between
paramagnetic and diamagnetic materials but this discrimination is not necessary here. For
magnetic materials (magnet materials) µ r >> 1 is valid. This holds for all iron and steel parts
and the permanent magnets of an electric guitar. Magnetic materials that can be magnetized
by weak magnetic fields are called magnetically soft. The opposite expression is
magnetically hard. The limit at which a material becomes magnetically hard can only be
described approximately (HC > 1kA/m, see later).
The permeability µ is the magnetic conductivity. A material with large µ has a high magnetic
conductivity and the magnetic flux density B can become very high even at low field strength.
In an electric circuit one would talk about a highly conductive, low resistance material. If
materials with different magnetic conduction are located next to each other in the flow
direction (parallel), the material with higher conductivity will carry the larger part of the flux.
In two parallel resistors the one with the lower resistance will carry the higher electrical
current, and if two parallel layers of iron and air are considered, almost the entire magnetic
flux will be focused in the iron, because its µ is considerably higher than 1.
The electrical current passing through a transverse area S is J⋅ S, or electrical current density
multiplied by the area. Likewise, the magnetic flux is the product of magnetic flux density
and the area. A scalar product has to be formed if the area is not located transverse to the flux
density. If the flux density depends on the location, one has to integrate:
r r r dΦ r
∫
Φ = B ⋅ dS B=
dS
⋅ eΦ Magnetic flux Φ
S
The flux density is the quotient of the flux and of the area it flows through. The flux density
vector points into the direction of the Φ-unit vector if the area tends to zero.

The permeability µ is a scalar constant only in very simple cases. µ shows a strong nonlinear
dependence on H in most cases for which µ deviates considerably from 1. Very large values
of µ can be obtained (above 10000) for small values of H. The material will become
“magnetically saturated” with increasing field strength and µ r decreases. Consequently, the
magnetic field can no longer be considered as a linear system, which has far-reaching
consequences: non-linear distortions emerge, the superposition principle is no longer valid
and there is no transfer-function and impulse response. In addition, time invariance can no
longer be assumed because the memory of permanent magnetic materials yields a hysteresis:
for increasing field strength the flux is different from the decreasing case. Finally, one has to
consider that, at least in strong magnets, the permeability becomes orientation-dependent; µ r
will become a tensor in these anisotropic materials.
The material is isotropic

r and
r linear for the simplest case.
r Then
r µ r is a constant and the field
directions of the B - and H - vectors are the same: B = µ ⋅ H . However, an approximately
linear description is also possible for a non-linear B/H relation (linearization, tangent
approximation, Taylor series) for small deviations from linearity. If the amplitude of the
signal can no longer be considered as small, an isotropic/non-linear model has to be used; in
that case µ is defined as an H-dependant series of curves.
For the anisotropic/linear model, µ is indeed independent of H, but depends on the spatial
orientation (relative to the crystal axes).
⎛µ µ xy µ xz ⎞ ⎛µ 0 0⎞
r ⎜ xx ⎟ r r ⎜ x ⎟ r
B = ⎜ µ yx µ yy µ yz ⎟ ⋅ H → B=⎜ 0 µy 0 ⎟⋅ H
⎜
⎝ µ zx µ zy µ zz ⎟⎠ ⎜ 0
⎝ 0 µ z ⎟⎠
The µ-tensor can be simplified by choosing a suitable coordinate system, so that only 3
elements remain. This can be achieved by orienting the coordinate system along the main
axes of the material (which is the direction of the largest µ); the other two corresponding µ-
values are then smaller and often equal.
Anisotropic/non-linear materials can only be described with enormous effort. For the simple
case, every one of the three µ-components is depicted as H-dependent curve or series of
curves. However, this case does not include the existence of non-linear couplings between the
three spatial directions. An exact modeling most often fails due to imprecise measurements
and a too large a diversity of parameters.
Materials with a large µ r are called ferromagnetic because, in most cases, iron (Ferrum) is
the root cause for the magnetisability. Cobalt and Nickel as well as some rare earths and
special alloys also show magnetic behavior. A single crystal of iron will show anisotropic
behavior. Its µ r yields the largest values in the direction of the cube edge. However, since all
magnetic domains are pointing into different directions in the unmagnetised (virgin) state of
iron, the macroscopic magnetic field can be considered as isotropic (quasi-isotropy). An
anisotropic behavior can be grown using particular production procedures, e.g. cool down
within a magnetic field or crystallization on a quenching plate.

4.3 Matter in Magnetic Fields 4-13
B B
d
e
a
p
b
H
g
H
c
Fig. 4.6: Ferromagnetic demagnetization curves (left).In the demagnetized state, the field strength H and the flux
density B are zero (at the origin). Increasing H e.g. towards the point a, B will increase according to the dotted
initial magnetization curve. However, if H is set to zero again, B will not return to zero but rather to the value at
point b. Applying a negative magnetic field, one will reach e.g. point c, and by reversing the field again point a.
Further increasing the field, one will reach point d via the initial magnetization curve. If now H is set to zero a
remanent flux density at point e will remain. The picture on the right shows reversible changes at very small
amplitudes (reversible permeability).
In Fig. 4.6 the nonlinear relationship between B and H for a ferromagnetic material is
depicted. This so called “hysteresis” is not only curved, it also splits into two sections:
approaching a certain value of the field strength by increasing H (from the left) will result in a
smaller B-value than by decreasing the H-value (from the right). The loop in Fig. 4.6 can only
be run through counter-clockwise.
Both the increasing and the decreasing section of the curve converge against a common
asymptote for high absolute values of the field strength – the material is magnetically
saturated. If the field strength is set to zero from one of these saturation points then a
permanent flux density remains at the crossing point with the ordinate axis. This is called the
remanent flux density or remanence. In Fig. 4.6 the remanence point is depicted by e. In
order to reduce the flux density to zero a counter field strength must be applied, which is
called the coercivity or coercive field strength. In older literature it is sometimes named the
“coercive force”. In Fig. 4.6 the coercivity point is the abscissa section of the outermost
hysteresis curve and is marked with g.
The flux density follows the curves in Fig. 4.6 only if H changes monotonically. If H is
decreased from positive values to the point p, as shown in Fig 4.6 (right), and is successively
increased again by a small amount, the return run will not take place on the large section of
the drawn hysteresis but rather on the lower part of the slanting branch. The return to p will
be realized on the upper part of the branch. For very small changes around the working point
p the branch sections will approximately coincide and their slope will yield the reversible
permeability. It is not given by the differential quotient of the B/H curve, but is smaller (see
magneto dynamics).

4.3.1 Magnetically Soft Materials
Magnetically soft materials are characterized by a slim hysteresis, i.e. a small coercive field
strength. It is easy to permanently magnetize them, but small external magnetic fields may
also change their magnetization to new values. The characterization “magnetically soft” is
chosen as to depict this easy (magnetic) access and does not necessarily mean reduced
mechanical hardness. Iron is the most common soft magnetic material. The crystal structure
is also responsible for the magnetic characteristics in addition to the chemical constituents:
cold work hardening as well as soft annealing will change the magnetic properties. Even
small amounts of additives will change the mechanical as well as the magnetic ‘hardness’.
The coercive field strength of magnetically soft materials is typically below 1 kA/m, in
special cases below 1 A/m. The remanent flux densities most often lie between 0.8 T and 1.5
T. In special cases they can be below 0.1 T. No single value can be given for the
permeability because it is strongly dependent on amplitude. The relative permeability of cast
iron is in the range of 50 to 500. Special metals may reach over 300,000.
Magnetically soft materials are used in pickups to guide the magnetic flux. The flux
originating from a permanent magnet is channeled and focused to the strings by magnetically
soft pole pieces. These pole pieces can be solid metal blocks but also laminated sheet
packages or height-adjustable screws. Some pickups (e.g. Fender, old Stratocasters) also may
have no pole pieces at all.
4.3.2 Magnetically Hard Materials
Magnetically hard materials should retain their magnetic field after magnetisation as long as
possible without external influence; they need a high coercive field strength. They are also
called permanent magnetic materials because their field will last for decades if handled
correctly. The coercive field strength of simple steel magnets is approximately 5 kA/m, for
the Alnico-alloys often used in pickups it is around 32 – 62 kA/m and up to 2000 A/m can be
reached with special magnets. The remanence is between 0.5 T and 1.5 T. The permeability
is, like in magnetically soft materials, strongly dependent on the working point Typical µ r
values are from 1 to 5. Magnets with a high coercive field strength tend to have a smaller µ r.
4.3.3 Non-Magnetic Materials
Only the vacuum is perfectly non-magnetic. µ r is slightly smaller than 1 for diamagnetic
materials, e.g. 0.99998 for Pb, and µ r is slightly higher for paramagnetic materials, e.g.
1.00002 for Al. Such small effects are completely unimportant for measurements at pickups
and also why materials like wood, copper, aluminum, all plastics (PVC, Nylon), varnish,
brass, bronze, are considered as non-magnetic (and also non-magnetizable).

4.4 Pickup Magnets 4-15
4.4 Pickup Magnets
There are several methods to detect the vibrations of a string and to transfer the movement
into an electrical current. One of these methods is based on the induction principle: A
magnetic field varying with time induces (produces) an electrical voltage in a conductor loop
(wire winding). Pickups based on this working principle are called magnetic pickups. The
magnetic field is produced by a permanent magnet and is time-dependently modified by the
vibrating string.
The magnets of most Fender and Gibson pickups are fabricated from AlNiCo alloys and this
basic materials is also of special importance for other manufacturers. Permanent magnets are
known from ancient times. However, efficient permanent magnets have been only available
since the beginning of the 20th century. At first C-steel magnets were in use and
improvements were achieved with Cr and Co-steel. In the middle of the 30s the Mishima-
Metal (13,5% Al, 28,5% Ni, the remainder Fe) was developed in Japan and a little later the
MK-Alloy (13Al, 25Ni, 4Cu). At the beginning these alloys were still called steels.
Nowadays, the word steel stands for carbon containing steel and carbon is undesirable as a
constituent in AlNi or AlNiCo alloys so today these are called magnet alloys. Alnico alloys
contain Aluminum, Nickel, Cobalt, copper, Titanium and other additives in addition to the
main constituent iron. The first alloys were fabricated without cobalt, so they are sometimes
called AlNi magnets, but sometimes also AlNiCo-magnets even though they do not contain
cobalt.
The history of the AlNiCo magnets begins around 1935, at a time when the first commercial
pickups were developed in the USA. Gibson built a magnetic pickup into the Hawaiian
Electric, which contains a huge 11 cm long horseshoe magnet made of steel. The developer
was Walter Fuller, however the pickup was known as the Charlie Christian pickup after the
artist who used it for the first time in public. Alnico magnets were first implemented at
Gibson in the 40s. At the end of the 40s Walter Fuller launched a new pickup with a bar
magnet and considerably smaller dimensions, the P 90, which is still in production. Nearly at
the same time Leo Fender started the production of the Broadcaster, which was renamed to
Telecaster shortly afterwards. It was also equipped with AlNiCo magnet pickups, however
the magnets were formed as cylinders.
One of the first Alnico alloys produced in the USA was Alnico 3 (or Alnico III). The Al
content is 12%, with 24 – 26% Ni and 0 – 3% Cu added. Co was not yet included. The
somewhat stronger Alnico 2 alloy contains 10% Al, 17 – 19% Ni, 12 – 13% Co and 3 – 6%
Cu. The even stronger Alnico 5 magnets were available at around the beginning of 1940 with
10% Al, 17 – 19% Ni, 12 – 13% Co and 3 – 6% Cu. In the following years a multitude of new
magnet materials was introduced which in the case of Alnico were supplemented with
numbers and additional letters. Patents and trademarks protect the mixing recipes and trade
names, which leads to an ever growing number of designations: Nialco, Ticonal, Alcomax,
Hycomax, Hynico, Ugimax, Columax, Coerzit, Oerstit, Gaussit and many others. In the 50s a
new type of magnet became available that does not require expensive alloy constituents.
Within a short time ferrite magnets make it to the top of the magnet market. With the
beginning of the 70s a new class of high-performance rare earth magnets is available with a
five times higher energy density. The pickup producers, however, soon realize that strong
magnets not only increase the volume but also change the sound. This is why, in the course of
a return to old values, it was necessary to declare Alnico as the favorite material again.

4.4.1 Alnico Magnets
Alnico alloys contain 7 – 13% Al, 12 – 18% Ni, up to 40% Co and up to 6% Cu as well as
possibly small amounts of Ti, Si, S and Nb. Alnico 5 (or Alnico V) is often mentioned in
connection with guitars. This numbering system (Alnico 1 -12), typical for the USA, should
classify the increasing BHmax value (volume-specific energy), however, a precise specification
of the magnetic characteristics and composition is not possible. Particularly, one has to take
into account that Alnico 2 is stronger than Alnico 3. Alnico 2, Alnico 3 and Alnico 5 are used
most often in pickups.
Br / T Hc / kA/m BHmax / kJ/m3 Al Ni Co Cu Ti
Alnico 3 0.65 - 0.75 32 - 45 10 - 11 12 24-26 0 0-3 –
Alnico 2 0.7 - 0.85 34 - 52 12 - 14 10 17-19 12-15 3-6 0.5
Alnico 5 1.1 - 1.3 50 - 62 30 - 50 8 12-15 23-25 0-4 0-0.5
Table: Magnetic characteristics and composition percents of Alnico-magnets; remainder = Fe.
Alnico magnets are differentiated in casted and sintered ones, which can be isotropic or
anisotropic, depending on their production method. The production of cast magnets consists
of melting the metallic constituents and casting the melt in the mold where it solidifies, e.g.
sand casting, chill casting, and vacuum precision casting. Untreated casted magnets have a
dark greyish-brown color. During sintering, the fine-milled constituents are baked under high
pressure and high temperature. Sintered magnets are shiny metallic, similar to nickel.
Contrary to the cast magnets, sinter magnets have improved mechanical but slightly worse
magnetic characteristics. In particular, their remanence is slightly smaller than that of cast
magnets. The coercive field strengths are similar. Sinter magnets can only be produced
economically with small dimensions and in large quantities. They exhibit fewer pores, shrink
holes and cracks than cast magnets and better retain their required composition. Alnico
magnets can only be ground due to their very high mechanical hardness (Rockwell hardness
45 – 60 HRC). The ground surfaces are shiny metallic.
Isotropic material characteristics are independent of direction. In contrast, anisotropy means

that a spatially predominant direction exists in which a certain characteristic, in this case
magnetic, is more pronounced (oriented material). Cast as well as sintered magnets without
special treatment are isotropic.
Magnetic alloys with Al, Ni and Co constituents were, and still are, produced world-wide
under different brands. As the first commercially successful pickups were developed and
wound in the USA, the American abbreviation Alnico became accepted. Seth Lover, the
developer of the Gibson “Patent Applied For” humbucker, answered the question whether he
always used Alnico V magnets with “We also used Alnico II and III, because Alnico V was
not always available. The only difference was that Alnico V did not lose its magnetization as
easily [13].”

There is something to add from a physical point of view and, obviously, also from a
commercial point of view: in 2002 Gibson communicated on their homepage: "BurstBucker
pickups now give guitarists a choice of three replica sounds from Gibson's original "Patent
Applied For" pickups – the pickups that give the '59 Les Paul Standard it's legendary sound.
... with unpolished Alnico II magnets and no wax potting of the coils, just like the originals".
However, one should keep in mind that Alnico II as well as Alinco V were produced in
different variations before wondering about the fact that today’s replica pickups are produced
out of a material that once was a stopgap. C. Heck [21] maintains four different Alnico II and
8 different Alnico V versions:
Br / T Hc / kA/m BHmax / kJ/m3 Al Ni Co Cu
Alnico II 0.73 46 12.8 10 17 12.5 6
Alnico II A 0.70 52 13.6 10 18 13 6
Alnico II B 0.75 46 13.6 10 19 13 3
Alnico II H 0.84 48 16.8 10 19 14.5 3
Br / T Hc / kA/m BHmax / kJ/m3 Al Ni Co Cu
Alnico V A 1.20 58 40 8 15 24 3
Alnico V AB 1.25 55 44 8 14.5 24 3
Alnico VABDG 1.31 56 52 8 14.5 24 3
Alnico VB (V) 1.27 52 44 8 14 24 3
Alnico VBDG 1.33 55 52 8 14 24 3
Alnico V C 1.32 46 44 8 13 24 3
Alnico V E 1.10 56 36 8 14.5 24 3
Alnico V-7 1.28 62 56 8 14 23 3
Table: Magnetic characteristics and percent compositions of Alnico magnets; remainder = Fe.
Obviously, a “typical” Alnico 5 material does not exist. The remanence values given in this
table vary by ±10% and the coercive field strength by ±11%. The variation of the respective
hysteresis curves is shown by Fig. 4.7. The units correspond to the CGSA-system common in
the USA: 1Oe = 80 A/m, 10 kG = 1T, 1 MGOe = 8 kJ/m³. When considering whether Alnico
5 “sounds” better than Alnico 2, one also has to investigate which special Alnico variation is
applicable. In addition, it is especially problematic that the magnetic characteristics of a
material not only derive from its chemical composition but also from the physical parameters
of its production process. In particular, the temperature treatments and external magnetic
fields can have lasting (permanent) impact.

AlNiCo 5
14
7 6 5 4 MGOe
12
10
kGauss
6
0
-800 -700 -600 -500 -400 -300 -200 -100 0
Oersted
Fig. 4.7: B/H-characteristics of various Alnico-5-magnets [22, 23]. 1Oe = 80A/m, 10kG = 1T, 1MGOe = 8kJ/m3.
The basics of material science are helpful in understanding of the characteristics of Alnico: In
solid metals the atoms arrange themselves in a regular periodic lattice. However, this crystal
lattice is not constructed perfectly, but also contains crystal defects which have a significant
influence on the material properties. The supply of energy (heating) results in a rearrangement
of the atoms in looser structure and the metal becomes liquid. During the subsequent cool-
down (solidification), crystallization begins at many different sites (the so-called nucleation
centers). The growth of these internally regular crystals, also called grains or crystallites,
persist until they hit a neighboring crystallite. At room temperature the metal has a
polycrystalline structure. Polycrystalline means that the entire metal volume is made up of
many single crystallites that butt up at their grain boundaries. Inside, every crystallite is
monocrystalline, i.e. all atoms are essentially arranged in a periodic lattice. However, the
orientation of each crystallite, which is only several micrometers in size, points into a
different direction.
The properties of a crystal lattice result from its constituents, the bonding conditions and the
lattice geometry. It is well known, that diamond as well as graphite consist of pure carbon.
Both materials, which in fact do not belong to the metals, have completely different
characteristics because their carbon atoms are arranged in different crystal configurations
(cubic or hexagonal). Likewise, some metals occur in different (polymorphic) crystalline
structures: iron, cobalt, manganese, titanium, tin and zirconium. At a certain temperature their
lattice system changes and so do their material properties. The change of material
characteristics is especially pronounced for alloys, i.e. metal mixtures. For instance, for an
iron-carbon alloy steel the physical properties can be changed by hardening and annealing,
although the chemical composition is not changed substantially. Also non-iron metals like
copper can change their stiffness by bending (strain hardening) although their chemical
composition remains unchanged. The root cause again is a change in the lattice structure
(lattice defects).

The magnetic material properties also depend on the crystal structure. In the iron atom the
electrons orbiting around the nucleus produce individual magnetic fields that are externally
completely compensated. On the other hand, the magnetic fields originating from the electron
spin are not completely compensated. Thus, every atom displays an elementary magnetic
dipole. Inter-atomic forces try to orient the dipoles parallel to each other as well as parallel to
the edges of the iron crystal lattice. In a cubic lattice there are six sign-dependent orthogonal
edge directions in which the elementary dipoles of a non-magnetized iron crystal are oriented.
Accordingly, regions of neighboring atoms with the same magnetization directions are
formed. PIERRE WEISS was the first to postulate these regions of equal magnetization
direction which were, hereafter, called WEISS domains, elemental domains or simply
domains. All domains are magnetically saturated; all domain atoms point in the same
magnetic direction. In general, a crystallite incorporates many domains. Their individual
orientation is statistically uniform with respect to the six lattice orientations. The entire piece
of iron is initially macroscopically non-magnetic as a result of this uniform orientation.
Initially the domain walls will move reversibly with application of a very weak external
magnetic field. These domain walls are called Bloch walls, named after FELIX BLOCH. As a
result, the domains that are parallel to the external field will grow. At higher external
magnetic field strengths the movement becomes irreversible, i.e. the Bloch walls will no
longer return to their initial position after removal of the external magnetic field, but will
remain in the nearest energetically favorable level. The movement of the Bloch walls may
even lead to a degradation (annihilation) of smaller domains in favor of the larger ones. Even
higher magnetic field strengths may lead to the reversible and/or irreversible orientation of the
elementary dipoles from the crystal axis direction to the direction of the external field. Once
irreversible changes have occurred, the magnetic orientations of the (newly formed) domains
are no longer equally distributed and a persistent (permanent) magnetization (remanence)
will be remain after the external field has been removed.
Permanent magnets are characterized by their excellent resistance of the domain

magnetization to external fields. One possibility to achieve this is to reduce the magnetic
particles to a size where no Bloch walls may be included and every magnetic crystallite may
contain only one domain. In this configuration only the more difficult, less accessible,
reorientation processes may occur without the easier movement of Bloch walls. Small
magnetic particles may be produced by milling (powder magnets) or by cooling down fused
alloys. Alnico magnets belong to the class of precipitation alloys, in which magnetic
particles can be grown to the right size by an appropriate temperature treatment (annealing).
Alloys are mixtures of materials with metallic properties. For Alnico, the main constituent
(the base metal) is iron with additional alloy elements (Al, Ni, Co, Cu). After heating (e.g. up
to 1670°C), all of the components are mixed up in a melt which solidifies during cooling. The
solidified alloy is single phase (phase = crystal class) at temperatures above 1100°C, which
means that it is made up of only one single cubic face-centered crystal class (α). Although the
alloy is already solidified, the miscibility of the components is described as the solubility
which, in this case, means a solid solution. There is, however, a maximum solubility of the
alloy components which is temperature-dependent: The maximum solubility becomes lower
with decreasing temperature.

The homogenous one-phase mixed crystals that exist at high temperatures dissociate into two
new phases which are also cubic space-centered: into the Fe-Ni-Al matrix (α2, basic
substance) and into an internally finely distributed Fe-Co phase (α1). The matrix is only
weakly magnetic. However, the ball- or rod-like Fe-Co particles are heavily ferromagnetic.
The change of texture from the mono-phase into the double-phase configuration which will
evolve during the cool down from 850°C to 750°C is called spinodal decomposition
(spinodal dissociation) [24]. Electron microscopic investigations have shown that the
developing (‘precipitated’) α1-particles are located along the cubic edges of the matrix. Once
the particles can be magnetized during their development, they can be influenced by an
external magnetic field so that they orient in a preferred direction. To achieve this, the
Curie-temperature has to be decreased by a suitable addition of Co so that it is lowered below
the spinodal temperature, because ferromagnetics can be magnetized only above the Curie-
temperature. Magnetic materials which have been cooled down in this way in an external
magnetic field will show a spatial anisotropy, i.e. their magnetic characteristics are direction
dependent. The size of the α1-particles developed during spinodal decomposition can be
changed to a large extent by a several hours long annealing (tempering) at 600°C – with
substantial influence on the maximum coercive field strength. Most effective are elongated
particles with lengths several times their diameter but with sizes well below the onset of
Bloch wall generation.
Cubic matrix crystallites with arbitrary orientation are formed during segregation; their edges
are pointing in uniformly distributed directions. (For one single crystallite the orientations of
the edges are, of course, orthogonal). During cool down in a magnetic field the α1-particles
are arranging predominantly next to the nearest edge orientation, but as the crystallites are
still directed in different orientations, the best result is not yet realized. To achieve this, all
crystallites in the matrix have to be oriented parallel to each other, which means they have to
be grown parallel to the lattice directions. Applying special treatments (unidirectional cooling,
homogeneous temperature gradient, quenching plate) it is possible to come close to the ideal
situation. Magnets produced in this manor are called grain oriented, crystal oriented,
preference oriented or columnar oriented. However, they can reach their optimum properties
only if the oriented crystal growth (crystal anisotropy of the matrix) is combined with a
proper magnetic field treatment (form anisotropy of the α1-particles).
In this short excursion into material sciences it should be pointed out that it is not sufficient to
simply characterize pickup magnets by their chemical composition. The description of
“Alnico V by 8% Al, 14% Ni, 24% Co, 3% Cu” does not provide information on the
remanence, coercivity or permeability. Moskowitz [23] summarizes this complex of
problems: There are 16 factors that determine the actual performance of a specific basic
magnet in a particular circuit. The magnetic and physical properties of the material are
directly dependent on the following factors in the manufacturing process: chemical
composition, crystal or particle size, crystal or particle shape, forming and/or fabrication
method, and heat treatment. Permeability, coercive force, and hysteresis loop are specifically
affected by gross composition, impurities, strain, temperature, crystal structure and
orientation. The effects of each of these factors are metallurgically complex and beyond the
scope of this book. After all, “this Book” is called PERMANENT MAGNET DESIGN AND
APPLICATION HANDBOOK. This book has most probably not been read by the author of the
2001 published book “E-Gitarren” who wrote: "The production of a magnet is quite simple.
The basic materials will only be exposed to a high electrical voltage … The field strength of a
magnet produced in such a way might be measured in Gauss.” ?? !!

Magnets with defined magnetic properties cannot be produced easily. Contrary to the constant
current resistance, magnetic parameters are not easily measured. The resistance variations of
±5% are discussed in depth in pickup literature and the sound difference between Alnico-5
and Alnico-2 is addressed in epic scope. However, the variation of magnetic parameters is
usually not mentioned.
Very pure components are necessary for the production of Alnico magnets. McCaig [26]
claims iron with a maximum carbon concentration of 0,02%, whereas Cedighian [25]
recommends aluminum with a purity grade of at least 99,6%. Moskowitz [23] claims very
close metallurgical controls and tolerances of, for example, ±0.05% for titanium and ±0.06%
for silicon. These tolerances must not only be realized during weighing but also for the melt.
Moskowitz [23] demands that Alnico-3 has to be homogenized at 1290°C ±5°C and McCaig
writes that a temperature deviation of only 10°C can lead to extremely poor results. Did all
magnet suppliers apply such a high precision, particularly in the forties and fifties of the last
century when the famous vintage pickups were produced? The scientists that were involved in
magnet production tried to gain insight into the crystal structures with the microscopes
available at that time. However, the optical microscopes could not resolve particles as small
as approximately 40nm x 8nm x 8nm. Electron microscopes as well as X-ray equipment were
already available, but not in large numbers. McCaig [26] notes: We at the Central Research
Laboratory of the Permanent Magnet Association became interested in the angular
distribution of crystal axes in the late 1950s. At this time we did not possess our own X-ray
equipment … Each crystal required an exposure of several hours, so the experiment was not
carried out on many samples. This statement was made in the late fifties. McCaig writes
further: Unfortunately the details of manufacturing processes are rarely sufficient to enable
you to produce magnets successfully yourself. Even when a process for making permanent
magnets is fully and honestly described, it may take several months for someone skilled in the
art to reproduce it successfully in a different environment. This was at the end of the seventies
- and is still valid.
In the early years (decades?) pickups not only had different numbers of turns but also
different magnets. Seth Lover, developer of the Gibson “Patent Applied For” humbuckers
answered to the question whether he permanently used Alnico-V-magnets: “We have also
used Alnico II and III because Alnico V was not always available. We have purchased
whatever was currently available, because they were all good magnets. The only difference
was that Alnico V did not lose its magnetization as fast [13]”. In contrast to this Gibson’s
advertisement claims: "BurstBucker pickups now give guitarists a choice of three replica
sounds from Gibson's original "Patent Applied For" pickups – the pickups that give the '59
Les Paul Standard it's legendary sound. ... with unpolished Alnico II magnets and no wax
potting of the coils, just like the originals". Right you are, if you think you are ...
“We have purchased whatever was currently available.” Obviously, the only important thing
was that it was marked “Alnico.” However, this name only means that an Iron-Aluminum-
Nickel-Cobalt alloy was used. The magnetic properties only develop during heat and, where
necessary, magnetic treatments and are manufacturer secrets. One would have to determine
the B/H hysteresis to reveal the characteristics of a certain magnet. However, to achieve this,
one would have to to demagnetize and remagnetize several times and what owner of a 1952-
Les Paul would like to perform such a treatment? Vintage pickups will therefore always be
surrounded by a mystical aura.

4.4.1.1 Alnico-III and Alnico-I
Alnico-I was derived from Alnico-III by replacing 5% Ni by Co [21]. Both alloys do not
differ significantly in their magnetic properties. Alnico-III is free of Co and, thus, is
sometimes called Alni. In the USA, however, Alnico-III is assigned to the Alnico magnets,
even without Co. Alnico-I is mainly used for larger magnets and is not important for pickups.
Alnico-III was the material of choice for smaller and cheaper magnets – and this is the reason
why it was used in the fifties by Leo Fender for the magnets of the Telecaster.
Most of the material science books quote the following composition for Alnico-III: 12% Al,
24-26% Ni, no Co, 0-3% Cu, remainder Fe. The maximum remanence which can be achieved
is 0.6-0.75 T, the coercivity is 32-45 kA/m, the maximum energy density is 9-12 kJ/m³. The
cool down procedure also has an influence on the magnetic properties, in addition to the
chemical composition, and subgroups are designated by additional characters, e.g. Alnico-III-
A. Alnico-III magnets are isotropic and are available as cast or sinter magnets.
AlNiCo 3
9
L 2 1 MGOe
/D
=
4
8
L/
D=
3
7
kGauss
4
0
-700 -600 -500 -400 -300 -200 -100 0
Oersted
Fig. 4.8: B/H-characteristics of different Alnico-III-magnets [21 - 23]. 1Oe = 80A/m, 10kG = 1T, 1MGOe =
8kJ/m3. L / D = length / diameter (cf. Fig. 4.11).
Fig. 4.8 shows the B/H-curves of several Alnico-III magnets. Their points of intersection with
the energy-hyperbolas are located close to 1.4 MGOe = 11,2 kJ/m³. The spread of coercivity
values, which is depicted as on the abscissa, is considerable.

4.4.1.2 Alnico-II
Alnico-II contains more cobalt as well as copper, which leads to a slightly higher price
compared to Alnico-I, -III and –IV magnets [21]. Alnico-II shows the highest BHmax value of
all isotropic Alnicos.
Most material science books quote the following composition for Alnico-II: approx. 10% Al,
17-19% Ni, 12-15 Co, 3 - 6% Cu, sometimes some per mills Ti and S, remainder Fe. The
achievable remanence is 0.7 - 0.85 T, the coercivity is 34 - 52 kA/m, and the maximum
energy density is 11-16 kJ/m³. In addition to the chemical composition, the cool down
procedure has an influence on the magnetic properties. Alnico-II is isotropic and available as
cast or sinter magnet. Alnico-II can be treated with external magnetic fields but the gain in
energy is only approximately 10% due to its relatively low cobalt content [21].
AlNiCo 2
9
L 2 1 MGOe
/D
=
4
8
L/
D=
3
7
kGauss
4
0
-700 -600 -500 -400 -300 -200 -100 0
Oersted
Fig. 4.9: B/H-characteristics of several Alnico-II-magnets [21 - 23]. 1Oe = 80A/m, 10kG = 1T, 1MGOe =
Fig. 4.9 shows the B/H-curves of several Alnico-magnets. The maximum specific energy is
located between 1.6 – 2 MGOe = 12.8 – 16 kJ/m³. The comparison with Alnico-III yields
somewhat higher values for coercivity and remanence.

4.4.1.3 Alnico-V
Alnico-V is anisotropic and reaches the highest BHmax-values of all Alnico-alloys [21].
However, its price is higher due to its considerably higher cobalt content. Alnico-V is the
material of choice for nearly all Fender pickups.
Most material science books state the following material composition for Alnico-V: approx.
8% Al, 12-15% Ni, 23-25% Co, 0-6% Cu, sometimes some per mills Ti, Si and S, and the
remainder Fe. The maximum remanence is 1.1 – 1.3 T, the coercivity is 50-62 kA/m, and the
maximum energy density is 30-60 kJ/m³. Besides the chemical composition, also the cool
down procedure and the application of magnetic fields has a significant influence on the
magnetic properties. Alnico-V is mostly anisotropic and is available as cast or sinter magnet.
Alnico-V can be entirely (Alnico-V-7) and partially (Alnico-V-DG) grain-oriented. Many
different brand names exist on the international market.
↓ 64 kA / m ↓ 40
AlNiCo 5 ↓ 20 ↓ 10
14
L/ 7 6 5 4 MGOe
D
=4
12
10
L/
D=
3
8
kGauss
6
0
-800 -700 -600 -500 -400 -300 -200 -100 0
Oersted
Fig. 4.10: B/H-characteristics of several Alnico-V-magnets [21 - 23]. 1Oe = 80A/m, 10kG = 1T, 1MGOe =
Fig. 4.10 shows the B/H-curves of several Alnico-V magnets. When compared to Fig. 4.8 and
4.9, one recognizes the much more pronounced cubic form of the hysteresis; the maximum
specific energy reaches values between 5 – 7 MGOe = 40 – 56 kJ/m³. It can be assumed, with
all caution, that the Alnico-V-alloys used for guitar pickups exhibit the lower BHmax-values,
for cost reasons.

4.4.1.4 Other Alnico-Materials [21]
Alnico-IV has, in comparison to Alnico-I to III, a relatively high coercivity which makes it
suitable especially for magnets with a small length-to-diameter ratio.
Alnico-VI was derived from Alnico-V. The coercivity increases with higher Ti content (up to
5%) while, at the same time, the remanence decreases. A further increase of this trend is
realized with Alnico-VII.
Alnico-VIII, -IX and -XII contain 35% Co. The expensive cobalt enables coercivities up to
130 kA/m, however production is difficult because the material is very brittle. The remanence
and specific energy density are smaller than for Alnico-V.
Alnico-V and Alnico-II is used mostly for guitar pickups, occasionally also Alnico-III.
4.4.1.5 Comparison of selected Alnico-Materials
Most guitarists want to play the guitar without considering whether their pickup magnets are
crystalline or form-anisotropic. This explains why pickup advertisement does not refer to the
material parameters but rather to the sound. The advertizing message sounds more competent
with the gleam of expert knowledge and the disclosure of proprietary information. This reads
as:
Alnico-II:
“For a vintage-oriented, warm sound. Since the magnetic field is somewhat weaker than for
an ordinary Strat-pickup, the string swings out more freely and naturally. The result is an
improvement of the sustain behavior.”
But also: “For the rather weak Alnico-II the tone literally breaks down.”
Or: “Pickups with Alnico-II-magnets are softer in their sound character, posses less treble,
are more quiet, more rounded and somewhat less dynamic.”
But also: “Due to its Alnico-II-magnet, the pickup does not loose treble.”
Or: “Alnico-2 corresponds rather precisely to a mature Alnico-5-magnet.”
Alnico-V:
“Alnico-V = clear/powerful sound, more wiry twang, more powerful bass.”
But also: “Alnico-V = bluesy base character with pleasantly rounded tone.”
As well as: “Alnico-V = fast attack and slightly undifferentiated reproduction.”
Or: “Stronger magnets will deliver less treble.”
But also: “The stronger Alnico-V-magnet sounds more brilliant.”
Alnico-VIII:
“The higher magnetic power of the Alnico-8-magnet results in a sustain loss.”
But also: “Louder pickups possess more sustain.”
As well as: “Alnico-8: The pickup produces high output power with little compression also
for hard plucking.”
Sources for chapter 4.4.1.7: Gitarre & Bass, Musik Produktiv; Rockinger; E-Gitarren (Day et al.).

Nearly no retailer who promotes pickups in his advertizing material makes an effort to
investigate differences in sound produced by the exchange of magnets. They may compare
two guitars, one of them sounding more trebly than the other, one of them with Alnico-V-
magnets in the pickups, the other with Alnico-II-magnets. Then the root cause is clear at once
and the advertizing text is ready. The rules of physics sometimes seem to be a real challenge
for textbook authors as well:
‘According to the information given by manufacturers of magnets, Alnico magnets are

supposed not to weaken over the course of the years, but to retain their Gauss-values and thus
their magnetic power over a long time. On the other hand, the pickup-industry claims that
Fender-type pickups noticeably loose magnetic power already after 2 years, and that Gibson-
type pickups do so after 3 years. However, this apparent discrepancy can be explained
because the supposed loss in power evidently seems to be a decline in the „retentiveness“ of
the pickup-magnetism. This means that with the vibrating string disturbing the magnetic field,
the particles of older magnets can be more easily thrown into disorder (and thus experience a
short-term loss of magnetic force) – compared to brand-new magnets.’ (E-Gitarren, Day et al.
– German text retranslated into English). Of course, a magnet does have a force – it can draw
an iron nail way from a table-top, for example. However this force is measured not in units of
Gauss but in units of Newton. The unit Gauss relates to the magnetic flux density but this is
not the definition applied by the above author-collective: „The field strength is measured in
units of Gauss”. Sorry, no agreement here – not with the scientific literature, anyway, which
defines the Oerstedt (in the US) or the A/m (in Europe), respectively, as the unit for the field
strength. Day et al. do have a quantity allocated to the unit Oerstedt, as well: “the resilience
against demagnetization”. That is not completely wrong if we think in terms of the coercive
field strength that actually is measured using the unit of Oerstedt and A/m, respectively.
However, the term “resilience” again opens the door to mix-ups. “Magnetic power”, as well,
is such a term that can be misunderstood easily, since power is measured in units of hp, or
Watt, or Nm/s. Any author trying to explain difficult technical context with simple, musician-
friendly terms runs the danger of being open to attack, and risks to be criticized in case of too
rigorous simplification. It does not really help, however, to assign a new meaning to
established terms just to achieve the simplification. Of course, a scientist will be criticized just
as much if he remains lost in his non-linear differential equations in an effort to maintain
exactness and full integrity. Accordingly, the journalist in the German magazine Gitarre &
Bass (4/2006) opines: ’Caution, if – in the matter of guitar speakers – somebody brings
science to the table. The man♣ will probably carry lots of misconceptions. It is in fact best to
give such people a wide berth.’ Another statement: ‘What’s all that scientific nonsense,
anyway?’ The same journalist does, however, also write: ‘A myriad of these prejudices exist
that seem to almost be set in concrete. Who actually decides on such bullshit? These theories
are supported by numerous books on guitars written by famous (or infamous) luthiers who
actually assume the right to stipulate how much a Telecaster may weigh, or how a
Stratocaster pickup should be adjusted.’ And once more a passage from the book “E-
Gitarren”: ‘If pickups remain close to AC-fields such as transformers or strong heat sources,
their magnetic structure becomes totally jumbled and they age more quickly.’ O.k. – yes,
above 500°C it will indeed start to be a critical situation – but it will not be only the magnetic
structure that becomes jumbled, but the tone-generating guitarist’s layout, as well: mighty
quick aging! (Paragraph translated by T. Zwicker)
♣
this would be Dr. Bose, loudspeaker designer and lecturer at the M.I.T. "with dubious formulas"

Obviously, the magnet is involved in the generation of sound: without the magnet there would
be no sound. It is also clear that the magnet itself does not have a sound. Alnico-II will not
sound different compared to Alnico-V. There is, scientifically speaking, no tone at all if the
string does not vibrate. However, one can moan less and talk less elaborately about the
“sound of the magnet” if one means its effect on the transfer characteristics. So how does
Alnico-V sound? Different from Alnico-II and, if yes, why?
The change of flux density is relevant for the induced voltage in the coil. A strong magnetic
field will not induce anything as long it does not change. For a change of the flux density the
string has to vibrate in a position-dependent, inhomogeneous magnetic field. If the magnetic
field would be constant at every position, no voltage would be induced. The inhomogeneity of
the magnetic field can be influenced by the magnet material as well as by the shape of the
magnet. Replacing the magnet might also change the permeability and, consequently, the
resonance of the pickup and/or the damping of resonances by eddy currents (resonance
quality). Thus, the behavior is by no means mono-causal, where one cause produces one
effect or rather that every effect can be attributed to one root cause. Rather, the relationships
are complicated and multi-factorial.
The difficulties start already with the material specifications. Fig. 4.10 shows, that there exist
several Alnico-V-alloys. In the pickup literature there are no indications on sub-groups, only
"Alnico-V", "The holy grail" or "The originally PAF". Not even Seth Lover was able to tell
which material was used during which time period, and how much turns were wound. Was
Eric’s favorite Paula equipped with Alnico-II or Alnico-V? Unfortunately she is no longer
traceable (or rather she is hanging in Japan in 17 safes – and every one of them an original!).
Does the transcendental sound of the Roy-B-guitar stem from the Alnico-III magnet or from
the fact that vintage Telecaster pickups with resistances beyond 11 kOhm have been spotted?
Or maybe it lies with the guitarist?
Fig 4.11 summarizes the hysteresis-scattering. In this graphical representation regions were
defined based on the trends of the hysteresis curves of many Alnico-materials. One can
recognize the scattering and the basic differences. Alnico-II is slightly stronger than Alnico-
III but considerably weaker than Alnico-V. In Fig. 4.11b the attempt is made to extract a
typical single curve from the many different possibilities, but without evidence that these
curves are the authentic or the most suitable ones.
When comparing different magnetic materials one first has to define into which magnetic
circuit the magnet will be integrated: single coil or humbucker (or special construction types).
Single coil pickups with cylindrical magnets, like those that were originally designed for the
Stratocaster, do not have ferromagnetic materials other than the magnet. The magnetic load is
defined by the shape of the magnet, or rather, strictly speaking, by the shape of the
surrounding air space. Frequently the length to diameter ratio is approx. 4 yielding a working
point near the knee of the hysteresis. Fig. 4.11 shows two straight load lines for L/D = 3 and
L/D = 4. However, one has to take into consideration that literature values may differ [23, 25],
and the slope of the lines is decreased by the neighboring magnets. Very roughly simplified,
for pickups with cylindrical magnets Alnico-V will produce a magnetic field twice as strong
as that of Alnico-II or Alnico-III. However, the flux density derived from the crossing of the
curves corresponds to the center of the magnet (neutral plane, chapter 5.4.1), not to the
location of the string.

If the magnetic circuit were a linear system this would result in a simple relationship: the
vibration of a string would change the magnetic resistance e.g. by 1% and consequently the
magnetic flux by 1%. Doubling the static flux, e.g. by exchanging the magnets, would result
in a doubling of the alternating flux and doubling of the induced voltage – the generated tone
will become louder and possibly more distorted. However, as the magnetic circuit is
nonlinear, doubling of the flux will result in an induced voltage slightly less than double. At
the same time the magnetic aperture will be decreased (Chapter 5.4.4) and the aperture
dependent treble drop becomes weakened, i.e. the pickup sounds slightly more brilliant. An
additional brilliance gain with cylindrical magnets might evolve from the fact that stronger
magnets possess a smaller reversible permeability – the inductance will become smaller, the
resonance frequency will increase and the figure of merit will likewise increase slightly
(chapter 5.9.3). On the other hand, a treble loss due to eddy currents will also be induced. The
electrical conductivity of Alnico-V is approximately 40% higher than that of Alnico-II. It is
hard to predict which effect will dominate; however, in most cases the stronger magnet yields
a gain in brilliance.
The magnetic field at the string location will be weaker for single coil-pickups with bar
magnets instead of cylindrical magnets. The magnetic aperture tends to be larger and the
aperture-dependent treble loss will be somewhat higher. The reversible permeability of the
magnet nearly does not play a role because it will hardly be penetrated by an alternating flux.
The frequency dependence of the impedance of the SDS-1, for example, will not measurably
change if both bar magnets are removed. For the P-90 the magnets have a small influence.
They increase the coil inductance by 10%.
The magnets have almost no influence on the pickup impedance for Gibson-type
humbuckers. The alternating magnetic field passing through them is negligible and, hence,
the reversible permeability and the eddy current damping play practically no role. As for the
single coil, the magnetic aperture and absolute sensitivity do depend on the magnet strength.
The working point of many Alnico equipped humbuckers is located below the hysteresis
knee, in a rather inappropriate region. The table depicted in chapter 5.4.1 shows that the static
magnetic flux densities of the investigated humbuckers are smaller than most of the single
coils.
Mechanical Characteristics of Alnico-Magnets:
Density: approx. 7g/cm³

Hardness: 45 – 60 HRC, brittle, risk of fracture, moldable only by casting and/or sintering
plus grinding.
Specific resistance: 0.45 – 0.7 Ωmm2/m. Alnico-V has a slightly better conduction than
Alnico-II. For comparison: nickel-silver = 0.3 Ωmm2/m, Cu = 0.018 Ωmm2/m, Fe = 0.1
Ωmm2/m. Ceramic magnets (ferrites) are, however, insulators.
Reversible relative permeability: approx. 4 – 6; usually lower for stronger magnets.
Alnico has a good corrosion resistance; however, it is not fully rust resistant.
Sinter magnets show a higher mechanical stiffness compared to cast magnets. Their magnetic
values are, however, somewhat worse. The quality of cast magnets is also reduced when they
have cavities.

Alnico-Magnets
14
12
Alnico 5
10
8
kGauss
Alnico 2
6
Alnico 3
0
-800 -700 -600 -500 -400 -300 -200 -100 0
Oersted
AlNiCo-Magnets
14
L/ PC 5 4 3 2 1 MGOe
D =2
=4
0
AlNiCo 5
12
PC
=1
5
10
L/
D=
3
8
2
AlNiCo
kGauss
PC
= 10
3
AlNiCo
6
CuNiFe
Bar magne
t without po 2
le pieces
0
-800 -700 -600 -500 -400 -300 -200 -100 0
Oersted
Fig. 4.11: B/H-regions of typical Alnico-magnets; the hysteresis curves are located in these regions (upper plot).
Lower plot: B/H-curves of Alnico cast magnets, data from old specification sheets. 1Oe = 80A/m, 10kG = 1T.
L / D = length / diameter. PC = Permeance Coefficient

4.4.2 Cunife-Magnets
Alnico is a very hard and brittle material, which can be machined only with considerable
effort. Cutting a screw thread is not possible with ordinary tools. However, this was exactly
what Leo Fender wanted when the former Gibson developer Seth Lover built the Fender
humbuckers for him: The cylinder magnets had to be adjustable in height by a thread. Cunife,
a copper-alloy with an addition of Fe and Ni, which was developed 1937 by Neumann,
Buechner and Reinboth in Germany was employed as an alternative to Alnico. The alloy
constituents are melted, rapidly cooled and cold-formed. Optimum magnetic parameters are
achieved with cold-formed 5 mm diameter wire; this is by chance exactly what is needed as
the diameter for pickups with single magnets. The cold-forming yields a heavily anisotropic
material with maximum field efficiency in the longitudinal direction. The magnetic
parameters are similar to that of Alnico-III.
Cunife (also called Cunife-1) consists of 60% Cu, 20% Ni and 20% Fe. The remanence
obtained is 5.4 – 5.7 kG, the coercivity 500 – 590 Oe (40 – 47 kA/m) and the maximum
energy density 1.3 – 1.85 MGOe (10 – 15 kJ/m³), which is somewhat higher than for Alnico-
III. In addition there is also a Cunife-2-alloy with a small amount of cobalt: 50% Cu, 20% Ni,
27,5% Fe, 2,5% Co. This alloy should not be mixed up with Cunico, which has a much higher
Co content. Cunife-2 will give higher remanence values at lower coercive field strengths and
is, thus, rather unsuitable for pickups.
The big advantage of Cunife is its low hardness: The specification sheets in [22, 23] state a
Rockwell hardness of B200. However, the B-Rockwell hardness is only specified up to a
maximum of 100, so maybe Brinell hardness is meant, instead of the designation ‘Rockwell
hardness’. The Brinell hardness measurement can only be used for measurements of soft and
medium-hard substances and 200HB is characteristic for the lower end of non-hardened
steels. The Rockwell Hardness employs a diamond cone (C = cone) and is adequate for harder
materials. 45 HRC characterizes the upper end of non-hardened steels, 60 HRC is
characteristic for hardened steels. Threads cannot be cut into hardened steel but they are
possible in non-hardened steel.
Cunife-magnets have not been widely used. The most famous protagonist is built into Fenders
Custom and Thinline Telecasters. It was developed by Seth Lover after he moved from
Gibson to Fender in 1967.
Spec. resistance of Cunife-1: 0.185 Ωmm2/m; Alnico has a 3 – 4 times higher resistance.
Density of Cunife-1: 7.8 g/cm3, comparable to Alnico.
The relative reversible permeability of Cunife-1 is close to 1, i.e. smaller than for Alnico.
The magnetic properties of Cunife are strongly dependent on the individual production
process (cold drawing, annealing), Fig. 4.11 shows approximate values for the B/H-curve.

4.4.3 Ceramic-Magnets (Hard Ferrites)
At the beginning of the fifties a new magnetic material was introduced, which is based on the
crystal anisotropy of barium oxide. This kind of magnet is called a ferrite, oxide or ceramic
magnet. Nowadays, mainly barium ferrite and strontium ferrite are employed. They can be
manufactured more cheaply than Alnico-magnets and achieve much higher coercive field
strengths, but smaller remanence values.
Ceramic magnets run through a powder-metallurgy production process and their magnetic
data can be tuned to a large extent. Their remanence is relatively small at 0.2 – 0.4 T, whereas
a coercive field strength of more than 200 kA/m can be achieved. The maximum energy
density, of up to 36 kJ/m³, is also much higher than for the Alnico magnets. In contrast to the
(comparatively long) Alnico-magnets, a typical ceramic-magnet is relatively short: the
optimum length/diameter ratio is close to two. This is the reason why it is employed in
(cheap) pickups as a bar magnet beneath the coil, nearly never as cylinder magnet within the
coil; for that application the geometry would be too unfavorable.
The relative permeability of ceramic magnets does not differ much from 1 and, thus, the
inductance of the coil is not increased much, even if the magnet is mounted inside the coil. In
contrast to Alnico magnets, ceramic magnets are insulators unable to produce eddy currents.
As a result, there is no eddy current dampening of the coil. However, if the field of the
underlying ferrite magnet is directed through the coil by iron rods, the eddy current losses are
higher as in the case of Alnico cylinder magnet pickups.
Even stronger magnets can be produced with cobalt/neodymium or cobalt/samarium with

maximum coercive field strengths of more than 2000 kA/m. These rare-earth-magnets are
very expensive – and for pickups only useful in “homeopathic” quantities.
Permanent magnets
14
5 4 3 2 1 MGOe
L/D=4
Alnico 5
12
10
L/D=3 Alnico 8
8
kGauss
Alnico 2
Alnico 3
6
Anisotropic Hard Ferrite 4
Isotropic Hard Ferrite

2
0
-800 -700 -600 -500 -400 -300 -200 -100 0
Oersted
Fig. 4.12: Comparison of Alnico- und Ferrite-magnets. The load curves (different for Alnico and Ferrite) specify
the length/diameter ratio of cylindrical magnets (chapter 4.6).

4.5 Magnet-Aging
Provided it is properly made and treated, the life of a modern permanent magnet is, to the
best of our knowledge, infinite. McCaig [26] will probably not be able to prove this statement
– but also does not have to. Modern permanent magnets will last forever. The magnetic field
will decrease measurably only during the first hours following the initial magnetization. At
the start some of the magnetic domains are in a meta-stable (unstable or weak) state and rather
small energy additions may cause a shift into a more stable energy level. As time progresses
these exchange effects will become increasingly less important. To avoid misunderstandings:
these processes are called after-effects or aging (ageing, relaxation, magnetic creep, magnetic
viscosity, time effect) and not demagnetization. Total or partial demagnetization means the
forced shift of the working point to smaller flux values as may be induced by load change or
the application of an external field. If a nail attracted by a horseshoe magnet is detached, the
flux density will decrease and the working point shifts down to the left on the hysteresis curve
(= demagnetization curve) in the 2nd quadrant. This is, of course, not what is meant by aging.
If, however, a magnet has lost 5% of its flux density after 10 years of storage without being
used, then it has aged. Between these boundaries there is, however, a grey area with
components from both worlds.
The main causes for aging are changes of load and temperature; other sources do not play any
role for pickups. Reversible aging can be compensated by new magnetization and the magnet
then appears “like new”. During irreversible aging, however, the internal crystal structure is
changed and the former values will not be reached again.
A quantitative description of aging processes needs the application of high-precision

measuring equipment and much patience. Prediction is difficult, especially into the future –
not different from stock prices. If the flux density has decreased by 0.1% in the first year and
the precision of the measuring equipment is of the same order, one cannot make exact
predictions for the next 10 years. On the other hand, a measurement covering 10 years is also
not without problems, because a great many parameters have to be kept constant during the
entire measurement period.
The natural aging without external interference is described by a logarithmic law:
B(t ) = B0 ⋅ (1 − k ⋅ lg(t / τ ) ) , t may not be too close to zero
where B(t) depicts the time-dependent flux density, k is a material constant (which can also be
dependent on geometry and size) and τ is a reference time, e.g. one day after production. For t
= τ one gets B = B0 the flux density after one day. k = 0.01 would mean that B has decreased
by 3% after 1000 days. A decrease by 4% would, according to this formula, happen only after
10000 days and a further decrease of B by 1% (to 5% in total) would happen in 105 days –
which is approximately 274 years. The actual k-values of good Alnico-magnets are still
considerably lower, after 10 years typically only 0.1 to 1% is missing. The natural aging thus
does not play any role for static magnetic parameters of pickup-magnets♣. Pickup-guru Bill
Lawrence is of the opinion that Alnico-5 decreases <5% in 100 years [Billlawrence.com].
♣
Effects on the permeability will be discussed in chapter 4.10.

4.5 Magnet Aging 4-33
Temperature changes will result in reversible as well as irreversible changes of flux and
field strength. The reversible changes of approx. ± 0.5% are negligible for typical temperature
changes. Irreversible changes will occur only beyond +500°C. Some authors, however, quote
this limit to be +200°C; every guitarist has to decide whether he considers this as critical.
Load changes in the magnetic circuit (chapter 4.6) and external magnetic fields may,
however, lead to dramatic changes. Since the hysteresis loop has two branches, which can
only be run through in different directions, successive changes of the field strength –∆H and
+∆H will not lead back to the original working point (Fig. 4.13). If, for example, the magnet
is removed from a speaker and afterwards built back in again, the flux density and
subsequently the efficiency will be reduced from that time on. For pickups, however, a
considerable load change (a disconnection of the magnetic circuit) is practically not possible
because of the large air gaps involved. They are exposed only to minor load changes by the
location and direction dependent magnetic field of the earth (approx. 0.5 Oe = 40 A/m) on one
side and by the vibrating string on the other side. Both effects will change the flux density by
less than 1‰ so, practically, not at all. In contrast, the magnetic field changes in electric
motors are much stronger – and do not lose their magnetization, either, do they? Magnets
used in this way are certainly stabilized, though, i.e. they are artificially aged. This procedure
should actually be applied to every permanent magnet.
During stabilization the freshly magnetized permanent magnet will be exposed several times
to a field and/or temperature change with absolute values slightly higher than the future
specifications. Usually between 10 and 20 load cycles are sufficient. The magnet will loose
flux density (e.g. 1 – 5 %), but will become less susceptible to external load changes. After
stabilizing, the reversible permeability (and also the pickup inductance) may slightly increase;
these details will be revisited in magneto-dynamics (chapter 4.10). If one considers small
changes in the lower per cent region, a very fundamental consideration may not be
overlooked: the magnetic parameters may vary considerably more due to the production
procedures. A scatter of ± 10% is not the exception but the rule.
a Fig. 4.13: Hysteresis-curve.

c Decreasing field strength from point a to b, and
b
d subsequently increasing to c, will yield a smaller flux
density, i.e. a is not identical to c. An additional change
in field strength will not lead to b, but rather to point d.
After several cycles, however, a lancet-shaped
H
equilibrium state will be reached that will be located
slightly below the d-c curve. (cf. also Fig. 4.6). The
two straight lines represent curves of equal load
(Chapter 4.6).
___________________
From guitar-literature: a) “Fender-type pickups will noticeably loose magnetic power after 2 years, Gibson-
type ones after 3 years”. This is physically not justified.
b) “The magnetizing values of Alnico-2 are matching that of aged Alnico-5 pretty much”. The forced aging thus
would have led to an extreme demagnetization, refer to Fig. 4.11.
c) "As time goes on, older magnets lose some of their power. The less power the magnets have, the better the
strings can vibrate. So maybe after 30 years, the magnets are at their 'ideal' power, thus producing 'ideal' tone."
Guitar collectors beware: Throw away your Les-Pauls and Nocasters from the fifties now – all mag-power lost!

Storage and handling of magnets needs special expertise. If one recognizes in photos, in a
professional journal, how the pickup guru has a handful of magnets laying in the drawer, one
hopes, of course, that these are non-magnetized blanks, which will be instantly magnetized
(behind the kitchen table also shown?) in a super-strong field. Since, if these were bar
magnets which are stuck together in a jumble and are mixed up daily, one cannot seriously
consider long-term stability and aging (cf. Fig 4.14)
Permanent magnets keep their polarization over a long time, but they are not indestructible.
Extreme temperatures and force or field impacts may weaken the magnetic field permanently.
One should not be worried that a magnet will become weaker when it only falls on a tabletop
but, rather, one should be careful with other ferromagnetic materials and other magnets in its
vicinity. The working point of an unloaded (open) magnet is located in the second quadrant:
negative field strength and positive flux density. If the working point is, however, pushed
beneath the “knee”, the kink of the hysteresis-curve, the original working point will not be
restored again after detachment of the other magnet. McCaig [26] reports on drop impact tests
where an Alcomax-III-Magnet has hit a hard wooden floor from a height of 1 m; the
measured change was much less than production variation (-0.5 %). In contrast, the magnetic
stray field of a second magnet can result in complete demagnetization (-100%).
The following advices may be helpful for the handling of magnets:

• Magnetized permanent magnets should only be shipped with a yoke (keeper).
Cautiously remove the keeper before using.
• Do not press magnets with the same poles to each other.
• If attracting parts need to be separated, do not slide them but rather pull them apart.
• The wall of a cylinder magnet should not come into contact with ferromagnetics.
• SmCo- and NFe-magnets may shatter if they collide.
• The attracting forces of strong magnets may possibly be unexpectedly high, resulting
in crush injuries.
Alnico-5, Flux density in 2mm distance Alnico-5, Flux density in 2mm distance
35 35
mT mT
30 30
north pole
north pole
south pole
25 25
south pole
20 20
15 15
10 10
5 5
0 0
0 10 20 30 40 50 mm 60 0 10 20 30 40 50 mm 60
Fig. 4.14: Magnetic flux density measured at a distance of d=2mm along the long side of two humbucker bar-
magnets from two different manufacturers. The operating point of the humbucker-bar-magnets is at a rather
disadvantageous position and need to be handled with care.

4.6 The Magnetic Circuit 4-35
4.6 The Magnetic Circuit
Magnetic fields permeate the whole space (Chapter 4.1). As they are invisible, one depicts
their distribution with field lines and performs modeling in analogy to flowing (immaterial)
currents. In contrast to electrical field lines, magnetic field lines do not have an origin and an
end. As a rule (for which exceptions are possible), they are closed lines with limited length.
The tangent to a field line is oriented in direction of the flux density propagation; vertical to it
is the penetratedr unit area. The corresponding descriptive
r field quantities are the (magnetic)
field strength H and the (magnetic) flux density B . The field strength is a length specific
quantity (unit A/m) which is determined along its length direction (direction of the flux), the
flux density is an area based quantity (Vs/m²), which is specified for the penetrated area.
r
The line integral
r along a space curve over H yields the magnetomotive force V, the area
integral over B yields the magnetic flux Φ. If the formula symbol V is already used for
volume, the magnetomotive force may also be denoted by Vm. For an infinitesimally small
volume element, the differential material quantity permeability µ depicts the r relationship
r
between the differential field quantities of field strength and flux density: B = µ H . In the
macroscopic region the integral field quantities magnetomotive force and flux are related by
the magnetic resistance Rm:
V = Φ ⋅ Rm = Φ / Λ Hopkinson‘s Law
In contrast to the electric field there is no “magnetic insulator”, even vacuum has a non-zero
permeability (µ 0). The magnetic resistance is also called the reluctance. As the formula
symbol R is also used for the electrical resistance, an index m is added sometimes. The
reciprocal value of the magnetic resistance is the magnetic conductance Λ. Here, no
confusion is possible (the electric conductance is Y), so there is no index m. The magnetic
conductance is also called permeance and, sometimes, the formula symbols P or λ are used
instead of Λ.
A magnetic circuit is defined in analogy to the current circuit, which naturally does not have
to be a circle. In fact, the magnetic flux flowing (“circling”) around the closed field lines is
what is meant. The impetus and cause of the magnetic flux is the magnetic source, e.g. a
current-carrying wire or a permanent magnet. The boundary of the source is sometimes
clearly visible (e.g. the surface of a permanent magnet) but sometimes chosen rather
arbitrarily (e.g. a permanent magnet with pole pieces) or does even not exist: The external
field of a current-carrying wire runs completely in air, which means outside the source. The
part of the flux that is defined as flowing inside the source flows through the source
resistance (source reluctance), the part flowing outside the source flows through the load
resistance (load reluctance). In general, each of these resistances consists of sub-resistances,
only in the simplest cases there is only one source with one source resistance and one load
(Fig. 4.15).
Φ R
i
Θ RL Fig. 4.15: Source with magnetomotive force Θ, source resistance Ri , load
resistance RL and flux Φ (cf. chapter 4.1).

A basic example is the horseshoe magnet (Fig. 4.16). Simplifying, one assumes that the
magnetic flux is restricted to the magnet itself, to both air gaps and to the yoke (with no
leakage flux). The internal, air gap and yoke resistances are successively penetrated by the
same flux and are, therefore, represented by a series connection in the equivalent circuit
diagram. The magnetic potential drops associated with every resistance will, when summed
up, result in the magnetomotive force.
A division of the flux into two parallel resistances is necessary if one would like to consider
that part of the flux, the leakage flux, bypasses the yoke (Fig. 4.17). The leakage flux is, of
course, spatially distributed and the “channeled” representation in the equivalent circuit is a
simplification. If, in addition, one would also like to consider the leakage flux in the magnet,
one has to divide up the source (Fig. 4.18). This is also a simplification and, if necessary,
more than two part sources and more than two part resistances have to be taken into account.
FEM-programs divide the field structure into thousands of small cells (elements) which are
then the basis for the numerical computations of flux and resistance, as well as for more
accurate representations of the field lines.
Φ R RL
i
Θ RA
RL
Fig. 4.16: Horseshoe magnet with yoke and no leakage flux.
Φ R RL
i
Θ RA
RS RL
Fig. 4.17: Horseshoe magnet with yoke and load leakage fluxes (schematic).
Θ2
R i1 R i2 2RL
Θ1 RA
RS1 RS2
Fig. 4.18: Horseshoe magnet with yoke, source and load leakage fluxes (schematic). The series connection of the
both air gap resistances is combined to a resistance 2 RL.

The determination of the elements of the equivalent circuit of Fig. 4.16-18 is complicated and
only approximately possible. For the magnetomotive force, one has to consider the product of
coercivity and magnetic length. The source resistance of the magnet is non-linear; it can be
determined from the hysteresis. The resistance of the air gap is linear, but it has to be
calculated over an inhomogeneous (position dependent) field. The resistance of the
ferromagnetic yoke is non-linear. A solution can only be determined iteratively: The spatial
field distribution is dependent on the non-linear resistances, but their working point, on the
other hand, is field dependent. In particular, it has to be stressed that the, otherwise so
powerful, superposition principle cannot be applied here.
However, Kirchhoff’s mesh rule can be applied, which means in its generalized form: The
flux quantity is equal everywhere in an undivided current circuit; the sum of all potential
drops is zero. When applied to a magnetic circuit this means that, in Fig 4.19, the magnetic
flux is equal everywhere and the sum of the magnetomotive force and magnetic potential
drops is zero: Θ + VM + VL = 0. Here Θ is the magnetomotive force, VM is the magnetic
potential drop inside the magnet and VL is the magnetic potential drop in air. For the sum to be
zero, all arrows have to point into the same direction. Alternatively, there are other arrow
systems, because each of the three circuit elements can be chosen to have a positive or
negative sign. The conventional measurement technique common for permanent magnets
yields an unusual sign convention: Θ = VL – VM. The arrows of Θ and VM,, on the one side,
and VL, on the other side, oppose each other. In addition, Θ and VL are assumed to be
negative. Two possibilities exist for the flux arrows; it has been defined in such a way that
(also unusually) the arrows of Θ and VL and Φ have the same direction inside the source. This
means that, at the air gap resistance RL, the potential and the flux arrows are opposite but at
the magnet resistance RM the direction are the same. Hopkinson’s law is, therefore, written: VL
= – Φ⋅RL , and VM = + Φ⋅RM. As mentioned, this needs getting used to.
VM
Φ Θ = VL − VM
RM
VM = Θ ⋅ RM
Θ RL VL
VL = −Θ ⋅ RL
Fig. 4.19: Horseshoe magnet without a yoke; no source leakage flux (simplified field line representation).
Hopkinson’s law can be represented by a line through origin for the linear air gap resistance;
according to the (arrow direction dependent) minus sign, the potential drop is negative for
positive flux. The magnet resistance (source reluctance) is non-linear and the VM /Φ
relationship is described by the hysteresis curve. As both RM and RL are penetrated by the
same flux, both functional graphs can be drawn in the same figure (Fig. 4.20). The sum of VM
and VL is Θ and this is the distance between the two vertical lines. If one lets the air gap
resistance go to zero, as the limiting case, this will yield the remanence flux density. This
would be the case when using a very high permeability material as a yoke instead of air.
Otherwise, if the air gap resistance tends to infinity the flux density will become zero and one
will get the coercivity point. However, a material with µ = 0 may not be realized: no
“magnetic insulator” exists.

Fig. 4.20: Graphical solution of the non-linear flux

AP
VM VL equation. VL increases proportionally to the left as a
function of Φ (ordinate!) (straight load line, striped
triangle on the right). VM increases to the right with
reference to the ordinate (!) (striped area curve on the
left). The sum of both values represents the
V magnetomotive force Θ at the working point (AP).
Θ
Fig. 4.20 contains two functional graphs, which are also called the working characteristics.
The non-linear B/H or Φ/V relationship is represented by the curved hysteresis line; the linear
air gap resistance is depicted by the load line. The slope of the load line depends on the
individual magnetic field geometry; its intercept with the hysteresis line marks the working
point (AP). It is only possible to approximately calculate RL since the air gap field fills the
entire (infinite) space.
Cylinder shaped magnets are employed for simple pickup constructions, without any pole
pieces (iron parts). Idealized, the magnetic flux will pass through the magnet cylinder in an
axial direction, exits the end face, diverging and flowing through the entire air space before it
re-enters at the other end face. In reality, however, a considerable source-leakage flux exists:
The magnetic flux will also penetrate the lateral cylinder surface which means that the source
must be partitioned (Fig. 4.18). The actual, effective air gap resistance, which is often
depicted in reciprocal form as the permeance, can only be computed by FEM programs with
sufficient precision. Indeed, the literature quotes permeance values [21-26] for some simply
formed bodies, but their precision is only moderate. This is aggravated by the fact that the unit
of permeance can be understood in the American literature only after some deliberation. The
permeance P is the quotient of flux and magnetomotive force; it should have the unit Vs/A =
H or Mx/Gb, respectively. Instead, P is given in the American literature with the dimension of
length, (e.g. cm), which stems from the incorrect definition of µ0. The correct value of µ0 in
the cgs-system is µ0 = 1 G/Oe. Instead, µ0 = 1 is used, with the consequence, that the units of
the derived quantities are also wrong.
In addition to the absolute permeance, a dimensionless standardized permeance is also

defined, which was introduced by Parker [22] as the permeance coefficient p = unit
permeance per centimeter or elsewhere as a dimensionless B/H ratio, again under the
assumption that µ0 = 1. McCaig [26] is more precise and speaks about µ0 = 1, but means the
same. Cedighian [25] again prefers the B/H ratio without µ0 but calls it the demagnetisation
coefficient (see later). The B/H ratio indicates the slope of the load line (load characteristic).
The dimensionless designation B/H = 12, for example, means that the a strength of 500 Oe
corresponds to a flux density of 6 kG. Transformed to MKSA units this would mean that 40
kA/m correspond to 0.6 T. For Alnico magnets B/H-ratios of about 15 are optimal in order to
run the load line through the point of maximum energy density (B/Hmax-point). This would
yield an optimum length/diameter ratio of about 4 for cylinder magnets if one uses the
published permeance diagrams of [e.g. 22, 23, 25, 26,]. In fact, Parker [22] talks about length-
to-area-ratio, but means length-to-diameter.

Many pickup magnets (e.g. Stratocaster) meet this optimum length to diameter ratio quite
precisely, but one should not make a dogma out of it; the precision of the permeance data is
not very high. The magnetic circuit theory, which is based on circuit analogies, is very well
suited to gain insight into the qualitative relations. Nowadays, with FEM computations, more
powerful and precise tools are available for quantitative conclusions.
The term demagnetization needs a further explanation. First of all, it means every process
that shifts the working point away from the remanence point of the hysteresis curve further
down to the left – this is the reason why sometimes instead of hysteresis curve one speaks of
the demagnetization curve. In addition, demagnetization also means the irreversible
destruction of the permanent magnetization, as will occur when the Curie-temperature is
exceeded or under the influence of a strong field (e.g. inside a demagnetization tool).
Demagnetization rarely refers to the aging processes, which are in fact also called aging,
after-effect, losses or similar expressions. Rather unexpected, however, is the description of
magnetomotive force drops at the source resistance as demagnetization: As RM in Fig. 4.19
could not be zero, a positive magnetic voltage drop Φ⋅RM will result for every flux Φ ≠ 0 at
RM, which - in series with a negative (!) magnetomotive force Θ - will decrease the value of
the magnetic voltage drop at the load resistance RL. The field strength H will also be
decreased; the magnetic circuit will thus be “demagnetized”. However, one should not force
the electromagnetic analogies too far, because otherwise one has to speak about “de-
electrification” in conjunction with a charged car battery.
In Fig.4.21 the load lines are given for typical Alnico curves. A flux density of 9 kG = 0.9 T
results for the center of an Alnico-5 cylinder magnet for L/D = 4; if one takes an Alnico-2
magnet instead, B decreases to 5 kG. The standardized permeance values for humbucker
magnets are, with p ≈ 4, surprisingly small; the flux density differences between Alnico-2 and
Alnico-5 for are accordingly minor.
AlNiCo
14
AlNiCo 5
12
Fig. 4.21: Load lines for different
10
standardized permeance values p = 8, 12
p = 16
& 16. The flux densities emerging at the
8
working point are achieved at the neutral
AlNiCo 2
surface, i.e. at the center of the cylinder
kGauss
p = 12
6 magnet. A length/diameter-ratio of
AlNiCo 3
p=8
approx. 2.5 corresponds to p = 8 for
4 cylinder magnets.
Accordingly: p = 12 → L/D = 3.2
and p = 16 → L/D = 4.
p=4
2
Alnico and ceramic magnets have

-700 -600 -500 -400 -300 -200 -100 0
0 somewhat different permeance values
Oersted
[23].

4.7 The Representation of Magnetic Fields
Magneticr fields are spatial vector fields. Alternatively,

r to be more precise: the magnetic flux
density B and the magnetic field strength H are vector quantities of a three-dimensional
field. In a graphical representation the location, as well as the field quantities, have to be
visualized three-dimensionally – impossible on a two-dimensional sheet of paper. Spatial
depth can be affected with perspective charts, but distances and angles are very difficult to
realize correctly. This is the reason, why cross-sections (and, for special cases, curved areas)
are used to draw the field profile. This method is especially suited for parallel-plane or
rotational-symmetric fields. For a parallel-plane field, the coordinates only depend on two
Cartesian coordinates (x, y). The field of an infinitely long, straight current carrying
conductor (Fig. 4.1) is an example if the axis of the conductor is chosen to be the z-direction.
However, the same field can also be considered to be rotationally-symmetric with the axis of
the conductor to be the symmetry line and with cylinder coordinates (r, ϕ) instead of
Cartesian coordinates. Still there remains the problem of representing the three-dimensional
field quantities in this cross-section. Every point in the represented area indeed stands for a
point on the cross-section – in fact no space is left to draw the value and direction of the field
quantity. So, compromises have to be found and less important information has to be omitted
in favor of more important ones. For instance, the field quantity might be depicted only at
discrete locations and not throughout the entire (continuous) area. Alternatively, one codes the
value of the field quantity by an assigned color and forgoes the direction representation.
4.7.1 Field Strength and Flux Density
The following field representations refer to two-dimensional fields. The field quantity is
depicted by an arrow, whose length characterizes the value and whose direction describes the
orientation of the field quantity. Scaling is necessary, e.g. 1 cm =ˆ 1 T for the values. The field
quantity depicted by an arrow is associated with its root point. This can easily lead to
misinterpretations, as the drawing area now has two functions: it represents the position and
also the field quantity. The observer is tempted to establish a local relation between the tip
and root points of the arrow, although only the root point is assigned to a point on the cross-
section. Fig. 4.22 explains this difficulty with the help of an example of rotating arrows:
>
>
< < Fig. 4.22: Velocity vectors, drawn

at two different times t2 > t1. In the
<
<
left picture a connected polygon line

< <
is drawn, however, without physical
meaning.
t1 t2

4.7 Representing Magnetic Fields 4-41
The human brain converts optical information into visual impressions. As a result, the
immense amount of data is reduced considerably and structured by shape laws. Thus, similar
objects in close vicinity are combined into higher-level units with smooth gradients. In the left
picture of Fig. 4.22 the tips of the arrows point to the bases of the next arrows in each case.
The perceived line is, however, irrelevant, as shown by the picture on the right taken at a later
point in time. Fig. 4.23 shows the upper left vectors of the magnetic field strength. An electric
current flows into the image plane at the point [0.5, 0.5], yielding a concentric field. All
arrows are tangents to a concentric array of circles; however, due to the large distances it is
hard to identify these circles. On the upper right picture the conductor has been moved
slightly to the upper right resulting in a giant arrow. This is correct because the base of the
arrow is now located very close to the wire and the field strength is actually very high – but
this representation is not very clear.
5 5
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
-5 -5
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
5 5
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
-5 -5
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
Fig. 4.23: Field strength vectors around a current carrying conductor.
In the lower left picture the density of arrows has been increased in order to increase the
resolution of the circles – not a good idea, either. In the lower right picture all arrows are
drawn with identical length; the visual impression here is the best but indeed the value
information is lost.

If the viewer of the lower right picture of Fig. 4.23 comes to the conclusion that the field is
slightly rotated clockwise because there is a slanting characteristic in every picture frame,
they perceive an optical illusion: the connection of single arrows to contiguous lines is
physically not meaningful for this arrow lattice. Moreover, a rotationally symmetric field
cannot be twisted!
Figure 4.24 depicts the field strength vectors of a two-wire conductor. Here, the clarity can
also be increased considerably if the value depiction is omitted. If there is a possibility for a
color print the value can be shown as colored arrow – on a black and white copy, however,
nothing more is visible.
5 5
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
-5 -5
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
Fig. 4.24: Field strength vectors of a current carrying two-wire conductor (left) and the normalized version
(right). In the right image, asymmetries between the upper and the lower part of the picture can be perceived,
again based on the misinterpretations shown in Fig 4.23.
In addition to the vector characterization, field line images convey a descriptive impression
of the spatial field characteristics. Field lines do not show locations of equal field strength –
they should not be mixed up with the isobars of a weather chart or the contour lines of a map.
r
Rather a curve will become a field line through the field strength vector H defined as tangent
vector at every point of this curve. The direction is defined at every point in space as the
differential quotient of the field strength. If looked upon geometrically, the integration of this
spatial differential equation means the connection of infinitesimally small direction arrows
into integral curves, i.e. into field lines (Chapter 4.1).
The field lines of a current carrying conductor are concentric circles. In this simple case one is
successful with this equation-analytical description. However, with more complicated real
fields, an FEM computation is necessary. Fig. 4.25 depicts the concentric field: for a current
flowing into the image plane the field lines proceed clockwise. The direction of the field
strength vector can easily be deduced as tangent to the field lines; however, its value cannot
be determined from a field line. Yet an estimate can be deduced from the distance of the field
lines: the closer the neighboring field lines, the higher the value of the field. The value is
depicted as gray tone on the right of Fig. 4.25, with limited success. The dynamic range that
can be presented is not sufficient for a linear relationship; for the 1/r decrease a special color
map would have to be defined.

Fig. 4.25: The magnetic field of a current carrying conductor. Left: Field lines. Right: Gray tone coded. The
gray tones produced by printers and copiers are not able to resolve the radial field decrease with sufficient
precision.
An analytical field-description is also possible for an infinitely long two-wire conductor (Fig.
4.26). Assuming opposite current flow, the field lines are eccentric circles with centers
located on the x-axis. This could already be inferred from viewing the arrow-description (Fig.
4.24), but in the line description it is obvious. Between the wires the field strength is the
highest (= maximum line density) and with increasing distance H decreases rapidly. A
contour-plot can be obtained, if all points of equal field strength are connected by lines. This
representation is known from maps: A contour line connects all points of equal altitude.
However, the expression “contour-line” only means that all points with equal functional
values are connected by lines; it has to be specified which value is shown. On an isobar the
pressure is constant, on an isotherm the temperature is constant, for the magnetic field one
could call it iso-field-strength-line, but this expression is not used. Instead, one talks about
curves of equal field strength. If one considers curves of equal flux density, sometimes they
are called iso-flux lines.
Fig. 4.26: Current carrying two-wire conductor. Field lines with direction arrows (left), contour lines of equal
field strength (right). Lines of equal field strength are not equipotential lines (→ Fig. 4.27).

4.7.2 Magnetic Potentials
In chapter 4.2 we introduced the magnetic scalar potential and the magnetic
r vector potential.
The negative gradient of the scalar potential ψ is the field strength: H = − grad ψ . The scalar
potential is a scalar quantity that only has a value and no direction. Hence, its characterization
(and computation) is simpler than that of a vector. The spatial change of the scalar potential
yields the field strength; the line integral over the field strength yields the scalar potential. If
one considers very simple fields e.g., that of the current carrying conductor in Fig. 4.25, the
field strength is constant on every one of the (circular) field lines. ψ increases proportionally
with the angle for rotation with constant angular velocity. Thus, the connecting lines of equal
scalar potential values, the so called equipotential lines, are rays originating from the center
of the conductor outwards. The potential does not change along an equipotential line,
perpendicularly it changes the most – this is the direction that the gradient points in.
Expressed differently, the field lines and equipotential lines cross at an angle of 90°, the field
strength vector is perpendicular to the equipotential line and its value corresponds to the
spatial density of the equipotential lines. The field strength is large in areas where the
equipotential lines are in close proximity.
Fig. 4.27: Field lines and equipotential lines of simple fields: single wire (left); two-wire conductor (right).
The term equipotentialr line depicts curves at which the scalar potential is equal. Considering
the vector potential A , equality is much harder to achieve because, as consequence of its
vector character, three components have to be equal. The vector potential of a two-
dimensional magnetic field, however, has only one component which is normal to the field
plane (Chapter 4.2); for a current carrying conductor it is, therefore, oriented parallel to the
direction of the conductor. If one defines the field plane as the x-y plane, the vector potential
consists only
r of an Az component, with the partial spatial derivative yielding the flux density
vector µ H :
r r r ⎛ ∂Az ∂y ⎞
µ ⋅ H = ∇ × A = rot A = ⎜⎜ ⎟⎟ 2-D vector potential
⎝ − ∂Az ∂x ⎠

Considering the parallel-plane field,

r which is conveniently expressed in Cartesian
coordinates, a flux density vector B can be assigned to every point in space. Both its
components are:
B x = ∂A ∂y , B y = − ∂ A ∂x , A = Az Az = 2-D vector potential
r r
The direction of B coincides with the direction of the flux lines, B is the tangent to the flux
line:
∂A ∂A
dy dx = B y B x → B x ⋅ dy − B y ⋅ dx = 0 → ⋅ dy + ⋅ dx = 0 =ˆ dA
∂y ∂x
The equation on the right depicts a total differential dA which is zero. The total differential
can be interpreted as increase in elevation above the x/y-plane, if a position change of dx or dy
is made. Since, as discussed above, this increase in elevation for the vector potential is always
zero if one runs along a flux line in the x-y plane, the value of A remains unchanged.
Consequently, flux lines are associated with constant A values or the other way round: in the
parallel-plane field, locations of constant vector potential are connected by flux lines. Hence,
for the computation of flux lines (or, after division by µ, of field lines) the vector potential has
to be determined and positions of equal vector potential have to be connected.
Cylindrical coordinates are more appropriate than Cartesian coordinates for rotation-
symmetric fields The computation of the rotation yields somewhat different differential
equations to those for the parallel plane field. The corresponding equation for a radius r,
rotational angle ϕ and axis direction z is:
∂ ( rA) ∂ ( rA)
⋅ dz + = 0, A = Aϕ Aϕ = 2-D vector potential
∂z ∂r
Such a field-symmetry would be adequate for a circular conductor; the vector potential would
be circular as well. Flux lines, however, are not positions of A = const, but rather follow r⋅A =
const.
4.7.3 Spatial Fields
All real fields are three-dimensional and, only for special cases, are they either restricted to
thin layers or are (in symmetric cases) characterized by a plane. Field or flux line projections
onto a plane are not helpful for the case of a general spatial field evolution – as a rule, the
spatial depth cannot be discerned. A last resort would be to define only cross-sections and to
depict the flux density or the field strength by color coding within them. The relationship
between the represented value and the associated color is given by color maps, which e.g.
assign a color gradient from blue to green and yellow to red for increasing functional values.

4.8 Field Distribution in Materials
There is a simple relationship between flux density and field strength in vacuum (or air):
B = µ0H. In ferromagnetic materials things are much more complicated; here, µ is much
bigger than µ0 and depends non-linearly on H. A larger permeability µ means a higher
magnetic conductivity, hence a smaller magnetic resistance. If one assumes that field lines
(flux lines) always “seek” the path of smallest resistance, one gets a descriptive explanation
for the effect of ferromagnetic materials within a field: they “suck,” so to speak, nearby field
lines into themselves and, thus, encounter – despite the somewhat longer path – a smaller
overall resistance. The specialist literature offers yet more comprehensive models, in which
e.g. microscopically circular currents are anticipated as the origin of any kind of magnetic
behavior. However, for simple cases basic models are completely sufficient.
The simplifying differentiation into magnetic and non-magnetic materials is very convenient
for the description of the magnetic behavior of common guitar pickups, because para- and
diamagnetism do not play a role. Air, wood, plastics, lacquers, brass, copper, aluminum,
German silver (nickel silver) (and many more) are, thus, non-magnetic. Steel♣, nickel and
iron are (ferro) magnetic. A copper plate cannot, as some suspect for the Telecaster pickup,
“reflect the magnetic field” because, for stationary fields, there is no difference between air
and copper. In fact, there is no copper plate underneath the Tele-pickup but rather a copper-
plated steel plate, which was also sometimes tin-plated, zinc-plated or something else – that is
a different story. Ignoring the thin copper layer and assuming a ferromagnetic behavior for the
rest means that we have a magnetic material located in the field of one (or several) permanent
magnets. How does this change the field profile and what is the effect of this material?
Ferromagnetics conduct magnetic fluxes better than air, comparable with a drainage pipe in
the soil that should drain its vicinity: “Mühsam sind des Wassers Wege, in der Erde fließt es
träge, doch in Rohren aus Schamott, rauscht's von hinnen mächtig flott♦.” Drainage pipes
have a lower flow resistance than soil, and this is similar for magnetic fields: The flux density
is increased inside the ferromagnetic material and is decreased in the neighboring air. In Fig
4.28 the field-focusing effect of a ferromagnetic material is depicted for three different
permeability values (= conductivities): the higher µ is, the larger is the effect. However, it is
not unlimited: even for a very high magnetic conductivity, the effect on the outermost field
lines is marginal – they are already too far away.
Fig. 4.28: Field line patterns in ferromagnetic materials for stationary parallel plane magnetic fields with no
eddy currents. For simplification, the permeability is chosen as locally constant: µ = 5, 50, 5e6 (left to right).
♣
Some steel grades are non-magnetic ♦German rhyme (no, not by Schiller) – very roughly translated: “Arduous
is the water’s path, in the earth it runs sluggishly, but in pipes made of clay, it very quickly runs away.”

4.8 Field Distribution in Materials 4-47
The efficiency of the field enhancement (or flux enhancement) of the ferromagnetic material
is described by the permeability µ (chapter 4.3). There are already clear differences between
µ = 1 (e.g. air) and µ = 5 (Fig. 4.28), also between µ = 5 and µ = 50. However, with
increasing µ the visible gain becomes smaller and, even with µ = ∞ (compared to µ = 50),
there is no further dramatic change. Fig 4.29 illustrates this with an example: As long as there
remain magnetic resistances (air-gaps) in front of or behind the ferromagnetic material, field
lines escape into the air space.
Fig. 4.29: Electrical current flow through a locally discrete model. A current still flows in the outer branches
even when the central resistance is decreased to zero, (middle picture). The outer branches are only free of
current flow for a complete short circuit (right picture). Analogy: electrical current ↔ magnetic flux.
The above assumed limiting case µ = ∞ in fact is not achievable, but it is helpful for boundary
layer considerations: under this limiting case the field strength within the metal turns to zero,
because B cannot become infinite. From the continuity condition of the boundary-parallel
tangential field strength it can be deduced that the air field strength along the metal boundary
also has to be zero. According to B = µ ⋅H the flux density in air also has to be zero. The
immediate vicinity of the (lateral) boundary layer is, therefore, approximately magnetic field
free. The quantitative value of the flux density, e.g. in Fig 4.28, can be visualized by the
distance between neighboring field lines: The denser they run, the larger is B. Since it was
assumed that air is the medium around the ferromagnetic material in Fig. 4.28, regions of high
flux density (above and below the square) are also regions of high field strength. Accordingly,
beside the square, there are regions of relatively low flux density and field strength. It is
arbitrary which figurative density of field lines (lines per centimeter) one associates with a
particular value of the flux density and it is dependent on the line width, the print quality and
angle resolution of average eyes. With only 1 line per cm one may give away space, with 100
lines per cm one may overburden printer and observer. Scaling hints in the picture (e.g.10
lines/cm =ˆ 1 T) may be helpful, but were left out in the figures because only the relationships
are of interest here. If one splits the field vectors at a material interface layer (wall) into a
wall-parallel (tangential) and a wall-normal component, it follows:
The wall-parallel field strength as well as the wall-normal flux density is always steady. If the
permeability on the one side and the other is different the wall-normal field strength and the
wall-parallel flux density are unsteady.
For the general case, i.e. for varying permeability, every field line will form a kink at the
material interface layer – it is broken, like a ray of light. The larger the difference of both
permeability values, the larger the kink. For the angle (α) to the normal of the interface and
the tangential flux density Bt this yields:
µ 2 ⋅ tanα1 = µ1 ⋅ tanα 2 µ 2 ⋅ Bt1 = µ1 ⋅ Bt 2 Interface conditions [7]
Generally, the permeability of ferromagnetic materials is large and so the field lines will exit
approximately vertical from them, i.e. normal to the wall.

Fig. 4.30: Parallel magnetic conductors; µ = 500 / 5 (left), µ = 500 / 5 (middle), µ = 500 / 50 (right).
Fig. 4.30 shows how two magnetic conductors influence each other. If positioned parallel in
the flux direction, each conductor bends the course of the flux density not only in the
surrounding air, but also in the neighboring magnetic conductor. The material of higher
magnetic conduction (left in the picture) diminishes the magnetic flux in the material with
lower conduction and, thus, forms a sort of shielding. The magnetic conductors act on each
other as flux enhancing when positioned in series with respect to the flux (Fig 4.31): the
magnetic conductor placed in front of the other seems to concentrate the magnetic lines like a
“convergent lens”.
Fig. 4.31: Serial magnetic conductors; µabove = 500, µbelow = 5, 5, 5e6 (left to right).
If two materials with good magnetic conduction are placed together, one has to pay attention
to the air gap in between them, especially in serial configurations: because of the possibly
large differences in permeability, very small gaps may lead to considerable magnetic
resistances. For pot cores, air gap clearances of e.g. 0.1 mm have to be met with a precision of
only a few µm.
A constant permeability was used for the calculations of Figs. 4.28 and 4.31, which may be
permitted for introductive treatments. However, accurate analysis will need precise material
parameters and, consequently, the calculation effort will increase. In fact, there is a mutual
dependency between flux distribution and conductivity: high conductivity will yield high flux
density but this will also (as a function of the declining hysteresis curve) lead to a decreasing
conductivity. This will yield a lower flux density and in turn a lower conductivity – it is a
very complicated coupled system.
Non-linear FEM-models approximate this iteration process by many (sometimes even very
many) calculation steps, which may occupy a PC for half an hour, or maybe even longer,
according to the complexity of the task and the performance of the computer. In addition, not
every material is magnetically isotropic. In fact, for a single crystal, magnetic anisotropy is

4.8 Field Distribution in Materials 4-49
the rule and not the exception. If a magnetic field is applied pointing in one of the preferred
directions, the magnetization energy is lower than for the other directions. Homogenous
spatial distribution of the magnetic crystallite orientation may lead to isotropic (non-
directional) macroscopic behavior, but the crystallite orientations are not always uniformly
distributed. In fact, for grain-oriented Alnico-5 magnets a special anisotropy is perfected, at
pole pieces and/or strings it can happen more or less accidentally as a side-effect. Fig. 4.30
has already shown that field lines in a material may run in very different directions. For
anisotropic substances the material parameter tensors would have to be specified for FEM
calculations, and they are often not available with sufficient precision.
As long as a parallel-plane field is assumed, like in Fig. 4.28 to 4.31, the computational effort
can be limited, to some extent, because one can calculate with plane mesh elements.
Rotational symmetry may also reduce the calculation effort but, for general 3-D models,
things become elaborate. Yet, this is exactly what is necessary for the computation of the
magnetic field of a pickup. All these challenges, such as non-linearities, inhomogeneities and
anisotropies, impede the calculation, but they do not prevent it entirely: The stationary flux
can be determined with sufficient precision. However, the real challenge is yet to come: the
magnetic flux which is important for the induction voltage is the alternating flux, i.e. the time-
dependent change of the steady flux. This part of the flux is only about 1% of the steady flux!
An FEM calculation with “only” 2% accuracy will suddenly lose its original attraction.
On the other hand, if only the basic magnetic flux characteristics need to be described, e.g. for
qualitative considerations, FEM calculations are mostly helpful tools. In Fig. 4.32 a U-shaped
metal bar within a magnetic field is shown. A similar arrangement can be found for
humbucker pickups (Chapter 5.2, 5.7), with a central bar magnet and adjacent pole pieces.
The left picture immediately shows that the humbucking effect might not be sufficient – even
though the magnetic flux representation may quantitatively not be drawn with particular
precision.
Fig. 4.32: U-shaped bar.
It is hardly possible to represent the magnetic fluxes of a guitar pickup with sufficient
precision. The calculation is very elaborate, because one has to deal with non-linear,
inhomogeneous, time-dependent and unsymmetrical fields. Further, measurements are
difficult, because the lateral dimensions are so small: the diameter of the treble strings is only
a few tenths of a millimeter. So, one would need very small Hall-generators which are moved
by micro-manipulators on defined paths to determine the spatial extension of the fields. The
measurements presented in chapter 5 are, presumably, the first of their kind, but surely not the
most precise ones. This makes a basic insight possible, but the differences between similar
pickups cannot be deduced from this data yet – to improve, one would need measurement
tools which are beyond the current college budget.

4.9 Mathematical Field Theory
A field is defined as a special domain in which a physical quantity is depicted as a function of

space and time. Mathematics supports the analytical field description by field and vector
analysis as well as by complex function and transformation theories. The following gives a
short overview of these theoretical field descriptions; detailed information can be drawn from
the books of e.g. Bronstein, Papula, Smirnow, Heinold/Gaede.
The general field theory is powerful, but complicated. Simplifications are possible at the
expense of generality which, in many cases, results in practically no restrictions with respect
to precision. Magnetic fields propagate with the speed of light (~300000 km/s). The distances
of approximately 10 cm relevant for pickups will, thus, be completed in 0.3 ns. Or the other
way round: The settling times of the fields are much shorter than those that are typical for
audio applications (µs – ms); this is why the time needed for the field set-up can be neglected.
Simplifying, one assumes that the field change in the entire area takes place at the same time
(phase-synchronous), that the magnetic field is free of memory effects, and that the current
change of the field state is caused only by the actual excitation, not by spillover from the past.
These kinds of fields are called quasi-static or quasi-stationary. A quasi-static field is in
equilibrium and no energy transport or energy transfer is taking place. A quasi-stationary
field is in a time-independent state of persistence, including energy transport or energy
transfer. The prefix quasi stems from the fact that the time derivative ∂/∂t is virtually zero.
r r
The magnetic field is described by both the vector field quantities B and H . A vector field is
called conservative, if the rotation is zero in the region under consideration; in this case the
line integral is only dependent on the start and end points of the integration line and not on the
integration path itself. Conservative fields are also called potential fields, because they
possess a scalar potential ψ and the vector field quantity is the gradient of the scalar
potential, which is the reason why it is also called a gradient field. The magnetic field is
conservative only in the regions without electrical current flow. In addition, these regions
have to be simply connected
r (Fig. 4.4). In mathematics a gradient field is often described
r
by x = grad ψ , with x being the vector field quantity and ψ being the scalar potential. The
gradient (of the scalar ψ) is a vector pointing into the direction of the largest field
r increase.
However, for the magnetic
r field the respective formula contains a minus sign: H = − grad ψ .
The field strength H of the magnetic field points into the direction of the largest potential
decrease.
r
The spatial characteristics of the field vectors (e.g. H ) can be visualized by field lines. A
curve C(x,y,z) is a field line if the field vector at every point of the line is a tangent vector:
dx H xr = dy H y = dz H z . The solution of this differential equation yields the r spatial vector
field H ( x , y , z ). The field lines do not cross, except for the points at which H is not defined
or becomes zero. For r the magnetic field it can be usefulr to differentiate between field lines
(originating from H ) and flux lines (originating from B ); flux lines sometimes are also called
stream lines. Here, the literature does not have a consistent terminology.

4.9 Mathematical Field Theory 4-51
The flux lines of a magnetic field are (as a rule) closed curves – without a beginning and an
end. This is the origin of the expression source-free field. Obviously a natural impetus exists
which is, however, not the origin of the flux or field lines. Within the framework of vector
analysis, source-free meansrthat the divergence is zero. Such a field sometimes also is called
solenoidal. However,
r divH = 0 is not valid for every magnetic field; if ferromagnetics are
incorporated, divH can be non-zero because of the field-dependent permeability.
The expression flux line implies that something is flowing (streaming). The flux (the stream)
can be objective (e.g. water circuit), or abstract (e.g. magnetic flux). As we just have
explained, the field changes propagate with the speed of light, so it would appear obvious that
the flow-rate corresponds to the speed of light. But one has to distinguish between the flow-
rate and the velocity. If one throws a stone into the water, a circular wave propagates on the
water surface. This, however, doesn’t mean that all the water molecules move outwards with
this velocity. The wave propagation velocity is much higher than the particle velocity which,
for distinction, is called the particle velocity. The entire water surface would elevate and drop
in phase for an infinite wave propagation velocity, without a visible wave pattern. The time
differential of every particle displacement is the particle velocity – which also would be in
phase across the entire water surface. The flux has the dimensions volume/time for flowing
water whereas the area specific flux density has units of length/time, which might be
interpreted as velocity. The dimension of the magnetic flux is Weber or Volt-second, the
dimension of the flux density is T = Wb/m² = Vs/m². Here, it doesn’t make sense to
strenuously search for the dimension m/s. In fact, one might draw analogous conclusions on
the basis of isomorphic (with identical structure) network graphs and the corresponding
equation systems between magnetic circuits and water circuits (and other flow circuits). At
the same time it should be clear, that the magnetic flux density does not correspond to the
constant wave propagation velocity, but rather to the signal-dependent particle velocity. As
the last analogy example we consider the sound field: the velocity of sound is constant at
approximately 340 m/s. The velocity of the air particles is very much smaller – depending on
the excitation. The sound flux q is derived from the particle velocity by multiplication with
the area, not from the velocity of sound.
r
Within the framework of the above explained analogy considerations, B can be interpreted as
a vector flux field, whose flux integral yields the fluxΦ :
r r r r
∫
Φ = B ⋅ dS ∫
Φ q = B ⋅ dS = 0 Flux integral over S
S S
The flux integral is the area integral over the area S. In case S is a closed, enveloping surface,
as it is for the integral on the right, the flux integral describes the source flux emanating from
the enveloping surface. As the magnetic flux lines do not have an origin and an end, it is clear
that the source flux emanating from the volume limited by S must be zero; the field is source-
free or solenoidal. If the envelope surface and the enclosed volume tend to zero, one obtains
the divergence:
r ⎛1 r r⎞
divB = lim ⎜ ∫ B ⋅ dS ⎟ Divergence
⎜
V →0 V
⎟
⎝ S ⎠

The divergence is applied to vectors; the result is a scalar that provides information on the
solenoidal strength at that position. Vector fields whose divergence is zero are called
solenoidal. If one forms the line integral rather than the area integral on a vector field value,
one ends up with the contour integral:
r r
∫
W = F ⋅ ds Work integral along the curve s
s
r
If F depicts a force, then the line integral results in the work performed (= force ⋅ path).
Within the gravitational field the work integral only depends on the start and end points, not
on the path taken. The value is zero in the gravitational field if the contour integral is
computed over a closed line s (starting point = end point),. These fields are called curl-free
fields. In analogy to the divergence, the curl (the rotation) can be depicted as the limiting
case of an infinitesimally small circulation path:
r ⎛1 r r⎞
rot H = lim ⎜ ∫ H ⋅ ds ⎟ Rotation (Curl)
⎜
S →0 S
⎟
⎝ s ⎠
r r r
Here we have introduced the magnetic field strength H instead of the force F whereby H is
not necessarily curl free. s is the closed circulation path and S is the surface limited by s.
Despite the possible confusion,
r the surface area is not depicted by the letter A as common in
mathematics, because A or A denote the vector potential.
The following sentences of the field theory are given here without proof, full particulars can
be found in the cited mathematics literature:
• A solenoidal vector field can always be depicted by the rotation of a vector potential.
• A curl free vector field can always be represented as the gradient of a scalar potential.
The relationships collected in Fig. 4.33 can be understood most simply if one starts with the
flux density. Magnetic fluxes are solenoidal because no magnetic monopole exists and, thus,
the divergence is zero. This statement is depicted as Maxwell’s Third Law, but sometimes r
also as Maxwell’s Fourth Law – physicists do not have a common rdenomination: Since B is
solenoidal one can always specify ar corresponding vector potential A . Using the permeability
µ, one gets from the flux density B to the field strength, the line integral of which will yield
the magnetomotive force V, which is coupled to the flux rΦ through the magnetic conductivity
Λ. The flux is the surface integral over the flux density B . If one integrates the field strength
over a closed loop s one will get the integrated magnetic field Θ. This equals r the current I
enclosed by s, which can be depicted as surface integral of the rcurrent density J over a surface
r
S limited by s. The relationship between the field strength H and the current density J is
described by Maxwell’s First Law: its integral form equates the integrated magnetic field and
the enclosed current (Ampere’s Circuital Law). Its differential form connects the current
density and the rotation of the field strength. In current-free areas the rotation of the field
strength is zero and, thus, the field strength is curl-free and can, consequently, be interpreted
as the gradient field of the scalar potential ψ.

Equi- orthogonal
Flux-
potential-
lines
lines
Complex
A = const
ψ = const
Potential
–ψ + jA/µ
2-D
r r
J =0 r div A =0
Curl- –ψ A Coulomb-
free Gauge
r
grad (−ψ ) rot A
r r r
r rot H r µ ⋅H r div B =0
J H B Solenoidal
r r r r r r r r
∫ J ⋅ dS ∫ H ⋅ ds ∫ H ⋅ ds ∫ B ⋅ dS ψ
r
= scalar potential
A = vector potential
2-D
r = two-dimensional
Jr = electrical current density
Λ ⋅V Φ
H = field strength
I = Θ V µr = permeability
B = flux density
I = el. current
Θ = integrated magnetic field
d d V = magnetomotive force
Λ = magn. Conductivity
dt dt Φ = magn. Flux
U = el. voltage
L = N ⋅Φ I L = inductance
N = number of turns
U
I& N S = area, surface
L = U I& s = line
Fig. 4.33: Formal relationships between the magnetic field values. The complex potential is only defined for 2-
dimensional fields. The scalar potential is only defined for current-free regions. This representation does not
include wave propagation processes.

r
Under
r the assumption that µ is constant, i.e. that it is not position-dependent, H as well as
B are curl-free and solenoidal. For the scalar potential this yields
∂ 2ψ ∂ 2ψ ∂ 2ψ
∆ψ = div (grad (ψ ) ) = + + = 0 Laplace Equation
∂x 2 ∂y 2 ∂z 2
This is the (homogenous) Laplace-Equation. If the Laplace-Operator ∆ is applied to the vector

potential instead of the scalar potential, the result is also zero:
r !
( ) ( )
r r r
∆ A = grad div ( A) − rot rot ( A) = 0 ; div( A) = 0 Coulomb Gauge
r r
Basically, an integration fromr B to A offers an optional integration constant; it is chosen
such that the divergence of A is zero (Coulomb-Gauge of the vector potential).
r Thus, the
above given scalar Laplace-Equations are valid for every vector component of A :
∆ Ax = 0, ∆ A y = 0, ∆ Az = 0 . Vector Laplace Equation
The Laplace-Equation is a very general linear homogenous differential equation. It can be

applied to describe a solenoidal and curl-free field with only a single equation. However,
within current carrying regions, the magnetic field is only solenoidal and not curl-free.
Consequently, a scalar potential does not exist. For this application one employs the
(inhomogeneous) Poisson-Equation:
r
∆A = − µ ⋅ J
r r
( r
⎯→ rot rot( A) = µ ⋅ J .
⎯ ) Poisson Equation
The Laplace and Poisson equations assume a locally constant µ. However, in ferromagnetic
materials the permeability is dependent on the field strength which is dependent on the
position, yielding a location dependent permeability – the Laplace and Poisson equations are,
consequently, not valid without restrictions.
A complex potential F can be defined in the two-dimensional parallel plane field. The
formula symbols f or F are used in the mathematical literature; both characters are
problematic because, in electrical engineering, f stands for frequency and F for force. In order
to distinguish, F will subsequently be used. The components of the complex potential are
differentiable complex functions, which are also called holomorphic, analytical or regular.
Regular potential functions are invariant with respect to conformal mapping and, yielding
simpler options for characterization and computation. The air flow around a complex airfoil
can be projected onto two simple circles; the magnetic field around a cylinder can be mapped
as a superposition of a parallel and a dipole field. Here, the orthogonality between
equipotential and flux lines is conserved, because the conformal projection is isogonal.
However, one may soon notice that the magnetic fields of pickups cannot be described
sufficiently in two dimensions. The theory of the complex potential is only useful as a starting
point to explain the basic relationships. The two cylinder axes are orthogonal for a cylinder
magnet and a string; this situation cannot be described by parallel plane or by rotational
symmetry. Rather, a 3-dimensional coordinate system is necessary – and the complex
potential is not defined within this boundary condition.

The point variable of the complex potential is z = x + jy. Here x and y are the abscissa and the
ordinate of a two-dimensional coordinate system, respectively. The scalar potential ψ is a
regular potential function of z, for which the differentiability and the regularity are given by
the applicability of the Laplace differential equation. The scalar potential is then considered to
be the real part of the regular complex function F or, in other words, the real part functionψ is
supplemented by an imaginary part to become an analytical function. This imaginary part is
defined by ψ in a unique way, because F should not become any complex function but a
regular (=analytical) one. The Cauchy/Riemann differential equations (C/R-DGL) are valid
for regular functions and, as a result, an imaginary part can be derived for every real part. As
for every integration, an additive constant can be freely chosen. It is determined by the
Coulomb Gauge.
The definition of the complex potential is, however, yet not entirely unique: One could
consider ψ also as imaginary part to which a real part is supplemented and also the sign can
be arbitrarily assigned. In general notation the complex potential is:
F(z) = u ( z ) + jv ( z ) Complex potential
The C/R-DGL have to be valid because F should be a regular function in the region under
consideration:
∂u ∂x = ∂v ∂y , ∂v ∂x = − ∂u ∂y Cauchy/Riemann
Mathematics interprets u to be the scalar potential of a curl-free vector field, which can be
depicted as the gradient of u. The gradient is a vector which points in the direction of the
largest field increase. However, physics defines the magnetic scalar potential ψ (also the
electrical scalar potential ϕ) as a vector pointing in the largest field decrease. It is, therefore,
obvious to assign a minus-sign to the real part of F:
F(z) = −ψ ( z ) + jv ( z ) Sign facultative
The gradient of the real part of F is the field strength vector, the components of which can be
translated into the imaginary part of the complex potential with the help of C/R-DGL:
r ⎛ − ∂ψ ∂x ⎞ ⎛ ∂v ∂y ⎞
H = −gradψ = ⎜⎜ ⎟⎟ = ⎜⎜ ⎟⎟ ψ = scalar potential
⎝ − ∂ψ ∂y ⎠ ⎝ − ∂v ∂x ⎠
A corresponding relationship can be derived for the vector potential, whose rotation yields the
flux density. In two dimensions (for which the current description is valid) the vector
potential has only one component A = Az. The index z here refers to the third Cartesian
coordinate of the x-y-z-system; it should not be mixed up with the complex space coordinate
z = x + jy of the two-dimensional field!
r 1 r 1 r 1 ⎛ ∂A ∂y ⎞ r
H = B = rotA = ⎜⎜ ⎟ , ν = A µ , µ ≠ µ (z) . A = vector potential
µ µ µ ⎝ − ∂A ∂x ⎟⎠
It follows for the complex potential of the magnetic field:
F(z) = −ψ (z) + jA(z) µ Complex potential

r
The complex potential is a combination of the scalar potential ψ and the vector potential A ,
which are, in turn, functions of the field strength. The rotation of the field strength is zero in
the curl-free magnetic field:
r
rotH = 0 ⎯
⎯→ ∂H y ∂x − ∂H x ∂y = 0 ⎯
⎯→ ∂H y ∂x = ∂H x ∂y
This is the so-called integration condition of a plane vector field, a necessary and sufficient
condition for the independence of the line integral on the integration path and for the
complete differential dψ :
∂ψ ∂ψ
− dψ = − dx − dy = H x ⋅ dx + H y ⋅ dy Complete differential
∂x ∂y
For dψ = 0 one obtains curves of constant scalar potential ψ = const., the so-called
equipotential lines. The slope dy/dx of these lines correlates with –Hx /Hy, i.e. the equipotential
lines are normal to the direction of the field strength vector.
In a solenoidal magnetic field the divergence of the flux density is zero and, hence, the
divergence of the field strength is zero for a location independent µ:
r
divH = 0 ⎯
⎯→ ∂H x ∂x + ∂H y ∂y = 0 for µ = const , i.e. µ ≠ µ ( z )
The complete differential dA of the vector potential A leads to:
∂A ∂A r r
dA = dx + dy = − µH y ⋅ dx + µH x ⋅ dy with µH = rot A
∂x ∂y
One gets curves of equal vector potential (A = const.) for dA = 0, whose slope dy/dx
corresponds to the slope of the field strength vector: dy/dx = Hy /Hx.
Hence, curves of equal vector potential form the directional field of the field lines, with
curves of equal scalar potential (equipotential lines) being normal to them.
Since the complex potential F is defined as a regular (analytical) function, every regular
projection of F must result in a regular function of z. The complex derivative d/dz is such a
regular projection (satisfying C/R-DGL); if applied on the complex potential this yields:
∂ ∂ ∂ψ ∂ψ
d F ( z ) dz = Re{F }− j Re{F } = − +j = H x − jH y = H *
∂x ∂y ∂x ∂y
The derivative of the complex potential corresponds to the conjugated complex field strength,
whose x and y components are interpreted as the real and imaginary parts. H* is also a regular
function of z. The complex integral over the conjugated field strength is the complex
potential. The additive constants for ψ and A are arbitrary for this integration; they define the
origin of the scalar and vector potentials.

4.10 Magnetodynamics
Dynamics is derived from the Greek word dynamos = force, which is why dictionaries like to
explain this word as “tenet of the force”. However, magneto-dynamics does not primarily deal
with forces but, contrary to magneto-statics, it deals with systems whose signals or system
variables are experiencing variations. In this case it can be changes of location or movement
as well as temporal changes of fields in a stationary system. Ultimately, forces may also be
involved, but they are not in the foreground.
4.10.1 Magnetic Voltage Induction
An electric voltage is induced in a conductor loop, which correlates with the temporal change
of the percolating magnetic flux. This is the basic principle of voltage generation in a
magnetic pickup. The flux is the product (in general the integral) of the flux density and the
area. A change can therefore arise as a change of the flux density and/or area. In a pickup the
conductor is formed by many loops of coiled copper wire. For high quality manufacturing the
single loops are glued together in such a way that the coil area remains constant so that the
only changes are in the flux density. The source and origin of the flux change is the vibrating
string, the effect of which can be described in two different ways: changes of location of the
string alter the magnetic resistance in the magnetic circuit, resulting in flux density changes
(magnetic transducer). Alternatively, the string can be considered as being magnetized by the
pickup; movements of the string are, therefore, relative movements between the string magnet
and the coil (dynamic transducer).
u (t ) = N ⋅ dΦ dt Induction Law
Φ (t)
Fig. 4.34: If the magnetic flux Φ(t) increases with time,
u(t) a positive voltage u(t) will be induced.
In Fig. 4.34 a wire loop is depicted which is penetrated by a flux that increases with time. The
voltage u(t) shown forms as a consequence of the change in flux. Sometimes, the induction
law is also written with a minus sign, depending on the reversed definition of the arrow.
For a guitar pickup there is not only one loop, rather the wire is wound to a coil with N =
5000 – 10000 turns. If one calculates from u~ = 1V back to the magnetic parameters, one will
get a change in flux density of 1 mT (peak value B̂ ) for f = 2 kHz, N = 6200 turns and a
magnet area of 18 mm2. The relative change in the flux density caused by the vibrating string
is, thus, only approximately 1% compared to the static flux density of the permanent magnet
(approx. 100 mT).

More accurate considerations show that it is difficult to compute the induced voltage in the
pickup coil. Not only does the string change its position in space, it also warps (bends) while
it vibrates. For the calculation of the flux change one would have to perform a three-
dimensional field calculation, including the non-linear B/H behavior of magnet and string.
Further, one has to take into account that not every single loop of the coil is penetrated by the
same magnetic flux: the field that is generated by the string diverges and loops that are
located closer to the string will experience a larger flux than those further away. Despite these
difficulties one can, with restrictions and approximations, realize a reasonable agreement
between theory and observation.
Two questions are particularly pertinent when using the Induction Law: How big is the
induced voltage and which type of curve is generated? As the movements of the strings are
non-linear projections of the change of flux, sine-like vibrations will not yield sine-like
voltages. Chapter 5.8 will address pickup-distortion in more detail. We will only investigate
the small signal behavior here. The flux change will, therefore, be assumed to have a single
frequency, be sine-like and be described by the frequency f and by the effective value of the
flux density B. The time differential, thus, simplifies to a product:
u = 2πf ⋅ N ⋅ B ⋅ S Induction Law for a sine-like flux density
where N is the number of turns, B is the effective flux density, and S is the area of the coil; u
is the effective value of the induced voltage. Since all turns of a coil are wound in the same
direction, the voltages induced in every single loop are adding up to the total coil voltage
(typically some 100 mV, max. approx. 5 V).
4.10.2 Self-Induction, Inductance
The voltage that is induced in the coil should be interpreted as a source voltage, not as a
terminal voltage. It evolves, so to say, in the interior of the pickup, just as if an alternate
voltage source is built in there. The voltage measured at the clamps (pickup cable) only equals
the source voltage for the case without load, i.e. in open source mode. Once a load is applied,
“the terminal voltage breaks down”, i.e. it becomes smaller than the source voltage. This
behavior can also be observed in the lighting main: If one switches on a 2 kW furnace, the
light dims. The reason for the voltage decrease is the voltage drop across the internal
resistance, which results in a voltage divider (Fig. 4.35):
U =U ⋅ Z / (Z + Z )
ZQ K Q L Q L
UQ ZL UK
Fig. 4.35: Voltage divider between load impedance ZL
and source impedance Z0.
For the general case the load and source resistors are frequency dependent, which is why one
speaks of load impedance ZL and source impedance ZQ. The induced voltage is UQ and at the
terminals UK builds up. Both voltages are only identical for an infinite load impedance (open
circuit).

Figure 4.35 depicts a circuit mesh in which a current can flow. The required energy is
delivered by the ideal voltage source, which is drawn as a circle. Of course, the energy cannot
emerge from nothing. If the pickup should deliver energy, it must be provided with energy.
And it is: energetically, the source of the pickup terminal voltage is the vibrating string. Its
vibration energy is diminished a little bit by the pickup (and the connected resistances), where
electrical energy is transformed into heat. In other words, the electrical resistances in Fig. 4.35
retroact on the string and change its vibration so that the source voltage is load dependent.
However, as this effect is only very small, it is neglected here. The source voltage UQ is
considered to be imprinted, it equals the induction voltage defined in the preceding chapter.
The load impedance is composed of the coil capacity (see pickup parameters), the guitar
electronics (volume and tone potentiometers), and of the guitar cable and amplifier.
Simplifying, this can be described by a parallel circuit of 300 – 1000 pF and 100 – 350 kΩ.
Further, the source impedance is formed by the coil. Here, one first deals with approximately
1 km of thin enameled copper wire with a DC resistance of approx. 5 -15 kΩ. Another effect
has to be considered for AC, and only AC flows as consequence of the induced alternating
voltage: as already shown in chapter 4.1 every alternating current produces a magnetic field in
its vicinity: an alternating current produces an alternating field, i.e. a field with alternating
polarity. This field also percolates through the pickup coil and induces a voltage. As this field
is produced by the pickup itself (in contrary to the field produced by the string), the voltage
which is built up as a result is called the self-induction voltage. The self-induction voltage
superimposes itself inversely phased on the voltage originating from the string vibration and
weakens it (Lenz’s rule). It is obvious that this voltage superposition cannot be in phase
because otherwise the current would increase with increasing voltage, yielding voltage
increase, yielding current increase … and the system would become unstable. In order to take
into account the self-induction induced voltage decrease, one could include an additional
regulated voltage source in Fig. 4.35, whose source voltage is dependent on the current
flowing in the mesh. However, it is common to instead draw in a component, whose voltage
drop is equal to the self-induction voltage. In the circuit it is symbolized either by a black,
filled rectangle, or by a symbolic representation of the wire coils (Fig. 4.36); this component
is an inductive two-terminal device (= inductor), the unit symbol for an inductor is L.
L RCu
UQ ZL UK
Fig. 4.36: Equivalent circuit of the coil with inductor L
and resistance of the copper wire RCu.
In a Gedankenexperiment (thought experiment) we will now allow a constant current to flow

through the circuit depicted in Fig. 4.36. This constant current can, however, not be generated
by induction as UQ, but it could be fed in at the clamps depicted on the right. As consequence
of this constant current, a constant magnetic field would be generated, whose flux differential
(with time) yields the induced voltage. The differential of a constant is, of course, zero –
consequently, the constant voltage drop at an ideal inductor must be zero too.

However, if an AC current flows, an induced voltage develops, the quantity of which is

depending on the current change:
u (t ) = L ⋅ di(t ) / dt; U = jωL ⋅ I Two terminal equations
A voltage drop u(t), which is proportional to L and to the current change with time, will be
generated in an inductor L in which a current i(t) flows. A representation with rotating
complex pointers is convenient for sinusoidal oscillations. Using this, the time differential
will transform into the factor jω. The imaginary unit j will yield a rotation (phase shift
between voltage and current) of 90°, ω = 2πf is the angular frequency. The derivative d/dt and
multiplication with jω are linear operations; they do not destroy the proportionality between
current and voltage. For direct current the proportionality coefficient between U and I is
called the resistance (U = RI), for alternating currents it is called impedance instead. The
impedance of the inductor is jωL; for direct current it is zero, with increasing frequency it
increases in proportion to the frequency. The impedance of an inductor is a positive imaginary
quantity (precisely: not negative), one could also say: the resistance of an inductor is
positively imaginary.
The unit of the inductance is Henry: 1H = 1Vs/A.
The letter H should not be mixed up with the formula symbol H, which stands for the
magnetic field strength! The quantity of inductance of L can be deduced from the geometry of
the coiled wires. Typical values for a guitar pickup are L = 2 - 10 H.
Equations for the calculation of simple coil inductivities are quoted in every book on
magneto-dynamics. Simple formulas are obtained for the toroidal coil and the long cylinder
coil. However, for the magnetic pickup the conditions are more complicated: the magnetic
field generated by the vibrating string is inhomogeneous, i.e. dependent on the position. Thus,
every turn of the pickup coil will be penetrated by different magnetic fluxes and an analogy,
as in Fig 4.35, with one single voltage source and one single inductor is not possible in the
first instance. This effect is substantial, it cannot simply be ignored: for a Stratocaster pickup
the magnetic fluxes in a turn near the string and in a turn away from the string differ by a
factor of 10 (chapter 5.4.3). For the calculation of the quantity of the induced voltage one has
to perform appropriate suitable averaging. In addition, one has to consider that the magnetic
field will be focused (enhanced) by ferromagnetic materials. The Alnico-magnets placed
inside the coil are ferromagnetic and focus the magnetic flux, which yields a higher
inductance in comparison to a coil which is free of magnetic fields (see chapter 4.10.3).
Since the alternating magnetic flux has its maximum strength in close vicinity to the string,
for efficient conversion (‘loud pickup’) it is recommended to locate the coil as near as
possible to the string. One may find Fender pickup designs with magnets ending right at the
edge of the flange facing the string; they are ‘loud’. However, there are also pickups (the ones
with ‘staggered magnets’) with magnets extending up to 4 mm; they produce (as a rough
approximation) only half of the voltage. Naturally, this rule of thumb presumes that all other
parameters remain constant. In particular, the contour of the coil can deliver an additional
degree of freedom: of equal height or conically wound.

4.10.3 Permeability
In chapter 4.3 the permeability was defined as the specific magnetic conductivity. It is a
material property like, e.g. the electrical conductivity in a current circuit. Air has a relatively
poor magnetic conductivity (µ0 = 1,257 µH/m) and the magnetic resistance of air is relatively
high. The permeability of magnetic materials is often depicted in relation to the permeability
of air♣ as the relative material permeability: µ = µr ⋅µ0. Here, µr is the relative permeability,
also called permeability number.
The permeability number of air equals one, to very good approximation The difference to
vacuum is not significant. In contrast to air, in which the relationship between the magnetic
field strength H and the magnetic flux density B in air is proportional: B = µ⋅H, in
ferromagnetic materials (Alnico, steel, iron, nickel) the permeability number is larger than one
and dependent on the magnetic field strength. The literature for iron claims maximum
permeability numbers of approx. 5000, but also up to 250000, if one deals with purest iron
heat treated in hydrogen. If one would multiply the field strength in the close vicinity of a
pickup of 40 kA/m with 5000⋅µ0 this would yield, in purely mathematical terms, a flux density
of 251T. But this value has of no practical meaning because, on the one hand, generally the
magnetic field strength changes when materials with magnetic conductivity are inserted into
the field, on the other hand the saturation value for the magnetic flux density of iron is
approx. 1 T. Saturation means that the field enhancement in iron reaches a limit that cannot
be exceeded. The origin of a magnetic field is supposed to be a moving charge and, hence, an
electric current. In all magnetically non-neutral materials internal circular currents (elemental
currents) align themselves with respect to external magnetic fields and form an internal
magnetic field, which is superimposed to the external one. The flux density B generated in the
iron can be considered as being the sum of an externally generated component B0 = H⋅µ0 and
an internal field enhancing polarization J. In other words, the component B0 forms
principally and everywhere in the magnetic field. In magnetic materials the polarization J is
added to it.
B = B0 + J ; J = χ m ⋅ µ0 ⋅ H ; χ m = µ r − 1 = relative susceptibility
The relationship between B0 and H is proportional; it is not subjected to saturation. However,

the polarization depends decreasingly on H and tends to a non-exceedable limit = saturation
polarization Jsat (Fig. 4.37).
1.2
T
B
1
J
0.8
0.6
0.4
0.2
Fig. 4.37: Relationship between field strength H, flux
B0
0
density B and polarization J. In this example the
0 20 40 60 80 kA/m 100
saturation polarization is 1 T.
♣
Actually, vacuum is the reference: µ air = 1.0000004 x µ vacuum.

In the pickup an electrical voltage can only be generated (induced) if the magnetic flux
density B changes as a function of time t and, as B depends on H through µ, the magnetic field
strength H also changes. The calculation of such field changes is highly complicated and only
possible as a rough approximation. The string and the pickup-magnet are ferromagnetic in
different ways and, in addition, the screws, pole pieces and plates may possibly influence the
field. In every differential metal volume a different B and, hence, a different µ can dominate.
Adding to the difficulty, small changes of the field ∆H lead to small changes in the flux
density ∆B which cannot be deduced from the slope of the magnetization curves (hysteresis
loops). In Fig.4.38 in the right picture we have depicted a section of the magnetization curve.
If the field strength increases from the working point HA by a small amount ∆H, the
corresponding flux density will adjust to BA + ∆B, which is not located on the large
continuous curve; this one can only be traversed in the direction of the arrow. The quotient
from ∆B / ∆H = µrev is called the reversible permeability. It is smaller than the differential
permeability, which can be viewed the as differential quotient or slope of the hysteresis
(dashed line in the figure). According to [7] the reversible permeability not only depends on
flux density B, according to Gans [Phys. Z., 12/1911], but also on on the polarization J. It also
has – and this complicates the numerical calculations – a tensor character for isotropic media!
B B
∆B
BA
∆H H
HA
Fig. 4.38: Difference between differential and reversible permeability. ∆H and ∆B represent small changes; in
the picture they are considerably exaggerated (cf. Fig. 4.6).
The reversible permeability µrev is maximum for B = 0 and decreases monotonically towards
µ0 with increasing value of the flux density. Consequently, ferromagnetics conduct alternating
magnetic fields much more poorly the higher the static percolating magnetic flux is. Thus, it
is incorrect to simply assign a better magnetic conductivity to a steel-string because steel is
depicted as ferromagnetic material in tables (µ >> 1). Only steel strings without static pre-
magnetization may exhibit permeability numbers which exceed 50. However, as soon as a
string is in the vicinity of a pickup magnet, a considerable static magnetic flux will flow
through it, and the reversible permeability will drop to values only slightly higher than that of
air (Fig. 4.39)
0
Magnet
-1
Fig. 4.39: Approximate progression of
the static magnetic flux for magnet, air
-2
-8 -6 -4 -2 0 2 4 6 mm 8 gap and string (→ Fig. 5.4.8).

In Fig. 4.39 one can recognize, how the static magnetic flux density (depicted as line density)
increases within the string along the string axis. Directly above the magnet axis the axial
string flux density is zero, but already after approx. 6 mm a maximum is reached which
practically means full saturation (1.7 Tesla, steel string with Alnico-5-magnet, chapter 5.4.2).
If now the string loses its good conductivity for alternating magnetic fields already after some
millimeters, they will have no reason to follow the string and, consequently, leave it. Thus,
vibrations of the string only influence the magnetic field in a relatively small volume and the
magnetic window (the magnetic aperture) is relatively short (chapter 5.4.4). The closer the
magnet is located to the string and the stronger it is, the shorter is the magnetic aperture (the
string sampling is more selective) and the corresponding damping of the treble frequencies is
lower.
Gans´ plot String Hysteresis
1 2
T
0.9
1.5
0.8
1
0.7
0.5
rel. Susceptibility
0.6
Polarization
0.5 0
0.4
-0.5
0.3
-1
0.2
-1.5
0.1
0 -2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -4 -2 0 2 4 6 8 10 12 14 16
J / J sat Field strength in kA/m
Fig. 4.40: Gans curve for the reversible susceptibility (left) and string-hysteresis (right).
In Fig. 4.40 we have depicted the “Gans curve“ in the left picture. It describes the
relationship between the normalized polarization up to the saturation polarization Jsat and the
normalized reversible susceptibility up to the starting susceptibility χA. Their formula [21]
reads, in parameter representation:
χ rev ⎛ 1 1 ⎞ J 1
= 3 ⋅ ⎜⎜ 2 − ⎟; = coth( x ) − ; Gans curve
χA ⎝x sinh ( x) ⎟⎠
2
J sat x
In this parameter representation x stands for the parameter (0 … 100), J for the polarization
and χrev for the reversible magnetic susceptibility. The starting susceptibility acts at the
beginning, i.e. at the origin of the new (first) curve. In the right picture of Fig. 4.40 the
hysteresis curve of a steel string is shown, together with little lines indicating the slope of
small signal changes. Fig. 4.41 shows the measured results of a typical steel string.
80 80
χ χ
70 70
60 60
50 50
40 40
30 30
20 20
10 10
0 0
0 0.2 0.4 0.6 0.8 Jrel 1 0 0.2 0.4 0.6 0.8 Jrel 1
Abb. 4.41: Measured reversible susceptibility (left) and theoretical simplification of the Gans curve (right).

4.10.4 Magnetic Losses, magnetic skin-effect
The field enhancing effect of ferromagnetic materials is caused, on the one side by the
internal shift of domain walls (Bloch walls, chapter 4.4.1), and on the other side by initially
randomly oriented elemental magnets that are turned into a common direction by the external
field. A small part of the energy that is necessary for shifting and/or rotating is irreversibly
transferred into heat. The thermal energy that is produced by a kind of micro-friction is “lost”
from the electromagnetic field and this is the reason why one talks about loss of
electromagnetic field energy, or in short about magnetic losses; the designation iron losses
is also common. Losses will decrease the voltage generated in the pickup – an effect which
may mainly affect higher frequencies as brilliance loss.
The two most important loss mechanisms are eddy current losses and hysteresis losses. The
field-energy per volume w can be derived from the relationship between the magnetic field
strength H and magnetic flux density B, as given by the hysteresis curve:
B2
∫
w = HdB
B1
Volume-specific magnetic field energy
If the hysteresis curve is a (curved) line, the magnetic energy would be increased by elevating
the flux density from B1 to B2 and likewise would be diminished by the same value by
decreasing from B2 to B1 – the process would be reversible. However, as each hysteresis loop
consists of two different branches, a complete circuit leading back to the origin does not yield
w = 0 but an energy density which is proportional to the enclosed area and which represents a
measure for the energy loss. For guitar strings, the specific energy loss for a boundary loop
circuit is about 10 µWs/mm³. If one multiplies this value with a 2 cm long 0.7 mm string one
ends up with an energy loss of approximately 77 µWs for the total hysteresis circuit.
However, the working point of a vibrating string does not follow the boundary hysteresis
(from negative saturation to positive saturation and back) but only a small fraction of it. The
fraction of it heavily depends on the distance of the string to the magnet and on the amplitude
of the string deflection. The steady flux is also high in the regions of high alternating flux and
– conservatively estimated – the alternating flux may reach about one tenth of the steady flux.
In addition, if one considers that the small signal changes yield relatively small areas, lancet-
shaped hysteresis loops (also called Rayleigh-Loops), it becomes clear, that the energy losses
caused by the string are of only marginal significance. As an order of magnitude one can
estimate 10 mWs for the string energy and 1 µWs for the iron losses per cycle. If the string
vibrates with 150 Hz with this assumption it will lose 1.5% of its vibration energy, which
would be negligible. A more precise computation of the iron losses would be laborious,
because one has to deal with a three-dimensional inhomogeneous field, for which material
tensor parameters would have to be known. In addition, measurements are difficult because
one has to discriminate from other damping mechanisms. But, even for the case that the above
approximation would be unrealistic and the string energy loss per second would be 26%
instead of 1.5% this would be equal to a level decrease of 1 dB/s – insignificant against other
damping mechanisms. The bottom line of these approximations is, therefore, (without proof):
the hysteresis losses (magnetization-change losses) emerging within a string are negligible.

Other than in the string, hysteresis losses are also possible in the magnet or in nearby
ferromagnetics. Although these losses are not generated within the string, nevertheless the
energy necessary for changing the magnetization of these ferromagnetics has to be delivered
by the vibrating string. The magnetic volume affected by relevant alternating fluxes for single
coils is larger, by more than one order of magnitude, compared to the above considered
volume of the string. However, the relative change in flux density inside the magnet is also
one order of magnitude smaller than inside the string so, on the whole, again an effect of
marginal importance. As long as one does not move a very strong magnet close to the string
(which is in contradiction to a large string displacement) the conclusion is: hysteresis losses
are negligible. This statement is indeed speculative but is supported by measurements which
yield, without doubt, that string vibrations are damped more intensely by the fretting hand of
the guitarist than by the magnetic field of the pickup (chapter 4.11).
A second source for losses are eddy current losses. The induction law discussed in chapter
4.10.1 generates a voltage and a current in the pickup coil but also in every conductor that is
in close vicinity to the pickup. As metals represent electrical resistances, an effective
electrical energy or thermal energy is produced which weakens the magnetic field or the
string vibration. The electrical voltage that causes the eddy current is dependent on the
change of the magnetic field and, therefore, eddy currents do not play any role at low
frequencies. With increasing frequency they become more and more important, however, the
skin effect has also to be taken into account as reverse effect (chapter 5.9.2.2): the magnetic
counter field induced by the current flow forces the current more and more into the boundary
regions and consequently increases the eddy current resistance (chapter 3.3.2)
Eddy current losses cannot generally be neglected, but indeed deteriorate the treble
reproduction of every pickup; not only marginal but possibly by 5 dB and more, if thick low-
ohmic metal plates are employed. One could interpret this fact as a sound characteristic which
is deliberately chosen by the developer, but one should take into account that very dominant
treble can be reduced easily by a potentiometer in parallel to the pickup, which is not possible
the other way round. A pickup with less eddy currents may sound brilliant as well as dull; a
pickup damped by eddy currents may only sound dull♣. Pickups that exhibit small eddy
current losses are the ones with 6 Alnico magnets as sole metal pieces (USA-type
Stratocaster). Soft-magnet pole-pieces with underlying bar magnets increase the eddy
currents, as do tin covers. If one wishes to have a shielding case with small eddy current
losses, thin-walled German silver cases are recommended. One pickup that sounds brilliant,
despite having a metal case, is the Gretsch humbucker.
Eddy currents are not only present in magnets, pole-pieces and shielding cases but are also
possibly in metal support plates and shielding foils. When replacing a plastic by an aluminum
pick-guard one experiences a small treble loss. However, the loss can mostly be avoided by a
small slit, which suppresses the circling eddy currents.
♣
In general, sound filters built into guitar amplifiers cannot compensate for eddy current losses.

In order to obtain quantitative data for eddy current losses, a thin walled measuring coil was
fabricated, into which cylinder-shaped ferromagnetics (∅ = 5mm) could be inserted. The 14
mm wide coil form was wound with 5500 turns of an 80-µ-CuL enameled copper wire (Fig.
4.42). In this representation the logarithmic impedance unit is depicted over the logarithmic
frequency – unusual but convenient. 0 dB = 1 kΩ.
30 te 30
rri
Fe l
ee
0 St
50
o
nic e l
Al Fe Ni
ck
10 kΩ Coil with metal core
uNi 10 kΩ Coil with metal core
20 C 20
er
pp
Co
Impedance / dB
Impedance / dB
oil
rC
Ai
10 10
1 kΩ 1 kΩ
0 0
H H H H
m m m m
5 5 5 5
43 12 43 12
-10 -10
.07 .1 .15 .2 .3 .4 .5 .6 .7 .8.91 1.5 2 3 4 5 6 7 8 910 kHz 20 .07 .1 .15 .2 .3 .4 .5 .6 .7 .8.91 1.5 2 3 4 5 6 7 8 910 kHz 20
Fig. 4.42: Logarithmic impedance of a measuring coil (N = 5500); different core materials.
The wire resistance (630 Ω) is measured without a core (“air coil“) at low frequencies and at
high frequencies the impedance increase proportional to the frequency, the inductance (125
mH). The two terminal device is, thus, perfectly described by an RL-series connection in this
frequency range. Insertion of an Alnico-500-magnet (5x14 mm) increases the inductance by
46%, insertion of a respective ferrite cylinder increases the inductance by a factor of 3.5. In
both cases a frequency-proportional inductance increase happens at high frequencies, so that
only one resistance is necessary in the equivalent circuit: the wire resistance♥. The
inductance increase, however, does not mean that the relative permittivity of ferrite is only 3.5
(or for Alnico is only 1.46). These materials cover only a part of the field space, their
effectiveness is, thus, substantially diminished. As an analogy one might think of two resistors
in series, e.g. 1000 Ω and 10 Ω. The total resistance in this example is 1010 Ω. It decreases to
1001 Ω if the second resistor has only 1 Ω. At a 1V-source a current of approx. 1 mA will
flow even if one will decrease the second resistor even further. This is similar for the
magnetic circuit: the magnetomotive force is dominated by the low-conductive air field. With
a little peculiarity: A change in the magnetic resistance of the core will also affect the shape of
the field lines and, thus, the resistance of the air field.
The reason, why the impedance of Alnico and ferrite can be represented by an ordinary RL-
two terminal device, is quite simple: in addition to the wire resistance no additional loss
resistance has to be taken into account: eddy currents do not yet play a role♣. Ferrites are
sintered out of oxide powder; they have a high electric resistance that prevents eddy currents.
Alnico-alloys are, in comparison to ferrites, already quite good conductors. The fact, that they
exhibit nearly no eddy current losses in the relevant frequency range, arises from their
relatively small permeability (2 – 5). Good conductors with high permeability should, thus,
produce enormous eddy current losses – and this is what they do, to be confirmed by the
following measurements. To achieve this, we have inserted cylindrical cores made of different
materials into the above mentioned coil: steel, nickel, copper (Fig. 4.42 right)
♥
The denomination copper-resistance is disadvantageous here, because copper is also used for the coil core.
♣
The (nonlinear) remagnetization losses are also insignificant.

Copper is diamagnetic, its permeability differs only marginal from µ0. Steel and nickel are
ferromagnetic, their permeability is considerably higher than µ0. Copper is a very good
electrical conductor, Nickel has higher resistivity by a factor of 4, steel by a factor of 10 – 20.
As can be seen clearly by the measured curves (Fig. 4.42, right) the high-frequency
impedance increase with these metal cores is shallower than in air or in ferrites. The origin of
this behavior is the eddy currents, which increasingly force the field lines out of the core
with increasing frequency and, thus, decrease the inductance. Fig.4.43 shows the internal field
distribution for a steel cylinder (∅ = 5 mm, length = 14 mm) for three frequencies as well as
the frequency-dependence of the magnetic impedance, which, as complex unit, consists of a
real part (magnetic resistance) and an imaginary part (eddy current losses). The losses have to
be imaginary because the magnetic resistance is generally defined to be real – different from
electrical networks, where loss-resistances usually are defined to be real. However, these are
only conventions, finally only orthogonality between effective and reactive power is
necessary.
field distribution inside the cylinder relative magnetic impedance of the core
1 10
l
ee
100 Hz
0.9
St
0.8
H 0.7 5
0.6
er
pp
0.5
Co
I 0.4
0.3 2
0.2
1 kHz
0.1
5 kHz
0 r / r0 1
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 1 10 100 1k 10kHz
Fig. 4.43: An axial magnetic field H growing with time generates the eddy current I in the metal cylinder. This
current will produce a circular magnetic field around itself, which is in opposite direction to the generating field
and forces it out of the cylinder. The picture in the middle shows the radial distribution of the axial magnetic
field in a steel cylinder (r0=5mm), on the right the magnetic impedance normalized to low frequencies is
depicted.
The basis of the calculations is Maxwell’s Laws in their differential form under the
simplifying assumption that the electrical conductivity σ and the permeability µ are constant.
For the conductivity this assumption is true, for the permeability actually not: the space and
time-dependent flux-distribution leads to a space and time-dependent µ. The exact calculation
in an anisotropic non-linear medium is, however, so complicated that simplification is
necessary. Both Maxwell Laws now read as:
r r r r
rot H = σ ⋅ E and rot E = − µ ⋅ ∂H / ∂t Differential form of Maxwell’s Law
In cylinder coordinates H only exists in the axial direction, the field strength E exists only in
the circular (azimuthal) direction and the rotation rot can, therefore, be simplified to:
σ ⋅ E = − ∂H / ∂r and − µ ⋅ ∂H / ∂t = E / r + ∂E / ∂r In cylindrical coordinates
Both formulas combined will yield Bessel’s Differential Equation, which can be solved for
harmonic signals with complex units:
∂2H 1 ∂H ∂H ∂2 H 1 ∂H
+ ⋅ = µσ ⋅ → + ⋅ = jωµσ ⋅ H Bessels Diff. Equation
∂r 2 r ∂r ∂t ∂r 2 r ∂r
The time differential operator ∂/∂t has been replaced by jω (see system theory).

The solution of Bessel’s Differential Equation for the radial distribution of the axially
directed magnetic flux density B(r) = µH(r) is:
B(r ) = µc ⋅ J 0 (kr ) with k = (1 − j ) ⋅ πfµσ or k 2 = − jωσµ
Here c is an integration constant and J0 is zero order Bessel’s function of the first degree. The
total magnetic flux that axially passes through the cylinder is given by the area integral over
the cross section with r0 = cylinder radius:
r0 r0 kr0
2πµc
∫ ∫ ∫
Φ = B ⋅ 2πr ⋅dr = 2πµc ⋅ r ⋅ J 0 (kr ) ⋅ dr = 2 ⋅ kr ⋅ J 0 (kr ) ⋅ dkr
k
Total flux
0 0 0
∫
The integration of Bessel’s Function is carried out with x ⋅ J 0 ( x) ⋅ dx = x ⋅ J1 ( x) + C , where J1
is a first order Bessel’s Function of first degree. For the magnetic flux this yields:
2πc
Φ= [kr ⋅ J1 (kr )]0kr0 = j 2πckr0 ⋅ J1 (kr0 ) Total flux
− jωσ ωσ
The magnetic resistance is defined as quotient out of magnetomotive force and flux, the
length-specific magnetic resistance R'm is the quotient out of field strength and flux:
Rm = Vm Φ ; R'm = Rm l = H Φ ; Magnetic resistance
The length-specific magnetic resistance is calculated by dividing the field strength H(r0) by
the flux Φ along the cylinder barrel; the result is complex and is, therefore, called the length-
specific magnetic impedance:
− jωcσ J 0 ( kr0 ) k J ( kr )
Z 'm = H ( r0 ) Φ = ⋅ = ⋅ 0 0 Length-specific impedance
2πckr0 J1 ( kr0 ) 2πr0 µ J1 ( kr0 )
For very low frequencies k tends to zero and, using a series expansion of Bessel’s Function,
one will get as a (real) limit value Z'm → 1 r02πµ , or the inverse of the cylinder cross-section
and the magnetic conductivity. This means, that for low frequencies, there is no field
displacement at all, the flux density is independent of position for the entire cross-section.
However, with increasing frequency the magnetic flux is forced from the center to the
boundary area (casing vicinity), the magnetic resistance (impedance) increases and the
cylinder will become ‘less magnetic’ (see also chapter 5.9.2.4).
The field displacement calculated by Bessel’s functions can qualitatively explain the
impedance/frequency relation shown in Fig. 4.42. However, precise quantitative data are not
possible because the (possibly tensor) magnet data are not known precisely and the metal
cylinder is not percolated exactly axially. In contrast to the pickup calculations for metal
cylinders, a finite elements (FEM) computation would be possible but, also for this case, the
problem of insufficient material data remains.

The (magnetic) field lines of the coil shown in Fig. 4.42 run partly in metal and partly in air.
As already shown in chapter 4.6, one may visualize these magnetic networks by block
diagrams, in analogy to electrical networks, in which resistors are displayed by rectangles.
For the depiction of magnetic loss resistors there are no commonly defined symbols; in the
following they are represented by rectangles enclosing a zigzag line (Fig. 4.44). The
magnetic impedance Zm (the inverse of which is the magnetic admittance Ym ) consists of
real and imaginary parts: Zm = Rm + jXm. It has to be kept in mind that the depicted loss
resistors are imaginary – different from an electrical network. In order to project the networks
onto one another using this analogy one has to define the flux quantity and the potential
(difference) quantity [3]. The flux quantity♣ in electrical networks is the current, in magnetic
networks it is the magnetic flux. The potential quantity is the electric voltage and the
magnetomotive force, respectively. The flux quantity divides at nodes, where Kirchhoff’s first
law (or Maxwell I, respectively) is valid; in analogy, Kirchhoff’s second law is valid for the
potential quantity (or Maxwell II. respectively). Analogies that project flux quantities onto
flux quantities create an isomorphic (equal structure) network; the projection of a flux
quantity onto a potential quantity generates a dual network. Which is valid for
electromagnetic analogies? Using the transformation mechanisms predominant for pickups
(chapter 5) as orientation, one may find a projection of the magnetic flux to the voltage and of
the current to the magnetic field strength – or duality. Written as equations:
U = N ⋅ dΦ dt and
∫
N ⋅ I = H ⋅ ds Electromagnetic transfer formulas [3]
The first formula represents the law of induction, the second the law of magnetomotive force
(Ampere’s law). Consequently, a magnetic series circuit will become a parallel circuit in the
electrical block diagram. The differential occurring in the law of induction will be replaced by
a multiplication with jω for complex (sinusoidal) signals, which yields:
U jω ⋅ N ⋅ Φ
Z el =
I
=
Θ/N
= jω ⋅ N 2 ⋅ Y m ∫
Θ = H ⋅ ds = Magnetic flux
The magnetic and the electric impedance are thus reciprocal: The higher the permeability, the
lower is the magnetic impedance and the higher the electric impedance. A real magnetic
resistor will be projected into an imaginary electrical resistor (inductance, Z = jωL), an
imaginary magnetic (loss-) resistor will be projected into a real electrical resistor. The series
connection of the magnetic real and imaginary part of the impedance Rm + jXm will become
the parallel connection of the electrical resistance R and the inductance L; both are frequency-
dependent. The magnetic in-series connection of the air resistance RmL will become the
parallel lying inductor LL.
Losses
Y el = Z m ( jωN 2 )
Θ
ZZ RmL I R L LL R = ωN 2 X m
Rm jXm
L = N 2 Rm ; LL = N 2 RmL
Fig. 4.44: Dual analogy between the magnetic (left) and the electric network (middle).
ZZ = magnetic (metal-) cylinder-impedance, RmL = magnetic air resistor.
♣
For the electromechanical FI-analogy [3] the electrical flux quantity “current“ will be projected to the
mechanical flux quantity “force“ in equal structure; the FU-analogy projects dually.

The magnetic cylinder impedance ZZ for a metal core with length l and radius r0 located in
the center of the coil is:
k ⋅l J ( kr )
ZZ = ⋅ 0 0 with k = (1 − j ) ⋅ πµσ ⋅ f Magnetic cylinder impedance
2πr0 ⋅ µ J1 ( kr0 )
Here, µ is the (absolute) permeability of the core and σ is the electrical conductivity. Both the
argument (kr0) as well as the resulting Bessel function are complex. Fig. 4.45 depicts the
frequency dependence of the real and the imaginary parts of the magnetic cylinder impedance.
If the metal cylinder were the only magnetic resistor in the (closed) magnetic loop, one would
obtain 5.9 H for the low frequency case, as shown in the Fig. 4.46 (left). However, as for the
cylinder coil under consideration, the field lines close over a long air distance and an air
resistor also has to be taken into account.
Nyquist plot of the cylinder-impedance Real und imaginary part

50 50
6 6
•10 A / Vs •10 A / Vs
20 kHz
40 40
10 kHz
30 30
5 kHz
20 20
10 10
1 kHz
real part
200 Hz
imaginary part
0 6 0
0 10 20 30 40 •10 A / Vs 50 10 100 1k 10kHz
Fig. 4.45: Frequency-dependence of the complex magnetic cylinder impedance ZZ, µr = 110, σ = 5e6 S/m.
If one considers, in a simple magnetic equivalent circuit, a series connection of core and air
resistance (Fig. 4.44), this will reduce the absolute value of the inductance as well as its
frequency dependence (Fig. 4.46). This simple model is well suited as long as the
ferromagnetic metal core can sufficiently focus the field running through the coil. For small
µr, however, a considerable part of the inner magnetic field flows within a kind of hollow
cylinder, i.e. between core and average coil diameter. The magnetic resistance of this hollow
cylinder is located parallel to ZZ in the magnetic block circuit, hence, in the electrical
equivalent circuit in series with the parallel connection of R and L (Fig. 4.47).
electric parameters, without air resistance electric parameters, with air resistance
80 80
kΩ kΩ
70 70
60 6 60
H
50 5 50 0.5
H
40 4 40 0.4
30 3 30 0.3
LΣ →
20 2 20 0.2
←R L→ ←R
10 1 10 0.1
0 0 0 0
10 100 1k 10kHz 10 100 1k 10kHz
Fig. 4.46: Frequency dependence of R and L (left) as well as R und L // LL (right) from Fig. 4.44 (N = 5500).

It is possible to explain every impedance/frequency curve in Fig. 4.42

LHZ LL
with a good precision using this extended equivalent circuit diagram (Fig.
4.47). The magnetic resistance of this “hollow cylinder” is real and it is
mapped onto the inductor LHZ – by definition its electrical impedance is I R L
purely imaginary. The values of R and L are, as explained by Fig. 4.44,

frequency-dependent.
Fig. 4.47: ECD
For the magnetic pickup the magnetic eddy current losses have the following consequences:
1) The pickup resonance is damped not only by the wire resistance of the coil but also the
ferromagnetic core inside the coil. 2) The inductance of the coil is frequency-dependent and
decreases towards higher frequencies. A higher order RL-circuit can be employed in the block
diagram as an alternative to the frequency-dependent inductance (see chapter 5.9.2.3). The
different geometries and the diversity of the material parameters produce different damping
and inductance frequency-dependencies. Using this, the pickup designer can purposely
influence the frequency transfer characteristics.
Fig. 4.48 shows the impedance/frequency curves taken with a measuring coil (N = 5500, Fig.
4.44). The highest inductance is created by the ferrite rod, whose isolated elemental magnets
do not allow eddy currents in this frequency range. The permeability of the humbucker screw,
made of undefined steel, is practically the same for low frequencies (300 Hz), however, due to
high eddy currents, its inductance decreases. The humbucker cylinder (“slug”) has a
somewhat smaller inductance for lower frequencies but also smaller eddy-current losses.
Alnico-magnets are practically free of eddy currents; the magnetically weaker Alnico 2 has a
higher permeability as compared to Alnico 5 (Alnico 500), resulting in a lower pickup
resonance (for otherwise equal parameters).
30 30
r
de
lin 2
cy w o
e nic
sc
r Al 50
0
o
Coil with metal core Coil with metal core nic
20
10 kΩ 20
10 kΩ Al
Impedance / dB
Impedance / dB
10 10
1 kΩ 1 kΩ
0 0
H H H H
m m m m
5 5 5 5
43 12 43 12
-10 -10
.07 .1 .15 .2 .3 .4 .5 .6 .7 .8.91 1.5 2 3 4 5 6 7 8 910 kHz 20 .07 .1 .15 .2 .3 .4 .5 .6 .7 .8.91 1.5 2 3 4 5 6 7 8 910 kHz 20
Fig. 4.48: Impedance/frequency curves, taken with the measuring coil; core dimensions 5x14 mm. “Cylinder”
means the metal cylinders (= slugs) commonly used for humbuckers, “screw” depicts the humbucker screw
(5.9.2.6).
Elaborate details for the construction of single-coil and humbucker pickups, as well as their
technical data are summarized in chapter 5.

4.11 Magnetic Field Forces
Magnetic forces are the most obvious effect of the magnetic field: If one places a
ferromagnetic material into the field of a permanent magnet, it will be drawn towards one of
the magnetic poles. Magnetic forces also act for para- and diamagnetic materials, however,
they are barely detectable. Only when the string is composed of a ferromagnetic material its
vibration can be effectively detected by a magnetic pickup, because only then will the string
significantly change the magnetic flux, so that a sufficiently high voltage is induced. At the
same time, however, the magnetic forces will change the vibration mode of the string – the
generation of a voltage in the pickup is, thus, not free of retroactive effects.
Theoretical physics does not view the electrical and magnetic fields as independent and self-
contained conditions in space, but rather combines both phenomena into a unique field theory.
Forces between stationary charges have to be treated differently from charges in motion: a
relativistic approach is necessary even for small velocities. However, for the pragmatist says
that veritable is only what is appropriate for the act and he gains the winning tender, in this
case. The unique field theory is elegant but, for the present considerations, classical electrical
engineering theory is sufficient and describes – as shown in the following – the force effects
as independent phenomena.
4.11.1 Maxwell’s Force
A ferromagnetic string brought into a magnetic field experiences a magnetic force. Here, it
does not matter whether the string approaches the north or the south Pole; in both cases it will
be attracted. The larger the field strength the larger the attractive force. The force, however,
does not generally act in the direction of the of the field strength – and likewise also not
generally in the opposite direction. Most simply one can interpret the magnetic force as a
surface force that affects the entire surface of the string. Hereby it is understandable why an
iron ball will stay at rest if brought into a homogenous magnetic field: the drag forces acting
on both halves of the ball are balanced and the resulting sum of forces is zero. The fact that an
iron ball in the field of a horseshoe magnet is nevertheless attracted by one of the poles is due
to the inhomogeneity of the field. Only in a very theoretical middle position could an instable
balanced condition be constructed; in every other position one of the two forces dominates
and will accelerate the ball. This is completely different for the Coulomb force (4.11.6): a
charged Styrofoam ball will also be accelerated in a homogenous electrical field.
The magnetic force effect may be very obvious; however, it still remains difficult to
understand its underlying mode of action. Around the beginning of the 19th century magnet
scientists still had the opinion that magnetized bodies would act on each other by a long-
distance effect. This fact, that even an intermediate vacuum could not prevent this long-
distance effect, lead to the conviction that the intermediate space was not involved and that
the magnetic forces would directly act on the bodies without changing the space in between.
The first person to define the concept of the field was Michael Faraday at around 1830,
which changes the space between the bodies by force lines (near-field theory): the space
itself will now become the medium and transmitter of the force.

4.11 Magnetic Field Forces 4-73
James Clerk Maxwell (1831 – 1879) extended Faraday’s ideas into a comprehensive
electromagnetic field theory. A field is assigned to every point in space, which is defined by
its field quantity. For the magnetic field these quantities are the field quantities H and B. The
permanent magnet of the pickup generates an electromagnetic field that acts on other bodies
(e.g. on a string) and produces forces there. However, the now magnetized string will also
produce a field that acts on the permanent magnet. The generation and changes of these fields
happen in the pickup practically without delay.
A mechanical stress state can be assigned to every point in the magnetic space within a very
general force effect theory. The theory of elasticity distinguishes between normal stress
(force perpendicular to the area) and shear stress (force runs within the area). For example, if
a steel cylinder is stressed in the axial direction a tensile stress is generated which will
elongate the cylinder. At the same time it will become a little bit thinner, because a
compressive stress acts in the radial direction (lateral contraction). On the other hand, shear
stresses are generated by shearing-off a whisker, which are also called shear strains.
A general quantity for the characterization of the mechanical state of stress is the stress
tensor: it describes the mechanical stress load that a differential small volume of the string is
exposed to. By integration over all these string volumes (mathematically formulated by
Gauss’s integral law) one will arrive at surface forces that act in a radial direction with respect
to the string’s surface. Since two magnetically very different materials converge at the string-
air interface, one can obtain a very simple approximation for the normal force per area (F/S):
F B2 Vs
= µ 0 = 1,26 ⋅10 − 6 1 VAs = 1 Nm.
S 2µ 0 Am
The area-specific force is proportional to the square of the flux density. For B one has to take
the value that will result at the string surface and not the value which will be measured
without the string. The value at a distance of approx. 2 mm in front of a pickup magnet
without the string will be 20 – 50 mT, while it will be approx. 200 mT including the string.
The string, as a consequence of its good magnetic conductivity, “sucks” the surrounding field
lines as it were and, thus, increases the local flux density. As a rough approximation one will
get 48 mN for the magnetic force for a string area of 3 mm² and 200 mT flux density. A
precise calculation is difficult, because in this case the three-dimensional field distribution in
two non-linear media would have to be determined. In contrast, measurements convey a
sufficiently precise picture: for this a magnetic pickup was moved towards a steel wire (0.7
mm diameter) and the resulting magnetic force was measured (Fig. 4.49). Forces of 10 to 40
mN are detected for common separations – a good confirmation of the theoretical estimation.
For a typical humbucker (e.g. Gibson ‘57-Classic) the forces are smaller.
In comparison to the string tension force (50 – 200 N) the magnetic forces are very small; the
lateral string displacement caused by them is less than 0.1 mm. Nevertheless, the effect of the
magnetic field must not be totally ignored, because its stiffness changes the frequency of the
string. The nearer the string comes to the magnet, the more it is pulled. The differentiation of
this force/distance relation will yield a distance-dependent stiffness of –1 ... –30 N/m; in
contrary to common springs it is negative. The numbers are to be interpreted as guiding value;
the measurement precision is only moderate.

140 0
-10
120
-20
100 -30
-40
80
mN / mm
mN
-50
60
-60
-70
40
-80
20
-90
0 -100
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
mm mm
Fig. 4.49: Magnetic force as function of the clearance. Alnico-5, singlecoil (–––), Gibson-humbucker (---).
In the right picture the differential stiffness is shown as a function of the clearance.
However, the negative magnetic field stiffness not only affects the vibrations normal to the
fretboard; for vibrations parallel to the fretboard the string/magnet distance is practically
constant, the magnetic field stiffness is therefore negligible♣. For vibrations normal to the
fretboard the negative field stiffness generates a decrease of the mechanical stiffness and,
hence, a decrease of the partial tone frequency of the string. The effect is not dramatic, but
audible for strong magnets: if the magnet is moved closer the tone frequency drops. However,
every vibration of the string will occur as spatial wave, not as plain transversal wave. Even if
a plain vibration is prevalent shortly after plucking, mode coupling in the supporting points
and, last but not least, the magnetic field will cause a rotation of the original vibration plane.
The rotation frequency is low (some Hertz) and beat-frequency-like amplitude variations
will evolve in the pickup signal (4.11.3).
In this way the magnet does not just change the tone frequency but also the tone color of the
vibrating string. Whether this is good or bad depends on subjective assessment criteria. Many
guitarists have the opinion that the chorus-like beat frequencies of a Stratocaster belong to the
typical sound of this guitar – as long as they are not too dominant. The assumption (or
“certainty” of expert authors) expressed in several books that “the harmonics are slightly
detuned in comparison with the fundamentals” is incorrect: the fundamental will be detuned
the most. One can ask, since 2001, why it is suddenly referred to in the plural, and who feels
that the announcement: “further handbooks are in preparation” is mere threat. However, this
is just how they are, those string fundamentals.
4.11.2 Field-Induced Deviations of the Tone Frequency
When a magnetic pickup approaches a string, three effects can be anticipated: The tone
frequency decreases, chorus-like beat frequencies evolve and the amplitude changes. The
frequencies, especially the fundamental, will decrease due to the negative field stiffness,
which can be audible for large values. The detuning between the fretboard-normal and
fretboard-parallel vibrations induces beat frequencies; the altered frequency relationship
between the partial tones in the subsequent non-linear systems causes additional partial tones,
which can further increase the chorus impression. True damping, i.e. removal of vibrating
energy, occurs only to a negligible extent. Firstly to the tone frequency:
♣
If the string is located substantially beyond the magnet axis, both vibrating planes are affected.

The resonance frequency of a vibratory mass-spring-system depends on the square root of the
spring stiffness. The stiffness caused by the magnetic field is negative because approaching
the magnet does not need a force in the direction of the movement (as it is for every common
spring) but in the opposite direction: the magnet does not need to be pushed towards the string
with force but, to the contrary, must be held back. The negative stiffness which is acting
thereby decreases the total stiffness of the string and reduces the frequency. The frequency-
dependence of this effect or which partial tones may be affected can be investigated with the
conduction-analogy. Here, the mechanical system is described by an analogous electrical
circuit with the analogies: force/current, velocity/voltage, spring/coil, mass/capacitor [3]. The
principle effect can be investigated for an undamped plain transversal wave which is ideally
reflected at a solid mounting. The string corresponds to an electrical conductor shorted at the
end and whose length is large in comparison with the wavelength [e.g. Meinke]. The
corresponding mechanical input impedance ZE depends on the wave resistance ZW, the
terminating impedance (Ztermination → ∞, because of velocity v = 0), the conductor length l, on
the frequency f and the phase velocity c. All of these system parameters can be attributed to
the mechanical quantities by the analogy-laws: the string tension forceΨ, the string densityρ,
the string length l and the string cross section area A.
ZW ω
ZE = ; β= = 2πf ρA Ψ ; Z W = ρA Ψ Conduction
j ⋅ tan βl c
If one assumes a solid mounting for both ends of the string (Zel = 0 =ˆ Zmech = ∞), the Eigen-
frequencies (partial tone frequencies) of the string are the poles of the tangents-function, i.e. at
integer multiples of the basic frequency fG = c/2l. The reciprocal of the base frequency is the
transit time over 2l, i.e. from the beginning of the conductor to the reflecting end and back. In
order to introduce the influence of the negative field stiffness into the conduction model, one
divides the string into two consecutive conductors: a first conductor of length l1 from the
saddle to the pickup and a second conductor of length l2 from the pickup to the bridge. The
mechanical termination impedance Z of the first conductor is the sum of the input impedance
Z2 of the second conductor and the stiffness impedance ZS: The input impedance Z1 of the first
conductor (viewed from the saddle) is thereby given by:
Z + jZ W ⋅ tan βl 2 s
Z1 = Z = Z2 +ZS ZS =
1 + j Z ZW ⋅ tan βl 2 jω
The Eigen-frequencies are located at the poles of the impedance function, i.e. at Z1 → ∞.
Of course the input-impedance can also be calculated from the location of the bridge with the
same result. For a first check the magnetic field stiffness can be taken to be zero (ZS = 0, Z =
Z2) which indeed gives the partial tone frequencies at multiples of 82,4 Hz using the data of
the E2-string. As can be expected, a spring with stiffness zero cannot produce any changes.
For every stiffness s different from zero the absolute value of ZS will tend to zero with
increasing frequency (ZS = s/jω), from which immediately follows, that the magnetic field
stiffness can only detune low-frequency partials. As the field stiffness is negative the partial
frequencies will decrease.

Figure 4.50 upper left shows the calculated (mechanical) input admittance of an E2-string
(82,4 Hz). The admittance is the inverse of the impedance; its zero points lie at the poles of
ZE. The upper right picture shows the absolute value of the impedance of a 16 cm long part of
the string located between bridge and magnet, as well as the absolute value of a magnetic
field stiffness (–180 N/m). Its value was chosen to be unusually high to depict the effect more
clearly. In the lower left plot the effects of the field spring on the admittance of the entire
string are shown: especially the first and the second partial are detuned. For the calculation a
characteristic wave impedance of 0.7 Ns/m was chosen; the length of the string is 65 cm and
the magnet is located at a distance of 16 cm from the bridge. The frequencies of the partial
tones are at the zero positions of the admittance. No dispersion was modeled.
Input Impedance (Magnitude)

Input Admittance (Magnitude) 10
10
9
9
8
8
7
7
6
6
5
5
4
4
3 3
2 2
1 1
0 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600
Frequency / Hz Frequency / Hz
Input Admittance (Magnitude)

10
9
Fig. 4.50: Frequency dependences of the magnitudes.
Upper left: Admittance magnitude of an E2-string.
8
Upper right: Impedance magnitude of the field
7
stiffness (---), absolute value of the impedance of a
6 16 cm long part (––); both impedances are to be
5 added.
4
Lower left: Admittance magnitude as in the upper left
plot (––); Admittance magnitude including additional
3
field stiffness (---).
2
Zero positions → frequencies of partial tones.
1
0
0 100 200 300 400 500 600
Frequency / Hz
In addition to the calculations measurements of an E2-string are shown in Fig. 4.51. A Fender
E2-string (3150, 1.1 mm diameter) was mounted in an Ovation solid-body guitar (EA-68,
piezo-pickup) and the piezo-signal was analyzed. The magnetic forces were generated by an
18 mm long Alnico-5-Magnet (5 mm diameter) positioned relative to the string at a distance
of 16 cm from the bridge.
Fig. 4.51 shows that a precise frequency analysis is problematic: the resulting detuning is only
several Hertz, so that a frequency resolution smaller than 1 Hertz would be desirable. The
string vibration, however, cannot be considered as stationary within the necessary time
window (more than 1 s). The chosen DFT-windows represent a compromise between time
and frequency resolution (analysis was done with the CORTEX-software Viper).

450
Hz
400
20dB
350
300
Fig. 4.51: Spectrogram (left) and

250
partial tone level progress (right) of
the vibration decay of an E2 string.
200
At 2.2s a magnet was approached to
the vibrating string.
150
The frequencies of the first and
second harmonic decrease from 2.2 s
100 onwards . For the third harmonic one
can mainly detect a level change, the
50 fourth and the fifth harmonic remain
unchanged (holds also for higher
0 harmonics).
0 1 2 3 4 s 5 0 1 2 3 4 s 5
4.11.3 Field-Induced Amplitude Variations
The measurements and conduction model show, concordantly, that the permanent magnet will
detune the lowest harmonics. The detuning will happen mainly for the fretboard-normal
vibrations; field changes parallel to the fretboard and thus parallel to the magnetic pole
surface only develop weakly. For the spatial vibration this means that there are two spatially
orthogonal string vibrations with different frequencies which, after superposition, produce
beat frequency-like level changes. If one denotes the fretboard-normal component with y
and the fretboard-parallel component with x, one gets for the total amplitude ξ in vector-
notation:
r ⎛ xˆ ⋅ cos(ω1t ) ⎞ x̂ = Amplitude of the x-component

ξ = ⎜⎜ ⎟⎟
ŷ = Amplitude of the y-component
⎝ yˆ ⋅ cos(ω 2t + ϕ ) ⎠
For single frequency vibrations ( ω1 = ω 2 ) a point on the string moves according to the
amplitude-relation yˆ / xˆ and the phase shift ϕ along a line in an ellipse or a circle♣ (Lissajous
figures). However, if both frequencies are not equal the figures above alternate with weak
transitions. The time-dependent change of the curve is apparent when one transforms for
small frequency differences:
ω 2t + ϕ = ω1t + ∆ωt + ϕ = ω1t + ϕ (t )
The x as well as the y-vibrations contain ω1t , however, for the y-vibration an additional time-
dependent (slow) phase-shift ϕ(t) exists. A sensor that only detects the vibration exactly
normal to the guitar body will, however, not be affected by the curve changes because ŷ is
time-invariant.
♣
Line and circle are special types of the ellipse.

Real sensors cannot be expected to exhibit such a perfect direction sensibility: common
magnetic pickups are indeed the most sensitive for fretboard-normal vibrations; however, for
fretboard-parallel vibrations the sensitivity will not be zero but approx. 1/10. The voltage
which is generated is, thus, a combination of x and y vibrations which can, for the simplest
case, be depicted as a linear combination:
u (t ) = U (cos(ω 2t + ϕ ) + k ⋅ cos(ω1t ) ) k = relative x-ratio
The commonly known formula for the beat frequency is obtained for k = 1 whereas for k << 1
the signal can be approximately regarded as a mixture of frequency and amplitude
modulation. A cosine-like frequency modulation for a small modulation index can be
represented, to a good approximation, by three spectral-lines [3]:
⎡ m m ⎤
u FM = U ⎢cos(ω 2 t ) − cos ((ω 2 + ∆ω )t ) + cos ((ω 2 − ∆ω )t )⎥
⎣ 2 2 ⎦
If this FM signal should become amplitude modulated, the AM has to be applied to each of
the three spectral components. By neglecting the m²/4-terms (because m << 1), the lines at
ω 2 + ∆ω compensate, while the lines at ω 2 − ∆ω add:
u = U [cos(ω 2t ) + m ⋅ cos((ω 2 − ∆ω )t )] ω 2 − ∆ω = ω1
This signal equates to the above mentioned linear combination for ϕ = 0; corresponding
transformations are possible for other phase shifts. Hence it has been shown that for x and y
vibrations with k = 1 a beat frequency, and for k << 1 a mixture of AM and FM, will result.
This result can also be derived from the projection of the sum of two pointers with different
frequency. If one assumes, for example, that the pickup for the y oscillations is eight times
more sensitive than for the x oscillations (k = 0.125) then, for yˆ = xˆ , the amplitude of the
pickup voltage changes by ±12.5%, or ± 1 dB. The modulation frequency corresponds to the
difference frequency, which is the detuning caused by the magnet (e.g. 1 Hz). The amplitude
relation yˆ = xˆ means that the string vibrates at an angle of 45° with respect to the fretboard.
The amplitude modulation effect will decrease for larger angles (normal to the fretboard ) and
for smaller angles (→ fretboard-parallel) it increases, until at arctan(1/8) = 7° a precise beat
frequency is reached: The level change here is theoretically unlimited.
The linear combination is only a simple model for the description of time-variant level
fluctuations. For the magnetic pickup the induced voltage depends on the non-linear
relationships of the x and y velocities, which will result in additional sum and difference
tones. However, as this will not result in completely different effects, we have dispensed with
a precise investigation. An additional effect, which has also not been taken into consideration
acts at both string mountings (bridge / saddle). Both mountings are idealized as rigid, but
show a direction dependent compliance. As a consequence, the reflection factor has to be
defined including all modes: a pure y vibration will also be reflected, to small extent, in the x
direction and vice versa. For example, if the string is plucked exactly normal to the fretboard,
after a certain time there will be fretboard parallel component which will yield amplitude
variations in the pickup; the magnetic field can enhance or diminish them.

In addition, the (predominantly) fretboard normal magnetic field can induce a rotation of the
vibrating plane when it is not exactly fretboard-normal or fretboard-parallel: for an inclined
vibrating angle the string will experience a stronger pull at the turning point closer to the
magnet than at the turning point further away. By reducing the magnetic force into coplanar
and orthogonal parts one will get an angular force that tries to align the string (along
fretboard-normal direction).
Finally, one has also to consider that the field stiffness is non-linear: the absolute value of the
stiffness increases with decreasing distance. A simulation with a non-linear conduction model
results in weak beat frequencies, even for exact fretboard-normal vibrations, whose fractional
variation amplitude is dependent on the input signal amplitude.
In summary: Already without magnetic fields, sound level deviations are generated that
develop differently for each partial tone. They originate from anisotropic mounting
reflections, i.e. mounting impedances that depend on the oscillation direction and mode-
coupling. The magnetic field detunes the fretboard-normal vibration component which might
enhance or reduce the existing deviations. Non-linearities occurring in the mechanics and
during electro-mechanic transformation will create additional sub-lines in the spectrum so
that, in summary, a complicated level characteristic might develop for each partial tone.
Fig. 4.52 shows the selectively measured characteristic of the partial tones of the E2-string.
The recordings were performed with the built-in piezo pickup without a magnetic field. The
differences between both pictures originate from the plucking technique as well as from the
non-identical guitar positions and, possibly, slightly different guitar tuning and temperature.
In the course of these first orienting investigations it became clear that the guitar should not
be placed somehow on the thigh but must be supported in a defined way. Appropriate frame
conditions are “in vivo” (guitar hanging from the guitar strap, fret-hand defined at the neck),
and “in vitro” (guitar attached at the strap-pin, no damping at the neck).
Fig. 4.52: Decay of the first three partial tone levels after plucking (left) or fretboard-normal excitation pulse
(right) for an E2-string with no magnet and an Ovation EA-68 piezo-pickup. The recordings were taken on
different days.

Fig. 4.53: Decay of the first two partial tone levels after fretboard-normal excitation pulse for an E2-string, with
an Ovation EA-68 piezo-pickup. Continuous lines: without magnet, dashed lines: Alnico-5-magnet in neck
pickup position for a 2.5 mm distance between the string and magnet. Left: first partial, right: second partial.
Figure 4.53 shows the decay of the first two partial tones. The continuous lines were taken
without a magnetic field. The upper curve, with the slowest decrease, shows the level decay
of the undamped neck whereas the lower three continuous curves belong to measurements
that were made with the fret-hand holding the neck in different ways without touching the
strings. The dashed line was taken without neck-damping but with a magnetic field (Alnico-5-
magnet placed 16 cm from the bridge). A strong influence of the fret-hand on the decay-
characteristic (sustain) is observed in both measurements. The hand primarily acts as a
damping resistance removing vibration energy. The level decays linearly with time for the
first partial tone (left picture) without a magnetic field (exponential tension envelope curve),
whereas a slight level oscillation occurs with a magnetic field. The second partial tone is
completely different: There are intense level oscillations without a magnetic field, whereas
there is a nearly oscillation-free decay with a magnetic field. Fig. 4.54 shows similar results
for fretboard-parallel excitations (both with magnetic field).
Fig. 4.54: Decay of the first (continuous) and third (dashed) partial tone after fretboard-parallel excitation using
an Alnico-5-magnet in the neck pickup position. The only difference between both pictures is a slightly different
plucking direction.

4.11.4 Field-Induced Damping
Pickup magnets are rumored to disturb the decay and to deteriorate the sustain of the string.
Indeed, as shown in chapter 4.11.3, the magnetic stiffness can induce changes in the
vibrational parameters; at small magnet distances these changes are also audible. However, an
(ideal) spring is not able to extract energy from an oscillating system. If one pushes an ideal
spring (with positive stiffness) it will store energy. However, after expansion this energy is
returned entirely and without loss. In information technology one speaks of reactive energy
in contrast to active energy, which is “lost” in a frictional resistance. The term “energy loss”
can, of course, not just be viewed globally: in reality energy cannot be lost; however, it will
be transformed irreversibly into thermal energy due to the frictional resistance and is no
longer available for the oscillating system.
However, energetic considerations at a pickup are dangerous and may lead to the wrong
conclusions: a pickup does not transform the vibration energy of a spring into electrical
energy; rather it partakes from one component of the oscillation. Customary pickups mainly
react to fretboard-normal vibrations. If a magnet would rotate the vibration plane of the string
from fretboard-normal to fretboard-parallel this would not affect the vibration energy –
nevertheless the pickup output voltage would decrease. Fortunately, this rotation occurs rather
in the opposite direction (from fretboard-parallel to fretboard-normal); in this case the magnet
will indeed increase the pickup output voltage, however, without increasing the vibrational
energy.
At one place, however, real power is necessary: the voltage delivered by the pickup heats the
ohmic resistors of the electrical circuit, and this real power has to be drawn from the vibration
of the string, because the magnetic pickup is a passive transducer [3]. In addition, the so-
called active pickups are passive with respect to their transformation process; in this case only
the first amplification stage is located at a different place. The ohmic resistors in the
electrical pickup load circuit are the volume potentiometer, the tone potentiometer, the
amplifier input resistance and the coil resistance. The cut-off frequency of the 250 kΩ and 50
nF series connection (tone-pot) is 13 Hz, for higher frequencies the capacitor is approximately
a short circuit. Both potentiometer resistances and the amplifier input resistance are in parallel
for the standard circuit and, therefore, the result for the total resistance is 100 – 200 kΩ.
Further, one has to add the coil resistance (4-15 kΩ). For the pickup/cable resonance one
would have to consider a load transformation for the exact calculation, the following orienting
calculation assumes 100 kΩ for simplicity. According to this calculation a pickup, that
generates 100mV produces a real power of P = U2/R = 0.1 µW. This is very small but must be
viewed relative to the string energy.
The kinetic energy of a mass differential dm is dmv2/2. Here, v is the velocity of the
differential mass. The kinetic energy of the string will be highest at the transit through the
rest position. Integration over the total length of the string (with sinusoidal length-dependent
velocity) yields W = m v2/4, with m = mass of the entire string and v = velocity at rest position.

A typical Stratocaster pickup will generate an effective voltage of U = v⋅0.186 V for a 0.66
mm solid string at a magnet-distance of 2 mm; the velocity v has to be inserted as an effective
value in m/s. However, the velocity is not the one stemming from the energy-formula, rather
is it the velocity of the string above the pickup. For an oscillation of the first partial the
maximum of the velocity is located in the middle of the string (12th fret); above the neck-
pickup v is only 0.69 times as big. In addition, one has to bear in mind that in the energy
formula the amplitude of the velocity is depicted, whereas for the computation of the voltage
the effective value of the velocity is necessary. This will yield for the mechanical energy W
and for the electrical power P:
1 2 U 2 (0.186 ⋅ 0.69 ⋅ v~ ) 2 Pel 3.3 ⋅ 10−7 kg

Wmech = mvˆ Pel = = = ⋅
4 R 100 kΩ Wmech m s
The power P is the quotient out of the energy loss dW and the duration dt (power is energy
over time), the relative energy loss is thus dW/W = Pdt/W. Using 1.78 g for the mass of the
string, the relative energy loss per second is 0.019 %. The time dependent damping value, the
decay-rate D thus will be:
W dB dB 10 ∆W dB ∆W dB
= − 10 lg (1 − ∆W / W )
♣
D = 10 lg ≈ ⋅ = 4.34
W − ∆W s s ln 10 W s W s
Here, ∆W is the energy-loss over 1 s, which will be computed as P ⋅ 1 s. With the above string
one will get a decay rate of 0.0008 dB/s. This is the level decrease resulting from the
electrical damping. Even if one assumes much more efficient pickups with e.g. ten times
larger transformation coefficient, this effect is still minimal and can surely be neglected
compared to other damping mechanisms.
This seems to result in very simple conditions: the magnetic field acts as a spring mainly on
the lower partials and the electrical losses are negligible. However, it is not quite that simple.
The problems are already present in the measurement of the decay curves. It is relatively
simple to choose the appropriate DFT-windows that enable a sufficiently fast and selective
measurement of single partials. For most of the measurements with the CORTEX-software
Viper the 50-dB-Kaiser-Bessel-window with N = 4096 and zero padding = 2 turned out to be
well suited. The decay lines of the partial tones, however, are often curved and, thus, hamper
the modeling. In Fig 4.55 the level trends for the E2 string are depicted, taken without and
with magnetic field. How can the decay (the sustain) be defined with one number? As a level
change within the first second? Every time interval chosen appears to be arbitrary. The
guitarist will not be fussed about the functional decay of the vibrational level, however, for
basic research it will rather play an important role whether the level decay will be caused by
dissipation or by exchange of vibrational energy. For the case of rapid beat frequencies (right
picture) it seems to be relatively simple to derive a time-dependent envelope function from of
the maxima. But if the beat frequency period lasts for ten seconds or longer, the measurement
can become impossible: Until the next beat frequency maximum the oscillation may possibly
have become too small due to other damping mechanisms. It is also not particularly practical
to extract average values from a 30 second level decay because in music tones are seldom
kept over this time period. OK, A Day In The Life. But that was one day. And not guitar but a
piano!
♣
Approximation for ∆W <<«W

Fig. 4.55: Decay of the first partial of the E2-string after different excitations; without magnetic field (left), with
magnetic field (right).
Since it is discussed over and over again in guitarist circles, if and how the pickup-magnet
may dampen the string vibration (and shorten the sustain), we will finally make an attempt at
clarification. For this a guitar (Ovation EA-68) was suspended at the strap-button and the E2-
string (Fender 3150) was reproducibly hit with a pendulum. An Alnico-5 magnet was attached
at a variable distance to the string using a bridge placed over the last fret. The measured signal
was generated by the piezo-pickup (Fig 4.56). Placing the magnet at 2.5 mm distance causes
only a minor level reduction in the 1st harmonic, which hardly stands out. For lower magnet
distances the level loss is considerable. At the 2nd harmonic there is an intense beat frequency
without the magnetic field; a weak magnetic field (b) will increase the level, a strong
magnetic field (c, d) will lead to a substantial level loss. Almost contrary is the 3rd harmonic:
here a weak magnetic field (b) will yield intense beat frequencies. The differences in the
higher harmonics are so small that they are of the same order as the reproducibility.
From these measurements it can be concluded that the magnetic field changes the decay of
the partials; the term dissipation is conditionally justified only for the first two harmonics –
the magnetic field is indeed extracting energy from them in a considerable amount. However,
one has to take into account that, in practice, the neck-pickup-magnet is never brought as
close as to a distance of 1 mm to the string: the string would otherwise impinge on the
magnet. The small distances were chosen for the measurements in order to generate a distinct
effect. Dissipation effects are only clearly visible for this atypical situation (Fig. 4.56 upper
left, curve c and d). During the first seconds the level of the first partial decays much faster
than later on. The cause for the higher fluctuation frequency of 4 Hz, as compared to curve b,
is the higher negative field stiffness, which will lead to larger detuning. The time-dependent
slope of the envelope curve has to be attributed to a non-linear dissipation effect or an
amplitude-dependent damping. This is probably due to hysteresis losses in the string. As the
magnetic field strength in front of the magnetic pole is very inhomogeneous (location-
dependent), the field strength and flux density will change within the string during decay. The
respective reorientation events within the microstructure are partially irreversible.

Fig. 4.56: Decay of the partial tone levels of the E2-string after consistently comparable fretboard-normal
excitations: Without a magnetic field (a), a magnet distance of 2.5 mm (b, dashed), a magnet distance of 1 mm (c)
and a magnet distance of 0.8 mm (d). The higher d-level trend at 1611 Hz has to be attributed to a slightly different
excitation, which shows up only at high frequencies. The results are typical for the guitar under investigation, its
specific mounting and excitation; they should not be generalized for other guitars.

The hysteresis losses are proportional to the frequency, in the first approximation. With every
cycle of the BH-hysteresis loop the magnetic field will lose an amount of energy ∆W; the
higher the frequency the higher the number of cycles per second and the higher the dissipation
losses. For the string, however, one has to consider that higher frequency partial tones are
damped more strongly by other mechanisms and that the strength of the magnetic flux change
depends on the displacement. However, the displacement decreases for higher frequencies.
The lower pictures in Fig. 4.56 clearly show that the magnetic field does not have any effect
in the high frequency range. In addition, for low frequency partials, one should not
overestimate the field-induced dissipations. Finally, for comparison, the influence of the
fretting hand on the decay of the partials is shown (Fig. 4.57, left picture). The upper curve
shows a measurement in which the guitar was suspended from a steel wire at the strap button,
whereas for both of the lower curves the guitar was clamped at the strap button. For the
remaining measurements the fret hand surrounded the neck with different tightness but
without touching the strings. All measurements were done without magnetic fields. One
recognizes that even without magnetic field a variable dissipation is generated – the heel of
the hand touching the neck has to be interpreted as damping resistance. Its energetic (!)
influence on the sustain is considerably larger than that of a common pickup-magnetic-field
(right picture).
Fig. 4.57: Decay of the first partial tone for different manners of hand-damping (left). On the right, with
identical scaling, the influence of an Alnico-5-magnet attached at a distance of 2.5 mm is depicted (neck-
position).
4.11.5 Indirect Effects on Sound
In professional music magazines magnet-characteristics are often published without physical

rationale. It is to be feared that the following citations are pure speculations resulting from
findings after the replacement of an entire pickup. In addition, one can only hope that the
author also did not replace the strings (… the new pickup delivers much more treble …). For
an old Stratocaster pickup, for example, it is impossible to solely change the magnets; the coil
rests directly on the magnets and as soon as one pulls them out one destroys the flimsy coil-
wire. If, however, the whole pickup is replaced by another, the number of turns may change –
and, consequently, it would be incorrect to attribute changes in sound only to the magnet.

In literature very different characteristics are attributed to the magnet material, as can be
seen by the following collection of citations:
a) “For a pickup with the rather weak Alnico-2 magnet the tone seems to virtually die out
after hard plucking. The output signal is not only more quiet but also seems to be less
dynamic and perceptibly compressed in the treble range – which is actually appreciated
by many guitarists.”
b) “As the magnetic field of an Alnico-II magnet is somewhat weaker than that of a common
Strat-pickup (Alnico-V), the string vibration decays more open and more natural. The
result is an improvement in sustain.”
c) “Alnico-5: Strong and clear sound.”
d) “Alnico-5: Fast responsiveness and slightly undifferentiated reproduction.”
e) “The stronger the magnet, the more treble.”
f) “As time goes on, older magnets lose some of their power. The less power the magnets
have, the better the strings can vibrate. So maybe after 30 years, the magnets are at their
'ideal' power, thus producing 'ideal' tone.” ☺
One might add: “If someone has some Les-Pauls lying around that are older than 30 years –
throw them away! Especially for the 50’s Les-Pauls the magnets are completely shot, all
power lost, get rid of them. The author will accept these guitars for research sake, at a small
waste disposal charge.”
Still, back to the physics. The pickup-magnet is part of a mechanic-electric transducer and as
such it influences both the mechanical as well as the electrical partial system. Mechanically
the magnet retroacts on the vibrational characteristics of the string; the result can be chorus-
like beat frequencies and – to a minor extent – dissipation. The electrical effect of the magnet
does not really belong in chapter 4.11 because the forces or mechanical effects are described
there. The following listing is, therefore, only a precis: The reversible permeability of the
magnet influences the inductance of the pickup coil and, consequently, the resonance
frequency. If the resonance frequency is shifted, partials with different decays may influence
the sound and the perceived “sustain”; however, this should not be mixed up with a more
openly vibrating string – changes in the cable capacity would have a similar effect. Eddy
currents within the magnet influence the resonance quality factor (Alnico conducts, ferrite is
an insulator). Stronger magnets may increase the output voltage of the pickup and overdrive
the amplifier in a different way; this may also change the sound and perceived sustain – as
well as by changing the input gain. A replacement of the magnets may also change the
aperture because the spatial flux distribution may change as a function of the (non-linear)
string saturation and because the anisotropy of the new magnets may be different from that of
the old ones.
The magnetic material can, thus, indeed influence the (“electrical”) sound of the guitar. A
hindrance to the free string vibration, however, is not to be expected if the string/magnet-
distance is chosen properly.

4.11.6 Coulomb-Force
An electric field with the field strength E = U/d is generated between two electrodes at
different potentials. Here U is the potential difference, also depicted as a voltage (or voltage
difference, voltage drop) and d is the distance of the electrodes. 100 V at a distance of 10 mm
yields E = 10 kV/m. If one inserts an electrical charge q into this electrical field, an
electrostatic force F is generated which is called Coulomb-force, after its discoverer (Charles
Augustin de Coulomb, 1736 – 1806).
r r
F = q⋅E Coulomb-force
The coulomb-force does not play any role for guitar pickups; it may lead, however, to
misinterpretations: other than for magnetic forces the coulomb-force also “acts” within the
homogenous field. While a positively charged Styrofoam ball between two parallel electrodes
is drawn to the cathode (negative electrode) an iron ball between two parallel poles of a
permanent magnet will rest (more precise: 4.11.1). Indeed, within the magnetic field there are
also attractive forces but they are balanced in this idealized example. The Coulomb-force is
only mentioned here to point out its differentness. Analogy-considerations between electric
and magnetic fields have model limits that have to be observed.
4.11.7 Lorentz-Force
With the Lorentz-force (Hendrik Antoon Lorentz, 1853 – 1928) we will explain another force
that has no direct importance for the magnetic pickup (but indeed for the dynamic
loudspeaker). Again, we want to eliminate misinterpretations. A force F acts on a conductor
of length l carrying a current I when the conductor is carried into a magnetic field with flux
density B. F is oriented normal to the plane defined by I and B. If I is directed parallel to B
then F = 0. In vector notation one will get the vector product (×) :
r r r
F =l⋅I ×B Lorentz-force
If one points with the thumb of the right hand into the direction of the technical current flow
(from plus to minus) and with the forefinger into the direction of the magnetic flux, the
middle finger will point into the direction of the force (right-hand rule). The value of the force
is given by the product l ⋅ I ⋅ B ⋅ sin α , where α is the angle between the current and field
directions. For the magnetic pickup the Lorentz-force, as given in the above form, does not
play any role. A small alternating current does in fact flow through the coil which, however,
with a value of 10 µA, will not exert any substantial force on it. A retroactive effect on the
vibrating string is described by the Maxwell and not by the Lorentz-force, because the string
is not carrying a current. If the string would be conductively suspended one could hypothesize
an induced current in the neighboring string – however, the effect of its force would be
negligible.

4.12 Magnetic Quantities and Units
The literature on magnetic fields refers to two different unit-systems: The MKSA-system as
proposed by Giorgi and the CGSA-System.
The MKSA-system emanates from the four basic units Meter, Kilogram, Second and Ampere
(SI-units, Système International). All other units are derived from them and occasionally
linked with the names of outstanding scientists:
1N = 1 Newton = 1 kg m / s2 1J = 1 Joule = 1 N m
1W = 1 Watt = 1 N m / s = 1 VA 1 Wb = 1 Weber = 1 V s
1T = 1 Tesla = 1 Wb / m2 1V = 1 Volt = 1 m2 kg /(A s3)
The CGSA-system uses the four basic units Centimeter, Gram, Second and Ampere and
derives further units from them:
1 dyn = 1 g cm / s2 1 erg = 1 dyn cm
1 Gb = 1 Gilbert = 1 Oe cm 1 Oe = 1 Oerstedt = 1 Gb / cm
1 Mx = 1 Maxwell = 1 G cm2 1G = 1 Gauß = 1 Mx / cm2
The following table enables the conversion between both systems:
B Flux Density T = V s / m2 1 G = 10-4 T

Induction
H Magn. Field Strength A/m 1 Oe = 1000 / 4π ⋅ A/m
= 79.577 A/m
BH Specific Energy W s / m3 1 MGOe = 7.9577 kJ/m3
Φ Magn. Flux Wb = V s 1 Mx = 10-8 V s
Θ Amperes A 1 Gb = 10 A / 4π = 0.79577 A
F Force N = kg m / s2 1 dyn = 10-5 N
P Power W = VA = N m / s 1 erg / s = 10-7 W
E Energy J = Nm = Ws 1 erg = 10-7 J
Rm Magn. Resistance 1 / H = A / (V s ) 1 Gb/Mx = 7.9577·107 1/H

Reluctance
Λ Magn. Conductivity H = Henry = V s / A 1 Mx/Gb = 1.2566·10-8 H

Permeance Instead of H also Hy for Henry
µ0 Abs. Permeability = 4π·10-7 H / m = 1 G / Oe

of the Vacuum
4π = 12.566; 10 / 4π = 0.79577.

5. Magnetic Pickups
In a magnetic pickup (TA) the vibrating string produces an alternating magnetic field which
generates a voltage in a wire winding (coil). The string in itself in its original state is not
magnetized; the magnetization is derived from a permanent magnet installed at a small
distance under the string. Consequently, the pickup consists of a permanent magnet and a coil
plus some housing components, which keep everything in place. Sometimes additional metal
parts are included to guide the magnetic field.
The magnetic pickup belongs in the category of passive magnetic transducers [3] and uses the
electromagnetic conversion principle. The vibrating string changes the magnetic reluctance
resistance in the permanent magnetic circuit, and due to temporary magnetic flux changes in
the coil an electric voltage is induced. The magneto-electric conversion must not be confused
with the electro-dynamic conversion – in the latter a voltage is induced in an electric con-
ductor moving in a magnetic field. Examples for electro-dynamic transducers are the dynamic
loudspeaker and the dynamic microphone. For both it is the coil which moves relative to the
magnet. In a guitar pickup coil and magnet are fixed relative to each other. Although a minute
induction voltage is generated in the moving steel string this effect is not exploited.
5.1 Single-coil pickups
Close to the bridge, the six strings of the guitar have a distance relative to each other about 1
cm. In order to generate the loudest possible signal, every string has to be subjected to a
strong magnetic field. For many pickups, this is achieved by the use of six cylindrical
permanent magnets positioned in parallel and having a diameter of about 5 mm and a length
of about 1 cm. They are oriented all in the same way i.e. such that all north poles point in the
same direction. The magnets are stuck in a bobbin for the coil wire. The coil is protected
against damage by insulating tape or by a proper housing. Two to four screws keep the pickup
at a short distance below the strings. Most electric guitars have two or three pickups; one or
four pickups are more rare. A special design form is the twin-coil humbucking pickup. For
these pickups two coils are positioned side by side in the same housing. This design reduces
the sensitivity against external noise.
Instead of the six individual magnets, alternatively a bar magnet positioned below the coil
may also be used. For improved field guidance six cylindrical iron slugs are inserted through
the coil in this case. On their lower side, these slugs (also called pole-pieces) touch the
magnet, or they are held in a metal bar which touches the magnet. Often these pole-pieces are
designed as screws such that the volume of each string can individually be adjusted. Over
time, different pickup designs came into existence – the most important ones are being
compiled at the end of the chapter
© M. Zollner 2002 Translation into English by Tilmann Zwicker

5-2 5Magnetic pickups
To shield the pickup against electrostatic interference, metal covers are sometimes installed
over the pickup coil. In practice, the effect of this shield is rather modest because the main
interferences are not due to electrostatic but due to magnetic interference fields (such as they
are e.g. generated by transformers). Against magnetic fields a pickup should not be shielded
due to its working principle. The string vibrations are generating also a magnetic field and it is
this field which the pickup needs to sense - a cover made out of magnetic material
consequently is ruled out. Moreover, even shielding covers made of un-magnetic material
(e.g. brass or nickel silver) can have undesirable effects on the magnetic field because eddy
currents are generated within the material which themselves again generate magnetic fields.
Many guitar players therefore remove the pickup covers and achieve a small change in sound:
the pickup resonance emerges a bit more strongly which usually which often is perceived as
sounding better
The magnetic guitar pickup has a long history. In 1831, the English physicist MICHAEL
FARADAY (1791 – 1867) made the fundamental discovery that an electric current is flowing
in a closed conductor loop penetrated by a magnetic field of changing strength. At the same
time – but independently – the American physicist JOSEPH HENRY (1797 – 1878) arrived at
similar conclusions. The quantitative correspondences between the changing magnetic flux
density and the voltage induced by it are described by the induction law (see chapter 4.10)
which is called the FARADAY-HENRY-law after their discoverers. About 100 years after its
discovery this law yielded the basis for the mechano-electric transduction of the sounds of the
guitar which up to that time were rather soft in nature: the electromagnetic pickup emerged.
Who in fact built the very first magnetic guitar pickup cannot be established with absolute
certainty. DeArmond, Rowe and Beauchamp are often mentioned, and likewise manufacturers
such as Rickenbacker, Gibson, Epiphone, Gretsch. National – and Fender, of course.
Leo Fender facilitated the commercial breakthrough of the solid-body guitar. Teaming up
with George Fullerton, he developed the prototype for an electric guitar of solid wood in
1949. This instrument was introduced to the marked in spring of 1950 under the name
Esquire. In autumn of 1950, the 2-pickup Broadcaster followed, being renamed Telecaster
shortly thereafter. Leo Fender's original guitar is seen as the archetype of all solid guitars,
even though Lester Polfuss, better known under his pseudonym Les Paul, had already been
working on a similar concept for more than 10 years. However, his ideas – picked up by
Gibson – made it into production only by 1952.
The first Fender guitars were fitted with simple single-coil pickups - a tradition which is
retained to this day. Leo Fender used an individual cylindrical magnet for each string,
according to his own words this was to minimize the interaction between neighboring strings.
Together with flanges pressed on to them, the 6 magnets form the coil carrier (bobbin) around
which very thin enameled copper wire is wound. The magnet diameter is 3/16" (approx. 4,8
mm); the length of the magnet varied over the years (and across various Fender guitar models)
from 12 to 19 mm. The magnets consist almost exclusively from Alnico-5 (also known as
Alnico-V) which is a magnetic alloy developed in the 1940's. The flanges were first made of
vulcanized fiber of approx. 2mm thickness; this is a high-strength, horn-like material. The
color and thickness of the flanges varied over the years, and from 1980, injection-molded
bobbins were also used. The diameter of the magnet wire wound around the 6 magnets is
measured according to the American Wire Gauge: most pickup coils are wound with AWG
#42 but in some cases the thinner AWD-#43 wire is used (see chapter 5.5.).
Translation into English by Tilmann Zwicker © M. Zollner 2002

5.1 Single-coil pickups 5-3
Fig. 5.1.1 shows the components of a Fender

single-coil pickup, in this case a Stratocaster
pickup recognizable from the plastic cover.
The characteristic feature is the group of 6
cylindrical magnets onto which the winding is
directly placed. The two Telecaster pickups
are structured in a similar way although there
are some minor different details. Using the
same basic construction principle, the
Jazzmaster pickup was developed in 1957 - it
however clearly departs from its predecessors
in its dimensions. Leo Fender sought a
different sound and widened the coil from 12
to 35 mm, while at the same time reducing the
pickup length. “The Jazzmaster pickup wasn‘t
so deep, and it was wider, thinner, more
spaced out. See, the more spaced out the coil
is – the wider the spectrum under the string –
the warmer the tone. But a broad spectrum of
tone places a lot bigger demand on the amp,
and the earlier tube amps we had were kind of
limited in the amount of power they could
handle. [Wheeler].“
Fig. 5.1.1: Components of a Fender-
Stratocaster-Pickup [Duchossoir].
From the point of view of today's systems theory, Leo Fenders above explanation is not
comprehensible. One could surmise that the thinking then was to sample a longer part of the
string vibration via a wider coil, i.e. longer magnetic window (the magnetic aperture) was
desired. However, as will be shown by the analysis in Chapter 5.4.4, the length of the aperture
is in practice only dependent on the diameter of the magnet irrespective of the coil. Fender's
referring to spectrum also remains unclear: surmising that he desired a larger window-length,
one would expect a narrower bandwidth since time and frequency have a reciprocal
relationship. Conversely, Leo Fender talks about a wider spectrum to which he attributes a
warmer sound. Again, this does not fit: warmer sounds result from attenuating the treble, i.e.
are connected to reduced bandwidth. Bandwidth reduction, however, cannot be what Fender
meant, either, because his statement that a broadband signal challenges the amplifier more is
correct. It seems wise not to expect a lot of theory behind the first pickups: the systems theory
was still the new kid on the block in those days, and the development objectives were not
governed by science but by an empirical approach and sales reports.
Fig. 5.1.2: Cross-sections of Fender pickups: Stratocaster, Telecaster (Bridge), Jazzmaster.

In Fig. 5.1.2 shows cross-sections of Fender pickups. The heavy line at the upper border
marks the position of the string while the cross-section of the coil is represented by the
crosshatched area. Length and protrusion of the magnets changed several times for the
Stratocaster and Telecaster – only the Jazzmaster-pickup with its relatively short lifespan
retained its geometry.
The magnets of the first pickups were mounted flush with the upper flange (the one closer tot
he strings), but as early as approx. 1954 staggered pole-pieces were introduced i.e. the
magnets were protruding unevenly from the upper flange. This allows for compensating for
different loudness of the individual strings – but NOT by the musician! Recommendations to
shift the magnets with light hammer strokes such that all strings have equal loudness may
meet with a rather unexpected success. If in this process the coil wire directly sitting on the
magnet is torn, indeed the loudness of all strings will be equal: it will be equally at 0! The
philosophy behind the staggering is unclear: the first "staggered" pickups had the D-Magnet
protruding the most, then this changed and the G-magnet was closest to the strings. Later,
pickups with magnets of equal length and protrusion were built (level pole-pieces), and then
again the D-magnet is most prominent "to eliminate the chorusy warble". A result of the
staggering is the overall lower loudness of the guitar because the coil needs to move away
from the string (5.4.5). There might be some reasoning in the fact that the G-string may be
wound or plain, but the multitude of pickups offered today proves that staggering is not a
mandatory requirement.
The material for the magnets of early Fender pickups was Alnico-V, an alloy from aluminum,
nickel, cobalt and iron. Although literature for magnets lists exact percentages of the alloy
components, considerable variations in the actual magnet data should be expected. The shape
of the hysteresis does not only depend on chemical composition but also very much on the
manufacturing process (4.4.1). Furthermore, it should be considered that due to the war effort
cobalt became scarce. Today nobody can remember exactly what was actually sold ... and
what was in fact used. Seth Lover, the man who developed the Gibson Humbucker, notes:
"We also used Alnico II and III, and the reason is, that you couldn't always buy Alnico V, but
whatever was available we would buy as they were all good magnets". And even if the same
magnet material is used in pickups: in modern times the data of simple cylindrical magnets
vary by ±10% – the likelihood is small that back in the good old vintage days this situation
would have been any better.
Finally we need to consider the magnet-parameter which no data sheet presents with much
precision: the reversible permeability of the magnet. This quantity describes how many
times the alternating flow conductivity in the magnet is higher that that in air. Typical values
range from 3 to 6; it is almost impossible to give exact data since the inhomogeneous
(location-dependent) magnetic flux-density leads to a total value which is application
dependent. The reversible permeability µrev determines by how much the magnet increases the
inductivity of the coil. However, µrev may not be applied directly; rather, a corrected, smaller
values need to be used because the major part of the field travels through air. Replacing the
magnets in a pickup may have several consequences: the string magnetization can change
which results in different loudness. Changes in the field geometry could change the length of
the aperture, although the connected change in treble reproduction will be mostly minor.
Changes in the reversible permeability moves the pickup resonance which determines the
sound, and changes of the eddy currents generated in the magnet change how much the
resonance is pronounced (i.e. the emphasis or Q-factor).

Since the construction principle of the first Fender pickups was as simple as it was efficient it
is still used today. Criticism was voiced only regarding two disadvantages: the sensitivity to
hum (chapter 5.2, 5.7), and the missing control over individual string loudness. Though
staggered magnets offered a kind of loudness balancing, adjustment by the musician was not
possible. Pickups with adjustable magnets provided a remedy. In old Schaller pickups, the
magnets are stuck in thin tubes which carry a screw thread on their outer surface (headless
screw, grub screw). Turning these magnets will shift them axially. This trick with the grub-
screw design was necessary because almost all magnetic materials (except CuNiFe) are so
hard and brittle that no thread can be cut into them. Old DeArmond pickups as they are found
in early Gretsch guitars have 6 adjusting screws which allow for axial movement of the
magnets.
Gibson designer Walter Fuller chose a different approach to individual string-loudness

control when he developed the P-90: this single-coil pickup utilized from 1946 featured two
bar magnets below the coil. 6 ferromagnetic screws supply the magnetic flow to the strings
(Fig. 5.1.3). Occasionally, these screws are called nickel screws but this does not indicate that
they are (or were) made from solid nickel. It is well possible that they are regular steel screws
with a nickel (or chrome) plating. The DiMarzio SDS-1 and the Fender-Mexico pickups
share their construction with the P-90. The common characteristic are the field-guiding pole-
pieces which bridge the (re. Fig. 5.1.2) larger distance between magnet and string. The high
permeability of these pole-pieces focuses the magnetic flow through the coil, however at the
same time a new material is added into the magnetic circuit on top of magnet and air. The
magnetic conductance of air is very small and frequency-independent. Steel (as well as
nickel) conducts the magnetic flow much better than air - but only at low frequencies. For
higher frequencies, which for pickups already include the kilohertz-range, eddy currents
appear which lead to an additional attenuation (chapter. 5.9). In comparison to a Fender
pickup with cylindrical magnets, a Fender pickup with bar magnets generates somewhat less
treble due to the eddy currents.
Fig. 5.1.3: Single-coil-pickups: Gibson P-90 (left), DiMarzio SDS-1 (middle), Fender Mexico (right).
Besides magnet and string, the coil winding is the third component in the signal generation.
The first electric guitars were connected to the simplest tube amplifiers having a low input
sensitivity. Consequently, the pickup had to deliver as strong a voltage as possible which
necessitated a high number of winding turns: typically 5000 - 10000 turns. Exact figure
quotes regarding pickup coils come with self-proclaimed guitar gurus just like year dates

come with Nostradamus. The first Stratocaster pickups had approx. 8350 turns. Or so one
reads. Or knows. Thus a tad more than the Texas-Special neck pickup. That one had "only"
8200 turns. As one extends the search, tolerance specifications pop up: the magic number of
8350 turns emerges as the mean value between 8000 and 8700 turns [Duchossoir] which
evidently represents the range of variation as it occurred back in the day. For years, pickups
were wound without the use of a counter, just according to "feel" until the bobbin was full. Or
belt-drive counters were used, in the hope that the inherent slippage wouldn't be too great.
How else could it be explained that Duchossoir specifies 7,5 kΩ for early Telecaster bridge
pickups, while Day/Rebellius report up to 11 kΩ. No sooner than 1960 does Fender introduce
precise automatic winders - and still they continued to experiment with winding numbers.
Duchossoir writes in the Stratocaster booklet that the finished coils were merely monitored by
measuring the coil resistance with an Ohm-meter having a tolerance of as high as ±20%!
Considering - on top of this - that the wire resistance per length is also subject to production
tolerances, it is easy to imagine enormous variations in the winding numbers. Very generally
the following holds: if for a specific pickup the number of coil turns is increased, an increase
in resistance and inductivity follows. The resonance frequency drops, and the pickup gets
louder. However, the resistance itself has little bearing on the transmission characteristic – the
dampening effect it has remains small compared to that of other components in the circuit. If
geometry and wire diameter are indeed known, it is possible to draw conclusions about the
winding number from the resistance. Only given these boundary conditions there is validity in
the rule: higher resistance = louder reproduction.
Next to single-coil pickups without any field-directing pole-pieces (e.g. the Stratocaster) and
pickups comprising pole-pieces between magnet and string (e.g. the P-90), there is a third
significant group which features guidance of the return of magnetic flux via pole-pieces. Fig.
5.1.4 explains the principle based on the Fender Telecaster bridge pickup. Here, a metal plate
positioned underneath the coil is - according to advertisements - supposed to shield, and to
"reflect" the magnetic field. Duchossoir describes the material of this approx. 1,2 mm strong
metal sheet as "tin" although this should not translate into actual sheet tin. "Tin" can also
stand for tin-plated steel which is more likely to have been used since solid tin is not
magnetic. Fender brochures refer to a zinc shielding plate, i.e. a galvanization. That's also
fine. From 1951 a copper-plated steel sheet is used which is dropped in 1981 without any
replacement. Presumably people at Fender realized, too, that the strengthening of the
magnetic field towards the strings is so insignificant that the plate may as well be dropped.
Another possible reason may have been movements of the plate which could lead to
microphonic noise and feedback. Measurements do not confirm any magnetic shielding
effect: the presence of the plate creates merely a difference of 0,1 dB in the interference in the
parallel field. The signal level is increased by the plate by only 0,6 dB which is too little to be
noticed much. Similar results occur for the resonance frequency (3% change) and the eddy
current dampening (approx. 1 dB difference).
Fig. 5.1.4: Telecaster pickup with metal plate under the

lower flange. The plate increases the sensitivity by 0,6
dB and reduces the resonance emphasis by approx. 1
dB; the resonance frequency drops by 3 % (due to an
increase in inductivity by 6%)

Thoroughly continuing to think along the lines of the idea

behind the Telecaster metal plate brings us very quickly to
the variant Leo Fender implemented in the Jaguar. Here, a
yoke sheet is bent around the coil winding in the shape of a
"u" and the toothed upper rim focuses the field towards the
strings. It didn't help much, though. At the time Fender's
most expensive guitar, the Jaguar was not a commercial
success. Incidentally, the pickup shielding was quite
efficient, but other features of the guitar (e.g. the bridge)
Fig. 5.1.5: Pickup with field guide mercilessly bombed on the market.
sheet: Fender Jaguar.
Since pickup coils are comprised of extremely thin wire, they should be protected against
damage by a housing. For the Telecaster's bridge pickup, this was accomplished by winding a
thick thread over the copper winding. Simple and effective. Mechanical – not magnetic.
Indeed, a magnetic field cannot be changed by a thread – but it may easily be by a metal
housing (Fig. 5.1.6) as found with the Telecaster neck pickup (eddy currents, chapter 5.9.2.2).
A motivation behind the metal housing was – on top of the physical protection – presumably
the desire to shield the pickup. Mind you: no remedy can be achieved via this approach
against magnetic interference; for that, the principle of construction would need to change
(chapter 5.2, 5.3)..
Fig 5.1.6:
Single-coil pickup with
protective metal casing. To
avoid losses due to eddy
current, the housing needs to
be made of nickel silver.
A pickup installed in Gretsch guitars merits particular consideration: the "HiLoTron" (Fig.
5.1.7). In order to pickup as many harmonics as possible, the magnet was installed with
horizontal orientation. This reasoning is elusive from a systems theory point-of-view, but
apparently only the resulting sound counted for the developer – which in itself is highly
purposeful, and the justification in the associated patent (chapter 5.10.5) does not really need
to be correct now, does it!? The horizontally oriented magnet is also found in the Attila-
Zoller-Pickup (US-Patent No. 3588311) – and to go with it again a rather unconventional
reasoning in the corresponding patent. The US patent examiner was apparently not phased by
that ...
Fig.. 5.1.7: Gretsch HiLoTron (see also US-Patent No. 2683388).

Fig. 5.1.8: Comparison of single-coil pickups

5.2 Humbucker 5-9
5.2 Humbuckers
The interference occurring with single-coil pickups motivated the development of the
Humbucker. Single-coil pickups do not only pickup the vibration of the strings and generate a
corresponding electric voltage, but they are also sensitive to magnetic fields as they are
radiated by transformers, fluorescent lamps, or mains cables. Instead of having one coil, the
"Hum-Bucker" consists of two coils connected to form a dipole and wired such that they are
out of phase. The magnetic field generated by external interference sources induces in each
coil the same voltage. Because of the anti-phase connection of the two coils the voltages
cancel each other out. If the field generated by the permanent magnet would also flow through
both coils with the same polarity, the signals generated by the vibrating string also be
cancelled – this of course must not happen. For this reason the permanent field flows through
the two coils in an anti-parallel manner such that the voltages induced by the vibrating strings
are out of phase. Because the coils are connected out of phase, the voltages are turned twice
by 180° i.e. they are again in phase (180° +180° = 360° corresp. to 0°). With this
arrangement the signal-to-noise ratio can be improved somewhat compared to single-coil
pickups (chapter 5.7).
As early as the 1930s designers sought to develop a marketable pickup based on

compensation principles which were generally already known. Seth Lover, technician with
the guitar manufacturer Gibson, achieved the commercial break-through. He is the designer
of the Gibson Humbucker, but he's not the inventor of the humbucking principle as he himself
noted: "People had been working on double coil pickups since the 1930s [13]". Lover's patent
application from 1955 cites a further seven earlier patents for pickups considered in the
procedure which also already had been referring to the compensatory principle. Lover was
thus not the first but he succeeded together with Gibson in creating a commercially highly
successful, even "mythical", pickup which in this respect far surpassed e.g. the Gretsch
humbucker appearing almost at the same time (FilterTron pickup developed by Ray Butts).
Gibson applied for a patent for their humbucker in 1955. The patent was granted in 1959,
however already in 1957 Gibson guitars fitted with humbuckers appeared on the market. Up
to the granting of the patent the pickups sported the sticker "Patent Applied For". This led to
the abbreviation PAF-pickup. In 1962 the PAF sticker was changed: instead of "Patent
Applied For" now the patent number 2.737.842 could be read. The correct number of the
"Humbucking"-patent from 1959 was however 2.896.491. Allegedly, the misleading number
was deliberately printed on the sticker to fool competitors. Or so says Seth Lover.
The humbucker uses two coils instead of one with the objective that hum voltages are
superimposed out of phase and thus cancelled while the voltages derived from the moving
string are added in phase and thus amplified. Single-coil and humbucking pickups differ not
only in the interference voltages they pickup. Their different construction results also in
different transfer functions in i.e. a different sound. Musicians often express the opinion that
single-coils are softer but have more treble while humbuckers are louder but sound darker.
This may have been a reasonable assessment correct statement regarding the early guitars of
Fender and Gibson, however this prejudice is not suitable as dogma. The pickups of a Fender
Telecaster and those of a Les Paul differ not only in the number of the coils but also in the
pickup's inductivity, resonance frequency, and resonance dampening. The following sections
explain how the pickup parameters influence the magneto-electric transmission, and how this
determines the sound

Fig. 5.2.1: Gibson-Humbucker [drawing: Mike McDonald].
Fig. 5.2.2: Cross section of Humbucker. Gibson Type 490 (left), Gretsch FilterTron (right).

5.2 Humbucker 5-11
Fig. 5.2.1 shows the construction of a Gibson-Humbucker. Mounted on a base-plate (8) we

find a wooden strip (5) serving as spacer, an alnico bar-magnet (6), and a metal block (7) with
multiple bores. Placed above this are the two bobbins (3) with their coil windings, fixed with
two screws. One of the bobbins carries 6 cylindrical metal pins (1) which are often called
slugs, the other 6 metal screws (2). The cross-section shown in Fig. 5.2.2 describes the
magnetic flux: the bar-shaped permanent magnet is polarized horizontally and causes a
circular flux flowing – from the north pole – through the pin (slug) and returning through
string, screw and metal block to the south pole. Only a small part of the overall magnetic flux
runs through the string while most of it circles back through air as flux leakage. In Gibson's
patent publication two similar coils with pins are shown. The production version included the
two different coils, with the second one carrying the screws for adjusting the volume of
individual strings. The FilterTron pickup installed in Gretsch guitars uses a similar
construction principle. With its two rows of screws it achieves full mirror symmetry and thus
a better hum suppression. Both humbuckers shown in Fig. 5.2.2 are sealed with a metal cover.
In the Gibson Humbucker, an alnico magnet generates the permanent field. Without it the
pickup would not work. However, the influence of the specific magnetic material must not be
overestimated: the alternating magnetic field (which exclusively induces the voltage in the
coils) oscillates predominantly in the vicinity of the string; only a very small part reaches the
magnet (chapter 5.4.3). We have a similar situation for the magnetic field generated by a
current flowing in the coil and determining the inductivity: measuring the pickup resonance
with and without magnet show merely a 3% difference in the inductivity (chapter 5.9.2.6)
which is negligible compared to other parameter variations. Whether a strong or a weak
magnet is incorporated will have slight effects on the sound, but a significant change is to be
expected only in the loudness. Regarding the question which magnetic material was (or is) in
fact used one finds comprehensive answers in literature. Not to mention the Internet! "You
have many more hits than there are magnetic materials!" BINGO!
"Up to 1950, there was no commitment to a specific alnico material at Gibson, and Alnico 2,
4, 5, and 8 were installed depending on availability and presumably also on most favorable
purchase cost. From 1950 (...) Alnico 5 prevailed as predominantly used magnet material.
Which however does not mean that it stayed that way. Even towards the end of the 1950's
humbucker specimen with by all appearances other Alnico magnets do surface [Day et al.]".
"The magnets in Burst-PAFs were made of Alnico II and IV [VG Magazine]". "This pickup
(SH-55) was re-introduced by Seymour Duncan using the specifications of PAF-inventor Seth
Lover to 100%: Alnico-2 magnets" [Musik Produktiv catalogue]. "The SH-55 is really
faithful to the original, it will have my stamp of approval on it [Seth Lover in VG Magazine]".
We also used Alnico II and III, and the reason is, that you couldn't always buy Alnico V, but
whatever was available we would buy as they were all good magnets [the same Seth Lover in
the book The Gibson]".
So there we have it: most probably anything that couldn't climb a tree fast enough was
installed by Gibson in their pickups. Add two coils with 4500 turns each ... or more .... or less.
Then: slap on the cover and – most importantly from today's point of view – stick that PAF-
sticker to the bottom. Done. Today it'll cost ya $3000.- per piece. That's per piece pickup, not
per piece guitar! Occasionally that could rise to $10000.-. Trend: upwards. But then ...
Rembrandt's legacy is not evaluated based on the cost of paint and canvas he incurred back
then, either.

The screws and pins (i.e the pole-pieces) focus the field and sample the vibrations of each
string in two sectors which are separated by about 19 mm. In particular for the bass-strings of
the guitar a loss in brilliance results, which however is not generally undesired in particular
for distorted sound (chapter 5.10.5). To counter the treble loss – which is due to interference
effects – the distance between the poles needs to be reduced to a few millimeters. At the same
times, this allows for mounting the humbucker (now reduced in size) into a housing foreseen
for single-coil pickups – it will now fit into the single-coil-routing in the guitar body. Fig.
5.2.3 shows an in-scale comparison between a Gibson Humbucker (here a special version
with 3 magnets) and a DiMarzio-Humbucker. The latter employs 2 1,6-mm-strong iron blades
of 6 cm length, which run at a distance of 7,5 mm across the strings. Instead of screws and
pins, narrow blade-shaped pole-pieces were used very early on by Willi Lorenz Stich, alias
Bela Lorentowsky, alias Billy Lorento, alias Bill Lawrence, they later show up in Joe-
Barden-pickups, and by now they are also offered by Seymour Duncan and DiMarzio – an the
are rejected rigorously by many guitarists just because of their look.
Fig. 5.2.3: Gibson 'Super'-Humbucker [acc. to Lemme] with 3 magnets, und DiMarzio-Humbucker with two
metal blades. The Super-Humbucker installed in the L6-S had coaxial coils, however [Billlawrence.com].
Different construction of the two coils (Fig. 5.2.4) influences in particular inductivity and Q-
factor. Humbuckers with identically constructed coils target a broad-band cancellation of the
interference. Differences in shape and/or material of pole-pieces, wire diameter and/or
number of turns allow for limiting the cancellation to specific frequency ranges (usually the
lower frequencies), and for modification of the transfer function in the remaining frequency
range. The typical humbucker interference notch (chapter 2.8.3) can be shifted or reduced in
this manner. The exact calculation of the transfer behavior gets complicated since the coils are
magnetically (and in some cases to a non-negligible degree even capacitively) coupled. This
coupling needs to be considered also if only one of the coils of a humbucker is connected
(humbucker in single-coil mode, split operation). The magnetic poles (or the pole-pieces) of
the unused coil still generate an alternating magnetic field which partially flows through the
used coil and induces a voltage there.
Fig. 5.2.4: Various humbucker construction types.

5.3 Hum-compensated single-coil pickups 5-13
5.3 Hum-compensated Single-coil Pickups
Magnetic pickups convert alternating magnetic fields into electrical alternating fields. If these
fields are generated by a transformer, an electric motor, a monitor working based on magnetic
deflection, or similar source of magnetic fields, the conversion happens nonetheless – how
should a pickup know that these are unwanted signals? A possibility to attenuate such
interference was discussed in chapter 5.2 – another approach is taken by so-called stacked
single-coils also known as stacked humbuckers or co-axial humbuckers. Viewed from the
direction of the strings, such a pickup looks like a normal single-coil. However, in its interior
two coils are at work, and the designation single-coil is therefore not entirely correct. Or
maybe it is, after all, because only one coil senses the vibration of the strings; the other coil
compensates the hum voltage. Thus: humbucker – but a special one, specifically a co-axial
one.
Co-axial indicates that both coils are wound around the same axis; they do not, however, lie
in the same plane (as they could if they had different diameters of the winding). Rather, two
similar coils are 'stacked" on top of each other: one closer to the strings, one further away. As
we will see in chapter 5.4.3, the alternating magnetic flux circulates only close to the strings,
i.e. it does not penetrate the whole magnet with the same strength. For this reason, only the
coil windings positioned close to the string receive a significant part of the alternating flux.
the interference field of an external interference source creates an entirely different situation:
its virtually parallel field lines penetrate the whole of the winding and therefore induce
approximately the same voltage in every turn irrespective of the distance to the string♣.
Dividing the coil into one half closer to the strings and a second half facing away from the
strings, and at the same time connecting the two partial coils out-of-phase, will result in a
compensation of the interference voltage while the useful signal is attenuated only a little.
Compared to the uncompensated single-coil pickup, the co-axial humbucker shows several
differences: there is no hum but more space is required plus the sound is different. The space
requirement it rarely problematic but the altered transmission characteristic continues to be
fodder for extensive discussions. In order to clarify the context, it is helpful to separate the
pickups into two groups: there are those with elongated, slim coil shapes (such as e.g. the
Stratocaster pickup), and those with wide, flat coils (e.g. the P-90, Fig. 5.3.1).
Fig. 5.3.1: different winding-shapes in

single-coil pickup
Let us assume that the winding of the Stratocaster pickup shown in Fig. 5.3.1 would be
divided at half its height such that two coils result. The induced voltages are, however, not
divided in a 50:50-ratio, but – due to the location-dependent alternating flux-density – by
75:25. The upper winding (closer to the strings) receives a voltage which is 3 times that of the
lower winding. Connecting both halves of the winding out-of-phase to compensate the hum
decreases the string-induced voltage by half. The pickup is softer in loudness than an
uncompensated single-coil would be.
♣
the voltage induced into the winding depends dB/dt and on the area of the winding

Not only the loudness changes but also the sound-spectrum. This is because the phase-switch
reduces the inductivity of the pickup. The inductivity is the quotient of coil flux and current
[e.g. 18]. If the two anti-phase-connected halves of the coil were in the same place – this only
works as a though-experiment – then the excitation current flowing through both coils would
generate no magnetic field at all and the inductivity of this 'bifilar'-wound coil would be zero.
In reality the two halves of the coil are at different locations and the magnetic flux (generated
by the excitation current) in one coil would not fully compensate the flux in the other. The
inductivity therefore is not zero but smaller than for an in-phase connection. Smaller
inductivity implies a higher resonance frequency (chapter 5.9), and consequently the
conclusion is: due to the phase reversal, the pickup sound softer and with more emphasis on
treble. Whether this is perceived as advantage or disadvantage is a matter of individual
assessment. However, often a direct comparison with the uncompensated original pickup is
done, and the scathing verdict is: the hum-compensation kills the sound.
So far our considerations showed two effects of the phase reversal: weaker output voltage and
smaller inductivity. Simple corrective measures for this are: increased number of turns and
improved decoupling of the two halves of the coil. The coupling of the coils is determined by
the distance in space of the coils, and the permeability of the coil core. Soft iron pole-pieces
passing through the coils are rather disadvantageous in this respect, while the relatively small
permeability of customary pickup magnets on the other hand diminishes the coupling of the
two coil fields and reduces the disadvantages of the phase reversal. An even better decoupling
is achieved by a metal plate with high permeability separating – as flux-guiding yoke – the
two coil halves. Optionally, this plate can be bent to a u-shape. With magnetic decoupling and
an increased number of turns, co-axial humbuckers achieve a similar transmission
characteristic as single-coils. Complete identity is however impossible: the spatial distribution
of the magnetic flux (inc. all skin-effects) is different, and due to the higher number of turns
(plus 50% or more) the dc resistance changes. The latter does not only have an effect at 0 Hz,
but may influence the resonance emphasis (chapter 5.9).
For pickups constructed like the P-90 (Fig. 5.3.1) the division of the coil as just shown is not
purposeful: the coil is more shallow than the one in the Stratocaster pickup and therefore both
halves of a divided coil would be close to the strings, i.e. in the alternating magnetic field.
Furthermore 6 screws (pole-pieces) would make for a relatively strong coupling of all
windings. Possibly for this reason, Gibson did not divide up the coil present in the P-100 but
installed a second coil below the magnets which now serve as a magnetic shield as well. A
series connection of the coils would have doubled the already rather high inductance (approx.
7 H) and reduced the resonance frequency by 30%; apparently this was not desirable. For the
P-100 the coils are therefore not connected in series but in parallel (and anti-phase). Of
course, this has consequences as well: the resonance frequency is now higher compared to the
P-90. Obviously the musicians were not excited – production has since ceased.
Fig. 5.3.2 shows cross-sections of well-known co-axial humbuckers. Almost all have received
patent protection. US patent protection, that is. The question regarding the necessary
individual inventive step would probably only have come up in pedantic old Europe.

5.3 Hum-compensated single-coil pickups 5-15
Kinman, 15.03.1996, US-Pat. Nr. 5668520, 5834999, 6103966.
Freeman, 21.12.1970, Bill Lawrence L-220. The axis of Stich, 05.08.1974, US-P. 3902394,
US-Pat. Nr. 3657461 the coils run horizontally (Stich). horizontal axis of the coils.
DiMarzio, 06.08.1982, Seymour Duncan, 15.08.1983,

US-Pat. Nr. 4442749. US-Pat. Nr. 4524667.
Anderson, 14.01.1991,
US-Pat. Nr. 5168117.
Fender, 28.01.1998, Devers, 17.05.1999,

US-Pat. Nr. 6291758 US-Pat. Nr. 6846981
Fig. 5.3.2: Various co-axial humbuckers; dates given are those of the patent filing

Fig. 5.3.3: One of the first co-axial humbuckers, US-Pat. 2119584.

Both coils contain a core of layered transformer laminates; the cores
are separated by a nonmagnetic spacer. Prior to use first a direct
current had to be fed to the upper coil to magnetize the strings. Day
of filing for the patent: 9.12.1935.
Fig. 5.3.3 shows that Gibson is not the inventor♣ of the humbucking principle: even before the
famed PAF, there was the trendsetting idea to interconnect two coils in anti-phase. Seth
Lover, designer of Gibson's humbucker, was himself informed about competing pickup
developments: "People had been working on double coil pickups since the 1930s [13]". As
early as 1935, Arnold Lesti filed an application for a pickup with tow side-by-side coils (US-
Patent 2026841 = Re.20070) and describes the principle of interference: "And since these
coils are wound in opposite directions, the interfering stray currents are neutralized". On tp
of that, it would be possible to imply that Gerald Tuininga attempted - in his patent
application filed in 1929 and leading up to US patent 1838886 – to compensate interference
with the use of two coils: "The advantage of using this style of transmitter is that no other
electric current caused by foreign sound or vibration can in any way enter into the circuit".
In 1929, descriptions of patent applications as this one comprised only little more that one
page in letter format, so we should not be too small-minded and start splitting coil wire ... er:
hairs. Still, the circuit included in the patent description seems incorrectly drawn. If both coils
indeed had the same direction of their winding the wanted signals would cancel each other out
while the interference would double. Inverting one of the coils – which would have been the
only way back then it would have worked in the breadboard setup – one obtains a fully
functioning humbucker. Mind you, it would still needed firing up an electromagnet via a
battery. That we are not burdened by such cumbersome procedures anymore today – that we
owe to inventors such as Seth Lover (patent application in 1955). Or Leo Fender, who filed
an application for his humbucker in 1956. Or Ray Butts, who filed the one for his Gretsch-
Humbucker in 1957. Or Oskar Vierling, who as early as 1927 published the basic principle
of the electromagnetic string pickup with the German patent office in Berlin.
Whether – as Day et al. surmise – Bill Lawrence put together already in 1948 the "probably
world's first humbucker" is questionable. It would be possible: Lawrence was born 1931. On
the other hand, he himself dates the beginning of his entrepreneurial activities to 1965:
"Electrosounds in Munich, Germany". Back then Bill Lawrence was still called Willi Lorenz
Stich, and one of his partners was Jzchak Wajcman. It was the same Jzchak who would later
push Lawrence into a $ 1.156.250,00 bankruptcy [Guitar Player, September 1979, cited in
billlawrence.com]. Incidentally, the St. Lorenz, alias Laurenz, alias Laurentius, alias
Lawrence was "burned to death on an gridiron". You lucked out, Bill! (using this expression
since B.L. later lived in the US .... for you British readers this would have to read: "You had a
lucky escape, William!").
♣
In their advertisements for strings, Gibson indeed merely claim to be the "inventor of the Humbucker" ... and
not the "inventor of the humbucking principle"

5.4 Magnetic Field of Pickup 5-17
5.4 The magnetic field of the pickup
5.4.1 Static magnetic field without string
The vibrating string causes a change in the magnetic flux; this change induces an electric
voltage in the pickup coil. The terminology of systems theory describes change as a dynamic
(i.e. time-dependent) process superimposed onto a static magnetic field. The alternating flux
is rather small and reaches merely about 1% of the static part of the field even for strong
excitation of the string.
The source of the magnetic field is a permanent magnet installed under the string in the
pickup housing. For a typical Fender pickup (for example the one in the Stratocaster) the end
surface of the axially magnetized cylindrical magnet is positioned a few millimeters from the
strings. For the Gibson P-90 a bar magnet is mounted underneath the pickup coil; for better
field focus ferromagnetic screws penetrate the coil surface and guide the magnetic flux to the
string. It is of interest to measure the strength of the static magnetic field since the efficiency
of the mechano-electric transduction depends on it: without magnetic field there is no induced
voltage i.e. the stronger the magnetic field the louder the pickup, although the
correspondences are not quite that simple, after all. Besides the absolute strength of the
magnetic field, its distribution in space is of importance as well. Moreover the static magnetic
field exerts attraction forces towards the string which influence the vibration behavior − for
this reason particularly strong magnets are not generally desirable.
To measure the static magnetic field, a Hall probe (after Edwin Hall) is suitable. This is a
small semiconductor plate in which an electric voltage dependent on the magnetic field is
generated. The effective measurement surface is about 0.4 mm in diameter. For the
measurements described in the following, such a Hall probe was moved along a straight line
by a spindle drive. At the same time the field-proportional electrical voltage was recorded.
The direction of the advance was either in parallel to the string axis or perpendicular to it.
With a parallel shift of the Hall probe an area could be sampled.
In contrast to the sound pressure measurements favored in acoustics, the magnetic flux
density is not a scalar but a vector in space. The electromagnetic field is a vector field, each
point of which in space is associated with three-dimensional field values. The Hall probe ,
however, reacts merely to the flux density component which is parallel to its surface vector.
For a complete description of the field it would be necessary to use three orthogonally
oriented Hall probes. Simultaneous operation of the three sensors results in a mutual
interference, sequential operation is problematic due to the limited accuracy of the positioning
in space. To make the overall measurement effort not too excessive, it was the axial
component which was recorded. What is meant here is not the axis of the string but the axis
of the cylindrical magnets or the pole-pieces; in other words the Hall probe is oriented in
parallel to the fretboard of the guitar and samples the magnetic field component perpendicular
to the fretboard. In the vicinity of the magnetic poles a flux density of between 10 and 100
mT is found while larger distances result in a very steep decrease of B. Figure 5.4.1 gives an
impression of the field pattern above the pole area. Of course, it needs always to be
considered that a pickup without string is without purpose. The field pattern with string is
more important, however this is also much more difficult to determine.

Fig. 5.4.1: Field vectors above a magnetic pole without string in the filed (measured results). For the Jazzmaster
pickup (top) the field diverges more strongly than for the SDS-1 (bottom). Length and direction of the individual
lines represent strength and direction of the magnetic flux density; the coordinates (given in millimeters) refer to
the middle of the pole-plates (shown as thick line on the lower border of the figure. For this representation the
abscissa- and ordinate-components of the B-vector were measured at distances of d = 1:0,5:6 mm to the magnetic
pole.

In Fig. 5.4.1, the individual lines of the dashed line field represent – with their length and
direction – the pattern of the field. Since the medium the field propagates in is air, both the B-
and the H-Patterns can be determined: . The magnetic field is a vortex-field: its flux
lines (field lines) are closed lines without start- or end-point. Nevertheless, a presentation as a
point-source-field is customary, as well, although this is a rather rough simplification. For
the point-source approximation, the magnetic flux is thought of as originating from a point-
source which is located within the interior of the cylindrical magnet on its axle. A first-order
approximation for the distance of this point to the to the front face is the radius of the
cylinder. Outbound from this source the magnetic field diverges equally in all direction. The
surface area of a sphere concentric with the source point increases with the square of the
radius, and thus the radially oriented flux-density will decrease with the square ( B ∼ 1/r2 ).
Fig. 5.4.2 shows the measured results for the flux density at the magnet axis m; for this, the
Hall probe was moving axially away from the pole-piece.
Fig. 5.4.2: Axial Flux-Density in absolute (left) and relative (right) representation,
d = distance to the pole-plate
The field of the Jazzmaster pickup is, in absolute terms, larger than that of the SDS-1 but does
decrease faster. If this decrease happens according to a power law, it should show up as a
straight line in double-logarithmic coordinates. Fig. 5.4.3 shows log(B/B0) over
log[(d+Δ)/d0]; the abscissa, however, is scaled for d and not for d+Δ. B0 and d0 are reference
values for the logarithms (such that they are without a dimension). + Δ is the depth of the
magnetic source: it amounts to Δ = 4.7 mm for the SDS-1, and for the Jazzmaster-pickup it is
Δ = 3 mm.
Fig 5.4.3: as shown in Fig. 5.4.2, but here in double-logarithmic scaling (measured ––––, 1/r2-dep. -----).

The measured data shown in Fig. 5.4.3 are located almost perfectly on the given straight lines
which approximates the 1/r2-dependence rather well. We still need to consider that only data
along the magnet axis are depicted; in contrast Fig. 5.4.1 lends itself to show that the
elongations of the field vectors do not met in a single source-point, after all. Here, the point-
source-approximation reaches its limit of validity.
For the humbucker both magnet poles are positioned close to the strings; this results in a
dipole field (Fig. 5.4.4). Directly in front of the pole plate (slug or screw) we obtain a
rotationally symmetric field similar to Fig. 5.4.1, with a dependency on distance as given in
Fig. 5.4.1. In the area between the pole plates (middle of the figure) the superposition of the
anti-phasic fields results in a compensation of the vertical field component such that the
magnetic flux runs horizontally i.e. parallel to the strings.
Fig. 5.4.4: Dipole-field of a humbucker (Gibson ES 335). The screw (right pole) is the south pole. No string.
Fig. 5.4.5: magnitude of the vertical field, measured on the axis of the magnet.
The distance-dependency corresponds well to a 1/r2-Funktion well.
ES335: ΔSlug = 5,1 mm, ΔScrew = 4,0 mm. 490R: ΔSlug = 4,1 mm, ΔScrew = 4,0 mm
Magnetic fields are vector fields; a complete characterization of the B-field would require a
special representation of all three B-coordinates which is impossible to accomplish with two-
dimensional figures. In order to still get an impression of the filed distribution, colored flux-
diagrams are shown in the following. The axial component of the B-vector (corresponding to
the vertical component in Fig. 5.4.1) was measured with a Hall probe at a distance of
2 mm from the pole plate. It was then recorded using color-coding. For single coil pickups
the areas of small flux density are shown in blue; in contrast, the same color blue
characterizes areas of high negative flux density.

a) Axial magnetic flux density for singlecoil pickups:
Gibson P-90, bar magnet + pole-pieces (screws)
DiMarzio SDS-1, bar magnet + pole-pieces (screws)
Fender Jazzmaster, cylindrical bar magnets
Fender Stratocaster, cylindrical magnets of varying lengths
Fender Telecaster (bridge), cylindrical magnets + metal plate
Fig. 5.4.6: The column on the left shows the distribution of standardized axial flux density in the plane of
the strings. The color-scaling is as given by the color bar on the lower right
The right-hand column depicts the absolute axial flux densities 2 mm above the pole plates. d = 2mm.

b) Axial magnetic flux density for humbucking pickups:
Squier Humbucker
bar magnet, 6 (pole-) screws, 6 pole pins (slugs)
Gretsch Filtertron
bar magnet, 12 pole-screws
Gibson ES335 (square window),

bar magnet, 6 (pole-) screws, 6 pole pins (slugs)
DiMarzio DP184
bar magnet, 2 (pole-) blades
Fig. 5.4.7: Standardized axial flux density (left-hand column),

magnitude of the absolute flux density (to the right);
bipolar color scaling (color bar lower right); d = 2mm

Figs. 5.4.6 and 5.4.7 show the axial component of the static magnetic flux (measurement
without string), i.e. the flux running perpendicular to the guitar top. The figure could create
the impression that the Jazzmaster Pickup generates a more focused field than the P-90.
However, the contrary is the case: a (locally) quick decrease of the axial component points to
a strongly diverging field. For the Jazzmaster pickup (Fig. 5.4.1), the vertical (= axial) field
component at 2 mm distance decreases quickly with horizontal movement of the measuring
point because the direction of the field changes strongly.
Still, one should not attribute too much significance to the geometry of the magnetic field. As
soon as a steel string is introduced in front of the pole-pieces the static magnetic flux changes,
and as the string starts to vibrate, again entirely new field shapes result (Ch. 5,4,3). Actually,
measurements of the static magnetic field are only undertaken to obtain hints as to the
magnet(s), and even there merely a rough classification is advised: very strong (50 – 60 mT),
strong (40 – 50 mT), medium (30 – 40 mT), weak (20 – 30 mT) and very weak (< 20 mT),
with all measurements taken at a distance of 2 mm. Sure, the class borders given here are a
subjective choice – if one so desires, 5-mT-intervals may also be used. Much finer steps are
not purposeful, though: the measurement results depend rather strongly on the measurement
position, after all, adjustment screws may be twisted, the measuring distance may be defined
differently in case of tilted magnets of bent carrier plates, the 6 magnets of a pickup may
result in different flux densities – with all these imponderables it is only possible to arrive at a
mean value to the best of ones knowledge.
In the following table the static field measurements are listed – each taken at a distance of
2 mm above the pole plate (i.e. the slug, screw and blade, respectively). Data were collected
using a Hall probe (Bell Technologies Inc., Model 5060, Gauss/Tesla Meter). The
measurement error is specified by the manufacturer to ± 4%, which is adequately precise
because the errors caused by inaccuracies of the sensor-positioning are – as a rule – bigger.
Thus it is not sensible in the framework of the results presented here to judge e.g. the Duncan
APTL-1 with its 36 mT as “stronger” relative to the Jazzmaster pickup (33 mT).

Pickup §)
Fender Telecaster Texas special (Bridge)
Fender Telecaster-70 (Bridge)
Fender Stratocaster (USA Standard, Middle)
Rickenbacker (Toaster-Pickup)
Fender Noiseless Stratocaster (Neck)
Fender Stratocaster (USA Standard, Neck)
Fender Stratocaster (USA Standard, Bridge)
Fender Stratocaster-72
Fender Jaguar (Neck)
Duncan SSL-1 (Strat-Type)
Duncan APTL-1 (Telecaster-Type Bridge)
Fender Jazzmaster-62 (Neck)
Fender Jazzmaster-62 (Bridge)
Schaller
Fender Vintage Telecaster (Bridge)
"Telecaster"-Fake (Bridge)
DiMarzio DP172 (Tele-Type Neck) with cover
Rockinger Strat (bar magnet)
Fender Stratocaster (bar magnet)
DiMarzio SDS-1
Duncan APTR-1 (Telecaster-Type Neck) with cover
Fender Vintage Telecaster (Neck) with cover
Lace-Sensor gold
Gibson P90
"Telecaster"-Fake (Neck) with cover
Rockinger P90
Ibanez Blazer (Strat-Type Type)
Gretsch HiLoTron
Gretsch Filtertron
DiMarzio DP107 Megadrive
Joe Barden (Strat-Type, Bridge)
DiMarzio DP184
Gibson Tony Iommi with cover
Squier Humbucker without cover
Gibson Burstbucker Neck with cover
Gibson Burstbucker Bridge with cover
Gibson 490R without cover
Gibson ES 335 (Neck, 1968) without cover
Gibson 57 classic with cover
Gibson ES 335 (Bridge, 1968) without cover
Table: static pickup magnetic field without strings. + = north pole; measured at 2 mm distance
(orientation values – the measurement precision is mere moderate).
§) the actual numbers are reserved for the printed version of this publication

5.4.2 Static magnetic field in the presence of the string
Magnetic pickups only work with ferromagnetic strings. A large part of the magnetic flux
exiting the face of the magnet (pole plate) penetrates the string, splits in both directions, runs
within the string for a few millimeters, and exits again after a short distance. Fig. 5.4.8 shows
the fundamental course of the flux for the example with a cylindrical alnico magnet. In the
neutral zone – this is the plane dividing the magnet into 2 cylinders of equal size – the flux
density amounts to 0,63 T; this corresponds to a magnetic flux of 12,6 µWb given a cross-
sectional area of 20 mm2. About 50% of the flux leaves the magnet via the cylinder side-wall
while the remaining 50% exit via the pole plate – again about half of which flows through the
string.
Fig. 5.4.8: Magnet, string, flux lines.

The shape of the field is not calculated
exactly but shown as a simplification
A direct measurement of the static magnetic flux travelling in the string is not possible.
However, the continuity conditions allow for conclusions about the axial flux; a small
measuring coil enclosing the string is moved axially along the string; the voltage induced in it
corresponds to the axial flux change the integration of which results in the axial flux. The
measurements presented in the following were done with a D'Addario-String (diameter =
0,66mm). The measurement coil had 64 turns of CuL-wire (∅ = 80µm) wound in several
layers to have an inner diameter of 1 mm and a length of 2 mm. Using a synchronous motor
powering a spindle drive, this coil was pushed with a speed of 6,35 cm/s along a string of a
length of 18cm. Halfway along this distance an alnico magnet was positioned perpendicular to
the string; the gap d between string and magnet was adjustable. For aiming the measurement
parameters there is a troublesome conflict: the coils should be as small as possible in order to
arrive at a good local resolution – given the overall dimensions even a length of a little as 2
mm is relatively long). Reducing the wire-diameter does diminish the coil dimensions .... but
also the motivation of the one carrying out the procedure as the barely visible wires break
again and again. The 80 µm CuL-wire proved to be a good compromise. 64 turns kept the
outer diameter sufficiently small such that not too much of the field in air was sampled as
well. The feed speed of the spindle drive should on the one hand be as high as possible to
generate a high induction voltage but on the other hand the motor needs to be given enough
time to reach a constant speed, which precludes very short measuring times. A precision
spindle (with a gradient of 2,54 mm) yielded a feed speed of 6,35 cm/s and an induction
voltage just short of 1 mV. These are manageable values.
Since the measurement coil is of low impedance and at the same time the coil voltage is
integrated, noise interferences are not critical. The offset of the amplifier, however, poses a
problem. Even though the offset voltage (approx. 18 µV relative to the input) appears rather
small, the resulting error would be too large (Fig. 5.4.9). An induction voltage of 18 µV
corresponds to a flux-density change of 0.8 T/s; given a measurement time of 2 s this would
result in an offset-based deviation of no less than 1,6 T! This error needs to and can be
compensated – but not entirely, because the offset voltage is not constant but drifts such that a
small residual error remains. In practice these deviations are insignificant. In Fig. 5.4.9, a
measurement with offset compensation is compared to one without it. The uncompensated
measured flux density switches „on the way“ – which is a no-go, of course.

Fig. 5.4.9: Induced voltage (left) and – calculated

from it via integration – magnetic flux density in the
string without offset-compensation (lower left), with
offset compensation (lower right):
Distance of magnet d = 3mm.
String : D'Addario 0.66mm.
There are two basic approaches to magnetize the string: either one starts from an un-
magnetized string and brings the pickup magnet – starting from a big distance – closer up to
the desired distance. Alternatively, one may first let the magnet touch the string (d = 0) and
then moves it away to the desired location. Due to the hysteresis-like B/H-connection these
two measurement approaches do not arrive at the same magnetic flux despite the equal
eventual distance. The string becomes a magnetic source proper because of the external
magnetic field. The overall flux through the string can be interpreted as the sum of an
externally generated und an internally generated flux. As the pickup magnet is brought closer
to the originally demagnetized string, an internal magnet is switched on, so to speak, and it
now supports the flux generated by the external magnet. Even as the external magnet is
moved away again from the string by a few millimeters, the string retains a remanent
magnetization, and a stronger magnetic flux remains.
Strong magnets (e.g. alnico-5) succeed relatively easily in magnetizing the string (almost) up
to saturation – hysteresis-effects not as pronounced: the string cannot be more than saturated
and this condition can only be attained one way. For humbuckers, this is different: while
between the magnet poles the string is – independently of history – saturated as well, the
outwardly directed flux (i.e. the flux directed away from the pickup) is strongly dependent on
the magnetic past. If a new or a demagnetized string is brought closer to the strings, the flux is
more concentrated to the area between the magnetic poles; if the string already had magnetic
contact a stronger flux divergence ensues.

Fig. 5.4.10: Axial magnetic flux through the string for d = 2mm / 5mm (left); string-hysteresis (right).
Fig. 5.4.10 shows the measurement results for a string mounted above the magnet of a single-
coil pickup. The strong flux densities clarify that even for d = 5 mm the string is almost
magnetized up to saturation. This has far-reaching consequences for the alternating flux
which we will discuss further below: the ferromagnetic material of a magnetically saturated
string cannot accept further magnetization and – behaving as if it were located in vacuum (or
air) – shows the same small permeability µ0. Just a few millimeters away from the magnetic
axis, the string already loses its good conductivity for alternating magnetic flux and barely
differs from air in that way! Consequently, the alternating magnetic flux is not transported in
the sting over any significant distance; rather, it leaves the string already after a few
millimeters. The magnetic conductivity of the string is only high in areas where the flux
density is small i.e. in the centre over the magnetic axis. Corresponding results are shown by
measurements relating to the magnetic aperture (Ch. 5.4.4, 5.10.5).
Fig. 5.4.11: Axial flux density through the string for a Gibson-Humbucker 490R. Left: a = un-magnetized
string. b = magnetic poles at 2 mm after touching (d = 0) the string, c = after magnetization of an extended part
of the string. Right: magnet/string-distance = 2, 3, 4mm, each after saturation.
With a humbucker, the string is subjected to two magnetic poles: in the Gibson Humbucker
and its many copies typically the screw is the south-pole while the slug is the north pole
(compare to Fig. 5.4.4). Without a string, a rather weak field (13 mT) exists between the
magnetic poles. However, in contrast to single-coil pickups, the string over a humbucker
bridges almost the entire air-space of the magnetic circuit such that a very strong
magnetization of the string happens between the magnetic poles.

In Fig. 5.4.11 we see the axial magnetic flux in the string for a Gibson Humbucker 490R
(static field, i.e. f = 0 Hz). At 0 mm i.e. between screw and slug there is a large flux density
with little dependency on the string-to-magnet-distance d. Both branches of the B/H-curve are
almost horizontal und indistinguishable, which makes for an independence of the magnetic
pre-history. However, moving beyond the limits of the magnetic poles we find a much smaller
flux density for the un-magnetized string (a). Another striking fact (for the present
measurements) is that although the screw is about 0,3 m closer to the string, it has a smaller
magnetizing effect than the slug,
Fig. 5.4.12: Dependency of the induced voltage level on distance. Engine bench testing, rotating crank .
At least for measurement technology, the hysteresis effects described above must not be
ignored. Fig. 5.4.12 picks up on that theme: starting at d = 2 mm, the distance between pickup
and string is first made smaller, then larger, and then again made smaller for the DiMarzio
DP-184 pickup. The voltage levels obtained on the engine bench show different values for the
same distance – as much as 3 dB in the extreme case. This difference would be well audible
in a direct A/B-comparison.
Now let us take a look at the real world …. for example a look at a test in a commercially
successful music magazine comparing humbuckers of relatively similar sound. The pickups
are installed one after the other in a guitar, and if, incidentally, the person doing the test
arrives at the conclusion that the loudness of the pickups is a little different ...... no, hold it –
the guy will SURELY have taken into account the individual string magnetization. Man, such
a string really goes through a lot in that process: slap it on, then off again, swap the pickups,
slap the string back on ... wait a second, of course first we got to demagnetize it because it got
stuck on one of the magnets of the pickups lying on the bench, now re-magnetize to a
predefined value, o.k. - now slap it on again, do the listening test, take the string off .... and so
on. And all the while keep that de-magnetizing coil (turdus amagneticus) humming. Surely
this ordeal – necessary from what we learned above – is always done? Isn’t it strange one
never reads about it in the tests …. On the other hand, the test description does go to great
lengths and notes that the test-guitar was loaded with the original 1959-Sprague-bumblebee-
foil-potatoes. Well then .....

5.4.3 The alternating magnetic field
The pickup magnet generates a static magnetic flux in the space (the air) around it. This flux
flows from the north- to the south-pole. As the string oscillates in front of the pole-plate of the
pickup, this static flux changes. This can be understood as a superposition of a static magnetic
field and an alternating magnetic field, an approach which is at least permissible in the linear
medium air. While the magnet is a nonlinear system, the relative flux changes are sufficiently
small (1%) to support a linearization with good approximation. Still, the above superposition
must not be misunderstood in the sense that the paths in space of the static and the alternating
flux would correspond! The source of the static field is the magnet; its two poles are separated
by 1 – 2 cm which results in a relatively large path of flux. The main source of the alternating
flux, on the other hand, is the air gap reluctance in front of the pole plate, this gap being
variable due to the string oscillation. Since the associated dimensions are significantly
smaller, the extent in space of the alternating flux is also limited to a smaller sector. (Strictly
spoken, the two flux components of course extend indefinitely – what is meant here are the
relevant field areas). Both the magnet and the string are made of ferromagnetic material – for
this reason one needs to consider the hysteresis when calculating the static component,
whereas calculations relating to the dynamic component require consideration of the
reversible permeability.
A first insight into the spatial distribution of the alternating flux is given by Fig. 5.4.13: along
the abscissa we have the alternating flux through a cylindrical magnet which has a string
vibrating in front of its pole plate. The distance between string and pole plate is 2 mm, the
amplitude of the excursion of the string is 0,15 mm with an excitation frequency of 85 Hz. A
small coil (25 turns of 80µm magnet wire) wound tightly around the magnet samples the
alternating flux. The ordinate in Fig. 5.4.13 presents the distance of this sampling coil from
the pole plate close to the string. Clearly, the alternating fields decreases quickly along the
magnet axis: less than 2% of the alternating field flowing into the pole plate under the string
arrive at the opposite end. The remaining field has exited the magnet ‘along the way’ through
the cylinder mantle. (The term flowing into applies during one half-wave – for the other half-
wave all flux directions are reversed).
For such a field-geometry the induction law should obviously be applied with caution. Not
every turn of a pickup coil wound around the full length of the magnet receives the same
amount of alternating flux! The section of the winding pointing away from the string
contribute much less to the induced voltage while not being without effect: every additional
turn increases the inductivity of the coil and reduces the resonance frequency (with everything
else being kept equal).
The left section of Fig. 5.4.13 schematically shows the flux paths for a Stratocaster-coil. The
static current flows through almost all of the coil but does not contribute to the induced
voltage. The alternating flux exits the magnet already within the first few millimeters and
does not even reach the coil – this actually is astonishing given the fact that this pickup is
considered as the ‚holy grail’ for electric guitars. However, a high efficiency is not the only
development objective for the mechano-electric transmission: the Jazzmaster-pickup with its
large, flat coil had a higher efficiency but was widely rejected due to its different resonance
behavior.

Fig. 5.4.13: magnetic flux for a Stratocaster-pickup (left, schematically). The static flux runs through the whole
cylindrical magnet (––––), the alternating flux mainly circles close to the strings (----). On the right the decrease
of the density of the alternating flux along the axis of the magnet is shown (measured data).
In Fig. 5.4.13 we see the alternating flux as it runs within the magnet. However, only the
innermost windings enclose only the magnet; the more outwardly positioned turns are also
penetrated by the magnetic flux that runs through the air. The flux density in air is somewhat
smaller than that within the magnet, but nevertheless the field in the air must not be
completely neglected. Fig. 5.4.14 shows the spatial distribution of the alternating magnetic
flux density as measured with concentric circular coils. At a distance of 2 mm from the pole
plate of a 5x18 alnico magnet, the steel string of 0,66 mm diameter follows a sinusoidal
movement with an amplitude of 0,15 mm and a frequency of 85 Hz. The local flux density
can be easily calculated from the measured induction voltage; for an improved visualization it
is smoothed via a spline-interpolation. In the left part of Fig. 5.4.14, the maximum of the
color scale corresponds to a flux density of 250 µT. This allows for a good representation of
the flux density within the magnet while the small values of the field in air remain
indistinguishable (green; ≈ 0). Changing the color maximum to 15 µT (right part of Fig.
5.4.14) pushes the values of the field running within the magnet out of range but the course of
the field in air becomes visible. We now see that close to the string (upper part of the figure)
anti-phasic field patterns happen already within a few millimeters. A coil winding enclosing
as well a blue field area does not, however, necessarily generate an anti-phase (i.e. unwanted)
voltage. Of relevance is in fact the whole alternating magnetic flux through each winding, i.e.
the integration of the flux density in the axial direction. Consequently, the induction voltage
generated by the whole coil is given by three spatial integrations: a radial integration (dS =
2πr dr) to include the total flux of one turn, a radial integration over all turns in one plane, and
an axial integration to consider the length of the coil.
Using color-coding, Fig. 5.4.15 shows the spatial distribution of the flux in the winding; its
temporal derivative results in the voltage induced per turn. Close to the string (upper section
of the figure) the alternating flux flowing through the winding increases with a growing radius
of the winding, because the polarity of the alternating field is the same both in the magnet and
the air surrounding it. However, as the radius grows beyond approx. 7,5 mm, the flux through
the winding decreases – the field-polarity in air is in anti-phase to the alternating magnetic
flux in this region. As one increases the distance between magnet and string to 4 mm, this
border shifts somewhat to a larger radius.

Fig. 5.4.14: Alternating magnetic flux density around an alnico-V magnet. The color-coding exemplifies the
distribution of the flux density: the scale for the left section is such that the flux-density distribution within the
magnet becomes visible; the scaling on the right clarifies the flux density in the surrounding air. In the ranges
colored in blue the alternating magnetic field is in anti-phase to the field within the cylinder of the magnet (d = 2
mm). The direction of the field is axial.
Fig. 5.4.15: alternating voltage in the winding dependent on the number of turns and the distance to the pole
plate. The coil cross-sections marked are those for Stratocaster- and Jazzmaster-pickups. On the left, the distance
between string and magnet is 2 mm, on the right it is 4 mm. The numbers entered in the figure have the
dimension µV / turn.
Fig. 5.4.16: for a nickel cylinder (5 mm x18 mm) with two bar magnets (3 mm x13 mm); voltage in the winding
(left) and magnetic flow density (right). d = 2mm.

In Fig. 5.4.15 we find – in addition to the voltages in the windings – the winding cross-
sections for two pickups, as well. It should be noted that for the measurements, circular coils
were used while the delineated pickup coils are of an oblong shape. For the Jazzmaster
pickup, the winding is about 4 – 7 mm away from the string and has a radius of between 2,5
and 17,5 mm. The average alternating flux through the winding was found to be approx. 4
nVs (Fig. 5.4.15, left section) for a distance of 2 mm between magnet and a 0,66-mm-string,
the latter vibrating with 85 Hz and 0,15 mm amplitude. Via temporal derivation
differentiation of the sine-shaped alternating flux a per-winding voltage of approx. 2 µV/turn
(root mean square value) is found. With this approximation, a coil of about approx. 8500 turns
would thus produce an overall voltage of 17 mV. Comparative measurements with an actual
Jazzmaster-pickup with the same excitation yielded 19 mV. In view of the differing coil
geometries and magnets, this difference is quite acceptable – especially since the number of
turns of the Jazzmaster-pickup is only approximately known (being a vintage 1962, i.e. pre-
CBS, it’s sacrosanct in any case).
For the Stratocaster pickup, the integration over the surface for 7650 turns yields about 7
mV. Here the difference between calculation and measurement (10 mV) is somewhat bigger –
however, we again have to deal with the already mentioned differences (magnets, shape of the
coil, number of turns). The aim of the measurements is not to determine the pickup-
transmission-coefficient; this can be done much better with the shaker-test-bench (chapter
5.4.5). Rather, we wanted to obtain an impression of the spatial distribution of the alternating
field which indeed can be seen quite well from the figures. As a comparison, Fig. 5.4.16
shows field measurements for which, instead of a cylindrical magnet, two bar magnets
generate the magnetic field (similar to an SDS-1, Fig. 5.1.3). Towards the string, the field is
focused by a cylinder made of nickel. The higher flux density obtained with this configuration
can be nicely seen, just as the fact that the alternating field penetrates more deeply. Both these
characteristics give a higher sensitivity; possible drawbacks should also be mentioned: higher
inductivity and stronger dampening due to eddy currents (chapter 5.9).
Besides the alternating flux penetrating the coil, the magnetic field of the string is the second
interesting quantity. The strong static flux density was already pointed to – as a consequence
of it the sting is all but magnetized into saturation. The permeability of a saturated
ferromagnetic material is only marginally higher than that of air which is why the string
looses its good magnetic conductivity and does not represent a focusing channel for the
alternating flux anymore. The alternating field leaves the string already after a few
millimeters and returns to the magnet. For a string of 0,66 mm diameter, Fig. 5.4.17 depicts
the axial flux density (f = 75 Hz, = 0,28 mm).
Fig. 5.4.17: String-internal axial alternating flux-density (RMS value). The diameter of the cylindrical magnet is
marked in grey. On the right the normalized drive-dependency of the alternating flux density.

At a string-to-magnet distance of 2 mm we obtain a maximum flux of 33 mT at 3 mm from

the axis of the magnet. Multiplied by the doubled string surface (the flux through the string
flows in both string directions) we obtain an alternating flux in the string of 22,6 nWb. It is
possible to compare this value with the alternating magnetic flux exiting the magnet (Fig.
5.4.13): there, the flux density amounts to 0,68 mT which combined with the magnet surface
yields an alternating magnetic flux of 13,4 nWb. The different string frequency (75 Hz vs. 85
Hz) and the different string excursion (0,28 mm vs. 0,15 mm) need to be considered – the
correspondingly corrected alternating magnetic flux amounts to 21,4 nWb which is a very
good correspondence and a confirmation of the model we have used.
In the right-hand part of Fig. 5.4.17 the dependency of the alternating string flux on the string-
excursion is shown. A linear dependency would lead to the dotted line, however the measured
data increase progressively i.e. in a non-linear fashion. In fact, it is not surprising that we do
not find a perfect linearity here: presumably this is less an effect of the non-linearity of the
magnet’s hysteresis but the distance-dependency of the reluctance of the field in air. In the
normalized presentation which is used in the figure, 10 dB correspond to a peak-excursion of
0,9 mm. The string therefore oscillates between a distance of 1,1 and 2,9 mm from the magnet
which is a relatively large range, but one that is not unusual in everyday guitar practice.
Fig. 5.4.18: left: string-internal axial alternating flux

density for a humbucker (measured RMS values);
above: approximate course of flux.
Fig. 5.4.18 shows the course of the alternating flux for a humbucker. The left part indicates
the RMS-values which by definition always have a positive sign; the direction of the flux is
indicated with arrows for an arbitrary moment. At the lower border of the figure slug and
screw are hinted to facilitate the orientation, however the alternating flux relates to the string
located 2 mm above. In the right part of the figure we see the approximate shape of the flux
which is, admittedly, unfamiliar in its angularity. But how would one make a better drawing?
Via the check-box method? That only works for the plane-parallel field. The field-lines exit
metals perpendicular to the given surface? That only holds for materials with a large µ. The
pickup-field is three-dimensional, without symmetry-planes or -lines. The ferromagnetic
materials in the field are almost saturated in some areas – this complicates an exact
calculation drastically, after all. For these reasons, the figure can only give a rough impression
of the spatial field shape. The humbucker „squints“ a bit outwardly; this was observed for
other measurements, as well. Possibly it is in particular the strong static flux between the
magnet poles which makes for asymmetric alternating-flux reluctances. Clearly observable is
the weak coil coupling: the alternating filed is focused predominantly towards the vicinity of
the pole plates; in the picture only one singe field line penetrates both magnetic poles.

5.4.4 Window of the magnetic field (aperture)
Magnetic pickups pick up the vibration of the string. Instead of pick up the term sample would
also be appropriate; however this is not a sampling in time but one in space: the place- and
time-dependent vibration of the string is captured discretely with regard to place and captured
continuously with regard to time, and it is then transformed into the pickup voltage. As is the
case for all real-world sampling processes, the place-discretization does not happen with
ideal, infinitesimally small extension in space but across a range of several millimeters which
is called the window of the magnetic field or aperture. The pickup so to speak "looks"
through this window onto the string vibration. We find an ongoing speculation about the size
of this window in literature: is it as big as the diameter of the magnet, or rather as big as the
coil extends? Do wide pickups (e.g. the one for the Jazzmaster) have a larger window than
thin ones (e.g. the one for the Stratocaster)? How does the window-width influence the
transmission characteristics?
System theory divides its "world" into linear and non-linear systems, i.e. in less complicated
and more complicated systems. Pickups belong to the latter, unfortunately. Therefore, the
following considerations – which all have their basis in the theory of linear systems – may be
understood merely as approximations. The principle of superposition holds in linear systems
only; it forms the basis for a comprehensive application of impulse response, convolution
integral and transfer function. For small string excursions at least this linearization is justified.
For large excursions of the sting, considerable non-linear distortion should be expected,
however the effects on the transmission frequency response nevertheless are on the small side.
The transmission characteristic of a linear system can equally be described in the frequency
domain and the time domain: in the time domain via impulse excitation and impulse response,
in the frequency domain via excitation by a sine function and by the transfer function [e.g. 6].
For the guitar string, both measurement principles are problematic. The excitation with a sine
function results – due to the almost perfect boundary reflections – in standing waves with
strongly frequency-dependent amplitudes. At the vibration nodes, the latter vanishes; the
pickup cannot be excited here. Simple absorbers such as cotton wool between string and
guitar neck do not give a satisfactory reflection-dampening while efficient absorbers require a
big development effort. An excitation with a short impulse delivers better results but due to
the dispersive propagation requires a dispersive convolution. Completely unusable results are
delivered by a „sampling“ of short, shaker-driven pieces of string: with a magnetic field of
entirely different shape compared to that of the regular long string, the measurements target
an entirely unrealistic situation having nothing in common with the regular operating status.
Motorized test bench
In order to measure the size of the window of the magnetic field without too much effort, the
following experimental setup was developed: in the middle of a string of approx. 12 cm
length and 0,7 mm diameter, a crank of about 2mm length is bent (Fig. 4.4.19). The string is
then fixed to the shaft of an electric motor, such that it can rotate around its longitudinal axis.
The pickup under investigation is mounted to a sledge and can be moved along the string. The
rotating string crank represents a place-discrete, time-periodic excitation, i.e. a local impulse.
The motor speed is immaterial as long as it can be kept constant during the experiment.
Moving the pickup delivers a local response-function a(z).
Fig. 5.4.19: Rotating steel string with crank.

If indeed the pickup were a linear system, and if the crank were limited to a very short range,
then a(z) could be interpreted as local impulse response. Since, however, the excitation
impulse (the crank) has a length clearly very different from zero, a(z) represents a convolution
of the crank k(z) and the impulse response h(z). The result is that there is a tendency to
measure too long a window of the magnetic field.
This measurement technique of course differs from the real excitation: the plucked string has
a transversal wave running along its length while with the method above a crank rotates. For a
freely vibrating string it is not possible to generate a singular impulse excitation, because
displacement location z (axial coordinate) and time t are mutually interlinked via the
propagation velocity. Every generated transversal impulse runs along the length of the string
with high velocity and consequently does not generate a stationary excitation. To generate an
impulse of only a few millimeters, it would – given a propagation velocity of 100000 mm/s -
be necessary to control a frequency range extending considerably beyond 10 kHz (2 mm are
passed through within 20 µs). The transversal wave equations require a predetermined
interconnection of place and time – however, using a rotating crank, we succeed in
decoupling place and time, and obtain a location change as slow as desired.
Fig. 5.4.20 shows measurement results of selected pickups. Stratocaster- and Jazzmaster-
pickups both feature cylindrical magnets; the Stratocaster coil, however, is narrow and tall
(WxH = 13x11) while the Jazzmaster coil is very wide and short (35x4). The P90 coil, as well,
is wide and short, but the magnetic field is generated by two bar magnets positioned on the
side of the coil pointing away from the string; 6 round-headed screws guide the field. The
SDS-1 is of similar construction but incorporates hexagon socket screws. Despite the different
pickup construction, the measurement results are similar. Obviously it is only those string
movements which happen directly in front of the cylinder magnet (or in front of the screw)
that induce a note worthy voltage – the coil geometry has no bearing on the length of the
window of the magnetic field. Still, one must not conclude from these measurements that the
coil geometry is generally insignificant; the transmission coefficient of the pickup (and thus
the vertical position of the normalized curves in Fig. 5.4.20) does depend on the coil-
geometry, but the window shape does not.
Fig. 5.4.20: local window-functions normalized to the same maximum. The width of the Jazzmaster’s cylindrical
magnet is included as a bar at the upper border of the figure. Pickups: P-90, SDS-1, Jazzmaster, Stratocaster.
Lace: cf. Ch. 5.4.7, Hershey-Bar cf. Ch. 5.4.8.

The window functions depicted in Fig. 5.4.20 are place functions. They can be recalculated
into time functions using the phase velocity valid for transversal waves – for accurate
considerations the dispersion would need to be considered. Assuming linear transmission
(which removes us a bit from the actual reality, see chapter 5.8: harmonic distortion), we can
interpret this window shape transformed via the phase velocity as impulse response. The
Fourier transform of the latter gives the magnetic transmission function of the pickup. The
magnetic transmission function is complemented by the electrical transmission function
mainly composed of pickup inductivity and cable capacitance.
The effect of the aperture can be demonstrated using the example of the scanning of a film.
This scanning involves the film (which is blackened depending on the picture) running
through a thin ray of light. The strength of the ray is correspondingly modulated and e.g. a
photodiode can detect this. The remaining brightness of the ray is the average value across the
sampled surface: the thinner the ray, the finer the resolution. If we assume that the film is
blackened with a sine-shaped place function, then the scanning with the ray of light represents
a local averaging which can be interpreted as a convolution in the time domain (as is the case
for every averaging process). The place function (divided by the velocity of the film)
transforms into a time function which – convoluted with the window function – yields the
output signal of the photo diode. In the case that the width of the ray of light corresponds to a
wavelength in the blackening, the averaging is done over a full period and delivers a zero in
the transmission. Systems theory calls the resulting (idealized) system a gap low-pass filter
[6, 7], the sin(x)/x-shaped transmission function is also designated gap function. A similar
situation is found with the magnetic tape [3, chapter 11.2].
For a guitar pickup, using a rectangular window (insensitive – sensitive – insensitive)

represents merely a rough approximation: indeed Fig. 5.4.20 reminds us more of a Gaussian
function. The latter is invariant regarding the Fourier-transform: a spectral Gauss function
(i.e. a Gaussian low-pass) pertains to a Gauss function in time. It would anyway not make
sense to spend too big an effort on the approximation, since the non-ideal impulse function
(Fig. 5.4.19) has an influence, as well. Fig. 5.4.21 shows typical field-transmission functions.
Clearly visible is a string-specific filtering resulting from the string-specific phase velocity cp.
Considering that the transmission range is limited to about 5 kHz due to the pickup resonance,
it is obvious that for a single-coil pickup the window of the magnetic field has little influence
on the transmission behavior.
Fig. 5.4.21: frequency response (real part) of the magnetic aperture function; dispersion is considered.
Left: Stratocaster; distance magnet/string d = 2mm. Right: P-90; d = 4mm.

On top of the axial shift of the offset (the variable of the abscissa in Fig. 5.4.20), there is a
second variable: the distance d between rotating string and pickup. Enlarging the distance
increases the length of the window of the magnetic field which leads to a slight dampening of
the treble. The main change is in the absolute transmission gain (the sensitivity, chapter.
5.4.5).
Fig. 5.4.22: Aperture-low-pass (E2-Saite) for 2mm and 4mm magnet/string distance. Normalized, dispersion is
considered.
Fig. 5.4.22 shows, for two particular pickups, the normalized aperture-filter frequency
response dependent on the distance d between the magnet and the string. As a rule, for
customary distances (approx. 3 mm) the voltage level drops by 3 dB per mm distance increase
(Fig. 5.4.23 ).
Fig. 5.4.23: Voltage level for variable distance d, the crank is directly above the magnet plate. The average
increase is - (3 ... 4) dB/mm. The specific dBV-values are bench-specific.
This figure is reserved for the printed edition.
Using a logarithmic division of the abscissa (as it is done in the right-hand section of Fig.
5,4,23) and adding a fixed value A to the distance d, we obtain straight lines with good
approximation. The distance function therefore is a power-function of the type:
dependency of voltage level on distance
The fixed value Δ came to 0,5....4 mm for the pickups depicted in Fig. 5.4.23; the exponent ψ
was 1,3 ... 2,7.

Contrary to the single-coil pickup, the classic humbucker samples the string vibration at two
sections; for this reason its local window function shows two maxima. In Seth Lover’s
Gibson-Humbucker (and its innumerable copies), a bar magnet located under the coils creates
the magnetic fields which is guided to the strings by 6 screws through one coil and by 6 pins
(or slugs) through the other coil. The distance of these poles directed towards the string
amounts to 18 – 19 mm, the screw-head has a diameter of 5 mm, the slug one of 4,8 mm. Bar
magnet, screws, string, and slugs form an annular magnetic circuit flowing through both coils.
The flux-change created by the string will thus affect both coils – however with different
efficiency due to the considerable degree of scattering. A movement of the string over the
screw induces a voltage predominantly in the coil carrying the screws. The coil with the slugs
receives a part of the alternating field, and also here a voltage is induced, but the latter is
smaller than the one in the coil with the screws. The two coils are connected in series so that
the voltages generated by movements of the same phase add up.
Fig. 5.4.24 shows, for selected humbuckers, the results of measurements taken on the same
test bench as used for Fig. 5.4.20. In all tested pickups the coil fitted with the screws yielded a
smaller sensitivity versus the coil with the slugs. On the right hand side of Fig. 5.4.24 further
measurements for humbuckers of other distances of the pole pieces are depicted.
Fig. 5.4.24: local aperture functions normalized to the same main maximum (coils in series).
Left: typical Gibson-Humbucker, e.g. 490R; to compare: Fender Jazzmaster (----).
Right: Gretsch Filtertron (18mm pole distance), DiMarzio DP184 (7,6 mm pole-distance, ----);
Besides the single-coil pickup having one maximum and the humbucker having two, the
measured aperture functions are very similar. The second coil allows for additional degrees of
freedom in the humbucker: the distance of the poles (abscissa) and the different sensitivity of
the individual coils (ordinate of the secondary maximum).
Customarily the two coils of a humbucker are connected in series and only the summed
voltage is evaluated. Picking up only the voltage of one individual coil (so called split mode
operation) makes the hum compensation disappear. As a general rule, the sound is still not
that of a typical single-oil pickup because the shapes of the magnetic field are different for
single-coils and humbuckers, and also because the pickup resonance is at a higher frequency.
For more details regarding the split operation see chapter 5.9.2.8 (coupling) and chapter 5.10 (measurements).

Fig. 5.4.25 shows a typical window function of the two coils of the Gibson Humbucker. The
solid line marks the level of the more sensitive coil with the slugs, the dashed line represents
the coil with the screws. The vertical distance of the main maximum is typically 2 – 3 dB; the
secondary maximum measured for the individual coil is about 14 – 20 dB lower than the main
maximum. It is not possible to be more precise regarding the secondary maximum. To
achieve a larger dynamic range of the measurement, the string with the offset would have to
rotate smoothly with tolerances within a range of 1/100 mm across a string length of several
cm – this cannot be achieved with elastic steel wire. External to the offset there will be small
eccentricities which would distort the measurement result. Supplementary measurements can
be found in chapter 5.9.4.5.
Fig. 5.4.25: same as in Fig. 5.4.24 but with the

humbucker in single coil (split) mode. The result
for the secondary maximum needs to be
interpreted as 'in principle'; the measurement
accuracy is mediocre at best here. Typically the
secondary maximum is about 14 – 20 dB below
the main maximum.
The frequency response of the humbucker-aperture-filter is obtained the same way we have
done it for the humbucker: via the Fourier transformation (linearity provided). The regular
humbucker setup (both coils in series) samples the string at two areas. The second maximum
(provided by the second coil) can be seen – in the time-domain – as a repetition of the first,
this leading according to the displacement law of the Fourier transformation to a comb-filter
frequency response (Fig. 5.4.26). The interference gap in the frequency response corresponds
to the two humbucker poles being at a distance of half a wavelength: the string moves away
from the one pole but moves towards the other. If both coils have the same sensitivity, the
cancellation (at the corresponding frequency) is complete. For the listening sensation it does,
however, not make any significant difference whether the gap is 15 dB or 25 dB deep.
Fig. 5.4.26: calculated frequency response (real part) of the magnetic transmission function, with dispersion
considered. Left: Gibson 490R, both coils in series. Right: DiMarzio DP-184, series connection.

More important is the characteristic between 0 dB and about -10 dB: here it is clear that in
particular for the bass strings (E-A-D) a highly significant treble loss happens. Humbucker
with a smaller distance (e.g. DP-184) between the poles show a reduced but still clearly
audible treble loss. However, it is noted here once again that a pickup is not a measurement
device which would have to display a frequency-independent transmission characteristic. The
comb-filter response must therefore not be seen as a fault and its effect can only be evaluated
on a subjective basis.
Fig. 5.4.27: left: normalized aperture window, string/magnetic-pole distance –––– 2mm, ------ 4mm.
DP-184 (top), Gretsch Filtertron (middle), Gibson 490R (bottom).
Transmission behavior for the E2-string (right column).

There is a two-fold influence of the string-to-magnetic-pole distance d on the magnetic

window: on the one hand the shape widens in the maximum (just like it does for a single-
coil), but on the other hand the distance between the maxima changes (due to the divergence
of the field). Both effects lead to an increasing treble loss with increasing distance (Fig.
5.4.27).
For the humbucker in single-coil configuration (Fig. 5.4.28) the interference is not as
pronounced compared to the series circuit but still measurable. Despite the disconnected
second coil the string continues to be scanned in two positions – because of the coupling via
the magnetic field.
Fig. 5.4.28: Humbucker in single-coil configuration.

The window-side-lobe is 14 dB below the main
maximum in this example.
The pole-screws were almost fully tightened flush with the coil bobbin for the measurements
presented so far. The distances between the string and the slugs were thus approximately
equal to the distances between the string and the pole-screws. Unscrewing individual screws
allows for adjusting the loudness of individual strings: the smaller the distance, the louder the
string. Fig. 5.5.29 shows the aperture functions for a Gibson Humbucker (490R). The
distance between the slug and the string was 3,8 mm for both measurements. First, the screw-
head protruded 0,3 mm out of the bobbin (solid line), then – for the second measurement
(dashed line) – the screw was un-tightened two full turns (leading to a protrusion of 1,8 mm).
The distance between string and screw decreased from 3,5 mm to 2,0 mm while the
sensitivity grew by 7 dB. Interestingly, un-tightening the screw increases the sensitivity of the
coils fitted with the slugs, as well (again due to the magnetic field coupling).
Fig. 5.4.29: change of the aperture-function

dependent on the position of the screw. The left-
hand maxima relate to the slug, the ones on the right
to the screw.

The dependency of the pickup output on the string-to-screw distance is shown in Fig. 5.4.30
for a Gibson Humbucker. Whether the rotating crank on the string is positioned – for the
measurement – over the coils fitted with the slugs (a) or over the coil fitted with the screws
(b) does not make a difference in principle; it is merely the absolute sensitivity which differs
by about 2 dB. If the crank rotates above the coil with the slugs, a 15 dB lower output level is
measured in the coil with the screws. As a comparison the dependency of a single-coil pickup
(Gibson P-90) is also shown (dashed line); the main difference is in the absolute sensitivity.
This must, however, not be interpreted such that the P-90 would have double the sensitivity of
the 490R; due to the chosen excitation location only one of the coils of the 490R receives an
input. For low-frequency transversal waves exciting both coils at the same time and in sync,
both pickups have approx. the same sensitivity (see shaker test bench).
Fig. 5.4.30: voltage level dependency on variable

distance; Gibson 490R (motorized test bench)
a = level of coil with slugs, offset over this coil.
b = level of coil with screws, offset over this coil.
To compare the distance dependency of the P-90
pickup is taken from Fig. 5.4.23 (dashed line)
The measurements done using the motorized bench test show without any doubt that the
width of the window of the magnetic window (the main aperture) is not determined by the
coil but by the pole of the magnet. An effective aperture width of approx. 1 cm creates a slight
treble loss for the single-coil pickup; the loss becomes larger as the string-to-magnet distance
is increased. Supplementary investigations suggest that the magnetic pole pointing away from
the string also creates a (secondary) aperture. The motorized test bench does, however, not
allow for a sufficient exactness to check this. Laser measurements in combination with
calculations (see ch. 5.10.5), on the other hand, resulted in robust results supporting the
assumption that the secondary aperture is responsible for a broad treble-loss (approx. 1 – 2 dB
above approx. 1 kHz. The effect of this secondary assumption is more pronounced (chapter
5.4.7) in pickups with field-focusing guides (such as the Fender Jaguar).

5.4.5 Absolute Pickup Sensitivity
The results obtained using the motorized test bench offer conclusive indications about the
local sampling of the string; the results are, however, based on a movement which is untypical
for a string (rotation rather than transversal wave). In order to obtain supplementary data
regarding the absolute pickup sensitivity, the same pickups were investigated again using a
shaker test bench. An electromagnetic shaker (B&K 4810) served as drive, causing a string
of 10 cm length and a diameter of 0,66 mm to vibrate with a sine movement. The string was
positioned orthogonally versus the magnetic axis, moving closer to and further away from the
magnet, respectively. This can be seen as an excerpt from a very low-frequency, level-
polarized transversal wave. An accelerometer served as sensor to capture the measurements;
most of the investigations were done in the frequency range between 80 and 95 Hz with a
displacement-amplitude of approx. 0,4 mm.
Fig. 5.4.31 shows the dependency of the measured voltage level on the width of the gap
between the magnetic pole and the string (the distance d); this gap was varied between 5 mm
and 0,5 mm. The results for the single-coil pickups can be divided in three groups: Telecaster
and Stratocaster (relatively shallow curvature), SDS-1 and P-90 (stronger curvature), and
Jazzmaster (flatter evolution). In this kind of measurement, the SDS-1 proves to be 10 dB
more sensitive („louder“) than the Stratocaster Pickup. On the right hand side of Fig. 5.4.31
we find the results for humbuckers. 490R represents the typical Gibson-humbucker; similar
dependencies could be found for the 57-classic and ES-335 pickups. The Toni-Iommi-pickup
differs from the 490R for small distances – this can be traced to a different construction.
The pickups are most sensitive with their magnetic pole axis pointing in the same direction as
the movement of the string. In the guitar this corresponds to a vibration plane perpendicular to
the to the fret-board. String-vibrations in parallel to the fretboard induce next to no voltage
(chapter 5.10).
Fig. 5.4.31: voltage level dependent o the variable distance d. Displacement amplitude = 0,4 mm, frequency =84
Hz. In practice, the distance d often 3 – 4 mm, the associated gradient is -2 ... -3 dB/mm.
These figures are reserved for the printed version.
The different sensitivities are predominantly due to the various types of coils and their
distance to the string. The distance d marks the clearance between string and magnetic pole;
large magnet protrusions (such as for the Stratocaster) require the coil to be further away from
the string compared e.g. to the Jazzmaster. The following table summarizes the measurement
results.

Tonabnehmer §)
DiMarzio SDS-1
Gibson P-90
Rockinger P-90
Duncan APTL-1 (Telecaster-Type, Bridge)
Rockinger Strat-Type (Balkenmagnet)
Schaller
Fender Stratocaster (Balkenmagnet)
Fender Stratocaster (USA Standard, Bridge)
Ibanez Blazer
Joe Barden Strat-Type (Bridge)
Fender Telecaster Texas (Bridge, D / A)
Fender Telecaster-70 (Bridge, mit Platte)
Fender Stratocaster (USA Standard, Middle)
Fender Telecaster-70 (Bridge, ohne Platte)
Fender Noiseless Stratocaster (Neck, G)
Lace-Sensor gold
Fender Stratocaster-72 (G)
Gretsch HiLoTron
"Telecaster"-Fake (Neck)
DiMarzio DP-172 (Telecaster-Type, Neck)
Fender Telecaster-73 (Bridge, D / A)
Duncan APTR-1 (Telecaster-Type, Neck)
Fender Telecaster-52 (Neck)
DiMarzio DP-107 Megadrive
Gibson Burstbucker #2
Gibson 57 classic
Gibson 490R
Squier Humbucker
Gibson ES 335 (Neck, 1968)
Gibson Tony Iommi
Gibson ES 335 (Bridge, 1968)
DiMarzio DP-184
Gretsch FilterTron
Table: low-frequency pickup transmission-coefficient TUv.

String diameter = 0,70 mm (plain), distance to the magnet pole d = 2mm.
§) The numeric values are reserved for the printed version.

5.4.6 Staggered and beveled polepieces
When picked with the same strength, the six strings of the electric guitar are supposed to
generate an approximately equal voltage in the pickup. This requirement is met by a piezo
pickup, but by a magnetic pickup not so much: if all 6 strings were constructed of solid
material, the E2-string would yield 4 times the output of the E4-string (chapter 3.2). However,
the winding around the lower strings is rather inefficient in terms of magnetism, and thus the
bass strings produce roughly the same loudness as the treble strings. For reasons of clarity, we
will in the following not look at loudness (which is dependent on numerous factors) but at the
level of the fundamental of the string: Fig. 5.4.32 shows the results for nickel-wound Fender
strings. Basis for the measurement is an identical picking strength for all 6 strings. For solid
strings the level is, for this case, only dependent on the fundamental frequency of the string♣
(dotted line). The E4-, B-, and G-strings are assumed to be solid, while the remaining strings
are taken to be wound, showing 4 – 10 dB less pickup output compared to the solid strings.
(chapter 3.2). The dashed line gives the level for a wound G-string matching within the set.
Fig. 5.4.32: level of the fundamental of the strings, Fender-150 (pure Ni-wrap): 42-32-24-16-11-09. Dashed line:
with wound G-string. Left: equal string-to-magnet distance and equal pickup sensitivity for all strings. Right:
convex string action across the neck as is typical for Fender; dotted line: boundary effects of the pickup.
When comparing the output level of the strings we need to weigh several effects: the
magnetic efficiency of the strings (chapter 3.2), the distance between string and pickup, and
the sensitivity associated with the individual pickup magnets. Due to the curvature of the
fretboard (with a radius between 18 and 30 cm), the strings are not located in a plane but
along an arch. In most scenarios the E2-magnet shows a 1mm-larger distance to the string than
the E4-magnet, this leading – as an example - to the following string curvature: 1,0 – 1,5 –
1,7 – 1,5 – 0,9 – 0,0. For the string-specific pickup sensitivity we need to consider, on the one
hand, the individual static magnetic field which can easily vary by 10%, and on the other hand
the reduced sensitivity of the pickups for the outer strings (E2 and E4) typical for Fender
pickups: this will be 1,5 – 2,5 dB less compared to the inner 4 strings, conceivably because
the coil winding captures only part of the magnetic field of the string in the edge region. In
summary, we arrive at individual level differences with small loudness deficits for the D- and
G-strings, and a B-string that is a bit louder. The level differences between the strings are not
dramatic but did lead to corrective measures: to compensate for level- and thus as well
loudness-differences, Fender modified – as early as the 1950’s – the magnet lengths such that
the softer strings are subjected to a stronger magnetic field. These magnets protruding more or
less far out of the pickup housing were called staggered polepieces, as opposed to flush
♣
Given these conditions, a set of higher-gauge strings is not louder, because the higher required tension reduces
the string displacement and thus also the string speed

polepieces which are also called level polepieces. Not all guitars received staggered pole-
pieces: the Jaguar and the Jazzmaster (then considered the flagships of the line) sported flush
polepieces while the Stratocaster had staggered polepieces. Opinions about the principle
according to which the magnet protrusions should be arranged seem to have differed over the
years: the D-magnet was longest at some point, then the D- and G-magnets were of the same
length but longer than the others, then again all 6 magnets were of the same length, then again
they were staggered. Fig. 5.4.33 shows some of the designs, without any claim to complete-
ness.
Fig. 5.4.33: different-length magnet protrusions in Fender pickups. Upper left: flush polepieces, upper right:
1972 Stratocaster, lower left: 1973 Telecaster, lower right: 2004 Stratocaster („noiseless“).
N.B.: „H“ (German) = „B“ (international)
The 1972-Stratocaster-pickup investigated for the example had extended D- and G-magnets, a
shorter B-magnet and a slightly shortened E4-magnet♣. This configuration leads to the level
dependencies shown in Fig. 5.4.34 – indeed a visible improvement over Fig. 5.4.32 –
especially with a wound G-string, as it was the standard in the 1950’s when the first Fender
guitars were built! As late as 1968, the Fender brochure indicates for the 1500 string set: 12–
16–26w–34–44–52 this set supplied on all new instruments except ¾. Alternatively the "light
gauge rock 'n roll" string set was already available (gauged 10-13-15-26-32-38 and with solid
„unwound“ or „plain“ G-string) – the wound G-string was still standard, however. When
thinner strings with a solid G-string became the new standard, the old magnet-protrusion-
profile did not fit anymore. The solution was typical for musicians: newer pickups have the
Fig. 5.4.34: level of string fundamental, Fender-150 (pure Ni-wrap): 42-32-24-16-11-09. Dashed line: with
wound G-string.. Left: '72-Stratocaster, right: Noiseless Stratocaster (2004). Convex string action.
♣
In old Stratocaster-pickups the magnets were mostly flush on the lower pickup-side. Here is an example in
which presumably 2 of the magnets were moved . NOT RECOMMENDED: RISK OF DAMAGE!

G-magnet protrude only little, but "vintage pickups“ using the old profile are available new,
as well. Not only a few guitar players request the vintage profile ... but still mount the light
gauge strings with the "plain“ G ....
How significant is a level difference of 3 dB? From a pure signal-theory point-of-view an

increase of 3 dB ties in with a doubled power i.e. 200 W instead of 100 W. That would be
quite substantial. On the other hand: Johannes Webers writes in his book on studio electronics
("Tonstudiotechnik“, Francis, Munich) that the attenuation-per-step in stepped level controls
typically amounts to 1,5 dB – this would be correspond roughly to the smallest discernible
loudness difference. 3 dB would thus be twice such a minimum step: perceivable in direct
comparison but not really a very big deal. Seth Lover, developer of the Gibson Humbucker,
remembers: "My PAF prototype ... worked well. When the salesmen saw this, without any
adjustment screws, it was like breaking their arms. They just didn't have anything to talk
about. So, next came the punched-out holes and the adjustment screws." [Vintage Guitars,
Feb. 1996]. Business as usual, then: sales has to straighten out mistakes made in R&D ... or
was it the other way round?? A later development in the Gibson product line, the Tony Iommi
pickup, lacks the adjustment screws again. The times they are a-changing. Or Greek-
orthodox: panta rhei.
Of course, the adjustment screws give power to the guitar player, and individuality to his or
her instrument: "only after I had turned the second screw a quarter-turn counter-clockwise I
suddenly got this awesome sound“. Immediately, however, the maestro runs into the next
problem: if he doesn't tell anyone, his genius remains unrecognized. If he does tell, they all
can copy his awesome sound. An improved statement, then: "of course I first need to fine-
tune every guitar I receive from the manufacturer: those guy deliver such shitty stuff – even
from the custom shop, it's unbelievable. However, with my extremely sensitive hearing I got
every Custom to sound great. It's just that there are so many years of hands-on experience
involved that can't really relate it all". O.k. then ... keep them screws turning. Incidentally,
Jimi Hendrix did not modify the pickups in his Stratocaster whether or not he had access to a
lefty and had to restring a righty. "We don't need another hero ..."
Next to staggered magnets the other specialty are beveled magnets. These are tapered like a
truncated cone on the side pointing towards the string (45°-bevel, Fig. 5.4.33, Noiseless
Stratocaster). One might speculate whether the pickup assembly (the press-in operation) could
be done more easily, or whether Leo Fender was hoping for a stronger magnetic field.
Measurements with turned magnets in a Noiseless-Stratocaster yielded practically no
difference: on average the "improvement" of the response of 0,2 dB is within typical
measurement tolerances and insignificant. For the harmonic distortion, as well, no difference
could be found relative to Stratocaster-pickups with strictly cylindrical magnets. The theory,
too, fails to point to any big differences: in the range of the facing edge (i.e. the intersection
between cylinder barrel and the end surface of the cylinder), the flux-density of the cylindrical
magnet is very high; the magnetic material is in saturation and consequently rather
inefficient
It is not recommended to "sharpen" the cylindrical magnets. The sole possible working-
method would be to grind them – however this would involve extreme heating of the
magnetic material which can lead to a lasting change in the magnetic properties (watch den
Curie-temperature!).

5.4.7 Fender: Jaguar and Lace
Both the Lace pickup distributed by Fender and the Jaguar pickup generate – with the aid of a
u-shaped yoke – a special magnetic field which shall be further investigated in the following.
That Lace advertising tireless tries to convince us that the Fender Lace Sensor is not a pickup
but an "acoustic emission sensor" is merely typical sales mumbo-jumbo: every pickup is a
sensor, anyway. However, at the same time the ads claim that the Lace has the ability to
accurately reproduce the sound characteristics of any existing conventional pickup. Now that
is going a few steps too far and would seem to be quite a put-on. Just looking at the Lace
patent (US 4,809,578), i.e. using Lace's own reasoning, casts some serious doubts: the patent
notes that all other pickups dampen the string vibrations – it's just the Lace that doesn't. If so,
it does NOT reproduce the characteristics of all other pickups – in fact it lacks at least that
one.
Over the years, the Fender company tried several times to bend the magnetic field of single-
coil pickups, starting with the base-plate of the Telecaster bridge pickup up to the pickup with
a ceramic magnet patented in 1980. Leo Fender was of the erroneous opinion that the more
string-length is sampled, the better the sound would be, and for this reason the Jazzmaster
receives a pickup with a particularly coil and the Jaguar pickup a special yoke. In agreement
with this kind of thinking, the patent for the Jaguar discloses that in regular pickups, the
magnetic field lines pass through only very small portions [of the string], with small
harmonic content. In contrast, the teethed yoke of the Jaguar pickup is supposed to magnetize
a approx. 2 cm long area (Fig. 5.4.35), and for the Lace pickup the magnet strips allegedly
push the magnetic field outward, i.e. they make it broader (Fig. 5.4.36). But aren't the aperture
width and the frequency bandwidth in a reciprocal relationship? Of course they are: the
shorter the sampled piece of string, the better the treble reproduction – that's also exactly why
the old tape recorders had the smallest possible magnetic gaps in the tape heads.
Fig. 5.4.35: Fender Jaguar pickup [www.guitar-parts.com, www.jimshine.com]; the teethed u-shaped "claw"
leads a part of the magnetic flow returning from the string back to the south-pole
Fig. 5.4.36: Lace-Pickup [Fender-Actodyne]. The ferromagnetic coil bobbin has a teeth-shaped upper side to
generate a magnetic field "as inhomogeneous as possible". The distance of the teeth has no regular relation to the
distance of the strings (in the middle section above two typical cases are hinted: top: 51 mm, bottom: 49 mm) .

Luckily, the magnetic field lines ignore the patent publication for the most part and instead
follow the laws of physics when seeking their path. Fig. 5.4.37 shows, in its left section, the
magnetic window of the Lace pickup measured with rotating string, while the right-hand
section depicts the aperture-frequency response derived from the window. Yes indeed, there is
a difference to the Stratocaster pickup, but the treble-loss is still limited, as also verified via
the transfer measurement using the laser-vibrometer (Fig. 4.5.38, measurement setup as given
in chapter 5.10.5).
Fig. 5.4.37: left: aperture of the magnetic field; right: aperture-frequ.- response (E2, with dispersion, b = 1/8000).
The impedance-frequency-response (Fig. 4.5.38) reveals further differences: the yokes lead to
stronger eddy current losses and consequently the emphasis of the resonance in the Lace
pickup is a bit less than that of the regular Stratocaster pickup. Similar differences can easily
be achieved as well via changes on the resistance of the connected potentiometers, and thus
Lace and Stratocaster pickups are very similar regarding their transmission. There are,
however, big differences in the sensitivity to hum (chapter 5.7) and in the strength of the
magnetic field – the latter is about 60% less than that of the customary Stratocaster pickup.
That's not really "Leo-compliant" since he thought it to be patentable to generate – in the
Jaguar pickup – a magnetic field stronger than that of conventional pickups. Conversely, the
allegedly patentable subject matter in the Lace is a magnetic field weaker than that of
conventional pickups. Who would have thought .....
Fig. 5.4.38: left: frequency response measured with the laser-vibrometer. right: impedance-frequency-response.
Two specimen of the Lace were analyzed. Noiseless. (Noiseless = Fender Noiseless-Strat-Pickup).
Now then: is the Lace good or bad? In a nutshell: the advertising may be dubious but the
pickup is quite o.k. It features a good hum rejection♣ with only a slight treble loss.
♣
However, the Fender Noiseless-Strat-pickup shows an improvement of another 13 dB in its hum rejection.

Incidentally, Mr. R. Blackmore responded to the question whether he was happy with the
Lace: "well ... sure – would I use it otherwise?" (in a German music magazine in May, 2005).
Seems the interviewer was actually in luck that he didn't get smacked. By the way, it appears
that both have fallen out of fashion a bit: Fender almost never installs the Lace anymore, and
that Blackmore guy ... who was that again ... anyway, there's probably enough lace of the
other kind in Blackmore's Night.
The Jaguar pickup is not in a front-row position anymore, either, despite it being better than
its reputation. If indeed the u-shaped yoke would generate a 2-cm-wide aperture window, it
would face a significant treble loss. As it is, nothing really changes much. That is connected
to the fact that, contrary to the patent, Fender does not mount the yoke directly and without
any gap to the magnet but leaves a 1-mm-wide annular air-gap (Fig. 5.4.35). Off to the patent
office right away, and only afterwards do some testing ... ain't that so, Leo? Without the air-
gap, microphonics could take over too much, and that's not what we want, do we? And then:
the magnetic field doesn't have to be that strong, anyway, and we can make the yoke a bit
thinner than in the patent, and shorten it by tow teeth, and change (1964) to staggered magnets
... it's a fit!!
Measuring the impedance (Fig. 5.4.39) shows 3,8 H with the yoke and 3,15 H without it; that
is more than for the "normal" Strat pickup which had approx. 2.2 H. The DC-resistance is
higher than that of the Strat (6,8 kΩ versus 5,7 kΩ) which indicates a larger number of turns.
The ferromagnetic yoke increases the inductivity but also reduces the emphasis of the
resonance due to the resulting eddy currents. The main differences to the Strat-pickup are: the
resonance frequency is lower, the resonance emphasis (Q-factor) decreased, but on the other
hand the Jaguar pickup is louder and receives significantly less hum (chapter 5.5, 5.7)
Fig. 5.4.39: Impedance-frequ.-response (--- = w/out yoke). The transfer is for a 333 kΩ load (amp = 1 MΩ).
Comparing both guitars divulges a further peculiarity: for the Strat the pots have 250 kΩ
each, for the Jaguar 1 MΩ each! That's why the resonance emphasis for the Jaguar in fact
even bigger. However, the 1-MΩ-pots are not really purposeful: turning down the volume just
a bit all the treble is lost (chapter 9). But that's not all, folks: the Jag holds a secret which has
occupied the fan community for decades: why are two teeth shorter and which way 'round
should the pickup be installed? It appears that even in the Fender company there was
controversy about this, and the shorter teeth were installed underneath the E2- und A-string ...
but also underneath the H- und E4-Saite. Had the issue been solely the loudness of the
individual strings, it would have been solved by the staggered magnets. Probably there was
the wish to give the two bass strings more brilliance. Not a bad thought in principle – however
the improvement is only a few tenth of a dB, and we can check off the issue. Myth busted ....

The comparison between the calculated transfer function (HUv chapter 5.9.3) and
measurement with a laser (chapter 5.10.5) show a slight treble loss (Fig. 5.4.40), the cause of
which quite surely is the special magnet aperture. The left hand part of the figure shows the
local weighting function belonging to the transfer function – it is obtained via the inverse
Fourier transform. The saddle-shaped drop around 5 kHz is a consequence of the secondary
maxima of the aperture function: without the secondary maxima the transfer function has the
shape of the dotted line.
Fig. 5.4.40: Jaguar pickup, left: aperture frequency response (–– with & --- w/out second. maxima); right: local
weighting. Wound string, outer diameter = 1,1 mm; string-to-magnet distance = 4 mm, f = 82 Hz for the 65-cm-
scale. The dimensions of magnet and heads of the "teeth" are indicated in grey at the bottom of the diagram.
For the analyzed Jaguar pickup, the magnetic field enters the string over the pole (N) and
exits it again from approx. 7 mm (compare to Fig. 5.4.8). The flow back to the south-pole
generate the secondary maxima of the aperture function which are located a small distance
outside of the "teeth". The u-shaped yoke including the teeth is able to focus these flows
somewhat; this causes the saddle-shaped treble loss – in addition to the reduced sensitivity to
hum. Of course, without the yoke with its teeth, the flow back to the south pole is also present
– but it is more distributed in space and thus with less attenuation of the treble. The secondary
maxima show up in measurements at -40 dB but can be determined only as an approximation
since the measurement accuracy is dropping considerably from 5 kHz up.
Fig. 5.4.41: left: Jaguar pickup without the teethed yoke, otherwise as Fig. 5.4.40; right: measurements with
teethed yoke, above the D-magnet (–––) and the A-magnet (----, shortened "tooth"), respectively.
In Fig. 5.4.41 we see the transfer function without the teethed yoke on the left; on the right
the treble gain caused by the shortened tooth shows but it's in fact, not worth mentioning. The
shielding, however, is quite a success and reaches second place of the investigated (true)
single coils.

5.4.8 DeArmond pickups
Harry DeArmond (Ohio) was one of the pickup-pioneers: as early as the 1930's he developed
magnetic pickups and sold them via his business partner H. Rowe to many guitar manufac-
turers. Common were at that time flattop and archtop acoustic guitars which could be
"electrified" with a pickup. If they sported a round soundhole, the pickup was mounted in
there, if they had f-holes, a pickup as flat as possible had to be installed between top and
strings (e.g. fitted to the pickguard or the end of the neck). DeArmond's FHC was attached to
a rod running parallel to the strings, its position could be correspondingly adjusted between
neck and bridge. A difficulty encountered with this retrofit of pickups related to the loudness
of the individual strings. The "plectrum guitars" used back then did already use steel strings
but he lower 4 strings (EADG) were wound with brass or bronze. Here, only the thin steel
core is magnetically active and the voltage induced in a magnetic pickup is much lower than
for the two solid top strings (chapter 3). DeArmond solved that problem with a very special
magnet design for which he even obtained a patent: the bar magnet under the coil is not
continuous but has a gap under the B-string. Above the coil two ferromagnetic metal strips
focus the field (A, C, Fig. 5.4.42); a metal bridge (B) attenuates the magnetic field further.
Fig. 5.4.42: DeArmond FHC (US-Patent 2,455,046).
The invention does meet its purpose: the B-string is picked up 8 dB weaker than the bass-
strings, the high E-string features a 5 dB drop. The static flux density (measured 2 mm above
the lid of the housing) is – at 17 mT – relatively weak; strong single coils easily reach triple
this value. There is another difference in that the aperture of these "other" single coils is
narrower (chapter 5.4.4). Fig. 5.4.43 shows the aperture windows compared to the
Stratocaster pickup. There is little effect of the extended width of the aperture of the B-string:
the wave velocity of the latter is relatively high (chapter 5.4.4). However, for the bass strings
there is a loss of brilliance. The dominant treble absorber is the ferromagnetic sheet mounted
below the pickup: the eddy currents generated in it (chapter 5.9.2.4) have hat effect of a
pronounced treble loss.
Fig. 5.4.43: Aperture window; DeArmond FHC (left), Rhythm-Chief and Hershey-Bar (right).

The relatively strong eddy-current losses also show up in the frequency response of the
impedance (Fig. 5.4.44). The inductivity is rather strong but the resonance emphasis only
weakly developed. The broken-line curve indicates that the (non-magnetic) cover of the
pickup housing reduces the Q-factor, as well. For the impedance the effect is small but fort he
transfer function strong. An even more dramatic treble loss results from loading the pickup
with a potentiometer – back in the day this device often had only 50 kΩ which killed the
treble completely: 8.2 H and 50 kΩ yields a 1-kHz-lowpass having its effect on top of the
aperture- and eddy-current-losses (in Fig. 5.4.44 it is not even considered yet). Still: that's the
"golden tone" for which these pickups are sought after and change hands for substantial
amounts of money.
Fig. 5.4.44: Impedance. No load (black), w/load of 330pF (blue); Jazzmaster for comparison (red). The broken
lines show the frequency response of the impedance as it is measured without pickup cover. The right part above
shows the transfer measured with the laser vibrometer (chapter 5.10.5).
A further development based on the FHC is the Rhythm-Chief. Early variants were given a
divided winding with a reduced number of turns below the B- and E-string to compensate for
loudness as indicated above. The next step, the Rhythm-Chief 1100, features adjustable pole
screws. Watch out: these work rather differently than e.g. in a P-90. The particularity starts
with the magnet: for DeArmond this often is a plastic magnet (also called rubber magnet).
Despite the name the magnetic active substance is a metal powder which is molded to shape
using plastic or rubber as binder. In the RC-1100 the magnet consists of the whole (oblong)
coil core and the screws are inserted into it. This is indeed very unusual, since the screws are
directed in parallel to the magnet and short-circuit it partially. Two cases are shown in Fig.
5.4.45: if the screws are deeply inserted (second section of the figure from the right), the
magnetic circuit is closed mainly via the screws and the external field is relatively small.
Unscrewing the screws to a large extent (as shown in right-most section of the figure) renders
them field-focusing and -amplifying. In the end the result matters, and indeed: yes – it works!
And even with a little less treble loss than in the FHC. The RC-1100, as well, has the
dampening effect of the eddy currents and also the non-negligible aperture dampening (Fig.
5.4.43). The connecting cord is fastened rather amateurishly and easily torn off – which the
collectors are not too unhappy about since the collectors value of surviving specimens
increases .... to presently approx. $ 1200. Trend: going up.
Fig. 5.4.45: DeArmond Rhythm-Chief 1100.

In Fig. 5.4.46 the impedance frequency responses are shown. The "naked" coil with the
plastic magnet inserted in it has a high Q-factor. Installing the 6 screws increases the
inductivity (right section of the figure). An even bigger push towards more inductance is
generated by the ferromagnetic bottom plate (left part of the figure), but this component also
reduces the Q-factor by a considerable amount (eddy currents)
Fig. 5.4.46: impedance DeArmond Rhythm-Chief 1100. Left: original condition, w/out load and w/330 pF load,
respectively. Right: coil w/out housing, w screws (––––) and w/out screws (-----).
The Rhythm-Chief has directly attached to it a small control unit (volume and tone controls
plus a lead/rhythm switch). In Fig. 5.4.47, the frequency response of the unloaded pickup is
shown in black while the condition with a load Cload = 330pF is shown in blue. In contrast to
the very low-impedance controls they used elsewhere, DeArmond suddenly switches to high
values here aiding a better treble response – as long as one does not turn down the volume.
Circuitry of the DeArmond control unit
Fig. 5.4.47: transfer frequency response for the Rhythm-Chief 1100; Telecaster Bridge-pickup fro comparison.
A much simpler representative of the DeArmond pickup line is the so-called Hershey-Bar
(named of course after the well known chocolate bar). Take a flat, rectangular plastic magnet
with a coil wound around it, fix it to a ferromagnetic base plate, slam on a non-magnetic
cover – done. Just 7 mm tall, no adjustment possibilities, no treble – perfect. O.k., not perfect
for everybody but this pickup, as well, found its fans. The magnetic window is about as broad
as the one of the Rhythm Chief (Fig. 5.4.43), and the flux density is (at 19 mT measured 2
mm from the pickup) about as weak as with the FHC, but the coil has either fewer turns or a
bigger wire: the DC-resistance is only 3,8 kΩ versus 9.7 kΩ (FHC) and 14 kΩ (Rhythm-
Chief), respectively. Interestingly, the Rhythm-Chief is the softest of the three: still about 2
dB more sensitive than the Strat pickup (used as reference, chapter. 5.4.5), but the Hershey-
Bar is 4 dB more sensitive and the FHC even 9 dB. This again shows that the DC-resistance
has little bearing on the transfer coefficient (see also Fig. 5.5.19).

Fig. 5.4.48:
DeArmond Hershey-Bar
Hershey-Bar-measurements are shown in Fig. 5.4.49. The original-accessory-volume-control

has merely 50 kΩ and significantly cuts the treble. Those desiring more treble can switch to a
250-kΩ-pot without any issues.
Fig. 5.4.49: DeArmond "Hershey-Bar". Left: impedance frequency response, w/out load and w/330 pF load.
Loaded w/original 50-kΩ-potentiometer (––––); w/out potentiometer (-----).
Right: transfer-frequency-response; Telecaster-bridge-pickup fro comparison (330 pF, 0 pF).
To complement the information about these rather special pickups: 1) designations such as
FHC or Guitar-Mike are not unambiguous, they specify merely a group of similar but not
identically constructed pickups 2) Such old pickups may have incurred shorts in the winding,
or a torn off connecting wire. 3) Because the pickups were often defective, there are many
that where repaired somehow but failed to regain the original state after the repair. 4) Some of
the pickups are attached to very long cables, and the latter may have significant losses
capacities (e.g. 250 pF/m). 5) The aperture attenuations measured via the laser-vibrometer are
string dependent! 6) And just to mention it: enthusiasts willing to pay in excess of $ 1000 for
a pickup might inspire obvious ideas ....
In closing here a look at the signal-to-hum ratios (chapter 5.7): FHC = 3 dB better than the
Strat used as reference, Rhythm-Chief 1100 = 4 dB worse, Hershey-Bar = 2 dB worse.
http://theunofficialmartinguitarforum.yuku.com http://www.harmonycentral.com
Fig. 5.4.50: DeArmond pickups: Rhythm-Chief and FHC.

5.5 Elementary Pickup Parameters
The market offers a large number of different magnetic pickups which differ in basic
construction, in their dimensions and in the transmission behavior. Some of the electric
parameters can be measured easily – these are therefore often listed in overview tables and
connected to sound attributes such as: brilliant, muffled, loud. Of course, the pickup in itself
does not generate a sound – that requires a vibrating string, an amplifier and a loudspeaker. In
fact, the sound attributes are absolute, categorical judgments, although they are meant as
comparing, ordinal judgments: calling a pickup "loud" actually reads: "louder than most
others". "Shrill" therefore stands for "this pickup generates – using a customary guitar
connected to a customary amplifier with a customary control setting – a sound with much
more treble-emphasis than most others". What then causes a pickup to sound louder or shriller
than others?
5.5.1 DC resistance
The DC resistance is seemingly the most important parameter. It can be determined very
easily with an Ohm-meter. Sometimes alternatively the term 'impedance' is used, other times
the term 'loudness '. The former use is not actually wrong since it is possible to connect DC to
an impedance – the frequency should, however, be specified. In other words, one should
either talk about 'impedance at 0 Hz', or simply of 'DC-resistance'. Statements like 'loudness =
8 kOhm' or 'Output = 8 K' are plain incorrect. For one, the quantity and the unit are already a
mismatch, and even more importantly there is no simple connection between loudness and
DC resistance. This is easily seen when taking the magnet out of the pickup: the DC-
resistance remains the same, but the loudness approaches zero fast.
The DC-resistance R is determined by the specific resistance ρ of the coil wire, the area SCu of
the wire cross-section, and the length l of the wire: R = lρ / SCu. Copper wire is almost
always the chosen material for magnetic pickups, for it we get: ρ ≈ 0.018 Ωmm2/m.
Depending on additions and impurities there will be small variations in ρ while larger
variations should be expected on the wire diameter. More recent data sheets specify AWG-42-
wire with a diameter tolerance (due to manufacturing processes) of ±5% an. Since the cross-
sectional area and the diameter have a quadratic interdependence, that cross-sectional area SCu
and thus the resistance value R has a spread of ±10%.
The diameter D is very small – often as thin as approx. 63 µm and thus thinner than a human
hair.
US-literature specifies the diameter as AWG (American Wire Gauge) an. AWG-42 – a wire
very often used in pickups – has a copper diameter of 2,5 mil = 63,34 µm. The following
approximation can be used for conversions in the range 30 < AWG < 50:
♣
e.g.: AWG-42 → DCu = 63.3 µm
The nominal value of the resistance per meter for this wire (AWG-42) is: 5,4 Ohm/Meter.
Manufacturing variations lead to a scatter of 4,9 to 5,9 Ohm/Meter (modern manufacturing).
It also depends on the temperature: R rises per °C by 0,39 %.
♣
more precisely:

5.5 Pickup-Parameters 5-57
A thin layer of varnish applied to the cylindrical copper wire serves as insulator. As a
consequence, the diameter grows by 10% for a single build wire (one coat of insulation) and
for a heavy build wire (2 coats of insulation) by 20%. Since merely very tiny voltages are
generated in pickups, one coat of insulation is sufficient.
The maximum applicable length of wire depends – other than on the wire diameter – on the
winding space on the coil bobbin and on the fill factor. Winding by hand results in the wire of
individual turns crossing the wire of other turns, and more air and less copper is in the coil.
Exactly positioning every turn next to the other is achieved via winding by machine; the fill
factor is higher. The pull has next to no influence: in order to firmly layer the turns, a small
braking force is applied. However, since such delicate wire breaks very easily, there is not
much margin here. One manufacturer recommends winding AWG-42-wire with approx. 0,33
N pull. The strain in this case is only about 0,1% and the transversal contraction even less.
Any resistance increase due to the pull is therefore negligible.
Fig. 5.5.1: Cross-section through a pickup coil. Winding width b and winding
height h define the interior dimensions of the bobbin. The wire diameter is
shown drastically enlarged.
Fig. 5.5.1 shows the cross-section through a pickup coil. For customary pickups the width b
varies between 4 – 12 mm and the height h between 5 – 15 mm; very small coils (e.g.
Gretsch) have a height of merely approx. 2.5 mm. Often the available winding area S = b x h
is between 30 und 60 mm2. For an AWG-42-wire the cross-sectional area including the
varnish is approx. 0.004 mm2. To calculate the largest possible number of turns from these
data we need to estimate the proportion of air in the winding. Fig. 5.5.2 presents two ideal
cases: the fill factor F is the quotient of circular wire-area to rectangular winding area.
The right-hand section of Fig. 5.2.2 shows the desirable objective: all turns fit tightly into the
notch between the wires below and the fill factor is in excess of 90%. A winding of such
precision is only achievable with a correspondingly precise feed rate control. Given that the
feed is merely 71 µm per turn, only smaller fill factors will be achievable in practice (F = 70
– 85%). The Stratocaster pickup, for example, offers a winding area of approx. 40 mm2. The
application of 7600 turns (a usual value for CBS-Fender in the 1960s) results in 30 mm2 wire
area and approx. 75% fill factor. Given an average length of 14 cm per single winding turn
the overall wire length comes to 1064 m which can be calculated to a DC resistance of 5.7
kΩ. This value is quite nicely confirmed with measurements.
Fig. 5.5.2: Fill factor F for ideal wire positioning. The cross-section of the copper itself is – for a wire with a
single layer of varnish – approx. 80% of the full wire cross-section area; for a double layer of varnish approx.
70%. Depending on manufacturer, insulation type and manufacturing method other values may result!

5.5.2 Inductivity of the coil winding
As electric current flows through a conductor, a magnetic field surrounding this conductor is
generated. Strictly speaking, a magnetic field is also generated within this conductor but this
effect is mostly neglected. Fig. 5.5.3 schematically shows a wire through which a current
passes from bottom to top. The technical current direction (from plus to minus) and the
direction of the magnetic flux (from north to south) are connected unequivocally: using your
right hand and pointing with the thumb in the direction of the current will make the remaining
(bent) fingers point in the direction of the magnetic flux.
Fig. 5.5.3: Magnetic field around a conductor through

which current passes
It has already been mentioned that magnetic flux must not be seen as a concrete means of
transport. Flux and flux density are assumed as analogies from fluids. The same approach is
found in other areas of physics (e.g. the flow of current). A straight wire of infinite length
through which a current I flows generates – at a radial distance of R – the magnetic field
strength of H and the magnetic flux density of B:
The quantity µr is called relative permeability and identifies the magnetic property of the
material penetrated by the magnetic field as a multiple of the permeability of air µ0
(strictly speaking µ0 is valid only for vacuum but the difference to air is negligible).
In Fig. 5.5.4 we see a rectangular wire frame though which electrical current flows. Again, a
magnetic field results which for this representation has an orientation perpendicular to the
paper plane. Fields running in the viewing direction are customarily shown as crosses while
the opposite direction is given by dots.
Fig. 5.5.4: Wire frame carrying a current, magnetic field. The field
strength decreases with increasing distance.
The magnetic flux density B specifies the area-specific magnetic flux. Integrating B over the
field-penetrated area S results in the overall magnetic flux Φ :
(scalar product)

In air, there is a linear correspondence between the current I and the magnetic flux Φ
generated by it. The coefficient characterizing this proportionality is the inductance L. Given
material and topology, L can be calculated from the build and is (in linear systems) not
dependent on the current. For the magnetic guitar pickup, the coil inductance L is the most
important electrical parameter. It has a major influence on the sensitivity, the impedance
frequency response and the resonance frequency.
If an alternating current is flowing through the wire frame shown in Fig. 5.5.4, a time-
variant magnetic field results. The law of induction tells us that an electric voltage is induced
in a current loop (through which a magnetic field is flowing). This voltage corresponds to the
variation over time dΦ/dt of the flux penetrating the coil. For a sine-shaped current and with
complex nomenclature the time-differential corresponds to a multiplication with jω :
, with: .
The quotient of the voltage U and the current I is called impedance Z = jωL in the framework
of complex calculation. Z is a system quantity and thus independent of the signal (as required
in linear systems). The impedance of the wire frame in Fig. 5.5.4 is proportional to the
inductance L and proportional to the frequency f.
Shown in Fig. 5.5.5a are two square wire frames through which the (same) current is flowing.
The magnetic fields generated by these two frames should not superimpose which can be
achieved either by a big distance between the frames or via fields with perpendicular
orientation. In Fig. 5.5.5b the two frames are laid on top of each other such that the magnetic
field penetrates both frames in the same way.
Fig. 5.5.5: two square wire windings connected in series carrying the same current.
a) separate location (left), b) on top of each other (right).
Each of the two frames in Fig. 5.5.5a generates the flux Φ, and in each frame the voltage U is
induced; the overall voltage induced in the series connection of the two frames therefore is
2U. Relative to one frame of the same size, the inductivity has doubled (assuming the
connecting wire to have no inductivity). In Fig. 5.5.5b the superposition of the magnetic flux
generated by the two frames results in double the overall flux. The voltage induced in each of
the two frames is double that found in the scenario of Fig. 5.5.5a, i.e. the series connection
results in the quadruple overall voltage and – correspondingly – the quadruple inductivity.
Thus, if a wire is wound with N windings, its inductivity may increase by a factor of N or by a
factor of N2 – depending on how the windings are coupled. Of course, wire windings can
never share the exact some location – the individual turns will in reality have to have a certain
distance and cannot be completely coupled. Still, real coils exhibit L ∼ N k with k > 2 because
an increase in the number of turns will also require an increase in the area.

The typical shape of the winding of a pickup is oblong (Fig. 5.5.6). Its inductivity can be
calculated with good approximation [Hertwig]:
Here N is the number of turns and the diagonal. The dimensions need to be
entered in cm. For a Stratocaster pickup (N = 7600, without magnets) the result is
L = 1,7 H. With magnets, the inductivity rises by approx. 30% to 2,2 H.
The above formula makes it also possible to calculate the effects of changes in the number of
turns. Since x, y, and h also changes, a power function with an exponent larger than 2 results:
L ∼ N 2,14. Increasing e.g. N by 10% pushes the inductivity by 23%. Since the resonance
frequency is dependent , the resonance frequency decreases by 10% in this example
(keeping the capacitance constant).
Fig. 5.5.6: Shape of a Stratocaster coil and cross section. Idealized rectangular coil.
The inductivity is dependent on number of turns of the winding and the geometry of the coil,
but also on the magnetic conductivity of the space penetrated by the magnetic field. Most
materials differ from air only marginally in magnetic terms; their magnetic conductivity, the
permeability µ = µr ⋅µ0, is µ = µ0 with very good accuracy, since the relative permeability of
these (diamagnetic or paramagnetic substances) is almost exactly one. Ferromagnetic
materials, on the other hand, react rather differently: their relative permeability is
considerably larger than one and moreover not constant but dependent on the field strength.
Alnico-pickups and polepieces, and also screws and shielding plates made out of
ferromagnetic material (e.g. iron or nickel) are ferromagnetic. By means of their better
magnetic conductivity (relative to air), such ferromagnetic materials decrease the magnetic
resistance and thus increase the inductivity. A particularly strong increase in inductivity is
possible if the complete magnetic field flows through the ferromagnetic material – this is,
however, as a matter of principle not possible in guitar pickups (5.4). Due to the fact that the
field running through air forms the largest part of the magnetic resistance, the magnets in the
Stratocaster pickup can increase the inductivity by only 30%, for example.
The non-linear permeability µ of a ferromagnetic material bends the magnetic flux into such
complex curves that an analytic description is not viable anymore. On top of this, eddy-
current- and skin-effects aggravate any calculations even further since they contribute an
additional inductive share which is dependent on frequency in a rather complicated manner.
For pickups without cover which contain on top of the coil only alnico magnets (e.g.
Stratocaster), stating one single inductivity is an acceptable compromise; here the losses in the
magnets are rather small.

However, as soon as a pickup comprises polepieces and/or mounting panels, the equivalent
circuit diagram contains either a single albeit frequency dependent inductance difficult to
interpret, or several inductances. Characterizing such a pickup with a single inductance is a
drastic simplification. For this reason, inductance-measuring meters are to be used only
with great caution for magnetic pickups (chapter 5.6). Such instrumentation determines e.g.
the inductive part of the complex impedance at one special frequency (e.g. at 1 kHz) and
implying a series equivalent circuit: Z = R + jωL. Since, however, polepieces subjected to the
skin-effect do not result in an imaginary part with an ω-proportionality, this measurement
approach is not suitable. More appropriate is to record a complete impedance-frequency-
response Z(ω) from which the components of a better suited equivalent circuit can be
calculated using methods of network synthesis (chapter 5.9).
5.5.3 Coil capacitance
The capacitance is defined as the proportionality between electric charge and electric voltage.
A small capacitance exists between two turns each of a coil; this capacitance is dependent on
the length, the distance and the dielectric constant ε. In vacuum (or air) we find ε = 8.9 pF/m,
insulators (such as the varnish and the bobbin) have 2 to 5 times that value. The overall
capacitance of a pickup can only be calculated as an approximation, because there are capaci-
tances between all turns of the coil. Given the height h, the width b (Fig. 5,5,1) and the
average length ξ of one turn, the result for the coil capacitance Cw is:
Coil capacitance using regular varnished wire [17]
Customary pickup coils have capacities in the range of 10 ... 150 pF. The capacitance of wide,
shallow coils (which seem to have a large surface area when observed from above i.e. from
the direction of the string) is smaller than the capacitance of compact coils with
approximately square cross-section of the winding. For example, Jazzmaster- or P90-pickups
have a smaller capacitance than Stratocaster pickups. For machine-wound pickup coils the
individual turns are closer together which results in a slightly higher capacitance compared to
hand-wound pickups. Increasing the thickness of the varnish layer has the opposite effect: the
individual windings have a larger distance, and the capacitance decreases.
Installing the pickup in the guitar leads to an increase of the capacitance. The main reason is
the pickup connecting cable the capacitance of which can vary between a few picofarad
(unshielded two-wire cable) and several hundred picofarad (old Gibson cables). The second
reason for the increased capacitance is the presence of stray capacitances towards metal parts
which are close-by, in particular towards shielding sheets.
Working in conjunction with the coil inductance, the capacitance is the basis for the pickup
resonance (at 2 – 5 kHz). However, much more important than the coil capacitance is the
cable capacitance (chapter 9) which has the main contribution to the overall capacitance.
Several components are involved in the resonance damping; of there the loss resistance
connected to the coil capacitance (chapter 5.5.4) has the smallest effect.

5.5.4 Resonance quality factor Q
The interaction between inductivity of the pickup (approx. 2 – 10 H) and the capacitance of
the cable (approx. 300 – 600 pF) forms a resonator with a resonance frequency in the range of
2 – 5 kHz. The quality factor Q is a measure for the resonance dampening. A strong
dampening results in a low quality factor, weak dampening makes for a high quality factor. A
small Q-factor must not be equated with 'bad'. A high Q-factor implies that the pickup
frequency response has a strong resonance emphasis at the resonance frequency. Its effect can
be equated to that of an equalizer boosting a certain frequency band (presence filter). Give
that the equivalent circuit of the pickup includes only one single coil, one single capacitor
plus resistors, the resonance quality factor can be stated unambiguously. If, however, skin
effects and eddy current losses require a more complex equivalent circuit, it is necessary to
define several poles with several Q-factors. A single value for the Q-factor can only be
specified as an approximation.
Fig. 5.5.7: Varying resonance emphasis for a Jazzmaster pickup. Different parallel resistors result in different
resonance dampening, or different resonance quality factors Q.
Fig. 5.5.7 shows the low-pass transmission of a Jazzmaster pickup. Setting the denominator-
polynomial to zero results in two poles of the transfer function (2nd-order low-pass).
Different resonance dampening can be achieved by varying the parallel resistance Rq. The Q-
factors associated with the 5 graphs in the figure are: 9,0 ; 2,5 ; 1,4 ; 0,9 ; 0,5. The highest Q-
factor (Q = 9) belongs to the lowest dampening with an emphasis of 19,1 dB. This behavior
can be achieved by loading the pickup exclusively with a 600-pF-capacitor. The results would
be, however, not very usable since it produces a shrill, whistling guitar sound. In normal use
the pickup is not just working in conjunction with a purely capacitive load but also with
parallel resistors constituted by the volume- and tone-controls plus the input impedance of the
amplifier. With these components, we arrive at a Q ≈ 3. The resonance emphasis seen in Fig.
5.5.7 can approximately be estimated via 20 lg(Q) in dB. For low Q-factors, this
approximation becomes increasingly inaccurate, though.
The resonance Q-factor is the second-most important transmission parameter right after the
resonance frequency. The above calculation shows, however, that the Q is dependent on the
connected circuitry. Already a change in length of the guitar cable results in a change of the
Q-factor (see also Fig. 9.14). Consequently, specifying a Q-factor value is problematic: the Q-
factor of the disconnected pickup does not allow for any conclusions regarding the Q-factor of
the installed pickup. Even the Q-factor of the pickup combined with the other components in
the guitar is not very meaningful. Only after additionally specifying cable and amplifier, a
value for the Q-factor can purposefully be interpreted.

Even more problematic is specifying the Q-factor for pickups which contain further metal
parts on top of magnet and coil. From a systems-theory point-of-view they represent systems
with an order of higher than 2. Stating a single Q-factor value is insufficient. The specifi-
cation of a resonance emphasis in dB is ambiguous, as well, since despite equal emphasis
different band-widths are possible. Fig. 5.5.8 compares a measured and a calculated
transmission curve. For both cases, low-pass behavior (not band-pass) was taken as a basis,
and one single coil of a Gibson Humbucker was measured. The slugs (polepieces) make for
pronounced eddy-current losses with skin-effects contributing, and thus a system of higher
order results. The 2nd-order transfer function shown in comparison has in principle a similar
shape but clearly differs.
Fig. 5.5.8: comparison of a measured

transmission curve (bold) with a
calculated curve (fine). Despite the
same emphasis height and equal
asymptotes, the shapes are different.
In closure it needs to be noted that – in contrast to the resonance quality factor Q – the quality
factor QL of the coil itself has even less significance. In the RL-series equivalent circuit of a
coil the Q-factor of the coil is defined by QL = 2πfL/R. It is dependent on the frequency and
therefore subject to an arbitrary frequency definition. For example, DUCHOSSOIR defines the
coil-Q-factor at 1 kHz and lists Q-factors of 2,1 to 3,5 for the Stratocaster pickup. Fig. 5.5.9
shows how small the effect of the coil-Q-factor is on the transmission behavior. Increasing the
coil resistance R by 50% decreases QL by 33% but changes the resonance emphasis only very
little.
Fig. 5.5.9: transfer Function.

Disconnected Stratocaster pickup
(resonance at 9 kHz), and with 111
kΩ load plus 600 pF cable
capacitance. The thin lines show the
transmission behavior with a coil
resistance increased by 50% i.e. a
coil-Q-factor reduced by 33%. For
comparison, a 33% reduction of the
Q-factor is shown by the dashed line.

5.5.5 Polarity
The polarity of the voltage generated by the guitar depends – on top of the string vibration –
also on the polarity of the magnet, the direction of the coil winding and the wiring. Old
Fender pickups sported a yellow (or white) and a black (or blue) wire; the yellow wire fed the
switch while the black went to ground. Very early Fender pickups had the north-pole of the
magnets pointing towards the strings but as a rule (from which there are exceptions) the
south-pole points "up". Stratocaster pickups are wound clock-wise, Telecaster pickups
counter-clock-wise. For the Jazzmaster pickup, the south-poles point "up" for the neck-pickup
and "down” of the bridge-pickup. Both are wound in opposite directions so that their signals
are added when they are both "on" but the hum-voltages cancel each other out – an advantage
which 1970s-Stratocaster pickups also profited from (the middle pickup was reversed in coil
winding and magnet orientation). From all this it can either be derived that pickup polarity
does not matter much for the sound, or that here lies a secret of the "vintage sound".
For a long time after the publications by G. S. Ohm (1843) and H. v. Helmholtz (1863), the
hearing system was seen as phase-insensitive: accordingly only the level of the partials define
the sound but not their phases. Initially there were contradicting experimental results regar-
ding this assumption until around the middle of the 20th century comprehensive psycho-
acoustical experiments could prove without doubt the phase-sensitivity in hearing. However,
not all phase changes are audible – which complicates matters. All following considerations
refer to diotic presentation (i.e. both ears receive the same signal) although, in fact, listening
to music involves dichotic conditions (i.e. there are different signals at the two ears).
However, switching the phase of a pickup results in a diotic signal change (i.e. the differences
at both ears are the same).
Fig. 5.5.10: time function, compound

from 1st and 2nd harmonic; can be
projected onto each other via reversing
the polarity.
Fig. 5.5.10 depicts two pure ac-time-functions differing only in polarity. With e.g. a
fundamental frequency of 200 Hz and a presentation loudness which is not too low, switching
between the two signals results in perceiving a small sound difference♣ . This indicates that
the ear can distinguish the absolute phase – in other words, an inward push of the tympanic
membrane gives a perception different from the one caused by an outward pull. Physiological
experiments measuring the potential in inner-ear-receptors (hair-cells) support this insight: the
hair-cells preferably react to an excitation of one polarity (bending of the stereocilia in the
direction towards the modiolus). This property of the hearing system alone would be reason
enough to consider the pickup polarity; still more important, however, is the fact that guitar
amplifiers almost always include non-linearities the effect of which is polarity-dependent.
Even in the so-called "clean mode" at least the attack of the sound is slightly overdriven, and
via "crunch" towards "lead" the harmonic distortion increases to an extreme degree – which
is not a deficiency but desired tone-shaping. Reversing the input signal would only result in a
pure reversal of the output signal in the case that the characteristic curve of the transmission
were symmetric re. the origin (odd-numbered distortion products). For even-order distortion,
the shape of the signal changes with polarity reversal and so does the level-spectrum of the
output (Fig. 5.5.11).
♣
However, at higher frequencies no sound differences can be perceived when switching.

The left section of Fig. 5.5.11 shows a non-linear characteristic as it can be found e.g. in a
tube pre-amplifier. Using one of the signals from Fig. 5.5.10 as input on the abscissa, the
ordinate (output) yields the time functions given in the middle and right-hand sections of Fig.
5.5.11. Even without formal and quantitative description one can directly see the polarity-
dependent unbalances resulting from the non-linearity. Depending on the polarity of the input
signal two different output signals are created. Only for special half-wave symmetries are the
sound differences due to the polarity-reversal limited to the signal attack phase (and therefore
remain insignificant); in the general case the phase reversal of a pickup can – depending on
the circumstances – lead to audible sound differences.
Abb. 5.5.11: nonlinear transmission curve (left), time functions of the signals from Fig. 5.5.10 (-----), after
having passed though the nonlinear transmission curve (–––). All signals without DC-component.
Next, we will have to look at the question whether – and if so to which degree – the voltage
half-waves of magnetic pickups differ. For this, the neck pickup voltage of a Stratocaster
(USA) was investigated. Above the magnet pole of the neck pickup, the E4-string was
depressed with a pick and let go abruptly (force step, chapter 1 and 2). The result is a
rectangular velocity curve (Fig. 5.5.12) to which a triangular displacement corresponds. Due
to the non-linear characteristic of the magnet (chapter 5.8), the tip of the flux-density-curve is
bent (the tip of the triangle belonging to the linear model is shown as a thin line in Fig.
5.5.12). A differentiation of the flux-density function results in the induced voltage: this is
rectangular in the case of a linear magnetic characteristic, and pointed for the non-linear
model. The measured voltages show a clear similarity with the slight oscillations being results
of the dispersion which is not modeled here (chapter 1.3.2).
Fig. 5.5.12: time functions:

velocity, displacement, flux density,
Voltage. E4-string picked above the
neck pickup, voltage of the neck
pickup. Fender USA-Stratocaster.

In order to strongly dampen the resonance formed by coil and cable, the Stratocaster pickup
was loaded with a 1-kΩ-resistor for these measurements. In the relevant frequency range, coil
resistance, coil inductance and the 1-kΩ-resistor act as a first-order low-pass the real pole of
which was mathematically compensated by a zero. Additionally, a real pole at fx = 9000 Hz
was included so that the induced voltage was in total filtered by a first-order low-pass with a
cutoff frequency of 9 kHz.
Both calculation and measurement show that with a dispersion-free model of the string pickup
voltage is created which remains symmetric to the time-axis – even if a non-linear
characteristic of the magnet is used as the basis. Dispersion-effects play no role for the thin
guitar strings, and consequently calculation and measurement are in good agreement.
However, on the E2-string the frequency-dependence of the wave-propagation velocity
(dispersion) leads to deformations of the time function already after one single period (Fig.
5.5.13); the half-cycles loose their symmetry and thus the possibility arises that the sound
changes when the polarity of a pickup is reversed. Still, changes in the time-function do not
always lead to audible sound changes. The hearing system in not an oscilloscope; rather, the
sound-signal is split up into frequency bands (critical bands), and only the output of these
analyzing band-filters are subject to the time-dependency analysis. Phase shifts occurring
between signals falling into different critical bands may not cause any changes in he
perceived sound. Phase shifts within a critical band may on the pother hand very well lead to
audible roughness- and/or pitch-changes [Fleischer 1978].
Fig. 5.5.13: measured pickup voltage (Stratocaster), normalized. For the E2-string the dispersion-caused
oscillations are particularly striking (compare to chapter 1.4). The E4-string is, however, not entirely dispersion-
free, either: after about 7 periods clear dispersion-caused unbalances are visible (not shown in the figure),
The superposition of a low- and a high-

frequency oscillation shown in Fig. 5.5.13 is
reminiscent of the measurements regarding
masked-period-patterns carried out by
Zwicker [12]. Whether the tone-burst is
audible depends on its position within the
phase of the lower-frequency tone. (Abb.
5.5.14). Despite equal magnitude spectra, the
three signals shown in this figure sound
differently – the masking effect of the lower-
frequency component is phase-dependent.
Fig. 5.5.14: test signal for masking-period-pattern experiments [12].

Before we apply the masked-period-patters to guitar signals, we need to rather consider that
the time functions shown in Fig. 5.5.12 and 5.13 are derivatives of the string velocity; i.e. the
signal will never reach the tympanic membrane in this shape. First, already the pickup
resonance effects changes on the signal, then guitar amplifier and loudspeaker add their own
considerable part, plus last the sound wave has to travel through the listening room until it
finally reaches the ear of the listener or the player. In Fig. 5.5.15 the pickup voltages as they
show up for a Stratocaster loaded with 513pF // 1MΩ. Along the time axis non-symmetries
can appear which appear significant o the eye – however the eye does not judge the sound. In
fact, the hearing system struggles despite the obvious non-symmetries to recognize any sound
differences. Even more explicitly this is shown by Fig. 5.5.16: both these impulses sound the
same!
Abb. 5.5.15: pickup voltages, USA-Stratocaster, neck pickup. E2-string pressed down above the neck pickup
with a pick and then released (top left and right); virtuoso-like" picked (bottom left and right). Left and right
show two different attempts.
Fig. 5.5.16: two impulses which can be projected

upon each other via all-pass filtering. Since the
group-delay distortions remain below the threshold
of 2 ms (as it is relevant to the ear), this filtering is
not audible.

As shown in Fig. 5.5.11, the nonlinearities occurring in amplifiers cause a polarity-dependent

limiting of the guitar signal, but the main differences most often happen with the short
impulse peaks – the limiting of which changes little in the overall sound. The signal is subject
to significant alterations only as it is radiated off the loudspeaker: in Fig. 5.5.17 the voltages
generated by two microphones positioned within a listening room with a distance of 50 cm.
Despite major divergences in the time function, these differences are perceived merely as a
general change in the treble content without any special significance.
Fig. 5.5.17: microphone voltages at 0,5m in front of the amplifier (top) and inclined at 1 m in front of the
amplifier. Fender-Stratocaster, Fender-Deluxe-Amp. Two different attempts shown left and right, respectively.
We note as intermediate result: reversing the polarity of a pickup leads to clearly visible
differences in the voltage-time function. Our hearing, however, does not observe these
differences at all, or just marginally. In no way are differences due to polarity-reversal
obtainable in the sense of clearly better or clearly worse. Consequently, no recommendation
is possible regarding which polarity would be preferable. Still, two special operating states
merit additional consideration: the combination of several pickups and the feedback via the
air.
As the guitar is played loudly using amplifier and loudspeaker, an air-wave emitted by the
loudspeaker strikes the guitar body and excites vibrations in it and also in the strings. These
vibrations are fed back to the amplifier and thus we obtain a feedback circuit. With a
sufficiently high gain within the feedback circuit the guitar starts to play "by itself" [literature:
control engineering]. The pitch of this self-oscillation depends on a number of factors
including the polarity of the closed-loop-gain: reversing the pickup polarity leads to a change
in the sound. However, the same happens as one changes the position of the guitar by e.g. 10
cm (i.e. the phase in the feedback loop changes); as such the pickup polarity is irrelevant even
when considering feedback.
The fact that the sound changes drastically as the polarity of one pickup in a combination of
pickups is reversed requires not a lot of explanation. More interesting is the question whether
there are audible differences if both direction of the turns in a coil and the polarity of the
magnet are reversed (e.g. for the middle pickup of a Stratocaster) to achieve hum-suppression
in the combination of pickups (RW/RP middle pickup). Indeed, we could expect an effect if
the two pickups were magnetically coupled to a significant extent. Measurements, however,
show merely a 0,6%-coupling-factor which is much too little to give audible effects. The
measured level differences are, with 0,05 dB, far below threshold. Incidentally, for the forces
of magnetic attraction the rule applies that they are not polarity dependent!

5.5.6 Time variance
In systems theory, resistors, inductances and capacities are initially taken to be linear and
time-invariant such that at every point in time the same laws of proportionality hold for the
quantities of voltage and current. For small drive values, every guitar pickup is indeed
sufficiently linear whereas time-invariance cannot be fundamentally taken for granted. The
pickup parameters introduced on the previous pages do change over time (intra-individual
scatter), and for another specimen of the same type, any specific values only hold with
reservations (inter-individual scatter).
The DC-resistance of each pickup is temperature-dependent, it rises by 0,39 % per °C.

Within the temperature interval of 17° – 30° the DC-resistance therefore changes by 5% (e.g.
from 6000 Ω to 6300 Ω). This needs to be considered for the values given in literature which
are sometimes surprisingly precise as seen e.g. in a specification for the Stratocaster: 6100 Ω
(Vintage reissue), 6210 Ω (Texas special). Due to manufacturing tolerances, the wire diameter
will have a scatter of typically ±10%, which renders the comparison of two pickups
problematic: do we have the same type but with slightly differing wire strength, or is it the
other type with a different number of turns?
The coil inductance is given – other than by the coil geometry – by the space filled by the
field. As ferromagnetic and/or conductive materials are introduced into this space, the
inductance may change. Not just pickup covers but also other guitar parts can change the coil
inductance. These include the metal mounting plate of the Telecaster bridge pickup just as
shielding foils under the pickguard which enclose the pickups and enable eddy currents to
flow. We should not expect dramatic deviations but for precision-measurements, the
environment should be clearly defined. For the time-variance of magnet-parameters see
chapter 4.5.
For the coil capacitance as well, the space filled by the field should be considered. If a
hygroscopic material able to absorb water is used for the insulation of the coil wire the
capacitance will depend on the give water-content. In case of potted coil-winding an increase
of the capacitance will happen because all potting material have a dielectric number larger
than 1. However, since the major share of the overall capacitance is given not by the pickup
but by the cable capacitance, the effects of changes in the pickup-capacitance are – as a
general rule – only of secondary importance.
An environmental influence which is often overlooked results from the acoustical

surroundings. As soon as the pickup signals are amplified and radiated by a loudspeaker, the
pickup becomes part of a feedback loop. While this does not change the parameters
mentioned above, we need to enhance equivalent circuit of the pickup by controlled sources.
A complete description requires a (as far as possible) complete description of the transmission
coefficients of air-borne and structure-borne sound, the coefficients themselves being
dependent on time-variant mechanical dampening factors. It is, for example, conceivable that
rubber bearings stiffen over the course of decades and influence the sensitivity to structure-
borne sound. Depending on personal preferences, such an effect can be either classed as
insignificant and ignored, or be defined as belonging to the guitar body, or be seen as effect of
the pickup aging.

5.5.7 Insulating varnish, wax
The pickup coil is wound from very thin copper wire onto which a film of varnish is
deposited in order to protect from short-circuit and aggressive substances in the air. The
substances most often used as insulating varnish are "Plain Enamel", "Formvar", "Polysol"
and "Polyurethane-Nylon". The resulting insulated wire is often called magnet-wire – it is of
course non-magnetic (or, more exactly, paramagnetic) and has the same magnetic resistance
as air has.
The noun enamel also stands for glaze, lacquers in general or special lacquers (e.g. synthetic
resin varnish). The verb to enamel also means to varnish. "Enamelled copper wire" therefore
is varnished copper wire, and as such every magnet wire used in pickups merits the
designation enamelled copper wire. The situation is, however, not that simple since enamel is
often used in a more specialized sense:
The name plain enamel designates one of the first industrially produced insulating varnishes.
It is an oil varnish manufactured with oil which oxidizes while drying and generates an
irreversible film. In order to increase hardness and gloss, resins are added. The also used
designations oleoresinous email and oleoresinous insulation are derived from this oil/resin
mixture. The plain enamelled wire used in old (i.e. "vintage") pickups has a brown or voilet
color.
Formvar (sometimes incorrectly spelled "Formivar") was a trademark of the Monsanto

Chemical Company (St. Louis, Missouri, USA). It was renamed from Formvar to Vinylec
after the sale of a business unit to Structure Probe, Inc. Formvar varnishes contain
polyvinyl-acetal = polyvinylformal. In a two-step process first polyvinylalcohol is
manufatured from polyvinylacetate; the polyvinylalcohol is then acetalyzed. To produce
magnet-wire, the phenolic resin polyvinylformal (also called modified polyvinyl acetal resins)
is added. Formvar magnet wire is of a glossy-gold color and cannot be soldered.
Polysol varnish is a polyurethan lacquer which can be soldered and is mixed as a two-
component varnish. It usually is of a glossy bright-red. Or it could be brown-violet if a
"vintage" vibe is asked for .
Polyurethan-nylon is a polyurethan insulation with a nylon coating.
It should not be assumed that the designations for varnish as given above seek to be a 100%-
correct material designation. While e.g. the chemical formula NaCl unambiguously designates
common salt, a term such as oil varnish merely indicates a group of substances which are
similar but individually chemically and physically different lacquers.
No big demands are placed on the insulation properties of the copper-varnish-wire used in
pickups since the voltages to be handled are very small. Even considering a peak voltage of 5
V (which is quite a high value) and a varnish thickness of 2x2,5 µm = 5 µm we obtain a
"worst-case"-field-strength of about 1 kV/mm – which is rather undemanding for an insulator.
Formvar, for example, is specified to handle up to 80 kV/mm – but such high field strengths
cannot be reached in a pickup.

The magnetic properties of the insulators mentioned above are highly similar; they all show a
permeability of very close to 1 and can be seen as non-magnetic as a good approximation.
Regarding the dielectric numbers, however, differences can be measured. The εr of such
insulators is typically between 2 and 5 – exact numbers are not published by the
manufacturers. Variations in the dielectric number correspondingly change the capacitance of
the coil. Considering a change in capacitance from 50 pF to 100 pF (which is very much on
the high side) would lead – in conjunction with a 450-pF-cable – to a 10% capacitance change
corresponding to a 5%-change in the resonance frequency. The same resonance shift would
occur with changing the cable length by 11% i.e. increasing it from 3,75 m to 4,15 m. It
cannot be excluded that such small changes are noticeable in a true A/B-comparison. The
internet is full of speculations regarding the contribution of the varnish insulation to the sound
of a pickup, or regarding the sound differences due to different lacquers. Since however rarely
any guitarist will consider (in order to obtain a different sound!!) whether he/she should today
use the 3,75-m-cable rather than the 4.15-m-cable, it seems rather excessive to attribute a big
significance to the type of varnish. Anybody in doubt is cordially invited to listen to the
difference caused by a 50-pF-capacitor connected in parallel to the guitar output ... and if
indeed it does sound much better with the capacitor: grab the soldering iron and install it!!
Apart from the potential dielectric differences, there are occasional reports that a specific
varnish was applied more thickly than another, this leading to a different coil geometry. Of
course the coil inductance and coil capacitance depend on the geometry – however the
thickness of the varnish is not generally typical for a type of varnish. It must not be
assumed that all manufacturers produce a 42AWG-wire with the exact same thickness of the
varnish – even if the insulating material would be the same. The dimensions of copper and
varnish are subject to manufacturing tolerances; it also should be considered that many
manufacturers offer a special wire (e.g. 42AWG, Formvar) deliberately with different varnish
thicknesses. For high-voltage installations a thicker (multiple) insulation layer is desirable
while for pickups a single varnish process is sufficient. Even though some manufacturers use
wire with multiple varnishing for pickups.
So: what changes if, instead of wire with a single coat of varnish, one with a double coat is
used? That depends on which parameter is kept constant. With an equal number of turns the
coils grows larger. Conversely, filling up a given bobbin with wire of a thicker insulation will
lead to a smaller number of turns. As an approximation we can assume that a double-insulated
wire will require 20% more cross-sectional surface than the single-insulated wire.
• For a constant coil cross-section (i.e. wind until the bobbin is full) we obtain a 17%
smaller number of turns – connected to a reduced inductance, diminished sensitivity and
smaller DC-resistance.
• Keeping the number of turns constant (i.e. wind unto the counter shuts down the
process) enlarges the surface of the winding. However, this does not necessarily lead to an
increased sensitivity because the turns are located also in the range of smaller flux density.
Sensitivity and inductance cannot be calculated in any simple manner; for the DC-
resistance we get an increase of about 2%. This increase is so relatively small despite the
20% surface area change because the coil is oblong, not circular.
The pickup parameters depend only little on the thickness of the varnish if the number of
turns is kept constant; larger effects will be connected to keeping constant the cross-section of
the winding.

Besides the varnish, there is another dielectric between the turns of the coil in many pickups:
they are immersed (potted, dipped) in wax in order to give more stability to the coil. In the
middle of the 1960s a mishap occurred in the guitar production at Fender [Duchossoir]: the
newly introduced polysol-insulation dissolved in the wax bath and the pickups suffered from
short circuits. From that point in time production continued without wax-potting (apparently
the differences were not that serious), and not until the 1980s did Fender (now post-CBS)
return to the old recipes. Wax can solidify the coil windings and reduce pickup self-
oscillations (microphonics) on the one hand but also increase the coil capacitance on the other
hand. However, compared to the all-dominating cable capacitance only marginal changes in
capacitance are to be expected. For microphonics see chapter 5.14.
The losses within the insulation between the windings of the coil do not play any role at all:
the loss resistance in parallel to the coil capacitance is larger than 10 MΩ and thus negligible.
However, depending on the material it may be necessary to consider hydroscopicity: the
insulators may be able to absorb water which can – due to its high dielectricity – cause a
noticeable capacitance increase (see table)
Material εr at 1kHz tanδ in ‰

Casting resin 4–8 20 – 80
Cellulose acetate 3,5 – 6 12 – 25
Cellulose ethyl 2,5 – 3,5 5 – 25
Vulcanized fiber 4 80
Polyurethane 3,0 – 5,5 5 – 50 Table: dielectric properties of
Paraffin 1,9 – 2,3 <5 insulating materials. The numbers
Shellac 3–4 10 should be taken as guide values, the
Bakelite 4,8 – 5,3 10 material compositions vary
Pertinax 4,8 – 5,4 25 depending on the manufacturer.
Water approx. 80
As a bottom line it should be noted that potting a pickup in wax on one hand, and the
material and the thickness of the varnish on the other hand, can lead to small, measurable
differences in capacitance. The significance of these differences is, however, subordinate in
practice. Microphonics can be efficiently fought by potting.
5.5.8 Bobbin, coil former
In old Fender pickups the 6 cylindrical magnets were pushed through 2 planar coil formers
made of vulcanized fiber (hydrate cellulose): these coil formers kept both the magnets in
position, and the would wire on the magnets. An urgent warning needs to be heeded: the axial
position of the magnets in these pickups must not be manipulated by "light hammer-blows".
Doing this will in many cases rupture the fine winding wire which necessitates replacing the
pickup (or rewinding it). It is inconceivable why some authors recommend this kind of
"adjustment" – possibly they are sponsored by the pickup manufacturing industry ..... Much
better mechanical protection is afforded by pickups with complete plastic die-cast bobbins.
Regarding any influence of the bobbins or coil formers on the sound, what was said for
insulators holds again: the materials used may have varying ε and thus potentially could have
an effect on the coil capacitance and the resonance emphasis. Compared to the cable
capacitance and the dampening afforded by the potentiometers, such differences are however
to be taken as highly secondary.

5.5.9 DC-resistance vs. loudness
In chapter 5.5.1 it was already noted that the dc resistance of a pickup has little bearing on its
loudness. While it is of course true for an individual pickup that unwinding a few thousand
turns will reduce both the resistance and the loudness, it must not be concluded that a 7-kΩ-
pickup is generally louder than a 5-kΩ-pickup. Unfortunately, this is however exactly what is
suggested in many tests which e.g. read: "the guitar is equipped with two different pickups.
While the alnico-2-magnet at the neck makes for singing highs, its colleague at the bridge
with the ceramic magnet yields a brutal punch. Our measurements reveal just how significant
this difference is: 12 kΩ (bridge) versus 8kΩ (Neck)." Does that mean the "colleague" at the
bridge is 50% louder? Another example from a guitar comparison: " the pickups of this guitar
have the lowest output power of all candidates: they show merely 8 kΩ; all others are at 10 –
18 kΩ." And one last example: "the neck pickup corresponds in its power to a Gibson PAF (8
kΩ)." From a physics point-of-view, such texts are more than problematic.
Pickups are almost always wound with copper magnet wire, and therefore only the wire cross-
section and the wire length figure for the DC resistance RDC. Measuring the resistance is easy
– even inexpensive RDC-instruments have a tolerance better than 1%, and 1‰ is achievable
without much effort. If indeed such a high accuracy is the objective, the temperature needs to
be specified exactly as well within a ¼°C. Test reports often include four-figure resistance
details: for the Gibson 498-T e.g. 12,23 kΩ, or – in another guitar – 13,40 kΩ. The reader is
however left in the dark about whether such differences are due to the instrumentation (which
is almost never specified at all), or due to manufacturing scatter ... or at least in part due to the
often applied practice not to disconnect the volume potentiometer for the measurement: this
changes the reading of a 13,00-kΩ-resistance to 12,67 kΩ (for a 500-kΩ pot) or to 12,36 kΩ
(for a 250-kΩ pot), after all. Such small differences would not be of any significance if they
were not the reason to draw the conclusion that with 13,40 kΩ that last bit of "punch" would
be achieved which the 12,23-kΩ-contestant unfortunately missed. Reading such a test report,
indeed not few guitarists will invest $200 to profit from that "punch".
The pickup industry happily picks up on this resistance diversification and offers a vast
variety of pickups. The Gibson BurstBucker is available in three versions: slightly
underwound, normal, and slightly overwound. The DC resistances differ by 7% each – and
these are not unavoidable manufacturing tolerances but deliberate production♣. Or so the
Gibson advertisements state. On the other hand, the Gibson 498-T is only available in a single
version. Tests in a German music magazine (in Summer 2003 and Winter 2005) report that
there are resistance tolerances of 9,6% between two specimen of this pickup.
In many test reports the DC-resistance of a pickup receives a multi-digit specification;

however, it is quite often not designated with "resistance" but with "output power". The
skillful reader will interpret this as loudness und is generally not entirely wrong with this
approach. Indeed an SDS-1 (9,1 kΩ) will yield more output voltage as a vintage Strat pickup
with its modest 5,8 kΩ. On that basis, a Gibson Tony-Iommi-Signature pickup would really
hit home, wouldn't it, having not less than 17,8 kΩ DC resistance! That's almost the double
"output power" relative to the SDS-1. However, given the same string vibration, the Tony-
Iommi generates less voltage than the SDS-1, its high resistance does not increase the
transmission coefficient TUv. The latter value – defined as quotient of pickup voltage and
string velocity (chapter 5.4.5) – is well suited to investigate correlations between the DC
resistance and transducer efficiency.
♣
The '57-Classic-Plus sports as little as "3% more winding" versus the '57-Classic [Gibson special issue of the
German "Gitarre & Bass" magazine].

The frequency dependence of the transmission coefficient HUv follows a more or less
complicated low-pass function (chapter 5.9.3). For the Stratocaster pickup we obtain a simple
2nd-order low-pass with a resonance emphasis of about 5 dB (Fig. 5.5.18). Increasing the
number of turns by 10% (e.g. from 7600 to 8360 turns) will increase the DC resistance by
10%, as well. Calculating more precisely and considering that the additional turns are located
on top of the coils and will therefore be a bit longer we obtain an 11% increase of the DC
resistance. The effects of the CD resistance on the transmission function are, however, so
small that over-exaggerated requirements as to the precision are not purposeful. The
inductivity of the pickup will rise by about 23% (chapter 5.5.2), the capacitance is determined
predominantly by the cable, and equally the load resistance. All these contributions combined
will result in an increase of the transmission factor (the log of the transmission function) by
0,9 dB while the resonance frequency drops by 10%. In a direct listening comparison these
changes will be just about noticeable with the increased number of turns the pickup features a
little less brilliance. In the treble range we even incur a very small loudness drop while a
minimal loudness increase happens in the low end. From a psychoacoustic perspective [12]
the most appropriate parameter to describe these changes would be the sharpness: it drops
with increasing number of turns.
Fig. 5.5.18: changes of the transmission factor GUv = 20⋅lg(HUv)dB for a 10% increase of the number of turns N.
To conclude any assessments of loudness based on the pickup DC resistance is difficult

because the former depends on so many parameters. Other than the frequency response of
amplifier and loudspeaker, the room acoustics also determine the final perception of the
sound, and added to this are subjective preferences (e.g. attack vs. sustain). For the following
analysis (Fig. 5.5.19) we will therefore not evaluate the loudness but the low-frequency
transmission coefficient TUv and compare it with the DC resistance (see also chapter 5.4.5).
The large scatter of the pairs of values clear-

ly shows that the transmission coefficient
and the DC resistance correlate only little.
For identical DC resistances the transmis-
sion coefficient can vary as much as a factor
of 4!
Fig. 5.5.19: Comparison between low-frequency

transmission coefficient and the DC resistance.
Transmission data as in chapter 5.4.5.
o = singlecoil, * = humbucker.

5.6 Instruments for measurements on pickups 5-75
5.6 Instruments for measurements on pickups
How do I measure an electric pickup parameter? In most cases presumably with an instrument
the inner workings of which are not really known to the user. A popular choice is the so-
called RLC-instruments able to meter R (resistance), L (inductance) and C (capacitance). If
the pickup were an ideal basic two-terminal network, there would be no objections against
this approach. Basic two-terminal networks consist either of an ohmic resistor, or of an ideal
inductor, or of an ideal capacitor. In a pickup, however, all three of these elements operate in
conjunction – the pickup thus is a composite two-terminal network.
In the simplest case the pickup impedance is modeled via an electrical resistor R connected in
series to the inductance L of the winding, with the capacitance of the winding connected in
parallel to this series connection. The resistor R is the DC-resistance of the wound copper
wire as was already described above – it is also called copper resistance. As has been
elaborated in the chapter Magnetodynamics, the DC-resistance of an ideal inductor is zero; the
DC-resistance of an ideal capacitor is infinite. Indeed, at f = 0 Hz only the value of R remains,
since the other two components in the network do not contribute anything at this frequency.
Consequently, if R is to be measured, this should be done at 0 Hz – kind of obvious, isn't it.
However, RLC-instruments do not work at 0 Hz but at other frequencies, e.g. at 1 kHz. They
will determine the real part of the complex impedance Z – which may well be different from
the copper resistance.
The formal description of the impedance works best with the aid of the complex notation [see
e.g. 18, 20]. The complex impedance Z of an RL-series-connection (i.e. to begin with without
the capacitance C) is:
Complex impedance
The real part of the complex impedance is R, the imaginary part is ωL (the imaginary unit j is
not a section of the imaginary part!). As evident, the real part is independent of the frequency
and can – for this specific two-terminal network (!) – measured at any frequency. As soon as
the capacitance is connected, however, this situation changes: the capacitance C is, in the
simple equivalent circuit, connected in parallel to the RL series circuit. The complex
impedance Z of this RLC is calculated as:
Breaking down this complex impedance as a sum of a real part and an imaginary part yields a
value which an RLC-instrument will show as loss-resistance if a coil is to be measured:
real part of the RLC-circuit
This real part is not constant anymore but dependent on the frequency! For DC i.e. at ω = 0,
the correct DC-resistance R is still the result, however for every other frequency a diverging
and thus incorrect value is measured.

These deviations are not always dramatic – BUT they should be looked against the
background that an "expert" for example has to know that the Texas-Special-Pickup sports
6210 Ohm whereas the 'Vintage Reissue Pickup" throws a mere 6100 Ohm (i.e. a full 1,8%
less!) into the ring. Incidentally, the expert hopefully is also aware of the fact that the same
1,8% resistance change can also be caused by a temperature change of as little as 4,5°C .
How large the differences can be due to the instrumentation is shown by Fig. 5.6.1: using an
RLC-Instrument working with 1000 Hz to measure the Stratocaster-coil-resistance will give a
value which is too large by 6%. Which amounts to about the difference between a '80s-
Standard-Pickup' and a 'Late-60s-Pickup. At the same temperature …..
Pickups with a relatively low resonance frequency (e.g. Gibson P90), on the other hand, show
significantly larger deviations (Fig. 5.6.1). Connecting a cable changes the real part, as well,
even if the cable is defined as ideal capacitance having exclusively an imaginary effect (in the
sense of the imaginary notation system ). For the P90 pickup, the addition of a cable of 600
pF has the effect that at 1 kHz the real part of the impedance increases by 40%.
Abb. 5.6.1: Real part of the pickup impedance referenced to RCu. Left: Stratocaster, right Gibson P90.
At f = 1000Hz the real part (without cable) diverges by 6% respectively 130% from the 0-Hz-value (–––––).
Narrow lines: with 600-pF-cable. To clarify at which frequency the impedance meter is operating, the measuring
frequency can be checked e.g. via an oscilloscope during the measurement.
Besides the DC resistance R, the inductivity L is the second important electrical parameter. If
only R and L were cooperating in a pickup, we could measure L without any issue as
imaginary part of the impedance – at any frequency except 0 Hz. However, the capacitance
connected in parallel has the effect that below the resonance frequency the normalized
imaginary part rises; above the resonance frequency it even becomes negative. An RLC-
instrument, which displays in the "coil measurement" setting merely the imaginary part of the
impedance divided by 2πf, will follow the curve shown in Fig. 5.6.2. Up to 1000 Hz the
deviations for the Stratocaster pickup are actually not too significant yet; for higher
frequencies, the error keeps mounting – as it does for pickups with lower resonance
frequency.
The real problem with inductance measurements starts if the pickup impedance should be
described by more than one inductivity. As we will see in the chapter about equivalent
circuits, this comes into play especially if eddy-current losses can not be ignored, i.e. for
pickups with slugs made of soft iron or nickel, and/or with metal covers. In complete analogy,
a mechanical system including 3 independent masses connected via 2 independent springs
could not be characterized for oscillations of every frequency by one and the same spring
stiffness, either.

5.6 Instruments for measurements on pickups 5-77
Imaginary part of the impedance of the RLC-circuit:
Fig. 5.6.2: Imaginary part of the pickup impedance

(Stratocaster), referenced to ωL.
Thin line = 600-pF-cable added.
In such a case a possible way would be to first define a suitable equivalent circuit, and then
to determine the element values in this equivalent circuit via measurements. RLC-meters in
fact use the same approach, and in some cases even include options: to measure a coil two
equivalent circuits are offered – an RL-series circuit and an RL-parallel circuit. These two are
however not compatible. For example, the series connection of a 6861-Ω-resistor and a 2-H-
coil may be described at 1 kHz by an equivalent parallel circuit of a 30-kΩ-resistor and a 2,6-
H-coil. The equivalence is valid only for 1 kHz; at every other frequency different values will
result for the elements. Both the RL-series circuit and the RL-parallel circuit are moreover too
simple for a pickup; suitable equivalent circuits should at least include a capacitance (see also
the chapter on equivalent circuits).
As an alternative to measuring the inductivity with an RLC-meter it is possible to draw the

frequency response of the amount of the impedance in a double-logarithmic representation.
Since the impedance is dependent on the frequency according to a power function, curves
result which – in sections – can be approximated by straight lines. Or so the theory according
to Bode says. However, this only holds for simple networks such as an RL-series circuit (Fig.
5.6.3). Or so the author says.
Fig. 5.6.3: Amount of the impedance of an RL series circuit (left), and of a RLCR-equivalent circuit (right).
Plotted in the left part of Fig. 5.6.3 is the amount of the impedance frequency response of an
RL-series connection (R = 7 kΩ, L = 2 H). The curve approximates towards low frequencies a
horizontal straight line Z = 7kΩ), whereas towards high frequencies we get an increase
according to the slanted straight (Z = 2πfL). The inductivity L can be determined graphically
from this measurement by shifting the approximative straight line to best match the curve.
The proportionality coefficient L is the inductivity.

However, as soon as a capacitance C (600 pF) and a dampening resistor Rp (1 MΩ) are also
included, this high-frequency approximation is not possible any more (Fig. 5.6.3 left). One
can try the approximation at medium frequencies (e.g. at 1,5 kHz) but this will result in a
inductivity result which is 20% too large. In fact, that would not be a dramatic error but it all
depends on the desired measurement accuracy. DUCHOSSOIR specifies the following, for
example: Late-60s-Strat: 2,2 H, Vintage-Reissue: 2,3 H, 1980s-Standard: 2,37 H, Texas-
Special-Neck: 2,47 H, Texas-Special-Middle: 2,50 H. If indeed such small differences (as far
as they are of any significance to begin with) are to be determined, 20% tolerance would be
unacceptable. As a precaution it is noted here that determining the intersection point with the
straight line dropping off at high frequencies (due to capacitances) brings an improvement
only in theory: in practice there are parasitic disruptive effects which falsify the theoretically
expected 1/f-drop-off.
Measuring the pickup quality factor Q with an RLC-Meter is even more misleading that the
measurement of R and L. What is actually measured here is the coil quality ( QL = 2πfL/R )
and thus a frequency dependent parameter. DUCHOSSOIR assumes, in his books on the Fender
Stratocaster and Telecaster, relatively arbitrarily f = 1000 Hz. He notes: a pickup with a
higher Q emphasizes a narrower frequency band, and vice versa a pickup with a smaller Q
emphasizes a wider frequency band. This clarification would hold if Q were meant to be the
resonance quality factor, however, DUCHOSSOIR does not list resonance quality factors, but the
coil quality. The influence of the latter on the resonance emphasis cannot be described by a
simple function. Fig. 5.5.9 shows how changes in the coil quality have only small effects on
the resonance emphasis. If on the other hand both R and L are changed similarly, e.g. both by
50%, QL remains constant but the resonance emphasis drops by about 2 dB on the
Stratocaster.
As a closing example a pickup from a Gretsch Tennessean is investigated. Its DC-resistance

amounts to 3260 Ω, however taking an impedance measurement at 1 kHz yields 7155 Ω in
series with 1,2 H. An equivalent circuit built from these two components indeed shows the
same impedance at 1 kHz (Fig. 5.6.4) but behaves much differently at other frequencies.
Fig. 5.6.4: Measured impedance amount of a Gretsch pickup (–––). A RL-meter operating at 1000 Hz
measuring frequency shows 7155 Ω and 1,2 H. Measuring such an RL-series circuit (-----) reveal, however, major
differences. The right figure depicts the impedance locus (50 – 5000 Hz). The reason for these considerable
deviations is the strong eddy current dampening of this special pickup.

5.7 Hum-sensitivity 5-79
5.7 Hum-sensitivity
Magnetic pickups generate an electrical alternating voltage from a magnetic alternating field.
This voltage is the wanted signal as long as the alternating field results from the string
vibrations. All alternating fields, which are not due to the string vibration, generate, in
contrast, undesired interference. In the environment of the electric guitar the most common
source of interference results from 50-Hz-fields caused by the 230-V-power-network (or 60
Hz at 110 V in the US, or other frequencies and voltages, depending on the country and local
power system). A 50-Hz-field coupled into a guitar pickup comes through as a low-frequency
interfering tone (49 Hz equals the pitch of G1) which is called hum. Hum interference rarely is
of a single frequency – more often it is a complex tone with harmonics at multiples of the
fundamental (50, 100, 150, 200 ... Hz – or in non-European power supply systems the
harmonics of the local supply frequency). Filtering the fundamental therefore does not help a
lot.
The principle of magnetomotive force provides us with the basis for the quantitative
interference: around a long, straight conductor a magnetic fields with the flux density of
is created. In this formula, I is the current strength, r is the radial distance,
and µ0 represents the permeability of air . Accordingly, a line carrying 10 A
generates a flux density of 1 µT at a distance of 2 m. This seems not to be a lot – however, in
a coil of 10 cm2 with 10000 turns, the resulting flux is 10 µVs, after all, and the corresponding
voltage at 50 Hz is 3 mV. For a signal of 100 mV, the signal-to-noise ratio is a mere 30 dB
i.e. not a lot. In practice, things are a little different, though – not so much because the
magnets on a pickup have a fields-amplifying effect (about +2 dB) but because power current
is supplied via two-wire lines. The forward and backward flow generates anti-phasic fields
which attenuate each other in their effect. For the situation as given above this results in an
improvement of the signal-to-noise ratio by about 50 dB to about 80dB. This would seem
adequate – a tape recorder would be very happy with such a dynamic range. Guitar players,
however, are no tape recorders (even if they tend to copy and repeat licks ...). They will
overdrive their amps, depending on the style of music, by 10 – 30 dB. This again reduces the
signal-to-noise ratio in our example to as little as 50 dB, and given e.g. an SPL of the music of
100 dB (VERY moderate Hardrock), a clearly audible hum interference remains. The 50-Hz
component is not the actual issue (it may eve be below the hearing threshold, but the almost
always present harmonics will be rather disturbing. Also, power transformers, CRT screens,
fluorescent lights, switched power supplies or electrical motors can create much stronger
interference.
Fig. 5.7.1 shows time function and spectrum of two typical interference signals: the one of a
CRT screen causes an impulse-like noise, while the stray field of the mains transformer of a
power amplifier generates a distorted sinus wave. The derivative of the sawtooth-shaped ray-
deflection in the CRT-screen results in the needle-shaped peaks in the upper signal shown in
Fig. 5.7.1; it reaches about 12 mV as a maximum. This interference was recorded with a
Stratocaster about half a meter away from the screen, and while this signal does not hold a lot
of power, it may already lead to overdriving the amplifier due to its high peak values. The
spectrum diminishes only little towards the high frequencies; the guitar amplifier generates a
hard, buzzing tone. The stray flux of the power amp transformer includes mainly the 1st, 3rd
and 5th harmonics (due to saturation in the core and the hysteresis), and the peak value of the
time function is about 0,9 mV. The sound of this interferer is a low hum similar tot he sound
of an electrical bass guitar.

Fig. 5.7.1: Time function (left) and spectrum (right) of interfering signals. The upper two graphs relate to noise
due to a CRT screen, the lower graphs show the interference by a transformer. The left graph for the CRT is
scaled in Volts; the maximum value is 12 mV. The time function for the transformer is scaled in mV with the
maxima being at 0,9 mV. Both level spectra are scaled in dBµ, i.e. relative to 1 µV.
In order to obtain quantitative data on the hum-sensitivity of typical guitar pickups,

measurements were taken in an artificial interference field created via a pair of Helmholtz
coils (Beff = 6,5 µT). For singlecoil pickups, the axis of the coil was oriented in parallel to the
direction of the field while humbuckers could be rotated. The interfering voltage (measured at
500 Hz) was 0,1 – 0,2 V for singlecoils; for humbuckers the maximum was 30 mV. Taken by
themselves, these numbers are not very meaningful – however, in combination with the
transfer coefficient of the pickup it is possible to give a signal-to-noise ratio (level of the
useful signal minus the level of the interference). Of course, a pickup boasting 10000 turns on
its coil will reproduce the interfering field more strongly (i.e. louder) compared to a pickup
having 5000 turns, but the former will also generate a louder useful signal than the latter.
Consequently, the individual relation between voltage of the useful signal and the voltage of
the interferer (or the difference between the levels of these signals in dB) is the purposeful
measure.
It was striking during most humbucker trials that – in contrast to the euphoric slogans in
advertisement – the hum-rejection is rather modest. Seth Lover's statement that "the 2 coil
pickup eliminates the hum" should not be taken literally. Indeed, the very plausible basic
principle of interference compensation using two inverse wound coils requires a design which
is symmetric relative to a single central point; this is not there for your typical humbucker.
The magnet positioned below the coils bends the magnetic field and downgrades the hum-
suppression substantially.

In Fig. 5.7.2 we see the directional patterns gathered with a Gibson 490R pickup. If only a
single coil (without ferromagnetic materials) were rotated in the magnetic field, a cosine-
shaped directional pattern would be seen (direction of field, rotation axis and coil axis all
being perpendicular to each other). Due to the ferromagnetic being positioned like a "u" the
field is bent such that both coils "squint" inwards; the highest sensitivity is found to be off by
9°, directed inward from the coil axis. Interfering fields directed through the pickup in parallel
to the axis of the coils can be compensated to good effect. However, if the direction of the
field is perpendicular to the coil axis, the compensation effect is only rather moderate.
The coil with the slugs is less sensitive by 0,6 dB.
Fig. 5.7.2: Normalized directivities D of the humbucker coils: 1 = coil with slugs, 2 = coil with screws. The
pickup (Gibson 490R) subjected to a parallel AC-field of 500 Hz with both coils disconnected and measured
separately. The right-hand directional diagram shows the directivity with both coils connected on series. ϕ =
rotational angle of the pickup relative to the magnetic field.
In Fig. 5.7.2 the minimum for the series connection does not occur at ϕ = 0° (i.e. axial
direction of field). This is not due to the magnet but to the differences between the coils which
are created on part by differences in the coil winding and in part by differences in the coil
core (i.e. the screws and slugs). In practice it is rather irrelevant for which interference
direction the pickup is least sensitive since interference can come from any direction. The
player's performance will not improve if he/she needs to hold the guitar horizontally to
minimize the hum♣. Therefore, it is best to concentrate on worst-case-scenarios and consider
those interferer directions which create the strongest hum. Magnetic fields with a direction
running in parallel to the coil axis are most disturbing for singlecoil pickups. Gibson-type
humbuckers are most sensitive to hum-fields running in parallel to the strings (i.e. axis-
normal). In coaxial humbuckers (see chapter 5.3) the coil-symmetry is the decisive factor for
the direction of strongest interference – normally these pickups hum the most for axis-parallel
fields.
♣
Come to think of: that may depend on the guitar player, as well, …

Fig. 5.7.3 schematically shows the field distributions for a humbucker. The magnetic flux in
both coils is opposed if the direction of the field runs in parallel to the coil axis. This shows
that a direction-independent compensation is in principle not possible. For the Gibson 490R
and axis-normal field direction (i.e. the worst case), the anti-phase connection of the coils
reduces the interfering signal merely to one third – referenced to the interfering voltage which
would be generated in one coil by the axis-parallel field. Since the coils are connected in
series and the useful signals are summed up, we could add another 6 dB and specify the
worst-case hum-suppression to 16 dB. On the other hand, it should not be forgotten that not
all useful signal components are in fact added up: a number of partials of the string vibration
are even cancelled out completely due to the sampling of the string at two points.
Fig. 5.7.3: Shape of magnetic field running through a Gibson-type humbucker. If the field runs in parallel to the
coil axis (left), in-phase voltages are induced. However, for an axis-normal magnetic field (right) anti-phase
interference signals are created (just like the signals induced by the strings).
The directionality of the interference suppression could be described in simple formulas in

Fig. 5.7.2. For the frequency dependency we get more complex relations, however. On the
one hand, this is due to the skin-effect but also due to the capacitive coupling between the two
coils at higher frequencies. Fig. 5.7.4 depicts the frequency responses of the transmission with
an excitation in the parallel Helmholtz-field. Compared to the single-coil operation the
Gibson 490R reduces the interference by merely 10 dB (or 16 dB considering the doubling of
the useful signal with both coils in operation). For Fender's coaxial humbucker the gain is 24
dB, after all.
Fig. 5.7.4: Frequency dependency in the parallel Helmholtz field (6,5 µTeff). The pickups were loaded each with
200kΩ // 330pF. For comparison, the right-hand section shows Stratocaster pickups with comparable sensitivity.

The transmission factors shown in Fig. 5.7.4 are not quite applicable to the real-world
operation: for the interference generation, the Helmholtz coils are a good standard; however,
this does not hold for the generation of the useful signal, because the air-gap changes caused
by the string vibration have a locally very limited effect. To achieve a comparable assessment,
the following measurement method was used. For each pickup, the interference voltage
level was measured at 520 Hz and an effective flux density of 6,5 µT (worst case). To
determine the sensitivity to the useful signal, a 0,66-mm-string was moved in front of the
magnet poles in axial direction at 84 Hz and an amplitude of 0,4 mm. The S/N-ratio (level of
the useful signal minus level of the interfering signal) derived from these readings was
arbitrarily increased by 11,5 dB such that for the Stratocaster pickup – which was used as
reference – a standardized hum rejection of 0 dB resulted. Using this definition, pickups
with positive hum rejection are less interference-prone that the reference pickup. The best
results were achieved by the Joe-Barden Strat pickup and the Gretsch FilterTron – due to their
symmetric construction.
The following table lists the data taken during the measurements done according to the above
approach. The double-digit representation for the hum-rejection of the singlecoils does not
imply that indeed an accuracy of 0,2 dB was achieved. Such a high accuracy is actually not
required, anyway, since differences of e.g. 1 dB are normally not detectable. However, the
low hum-sensitivity of the Gretsch HiLo-Tron is noticed. This pickup shows that one level
alone does not have much informative value: the hum level of the SDS1 is actually 2 dB
higher – the SDS1 delivers a much higher useful signal level. On the other hand, the DP172 is
even less sensitive that the HiLo-Tron – its hum-level is however lower by almost 8 dB. We
must moreover also not forget that factors other than these pickup-parameters do play a role:
the HiLo-Tron is known for its brilliant (i.e. bright) sound and will presumably be used by
most guitar players for a more "clean" sound using little distortion in the amplifier. This is
rather different for the SDS1: with its high output and mid-range emphasis, it is predestined
for "crunch" i.e. a distorted reproduction. Distortion, however, implies high gain, and thus
relatively loud hum.

Tonabnehmer §) Hum-level /dBV Signal-level / dBV S/N-ratio / dB
"Telecaster"-Fake (Neck)
Duncan APTR-1 (Telecaster-Type, Neck)
Schaller
Fender Stratocaster (bar magnet)
Fender Stratocaster-72 (G-Magnet)
DiMarzio DP172 (Telecaster-Type Neck)
Fender Telecaster-73 (Bridge, D-Magnet)
Rockinger P-90
Fender Telecaster-70 (Bridge), w/out plate
Rockinger Strat-Type (bar-magnet)
DiMarzio SDS-1
Fender Texas-Tele (Bridge, D-Magnet)
Fender Stratocaster (USA Standard)
Ibanez Blazer
Gibson P-90
Duncan APTL-1 (Telecaster-Type, Bridge)
Gretsch HiLoTron
Lace-Sensor gold
Squier Humbucker
Gibson 490R
Gibson Burstbucker #2
Gibson ES 335 (Neck, 1968)
Gibson 57 classic
DiMarzio DP184
Gibson Tony Iommi
Gretsch FilterTron
Joe Barden (Strat-Type, Bridge)
Table: hum-rejection. Interfering field: parallel single-frequency magnetic field, f = 520 Hz, Beff = 6,5 µT.
String vibration: f = 84 Hz, amplitude 0,4 mm, distance of string to magnet = 2mm, D'Addario PL-026.
The pickup was loaded with 50 kΩ fro this measurement.
§) The actual values are reserved for the printed version of the book

The levels of hum in the Jazzmaster- and the Stratocaster-pickups differ by 7 and 8 dB,
respectively. This difference matches the ratio of the surfaces (2:1) and the assumed number
of turns (ca. 1:0,9). The stronger hum-sensitivity of the Jazzmaster-pickup would be
compensated if the useful signal level would also be stronger by 7 – 8 dB. However, the gain
relative to the Stratocaster pickup is only about 5 dB i.e. the Jazzmaster-pickup "hums more".
The difference of barely 3 dB is however not dramatic, plus the spectrum of the interference
plays a role, as well. Connected to the typical circuitry, the Jazzmaster pickup has a stronger
resonance peak than the Stratocaster pickup: in case of a broadband interference (e.g.
fluorescent lights, or phase angle control) the Jazzmaster carries both the useful signal and the
interference equally. However, if the interference has its emphasis at low frequencies
(incandescent light, or power transformers), the Jazzmaster wins out because the useful signal
is emphasized. As long as the differences in the signal-to-noise ratios (measured at 84 Hz /
520 Hz) are merely a few dB there will be no big effect noticeable in practice.
Humbuckers do play in another league: relative to a singlecoil they show a significant S/N-
gain of 19 to more than 40 dB as long as they are subject to interferers generating parallel
magnetic field lines (as generated by distant hum sources or by Helmholtz coils). A power
transformer operating close to a humbucker will generate a strong field with bent field lines
and may cause strong disturbance despite the hum-rejection effect. Moreover, the two coils of
a humbucker may not have the same sensitivity: if the number of turns or the core materials
are different, the compensation effect may be incomplete.
Many singlecoil-guitars fitted with more than one pickup feature a hum compensation via
different direction (cw or ccw) of the winding and opposed magnet polarity of the pickups. As
two pickups are selected in combination, a humbucking-effect happens. Occasionally a
compensation coil is built into the guitar – it includes no magnet and reacts only to the
interference. Connected in series with the pickups coil, and given correct dimensioning, it
reduces the interference. Since the useful signal has to travel through an enlarged inductance,
the resonance frequency decreases, as well. Changes in sound are possible. Connecting the
compensation coil in parallel (as it was tried with moderate success e.g. in the P100) increases
the resonance frequency.
In closing it should be mentioned that magnetic shielding is possible but is inefficient and
impractical. Fully encapsulating the pickup would be pointless since it could not sense the
string vibration anymore. Shielding covers around both the pickup and the string do exists,
but the musicians see them as obstacle (or best as transport safety) and remove them
(sometimes to use them as ashtrays ....) . However, shielding against electrical fields which
are capacitively coupled to the pickups from voltage-carrying lines, is possible and
purposeful. Shielding foils and conductive paint serve well for this. Still, it should be
considered that the magnetic fields will induce an eddy current into any conductive surface
which may dampen the pickup resonance. For this reason, high quality shielding covers are
made from nickel silver (German silver) and possibly in addition can include slots (see
chapter 5.9).

5.8 Non-linear Distortion
Both the measurements with the motorized test bench (chapter 5.4.4) and the measurements
with the shaker (Fig. 5.4.23) make us surmise that the functions of distance relationships are
in fact (non-linear) power functions – in the framework of a diverging magnetic field this
would not be surprising. However, sinusoidal excitations fed into non-linear functions will
lead to non-linear distortion i.e. to the generation of new frequencies. Such a system can be
seen as linear only for very small (string) excursions; the amplitudes occurring in the practical
musical application are rather strong such that the large-signal-behavior needs to be
investigated, as well.
In order to explain the basic relations, let us first look at a system with a transfer characteristic
including a linear term and a square term:
Transfer characteristic; signal
The squared sine-function can be seen as a superposition of a (constant) DC-component and

an oscillation at double the frequency:
Nonlinear distorted signal
The spectral representation of y(t) shows three components: the DC-part at 0 Hz, the first
harmonic at ω and the second harmonic at 2ω. Form this we obtain the 2nd-order harmonic
distortion k2 at:
2nd-order harmonic distortion
A value used often instead of harmonic distortion is the (2nd-order) harmonic distortion
attenuation:
Harmonic distortion attenuation
L1 is the level of the 1st harmonic and L2 the level of the 2nd harmonic. The approximation
holds, strictly speaking, only for small signal levels but will be used without constraints in the
following.
In the general case the transmission curve does not only include a 2nd-order distortion
component but further series components of higher order:
General transmission characteristic
Any continuous function can be expanded into such a series (Taylor-MacLaurin). The
corresponding spectral representation includes not only the additional 2nd order harmonic but
also further lines (higher harmonics) at integer multiples of the fundamental frequency. The
distortion components in power function decrease with the order and therefore we will regard
only the dominating 2nd-order distortion as a simplification.

5.8 Non-linear Distortion 5-87
If a nonlinear system is excited not with a mono-frequency signal but with a mixture of
frequencies, not only multiples of the fundamental frequencies result but also summation and
difference frequencies. For the ideal, dispersion-free string, exactly harmonic partial tones
are assumed, i.e. for example 100, 200, 300, 400 Hz. The difference tones generated by the
non-linearity (as described above) correspond exactly to already existing frequencies. The
flexural rigidity of real strings, however, generates dispersion and a frequency-spreading
resulting in complicated spectra. For every primary tone (e.g. 100, 201, 302.3, 404 Hz),
neighboring lines come into being which lead to additional beat-like modulations.
The typical transmission curve shown in Fig. 5.8.1 describes the correspondence between the
distance of string to magnetic pole and the magnetic flux. With the string in the still position,
the distance between magnetic pole and string is d = 2 mm in this example (operating point).
A sinusoid movement of the string with an amplitude of 1,5 mm leads to a non-linear flux
change, in which the negative half-waves have smaller value that than the positive half-
waves. The induced voltage is proportional to the flux change over time (law of induction, dΦ
/ d t), and a saw-tooth like curve results for the voltage. In this example the square harmonic
distortion attenuation is about 12 dB, corresponding to a 2nd harmonic distortion of about
25%. The 3rd-order harmonic distortion attenuation amounts to about 26 dB (k3 = 0,5 %).
The square harmonic distortion is approximately proportional to the amplitude of the string
vibration. For the above example and an excursion of 0,5 mm, k2 decreases to about 8 %, and
k3 to about 0,055 %.
Fig. 5.8.1: Curved transmission characteristic and

sinusoid excitation (left). The time-derivative of the
distorted magnetic flux has a saw-tooth shape (lower
left, the spectrum contains all whole-numbered
harmonics (below).

The following measurements were taken on the shaker test bench at 84 Hz. As was
established with an acceleration sensor, initially the shaker itself had a harmonic distortion of
k2 = 2%. This value could be improved to 0,1% via compensation – a base line which is more
than adequate in view of the much higher pickup distortions. In Fig. 5.8.2 the results for
singlecoil and humbucking pickups are shown. The string excursion was 0,4 mm for all
measurements, the clear span (distance d) between the (still) string and the magnetic pole was
varied between 0,5 mm and 5 mm.
Fig 5.8.2: Harm. distortion attenuation ak2, f = 84 Hz, excursion amplit. = 0,4 mm. String diameter = 0,66 mm.
Abscissa: distance string/magnetic pole d. For the T-Iommi-pickup, the distance string/cover is used, the
distance to the magnetic pole is thus larger, and the curve needs to be shifter right for comparison.
For all pickups the distortion decreases with increasing distance; within the relevant range of
d the 2nd order harmonic distortion amounts to 4 – 5% for 0,4 mm string excursion.
Considering that with strong picking 2 mm excursion can easily be reached, a harmonic
distortion of above 10% is possible. This is, however, not a characteristic of a special pickup
but occurs similarly in all investigated pickups. As with comb-filter responses, it is necessary
to take into account that every pickup is part of a musical instrument: one can objectively
describe its transmission characteristic but an evaluation remains a subjective affair. Since the
vibration of each string is distorted individually (without interaction with neighboring strings,
see below), the effect of the distortion is much less spectacular than the numbers would
appear to indicate. Clearly audible distortion is generated mainly in the electronics to which
the pickup is connected but not in the pickup itself.
Fig. 5.8.3 compares measurements and calculations. As a good approximation, the field
transmission characteristic of a Stratocaster pickup follows a simple power function:
Field-transmission characteristic
The levels of the first and second harmonics dependent on the distance d (left) and the
excursion amplitude (right) agree very well with the measurements. The static magnetic
flux (no string excursion) can be defined via the constant K0; for AC-considerations its value
is without importance since it disappears in the process of differentiation. The constant K1
determines the transmission coefficient. For small string excursions it is (for 7600 turns on the
coil) . Taking the magnet cross-section as the area through which the field
penetrates yields – with d = 2 mm and – a flux-density amplitude of 0,5 mT.

5.8 Non-linear Distortion 5-89
This is only a coarse estimate since the magnetic flux is not concentrated on the magnet cross-
section but spreads into neighboring areas. The area is therefore larger than assumed. At the
same time, it is necessary to consider that not all turns of the winding are penetrated by this
magnetic flux. The number of turns therefore is smaller than assumed. These two errors are
conveniently opposed and the overall estimate should not be too far off. Compared to the DC-
component of the flux density (which amounts to about 100 mT at the end face of the magnet)
the AC component is very small for the parameters as given above, and a linearization for the
calculation of the fundamental oscillation is possible without large errors. The nonlinear
behavior is described by the given characteristic with sufficient accuracy.
Fig. 5.8.3: Dependency of the level of the 1st and 2nd harmonic on the distance (left) and the amplitude (right).
The dots are measured values, the lines result from a calculation of the power function as discussed in the text.
The real string vibration does not include only a single frequency but is a frequency mixture
from many partials. If all these frequencies were exact multiples of the fundamental, the non-
linearities of the pickup would create new components exclusively at the already present
frequencies. For example, the 3rd-order distortion of the fundamental generates (amongst
other components) a tone at three times the fundamental frequency – i.e. exactly at the
frequency of the 3rd partial (3rd harmonic). However, the partials of the string are not exactly
harmonic: flexural rigidity, magnetic filed of the pickups and frequency-dependent bearing
impedances lead to a spreading of the frequencies of the partials. An E2-string tuned to
exactly 82 Hz could e.g. have a 3rd harmonic which is shifted from 246,0 Hz to 246,3 Hz. A
3rd-order distortion of the fundamental will create (amongst other components) a distortion
product at 3 ⋅ 82 Hz = 246,0 Hz which creates a beat-like amplitude change with the 3rd
harmonic (246,3 Hz). Since, however, every string vibration in reality includes modulation of
the partials anyway, the additional modulation generated by the pickup is insignificant.
Completely immaterial are nonlinear string interactions. The primary tones (f1, f2) generated
by two strings interact and produce sum- and difference-tones ( ) due to the
nonlinear characteristics. To measure this effect quantitatively, 2 neighboring strings were
strongly plucked and the pickup output voltage was analyzed. This was done for the following
pickups: Gibson '57-Classic, Gibson Tony Iommi, DiMarzio DP184, Fender Texas-Special-
Telecaster. Even with a mere 1-mm-distance between string and magnet, the intermodulation
remained below 0,1%. The main reason for pickup distortion is the magnetic resistance of the
field in air between string and magnet – this resistance being nonlinearly dependent on the
string position. The neighboring string vibrating at a distance of about 1 cm has practically no
influence on this process.

The magnetic flux changes generated by individual strings do superimpose in the magnet (or
in the field-shaping pole pieces) – but the relative flux changes are so small that the – in
principle non-linear – hysteresis may be linearized, after all. String interactions and string
intermodulation starts to play a role only as non-linear distortions appear in the amplifier.
On top of the interactions resulting from two strings, the term intermodulation could
however also be considered regarding the combination tones generated by individual partials
of one string. Strong low-frequency string excursions shift the operating point on the non-
linear transmission characteristic (Fig. 5.8.1), and as a consequence the amplitude of the
higher frequency partial changes. Again, the shaker test bench delivers quantitative data: a
D'Addario string (0,66 mm diameter, PL026) was adjusted to 2 mm distance to the magnetic
pole. A low-frequency vibration (20 Hz, 0,55 mm amplitude) was added to a higher-
frequency vibration (80 Hz, 0,23 mm amplitude), with the vibrations oriented in parallel to
axis of the magnet. As a result of the non-linearity, new spectral components appear with the
60-Hz- and 100 Hz-lines being of particular interest. In the idealized model, the two-tone-
mixture receives a 2nd-order distortion: .
. Non-linearity
With and , the new

frequencies resulting from the non-linearity can be easily calculated: next to the DC
component (0 Hz, unimportant in this context), the double primary frequencies (2ω, 2Ω) and
the sum- and difference-frequencies (Ω + ω, Ω – ω) occur. The 3-tone-mixture of Ω – ω, Ω
and Ω + ω can be interpreted as classical amplitude modulation [e.g. 3]. A more descriptive
approach: the low-frequency primary tone (20 Hz in the example) shifts the operating point
back and forth along on the curved (non-linear) characteristic, and the additionally present
higher-frequency signal (80Hz) finds a time-dependent steepness of the characteristic curve.
In the ranges of higher steepness, the output signal is stronger, and for lower steepness
correspondingly weak (Fig. 5.8.4). For the measurement, the 20-Hz-amplitude was 0,55 mm,
and the 80-Hz-amplitude amounted to 0,23 mm. For the calculation to be compared to the
measurement, the same characteristic as in Fig. 5.8.3 was used, and the correspondence is
acceptable (Fig. 8.5.4, left). The harmonic-distortion model therefore fits also well for
describing intermodulation distortions.
Fig. 5.8.4: curved characteristic with two-tone signal (left). The right-hand section shows measurements (o) and
the correspondingly calculated string-velocity-spectrum; characteristic as in Fig. 5.8.3. The shape of the curve in
the left section shows the basic relations but does not correspond to the data of the right-hand figure.

5.9 Equivalent Circuit Diagrams 5-91
5.9 Equivalent Circuits
The vibration of a string is always a composite from partials of different frequencies. The
conversion of mechanical into electrical vibrations in the pickup includes a frequency-
dependent weighting: spectral components in the vicinity of the resonance frequency are
emphasized, and those of higher frequency are attenuated. The pickup can therefore be
considered as a system with frequency-dependent transfer-function i.e. as a filter. According
to the teachings of the theory of electrical system (systems theory, e.g. [7]), the transfer
behavior of a linear system is unambiguously described by its transfer function. Linear
systems with identical transfer function have an identical filter effect, even if they are
differently constructed. The construction of a magnetic pickup seems to be simple (a wound
wire) – the frequency-dependent filter effect can, however, not be visualized this way. The
telecommunication engineer is more familiar with passive filter networks, i.e. networks
consisting of coils, capacitors and resistors. In the case of the pickup we use such networks as
replacement for the original system: as long as the network in its transmission parameters is
equivalent to the pickup, it represents a replacement (a model) the behavior of which is
investigated in place of the pickup.
5.9.1 Models and analogies
The approach in physics is to try to explain natural phenomena and make them accessible by a
mathematical description. Influencing factors, states and effects of real processes are,
however, so diverse that a complete description is impossible. For this reason, simplified
systems (models) are developed which are equivalent to the reality in a number of (but not
all) characteristics. The model-boundaries define what is to be reproduced and what is not.
Famous examples of physical models are the simple law of inertia (Newton) which is not
valid for relativistic considerations, or models of atoms. The model is the compromise
between the exact mathematical description which cannot be realized, and the variations
existing in the real world which are, however, to complex to be describable.
The analogy (the analogon) is a model-like description in a related area which is usually
better understood. To understand the resonance effects in a spring-mass-system, the electrical
engineer finds an illustration via an electrical resonance circuit; the mechanical engineer, on
the other hand, will probably prefer the opposite approach and look at an electric resonance
circuit using an electro-mechanical analogy. For all models and analogies, the respective
ranges of approximations and definitions need to be considered: if, for example, an electro-
mechanical analogy models merely the transversal movements in one plane, then torsion-
vibrations will remain without consideration. The idealized spring has a single elasticity
(Hooke) - and therefore is without any mass, which is in sharp contrast to the real spring. In
every model it is necessary to find a compromise between complexity and accuracy. Strong
simplifications lead to clear, simple structures; however, these may not be able to reflect the
effect under investigation with sufficient accuracy, or even to describe it at all. On the other
hand, a highly exact model may result in too great a variety of parameters the calculation of
which may take too long, or the complexity of which goes beyond the imagination.
Purposeful limits for the accuracy of approximation (and therefore for the complexity) are
given by the reachable measurement accuracy, the reproducibility, or whether or not a model-
specific inadequacy is in fact audible.
© M. Zollner 2002
5-92 5. Magnetic Pickups
Two approaches are often found when models are put together: inductive, concluding
generally applicable statements from a single finding (bottom up), and the deductive
conclusion from the general finding to the single event (top down). Both approaches may be
used in parallel. The laws of magnetism and those of electrical networks can be applied to all
magnetic pickups. The specific equivalent circuit, which would be not sufficiently accurate to
generalize, may be supplemented by additional components and thus be improved in its
precision (and general applicability).
Models and analogies have led to an adaptation and broadening of the meaning of familiar
terms. For example, the term ‘flow’, as it would be used in the context of water circulation,
relates to a macroscopically visible matter movement. In an electrical circuit, however, the
term is directed to the microscopic movement of electrons, while in a magnetic circuit there is
no movement (flow) at all (panta rhei ?). Nevertheless, we imagine a magnetic flow
including flow-lines, flow-density, turn-offs and junctions - very much in analogy to the
electrical circuit … which in itself is not always happening in a circle, anyway. 
5.9.2 Equivalent Circuits for Electrical Impedances
Regarded from the point of instrumentation, there are two pickup characteristics which are of
fundamental significance: its electrical impedance and its transfer behavior. Naturally, in the
end only the latter is of interest but the corresponding required system parameters can only be
determined with much effort. The impedance, on the other hand, can be determined easily and
accurately, and forms a good starting point to arrive at the transfer parameters via
calculations.
The impedance Z is the complex, frequency-dependent resistance of a two-pole element. The

term two-pole points to the fact that Z is to be determined in relation to two poles (junctions,
terminals) in an electrical circuit. If a circuit includes more than two terminals, it is possible to
define two-terminal-network impedances between any two of these terminals. Using the
complex number terminology (designated by underlining the respective character in a
formula), magnitude and phase of the impedance can be described elegantly and
economically. For this purpose, two different but each individually complete ways exist: the
polar and the Cartesian form. In Cartesian coordinates Z is constituted from a real and an
imaginary part, while in polar coordinates a radius (= amount, magnitude) and a phase-angle.
The impedance of a purely ohmic resistor is real and independent of frequency: ZR = R. The
impedance of an ideal inductance (wound up wire, coil) is imaginary and dependent on
frequency: ZL = jωL. j is the imaginary unit , which in mathematics also is designated
with i. The product jω is the complex frequency , often also termed p or s. The impedance of
♣
an ideal capacitance is imaginary and frequency dependent: ZC = 1 / jωC. If two two-terminal

networks are connected in series their impedances add up; if they are connected in parallel
their admittances are added. The Admittance Y is the inverse of the impedance Y = 1 / Z.
(This is elaborated on e.g. [7, 18, 20]).
For a real resistor and an inductivity connected in series, their impedances have to be
added up: Z = R + pL. In this example R is the real part of the impedance and L is the
imaginary part. The j contained in p is not counted as part of the imaginary part.
♣
p = σ + jω; here: σ = 0, i.e. p = jω (steady state).
© M. Zollner 2002
The magnitude of the impedance is calculated taking the square-root of the sum of the
squared real and imaginary parts. To mark the result as magnitude, two vertical lines are
added, or the symbol in the formula is written without underline. The latter form is
unfortunately somewhat dangerous, because complex values are sometimes found in
literature where the underline has conveniently been dispensed with.
Magnitude of the complex impedance Z
Equivalent circuits for impedances have the same impedance as the system they replace.
Let us take, for example, an ohmic resistor: ideally its impedance♣ should be purely real (Z
= R). However, at high frequencies the contact caps of the resistor act as a small
capacitance; the magnitude of the impedance decreases with increasing frequency. This
behavior could be reproduced by an equivalent circuit with a capacitor connected in
parallel to the resistor. On the other hand, the resistor may be built of a wound-up wire; in
that case the resulting inductive component would have to be reproduced by a coil
connected in series – possibly in conjunction with the capacitor mentioned above. For DC-
considerations neither coil nor capacitor are required; they do not hurt either, though, since
the DC-resistance of the coil is zero (i.e. no effect in a series connection), and the one of
the capacitor is infinite (i.e. no effect in a parallel connection). At 50 Hz, the contribution
of the coil and the capacitor may be so small that it can neglected. Starting from which
frequency the consideration is required depends on the desired accuracy.
In a guitar pickup the inductive component of the wound-up wire has a significant effect
already from 100 Hz. Consequently, the equivalent circuit will require at least an
inductance on top of the pure wire-resistance. The calculation of the inductance is in fact
not entirely trivial. Even very simple, symmetrical structures will require an extensive
integration which can quickly reach an undreamed of scope. Strictly speaking, every one of
the 10000 turns?? would have to be broken down into differentially small wire-pieces
which all interact with all other wire-pieces and form a complex inductance and
capacitance. As a first approximation, however, it is sufficient to supplement the ohmic
pickup resistance with an ideal coil and an ideal capacitor. The quality of the modeling
can easily be checked by measuring the frequency characteristic of both impedance of the
pickup and that of the equivalent circuit (put together from a resistor, a coil, and a
capacitor). The two measurements should agree within the desired measurement accuracy.
Even simpler is calculating the impedance of the equivalent circuit using methods of
systems theory. This approach removes the obligation to acquire an ideal coil which does
not include the inadequacies which are the reason we are making the effort to start with! If
we find that the object under measurement and the model differ too much, the model needs
to be improved with other parameters or with another – and possible more extensive –
structure. Moreover, the purpose of the model must never be forgotten: it is supposed to
reproduce the frequency response of the impedance. The model cannot and will not
reproduce the transfer behavior or the non-linear distortions.
♣
Purely real numbers are a sub-set of the complex numbers, as are purely imaginary numbers.
© M. Zollner 2002
5.9.2.1 Singlecoils with weak dampening of eddy-currents
The typical Stratocaster pickup consist of 6 magnets, two coil flanges, and would-up wire.
The impedance of such a pickup can be reproduced in the frequency range up to 20 kHz with
good accuracy with few circuit elements. As already discussed, the term „reproduce“ means
putting together an impedance-equivalent circuit consisting of few ideal, concentrated
elements (R, L, C). This equivalent circuit, described in the form of an equivalent circuit
diagram, approximates the pickup impedance in the framework of purposeful accuracy
limits, e.g. 5%.
Fig. 5.9.1 shows an electrical equivalent circuit diagram (ECD0) of a Stratocaster pickup as
well as the corresponding impedance frequency plots. For the measurements, the pickup was
connected via short leads and without further components (i.e. without potentiometers) to an
impedance meter. At very low frequencies the impedance is determined by the copper
resistance R; at a few kHz there is a resonance maximum and at high frequencies an
impedance drop-off occurs which is due to the capacitance. There is a general correspondence
between the measurement (of the real pickup) and the calculation (based on the equivalent
circuit diagram). The emphasis of the resonant peak is different, however. Obviously, the
pickup contains an additional dampening which the simple equivalent circuit ECD0 does not
model (eddy currents in the magnet, see chapter 5.9.2.2).
There are several possibilities to extend ECD0 by real dampening restores. Seen from the
point of networks theory, the corresponding impedance function is a second-order broken
rational function, since the network includes two independent storage elements (namely L and
C). In a fraction, containing (in numerator and denominator) the complex frequency variable p
with not more than the power of 2, five polynomial coefficients can be chosen freely –
corresponding to five components which may be selected freely. Therefore, apart from L and
C, a maximum of three dampening resistors may be determined independently from each
other in a 2nd-order system. It is not difficult to draw more than two additional resistors into
ECD0; the resulting new circuits can, however, be transferred into simpler circuits (with a
maximum of three resistors) using equivalence-transformations. There are in fact even several
possibilities to extend ECD0 by merely one single resistor – these support various physical
interpretations differently, although they are equivalent regarding the impedance modeling.
Equivalent circuit diagram
Fig. 5.9.1: Stratocaster-impedance, ECD0

Measurement (…), ECD-calculation (----).
© M. Zollner 2002
In Fig. 5.9.2 we see two equivalent-circuit diagrams containing two resistors each. To
arrive at the left ECD, the overall circuit of Fig. 5.9.1 was supplemented by a resistor
connected in parallel, while for the right-hand ECD, the resistor is connected in parallel to
the coil. The impedance of both ECDs approximates the measurement curve of Fig. 5.9.1
so perfectly that no difference at all can be seen anymore. Several questions result: do both
ECDs have the exact same impedance? Which ECD is correct? How do we arrive at the
values of the components?
Fig. 5.9.2:
Extended Stratocaster ECD (ECD1)
The following considerations require simple knowledge in network analysis as it is e.g.

imparted in [18, 20]. The impedance of a resistor is R, that of an inductance is pL, and that of
a capacitor is 1 / pC. For the series connections of two-poles, their impedances are added; for
parallel connections the sum of the admittances (inverse) is used. In both the ECDs a section
of the circuit consisting of two resistors and a coil is connected in parallel to a capacitor. If
both ECDs are supposed to have the same impedance, and if the capacitance is supposed to be
of the same value for both ECDs, then these two partial circuits need to have the same
impedance, as well. Without compromising the accuracy, it is therefore possible to limit the
issue of identifying the ECD-impedances to calculating the impedances of the respective
sections of the circuit mentioned above:
Fig. 5.9.3: As above, without C.
Impedance functions
It is a necessary (but normally not sufficient) requirement for the identity of the impedance
function that the impedances for f = 0 and for f = ∞ must be equal. It follows that:
Introducing these requirements into the impedance function described above, we obtain the
still missing requirement for the relationship between the inductances:
These equations enable us to set up an impedance-equivalent ECD from the respective other
ECD.
© M. Zollner 2002
It may be surprising that the two ECDs need different inductances for identical
impedances. However, surmising that the imaginary part of an impedance would be
determined solely by the reactances (L and C) is only correct for a series connection. As
soon as there is a parallel connection, the imaginary part is determined by the real
resistances, as well. Understanding this has strong implications on how the component
values need to be interpreted. Let us use a simple concrete example to exemplify this
problem:
Fig. 5.9.4: Impedance-equivalent circuits
Regarding their impedances, the equivalent-circuit diagrams shown in Fig. 5.9.4 are fully
identical, although the inductances differ by 21%. For the Stratocaster pickup, the differences
are significantly smaller, but as soon as additional iron parts are introduced into the magnetic
circuit (as e.g. in the P90), considerable differences result. In the end, this means that a precise
pickup-inductivity can only be given if the corresponding equivalent circuit is specified. It is
merely for very simply constructed pickups that the component values differ so little that
stating the ECD-topology may be dispensed with.
We can now answer the question posed above: both ECDs shown in Fig. 5.9.2 feature exactly
the same impedance, both ECDs are correct, and he values of the components can be derive
via a regression-process. If you want avoid deploying the big guns, you may vary the ECD-
component values until the measured impedance curve and the ECD-impedance correspond
with the desired accuracy. It is not as easy to answer the question which ECD is more
purposeful. An additional question could lead the way: which is the purpose of the ECD?
Normally, an ECD is put together in order to obtain a clear basis for calculations, and to be
able to recognize simple correspondences at a glance. The real use of an equivalent circuit for
the impedance only reveals itself once the equivalent circuit for the transmission has been
derived from it. Since the theory required for this process is only discussed later (see chapter
5.9.3), we will just take a quick look here: the pickup parameter measured in the easiest way
is the DC-resistance RDC. With R = RDC, it can be directly included in the equivalent circuit if
the right-hand version shown in Fig. 5.9.2 is preferred. It may also be explained using an
equivalent circuit diagram of a transformer that resistive losses are considered with a
connection in parallel to L and not with a connection in parallel to the series connection R’L’.
May be … but doesn’t have to be. From a network-theory point-of-view both circuits are
equal, and the preference is a matter of taste. The following considerations use the (Rq //L)+R
-structure (Fig. 5.9.5), and the result perfectly approximates the measured curve (Fig. 5.9.1) –
to the width of a line.
Fig. 5.9.5: Equivalent circuit diagram for the

impedance of a Stratocaster pickup
© M. Zollner 2002
5.9.2.2 Eddy currents in non-magnetic conductors
As a time-variant magnetic flux penetrates a conductor, it induces an electric circulating

voltage which results in annular eddy currents. The vibrating string changes the magnetic
resistance in the magnetic circuit and modulates the magnetic flux such that an alternating
field is superimposed over a constant field. According to the law of induction the flux dΦ/dt
changing over time leads to a voltage U which – depending on the electrical conductivity σ =
1/ρ – causes a current I.
Non-magnetic (i.e. non-ferromagnetic) conductors are found in pickups predominantly in the

form of assembly and shielding sheet metals. Not all pickups are fitted with them: the typical
Stratocaster pickup has a plastic cover but the Gibson Humbucker is mounted to a metal plate
and shielded with a metal cover. These metal sheets do have an influence on the magnetic
field even if they are not ferromagnetic (!), and therefore also on the transfer characteristic of
the pickup. The eddy currents flowing within them draw their energy from the magnetic
circuit which receives a corresponding dampening effect.
Fig. 5.9.6: The electrically conductive plate is

penetrated by a magnetic flux Φ(t) which increases
over time. The result is an annual eddy current I(t)
in the direction as drawn. This current flows in the
plate as a whole, not only on the indicated circle. It
causes a secondary magnetic field (as indicated at
the upper edge) which attenuates the primary field.
A time-variant magnetic flux Φ(t) inducing an eddy current I(t) in a current-carrying plate is
shown in Fig. 5.9.6. This eddy current again generates itself a magnetic counter-field which
attenuates the primary field such that a smaller voltage is generated in the pickup coil (not
shown). Since eddy currents depend on the temporal change of the magnetic field, they have
an effect particularly at high frequencies (skin effect, see below). Therefore, metal sheets do
not only provide shielding against electrical interference but they also deteriorate the treble
response. This is not necessarily a disadvantage – a full, warm sound may in fact be the
objective of pickup design. For a brilliant, treble-laden tone, however, eddy currents must not
have too big an effect.
There are several measures to get a handle on the eddy-current dampening. Size, thickness
and distance of the dampening metal sheet play a role, as does the material used. The eddy
current emerges as the quotient of induced voltage and electrical resistance. The induced
voltage depends on the magnetic flux; metal sheets in areas of weak magnetic alternating flux
attenuate less than sheets in areas of strong alternating flux. Sheet metal bent into a ring-shape
(e.g. for covers) may enclose a large surface with strong alternating flux; in such a case it
should be checked whether a slot could not interrupt the current flow. Thin sheets offer higher
resistance than thick ones; German silver (nickel silver) has a higher resistance than brass.
Gold-plated covers have better conductance (dampen more) than chrome plated covers - if the
gold layer is thick enough.
© M. Zollner 2002
The following table offers an overview of the specific resistances of common sheet metal
materials. German silver is often used in higher quality pickups. This metal of a silvery shine
is corrosion-resistance and of relatively high resistance.
Material ρ in Ωmm2/m Material ρ in Ωmm2/m
Copper 0.018 Brass (Cu, Zn) 0.08 (0.06 – 0.12)

Gold 0.022 Bronze (Cu, Sn) 0.08 (0.02 – 0.14)
Aluminum 0.029 Steel for strings (ferromagnetic!) 0.20
Nickel 0.070 (ferromagnetic!) German silver (60 Cu, 17 Ni, 23 Zn) 0.3
Iron 0.098 (ferromagnetic!) Alnico-Magnet (magnetic source!) 0.6
Chrome 0.12 Chrome-nickel (70 Ni, 30 Cr) 1.2
Table: Specific resistance ρ of metals
Fig. 5.9.7 schematically shows a pickup winding next to which a sheet metal forms a short-
circuit winding. The elements of the pickup winding are the DC resistance R (copper
resistance), the winding capacitance C, and the winding inductivity L. The Short-circuit
winding is characterized by RK and LK. Due to the incomplete flux-coupling k we will not find
the same flux Φ(t) penetrating both windings i.e. k < 1. The eddy-current resistance RK is
transformed up as Rw and attenuates a part of the winding (in the transformer-free equivalent
circuit on the right-hand side). The eddy currents induced into the sheet metal thus reduce the
coil inductance and increase the coil losses. The stronger the coupling (i.e. the closer the sheet
is positioned to coil) the larger the part of the coil which is shorted by Rw and the larger Rw
itself. In addition, Rw depends on the specific resistance of the sheet metal.
At low frequencies, the parallel-connection of Rw and (1 – σ)L has the effect of an

inductance, at high frequencies it has the effect of a resistance. The cutoff-frequency between
inductive and resistive behavior is . As a simplification, the eddy-
current losses can be neglected below fg while above fg the resistance increases from R to R +
Rw, and the inductivity decreases from L to σL (compare to Fig. 5.9.8).
N = number of turns of the winding σ = 1 – k2 = degree of scatter
Fig. 5.9.7: Pickup coil with short, equivalent circuit diagrams [4].
© M. Zollner 2002
Fig. 5.9.8 shows the frequency response of the impedance-magnitude for the equivalent
circuit of Fig. 5.9.7 – without capacitance, though (i.e. C = 0). For the left section the degree
of coupling k was varied. As the short circuit winding is brought closer to the pickup coil, the
coupling increases while at the same time the degree of scatter decreases. The treble loss
becomes stronger and the impedance level of the circuit is reduced. The left section of the
figure shows the effect of the variation of the resistor Rw for fixed coupling, which is
equivalent to a change in the thickness of the sheet metal, or of its material type. This measure
changes the cutoff frequency fg: with increasing resistance (thinner sheet metal, higher
material resistance) fg increases.
Cutoff frequency of the parallel connection of Rw and (1 – σ)L
Fig. 5.9.8: Effect of changing the coupling (left) and varying Rw (right). The cutoff frequency is marked by a
circle.
Fig. 5.9.9 shows impedance measurements for a Jazzmaster pickup. First a 1mm strong
sheet metal made of brass was brought in close proximity (2,5 mm) to the pickup and the
impedance measurement taken. Subsequently, the brass plate was replaced by an equally
strong copper plate (positioned at the same distance). The differences in the impedance
frequency plot are relatively small but still readily identifiable (at 1 – 3 kHz), and
moreover in good agreement with the results from the equivalent circuit diagram (Fig.
5.9.10).
Fig. 5.9.9: Measured impedance chart for a Jazzmaster pickup. The pickup was unloaded (high resonance
frequency), or loaded with 1 nF (resonance at 2,5 kHz. With the sheet of brass (left) or copper (right)
respectively, the impedance drops and the resonance frequency increases.
© M. Zollner 2002
The equivalent circuit diagram for the impedance of the Jazzmaster pickup with eddy-
current dampening is shown in Fig. 5.9.10. The ECD on the left refers to the pickup
without any dampening sheet metal. The only eddy-current losses are due to the six alnico
magnets; they can be modeled by a 72-kΩ-resistor (see the next chapter). The additional
dampening effected by the brass sheet (middle section of the figure) is modeled by the 4-
kΩ-resistor shunting about 1/6 of the overall inductivity (0,8 H). As is obvious, magnetic
losses cannot be always modeled by the same RL-element. This is because the magnets and
the brass sheet influence each other. Every eddy current changes the field geometry and
with it the individual coupling effects. The dampening caused by the copper sheet is
modeled via the right hand section of the figure. The conductivity of copper is about four
times higher than that of brass, and consequently the 4-kΩ-resistance needs to be decreased
to 1 kΩ. The partition of the coil remains since the coupling effects to the brass sheet and
the copper sheet are about the same.
Fig. 5.9.10: Equivalent circuit diagrams for the measurements of Fig. 5.9.9. Left: pickup without sheet metal;
middle: with brass sheet; right: with copper sheet. Capacitances are in pF, resistances in Ω, inductances in H.
During the experiments just elaborated sheet metals were brought into proximity of the
pickup since their geometry quality could easily be established. Of course, there is no 1-
mm-sheet-metal over or under to the Jazzmaster pickup in reality because the pickup is
housed in plastic. However, many pickups do have metal bases or metal covers which
indeed change the electrical pickup characteristics. The effects of (per se non-magnetic)
shielding materials are shown by impedance measurements with a Hoyer-pickup (from the
1960s). The P90-like coil of this pickup is shielded by a metal cover on the surface towards
the strings. Fig. 5.9.11 shows the effects of this shielding on the impedance frequency plot.
The eddy currents do not only dampen and attenuate the resonance peak more strongly; the
resonance frequency increases, as well.
Fig. 5.9.11: Frequency response of the

impedance of a Hoyer pickup. The bold lines
represent the original condition while the thin
lines refer to the cover taken off. The pickup is
loaded with 4700pF, 700pF, 0pF, respectively.
The eddy-current dampening effect due to the shielding cover attenuates the high
frequencies and reduces the reproduction brilliance. If this is thought to be a disadvantage,
the cover may be replaced by one made of plastic.
© M. Zollner 2002
5.9.2.3 Equivalent two-terminal networks
Magnetic pickups may be represented as two-terminal network or as four-pole network.

The basis for the design of a two-terminal equivalent circuit diagram is the (frequency
dependent, complex) resistance – the impedance – measured at the two terminals. In
addition, the transfer characteristic may be described by way of this equivalent circuit
diagram being extended by two further terminals yielding the four-pole equivalent circuit
diagram (chapter 5.9.4). Circuits (networks) are equivalent with respect to impedance if
their impedance-functions Z(f) correspond; topology and component values may in fact be
rather different.
The higher the order n (the number of independent storages) of the network, the larger the
number of possible impedance-equivalent but structurally different networks is. In network
synthesis three topologies are of particular significance: the resistance partial fraction
circuit (RPFC), the conductance partial fraction circuit (CPFC) and the continued fraction
circuit (CFC). For the RPFC (Fig. 5.9.12), the network analysis is done via series
connection of individual impedances, for the CPFC (Fig. 5.9.13), this is done via a parallel
connection of individual admittances, and via alternating series and parallel connections
for the CFC (Fig. 5.9.14).
Fig. 5.9.12: Resistance partial fraction circuit (RPFC)
The RPFC is highly suitable to describe magnetic pickups. The DC-resistance is - as R0 -

directly found in the diagram, the values Li are easily interpreted as components of the overall
inductance, and the loss resistances can be attributed via the transformer-equivalent. If all
resistances Ri (i ≥ 1) are finite, the impedance approaches a real, constant value at high
frequencies. If one of the resistors is omitted (Ri=∞, i ≥ 1), the impedance approaches - for
high frequencies - a straight line increasing proportionally with the frequency. The CPFC
delivers the same impedance, but the large inductance values occurring here are more difficult
to interpret, and RDC is not immediately evident, either. The CFC shown in Fig. 5.9.14 yields
RDC in a straightforward manner but is not used due to the high inductivity values.
Fig. 5.9.13: Conductance partial fraction circuit (CPFC)
Fig. 5.9.14: Continued fraction circuit (CFC)
© M. Zollner 2002
Understanding the impedance frequency plot of a resistance partial fraction circuit (Fig.
5.9.15) is made easy by - as a first step - assuming all resistances except R0 to be infinite.
What remains is merely a series-RL-circuit the impedance value of which can be
approximated by R0 at low frequencies and by ω(L1 + L2 + L3) at  high  frequencies.  For 
magnetic pickups the impedance growth towards high frequencies is not proportionally 
to  ω  but  with  a  shallower  slope,  and  this  behavior  can  be  modeled  by  decoupling  the 
partial inductances using the resistors coupled in parallel. The effective inductance now 
decreases  with  increasing  frequency  and  the  phase  angle  does  not  approach  90°  but  a 
smaller value. 
Fig. 5.9.15: resistance partial fraction circuit,

magnitude of frequency response. For the upper
curve, R1 = R2 = ∞ was assumed, for the middle
curve R1 = 2 kΩ, R2 = ∞ , and for the lowest
curve R1 = 2 kΩ, R2 = 40 kΩ.
The differences between the curves shown in Fig. 5.9.15 may seem rather small. However,
the magnitude by itself is not adequate to unambiguously describe a network. As soon as a
capacitor is connected (capacitance of the coil, or of a cable), real and imaginary part
change in different manner. It is therefore necessary to precisely model both real and
imaginary part and not only their magnitude. Depicted in Fig. 9.5.16 are impedance
frequency plots as they result from a capacitor of 1 nF being connected to the terminals of
the CPFC according to Fig. 9.5.15. For the dashed line, again R1 = 2 kΩ, R2 = ∞ was set,
the solid lines refer to the unchanged circuit (R1 = 2 kΩ, R2 = 40 kΩ). Although the
magnitudes of the impedances of the circuits without capacitor are almost identical at 3
kHz, there are large differences with a capacitive load. This clearly demonstrates that a
high precision is necessary when putting together an impedance-equivalent circuit
diagram.
Fig. 5.9.16: Frequency responses of the impedance magnitude of the circuit according to Fig. 5.9.15, with and
without capacitive load; Nyquist plot of the impedance without capacitive load (right, marking at 3 kHz.
© M. Zollner 2002
5.9.2.4 Eddy currents in the magnetic conductor
As a ferromagnetic, electrically conductive material is brought into a time-variant magnetic

field, there are two effects: the (relative to air) higher permeability of the material increases
the magnetic flux density, and at the same time eddy currents diminish this permeability.
Since eddy currents are proportional to the temporal changes of the magnetic field, the
effective permeability (and thus also the inductance) decreases with increasing frequency.
As we have already shown for the non-magnetic conductor, the eddy currents generate
active power drawn from the primary field – the pickup receives a dampening♣.
Fig. 5.9.17 depicts a cylindrical magnetic conductor axially permeated by a magnetic field
H. Examples for such a scenario are the magnets as they are found in typical Fender
pickups under each string, or the pole-pieces (slugs) of a pickup with a bar magnet. If the
field is flowing in the direction as indicated and increases over time, it induces a clockwise
flowing eddy current I. This eddy current weakens the primary (generating) field,
especially close to the axis. As a simplification we can imagine that the axial area is left
without any field at all, and a magnetic flux remains only in a thin border layer with a
depth penetration δ (skin effect). δ depends on the electrical conductivity ρ, on the
frequency f, and on the permeability µ. Permanent magnets show practically no skin effect
in the audio range due to their small reversible permeability (µr = 1.1 – 5) and their
relatively bad conductivity (≈ 0.6 Ωmm2/m). Steel behaves less favorably: at 2 kHz we get
merely δ ≈ 0.4 mm (with µr = 100). The magnetically effective cross-section is thus
reduced to 1/7th!
; [5]
Fig. 5.9.17: Cylinder with axial magnetic field H, eddy current I, and border layer with penetration depth δ.
The penetration depth δ (also called conductive-layer thickness) determines both the cross-
sectional area for the eddy current and the magnetically effective cross-section area. Both
areas are reciprocal to the square-root of the frequency, and since the square-root is an
irrational function, the impedance cannot be described with a rational function (i.e. a
function with a finite number of polynomial sections) – and therefore cannot be modeled
with a finite number of components. An equivalent circuit diagram as given in Fig. 5.9.7 is
only possible below the approximated cutoff frequency. For magnets (with small µr), this
cutoff lies above the relevant frequency range, and consequently a single loss resistor is
sufficient. The common pole pieces (with a larger µr) require a more elaborate modeling
including several R//L-two-terminal networks. Of course, the desired accuracy plays a role,
as well: the circuit behavior can always be reproduced in principle with one coil, one
capacitor and two resistors – however depending on the situation there may be
considerable differences to the original.
♣
See the theoretical derivation in chapter 4.10.4
© M. Zollner 2002
Apart from eddy currents there is a further source for losses: the remagnetization of the
iron and magnetic parts requires energy which is taken from the magnetic field, as well,
and thus requires a load resistor in the equivalent circuit diagram. Since the magnetic field
changes direction twice with every period of the signal, the remagnetization losses increase
with rising frequency. Other lossy mechanisms do exist – however, these are of minor
importance.
The following measurements were taken from the „screw-coil“ of a Gibson humbucker
(PU490). The pickup was disassembled and the screw-coil removed. Unscrewing the 6
screws leaves a coil without ferromagnetic parts. Its impedance can be described rather
perfectly by a resistor (4379 Ω), an inductance (1125 mH) and a capacitor (43 pF). Fig.
5.9.18 shows the magnitude frequency response of the impedance with and without coil
capacitance. In conjunction with the inductance, the capacitance causes a resonance
maximum at 23 kHz.
Fig. 5.9.18: Calculated magnitude of

the impedance of a coil, with (---) and
without (––) coil capacitance. The
curve with smaller impedance
belongs to the coil with all iron
removed, the curves above it refer to
the coil with mounting block and 6
screws.
Adding in the mounting block located beneath the coil does not change the impedance
frequency plot much. Not until the 6 screws are moved in place does the inductance
increase significantly: the impedance curve slides upwards. However, it does not run in
parallel with the original course. The reason is the appearance of the eddy current which –
with rising frequency – increasingly displace the magnetic field out of the screws and
partially undo the inductivity gain. Since the impedance increase of the upper curve (with
iron) is not anymore proportional to the frequency, and approximation with several RL-
sections is required (Fig. 5.9.19). The influence of the iron screws is considerable, as the
transfer frequency responses shown below in Chapter 5.9.3 also show. If eddy-current
losses are undesirable, it is possible to moderate their effects by lamination of the sheet
metals or the ferrite materials.
For a high degree of accuracy of the

approximation, more than two R//L-
two-terminal networks are required.
Fig. 5.9.19: Equivalent circuit diagrams for a PU-490 coil without (left) and with (right) 6 pole-screws and
mounting block. Capacities given in pF, resistances in Ω, inductances in mH. The frequency response of the
impedance is shown in Fig. 5.9.18.
© M. Zollner 2002
As already mentioned, the ECD of Fig. 5.9.19 is only one of several equivalent options.
The more components an ECD includes, the more topologies are possible. It is purposeful
to divide the total inductance into a series connection of R//L-two-terminal networks
(RPFC, Chapter 5.9.2.3). Alternatively, a parallel connection of RL-series circuits could
also be used (CPFC), but here the inductances require very large values, e.g. 100 H.
Although the total impedance can be perfectly approximated that way, it is difficult to
interpret such a circuit. The series circuit above mentioned series makes more sense: at
first glance the inductivity decreasing with increasing frequency is evident.
While the magnetic losses were already extensively discussed, the dielectric losses may be
looked at in a more concise manner. Non-conductors cannot carry any current – and thus
no eddy currents, either. In the case of pickups, non-magnetic non-conductors are all
insulators, i.e. coil bobbins and the wire insulation. These materials are in fact the source
of dielectric losses – this effect is rather indistinct, however (Chapter 5.5).
Magnetizable non-conductors (µ >> 1) are, for example, ferrites i.e. ferrimagnetic

materials. They may be (but don’t have to be) used for field-guiding parts (polepieces)
and/or magnetically hard ferrite magnets. The electric conductance of ferrite magnets (e.g.
barium ferrite) is very small which makes for almost no eddy currents at audio frequencies.
The resonance dampening consequently is less compared to alnico magnets. Since the
reversible permeability µrev of alnico magnets is – by a factor of 3 to 4 – larger that that of
ferrite magnets, it is possible to create a larger coil inductance with alnicos ... but: the
much higher conductance of alnico leads to eddy currents and thus again to a reduction of
the inductivity (see Fig. 5.9.9). How large the differences individually are depends on
where the magnets are mounted and which alternating flux penetrates through them. For
example, the alnico magnets mounted underneath the coil of the P-90 increase (!) the
resonance frequency by as little as 5%. Therefore no big differences could be expected if
the alnicos were to be replaced by ferrite-magnets. A stronger effect would occur, on the
other hand, from exchanging the cylindrical alnico magnets (penetrated by an alternating
field) of a Stratocaster pickup for ferrite magnets: the resonance frequency would rise by
about 10 – 15%. However, for many pickups of this type, these considerations have to
remain a pure thought experiment, because pushing the magnets out of the coil is
dangerous, and the pickup may be irreversibly destroyed.
Material ρ in Ωmm2/m Material ρ in Ωmm2/m
Steel for strings 0.20 (ferromagnetic) Hard ferrite (oxide magnet) about 1012
Nickel 0.070 (ferromagnetic) Alnico-Magnets about 0.4 – 0.7
Iron 0.098 (ferromagnetic) Magnetically soft ferrites about 106 (up to 1012)
Table: Specific resistance ρ of magnet materials.
© M. Zollner 2002
5.9.2.5 Singlecoils with strong eddy-current dampening
As soon as pickups contain other metal parts in addition to the magnets it is necessary to
check whether the equivalent circuit diagrams introduced in Chapter 5.9.2.1 are still of
sufficient accuracy. The magnetic alternating flux does not only induce a voltage into the coil
but into all other metal parts as well, and this leads to eddy currents. In this process, the
metal parts act like a shorted secondary coils. The resistance of this short (a few milliohm) is
transformed upward with the squared winding transmission ratio (e.g. 55002) and results in a
non-negligible cross-resistance in the equivalent circuit diagram. Fig. 5.9.20 shows an
impedance measurement for a pickup from Hoyer guitar (made in the 1960s). Underneath the
coils there are two bar magnets held by the base-sheet, and in addition there is a shielding cap
put over the pickup. In Fig. 5.9.21 we find a simple ECD containing the winding resistance R,
the winding inductance L, the winding capacitance C as well as an additional dampening
resistor Rq. Using this diagram, the measured curve can be approximated at 0 Hz and around
the resonance; the agreement at 1 kHz is merely moderate, however.
Fig. 5.9.20: Hoyer-pickup, impedanc frequency response. Measurement (  ), ECD1-calculation (−−−−).

On the left readings for the unloaded pickup are shown, on the right loads are connected: 4700pF, 707pF, 0pF.
L = 2,17 H R = 9850 Ω
Fig. 5.9.21: Hoyer-pickup with metal cover.
Rq = 540 kΩ C = 125 pF Equivalent circuit diagram ESB1.
The differences grow more noticeable as a customary guitar cable is connected to the pickup.
Its effect is purely capacitive in the audible frequency range; depending on the length there
will be a cross-capacitance of 300 – 1000 pF. The instrument used for the measurement
allows for a connection of 0 pF, 707 pF, 4700 pF. The larger the capacitance, the lower the
resonance frequency is. In the right part of the figure curves for different capacitive loads are
given: the eddy-current losses lead to clear deviations between measurement and calculations.
The equivalent circuit diagram presented in Fig. 5.9.21 (with a topology designated ECD1)
needs to be extended by additional components in order to achieve better agreement.
© M. Zollner 2002
To obtain a better approximation it is necessary to model the eddy-current losses with the
equivalent circuit diagram of a loosely coupled transformer (Chapter 5.9.2.2). There are
several equivalent possibilities for this. As already shown in Chapter 5.9.2.3, the series
connection of R//L-two-terminal networks is particularly easy to interpret; it is used again
here. Fig. 5.9.22 shows the extended equivalent circuit diagram (ECD2).
Fig. 5.9.22: Hoyerpickup, impedance frequency plot; 4700, 700, 330, 0 pF load; measurment and calculation are
not distinguzishable anymore. ECD2 on the right.
Whether the eddy-current losses indeed need to be modeled depends on the construction of
the pickup and the desired accuracy. In many cases (such as for the Stratocaster pickup)
already ECD1 delivers very good results. On the other hand, for pickups with additional
metal parts more or less significant discrepancies between measurement and calculation
should be expected. Fig. 5.9.23 shows the impedance frequency plots of a P-90 pickup with a
capacitive load (0pF, 330pF & 1000pF): with ECD1 there are clearly visible deviations, while
for models of higher order a perfect agreement between measurement and calculation can be
achieved for the p-90 as well.
Fig. 5.9.23: Gibson P90, impedance. Measurement (  ), ECD-calculation (−−−−−−−). Left: ESB1; right: ESB2.
Pickup without coaxial cable. Component values in H, Ω, F.
© M. Zollner 2002
At this point we will take another look at the effect of load capacities. Every pickup is loaded
with a capacitance – in fact only this way it receives its charachteristic resonance. When
putting together the equivalent circuit diagram and during the approximnation process, it is
necessary to keep this load in mind. The following example includes a simple circuit: a coil of
2 H having a copper resistance of 5 kΩ has a dampening resistor connected in parallel to it. In
one case, this resistor has 3 MΩ, in the other it has 300 kΩ. The corresponding frequency
response of the impedance magnitude (Fig. 5.9.24) shows a difference only at higher
frequencies; at 3,5 kHz both magnitudes are almost identical. However, connecting a 1-nF-
capacitor in parallel causes considerable deviations between the two magnitude curves.
Without the parallel-connected load capacitor, the impedance magnitude at 3,5 kHz is mainly
formed by the imaginary coil-impedance – compared to it the parallel dampening resistor is of
high impedance and negligible. A load capacitor connected in parallel compensates the
imaginary part created by the coil, and the real part becomes dominant. In the Nyquist curve
on the right (showing the real part of the impedance on the abscissa and the imaginary part of
the impedance on the ordinate – with the frequency as parameter) sections of two clock-wise
curved circles are shown; due to the different coordinate scaling they are distorted to ellipses.
For 0 Hz both circles start at about 5 kΩ (more precisely at 5//300 and 5/3000, respectively)
and turn upwards in a clock-wise manner. In both curves, f = 3,5 kHz is marked as a dot. The
points for which the distance to the origin is constant are indicated with a dashed line (this is
in fact a circle, but again the different scaling on the coordinates distorts it to an ellipse). As is
clearly evident, both indicated 3,5-kHz-points have an almost equal distance to the origin –
the magnitude of their impedances therefore is almost identical, but the real parts of their
impedance differ by almost a factor of two. This shows that the magnitude of the impedance
alone does not give a complete description.
Rq = 300kΩ and 3 MΩ. C = 1 nF.
Fig. 5.9.24: circuit (above), frequency response of the impedance magnitude )left); Nyquist curve of the
impedance (right). Thin line: 300 kΩ, bold line: 3 MΩ.
© M. Zollner 2002
5.9.2.6 Gibson Humbucker: coil with screws
Seth Lover, developer at Gibson, reports in [13] that every pickup coil had 4500 turns of thin
enameled copper wire. He states an AWG-#42-diameter which equals 64 µm. On the other
hand, he also notes that the wire diameter was subject to tolerances dictated by manufacture:
"The DC resistance varied, because the diameter of the wire was not constant. As the
diameter decreases the resistance increases, but if the inductance remains within certain
tolerances, then it's OK" [13]. A further manufacturing issue with early Gibson pickups were
shorts within the winding (short turns) which apparently were due to insufficient insulation;
these reduced the resonance emphasis and thus the treble content in the signal. Regarding the
magnet material, Set Lover remarks [13]: "We also used Alnico II and III, and the reason is,
that you couldn't always buy Alnico V, but whatever was available we would buy as they were
all good magnets". ISO 9000 hadn’t arrived yet.
We can assume that the early Gibson pickups were subject to substantial manufacturing
tolerances, and that therefore their transmission characteristics (i.e. their sound) included
inter-individual variances. Tom Wheeler writes in his Guitar Book [14]: "Later Humbuckers
have slightly smaller magnets and other minor differences in construction", and he does add
about the color (!) of the cosmetics "The color of the bobbin has no direct bearing of the tone"
[14]. Seth Lover, however, does not remember any changes [13]: "No, we kept the pickups
pretty much the same, they were all identical. ... Actually the PAFs weren't any better than the
later pickups that were built right." Seems to be all a question of the point of view. Tom
Wheeler [15]: "The PAF's popularity, which is unsurpassed, is a blend of performance and
snob appeal".
The patent for the Gibson-Humbucker talks about two corresponding coils. A magnetic field
emanating from an interference source induces the same interfering voltage into both coils,
and the out-of-phase (reverse poled) connection of the two coils causes the two interference
voltages to cancel each other out. Production units of the pickup, however, sported two
different coils: they included a “slug”-coil and a “screw”-coil. The slug-coil contains 6
cylindrical pins (pole pieces) of a diameter of 4,8 mm. The pins are positioned such that their
upper surface is flush with the string-facing surface of the coil while their lower surface
extends 3 mm out of the bobbin. The screw-coil holds, instead of the pins, 6 round-head
screws of a length of 21 mm and a diameter of 3,2 mm = 1/8”. The sensitivity of the pickup
can be adjusted for each string individually by rotating the screws.
The following measurements and calculations refer to a bridge pickup of a 1968 Gibson ES
335 TD – it was taken out of the guitar and disassembled (with rather mixed feelings!). The
screw-coil (Fig. 5.9.25) is penetrated by 6 screws which are screwed into an iron block at the
lower side of the coil. Unscrewing all 6 screws allows for taking off the iron block such that a
coils without any ferromagnetic parts remains. The impedance frequency plot of this coil is
shown in Fig. 5.9.26. In the low frequency range, the impedance is determined mainly by the
copper resistance, and in the middle frequency range by the reactance of the coil ωL; in the
high frequency range the reactance of the capacitance 1/(ωC) is the main factor. At the upper
end of the frequency range a pronounced resonance is clearly visible.
Resonance frequency
© M. Zollner 2002
Fig. 5.9.25: Exploded view of a Gibson-Humbuckers (according to Mike McDonald)

1 = fixed pole pin, south pole (not accessible); 2 = adjustable pole screw, north pole;
3 = bobbin; 4 = coil; 5 = wooden spacer; 6 = alnico bar-magnet;
7 = block with threads for the pole screws; 8 = metal base plate.
© M. Zollner 2002
R = 4420 Ω
L = 1125 mH
C = 43 pF
Measurement: ––––––
Calculation -------
Fig. 5.9.26: Magnitude of pickup-impedance frequency plot. Screw-coil without metal parts. Measurement and
calculation (equivalent circuit diagram) are practically identical. The pickup was loaded with 0/330/700/1030 pF.
The measurement gives a DC resistance (copper resistance) of 4420 Ohm. The inductance of
the winding determines the impedance increase at middle frequencies. The cutoff-frequency
at which the inductive reactance corresponds to the ohmic copper resistance is fg = 625 Hz.
Below fg, R dominates, above, L dominates. The wound-up coil wire does not only show
inductive but also capacitive behavior (winding capacitance), due to the neighboring coils.
Inductance L and capacitance C result in a resonance maximum in the frequency response at
the resonance frequency . The larger the capacitance is, the lower the
resonance frequency is located. The measurements were taken without and with a load (in the
form of an external additional capacitor) connected to the pickup; this way the resonance
frequency of 21 kHz at 0 pF could be lowered to 7.7 kHz (330 pF), 5.5 kHz (700 pF), and 4.6
kHz (1030 pF). The resonance shift gives additional information about the quality of the
modeling. As can be seen from Fig. 5.9.26, the measurement and the calculation agree to the
line width. The shown equivalent circuit diagram (ECD) is therefore well suitable to model
the impedance behavior of the pickup coil.
In the real pickup coil, resistance, inductance and capacitance are of course not concentrated
into one single point but differentially distributed. Every little piece of wire of the length dl
contains a partial resistance dR and a partial inductance dL, and forms a partial capacitance
dC with all other pieces of wire. Measurement and simulation (calculation) do however show
that a modeling of the impedance by concentrated elements (R, L, C) is fully sufficient.
Whether the transmission characteristic of the pickup can equally well be described this way
needs to be investigated separately (see below).
Next, the screw-coil is to be looked at in conjunction with the iron block, but still without
screws. For the measurements, the block was fixed in its normal position underneath the coil
using sticky tape. The ferromagnetic behavior of the block reduces the magnetic resistance in
the magnetic circuit; permeability and inductivity are increase that way.
© M. Zollner 2002
ECD: Screw-coil with iron block

(compare to. Abb. 5.9.26).
Fig. 5.9.27: Frequency response of the pickup-impedance magnitude. Screw coil with/without iron block. ECD.
Fig. 5.9.27 shows a comparison of the impedance frequency plots with/without the block. The
inductivity is increased by the presence of the block, which leads to a lowering of the
resonance frequency. This is not a pronounced effect, though, and could be ignored for a
simple model. A precise model requires that on top of the inductance increase, the iron losses
are reproduced, as well. To re-magnetize the iron, energy is necessary which is taken out of
the electrical circuit (re-magnetization losses). In addition, the time-variant magnetic field
causes eddy currents to be induced in the iron – again these are fed energy from the resonant
circuit. Coil and block may be though of as a transformer coil: the block represents a short-
circuit winding withdrawing energy from the pickup coil. In the end, the block is made
warmer; this effect is however so minute that the temperature-rise will not be noticeable. Still,
the dampening effect in the resonance circuit can be seen in the frequency response of the
impedance: the impedance maximum is slightly reduced.
Modeling the iron losses is complicated. In the above example it could be dispensed with, as
well, since the effect is so weak. However, when inserting the screws into the coil, substantial
iron losses occur which may not be ignored anymore. A fundamental discussion is therefore
necessary. In every electrically conductive material a time-variant magnetic field causes eddy-
currents; these flow on a circular path within the conductor. Since the cross section of the
block is relatively large, the eddy currents meet merely a rather small resistance, and the load
is of relatively low impedance. With increasing frequency, however, a displacement of the
current happens which is called ‘skin effect’. The eddy currents do not flow in the whole
block anymore, but only along the surface of the block. The cross-section available to the
current flow is reduced and the resistance increased. This resistance increase has a
dependency on frequency described with the square root. Since the square-root, however, is
not a rational but an irrational function, it is not possible to model the corresponding behavior
with a finite number of RLC-elements. An infinite number of elements are not a practicable
solution, though. Again, an approximation presents itself as a way out: the skin effect may be
modeled by a special RL-two-terminal network. The higher the requirements regarding
accuracy, the more components need to be put into his two-terminal network. For most
pickups, however, 3 – 7 elements will suffice (Chapter 5.9.2.2).
© M. Zollner 2002
The block is modeled in Fig. 5.9.27 by three additional components (5 mH, 50 mH, 10 kΩ).
The two coils increase the overall inductivity (ferro-magnetic effect of the block), at high
frequencies, however, the 50-mH-coil does not have its full impact anymore due to the
resistor connected in parallel (modeling of the skin effect). With this ECD, the reproduction
of the impedance is of such accuracy that it agrees with the measurement up to line-width.
Fig. 5.9.28: Comparison: screw-coil with block, without (––––) nd with screws (-----).
The impedance changes clearly once the 6 pole-screws are inserted into the bobbin (Abb.
5.9.28). The ferromagnetic screws are positioned in the area of strong magnetic alternating
flux; they substantially reduce the magnetic resistance and thus increase the coil inductivity.
On the other hand, the screws also produce losses, which is why the ECD requires additional
(ohmic) loss resistors.
Further difficulties arise from combining coil, screws, and block: the screws change the
distribution of the magnetic field in space. With the screws, the magnetic field permeates the
block in a different manner than without the screws. Consequently, the block-ECD, which is
valid for the setup without screws, cannot be used once the screws are added. Let us be
reminded at this point that the equivalent circuit diagrams present here do not model the
distribution of the field in space but are equivalents for the impedance. It is only possible to
explicitly identify the resistor R effective for DC. All other elements of the ECD are the result
of an approximation without physical correspondence.
Fig. 5.9.29: Equivalent circuit diagram (ECD) for “screw-

coil with blocjk and screws”, Gibson-Humbucker.
Compare to Fig. 5.9.30.
© M. Zollner 2002
In Fig. 5.9.29 we see an equivalent circuit for the screw-coil with block and screwed-in
screws. The impedance frequency plot is very well approximated with it, as shown by
measurement (––––) and calculation (-----) in Fig. 5.9.30; both families of curves are
practically identical.
Fig. 5.9.30: Frequency response of impedance: “screw-coil + block+ screws”.

Measurement (––) = calculation (----).
The circuit according to Fig. 5.9.29 is not the only one possible – there are several equivalent
replacements with the same high quality of approximation. Fig. 5.9.31 shows two simple
equivalent circuit diagrams, both including a resistor and a two independent coils each.
Consequently, the impedance functions are of 2nd order and include the frequency to the
power of 2, 1 and 0. Equating the corresponding polynomial coefficients (comparison of
coefficients), we obtain 3 requirements for the 3 components; this enables the conversion of
the component values from one circuit into the components of the other. For circuits of higher
order (i.e. additional coils), there are still more different topologies with the same impedance
frequency plot. Which equivalent circuit is used in the end remains a matter of taste. The
supplement via series connection of RL-parallel-circuits as proposed in Fig. 5.9.29 appears
more purposeful than the parallel connection of RL-series circuits, though (see also Chapter
5.9.2.3).
Fig. 5.9.31: Two equivalent circuit diagrams wth equal imepdance frequency response.
The only purpose of the equivalent circuit diagrams present here is to deliver impedance
frequency plots equal to those obtained in measurements, and to create the basis for
equivalent circuit diagrams of the transmission. As soon as a purposeful compromise between
component-complexity and accuracy was found, the approximation was successfully applied
and not optimized further.
© M. Zollner 2002
The consideration about impedance related to the coil with screws and block. Now we add the
bar-magnet, finally transforming the coil which by itself is insensitive to string vibrations into
a pickup. The effects of the magnet on the transmission behavior are of existential
importance; its influence on the impedance is, however, very small. It does belong to the
group of ferro-magnetic materials but its permeability at the operating point is relatively
small, and moreover it is not within the coil but positioned to the side of it. Fig. 5.9.32 shows
the frequency response of the inductivity with and without magnet. Adding the magnet makes
for a slightly larger inductivity (compare to Fig. 5.9.33), for a quality factor which is a bit
reduced, due to the eddy currents induced in the magnet, and in the end for a slightly better
coupling of the two coils.
Fig. 5.9.32: Frequency response of the impedance: “screw-coil + block+ screws”;

without (–––) / with (-----) magnet
The effects of the magnet on the inductivity show most at mid-range frequencies. Fig. 5.9.33
depicts an enlarged part from Fig. 5.9.32. The inductivity increase amounts to merely 2 – 3 %.
Fig. 5.9.33: Enlarged section from Fig. 5.9.32, without (–––) / with (-----) magnet.
© M. Zollner 2002
Fig. 5.9.34: Impedance: screw-coil + block+ magnet; without (–––) / with (-----) sheet metal
As a next step, the sheet metal forming the base plate of the pickup is added in. It consists of
German silver, which is a non-ferro-magnetic material. Its specific resistance is 0.3 Ωmm2/m,
compared to 0.018 Ωmm2/m for copper. Still, even with this higher resistance eddy-current
losses cannot be completely avoided, as Fig. 5.9.34 shows. The resonance quality factor has
again gone down compared to that of Fig. 5.9.32.
Fig. 5.9.35: Impedance: screw-coil without any metal parts (–––);

screw-coil + block + screw + mgnet + base plate (-----); loaded with 330 pF and 1030 pF.
Fig. 5.9.35 indicates a summary of the influences of the metal parts on the impedance. The
increase of the inductance is due to the screws, the decrease of the emphasis is caused by the
screws and the base plate. The metal pickup cover had been removed in the past – it would
have further reduced the emphasis.
© M. Zollner 2002
Now, the fully assembled slug-coil is mounted next to the screw-coil, but it is not yet
electrically connected. The measurements are still directed to the screw-coil; the objective is
to clarify whether the magnetically soft slugs increase the inductivity of the screw-coil. As
Fig. 5.9.36 shows, this is not really the case; the two measurements differ merely by the line
width. The ferromagnetic slugs are positioned so far away from the screw-coil that they have
little disturbing effect on the latter’s magnetic field. The magnetic coupling of both coils is,
however, not zero (Chapter 5.9.2.8)!
Fig. 5.9.36: Impedance: complete screw-coil without (–––) and with slug-coil next to it (-----);
loaded with 330 pF and 1030 pF.
Finally, Fig. 5.9.37 shows the equivalent circuit diagram for the impedance of the complete
screw-coil. We thus have arrived at an impedance model for an individual specimen of a
humbucker taken from a ES-335-TD made in 1968. On the basis of this pickup we could
exemplify the influences of various pickup components; however, we must not expect that
every Gibsom Humbucker can find its exact description in this equivalent circuit diagram.
Manufacturing tolerances and modifications have led to different versions and thus to
different data.
Fig. 5.9.37: Equivalent circuit diagram for the impedance of the screw-coil of the fully assembled pickup.
© M. Zollner 2002
5.9.2.7 Gibson Humbucker: coil with slugs
The Gibson Humbucker contains two coils: a coil with screws, and a coil with slugs. In order
to accomplish the hum-suppression, both coils should feature the same electrical
characteristics. Comparative measurements yield an inconsistent picture (Fig. 5.9.38): the
neck pickup of the ES-335 under investigation indeed had two equivalent coils, while for the
bridge pickup (of the same guitar) there were differences. The slug-coil featured an 11%
smaller DC resistance and a 13% smaller inductivity compared to the screw-coil; most
probably the coils differ in the number of turns. We cannot determine anymore whether this
lack of symmetry was on purpose, or happened by accident during manufacture (in 1968). In
any case it seems not entirely undesirable; otherwise Gibson would not offer, in the form of
the new “Burstbucker”, a pickup which replicates the uneven number of turns found in old
humbuckers.
Slug-coil:
0,98 H, 3928 Ω, 47 pF.
Screw-coil:
1,13 H, 4420 Ω, 47 pF.
Fig. 5.9.38a: Comparison

screw-coil (---) vs. slug-
coil (–––). Only bobbin and
wire for a ES335-bridge-
pickup.
1000pF, 700pF, 330pF, 0pF.
Slug-coil:
0,95 H, 3693 Ω, 40 pF.
Screw-coil:
0,96 H, 3660 Ω, 47 pF.
Fig. 5.9.38b: Comparison

screw-coil (---) vs. slug-
coil (–––). Only bobbin and
wire for a ES335-neck-
pickup.
1000pF, 700pF, 330pF, 0pF
The impedance frequency plots shown in Fig. 5.9.38 were measured with the coils taken off
the pickup assembly, i.e. there were only the bobbins wound with the wire, and no metal parts
were included. In the assembly process, the inductances increase due to the effect of the ferro-
magnetic metal components (Fig. 5.9.39).
© M. Zollner 2002
6 ferro-magnetic metal cylinders (“slugs”) are inserted into the slug-coil (∅ = 3/16" = 4,8
mm), and 6 ferro-magnetic metal screws are screwed into the screw-coil (thread-∅ = 1/8" =
3,2 mm, head-∅ = 3/16" = 4,8 mm); in both coils the inductance is increased by these metal
parts, while the resonance emphasis is decreased – but not in the same way. With the neck
pickup (having rather similar “empty” coils, Fig. 5.9.38b) there is a larger inductivity increase
at low frequencies in the screw-coil; from 1 kHz, however, the slug-coil features the larger
inductivity. The resonance emphasis of the screw-coil decreases more strongly that of the
slug-coil (Fig. 5.9.39b). These effects also appear for the coils of the bridge pickup (Fig.
5.9.39a) but the different numbers of turns hamper any analysis.
Fig. 5.9.39a: Comparison:

screw-coil with screws, with
block (---); slug-coil with
slugs (–––). Without bar
magnet and base plate
ES335-Bridge pickup.
1000pF, 700pF, 330pF, 0pF.
Fig. 5.9.39b: Comparison:

screw-coil with screws, with
block (---); slug-coil with
slugs (–––). Without bar
magnet and base plate
ES335-Necl pickup.
1000pF, 700pF, 330pF, 0pF.
Screw-coil Slug-coil
Fig. 5.9.40: Equivalent
circuit diagrams for the
impedances in Fig. 5.9.39
Bridge pickup (top)
Neck pickup (bottom)
© M. Zollner 2002
5.9.2.8 Gibson-Humbucker: coil-coupling
The two coils of a typical Gibson Humbucker (e.g. 490R; but not P-100) are connected in
series. Therefore, it could be expected that their impedances add. For the DC resistance this
assumption is correct; for frequencies which are not zero, however, there are deviations. The
measured impedance of the series connection is larger than the sum of the individual
impedances. The reason for this is the magnetic coupling of the two coils, and a different
provision for the addition results (compare Chapter 5.5.2).
Two series-connected inductivities L1, L2 carry the same current I. If there is no magnetic
coupling, a voltage jω⋅I⋅L1 and jω⋅I⋅L2 across them is created, respectively. However, if the
magnetic flux generated in one of the coils partially or completely penetrates the other coil, it
will induce an additional voltage there, depending on the coupling factor [20, Band II]. The
coupling factor k is zero for non-coupled coils; it is +1 for coils ideally coupled in the same
sense and -1 for coils ideally couple in the inverse sense. Ideal coupling is not possible in
reality; therefore, the magnitude of the coupling factor needs to be always smaller than 1.
Equal-sense coupling implies that the coil voltage is increased by the coupling – this is the
case for humbuckers (0 < k < 1). As two coils coupled by a joint magnetic field are connected
in series, the overall inductivity of this series connection is
Summed inductivity
In Fig. 5.9.41 we see the frequency responses of the impedances of the (separately measured)
coils of a Gibson Humbucker. The pickup was fully assembled and the coils isolated against
each other. A distinct non-symmetry is recognizable with the screw coil having a higher
impedance. Connecting the two coils with the customary polarity in series results in an overall
impedance which is not the sum of the individual impedances but one which has an about
20% larger value. Only at very low frequencies the overall impedance corresponds to the sum
of the sum of the individual impedances (Fig. 5.9.42). The coupling factor of the two coils
thus amounts to 20%, i.e. a fifth of the magnetic flux created by one coil penetrates the other
coil as well and induces a voltage there. If one coil would be connected with reverse polarity,
the overall voltage would decrease by 20% – no measurements for this are shown here.
Fig. 5.9.41: Frequency repsonses of impedances: ES-335-bridge-pickup, slug-coil (left), screw-coil (right).
Capacitive load: 1030 pF, 700 pF, 330 pF, 0 pF. For the measurement, the coils were insulated against each
other.
© M. Zollner 2002
In the left part of Fig. 5.9.42. the calculated sum of the individual impedances is shown; the
solid curve depicts the measurement results. In the right-hand section of the figure, the
equivalent circuit diagrams of the individual coils formed the basis of the calculated curve for
the sum (Chapter 5.9.2.6 and 5.9.2.7), with all inductances and loss resistances being
increased by k = 20%. With this, the agreement with the measurement results is much better –
the only discrepancy shows for the connected coils without capacitive load at the 15-kHz-
resonance. Apparently there is also a capacitive coupling of the two coils which, however, can
be ignored in the framework of practical operation.
Fig. 5.9.42: Frequency response of the impedances: ES-335-bridge-pickup. Model wthout coupling of magnetic
fields (left), with 20%-coupling of the magnetic fields (right). External capacitve load: 1030 pF, 330 pF, 0 pF.
Replacing the bar magnet (Alnico 2 vs. Alnico 5) may change the coupling factor; no measurements shown here.
In principle all pickup coils of a guitar are magnetically coupled. However, since the coupling
factor decreases rapidly with increasing coil distance, the magnetic field coupling has any
significance only for the neighboring humbucker coils. Just to be safe, a Stratocaster (3x
singlecoil) was also analyzed: indeed, the coupling factor could be measured but at k = 0,5%
has no significance at all.
© M. Zollner 2002
5.9.3 Equivalent circuit diagram for the transmission
In the previous sections we presented two-terminal equivalent circuits modeling the frequency
response of a pickup. The actual aim is, however, to describe the sound of the pickup – or,
more precisely, its transfer characteristics. Using the two-terminal-network theory discussed
in chapter 5.9.2, the electrical circuit (the network) is investigated regarding two terminals
(network nodes); for a pickup, these are the two connecting terminals at which a complex
impedance is measured. In contrast, for the quadripole theory two of the network nodes are
defined as input port and the other two as the output port; therefore occasionally the term two-
port-theory is found instead of quadripole theory. At each of the two ports two-terminal
impedances may be defined, but more important is how the signals at one port depend on the
signals at the other. For an electrical network, the signals at the ports are voltage and current.
Between the nodes of every port we find a voltage U that as a special case may be zero, while
the currents I flow in the connecting wires to external systems; again I may be zero as a
special case.
In order to make the complicated transfer behavior of a pickup describable, it needs to be

simplified. This is achieved by defining the pickup as a linear, time-invariant system of finite
order. Linearity implies among other things lack of sources and proportionality between
input and output signals. “Lack of sources” means that the pickup does not contain any signal
source – which is a matter of course in the framework of the usual approach, save for noise
disturbances (Chapter 5.12). “Proportionality” stands for an output signal multiplied by k if
the input signal is changed by the same factor of k. This is only approximately true for a
pickup, as distortion measurements show (Chapter. 5.8). For small levels, a pickup is a linear
system but for strong string vibrations a non-linear model is required. Time-invariant means
that the pickup always behaves in the same way: a condition generally met with good
approximation. The order n of the model marks the number of free energy storages, in other
words the number of independent capacitances and inductances within the equivalent circuit.
The more precise the model is supposed to be, the higher the order will be (it does – in this
context – not indicate the contrary of disorder). For usual equivalent circuits of pickup an
order of n = 2 … 5 is to be expected.
Using the simplifications mentioned above we can determine a broken rational transfer
function of n-th order, which maps the input signal onto the output signal. The output signal is
the voltage at the output terminals – but what is the input signal? If one does not want to
immediately go back to the action potentials of the guitarist, we could define the string
vibration as the input signal. This, however, is a spatially distributed vector field difficult to
describe. Again, several simplifications are required: measurements of the aperture (Chapter
5.4.4) suggest that only the string section vibrating directly in front of the magnet causes the
significant change in the field, and within this again predominantly the magnet-axial
(fretboard-normal) component. Movement of the string changes the magnetic resistance and
modulates the flux generated by the permanent magnet. A DC-source and a time-variant
impedance may model this process in the equivalent circuit, or one can imagine the magnetic
AC-flux as being generated by an AC-driven transmitter coil. This transmitter current is then
imagined to be proportional to the string movement; the latter in turn can be presented in
several ways: as deflection, velocity or acceleration.
© M. Zollner 2002
The transfer function describes the projection of the string movement (the input quantity) on
to the pickup voltage (output quantity) as a complex frequency function H(jω). The spectrum
of the pickup voltage U(jω) is a complex frequency function, as well, in contrast to H(jω);
however, it is not a system quantity but a signal quantity. U(jω) is dependent on the input
signal but H(jω) is not. For an excitation with a known input spectrum E(jω), the
corresponding output spectrum U(jω) can be calculated in the linear model:
From a systems theory perspective, each one of the three motional quantities of the string
(deflection, velocity, acceleration) could be defined as input quantity, but using the string
velocity is particularly purposeful and can be interpreted well. The pickup-transfer-function
used in the following is thus the velocity→voltage-transfer-function H = HUv, the first index
(U) of which points to the generated output-quantity while v yields the input quantity which
causes the effect. For the magnetic pickup, HUv is a low-pass function which sometimes
raises the question whether such a pickup can actually generate a “0 Hz”-signal. In this
respect, we need to consider that – as pointed out above – the output spectrum is not only
dependent on HUv, but also on the input spectrum. At 0 Hz, the string is without movement, its
velocity is thus zero and therefore the output voltage is zero as well – although HUv is not
zero.
In order to now put together a purely electrical transfer equivalent circuit we need to find an
electrical input quantity matching the string velocity. The cause for the induced pickup
voltage is the changing magnetic flux, the instantaneous value of which depends on the
distance of the string to the magnet: the closer the string to the magnet, the larger the flux.
Since the distance is equivalent to the integral of the string velocity over time, and since the
spectral operation corresponding to this is a division by jω [6], it is possible to use as
equivalent model a transmitter coil positioned on the pickup. This coil – excited by a current
source with a 1/f-characteristic - generates a magnetic alternating field.
In other words: to obtain a frequency-independent velocity-amplitude, the amplitude of the

deflection is reciprocal to the frequency, and so is the amount of the magnetic AC-flux.
Whether this AC-flux is generated by a moving string or instead by a current-excited
transmitter coil is – in the framework of this model – equivalent.
In the following the velocity → voltage-transfer behavior of the pickup is presented with a
low-pass-model. The input quantity is generated by an ideal current source with frequency-
reciprocal amplitude I ∼ 1/f while the output quantity is the pickup voltage.
Special emphasis is put on the fact that per pickup in the two-terminal- and in the quadripole-
equivalent-circuits one and the same components are used. If the basic models are viable it
needs to be possible to derive a two-terminal-equivalent circuit from an impedance
measurement (doable with little effort), and to further determine – with the components
calculated for the two-terminal-equivalent – also the quadripole-equivalent-circuit and the
transfer function HUv. As a precaution it is mentioned again that HUv is not identical to the
spectrum of the pickup voltage; for the latter the velocity spectrum is required on addition
(Chapter 1 – 3).
© M. Zollner 2002
Fig. 5.9.43 shows the quadripole-equivalent circuit for a Stratocaster pickup. Structure and
component values are taken from the two-terminal equivalent circuit (Fig. 5.9.5). The
alternating field caused by the string may be imagined to be generated by a transmitter coil
driven by an impressed current I. In the left-hand picture, the transmitter coil is the primary
winding of the transformer with the current source marked by the broken circle. The current
source may be transformed over to the right-hand side of the transformer (right-hand picture);
this merely changes the amount of the current, and the primary winding becomes redundant
and is dropped. From the transformer, only the pickup inductance (2,2 H) remains.
Fig. 5.9.43: quadripole-equivalent-

circuit of a Stratocaster pickup.
We can now define the quotient of output voltage and input current as the transfer function:
H(f) = U2(f) / I1(f). If the current amplitude is equal at all frequencies, the result is a band-
pass-characteristic (HUξ, Fig. 5.9.44, left-hand side). However, an excitation with frequency-
reciprocal current amplitude is more easily interpreted since it yields a low-pass-
characteristic (HUv). The left-hand section of Fig. 9.5.44 shows the results of measurements
taken for field-coupling with a small transmitter coil (constant current amplitude) wound
around the 6 magnets extending from the Stratocaster pickup; the pickup was loaded with
4700, 1000, 330, and 0pF resp. In addition, calculations based on the equivalent circuit shown
in Fig. 5.9.43 are included. The two sets of curves are practically identical; any differences are
hardly noticeable. Using a frequency-reciprocal current amplitude I ∼ 1/f instead of the
frequency independent current brings us to the right-hand section of the figure (HUv, low-pass
ECD).
Fig. 5.9.44: Transfer function for current impression (= field impression); left: band-pass, right: low-pass
The low-pass transfer function shown on the right are normalized such that for low
frequencies we obtain a transfer factor of 0 dB. Using the field coupling, only relative
frequency responses can be measured since the strength of the static magnetic flux is not
captured. The absolute scaling (i.e. the vertical position of the plots) may be determined on
the shaker test stand (Chapter 5.4.5). Given a capacitive load, every magnetic pickup shows a
low-pass characteristic (LP-ECD) including a resonance arising from the pickup inductance
and the capacitance of the pickup plus cable. The resonance emphasis is high for a purely
capacitive loading; it drops off with the resistances of typically connected potentiometers
(Chapter 9) coming in.
© M. Zollner 2002
As already shown, various transfer functions (HUv, HUξ, ...) may be defined from a systems-
theory point-of-view. However, besides the analytical description also a visual recognition
and evaluation of the shape of the frequency dependency is desirable, and here we find a
similarity between the third-octave spectra often used in acoustics on the one hand, and the
velocity→ voltage-transfer-function HUv.
In Fig. 5.9.45 we see the third-octave spectrum of the pickup voltage of a Stratocaster. The
guitar was in its original condition and externally loaded with 700 pF (cable). All strings were
plucked repeatedly in quick sequence while being slightly dampened at different neck
positions with the left hand. The left hand at the same time fingered various bar chords
without pushing the strings fully down onto the frets. This generated a wide-band noise-type
signal without too many tonal components (which could have disturbed the spectrum). The
third-octave spectrum obtained via main- and auxiliary-third-octave analysis (according to
DIN) is shown as a polyline. Between 100 Hz and 4 kHz there is a nearly horizontal
characteristic while below 100 levels drop off – the fundamental frequencies of the strings do
not cause us to expect anything else. Above the pickup/cable resonance (3 kHz) we find a
treble attenuation. Except for the low frequency range, this analysis matches a low-pass model
quite well while no similarities are recognizable to a 3-kHz-band-pass.
Fig. 5.9.45: Fender Stratocaster.

Measured levels of third-octave spectrum
(line); calculated LP-transfer function
(dots). The pickup was loaded with an
external capacitance (700 pF).
As a supplement, the HUv-transfer function is indicated as dots in Fig. 5.9.45. It was

calculated on the basis of the quadripole equivalent circuit (Fig. 5.9.45) to which a slight
treble attenuation of 1,8 dB/Oct. was added. Both plots globally show strong similarities;
moreover the two do not have to correspond in detail since these are two fundamentally
different quantities. Of course, the third-octave-level spectrum depends on the strings and the
way the playing style while the transfer function does not. The latter is a system quantity and
as such gives the transfer characteristic in a time-invariant and signal-independent manner:
the pickup resonance is, for example, independent of which music the guitarist is playing at
the time. The voltage spectrum, however, is dependent on the string excitation and the transfer
characteristic.
That the correspondence in Fig. 5.9.45 nevertheless is so pronounced indicates that the choice
of the low-pass transfer function is a good one. Although the other choice for the band-pass
transfer function would be scientifically also possible, a visual analysis is much more
difficult.
© M. Zollner 2002
The quadripole equivalent-circuit-diagram describes the transmission from the magnetic field
excitation (input port) to the connectors (output port); it can serve to determine – with little
effort – the transfer function (“frequency response”) of the pickup. The components of the
quadripole-ECD are the same as those of the two-terminal equivalent circuit and may be
determined via impedance measurement and curve fitting, or via special methods of network
analysis. The magnetic alternating-field-source corresponds, in the electrical quadripole-ECD,
to an AC current source fitted in parallel to the inductance dampening the eddy-currents (Fig.
5.9.46, right hand section).
Quadripole equivalent circuit diagram.
Fig. 5.9.46: Calculated and measured low-pass transfer-function. Stratocaster pickup, 700pF.
Measurement: using coupling of themagentic field; calculation: using the quadripole ECD shown on the right.
Comparing the measurement and the calculation shows how accurate the quadripole-ECD is.
To achieve the coupling of the magnetic alternating field there is a choice from several
possibilities:
1. Using a pair of Helmholtz coils we can generate a homogeneous, quantitatively well

defined field which however is in its shape very different from the locally limited AC-field
caused by a string.
2. A small coaxial coil positioned on a magnet (e.g. Stratocaster) or pole-screw (e.g. P-90)
yields a locally limited field, which however does not yet correspond to the field generated
by a string.
3. Much closer to reality is the excitation with a tripole-coil. For this, we wind …..
This text part is remains reserved for the print version of this book.
Fig. 5.9.46 contains 4 plots: one each for the measurement with Helmholtz-, coax- und
tripole-coil plus one calculated from the quadripole equivalent circuit diagram. Up to 10 kHz
all four correspond perfectly (ΔL < 0,5 dB). Only in the highest octave we find differences:
the excitation with the Helmholtz-coil (dashed line) yields the highest levels while the lowest
levels are given by the other two excitation coils with the calculation (solid line) positioned in
between. The pickup ware loaded merely with a capacitance; further loading (controls,
amplifier) would reduce the resonance emphasis for all plots in a similar manner.
© M. Zollner 2002
The transfer functions depicted in Fig. 5.9.46 are normalized in such a way that their position
accommodates a level of 0 dB at low frequencies. While the shape of the curves is thus set,
the absolute scaling remains undetermined. To know the vertical position of the transfer
curves, measurements with a real vibrating string are required – these need to be taken at
merely one single frequency, though. It is purposeful to choose a frequency at which few
artifacts can be expected.
As has been shown, the results of measurement and the calculation agree very well for the
original Stratocaster pickup, which features only little eddy-current losses. This agreement is
not as good for the pickup variant manufactured in Japan (Fig. 5.9.47) the construction of
which is based on a bar magnet and 6 iron slugs rather than on 6 cylindrical alnico magnets.
In the low and middle frequency ranges a perfect agreement does remain between the
measurements with a tripole excitation (----) and the calculation while for higher frequencies
there is a larger difference although this range is less important for electric guitars.
Fig. 5.9.47: Calculated and measured lowpass-transfer function. Fender-Japan-Strat, 670pF load.
Measurement with tripole-coupling (---), calculation with the quadripole ECD shown on the right.
The situation turns out differently for the Telecaster neck pickup (Fig. 5.9.48): there is a clear
divergence between measurement (----) and model calculation. We can pinpoint the reason in
the metal cover the eddy-currents of which necessitate a modified equivalent circuit diagram
(see Chapter 5.10).
Fig. 5.9.48: Calculated and measured lowpass-transfer function. Telecaster (S. Duncan), 680pF load.
Measurement with tripole-coupling (---), calculation with the quadripole ECD shown on the right.
© M. Zollner 2002
The equivalent circuit diagrams shown in Fig. 5.9.47 and 5.9.48 contain three different
inductances. Whether this effort is justified can be determined on the basis of the desired
accuracy. Using the Gibson screw-coil discussed in Chapter 5.9.2.6 as an example, Fig. 5.9.49
compares the transfer function derived from a 4th-order model (n = 4, Fig. 5.9.37) with the
transfer function calculated on the basis of a simple 2nd-order low pass (n = 2). Whichever
way we approach the alignment (whether going for identical maxima – left-hand section – or
for equal high-frequency asymptote – right-hand section): the resonance of the 2nd-order plot
comes out too broadband. In other words: if we are looking for more than just a coarse
approximation, the more exact modes should be preferred – the effort is manageable.
Fig. 5.9.49: Transfer functions calculated on the basis of models of different complexity.
5.9.4 Pickups connected in combination
Connecting two pickups in parallel changes the sound in two ways: due to the halving of the
inductance the resonance frequency rises by 40%, and in addition we get an interference filter
similar to a humbucker – although with a larger distance of the coils and negligible coil
coupling. Fig. 5.9.50 shows the frequency responses for a coil distance of 6 cm; this
interference filter is shifted further towards lower frequencies for the combination of neck-
and bridge-pickups (d = 12 cm). The resonance emphasis grows because the source
impedance is cut in half.
Fig. 5.9.50: Transfer frequency response and interference filter (d = 6 cm), pickups cinnected in parallel. The
interference effect is specific to the string; the bold line holds for the E2-string.
© M. Zollner 2002
5.10 Transfer characteristic 5-129
5.10 Determining the transfer behavior
In Chapter 5.9 the purpose of an equivalent circuit diagram has already been explained; it is
supposed to represent the transmission characteristics of a pickup in a form easily
understandable to the electronics expert. How far this is successful shall now be examined for
some selected pickups via control measurements. First, we have to specify which
transmission is meant. The generated effective quantity is the electrical voltage created at the
pickup terminals for a defined electrical load, and the source quantity generating this effective
quantity is the string velocity perpendicular to the fretboard. These two quantities give the
transmission coefficient HUv. We could, in addition, also use the string velocity in parallel to
the fretboard, or the strain-wave velocity running along the string – but for the following, let
us limit ourselves to the string velocity oriented perpendicular to the fretboard (fretboard-
normal velocity).
To model the transmission behavior the string-plucking guitarist is advantageously replaced

by a source which can be described more precisely. Such a source could be a generator coil
creating the magnetic field, positioned coaxially with the pickup coil or orthogonally to it, or
it could be a short string moved by a shaker, or an impulse-excited, laser-monitored long
string. It may be noted already in advance that all control measurements confirm the
suitability of most of the equivalent circuit diagrams – modifications are required merely for
pickups with strong eddy-current dampening.
5.10.1 Measurements using a shaker
For the shaker-measurements, a string of 10 cm length is driven by a B&K-Shaker (Type

4810) such that it vibrates – along a sinusoidal curve – orthogonally to its longitudinal axis
while keeping its shape (no string bending). The string acceleration is frequency-selectively
monitored via a PCB-impedance-measuring head (Type U-288) connected to a DFT-analyzer.
In most cases a D'Addario PL-026 string of 0,66 mm diameter was used; the measurement
frequency was in the range of 50 – 100 Hz with a deflection amplitude of 0,2 – 0,5 mm. Any
non-linearity of the drive-system was suitable compensated for if necessary.
Shaker-measurements allow for a relatively precise determination of the absolute pickup-

sensitivity, but can only be carried out in the low-frequency range due to structural reso-
nances. While the passive two-terminal networks of an equivalent circuit diagram (R, L, C)
may be identified via measurement of the pickup impedance, the shaker-measurement enables
us to calibrate the active source contained in the ECD. Indeed, pickup excitations via
alternating magnetic fields (Chapter 5.10.2 – 5.10.4) merely allow for determination of the
magneto-electric transmission coefficient; conversely, the shaker-measurements described
here make possible the identification of the mechano-electric transmission coefficient HUv -
though limited to the low frequency range where HUv is independent of frequency. Typically,
we find HUv to be about 0,1 – 0,3 Vs/m for the Stratocaster pickup. The precise value is of
course dependent on the individual pickup and on the distance between string and magnet;
moreover the string diameter needs to be considered.
From the point of view of systems theory we could state: the impedance-measurement allows
for the specification of poles and zeros in the transmission-function; the shaker measurement
adds the basic amplification. Or we could say: with the impedance measurement the shape of
the transmission frequency response can be determined while the shaker-measurement yields
its absolute position.
© M. Zollner 2002
The string velocity is not only applicable as source quantity at low frequencies but also across
a broad frequency range. Proof is found in Fig. 5.10.1 comparing the time-function of the
string velocity (left-hand section) to the voltage generated by a Telecaster-Bridge-pickup
(right-hand section). The pickup was mounted at a distance of 2,5 mm below the string, and
loaded with 110 kΩ and 330 pF. The ray of the laser-vibrometer struck the string on the
(extended) axis of the magnet with the string vibrating in the direction of the axis of the
magnet, and thus also in the direction of the laser beam. The velocity-signal generated by the
laser-vibrometer was filtered with a 2nd-order low-pass in order to model the filtering
happening within the pickup. The sameness of the two plots is impressive proof that the
pickup indeed does detect the string velocity, and allows for an absolute scaling of the
transmission coefficient of 0,29 Vs/m in the low-frequency part of the oscillation.
As a comparison, shaker measurements were available which, however, were taken with 2,00
mm string/magnet-distance and with a 0,66-mm-string. They had yielded a transmission
coefficient of 0,31 Vs/m. Matching the string diameter (0,66 mm → 0,70 mm) increases this
value to 0,31⋅(0,70/0,66)2 Vs/m = 0,35 Vs/m, and matching the string/magnet-distance (2,0
mm → 2,5 mm) decreases it to 0,30 Vs/m (Chapter 5.4.5). Consequently, the absolute
sensitivity determined with little effort via the shaker is a very good match to the value
obtained from the laser-vibrometer-setup.
Fig. 5.10.1: String velocity obtained via the laser-vibrometer (left); corresponding pickup voltage (right).
For string diameter and string/magnet distance see the text. Pure transversal wave.
Measurements using the shaker make possible a relatively effortless determination of the
transmission coefficient. The following hints are helpful to limit measurement errors to an
acceptable level:
- the pickup needs to have sufficient distance to the shaker to avoid direct magnetic coupling;
- the string needs to vibrate with a constant shape and must not develop any “life of its own”;
- the pickup needs to be mounted with non-magnetic materials to avoid any eddy-currents;
- the “magnetic history” of the string influences the result and should be recorded exactly;
- a DFT-window with small level-error (picket-fence-effect) may be dispensable for
transmission measurements because the resulting error compensates itself (it shows in both
channels in the same way), but it is still strongly recommended to make comparisons with
other measurement approaches;
- the pickup should be mounted as rigidly as possible since, with a string-excursion of e.g.
merely 0,5 mm, a vibration of the pickup of as little as 50 µm in amplitude can already cause
ugly measurement errors.
© M. Zollner 2002
5.10.2 Measurements with the Helmholtz-coil
Using a pair of Helmholtz-coils, it is possible with only little effort to generate a parallel
magnetic field the strength and flux density of which can be precisely calculated. We need to
consider, however, that magnetic fields around strings are everything but parallel – it is
therefore easily possible that measurements employing the Helmholtz-coil yield other results
compared to measurements where the pickup is excited by a vibrating string.
For the following measurements two oval Helmholtz-coils were wound; they had a size of 33
cm x 27 cm and 175 turns. The resistance of both (in parallel connection) is 7 Ω, they were
driven by the AF-100 frontend of a Cortex workstation at LU = 18 dBV for 600 Ω source-
resistance. With these values a flux density of 6,5 µTeff resulted in the low frequency range at
the measuring position. Since the (inductive) coil impedance could not be expected to remain
small relative to the source impedance across the whole measurement range (which would
have resulted in perfect current impression), the actual current was monitored (Fig. 5.10.2)
and any deviations were compensated for arithmetically. Still, we are confronted with
differences compared to the results obtained with other measuring methods. A more in-depth
analysis of the coil impedance showed a resonance at 44 kHz, i.e. capacitive currents having
a field-amplifying effect♣. Given this situation, the share of the inductive current was now
determined for a coil-equivalent circuit (Fig. 5.10.2, ----), and only the deviation of this share
was arithmetically compensated for the subsequent pickup measurements.
Fig. 5.10.2: Current flowing into the terminals (600 Ω source impedance, –––), inductive current (----).
The right-hand graph shows the frequency response of the amount of the Helmholtz-coil impedance.
Measurement and calculation are compared in Fig. 5.10.3. The measurements took place
within the parallel field of the Helmholtz-coils; the axis of the Helmholtz coils and that of the
pickup coil coincided. The compensation mentioned above resulted in operating conditions
which were equivalent to the operation with impressed magnetic flux density. The results of
the calculations were obtained using a quadripole equivalent circuit diagram. The component
values of this ECD were derived from the impedance frequency responses, as they were
determined in Chapter 5.9.2. The basic correspondence of the curves shows that the transfer
behavior of the magnetic pickup, from magnetic field to voltage, can approximately be
derived from a simple measurement of the impedance frequency response. The absolute
scaling cannot be determined that way but can be achieved at a single low frequency using the
shaker (Chapter 5.10.1).
♣
For a parallel-resonance-circuit, the current in the terminals is smaller (!) than the current in the res. circuit.
© M. Zollner 2002
Fig. 5.10.3: Pickup transmission factor GUv: measured with Helmholtz-excitation (–––), calculated with
ECD (---). Fender Stratocaster (left), Gibson P-90 (right), each loaded with 700 pF.
The Stratocaster-pickup has little eddy-currents and for it we find a very good
correspondence between measurement and calculation, while significant differences are
observed for the P-90. Other than the capacitive coupling which could be the reason for small
differences in the highest octave, it is in particular the different field geometry that is
responsible for the discrepancies. The brass plate used as mounting base below the P90-coil
influences the coupling factor stronger for the parallel Helmholtz-field incident than for the
focused string-field. As one removes the brass plate from the P-90, the Q-factor increases, and
measurement and calculation correspond (left-hand section of Fig. 5.10.4).
Fig. 5.10.4: Pickup transmission factor GUv: Helmholtz-measurement (–––), ECD-Calculation (---).
Gibson P-90 w/out brass plate, 700-pF-load (left). Gibson 490-R, 330 pF // 200 kΩ load (right).
The Helmholtz-field turns out to be totally unsuitable to measure the velocity/voltage-

transmission coefficient of a humbucker: the out-of-phase connection between its two coils
is designed to render such parallel fields ineffective. Fig. 5.10.4 shows how well or how badly
the design succeeds in that respect. The upper curve was taken in single coil mode, the lower
in humbucking mode with an axis-parallel field (the direction of the magnetic field was in
parallel with the coil axis). At least in the lower-frequency range the compensation works
well. However, as the pickup is turned by 90°, barely 9 db of compensation dampening
remain. This appears somewhat weak, especially since the pickup is manufactured by the
inventor of the humbucker (Gibson advertisement). More details regarding hum suppression
are explained in Chapter 5.7.
© M. Zollner 2002
5.10.3 Measurements with a coaxial coil
For these measurements the pickup is excited by a generator coil the axis of which coincides
with the axis of the pickup (thus the designation coaxial coil). Fig. 5.10.5 shows a cross-
section through the setup: a small coil (e.g. 6 mm ∅) carrying a sinusoidal current is
positioned over the magnet of a singlecoil-pickup (e.g. Fender Stratocaster). The magnetic
field of the small coil is – as a contrast to the Helmholtz-field - focused and thus more similar
to the magnetic-field of the string. For the Stratocaster pickup already the Helmholtz-
measurement was useable – the coaxial excitation works similarly well. We do get effects of
capacitive coupling above 10 kHz but these are negligible. For the P-90 the Helmholtz-
excitation gives clearer divergences re. the ECD-model; the coaxial excitation results in a
better agreement because the magnetic AC-flux is mainly concentrated on the upper side of
the winding und therefore the brass plate located below the pickup has merely a weak effect.
Fig. 5.10.5: Pickup with coaxial generator coil (cross-sectional drawing for a Stratocaster pickup).
For the Telecaster neck pickup, we see clear differences between the measured curve and the
transmission function derived from the two-terminal equivalent circuit diagram. These
differences are due to the eddy-current dampening of the metallic pickup cover. Apparently
one does arrive at a limit regarding modeling for pickups when confronted with such strong
dampening, and a modification of the simple quadripole-equivalent-circuit-diagrams
introduced in Chapter 5.9.3 is required (see Chapter 5.10.5).
© M. Zollner 2002
5.10.4 Measurements with the tripole coil
The coaxial coil introduced in the previous chapter generates an alternating magnetic field
which is a much more locally effective field than the field of the Helmholtz-coils. However,
there are still differences compared to the field distribution of an oscillating string. A further
optimized approximation of the field geometry of the string can be achieved with a tripole-
coil, i.e. a layout …..
The text remains reserved for the print version of this book.
Fig. 5.10.6: Tripole coil excited by alternating

current, measurement (–––), ECD-model (---).
While the correspondence between measurement and model-calculation is again good for the
Stratocaster- and P90-pickups, significant divergences appear for the Telecaster-neck-pickup.
The cause is found in the metal shielding cover. Although it is made of non-magnetic
material, this cover introduces a dampening due to the eddy currents induced into it. The
effect is mainly felt in the treble range. The equivalent circuit diagram (ECD) derived from
the impedance obviously requires modifications in order to account for eddy-currents close to
the strings, and to better model the frequency dependence of the mechano-electric coupling.
At this point we can also look into the question whether the tripole-excitation always yields
results which are equivalent to normal operation (string oscillation). It may be as well that the
ECD-model given above is closer to reality. Measurements with the laser-vibrometer give
clarifications regarding these issues (Chapter 5.10.5).
© M. Zollner 2002
5.10.5 Measurements with the laser-vibrometer
The measurement methods presented in the previous three chapters (5.10.2 – 5.10.4) deliver
nicely agreeing results for pickups that feature a small level of eddy-currents. However, as
soon as pickups with eddy-current-dampening (e.g. P-90) are the subject of the
measurements, we see differences at higher frequencies. We could ignore these differences
because the transfer behavior at 10 – 20 kHz is not really that important due to the lowpass
filtering. On the other hand, we could consider the divergences as an indicator that an
extension of the overhead towards the limits of our models might be in order and that
decreasing the differences would be worth the effort. Still, none of the measurement
approaches proves that the velocity/voltage-transfer-function indeed shows the identified
frequency response. In the end, all three methods yield merely the magnetic-field-to-voltage-
transfer-function, or – to put it even more radically – the current/voltage-transfer-function.
Considerably more insight is offered by measurements with the laser-vibrometer which do
require a higher effort regarding instrumentation but directly capture the desired source
quantity (the string velocity).
Laser-vibrometers take advantage of the Doppler-Effect: the frequency of a reflected wave

changes if source and reflector move relative to each other. If source and reflector get closer,
the reflected beam of light has a higher frequency than the ray emitted by the source (laser). If
source and reflector move away from each other, the frequency is lower. Given v = speed
difference between source and reflector, the relative frequency change corresponds
approximately to Δf / f0 ≈ v/clight. As one points the laser-beam at the oscillating string, the
voltage generated by the laser-vibrometer corresponds to the string velocity in the direction of
the beam. String movements along the string (i.e. perpendicular to the laser beam) are not
detected. This means that strain-waves (if we discount a minimal transversal contraction) are
not detected by the laser-vibrometer – but they are by the pickup. When performing laser-
based control measurements on a pickup, we need to ascertain as a consequence that either
exclusively transversal waves are generated, or that the two wave-types clearly happen
separately. For the following experiments using a string of a length of 28 m could ensure a
sufficient mode decoupling. This string is deflected on one end by a short transversal impulse.
Below the string the pickup is positioned at the regular distance (2 – 5 mm), and above the
string we have the laser-vibrometer. Fig. 5.10.1 could already drive home the point that a
magnetic pickup indeed samples the transversal string velocity at the given point – this is the
same with a laser-point. Transforming the voltage given by the pickup as well as that given by
the laser-vibrometer into the frequency domain puts us in the position to determine the
transmission function of the pickup. In principle, that is ……
Unfortunately, we may not conclude from the fact that the laser-vibrometer can carry out
highly precise measurements that the pickup measurement automatically is also free of
measurement artifacts. It is necessary that exclusively a transversal wave of plane polarization
propagates in the string – but this is not easily accomplished. Structural resonances in the
string bearing continuously lead to undesired strain-waves which falsify the measuring result.
After carrying out extensive pre-experiments, an experimental setup could be developed
which includes strain-waves artifacts only at very high frequencies. Since the pickup operates
as a low pass, the remaining interferences are tolerable or insignificant. A similar situation
exists for the inevitable offsets in the circuits, which cause voltage drifts at low frequencies.
The offset compensation chosen for the experiments was sufficiently potent, and the
remaining error was insignificant (Cortex-Workstation CF-90, CF-100). The TUv-values given
in the figures belong to individual measurements which were not always carried out with a
string-to-magnet distance of 2 mm.
© M. Zollner 2002
Fig. 5.10.7 compares the results of the laser-measurements with the ECD-model-calculations.
During the measurement the pickup was loaded with an RC-circuit; this was considered
correspondingly in the calculations. While the resonance-emphasis is, generally speaking,
correctly reflected, all measurements give relative to the calculations a characteristical treble-
loss amounting to about 1 dB at 10 kHz.
Fig. 5.10.7: Transversal-wave transmission-factor: laser-vibrometer (–––), ECD-model-calculation (----).
In Fig. 5.10.8 corresponding measurements and calculations are depicted for a coaxial
humbucker (Fender Noiseless Stratocaster). Despite the different construction and the so-
called “beveled magnets” (Chapter 5.4.6), the level differences in the treble range turn out to
be analog those in Fig. 5.10.7.
Fig. 5.10.8: Transversal-wave transmission-factor: laser-vibrometer (–––), ECD-model-calculation (----). This

figure remains reserved for the print version of this book.
© M. Zollner 2002
Further measurement results are shown in Fig. 5.10.9; they feature similar divergences
between measurement and model-calculation in spite of different pickup build. The Duncan
APTL-1 and the Fender Telecaster bridge pickup are seated on a ferromagnetic assembly
plate; their cheapo-counterpart uses a ferrite bar-magnet instead of 6 alnico magnets; the
Jazzmaster pickup sports a relatively large winding-surface and short magnets; for the P-90
two bar-magnets are located beneath the coil – the differences between measurement and
calculation still amount merely about 1dB in the relevant frequency range (only slightly more
for the P-90). Consequently, it is possible to derive the transmission-behavior of all pickups of
this simple singlecoil-type from the impedance frequency response. It may be – if necessary –
supplemented by a slight treble attenuation the cause of which can be found for the most part
in the aperture window (Chapter 5.4.4).
Fig. 5.10.9: Transversal-wave transmission-factor: laser-vibrometer (–––), ECD-model-calculation (----).
© M. Zollner 2002
But not all singlecoil-pickups show the aperture-induced differences between measurement
and model-calculation as depicted on Fig. 5.10.9. The Gretsch HiLoTron has a striking comb-
filter-like interference curve, and the Telecaster-neck-pickup gives serious discrepancies at
high frequencies – in these cases it is not possible to draw conclusions regarding the
transmission behavior from the impedance-equivalent-circuit-diagram. The reasons are found
in the magnetic field – but they are highly individual.
Let us first take a look at the HiLoTron-pickup which first entered service in the late 1950s in
Gretsch guitars (e.g. Tennessean). Tom Wheeler writes in his book “American Guitars” that
the pickup was developed by “fulltime Gretsch personnel”. However, with No. 2683388 there
was a patent already in 1954 which shows the exact same construction. Inventor is Ralph
Keller who is designated as “assignor to Valco Manufacturing Co.”. Valco (the successor to
National and Dobro) manufactured guitars for other companies in the 50s, including Gretsch.
Maybe somebody among the “fulltime Gretsch personnel” took off the vinyl-cover and
checked out the pickup? Or – as it does happen now and again – the time was ripe and two
inventors had the same idea at the same time without knowing from each other (they were
both from Chicago, though ….).
Anyway, it’s all Ralph Keller’s glory, who on the other hand also has to take the rap for the
justifications he gives in his patent: The most important advantage stems (…) from the
generally parallel relation of the magnetic lines of force with the instrument strings as com-
pared with the perpendicular relation between the magnetic field and the strings which is
common in many currently used pickup devices. (…) As a result, a wide area of the magnetic
pattern is efficiently activated by the moving strings whereby to produce substantially greater
and more effective variations in the reluctance of the magnetic field. (…) Consequently, (…)
the pickup produces substantial improvements in the tone color of the instrument due to the
capturing of additional overtones or harmonics which are not ordinarily reproduced when the
pickup point is limited to a single point or relatively restricted area on the strings. In short:
according to Ralph K. the aperture-window should be as long as possible in order to capture
as many harmonics as possible. This assumption (which was also taken up by Leo Fender♣ at
the beginning of the 1960s when he designed the Jazzmaster pickup) is however not in
agreement with systems theory: the longer the (actually effective) impulse response, the more
the system has a narrow-band character. This is a fundamental aspect of the reciprocity of
time and frequency as elaborated e.g. by Marko or Küpfmüller. In the case of the HiLoTron-
pickup, the measurement of the transmission function shows – compared to the curve derived
from the impedance ECD – a string specific interference gap at 5 kHz (Fig. 5.10.10).
Fig. 5.10.10: Gretsch HiLoTron. Laser-

vibrometer (–––), ECD-model (----).
♣
Compare to Chapter 5.1
© M. Zollner 2002
The horizontal position of the bar-magnet indeed lets the string be sampled “across a wide
range”, or more precisely at two relatively distant points – this leads to comb-filter-like
superpositions (Fig. 5.10.11). As the string vibrates perpendicular to the fretboard, two air-
gaps change: one (as usual) over the pole-screw, and a second one over the south pole of the
bar magnet. The air gap bordering the pole-screw is the smaller one and therefore the
transversal wave occurring here causes a larger relative distance change. In other words: at
this position the pickup is more sensitive. Increasing the string-to-pickup-distance makes the
pickup become less sensitive, as is to be expected; the interference effect becomes stronger,
however (due to the air gaps becoming more similar).
Fig. 5.10.11: Block diargam (above). Difference

between laser measurement and ECD-model (–––),
frequency response of interference filter (----, right).
Fig. 5.10.11 models the delay-time between the two air-gaps with τ(ω) (due to the dispersive
wave-propagation τ is frequency-dependent). Optimization of the parameters resulted in an
effective distance of the two sampling points of 23 mm which is in good agreement with the
dimensions. A real factor takes care of the smaller sensitivity of the second “channel”; it
amounts to 0,23 in the example. Although the magnetic polarity at the two sampling points is
opposite, the two channels need to be added (constructive interference): bringing the string
closer to the pickup decreases the magnetic air-gap resistance in both cases and thus increases
the magnetic flux. Of course, in reality the sampling does not happen at two ideally small
points but in two areas with finite dimensions each. Model and measurement will therefore
not match exactly. That for the chosen example the differences are nevertheless as small as a
few tenths of a dB (Fig, 5.10.11) is a nice confirmation of the model. Fig. 5.10.12 shows the
measurement results compared to the complete model, and also the dependency on the string.
Fig. 5.10.12: Comparison measurement/model (left), string-specific transmission function HUv (right).
© M. Zollner 2002
The “two-location-sampling” gives the HiLoTron-pickup its very own transmission

characteristic which is unique in this form. Compared to a humbucker, the interference gap is
much less pronounced, and the low coil inductance results in a brilliant, treble-rich sound. We
find an entirely different situation for the Telecaster pickup. While here, as well, the
measurement results differ from the model-calculations based on the impedance equivalent
circuit diagram, they do so in a different way and due to different causes (Fig. 5.10.13).
Fig. 5.10.13: Transversal-wave transmission factor: laser-vibrometer (–––), ECD-model calculation (----).
Right-hand column: difference between measurement and ECD-model calculation.
1st line: original Fender pickup, Telecaster retrofit set. 2nd line: Duncan APTR-1 ("for Tele®").
3rd line: “cheapo” copy, guitar with Telecaster-like body and similarly looking pickup
© M. Zollner 2002
Measuring the three Telecaster neck pickups first resulted in three different impedance
frequency responses. From these the different transmission frequency responses can be
calculated (as shown in Fig. 5.10.13); however, the measurement results diverge significantly.
The reason for the discrepancies are eddy currents induced by the alternating magnetic field
into the metal cover (Chapter 5.9.2.2 and 5.9.2.5). While the measurement of the pickup
impedance does capture eddy-current dampening, it only succeeds so with regard to the
impedance – and not (or only partially) in terms of the effect on the transmission.
An experiment exemplifies this: the influence of the cover on the impedance frequency
response is shown for a Duncan pickup (APTR-1, "for Tele®"), and we can see effects merely
in the range of the resonance. Operating the pickup without cover we can calculate HUv in the
usual way; any differences to the calculation can again be explained by the aperture
dampening. With cover, two measurement conditions can be distinguished: normal (string
above the cover) and upside-down (pickup turned over). Of course the impedance frequency
response will be identical in both cases; the string has no measurable effect. Not so for the
transmission frequency response where differences appear. Upside-down, when no cover
metal comes between string and coil, we do get an entirely different measurement curve
compared to “with cover”, but the differences between calculation and measurement are
similar for both cases. In normal configuration (Fig. 5.10.13) the difference between
measurement and model calculation amounts to 10 dB already at 10 kHz. Apparently the
positioning in space offers another degree of freedom which the model calculation does not
cover. An extended model with three coupled, lossy coils would have to be supplemented (the
cover acts as a shorting ring), but the usefulness is not in any reasonable relationship to the
required effort.
Normal position: upside-down position:
Fig. 5.10.14: Duncan APTR-1, with and without shileding cover (operation of normal position; see Fig. 5.10.13.
© M. Zollner 2002
All three pickups mentioned in Fig. 5.10.13 find use in similarly looking guitars but their
transmission characteristics differ tremendously: the treble reproduction diverges by 12 dB!
Where do these differences come from? Without cover, the vintage Tele and the APTR-1
show a similar behavior (Fig. 5.10.15) – as one would expect it due to the similar build. The
cheap copy fitted with a bar-magnet does not entirely reach the resonance emphasis presented
by the competitors (due to the eddy-currents in the iron slugs), and arrives at a somewhat
worse treble reproduction. However, only as the covers are mounted, the large differences
arrive: the similarity is only in the looks but the electrical characteristics differ significantly.
The cover of the Fender pickup is made of 0,5-mm-thick German silver; the other two are
made of chrome-plated brass.
Fig. 5.10.15: Impedance- und transmission-frequenca-responses: pickups with/without cover.
© M. Zollner 2002
The different conductivity of these metals (Chapter 5.9.2.2) gives varying eddy-current-
dampening. The cover of the cheap imitation has a thickness of 0,8 mm and is even more
efficient than the one of the APTR-1 (which has 0,5 mm). Of course, it is now a matter of
individual evaluation whether one prefers shining treble or boxy mids. However: for the
original Fender pickup it was possible to attenuate the treble if so desired. That does not work
the other way ‘round. In the Seymour-Duncan brochure the phrase "For tone that sets you
apart" is found above the picture of the APTR-1. Apart … to where, now? Be careful what
you wish for …. 
Humbucker
Humbuckers sample the string vibration at two positions using their two pole pieces per
string. Due to the delay between these two points, phase shifts occur and interference
cancellations happen if the delay matches half a vibration-period. Since the phase-velocity of
the propagating transversal wave is different for each string, the humbucker interferences are
string-specific. Fig. 5.10.16 compares laser measurements and model calculations. The
curves shown in the first line of the figure are practically congruent which is impressive proof
for the high quality of the model. Three components were considered in the calculation: the
aperture filter (Ap), the interference filter for a pole distance of 19 mm (Notch), and the low
pass transmission (RLC) derived from the impedance frequency response. Recalculated for a
scale of 63 cm, the fundamental frequency amounts to fG = 130 Hz, the string diameter is
0,7mm.
Fig. 5.10.16: Gibson-Humbucker ('57 classic). Laser-measurement (upper left), model-calc. (upper right.).
Components of the model-calculation (lower left). The treble attenuation of the metal cover is shown lower right.
© M. Zollner 2002
The measurements and calculations depicted in Fig. 5.10.16 were done for the string used on
the test-bench; Fig. 5.10.17 gives the transmission frequency responses for real guitar-
conditions. The left-hand part of the figure relates to a long-ish guitar cable (700 pF), the right
hand part holds for a load capacity of 330 pF (Chapter 9.4 and 9.6). The interference gap for
the low E-string (E2) is located at 3 kHz i.e. just at the range which would be particularly
emphasized by a Fender singlecoil. In conjunction with the treble attenuation caused by the
LC-low pass, a transmission frequency response results for the Gibson Humbucker (and its
innumerable copies) which evokes a Tschebyscheff-low-pass: that’s how treble is efficiently
cut. Another 5 dB are lost in the treble range if (as it was the case for many Gibson guitars in
the 70’s and 80’s) potentiometers of 100 kΩ (“Tone”) and 300 kΩ (“Volume”) are utilized
rather than 500-kΩ-versions. Tone is in the eye of the beholder ….
Fig. 5.10.17: Gibson-Humbucker ('57 classic). Stringspecific transmission function, with varying C. Due to the
dispersive wave propagation the second interference gap is not at trice the frequency but considerably higher:
high-frequency components run disproportionally faster (Chapter. 1.3.1).
Not all humbuckers feature the pole-distance of 19 mm as given by the Seth-Lover-developed

Gibson Humbucker. For the Fender Humbucker (incidentally also developed by Seth Lover)
we find 20 mm and for the Gibson Mini-Humbucker 13 mm, while for humbuckers in a
single-coil format the distance is as small as 6 – 9 mm. In the same way that the pole-distance
decreases, the notch-frequency increases. If the magnet poles are reduced to narrow blades
with a separation of the (middle of the) blades of 7,5 mm – as it is the case e.g. for the Joe-
Barden-pickup or the DiMarzio DP-184 – the notch frequency is pushed to higher ranges.
Fig. 5.10.18: Transmission frequency repsonse: humbucker of 7,5 mm blade-distance (Joe Barden, DiMarzio).
Laser-measurement (–––), model-calculation (RLC, notch, aperturefilter ----). String diameter 0,7 mm, fG = 130
Hz.
© M. Zollner 2002
Without dispersion a 0,4-fold pole-distance-decrease would increase the notch-frequency by a

factor of 2,5, in reality, however, it increases (string-specifically) by a factor of 4 to 5. The
corresponding theory (dispersive wave propagation, Chapter 1.3.1) is a good match to the
measurement results. In addition there are effects that are more difficult to model such as non-
negligible inductive and capacitive coupling between the two coils – in particular relevant for
humbuckers with small pole- or blade-distance. Under some conditions the transmission
characteristic can be dependent on the propagation direction of the transversal wave (Chapter
5.11). Simple transfer-models give the resulting complicated frequency responses merely with
modest accuracy (Fig. 5.10.18). The maximum showing in the 18-kHz-range for the Joe-
Barden-pickup (which could be interpreted as a dipole-resonance) is indeed due to the
coupling-resonance of the two coils. For the sound this side-maximum is insignificant.
5.10.6 Measuring accuracy (or rather measurement inaccuracy)
Test-stand-measurements support objective measurement data but they are carried out in an
un-typical situation (“in vitro”). There are differences to the behavior of the actually played
guitar (“in vivo”), and in addition we need to consider that all measurements contain errors.
Having doubts about the value of measurement results is therefore permitted. On the other
hand, subjective evaluations (e.g. given while playing a guitar) need to be questioned, as well:
they may have been expressed by a hard-of-hearing guitarist, or by somebody playing under-
the-influence, or may have been written up by a guitar-tester in dire need of money. Even a
combination of all three conditions is conceivable. An evaluation may also simply have been
given in a special (possibly non-reproducible) mood; thus it expresses a subjective opinion of
little relevance to the public. Chapter 8 addresses the world of psychoacoustics while the
following paragraphs deal with the measurement errors occurring with bench-tests.
1) Most of the pickups investigated show a rather poor production quality: each single magnet
generates a different flux density, the air-gaps are different, the mounting plate is distorted, or
the magnets are lopsided. Even simple distance measurements become problematic – plus
regular measuring tools made of steel cannot be used due to the magnetic attraction forces.
For dynamic measurements non-ferrous metals need to be excluded as well, since eddy-
currents generated within them lead to undesirable dampening. On the other hand, using a test
bench put together exclusively from regular plastic easily leads to measurement tolerances of
0,1 mm – for critical distance measurements this may often be already too much. Even much
more difficult are diameter measurements of strings: to determine the cross-sectional area of a
10-mil-string with an accuracy of 1%, we need to measure the diameter with an accuracy of
1,3 µm. Normal micrometer screws reach 10 µm tolerance which leads already to an error of
8% for the area.
2) For electrical measurements, the situation is somewhat more positive: voltage- and current-
values can be taken with an error of about 1%; in individual cases even more precisely.
Measuring magnetic quantities brings again a decrease in accuracy: errors of 5% are probably
the typical range.
3) An even greater problem resides with the string-magnetization: magnetic pickups work
only with ferromagnetic strings, and their operating point moves along a hysteresis loop. The
transmission factor of a pickup can change by as much as 3 dB (!) if the string is brought
close to the pickup magnet and then moved away again to its rest position … plus there are
many paths within the three-dimensional space to get to a specific location! An example shall
exemplify the associated difficulties (Fig. 5.10.19): in the left-hand section of the figure the
© M. Zollner 2002
pickup (a Gibson Humbucker) is moved along the strings across the cranks of the rotating
string (motor-test-bench). The paths “to and from” yield two different curves (––– d = 2 mm,
------ d = 4 mm). The right-hand section shows the voltage-level curve as a function of the
distance d; the crank was rotated across the slug-coil of the humbucker (left maximum in the
left section). It required many complementing measurements to arrive at meaningful and
practically relevant measurement curves – and to arrive at the “subjectively correct
objectivity”.
Fig. 5.10.19: Dependency of the voltage level on the pickup shift along the string (left) and on the string-to-
magnetic-pole distance (right). Motor-testbench (Chapter 5.4.4).
4) In order to see whether the data collected from an individual pickup are indeed
representative for this type in general, it would in fact be necessary to pick a sample
including more than just one single element. However, the price per unit of € 186 (in 2003) is
prohibitive for such an approach, and – as “extrapolation” – the speculation remains that all
pickups of this type received the same more or less crooked and messy assembly. (Here, the
business administrator nods, and with a stern expression points to the fact that a single alnico
magnet is already as expensive as 40 cents: “one needs to make a little profit, after all”.)
5) We may be critical regarding the motor-test-bench, and note that

- the string-crank we used does not represent an infinitesimal short impulse, that
- a rotational movement is happening, and that
- transformation in the frequency domain requires a linear system.
Moreover, the string does not maintain its cylindrical shape at the crank but gets minimally
bent (steel wire cannot be cranked in another way). For 1 mm crank amplitude and a desired
measuring dynamic of 40 dB, the required production tolerance is as small as10 µm. …. Or
maybe 60 dB were desired – in that case you need to bump it all up to 1 µm tolerance. … Oh,
right: and please do install the whole shebang on plastic bearings free of friction and
mechanical play …..
6) The shaker-test-bench, as well, includes typical artifacts: the string does not vibrate in one
plane but along a slightly elliptical path; the magnetic drive of the shaker generates a
magnetic crosstalk; measurements are limited to the low-frequency range due to self-
resonances; the drive is non-linear and time-variant due to it heating up. Plus much more.
© M. Zollner 2002
7) The laser-vibrometer operates with sufficient accuracy if mounted on a heavy stone-table

(this was the case for our experiments). In order to keep the noise low, a suitable narrow-
band-filtering is required, as discussed e.g. in chapter 6.
8) In order to be able to exactly specify the effective coil surface, a wire as thin as possible
should be used for all measurement coils. However, the thinner the wire, the greater the risk
of damage (as a compromise, 60 – 80 µm magnet wire could be used).
9) Measurements which include forming the integral of the measurement signal (e.g. to derive
the velocity from the acceleration) are falsified by amplifier offsets. Even for low-offset op-
amps sometimes just moving the air above the housing of the op-amp is sufficient to produce
measurable drift-effects. With a corresponding effort, this problem proves to be just about
manageable.
In the end, we obtain some relief from comparing the individual measurement results: it ain’t
all that inaccurate. As long as one does not approach the problems with excessive
expectations, and as long as one avoids real blunders (which do wait to happen, though), the
test–bench–measurements yield reliable results. The comparison of measuring results taken
over many years corroborates the (subjective) assumption that the typical accuracy of a test-
bench is comparable to that of a precision-SPL-meter, and amounts to about 1 dB.
© M. Zollner 2002
5.10.7 FEM-calculations
Besides unwavering faith (“ONLY pre CBS”), listening experiments (“vintage vs. new”) and
measurement (“3D-laser”) – and maybe sheer ignorance – the only other way to describe the
function of a pickup “exactly” seems to be mathematical/physical finitization. The magnetic
field is cut into hundreds of thousands of fragments (the finite elements) for which a computer
calculates (for hours) the exact field-distribution. The more expensive the software and the
larger the number of elements, the better this works. And so the ambitious hobby-scientist
enthusiastically posts his colorful fluxograms on the Internet – to deliver the definitive proof
why the Strat pickup sounds differently than the XY-90. Please heed two words of wisdom
from a professional scientist who threw in the towel after half a year: forget it.
It isn’t that these finite models are generally bad – the problems lie in the data entered by the
user. But first things first. Field calculations are simple if direct current flows through a
straight wire of infinite length. That ain’t the case for a gittar pickup? Okay then. Let’s open
the toolbox for permanent magnets and click on “cylindrical magnet”. Go to “Mesher”, on to
“Solver”, do a color plot …. and off into the bin it goes. Any questions? Sure – lots!
The field of one cylindrical magnet is a relatively simple one because it features rotational
symmetry. One might be able to ignore that a pickup includes 6 of them (let’s
“approximate”), and as well that there is a string; oh … 6 strings, even. Didn’t someone say
checking out a string-less guitar: “for a beginner that will suffice.” But seriously, the
magnetic field without string is of course only of any use as a starting point. Statements
regarding the transmission behavior work only with inclusion of the string. That, however,
makes the rotational symmetry go out the window; the calculation effort rises dramatically.
We could see that as a challenge: those with lots of experience with the cross-linking should
be able to master the geometry, set the boundary elements correctly (a job not at all trivial),
and have the field calculated with string. Now, this field needs to travel through air, with the
relative permeability µr = 1. And through a steel string with a high permeability (due of the
ferromagnetism). And through a permanent magnet with a rather small µr. Limiting the effort,
a search leads µr = 4 for alnico and to µr = 40 for steel. Again: off into the bin it all goes.
Your regular FEM software will model ferromagnetic materials with a BH-characteristic. One
BH-characteristic, that is. That’s because that way you don’t have to distinguish which branch
of the hysteresis-curve holds the operating point. So, the simulation gives us an approximate
impression of the field shape – but what about being accurate? Will the big effort bring
ultimate perfection? Well, both the string and the magnet are magnetically hard, and therefore
the two hysteresis-branches differ considerably: off then to get more computer power (if one
gets that kind of support to begin with ….) and to calculating the two branches. Two? I.e.
merely the boundary curves which in fact each hold an infinite number of BH-pairings? Now
the “Solver” can’t manage it anymore, the iteration fails to converge, the software capitulates.
Ah – but the newest release takes care of this issue as well! Super! So now we have two
different non-linear performance maps of the materials, and can only hope that magnet and
string abide by these. Does the newest release include a button for anisotropy, i.e. the fact
that the magnetic characteristics of metals are dependent on direction? In the simple model,
isotropy is used as a basis, but the string was drawn during the manufacturing process and
therefore subjected to significant mechanical stress in one direction – so at least we should
check whether it really acts isotropically. Alnico-V-magnets are certainly anisotropic, ceramic
magnets often as well. Plus, unfortunately only a small part of the magnetic flux flows
through the cylinder in the axial direction while a significant part penetrates the cylinder
© M. Zollner 2002
mantle at an angle. In conclusion: no final perfection, and in spite of mathematical overkill we

merely have a rough approximation.
As the proud owners of colorful fluxograms with increasingly fine resolution and barely
visible finitization we now believe we have the license for carrying out the final step: the
dynamic analysis. Now the sting is to vibrate i.e. it changes its distance to the magnet. We
suggest our desire to the FEM-software in the form of discretization: instead of one
string/magnet distance we do an overnight calculation-run for 10 of them (not more, let’s start
small and not exaggerate). This yields 10 different magnetic fluxes, and as difference between
them we obtain the quantity on which the induced voltage depends: the change of the
magnetic flux over time.
The highly optimistic assumption at the basis of this result is that material data we use are
close to reality – although rarely anybody will have tensorial data of anisotropic
ferromagnetics at their disposal. Let us imply that we would indeed have access to such
numbers – how do we continue? Eddy currents form in the magnet, and they displace the
magnetic field which we have just calculated with much effort! So: either we limit ourselves
to 82,4 Hz (that’s at least something, isn’t it?), or we justify even (much!) more calculation
effort and do a truly dynamic run. But then we realize at some point that the difference of two
approximately correct numbers will have a mightily big error margin. So again and finally:
off it all goes into the bin.
For all those still not convinced (because 5 months of work have been already invested, and
because the nice software support people were so kind to eradicate all programming errors):
as the magnetic field changes (which it does due to the vibration of the string), the BH-point
does not run along the hysteresis-family of curves. The gradient of the hysteresis curve is the
differential permeability but what we need is the reversible permeability which is smaller
that the differential one (Chapter 4.10.3). In the end we have therefore now a hodge-podge
collection of approximated dependencies the difference of which is not the mathematical
“ultima ratio” but still remains a coarse estimate. Not bad if you have nothing else to do …
but don’t show it, will you?!
Fig. 5.10.20: FEM-graphs
© M. Zollner 2002
5.11 Pickup-directionality
Magnetic pickups predominantly detect string vibrations perpendicularly oriented to the

surface of the fretboard. The string swings back and forth between areas of higher and areas
of lower magnetic field strength causing flux changes in the pickup coil. A motion in parallel
to the fretboard makes the sting move merely in areas of approximately equal field strength
such that only a small voltage is induced. Other than the polarization of the string vibration it
is also the propagation direction of the wave, which we need to consider (in particular for
humbuckers).
5.11.1 Polarization-plane of the string
The polar diagram shown in Fig. 5.11.1 tells us about the dependency of the pickup voltage
on the angle of the oscillation plane of the vibrating string. To do the measurement, a
D'Addario string (PL-026, diameter 0,66 mm) was sinusoidally deflected. The amplitude was
0,4 mm and the distance between string and magnetic pole was 2 mm. The string was
centered above the magnetic pole (on the magnetic axis) of a Telecaster bridge pickup.
Fig. 5.11.1:
Polar diagram of a
magnetic pickup (Fender
Telecaster, bridge). The
line represents the model
calculation, the dots show
the measured results.
1st harmonic (left),
2nd harmonic (right).
Above the centre of the magnetic pole the 1st harmonic shows a cosine-shaped dependency
with a maximum sensitivity for fretboard-normal oscillation and complete cancellation for
fretboard-parallel oscillation. The angle-dependency of the 2nd harmonic has a zero at 63°
and a secondary maximum at 90°. Let us consider for an axially symmetrical magnetic field,
a sinusoidal, centered oscillation oriented normally to the axis of symmetry yields. It will
yield a field shape which can include – due to the symmetry (even function) – exclusively
even powers of the series expansion. In other words it holds exclusively even-numbered
harmonics save for a DC component that remains unimportant for the present point of view.
However, this changes as soon as the string does not follow a centered path of movement
anymore – it may be shifted by string bending or, it may have been positioned eccentrically
already during production. Fig. 5.11.2 indicates string positions of an American Standard
Stratocaster built in 2002. The distance of the magnetic axis is 10,4 mm uniformly for all
three pickups. Since the strings are nut running in parallel but diverge from nut to bridge, it is
not possible that all 6 strings are centered above the magnets, and therefore the transmission
coefficient is not only dependent on the oscillation direction of the string but also on the string
position. The latter may be shifted in two dimensions.
© M. Zollner 2002
5.11 Pickup directionality 5-151
Fig. 5.11.2: String positions over the magnetic poles: neck, middle and bridge pickups (left to right). The strings
do not run straight across the bridge pickup but at an angle.
Polar diagrams such as the one given in Fig. 5.11.1 are therefore valid only for their
individual case – a generalization is possible to a limited extent only. To detect the
dependency of the transmission coefficient on the string position, a number of pickups were
measured on the text bench (Fig. 5.11.3). The string was subjected to a sinusoidal deflection
at a frequency of 85 Hz and shifted while keeping a constant distance over the magnetic
poles. Despite the fact that the pickups are of different build (Chapter 5.1 – 5.3), the resulting
curves are similar. The pickups differ in their absolute sensitivity (loudness), and moreover
the transmission coefficient is dependent on the string position. It is not surprising that the
voltage level is largest right on top of the magnetic pole and smallest in between two poles –
for a Telecaster the level difference for these two conditions amounts to 5 dB, and it is
somewhat smaller for the other pickups. In the figures only one half of the respective pickup
is shown since the curves are symmetric relative to the middle of the pickup. The Stratocaster
pickup is the one exception since its magnetic poles protrude – in the original condition –
differently from the pickup housing (staggered magnets). They were, however, adjusted for
equal height during the measurements.
Fig. 5.11.3: Location dependency of the level of the 1st and 2nd harmonic; string motion normal to the fretboard
plane. Amplitude 0,4 mm, string-to-magnet distance 2 mm, intrinsic distortion of the shaker compensated for.
Fig. 5.11.4 depicts corresponding results for string vibration polarized in parallel to the
fretboard plane. There are general similarities but also significant differences to the fretboard-
normal motion. However, we must not be tempted to explain inter-individual sound
differences from these diagrams. In listening experiments we found neither a sideways shift
in the string position (keeping the string-to-magnet distance constant) nor a change in the
string vibration polarization to have any significant effects. That does not exclude certain
smaller effects becoming apparent in the sound for individual cases but it does qualify the
significance (or rather insignificance) of such effects. For example, the sound changes much
more, as the string is plucked at a different position of with a different pick. A big difference
is heard between picking a string in a fretboard-parallel motion and “plucking” it
perpendicularly such that it hits the fretboard; this difference, however, stems from the strings
hitting the frets i.e. from the mechanical vibration (and not the directionality of the pickup).
© M. Zollner 2002
Fig. 5.11.4: Location dependency of the level of the 1st and 2nd harmonic; string motion parallel to the fretboard
plane. Amplitude 0,4 mm, string-to-magnet distance 2 mm, intrinsic distortion of the shaker compensated for.
Fig. 5.11.5 shows the results for the measurements of all 4 pickups on top of each other. The
curves were vertically shifted such that the divergences are minimal. For fretboard-normal
oscillation there are only minute differences for the 1st harmonic; larger differences exist for
the minima of the 2nd harmonic but this is insignificant due to the much lower level compared
to the 1st harmonic. Fretboard-parallel string vibration causes more pronounced differences
since zeros in the transfer function (i.e. cancellations) are passed through and thus small
imbalances in the magnetic field can have effects on the voltage level. The difference in the
height of the maxima in the right-hand figure is due to the finite number of magnets, which
causes a magnetic field diverging toward the outer range. If a large number of magnets were
lined up we would see maxima of equal height (save for the outer magnets).
Fig. 5.11.5: Location dependency of

the level of the 1st and 2nd harmonic;
string motion normal to the
fretboard plane (left), parallel to the
fretboard plane (right).
In the range of the E-string all curves have a similar shape, despite the fact that:
• the Telecaster magnet is cylindrical with a plane face,
• the Stratocaster magnet is beveled with a wraparound,
• roundhead screws focus the field of a bar magnet for the P90,
• for each string a slug and a screw focus the field for the humbucker.
Towards the middle of the pickup we do see (for the string vibration in parallel to the
fretboard plane) difference of in excess of 10 dB but their relevance immediately needs to be
put in question again: 1) the real string does not vibrate exclusively in the fretboard-parallel
plane; 2) if indeed this difference were of any significance, the A- and the D-string would
sound differently than the E-string which could not be confirmed at all in listening tests. In
fact, all three strings generated e.g. for the Stratocaster the expected sound without any string-
specific special feature (of course, the pitches did differ).
© M. Zollner 2002
5.11 Pickup directionality 5-153
5.11.2 Direction of the wave-propagation
As discussed in the previous chapter, magnetic pickups have directionality with regard to the
orientation of the string-oscillation. However, under certain conditions the transmission
characteristic of the pickup also depends on the propagation direction of the wave travelling
along the string. This effect does not manifest itself in axially symmetrical pickups, but as
soon as the design is asymmetrical, the wave running in one direction can generate a different
induction voltage compared to the wave running in the other direction.
Fig. 5.11.6 explains the context with a simplified block-diagram. A string is sampled at two
locations. A transversal wave propagates along the string; its direction is defined as “forward”
with the index V (from the German “vorwärts”) and “reverse” with the index R. Between the
two sampling points we find a phase-delay τ that may show any kind of dependency on
frequency (dispersion). A humbucker readily serves as practical example – with it the distance
between the sampling points typically amounts to 19 mm.
Fig. 5.11.6: Block-diagram for a string oscillation sampled at two locations.
The two sampling signals (induction voltages) each run through a filter, the transfer-function
of which is defined by H1 and H2, respectively; subsequently the two signals are added. In a
first step the overall transfer function can be set to:
however this would lead to different points of reference: for the forward traveling wave this
would be the input of H1 and for the reverse wave it would be H2. It is conducive to chose the
pickup mid-point as reference for the time, and thus to reformulate the transfer functions
accordingly:
; .
For the filters (pickup RLC low pass) several special cases need to be distinguished: H2 = H1,
H2 = k⋅H1, and H2 ≠ H1. The case of identical filtering (H2 = H1 = H) correspond to axially
symmetrical structure, and the two transfer functions are identical: HV = HR. The Gretsch
humbucker “Filter-Tron” is an example for a real case of this kind:
; H1 = H2 = H
The second special case comes with the two transfer functions differing by a real factor k: H2
= k⋅H1. The two expression in brackets are the conjugate complex of each other, and the two
transfer functions are equal in their magnitude; |HV| = |HR|. The phase functions differ,
however, which needs to be considered when superimposing waves:
; ; H2 = k⋅H1
© M. Zollner 2002
The third case can be described by a complex factor (H2 = k⋅H1) and includes different overall
transfer functions for each of two waves travelling in two directions. This scenario is the one
of all humbuckers with different coils, and for humbuckers in singlecoil-mode for which the
coupling of the magnetic field is non-negligible.
; ; H2 = k⋅H1
Fig. 5.11.7 shows an example for strong differences between the two filter functions: a
DiMarzio DP-184 was operated in “split mode", i.e. only one of its coils was connected while
the other was left open (electrical idle). We may not conclude from this type of operation that
the coil in idle does not contribute: due to the unavoidable winding capacitances (in this case
around 500 pF), currents are also flowing in the disconnected coils – they generate a magnetic
field which has effects on the other (connected) coil. The split-mode distinguishes itself from
true single-coil operation in particular in the range around the resonance frequency. While the
connected coil represents a low-pass system (e.g. H1), the idle coil works as a band-pass (H2),
and consequently H1 and H2 differ significantly, with a resulting strong dependency on
direction.
Fig. 5.11.7: Transmission measurement (laser-vibrometer): DiMarzio DP-184 in singlecoil-mode.
We observe directionality, as well, when

connecting the two coils in series (Fig.
5.11.8), although the differences are smaller
than those experienced with single-coil
operation. Based on the measurement
results, we may assume that both coils have
the same number of windings, but
obviously the wire diameter is different
leading to different DC-resistances (3757
vs. 5100 Ω) and different winding- and
coupling-capacitances.
Fig. 5.11.8: : Transmission measurement (laser-
vibrometer): DiMarzio DP-184 in humbucking-mode
© M. Zollner 2002
5.12 Pickup noise
Every pickup generates undesired noise. The magnetic pickup converts magnetic interference
fields into noise voltages (Chapter 5.7), but it also creates already without the presence of any
magnetic field a broadband noise. As the term “broadband” indicates, the spectrum of this
noise is distributed continuously from very low to very high frequencies. The reasons are
conducting electrons in the copper wire which can move freely and perform stochastic
movements. The superposition of all these charge-movements leads to a noise voltage having
a normal (= Gaussian) distribution and being measured as an RMS-value . For a given
absolute temperature T, a measurement bandwidth B, and with Boltzmann’s constant k, the
resistance R yields:
Thermal noise voltage
For example, a 10-kΩ-resistor would generate a noise voltage of 1,27 µV for a measurement
bandwidth of 10 kHz. However, this calculation is only valid for real (i.e. purely ohmic)
resistors without any loading. As a first consideration it will suffice to model the magnetic
pickup via its coil resistance R, its inductance L, the load capacity C (predominantly
contributed by the cable, and the cross-resistance Rq (Fig. 5.12.1). Rq is made of three parallel
resistors: the amplifier input resistance (typically about 1 MΩ), the volume potentiometer of
the guitar, and the tone potentiometer connected via a “tone”-capacitor. With respect to noise
voltages, this “tone”-capacitor may be seen as a short so that all three resistances are
connected in parallel. For a typical Stratocaster pickup we find, for example: R = 6000 Ω, L =
2,2 H, C = 0,7 nF, Rq = 111 kΩ. Merely the ohmic resistances in the circuit generate thermal
noise; it is modeled for R via the series-connected noise voltage source U, and for Rq via the
parallel-connected noise current source I. Both noise processes run independently of each
other so that their effects can be superimposed after separate calculation. We do need to
consider that the two RMS-voltages have to be added according to the Pythagorean law – as
it is required for interacting incoherent signals. For the calculation we first omit the current
source and obtain the terminal voltage generated by U; subsequently a short replaces the
voltage source and the terminal voltage generated by I is calculated. The two terminal
voltages are each squared and then added; the square root of the result is the actual noise
voltage.
The spectral distribution of the noise is shown in the noise spectrum with the frequency
running along the abscissa. Along the ordinate we find the noise power spectral density
(W/Hz), or the normalized square-root of it – the so-called noise voltage density
Fig. 5.12.1: Equivalent circuit diagram of pickup; left without, right including noise sources. The ECD is the
basis for the noise spectra shown in Fig. 5.12.2.
© M. Zollner 2002
The thermal noise power of an ohmic resistor R amounts to 4kTB and is independent of the
resistor value. The noise power referenced to the bandwidth B is called the power spectral
density PSD = 4kT = 10-20 W/Hz, with 293K (room temperature) used for T. Since the
power spectral density has the same value for each frequency region (i.e. it is independent of
f), this noise is termed white noise; as is the case for white light, “all frequencies are
contributing”. In circuit technology, the noise voltage density en is normally used instead of
the PSD; it is obtained by dividing the noise voltage by the square root of the bandwidth:
. A 6-kΩ-resistor generates white noise with a noise voltage density of
9,85 . In Fig. 5.12.1, this value characterizes the noise voltage source designated U.
However, the noise arriving at the terminal is not a white noise anymore but it is low-pass
filtered by L and C. The left-hand section of Fig. 5.12.2 shows, in the lowest curve, the
frequency dependency of the noise voltage density generated by R and found at the output
terminals. The resonance emphasis caused by the low pass is clearly recognizable at 3,8 kHz.
The second noise source is the cross resistance Rq. Its noise is expediently modeled via a
parallel-connected noise current source, the spectral noise current density in of which is
en / Rq. 111 kΩ yields 382 . To calculate the terminal voltage generated by this
resistance, in needs to be multiplied with the absolute value of the circuit impedance; this
result is given in the left part of Fig. 5.12.2 by the middle curve. Below 1,8 kHz, the noise
contributions of the coils resistance R dominate, and above this frequency the noise
contributions of the potentiometer/amplifier-resistances Rq. The latter in fact deliver the
overall largest share of the noise. For the figure, a constant percentage bandwidth of 1/12th of
an octave was chosen. The relative 1/12th-octave-bandwidth is 5,8% corresponding to 5,8 Hz
absolute bandwidth at 100 Hz, and to 58 Hz at 1 kHz. A noise voltage density of
9,85 generates – for a bandwidth of 5,8 Hz – a noise voltage of 23,7 nV
(corresponding to a voltage level of –152,5 dBV). In the right-hand section of Fig. 5.12.2, the
white-noise-levels of amplifying devices (tube, FET, operational amplifier) for
5,5 (ECC83, LT1113), and 18 (TL071), respectively, are shown as dotted
lines. The TL071 downgrades the pickup noise below 2 kHz while the ECC83 and LT1113
add almost no noise at all. For FET-operational amplifiers and tubes, the effects of noise
currents (10 ) can be ignored. Using a bipolar-transistor op-amp such as a NE5532
having about ca. 500 would, however, not be purposeful, despite the good en-value.
Fig. 5.12.2: Noise voltage denstity of a pickup (left), 1/12th-octave-level (right). Calculation done for the ECD of
Fig. 5.12.1 with R = 6kΩ, L = 2,2H, C = 700pF, Rp = 111kΩ. => Uoverall = 2,2 µV.
© M. Zollner 2002
5.13 Pickup microphonics
That a pickup is “microphonic” means that it is susceptible to air- and structure-borne sound.
Actually, a pickup should only react to string oscillations, but whether it was conscious or
subconscious, some designers have included rather efficient microphones into their guitars: as
one speaks to one’s instrument (“god-awful acoustics” …”dropped that bloody slide AGAIN”
… “who the **** came up with these lyrics”), everyone can hear it coming over the speakers.
In most cases it was probably an overambitious developer who sought to shield against hum,
but in fact added – in the form of a sheet-metal housing – a microphone membrane. Which
on top of everything is totally a lame duck to keep out magnetic fields at low frequencies.
Maybe we could simply pass over the whole subject with an “I never talk to my guitar”-
approach, but all too often the issue develops a life of its own and ends in a high-pitched
whine. Like it is the case with (proper) microphones, feedback develops as soon as the loop-
gain increases over 1. Especially shielding covers made of steel sheets are dreaded. Even Seth
Lover, famous developer with Gibson, sought to encase his PAF with a housing made from
steel sheet metal. He had very correctly recognized that the relatively bad conductivity of this
material leads to low eddy-current losses. Because steel is difficult to solder, German silver
was in the end used – the standard in the premium range. It is not known exactly which type
of steel Seth Lover originally wanted to use: there are indeed non-magnetic steel variants, but
most are ferromagnetic and covers made from them – as they are stimulated by airborne
sound – would induce significant voltages in the pickup coil, just as the string does. In the
early days, when guitar amps were not operated much in overdrive mode, this would have not
grown into too much of a problem. However, amplifier power and loop gain increased rapidly
and significantly … and suddenly guitars had pickups whistling catastrophically. "There was
a pickup for sound-hole-mounting designated ‘GM100’ which had – probably due to blatant
ignorance – the whole housing made of steel sheet metal. In terms of microphonics, it broke
all records [Lemme]". Come to think of that this pickup was supposed to be mounted on a
feedback-prone acoustic guitar ….. it’s howl-scream-whistle-city ….
Even using ‘non-magnetic’ brass sheets would have been – cosmetically – rather counter-
productive due to the yellow, quickly oxidizing color. Such covers were therefore nickel-
plated (yellow-ish color) or chrome-plated (blue-ish color). Nickel, however, is ferro-
magnetic and a similarly good conductor for magnetics as is electric sheet metal. On the other
hand, chrome indeed is paramagnetic i.e. practically non-magnetic, and so is aluminum. Still,
it is not sufficient to just use nonmagnetic materials: moving a conductor (the sheet metal)
within a magnetic field induces an eddy current in this conductor – and this eddy current
again generates a magnetic AC-field which generates an AC-voltage in the pickup coil.
In a much-simplified model we can describe the cover mechanically as a spring-mass-system.

Below the resonance frequency, the spring is contributing more, while above the resonance,
the mass does. Together, sound-pressure and surface area of the cover generate a surface-
normal force which below the resonance has (cooperating with the spring) the effect of a
frequency-independent pressure-displacement-function. Above the resonance, the system is
mass-inhibited and the pressure-displacement-function is proportional to 1/f 2. Since for a
ferromagnetic cover it is not the displacement but the velocity which determines the induction
effect, an overall band-pass-shaped transmission results. The maximum voltage happens at
the resonance of the cover. If that has low dampening (this seems to be the normal case for
sheet metal), tremendous amplification factors (Q-factors) can appear. For non-magnetic
metals a velocity-proportional eddy-voltage is generated which is transformed upwards
according to the number of turns in the coil.
© M. Zollner 2002
A simple experiment can help us estimate resonance frequency and Q-factor: given there is
enough gain set in the amp, tapping on the pickup cover with a non-magnetic item (such as a
pick) will generate a noise from the speaker. A short “tock” or “tuck” advantageously
indicates a low resonance frequency and a strong dampening; less desirable is a higher
pitched “bing”, because it would stand for low-dampening and a resonance frequency in the 2
– 3 kHz-range: rather fatal since here also the pickup resonance resides.
Everything has at least two sides to it, though: such a pickup casing resonance may give a
guitar a characteristic sound, as long as it does not (yet) lead to unwanted whistling. The
guitarist may indeed have bought that specific guitar due to that specific sound. Moreover,
since a pickup casing has 6 walls, there is a good probability that not just one but several
resonances are in the game. Although designed to be of wondrous shielding quality, a pickup
with a complete metal sheet surround may reveal undreamt-of sound qualities … again: only
as long as the amp is not turned up too far. A wah-pedal in facts does something quite similar
in that it creates a resonance emphasis (which can be altered with the pedal position). We
have now arrived in an area where it is the turn of the “vintage guru”: “the original PAF-
pickups did not have potted coils so that the resonances and all that were much stronger,
much more authentic; the harmonics could unfold much more freely. Everything breathes and
sings, and is not as clean as the later high-tech-replicas behave.” All bullshit? Well, there may
be a grain of truth in there, or a grain of salt. Either way, about 3000 PAF pickups were
installed on the ’58 and ’59 Les Paul’s alone. They will not generally be without housing
resonances, and among them there may well have been one with optimal structural
resonances. Whether this is audible, remains speculation, plus: what is “optimal”? To be on
the safe side, Gibson does today pot the ’57 Classic Humbucker with wax. The BurstBucker,
however, comes with “non-potted” coils; just like back-in-the-day without wax. In particular
if wax gets between the coil bobbin and the metal cover, it can dampen resonances.
Speculations about the relevance of resonances in the pickup housing can find support or be
rebutted to a fair extent by measurements. An experiment carried out in the anechoic chamber
should give objective transfer data. Several different pickups were mounted 1 m in front of
the mouth of a horn loudspeaker and the transfer coefficients were determined with the
substitution method. A Brüel&Kjaer-microphone (4190) served as reference for the sound
pressure level measurements. It became quickly apparent that the pickups did not only react to
airborne sound but also to the electromagnetic fields originating from the speaker and the
speaker cable. While this is certainly also a noteworthy characteristic, it was undesirable for
the given experiment, and a grounded grid between speaker and pickup ensured that only the
airborne sound had any substantial effect on the pickup.
Fig. 5.13.1 shows the free-field transfer factor of a select number of pickups. The smallest
sensitivity to airborne sound is found in the Gibson Toni Iommi; obviously it was designed
for high-gain applications i.e. strong overdrive. The whole pickup housing is potted with a
hard material: there are practically no vibrations in the metal sheets. The Gibson ’57 Classic
proves to be already more sensitive and yields about 50 nV/Pa at 3,4 kHz; i.e. at 1 Pa sound
pressure (= 94 dBSPL) the pickup generates 50 nV. Without cover, that is! Putting a cover in
place, we get – depending on the way the cover is fastened – a serious increase of the
sensitivity to airborne sound.
© M. Zollner 2002
The main resonance (1,35 kHz) depends strongly on the individual mounting. It is easy to
imagine that over the decades almost every frequency had the honor of being the dominating
pickup-housing resonance of a Les Paul – which raises the question about a pickup-housing
main-resonance.
Fig. 5.13.1: Transfer factor for airborne sound (free field). The plots characterize individual pickups, with the
inter-individual differences for pickups of the same type being considerable.
In normal operation and depending on string-material, action, and playing style, pickups
generate induction voltages up to about 2 V. Typical guitar loudspeakers are rated by their
manufacturer at e.g. 100 dB (1W, 1m) which implies as a rough approximation SPL levels of
e.g. 114 dB and a voltage of 10 mV induced by the corresponding airborne sound. This
voltage generated by airborne sound is therefore merely 1/200 of the voltages induced by the
© M. Zollner 2002
string vibrations. Still, it would be premature to conclude that resonances in the pickup
housing are generally insignificant. Depending on the specific playing scenario entirely
different relationships may arise. An opinion often expressed amongst guitar players is that
“pickup-whistling” (i.e. unwanted pickup feedback) would be a problem only for guitar-
amplification systems which generate very high SPL values: “In front of two Marshall stack
it’s gotta whistle”. This is however not correct as such. The determining factor here is the
loop gain i.e. the amplification-gain which signal is subjected to after having gone through
the loop once: from the guitar through amp and speaker, and through the room (as airborne
sound) back to the guitar.
Some numbers shall be given to exemplify: a guitar generates e.g. a voltage of 0,1V (due to
the movement of the string) which is amplified to 2V by the amp. For the 8-Ω-speaker, this
translates into 0,5W and results in an SPL of 97dB at a distance of 1m in front of the speaker.
If this sound now hits the pickup, the latter will generate e.g. 1.4mV due to its sensitivity to
airborne sound – in addition to the 0,1V mentioned above. The loop gain is 0,014 and thus
substantially smaller than 1. The guitarist may now turn up the amp, either to get more
loudness or to obtain more distortion, or he/she may add frequency-selective additional
amplification with tone controls or an equalizer. The loop gain will increase and approach 1;
in fact it may easily exceed 1. This is when pickup-feedback (i.e. whistling) occurs. Looking
at the system very theoretically, an additional special phase condition would also need to be
met, but that is always possible because of the manifold sound paths in regular room.
The sensitivity of the pickup to airborne sound will become apparent as sound coloration
already at a gain which does not generate feedback. One can look at this as if there was, on
top of the desired signal path, a signal decoupling into an additional effects channel. As there
is a sufficient level in that effects channel, changes in sound will become audible. A model
including a forward signal loop (sound pressure results in a voltage, HUp) and a feedback path
(voltage results in sound pressure, HpU) is shown in Fig. 5.13.2). The feedback path contains 5
places of resonance (dotted line); in the curve on top we see the consequences on the overall
frequency response. For a maximum loop gain of 0,1 ( dB) there will be no audible
effect; however, for a loop attenuation of merely 5 dB (corresponding to a loop gain of -5 dB)
pronounced effects will appear. It is not possible to generally determine how clearly the
resonance will bear down in the specific case, because sufficient signal energy needs to be
present in the respective frequency band – and more room remains again for speculation.
Fig. 5.13.2: Model of a signal loop and consequences of resonances on the overall transmission. The loop gain
SV is the product of the forward- and the feedback-amplification.
© M. Zollner 2002
In order to obtain at least some very rough data under regular operational conditions, a guitar
amplifier (VOX AD-60-VT) was analyzed in the anechoic chamber. A guitarist had set the
control such that, with a Les Paul (Historic Collection), a “slightly distorted, crunchy sound”
resulted. All effects incorporated in the amp were switched off. A measurement microphone
(B&K 4190) was positioned 1 m in front of the loudspeaker incorporated in the amp and
captured the sound resulting from a signal of 1 mVeff fed into the input "High" (Fig. 5.13.3). A
sound pressure level of just 1 Pa (94 dB) was generated at 2,5 kHz; the voltage gain was
about 500 (54 dB). Voltage-to-SPL transfer coefficient was HpU = 1 Pa/mV at this frequency.
In combination with the Duncan APTR-1 described in Fig. 5.13.1 the condition for an
oscillation ( ) would already be almost met – the 2,5-kHz-spike of this pickup
almost reaches HUp = 1 mV/Pa. To be fair, we need to remember again that all the wax with
which the potting was done had been removed. With the wax in place the sensitivity to
airborne sound would be less.
Fig. 5.13.3: SPL generated at 1 m distance for 1 mV input (left); gain factor up to the loudspeaker (right).
The sensitivity to airborne sound of the Les Paul guitar mentioned above was also determined
in the anechoic chamber; see Fig. 5.13.4. All strings were removed and the guitar was
positioned in 1 m distance in front of a horn loudspeaker. The guitar was loaded with 670 pF
(cable) and 1 MΩ and all control set to “10”. For the somewhat more sensitive bridge-pickup
(Gibson BurstBucker #2), we found a maximum sensitivity to airborne sound of just short of
0,1 mV/Pa. The neck pickup was loess sensitive by 5 dB. Unwanted feedback will not appear
under this condition since the smallest loop attenuation is 27 dB. For the same reasons, any
influence on the sound is not to be expected, either! (Compare to Fig. 5.13.2)
Fig. 5.13.4: Gibson Les Paul '59: Transfer factor for airborne sound (left), loop gain factor (right).
© M. Zollner 2002
As the gain is increased by 6 dB (relative to Fig. 5.13.3), a much more distorted guitar sound
is already generated for normal playing – well usable for lead-Sounds a la Beano Blues-
Breaker♣. The maximum loop attenuation is 21 dB and thus still well in the green. Not unless
all three volume controls (gain, volume, master) of the VOX amp are maxed out (introducing
an additional 26 dB of gain), the setup generates nothing but piercing whistling noises. Under
this condition it makes moreover no difference where the guitar is positioned in the room
relative to the amp – it remains “mission impossible”. The guitar would have to be removed
from the room, or the pickup selector switched to the neck pickup – the slightly less
sensitivity to airborne sound of that pickup♦ enables the guitarist to find a few positions which
are not subject to feedback. From the point of view of the conservative musician the resulting
messy sound is not actually desirable. Although: only now – so the control engineering
approach says – do the resonances of pickup-housing have an effect on the overall frequency
response.
From the carried-out experiments the following results can be derived:
1) A metal pickup housing with well-done dampening has little resonances and does not
change the pickup transfer characteristic in the case of distortion-free reproduction (clean
sound, stage volume) at all. (here we are not considering that a cover may cause eddy-current-
dampening → Chapter. 5.9.2.2). Even at “normal distortion” there are no effects on the sound.
At extreme gain-settings (ultra-distortion) some effects are conceivable – however: an entirely
distorted guitar sound is not really the right condition to be able to discern subtle sound
differences.
2) Pickup covers with weakly dampened resonances may have a sound-altering effect,
depending on the amplification. However, going on stage with such a caterwauler is a bit of a
ride on a cannonball: you never know at which point things will go sideways. If the loop
attenuation is high enough, you won’t hear any difference, there will be no effect of the
housing resonances, but as they become audible, the limit towards uncontrollable pickup
feedback is just a hair away, as well. This is of double (or triple) validity for thinline and full-
bodied electric guitars: their sensitivity to airborne sound is even larger than that of a badly
dampened pickup. Now, there are guitarists who are looking exactly for this borderline
situation, and some have even reached true mastery in that battle with the unbridled
resonance-power. So if a special guitar is said to have that very special unique sound drawing
upon the pickup housing resonances: impossible it is not from the point of view of physics.
With the single exception of the Gibson Toni Iommi, all the potted pickups examined in the
framework of this book showed – upon opening them up – a interior distribution of the wax of
… shall we say (to remain safe from the attorney assault): the wax distribution was following
artistic considerations. As such, every guitar again is a unique specimen. But we already knew
that, didn’t we … even without the whole physics shebang.
What remains is a matter of faith. Thesis: “The pickup covers, as well, add a material-specific
resonance to the sound. If you love that throaty and nasal PAF-sound (a la Allman Brothers
or even Peter Green), you should absolutely use covers on the pickups” [U. Pipper, Gitarre &
Bass, 9/2005]. That’s one way to look at it. Anti-thesis: "You may have heard that I remove
the covers from my pickups; the improvement in the sound is unbelievable” [Eric Clapton, in
Bacon/Day]. That’s the other side of the faith.
♣
Back in the day, the original setup included a Marshall combo amp.
♦
All these statements relate to one specific individual guitar.
© M. Zollner 2002
5.14 Pickups with shorts in the coil-winding 5-163
5.14 Pickups with shorts in the coil-winding
The coil of a magnetic pickup is made of very thin copper wire carrying an even thinner layer
of varnish for insulation. The insulation resistance of the varnish would still be sufficiently
high with a thickness as small as 4 µm, but to keep the insulation layer undamaged is
somewhat of a challenge. This was especially true in the old days when the magnet wire was
often directly wound onto the magnet rods and it could happen that the insulating layer was
abraded and shorts were introduced. Moreover, some of the insulating varnishes used back
then became brittle over the decades and came loose from the copper. It is also conceivable
that already the application of the varnish sometimes was sub-par or even faulty. Last, if the
quality control was done merely using an ohmmeter with a tolerance of 20% [Duchossoir,
Strat], much room remains for undetected shorted turns in the coils.
How does the transfer behavior of a pickup change it one or several windings are shorted out?
If indeed merely a single winding is shorted, the effects are negligible, but in case a wire
establishes contact to the next whole layer of the winding (or even the layer beyond that) we
would be confronted with a possibly substantial defect. It shows some naïveté if a pickup
manufacturer still writes in the year 2011 that it’s ok if a few hundred of 8000 turns of a
pickup are shorted out: indeed a few percent change in the DC resistance may be
insignificant, but the pickup operation is based on AC. And with AC a short in the winding
brings with it a resistive load and therefore a treble-loss.
It is purposeful to interpret a partially shorted inductance as a transformer (Fig. 5.14.1). Of

the N turns of the winding, n are shorted; they N-n non-shorted turns form the primary
inductance L1, while the remaining (shorted) n turns form the secondary inductance L2.
; ;
;
;
:
Rk = 0 yields an input
impedance Z of:
Fig. 5.14.1: T-equivalent circuit diagram of the transformer with hard coupling (top); ECD with short in the
winding (bottom)
© M. Zollner 2002
The copper-resistance (DC-resistance) of the full winding is R. R1 belongs to the primary

winding and R2 to the secondary winding. The resistance occurring between two turns (the
short resistance) is Rk. For a perfect short, Rk will be zero, but in the general model we will
assume an arbitrary value. A hard-coupled transformer without flux leakage can be described
by its T-equivalent-circuit-diagram, with L12 being the mutual inductance. The latter is
positive for a concordant coupling, and negative for a inverse coupling. Interpreting the
shorted inductance as transformer leads to a inverse coupling: L12 is negative.
From the ECD shown in Fig. 5.14.1 we can calculate the pickup impedance Z. For an ideal
short (Rk = 0) it may be simplified to the given formula. The DC-situation (f = 0) yields Z =
R1; however, towards high frequencies ( ), Z does not remain inductive but converges
to a real final value. With a further simplification (for n << N) we get R⋅N/n for this final
high-frequency end value. If e.g. 4% of a pickup winding is shorted, the end-value is 25 x R
(i.e. 25 x 6 kΩ = 150 kΩ for a typical Strat pickup). This only seems like a sufficiently high
resistance – for a capacitive load, the effect is substantial and the resonance emphasis drops
strongly (Fig. 5.14.2). As a consequence of the reduced Q-factor (compare to Chapter 5.9.3)
the resonance emphasis of the transfer function goes down, as well. This is depicted in Fig.
5.14.3 with a Stratocaster pickup serving for the example. A short across 2 layers of winding
is approximately equal to n = 280; the corresponding loss in brilliance is not negligible
anymore.
Fig. 5.14.2: Short in the winding. Left: without parallel capacitance; right: with parallel capacitance (850 pF).
Stratocaster-Pickup: R = 5700 Ω, L = 2.2 H, N = 7600, n = 280. Without (----) and with (–––) short.
Fig. 5.14.3: Without (----) and with (–––) short in the winding, data as in Fig. 5.14.2. Left: pickup with purely
capacitive load (850 pF), right: 110 kΩ load resistance added (potentiometers + amp).
© M. Zollner 2002
5.14 Pickups with shorts in the coil-winding 5-165
So. More than 160 pages about magnetic pickups – quite a heavy load. To conclude, let’s
bring in a goodie for those who persevered (no, not that Thorben-guy, he was not available –
and he’s had it, anyway). But we have Mr. Chris Kinman, well-known pickup manufacturer.
He had some news for his followers published on his website around Christmas in 2010
which we may look into here:
Chris K. was repairing two ’64 Strat pickups both of which had succumbed to broken coil
wiring. For one of the two, the fracture had occurred right on the outside of the pickup; so that
one is dealt with easily but the other’s gonna be a lot of work: it has to be rewound entirely.
Some original wire (i.e. the real Voodoo-stuff) was brought in, the rewinding done …
however the two pickups sounded differently. That remained the case even after the magnets
had been re-magnetized. Writes Chris: "This experiment exploded the myth that aged magnets
were the reason for this massive difference in sound. Another well known pickup
manufacturer claims weaker magnets are the reason that old pickups sound sweet, but I can
not confirm that claim when I deliberately degauss magnets." Well, he’s right on target:
magnets do not age (he could have read up on that in Chapter 4, by the way). That vintage
sound must still have some reason, though, and here it comes: "It turns out that Formvar
insulation is not age stable, it's an unsophisticated old technology coating that degrades over
time, unlike modern Polyurethane coatings which seem to go on forever. … So there you have
conclusive scientific proof for aging of old Fender pickups, Formvar wire degrades in time. It
definitely is not due to aging of magnets." The “scientific proof” then uncharitably hides
behind an impedance plot which indicates at the resonance frequency (3,2 kHz) a maximum
value of merely 41.25 kΩ♣, but even given this there are still differences between the two
pickups: the "1964 original Strat pickup that has aged excessively" indeed shows only 36
kOhm at the most. Approximately, that is – since the 4,46 kΩ per scale-division chosen by
Chris K. makes it difficult to interpolate. Anyway, the older the pickup, the smaller the Q-
factor will get because the aging insulating varnish encourages shorts in the winding. With the
decreasing Q the "ice-pick brittleness" goes away and the aged sound (less treble) is in reach.
That sound is – according to Chris K. – simply due to shorts in the coil winding. Conclusion:
anybody who would like to play a 1954 Strat but would rather invest money in old Aston
Martins does not really have a problem. Just buy a new Strat, turn down the “Tone” control a
bit: voila – aged sound. However: it is now psychologically prohibitive to ever again read
music “trade journals” because there the investor may find the statement that the old Strats
have an unequalled brilliant sound (more treble).
Good advice? You are very welcome. As a return service, someone could pay a visit to Chris
Kinman and show him how correct impedance measurements are done. Having said that, he
actually deserves much credit because he does make the effort and takes some decent
instrumentation to the pickups. Many manufacturer seem not to do even that …
♣
With a purely capacitive load, that should be (without the potentiometers) about 300 kΩ sein, and with the pots
still about 88 kΩ.
© M. Zollner 2002
empty page
© M. Zollner 2002
5.15 Collection of data 5-167
5.15 Collection of data
The following pages compile the most important parameters of a selection of pickups. Some
of the latter were bought new right before the measurements were taken while others already
looked back on a long service life of 40+ years in musical action. None of the pickups
necessarily is a typical representative of its kind but there was no indication of the contrary,
either. It is safe to assume that for pickups produced by way of modern manufacture the
production tolerances are small. This qualitative assumption is, however, not secured via an
analysis of variance, and it therefore has a speculative character. For old pickups, there
might have been a considerable spread over the years of production although again there is no
proof for this. Some old pickups are traded for more that $ 10000, - per piece; this supports
the doctrine of correlation between price and demand. It did not seem justifiable, though, to
enter this control loop and purchase e.g. 20 Gibson PAFs just to verify the parameter
variances in these pickups. Even for the brand-new and therefore relatively “inexpensive” P90
(compared to vintage pickups), the required sum (290. - Deutschmark in 1998) was invested
only once. In view of the rather plain construction, comments that come to mind regarding
this stately price are exclusively non-printable.
During all measurements, we made certain that errors due to instrumentation contributed
merely in a negligible manner, if at all, to the result. Likewise, we went to great lengths to
ensure that interference effects were insignificant, or – if unavoidable – at least clearly
recognizable. For level measurements, the tolerance was typically 0,1 dB which is more than
adequate since changes become only audible if they exceed 1 dB. Position measurements
close to magnetic poles proved to be more difficult. From a scientific point of view,
tolerances of below 0,1 mm would have been desirable but could not always be achieved. For
the user, this error is tolerable, since guitarists do not actually adjust the distance between
string and magnet with a higher precision. The temperature of the measurement objects was
not taken; in the vast majority of the cases it should have been room temperature between 20
and 24 °C.
Impedance measurements on pickups were done with imprinted current (3 mAeff); the load
by the meter connected in parallel was, at R > 10 MΩ, insignificant. In order to emulate the
shift in resonance frequency due to cables, various styroflex capacitors were connected in
parallel; in the framework of impedance measurements, the losses of these capacitors are
negligible.
To pickup remained load-free for the measurement or the calculation of the transfer behavior,
or it was loaded with a typical circuitry. Fig. 5.15.1 shows the complex loading model, and a
simplified equivalent circuit of this typical loading – the differences are insignificant in the
relevant frequency range.
Fig. 5.15.1: Pickup-loading. T = pickup, Zi = source impedance, RV = Volume control, RT = Tone control, CT =
Tone-Cap, L, RK, CK, GK are elements of the guitar cable, Ra, Rb and Ca model the guitar amplifier. The R||C-
circuit shown in the right represents a well-suited equivalent circuit.
© M. Zollner 2002 Translated by Tilmann Zwicker

5-168 5. Magnetic pickups
In Fig. 5.15.1, the pickup T is modeled by its source-impedance Zi , volume and tone controls
are assumed to have their maximum resistance value i.e. both potentiometers are turned fully
clockwise. For guitars with singlecoil pickups, these potentiometers frequently have a value
of 250 kΩ, and for humbucker-equipped guitars usually the 500-kΩ-version is used. The tone
control capacitor CT often has 22 nF. The guitar cable is modeled via a series-inductance
which – having a value of about 0,3 µH/m – may be ignored in the audio range with good
approximation. Similarly, series resistance (RK < 1 Ω/m) and parallel conductance (GK < 5
nS/m) may be left without further consideration. The parallel capacitance CK, however, needs
to be taken into account; it shows about 100 pF/m (see also Chapter 9.3). For a typical tube
amplifier, the parallel resistance Ra at the input usually amounts to 1 MΩ, and between the
input jack and the first tube we frequently find two 68-kΩ-resistors connected in parallel (Rb).
The tube input capacitance is enlarged by the Miller-effect; in combination with the wiring in
the amplifier we obtain about 150 pF for Ca.
Although the load circuit shown in Fig. 5.15.1 looks complicated, it substantially has no other
effect in the audio range than the R||C-circuit also given in the figure. Fig. 5.15.1 exemplifies
the differences between the two circuits. The R||C-circuit was optimized for 1 kHz; the small
differences occurring at low and high frequencies are inconsequential in practice. As an
orientation, each figure also shows the plots for two further R||C-circuits in which the resistor
differs by 5 kΩ and the capacitance by 5 pF. Considering that typical tolerances found in
potentiometers are much larger (> ± 25 kΩ), and that a capacity difference of 5 pF
corresponds to a difference in length of 5 cm of the guitar cable, we clearly see that the
equivalent circuit given here generates much less of a difference that the tolerances of regular
component do. The equivalent circuit is therefore well suited to model the pickup load. With
respect to the phase, differences are also negligible.
RV = 250 kΩ, RT = 250 kΩ. RV = 500 kΩ, RT = 500 kΩ.

R = 111,4 kΩ, C = 755 pF (-------) R = 200 kΩ, C = 750 pF (-------)
R = 110,4 kΩ, C = 760 pF (–––––) R = 199 kΩ, C = 755 pF (–––––)
Fig. 5.15.2: Differences of the amount of the impedances (ZRC/Z) of the load circuit of Fig. 5.15.1. CK = 600pF
was taken for the guitar cable, 1 MΩ, 34 kΩ, 150 pF for the amplifier input. The differences at low frequencies
are caused by the tone capacitor (CT = 22 nF), the ones at high frequencies by the tube-capacity. Since all
component values have a tolerance of at least 5% (even 20% are not unheard of), the accuracy of the model is
easily sufficient. In the low-frequency region, the source impedance of the pickup is less than 20 kΩ – any
discrepancies in the transfer parameters are therefore reduced to much less than 1%.
Translated by Tilmann Zwicker © M. Zollner 2002

5.15 Collection of data 5-169
Legend to the figures
For the measurements, all pickups were given an electrical “environment” as they would find
it in their typical real-life situation: connected to a volume- and a tone-control (presenting a
load consisting of two resistors in parallel), and to an amplifier input with an input impedance
of 1 MΩ. This meant that all Fender-type single-coil pickups (typically connected to 250-kΩ-
pots) faced a resistive load of 111 kΩ, and all Gibson-type humbuckers one of 200 kΩ (due to
the typical 500-kΩ pots). The latter “Gibson-scenario” was also used for P-90-type pickups,
the Gretsch pickups, the Rickenbacker humbucker, and the VOX pickup. Special cases were
the Fender Jazzmaster, Jaguar, and original “Wide-Range” humbuckers (where 1-MΩ-pots
were used): here the load was 333 kΩ. The Rickenbacker “Toaster” faced 91 kΩ (due to the
additional series capacitor and volume control in the circuit), and the DeArmond Rhythm
Chief 1100, FHC, and “Hershey-Bar” pickups with their integrated controls were connected
to 167 kΩ, 167 kΩ, and 50 kΩ, respectively. For the connection to the amplifier, a cable is
required, and the amplifier also has a capacitive input impedance – this was simulated for all
pickups by a capacitive load of 450 pF and 750 pF (to take into account different cable
lengths); an additional measurement was made with no capacitive load (but with the ohmic
load)..
Impedance frequency response: the left-hand figure depicts the amount of the pickup
impedance as a function of frequency – once without any load (open), and twice for purely
capacitive load (450 pF and 750 pF, respectively). The impedance level is referenced to 1 kΩ:
; 40 dB correspond to 100 kΩ. It is noted that looking at the
impedance level is not a customary approach – it is however very advantageous since it can
be interpreted easily. The angled orientation lines belong to purely inductive impedances. The
cable connecting the pickup to the potentiometers is only considered if it is soldered internally
to the pickup (as is the case e.g. for Gibson Humbuckers). The potentiometers are not
considered (as load).
Transfer frequency response: the right-hand figure shows the transfer behavior (Chapter
5.9.3) of the pickup under load. The latter is the parallel connection of a resistor R and a
capacitor C (Fig. 5.15.1). R models the usual guitar potentiometers and a high-impedance
amplifier input (1 MΩ); C is the capacitive load with the same values as given in the left-hand
figure (0 pF, 450 pF and 750 pF, respectively). The absolute sensitivity is indicated at the left
margin with a dot; reference is a Stratocaster pickup (see Chapter 5.4.5). The ferromagnetic
flux density B is indicated at the upper right (at 2 mm distance, Chapter 5.4.1).
Equivalent circuit diagram (ECD): the ECD shown was derived from impedance
measurements; in terms of its complexity it represents a compromise between effort and
accuracy. For most pickups, this ECD does serve well to model their transfer behavior; only
for pickups with strong eddy-current losses it is of limited accuracy. The comb-filter
frequency response found in humbuckers is not modeled with this ECD (for the impedance-
equivalence see Chapter 5.9.2.3).
Pickup-data depend on the individual production processes; pickups of seemingly equal

build can differ substantially in their electrical and transfer characteristics.
For the on-line publication only a limited number of pictures is released. The complete
documentation is for the time being reserved for the print-version only. The text “Diese
Abbildung bleibt der Druckversion vorbehalten” is the German explanation for a figure
reserved for the printed version and thus excluded here.

Fender Stratocaster
The basic construction of this pickup was developed by Leo Fender as early as the late 1940’s
when he put together the Broadcaster (later Esquire, Telecaster): 6 alnico magnets are stuck
into 2 flanges of vulcanized fiber, a lot of thin wire is wound between the flanges around this
assembly … and done. This approach did prove itself, and the Stratocaster guitar issued 1954
received three identical pickups of the type. The 16 - 19 mm long alnico-5-magnets have a
diameter of 4,8 mm (3/16”), and the enameled copper wire of 0,063 mm thickness was first
insulated by formvar, then with plain enamel and later (after 1980) with polysol. The first
Stratocaster pickups sported a number of turns between 8000 and 8700 – the belt-driven
counters did not allow for a higher precision. In the 1960’s, new automatic winding machines
were introduced, and after some to-ing and fro-ing (or hither-and-thither, back in the day) a
new standard was agreed upon: 7600 turns. Later came 7800, 8200, 8500, 9000, 9600 … it
seems to be an endless story of variation and change. An abundance of speculations exists
about the number of turns on the old pickups, but it is unlikely that this “secret” will be ever
unlocked because nobody is going to unwind a 1954-Stratocaster just to check the winding
and the number of turns (and even if somebody did that: what would we know about every
other pickup given the variances in production back then?). A resistance measurement does
not help, either: the wire diameter varies too much.
Fid. 5.15.3a: Fender Stratocaster [www.fender.com].
Fig. 5.15.3b: Strat-like pickup [www.phousemusic.com, www.guitar-letter.de].

5.15 Fender Stratocaster 5-171
Fender Stratocaster '72

Singlecoil pickup, 6 cylindrical alnico magnets. #17492, probably from 1972.

Coaxial "Singlecoil"-pickup, 6 cylindrical alnico magnets.

Fender USA-Standard Stratocaster (Neck)

Singlecoil pickup, 6 cylindrical alnico magnets.
Fender USA-Standard Stratocaster (Bridge)

Singlecoil pickup, 6 cylindrical alnico magnets, 2 field-amplifying screws.

Fender Japan-Strat
Singlecoil pickup, ferrite bar-magnet.
Rockinger (Strat Type)


Ibanez Blazer (Strat Type)

Seymour Duncan SSL-1 (Strat-Type)

Singlecoil pickup, 6 cylindrical alnico-magnets.

Lace gold
Singlecoil pickup, special guidance for the magnetic field
DiMarzio SDS1
Singlecoil pickup, 2 ferrite bar-magnets.

Fender Telecaster
Fender Telecaster Bridge-Pickup: the bridge pickup of the Esquire (the predecessor of the
Broadcaster and the Telecaster) was developed in 1948 and 1949. It is based on the same
construction as the Stratocaster pickup with 6 alnico-5 magnets and 2 flanges of vulcanized
fiber. About 9000 turns of AWG-43 wire were wound onto the assembly. At the beginning of
the 1950’s a change to the thicker AWG-42 wire occurred, with the basic construction
remaining the same. A special feature of the Telecaster bridge pickup is a metal base-plate – it
was originally zinc-coated but later received copper coating. Dropped in 1983 to lower
expenses, it was later re-introduced for special versions of the Tele. The “Texas-Special” has
more turns than the normal version. The pickup is fastened in the cutout of a metal support
plate that has the effect of an eddy-current-dampener and reduces the resonance emphasis.
Fender Telecaster Neck-Pickup: the Telecaster’s neck pickup was developed in 1950 –it
was first used in single specimen of the Esquire and then predominantly in the Broadcaster
and Telecaster. It included 6 alnico-5 magnets, two flanges of vulcanized fiber and about
8000 turns of AWG-43 wire. Over the years there were many changes in some details but the
basic construction remained the same. The Telecaster neck-pickup also has a special feature:
a shielding cover made of nickel silver (German silver, Cu-Ni-Zn). For cheap copies nickel-
plated copper is used, as well. The cover acts as an eddy-current-dampener.
Fender Telecaster Humbucker-Pickup: the same Seth Lover who developed the Gibson
Humbucker also designed this pickup. Compared to the Gibson it features a slightly larger
distance of the pole-pieces, and 12 adjustable CuNiFe individual magnets. Since CuNiFe has
lost all significance as magnetic material and is difficult to obtain, new versions of this pickup
are produced with a ceramic or alnico bar-magnets.
Fig. 5.15.4a: Fender Telecaster [www.fender.com].
Fig. 5.15.4b: Fender Telecaster-pickups [www.fender.com, http://img3.musiciansfriend.com].

5.15 Fender Telecaster 5-177

Seymour Duncan APTR-1 (Tele-Type, Neck)


DiMarzio DP-172 (Tele-Type, Neck)

Telecaster-Fake (Neck)




Fender Texas-Special Telecaster (Bridge)



Seymour Duncan APTL-1 (Telecaster-Type, Bridge)

Telecaster-Fake (Bridge)

Fender Wide-Range Humbucker

Humbucking pickup, 12 CuNiFe magnets.
Two potentiometers 1 MΩ each i.e. load impedance = 333 kΩ
Fender Thinline-Reissue Humbucker

Humbucking pickup, alnico bar-magnet.
Two potentiometers 250 kΩ each i.e. load impedance = 111 kΩ

Fig. 5.15.4c: Fender Telecaster Thinline [www.fender.com].
Three variants of the Telecaster received the "Wide-Range"-humbucker (as it was called):
Thinline (second generation), the Custom (second generation) and the Deluxe. Originally
fitted with CuNiFe-magnets, later copies are of a completely different build and fitted with
bar magnets (both of the alnico- and ceramics-variants). On top of this, there are different
circuits with potentiometers of 250 kΩ or 1 MΩ, and capacitors of 10, 22 or 50 nF.
Two different Fender humbuckers could be examined: one original Wide-Range from the
1970’s and a reissue pickup from a 2012-Thinline. The original has surprisingly special
magnets. The reversible permeability of the CuNiFe-magnets is very small (compare to Fig.
4.42), and their static magnetization is decidedly peculiar. That may be a special characteristic
of this individual pickup – we could not establish an assembly-line-production standard with
just one representative. In any case: two of the magnets feature south poles on both facing
surfaces! Yes, this is indeed physically permissible if a north pole is located in between the
two south-poles. Also, the pickup does not become totally unusable, either, because the large
differences occur on the lower side of the pickup. This still is rather peculiar …vintage, in any
case and after all….
Fig. 5.15.4d: Fender Wide-Range humbucker [www.fender.com]. In the right-hand diagram the magnetic flux
density measured at a distance of 2 mm is given; within the circle = top, numbers below: bottom of pickup

Fender Jazzmaster
Leo Fender developed this pickup for the Jazzmaster guitar introduced of in 1958. The pickup
included 6 alnico-5 magnets, 2 flanges of vulcanized fiber and about 9000 turns of AWG-42-
wire. Production was shut down in 1982, and later re-opened for replicas. Due to the larger
coil the Jazzmaster-pickup is relatively prone to interference fields. At the same time, the
larger size does not make its sensitivity profit in the same way because the magnetic field of
the strings is locally limited, and thus hum-rejection is only moderate (Chapter 5.7).
The larger-size coil was chosen because Leo Fender sought to sample a section of the
vibrating string as long as possible in order to obtain a wider spectrum. This approach holds
two conceptual errors: first, the width of the aperture does not depend on the coil but on the
magnet, and second, the width of the aperture and the frequency bandwidth are reciprocal to
each other i.e. there is an inverse dependency.
The Jazzmaster pickup has relatively short alnico magnets creating next to no eddy-current
dampening. In combination with the high-resistance 1-MΩ-potentiometers, the result is a very
treble-laden, almost shrill sound. That the Jazzmaster would have a soft sound “due to its
wide coils” is pure fantasy (Chapter 5.4.4). Despite the similar look, the Jazzmaster pickups
must not be confused with the Gibson P-90, either.
Fig. 5.15.5a: Fender Jazzmaster [www.fender.com]
Fig. 5.15.5b: Jazzmaster pickup [www.petesrareguitars.com]

5.15 Fender Jazzmaster 5-185

Large singlecoil pickup, 6 cylindrical alnico magnets.

Large singlecoil pickup, 6 cylindrical alnico magnets.

Fender Jaguar
Designed to be the ultimate flagship, the most expensive of all Fenders, with all of the top
features. Meaning: Floating Bridge and Tremolo (from the Jazzmaster), Lead/Rhythm-
Selector (also from the Jazzmaster), two wide-range high-fidelity pickups (its only very
special feature), a String-Mute (jeez-no …). Plus, often forgotten: a 61-cm-scale! In the
meantime, one could do very well without the string-mute, the Stratocaster already had a
much better working vibrato-system, the rhythm-circuit – involving merely the neck pickup –
was rather limited … so why invest all that money, thank you very much? Because of the
pickups? O.k., these were indeed different from all the other Fenders, featuring a u-shaped
metal shield. Allegedly, that had the effect of “focusing the magnetic field”, of “increasing
sustain” (!), of “reducing sensitivity to hum”, of “improving the sound” – the patent
application (issued as US 3,236,930) wallows in superlatives: the Jaguar would be highly
superior compared to all the junk produced up to (that) date.
It would not be fair to suggest that icons such as Leo Fender might have not been playing with
a full deck – no, in case of doubt it was the patent attorney who formulated that kind of
nonsense, with a lot of tolerance on the side of the patent examiner. Pay attention, guitarists
near and far: the guitars built until 1962 (i.e. pre-Jaguar, such as the Stratocaster) had,
according to the Jag-patent, undesirably characteristics, poor response to string vibrations,
much shorter sustain, were profoundly insensitive in the bass-range, their sound was with
small harmonic content. Why on earth then did they have to so soon reduce and finally shut
down the production of the highly superior Jaguar? By the way: we also find in the patent
description that the Jaguar pickup would be highly simple and economical ... that again does
sound like Leo.
An effect of the u-shaped magnetic-field guide (which Leo Fender hat patented once again in
US 4,220,069) cannot be disputed, though: the treble loss created by it as well as the magnetic
shielding is examined closely in Chapter 5.4.7.
Fig. 5.15.6a: Fender Jaguar [www.fender.com]
Fig. 5.15.6b: Fender Jaguar pickup [www.guitar-parts.com]

5.15 Fender Jaguar 5-187
Fender Jaguar
Singlecoil pickup, 6 cylindrical alnico-magnets, u-shaped, serrated metal shield.

Gibson P-90
A large singlecoil pickup, the P-90 was developed by Gibson technician Walter Fuller and
produced from 1946. It features 2 alnico-5 bar-magnets, 6 pole-screws, a bobbin and about
10.000 turns of AWG-42 wire. From 1957 onward it was installed only in the lower grade
Gibson models – but again from 1968 also in Les Pauls. It has a black or cream-colored
housing and also came (as the “dog-ear” version) in a metallic housing with two flanges.
P-100-L (L = Lead, bridge position, about 10 kOhm), P-100-R (Rhythm, about 6,5 kOhm).
Coaxial pickup with two coils on top of each other and connected in parallel. Not a big
success, it is not produced anymore.
P-94 = P90 in the format of a Gibson Humbucker. The bobbin is shorter and slightly higher
than that of the p-90.
Fig. 5.15.7a: Two "Soap-Bar" P-90 in a Gibson Les Paul [www2.gibson.com]
Fig. 5.15.7b: Two "Dog-Ear" P-90 in an Epiphone Casino [www2.gibson.com]

5.15 P-90 5-189
Gibson P-90 (Soapbar)

Large singlecoil pickup, 2 bar magnets, adjustable pole-screws.
Rockinger P-90 (Soapbar)

Large singlecoil pickup, 2 bar magnets, adjustable pole-screws.

Rickenbacker
From about 1931, Adolph Rickenbacker installed electromagnetic pickups on Hawaiian

guitars in California. The instruments were made of metal and later also of Bakelite. From
1936, various stringed instruments were fitted with a horse-shoe-magnet-pickup, and from the
mid-1950’s solid-body and hollow-body electrical guitars were available. When at the
beginning of the 1960’s the Beatles were sighted with Rickenbacker guitars, the brand
became popular in Europe and the UK, as well.
Fig. 5.15.8a: Rickenbacker 325, three "Toaster"-pickups [www.fatendfirst.com].
The pickup examined in the following was installed in a 1966 Capri. In the US, the
corresponding model range was designated Model-335 while the export-version was named
Model-1996. The US-version had a wedge-shaped sound-hole (“cat’s eye”) while the export-
version sported an f-hole.
Fig. 5.15.8b: Rickenbacker "Toaster"-pickup.
The humbucker internally has 2 narrow blades at a distance of 12 mm.
Fig. 5.15.8c: Rickenbacker Humbucker.

5.15 Rickenbacker 5-191
Rickenbacker "Toaster"
Large singlecoil pickup, 6 cylindrical magnets (∅ 6,3 mm).
Load impedance 81 kΩ
Rickenbacker Humbucker
Humbucking pickup, bar magnet with 2 blades.

Gretsch
The first Gretsch guitars (from 1949) were fitted with DeArmond pickups; in 1957 the
FilterTron was introduced, and from 1961 the HiLoTron pickup was available. Its eccentrical
pole-arrangement is similar to that of a humbucker, but in fact this is a singlecoil with a very
special guide for the magnetic field (Chapter 5.10.5). The FilterTron, however, is a true
humbucker (Chapter 5.7). The replicas built today have at least the cosmetics in common
with the originals. Approximately, anyway ….
Fig. 5.15.9a: Gretsch guitar with 2 FilterTron pickups [www.gretschguitars.com].
Fig. 5.15.9b: Gretsch-guitar with 2 HiLoTron pickups [www.gretschguitars.com].
Fig. 5.15.9c: Gretsch-pickup [www.gretschpages.com].

5.15 Gretsch 5-193
Gretsch HiLoTron
Large singlecoil pickup, alnico bar-magnet.
Gretsch FilterTron
Humbucking pickup, alnico bar magnet.

Gibson Humbucker
Developed around 1955 by Gibson technician Seth Lover as an alternative to the P-90, it
targeted the reduction of the sensitivity to magnetic interference. It includes two side-by-side,
serially-connected coils, an alnico bar-magnet, 6 polepieces (slugs) and 6 pole-screws
(Chapter 5.9.2.6). The Gibson Humbucker is produced in a number of variants: with or
without cover, with different magnet materials, with more or less winding on the coils, and
equal or different numbers of turn between the two coils. Anyone seeing 5% difference
between the coils as substantial will happily fork over the extra $.
The Gibson Humbucker samples the string in two locations – this leading in particular to
considerable treble-loss in the sound of the low strings (Chapter 5.4.4). However, this effect is
not necessarily undesirable: with strong distortion a special sound does result.
Fig. 5.15.10a: Two Gibson Humbuckers in a Gibson Les Paul [www2.gibson.com]
Fig. 5.15.10b: Gibson Humbucker [www.boutiquemusicinc.com, rainbowguitars.com].

5.15 Humbucker 5-195
Gibson Burstbucker (Neck)

Gibson Burstbucker (Bridge)


Gibson '57 classic

Gibson Tony Iommi

Humbucking pickup, bar-magnets.

Fender Squier (Neck)

Humbucking pickup by Squier (Fender), similar to the Gibson Humbucker but with smaller inductance.
Fender Squier (Bridge)

Humbucking pickup by Squier (Fender), similar to the Gibson Humbucker but with smaller inductance.

Gibson Firebird
Built in the 1960’s, the Firebird featured humbuckers with two side-by-side coils; in the
1970’s there was also a Lawrence-variant with coils rotated by 90°. The originals had an
alnico bar-magnet inserted fully within the coil and operating as a sort of magnetic blade. The
build is not visible since there are no openings in the top of the metal cover. The distance of
the poles is 12,5 mm and therefore smaller than that seen in the standard Humbucker (19
mm), and consequently the interference gaps are found at different places. The resonance
frequency is higher than that of the standard Humbucker; some specimens with somewhat
fewer turns can almost reach values found in Stratocasters.
Fig. 5.15.10c: Gibson Firebird with 2 Firebird pickups [www2.gibson.com]
Gibson Mini-Humbucker
Gibson-Mini-Humbuckers were installed in the Les Paul Deluxe from 1969 to 1984. It is
tempting to surmise that this is simply a reduce-in-size version of the standard Humbucker,
but in fact it is an alternative (probably inspired by Epiphone). One coils carries the usual
pole-screws, but the other does not have slugs but a thick steel-blade running the full length.
Like in the Firebird pickup, the pole-distance at 12,5 mm is smaller than that seen in a
standard Humbucker (19 mm). As a consequence, and also aided by the inductance bearing
towards lower values, the Mini-Humbucker sounds more brilliant than a standard Humbucker
Fig. 5.15.10d: Gibson Les Paul Deluxe with 2 Mini-Humbuckers [www2.gibson.com]

Gibson Firebird
Humbucking pickup: 2 alnico-5 bar-magnets located within the coils, pole-distance merely 12.5 mm.
Gibson Mini-Humbucker
Humbucking pickup: bar magnet underneath the coils; pole-distance merely 12.5 mm.

VOX pickup
VOX became famous in the 1960’s with their tube amplifiers but also distributed electrical
guitars. The HDC-77 shown below is not from the old days but a new development. The
pickup is new, as well – allegedly. In an interview, VOX chief designer Eric Kirkland stated
that “Planetz” were the "inventor of the CoAxe Pickup", and continues to explain: While there
are many good-sounding humbucking and stacked "single coil" pickups available, they are
generally limited to one good noise-canceling sound. The unique magnetic structure and
tapped coils of the VOX CoAxe offer both the power of a traditional humbucker and the high
frequency detail of a true single coil – and the most effective noise canceling available in a
passive pickup. This is not entirely incorrect: indeed the hum-cancellation is impressive and
so is the loudness of the pickup. What about the sound? Supposedly, the pickup should sound
like a Fender singlecoil, or a P-90, or a humbucker. That’s what it should do, but apparently
the pickup is oblivious of this job it was given.
The string-sampling at two positions is the characteristic of a humbucker. The P-90 and the
Fender singlecoil sample at one position but the CoAxe does so at three. How is the CoAxe
supposed to emulate a humbucker if its magnetic filed cannot not be switched? Only its coils
are switchable (they are tapped) – which gives merely a meager effect: 2,5 dB change in level
and 20% shift in resonance frequency. In the humbucker mode a series-circuit of a resistor
(100 kΩ) and a capacitor (1,5 nF) is connected to the pickup – that’s all (although achieved
via a monster switch with no less than 28 terminals). How could this turn a P-90 into a
humbucker? Sorry: close but no cigar. The CoAxe pickup is nice, but it remains different
compared to the targeted role models. Also: it ain’t something new, either – see the patent by
Aaroe, applied for in 1981.
Fig. 5.15.11a: VOX guitar with 2 CoAxe pickups [www.korg.co.uk].
Fig. 5.15.11b: CoAxe pickup [www.planetz.com].

VOX CoAxe-Pickup
Humbucking pickup: 2 co-axial coils, 6 slugs, 2 blades, 2 ceramic magnets.
Fig. 5.15.11c: top: impedance; bottom: normalized transfer (laser measurement, string-specific!)
Fig. 5.15.11d: circuitry (left: innere Spule = inner coil, äußere Spule = outer coil), cross-section of the CoAxe
pickup (right)
Fig. 5.15.11e: pictures taken from US-patent 4,372,186 (Aaroe, 1981).

Conrad pickup
The globally operating sales chain Conrad Electronic sells two pickups at an unbelievably low
price (date: 2012): there is a so-called “P-90 Soapbar” for as little as € 12,75, and on top of
that a so-called “PAF-Custom” for a staggeringly small amount of € 10,95. This is a blatantly
obvious demonstration that neither the cost for the materials nor the cost for the production of
a pickup needs to be very high. At the same time (2012), Gibson (USA) charges $ 141,59 for
a P-90 and no less that $176,99 to $212,39 for a Humbucker. Of course everybody expects
Gibson pickups to be much better that the low-cost Conrads – but are they? O.k. – at least in
Germany the Gibsons are not as expensive as the official US price list suggests, but a P-90
will still set you back in the order of € 70, and a Gibson Humbucker will drain € 85 – 100
from your pocket. The Conrad humbucker is € 10,95 - …. isn’t that asking for trouble?
Well, at least it’s not a complete disaster. The PAF-Custom does not carry any pole-screws
but 12 slugs – a fate it shares with many a colleague. The associated adjustability is however,
not that important. We could establish the fact that different eddy-current losses are present as
a deviation, and even as a deficiency, but the PAF-Custom does not seek to replicate your
regular Gibson PAF. Rather, it shows autonomy: its transfer factor is about 2/3rds higher than
that of a Gibson Humbucker (e.g. a Burstbucker) and its resonance frequency is about 20%
lower. It is thus darker and louder in sound than the original.
The P-90 is by about 40% less sensitive than the Gibson P-90 and its resonance frequency is
about 50% higher than that of the original – in comparison it is therefore softer and brighter.
One might like that sound but it is not that of the original. The Conrad P-90 sounds like a
strong Strat pickup even though it does not look like it. We can take the description “a
singlecoil just like the original” to be misleading, or we accept it grudgingly, because the
original is indeed a singlecoil.
Bottom line: neither Conrad pickup replicates the sound we would expect them to have on
the basis of the designation. If, however, we see them as “their own pickups”, they are a fully
fit alternative at an unbeatably low price. Of course, we know nothing about their longevity –
they may evaporate without residue after 5 years …
Fig. 5.15.12: Conrad pickups [www.conrad.de]

5.15 Conrad 5-203
Conrad P-90 Soapbar

Large singlecoil pickup, 2 ceramic bar-magnets, 6 adjustable pole-screws.
Conrad PAF-Custom
Humbucking pickup, ceramic bar-magnet, 12 slugs; very high sensitivity.

DeArmond
In cooperation with Rowe, the DeArmond company has produced and distributed a large
variety of pickups since the 1930’s. They were installed e.g. in guitars made by Gretsch,
Guild, Epiphone, Eko and Höfner, and of course into big-bodied acoustic guitars … that’s
how it all started, anyway. A pickup sought after even today is the Rhythm-Chief (see
Chapter 5.4.8). It was available in different variants – just as several realizations existed under
one and the same name for other DeArmond pickups. All the DeArmonds examined here are
fully encased in sheet metal provoking substantial eddy-current losses. Thus, the transfer-
function derived from the two-terminal equivalent circuit is not adequate; additional
laser measurements are shown in Chapter 5.4.8.
http://theunofficialmartinguitarforum.yuku.com http://www.harmonycentral.com
DeArmond Rhythm-Chief 1100

Singlecoil pickup, probably with plastic magnet within the coil.
Load impedance 167 kΩ (integrated controls)

5.15 DeArmond 5-205
DeArmond FHC
Load impedance 167 kΩ (integrated controls)
DeArmond "Hershey-Bar"
Load impdedance 50 kΩ (integrated controls)

5.16 Patents und inventions
5.16.1 American Patents (selection)

1890 435679 Breed: the first guitar pickup? 1890!!
1927 1933299 Vierling: piano-pickup (PU)
1929 1839395 Kauffman: tremolo (vibrato-unit)
1929 1838886 Tuininga: violin-HB
1930 2027073 Vierling: piano-PU (see also 1933299)
1931 1906607 Jacobs: piano-PU
1931 1915858 Miessner: piano-PU
1931 1978583 Kentner: piano-PU
1934 1941870 Severy: synthesizer
1934 2020557 Loar: guitar with structure-borne-sound-PU
1934 2025875 Loar: guitar with structure-borne-sound-PU
1934 2089171 Beauchamp: Rickenbacker Frying Pan, Horseshoe-PU
1935 2026841 Lesti: PU w/out permanent magnet
1935 2119584 Knoblaugh: stacked HB w/out permanent magnet
1936 2087106 Hart/Fuller (Gibson): Charlie-Christian-PU
1936 2170294 Dopyera: National Hawaiian guitar, Blade-PU
1936 2152783 Beauchamp: Rickenbacker Electro Spanish Guitar, Horseshoe-PU
1937 2175325 Sunshine/Epiphone: "Oblong Pickup"
1938 2145490 Miller (Gibson): further development of 2087106
1938 2241911 Kauffman: motorized tremolo for steel-guitar
1939 2262335 Russell: HB w/horseshoe-magnet
1940 2261358 Fuller (Gibson): retrofit-PU
1940 2294861 Fuller (Gibson): retrofit-PU
1944 2455575 Fender/Kauffman: Solidbody guitar w/PU
1946 2455046 DeArmond: PU "Type-1000"
1948 2542271 Alvarez: Piano-HB
1948 2567570 McCarty (Gibson): PU within the pickguard
1948 2686270 Ayres: piano-HB
1949 2557754 Morrison: Solidbody guitar w/PU potted in wax, 6 cylindrical magnets
1950 2573254 Fender: Telecaster-precursor
1950 2612072 DeArmond: PU w/6 adjustable cylindrical magnets
1950 2612541 DeArmond: PU w/6 adjustable cylindrical magnets PU and u-shaped yoke
1952 2683388 Keller: HiLoTron PU
1952 2740313 McCarty: McCarty-bridge
1952 2737842 Les Paul: bridge & tailpiece
1953 2784631 Fender: tone control
1954 2741146 Fender: Stratocaster
1954 2911871 Schultz: similar to P-90
1955 2896491 Lover: Gibson-HB
1956 2817261 Fender: first Fender-HB (for steel guitar)
1956 2909092 DeArmond: cylindrical magnets w/threaded sleeves
1956 2964985 Webster: slideable stereo-PU
1957 2892371 Butts: Gretsch-HB
1957 2968204 Fender: 7 magnets for 6 strings
1959 2976755 Fender: split-PU (for Precision-Bass)
1960 3035472 Freeman: bar magnets, out-of-phase-circuit
1961 3147332 Fender: split-PU
1962 3236930 Fender: Jaguar-PU (with u-shaped yoke)
1962 3249677 Burns: low-impedance-PU, sampling of individual strings
1964 3290424 Fender: Marauder
1968 3541219 Abair/Rowe: cylindrical magnets w/threaded sleeves
1969 3588311 Zoller: bi-directional PU
1970 3657461 Freeman: stacked HB
1971 3571483 Davidson: omni-directional Pickup
1971 3715446 Kosinski: right you are if you think you are
1974 3902394 Stich/Norlin: HB w/standing coils

5.16 Patents and inventions 5-207
1974 3916751 Stich/Norlin: HB w/standing coils

1974 3983778 Bartolini: wide magnetic field
1975 4026178 Fuller: PU with u-shaped yoke
1975 4051761 Nylen: magnetize strings by hand
1977 4133243 DiMarzio: SDS-1
1978 4145944 Helpinstill: ´magnetic field & structure-borne sound
1979 4220069 Fender: PU w/ceramic bar magnet and yoke
1979 4283982 Armstrong: HB, ceramic magnet in between the coils
1980 4320681 Altilio (DiMarzio): PU with top-position plastic magnet
1981 4364295 Stich (Lawrence): 2-blade-HB
1981 4378722 Isakson: string-penetrating coil
1982 4442749 DiMarzio: stacked HB
1983 4463648 Fender: angled HB
1983 4524667 Duncan: stacked HB
1985 4624172 McDougall: tube-like polepieces
1985 4686881 Fender: serrated polepieces
1986 4581974 Fender: dummy-coil
1987 4809578 D.A. Lace: Lace-"sensor"
1991 5168117 Anderson: stacked HB w/neodyne-magnet
1995 6372976 Damm (Gibson): P-94
1996 5668520 Kinman: stacked HB
1997 5792973 Riboloff (Gibson): HB w/3 magnets
1998 6291758 Fender: stacked HB
1999 6846981 Devers: stacked HB
Breed: Is this the first guitar pickup? In a way, yes, even though the idea was to obtain an actuator – but
reversible transducers work on both directions. Pity that the first amplifier tube was still 16 years away …

US-Patents, numerical list
1215973 1557476 1838886 1839395 1906607 1915858 1929027 1933299

1978583
2001392 2015363 2020557 2020842 2025875 2026841 2027073 2027074

2048515 2087106 2089171 2119584 2145490 2152783 2170294 2175325
2179237 2209016 2222959 2225299 2228881 2235983 2236946 2239985
2241911 2261358 2262335 2263973 2293372 2294861 2310606 2323969
2327277 2340001 2413062 2455046 2455575 2503467 2542271 2557754
2567570 2573254 2581653 2612072 2612541 2628524 2683388 2686270
2725778 2737842 2740313 2741146 2764052 2784631 2793293 2817261
2892371 2896491 2897709 2909092 2911871 2958249 2961912 2964985
2968204 2976755 2989884
3003382 3035472 3066567 3079535 3084583 3147332 3177283 3183296

3207976 3236930 3249677 3288906 3290424 3417268 3435610 3472943
3483303 3530228 3535968 3541219 3544694 3571483 3588311 3602627
3657461 3657481 3668295 3711619 3715446 3725561 3902394 3911777
3915048 3916751 3962946 3963975 3969771 3983777 3983778 3992972
4010334 4026178 4050341 4051761 4056255 4096780 4133243 4138178

4143575 4145944 4151776 4164163 4171659 4175462 4188849 4201108
4220069 4222301 4254683 4261204 4268771 4269103 4283982 4319510
4320681 4364295 4372186 4378722 4379421 4408513 4412454 4442749
4463648 4472994 4480520 4499809 4501185
5031501 5111728 5136918 5136919 5153363 5155285 5168117 5204487

5221805 5252777 5290968 5311806 5336845 5354949 5376754 5389731
5391831 5391832 5399802 5408043 5418327 5422432 5430246 5438157
5438158 5455381 5463185 5464948 5530199 5569872 5602353 5610357
5641932 5668520 5670733 5684263 5789691 5792973 5811710 5834999
5894101 5898121 5908998 5949014
6103966 6111185 6162984 6271457 6291758 6291759 6372976 6414233

6476309 6525258 6846981 7227076

5.16 Patents and inventions 5-209
5.16.2 Discoveries, inventions, and other milestones of technology

1694 Voltaire: France’s precursory experiment. Even the name failed.
1800 Volta: the Italian original. Electrochemical series, electrical battery
1820 Ampere: current-force-law
1820 Oersted: electric current deflects a magnetic needle
1823 Sturgeon: electromagnet
1826 Ohm: Ohm’s law: U = R ⋅ I
1830 Henry: law of induction
1831 Faraday: law of induction; dynamo, magnetic field
1833 Gauß/Weber: electromagnetic telegraph
1855 Maxwell: electromagnetic field theory
1860 Philipp Reis: contact microphone
1861 Philipp Reis: telephone (w/out commercial success)
1863 Thomson: theory for the condenser microphone
1863 v. Helmholtz: teachings of sensations of sound
1876 Bell: improved telephone (see also Gray)
1877 Berliner: carbon microphone
1877 Lord Rayleigh (John W. Strutt): Theory of Sound
1878 v. Siemens: patent for the electro-dynamic loudspeaker
1878 Hughes: improved microphone (USA)
1878 Lüdtge: improved microphone (Germany)
1878 Edison: Phonograph (precursor of the turntable)
1881 Gaulard/Gibbs: transformer
1883 Edison: Edison-effect (electron emission from hot cathode)
1886 Hertz: generation of artificial electromagnetic waves
1887 Berliner: grammophone, sound disc (record)
1890 Breed: US-Patent (Nr. 435679) Horseshoe-Actuator for guitar
1890 White: carbon-gain microphone (series production standard)
1906 – 10 von Lieben, deForest: amplifier tube (Triode)
1913 Langmuir, Schottky: double grid tube (Tetrode)
1915 Jensen/Pridham: Magnavox-loudspeaker
1923 E. Reiß: carbon-powder microphone
1925 Rice, Kellog: moving-coil cone-loudspeaker
1925 Lilienfeld: field-effect-transistor
1926 Tellegen: thee-grid-tube (Pentode)
1927 Vierling: electromagnetic piano-pickup (w/permanent magnet)
1928 Pfleumer: tape recorder
1929 Tuininga: electrical violin w/structure-borne-sound-sensor)(humbucker)
1931 deArmond (Rowe): retrofit guitar pickup
1932 Beauchamp (Rickenbacker): solidbody gitarre "Frying Pan", horseshoe-pickup
1932 Celestion: permanent moving coil loudspeaker
1933 First artifical head recording (no, not yet the "Aching Head")
1934 Loar (Vivi-Tone): guitar w/magnetic structure-borne-sound-PU
1935 Lesti: solidbody guitar w/string-humbucker (no permanent magnet)
1935 Knoblaugh: stacked humbucker (no permanent magnet)
1935 Telefunken: magnetophone
1935 Gibson: Electric Hawaiian Guitar and Amp EH-150
1936 Beauchamp (Rickenbacker): Electro Spanish Guitar (Bakelite), Horseshoe-Pickup
1936 Hart/Fuller (Gibson): Charlie-Christian-Pickup
1936 Dopyera (National): aluminum-Hawaiian- guitar wit "Blade-Pickup"
1936 Alnico-Magnets, Rola G12-Speaker
1937 Sunshine (Epiphone): steel-guitar with "oblong-shape-pickup" (patent application)
1938 Fender: Radio Repair Shop
1939 Russell: guitar-humbucker
1940 Gibson: ES-125, ES-300 w/singlecoil pickup
1941 Zuse: first digital computation machine Z3 (elektromechanical)
1941 Hewlett&Packard: Wave-Analyzer HP-300A
1941 Les Paul: guitar prototype "The Log"
1944 Fender: Leo’s first electric guitar "Electro Spanish"

1946 Fender: Leo’s first guitar amplifiers: Princeton, Deluxe, Professional

1946 Fuller (Gibson): singlecoil pickup P-90
1946 DeArmond: "Type 1000" guitar pickup
1947 Shockley, Bardeen, Brattain, Pearson: bipolar transistor
1947 Paul Bigsby: Merle-Travis-guitar, plus further custom guitars
1948 CBS: vinyl-LP
1948 Alvarez: piano humbucker
1948 McCarty (Gibson): pickguard with pickup and potentiometers
1949 Morrison: solidbody guitar, potted pickup with individual magnets
1949 Brüel&Kjaer: first level meter
1950 Fender: Telecaster precursor (US Patent 2573254)
1950 DeArmond: pickup with 6 adjustment screws for the magnet
1951 Fender: Precision-Bass
1952 Keller: pickup as seen later with Gretsch (HiLoTron-Pickup)
1952 McCarty (Gibson): McCarty-bridge
1952 Les Paul (Gibson): bridge & tailpiece
1952 Gibson: Les Paul guitar
1952 world’s first hydrogen bomb misses Los Alamos by almost 10000 km
1954 Fender: Stratocaster (US Patent 2741146)
1955 Seth Lover: Gibson Humbucker (US Patent 2896491)
1956 Fender: humbucker for steel guitar
1957 Butts: Gretsch Humbucker
1957 Fender: Precision-Bass with split humbucker
1958 Jennings: VOX AC-15
1959 Hank Marvin (The Shadows) starts grinning – and never stops again …
1959 Berry B. Goode
1960 v. Bekesy: Experiments in Hearing (Nobel-price 1961)
1960 WEM Copicat 'D.T.S. Model'. D.T.S stands for "Death to Selmer" [Elyea]
1962 Fender: Jaguar-Pickup w/u-shaped yoke
1962 Bran: first Marshall-Verstärker
1962 Beatles: first VOX AC-30
1964 Early prepatory work for “Physics of the Electric Guitar”
1965 Cooley, Tukey: FFT-Algorithm
1970 Hendrix pulls the plug
1970 Freeman: stacked humbucker
1970 Lover (Fender): CuNiFe-humbucker (Telecaster-Thinline und -Custom)
1974 Stich (Gibson): humbucker with upright coils
1976 Apple I, personal computer
1977 Brüel&Kjaer: first digital third octave analyzer
1978 Brüel&Kjaer: single channel FFT-analyzer
1979 Armstrong: humbucker
1981 IBM-PC, personal computer
1981 Stich: 2-blade-humbucker
1982 Commodore C-64, home computer
1982 DiMarzio: stacked humbucker
1983 Duncan: stacked humbucker
1987 D.A. Lace: Lace-"sensor"
1990 Gibson: stacked humbucker P-100
1991 Anderson: stacked humbucker
1996 Kinman: stacked humbucker
1998 Turner (Fender): stacked humbucker ("Noiseless")
1999 Devers: stacked humbucker
1999 Zollner: Physics of the Electric Guitar 

6. Piezo pickups
Around 1880, the French physicist Pierre Curie discovered the piezo-effect: as crystal sheets
of special material are deformed, an electric voltage occurs on their surface. If we mount a
small plate of crystal to a guitar top or to a guitar bridge, the vibrations of the guitar lead to
deformations of the crystal and therefore to electrical signals. In contrast to the principles of
the electro-magnetic transducer, it is not the original string oscillation that is tapped into, but
the effect of the latter onto the section of the guitar that can vibrate. This has the advantage
that the vibrations of non-magnetic strings can also be captured. However, piezo pickup and
magnetic pickup differ not only in the transducer principle but also in their transmission
function: the magnetic pickup captures the string velocity (particle velocity v) at one location
(single-coil pickup), or at two locations (humbucking pickup) – with position-dependent
comb-filter effects (Chapter 2.8). Conversely, the piezo pickup converts the force acting
within the bridge (bridge-insert pickup), or it captures the movement of a small area of the
guitar top (top-mounted pickup). In the 1960’s, top-mounted pickups manufactured by
Barcus-Berry started to capture the market as retrofit kits, while in particular the Ovation
company fitted their guitars ex-factory with bridge-insert pickups. By now, some solid-body
guitars are also feature a piezo pickup as alternative or as supplement to magnetic pickups.
6.1 The piezo-effect
As external forces act on special crystals, an electrical polarization results in addition to the
deformation – this is due to shifts in the charges located in the material. The descriptive
piezo-electric material parameters in fact have a tensor-characteristic, since both the
mechanical and the electrical tensions act in three dimensions. For the guitar pickup, however,
a simplified description will suffice. A scalar material-characteristic connects both force and
electrical voltage, and particle velocity and current with each other in the sense of a two-port
(quadripole) electro-mechanical transducer.
Today, most piezo sensors are manufactured from artificially polarized ferro-electric crystal
mixtures (lead zirconium titanate) that can be optimally matched to the specific applications
by suitable doting and composition. PVDF-foils (polyvinylidine fluoride) are also deployed.
The piezo ceramics are formed from mixed crystals, and need to be polarized after
manufacture (sintering, sanding, metalizing) at high temperature by a strong electric DC-field.
Over the years, this polarization will decrease again – but only to a relatively insignificant
extent so that long-term stability is, as a rule, very good. Thermal or mechanical overload may
lead to substantial deterioration of the transmission behavior, but such irreversible changes
must not be feared after the sensor has been mounted in place.

6-2 6. Piezo pickup
The transducer constant of a piezo pickup depends on the geometry of the piezo sheet
(surface S, thickness h) and on the material-specific piezo-constant e. For an unobstructed
thickness-mode transducer, we obtain simple correspondences between the mechanical and
electrical quantities that are operating in parallel [3]:
; ; Transducer equations
The voltage U generated at the crystal is proportional to the force F, the current I is proportio-
nal to the velocity v; the transducer constant α is the coefficient of proportionality. In terms of
manufacture, the piezo constant e can be trimmed over a wide range; typical values are
e = 20 ... 50 N/Vm. As a first value for orientation, α is 1 N/V.
Moreover, the piezo crystal does not only convert mechanical quantities into electrical ones
but it also includes mechanical and electrical elements. As a simplification, we need to
consider the stiffness s on the mechanical side, and the capacitance C on the electrical side. It
is not surprising that a small sheet of glass-like hardness and a modulus of elasticity of around
5⋅1010 Pa features a high stiffness. However, a significant share of the stiffness is caused by
the electromechanical coupling: we obtain from the capacitance equation I = C⋅dU/dt, via the
transducer equations:
The stiffness sC (caused electrically and converted to the mechanical side) contributes
significantly to the overall stiffness that results from the sum of the crystal stiffness sk and the
capacitance stiffness sc: s = sK + sC. This sum also shows up when considering the energy
scenario: when compressing a small crystal sheet, potential field-energy is stored (even
without the piezo-effect). The piezo-effect causes an electrical voltage – and thus potential
electrical energy – to appear across the crystal sheet (a dielectric with high dielectric
constant). This potential electrical field-energy is supplied by the mechanical side and
generates a load to the mechanical source just like an additional spring.
Fig. 6.1: Equivalent circuit of the piezo-transducer.

The driving force F is divided up into F1 and F2.
Fig. 6.1 shows the electro-mechanical equivalent circuit diagram. The partial forces F1 and F2
act in parallel; the representation in the figure (a scalar flow-diagram of the forces) disregards
directions in space. As we short the output (U = 0), the input force of the transducer becomes
zero, as well, (F2 = 0), and sK now is the sole active spring. Literature designates this special
load case with a superscript E (clamped field-strength E). The modulus of elasticity
measured under these conditions is symbolically termed with EE (the superscript E is no
mathematical exponent here). Given this, we find for sK:
Crystal stiffness
Typical values are: EE = 5⋅1010 Pa … 8⋅1010 Pa (modulus of elasticity without piezo-effect).

6.1 The Piezo-Effect 6-3
The modulus of elasticity EE describes the mechanical elasticity of the piezo-material for full
decoupling, as it is achievable for any value of α if the electrodes on the piezo sheet are
shorted. For electrical measurements, on the other hand, the decoupling occurs for every value
of α if any movement is prevented. As a thought experiment: for v = 0, the secondary current
of the transducer is zero, and therefore only the capacitance of the crystal CK remains. In
literature, this special case is designated with a superscript S (clamped mechanical relative
deformation S). Typical values for the relative dielectric constant are = 1000 … 4000.
Given the above, the capacitance of the crystal CK can be calculated:
; Crystal-capacitance
For regular (unclamped) operation, we measure – on the electrical side – two capacitances
connected in parallel: the crystal-capacitance CK, and the spring capacitance Cs caused by the
mechanical side:
Spring-capacitance
In summary: For electrical no-load, i.e. the open-circuit situation with only CK having an
effect on the electrical side, we measure two stiffnesses on the mechanical side: s = sK + sC.
For mechanical no-load with only sK having an effect on the mechanical side, we measure two
capacitances on the electrical side: C = CK + Cs. The reactive load that is transformed by the
transducer to the respective other side is: . Given high-grade
electro-mechanical or mechano-electrical linkage♣, we may assume ; depending on
the piezo material, lower values are possible, as well.
6.2 Electrical loading
An open-circuit connection at the transducer represents an idealization that does not really
exist in this form. For a piezo-electric guitar pickup, the electrical side is loaded via the cable
(acting as a capacitance) and the input impedance of the amplifier, while on the mechanical
side, the bridge and the strings need to be considered. Let us assume as a first approach an
imprinted force F to be the mechanical source. In order to calculate the output voltage, the
simplest approach is to transform both this force and the crystal stiffness onto the electrical
side [3]:
Transformed quantities
This transformation yields a purely electrical network that may be investigated using the
known approaches for network analysis. Of particular importance is the effect of the electrical
load impedance on the transmission function H. The input impedance of a guitar amplifier
typically amounts to 1 MΩ; relative to this, a line input can be of much lower impedance
(e.g. 50 kΩ). The capacitive internal impedance of the pickup forms – on cooperation with the
input impedance of the amplifier – a first-order high-pass (Fig. 6.2).
2
♣
Linkage-factor: k = Wmech / WΣ = Cs / (Cs + CK) = 0,3 ... 0,5.

6-4 6. Piezo pickup
Seen from the side of the load resistance, the two capacitances are connected in parallel, and
they therefore are added up to calculate the cutoff-frequency of the high-pass (3-dB-
frequency): C = CK + Cs.
Cutoff-frequency of the high-pass
With C = 1,5 nF and R = 1 MΩ, we obtain fg = 106 Hz, which is a value matching the
frequency range of the guitar. For R = 50 kΩ (line input), the cutoff frequency would rise to
2,1 kHz, corresponding to a complete loss of the lows. Even less suitable would be a
microphone input: its input impedance usually amounts to only about 2 kΩ.
Fig. 6.2: Equivalent circuit of the transducer
For an active piezo pickup, the input impedance of the guitar amplifier is immaterial – a
battery-operated preamplifier with low output impedance is integrated into the guitar, and
makes for a problem-free connection to line inputs. A passive piezo system, however, does
not include a preamplifier, and the pickup-signal needs to be fed to the guitar amplifier using
a shielded cable. High-quality cables represent merely a capacitive load (Chapter 9.4), and
therefore increase CK. On the one hand, this has the effect of a broadband signal attenuation
(capacitive divider), on the other hand it decreases the cutoff frequency of the high-pass. The
ubiquitous assumption that a long cable would attenuate in particular the treble range does
not hold for the piezo pickup – the internal impedance of the latter does not have a resistive
character but a capacitive one.
6.3 The piezo-transducer as a sensor
In its operation as a sensor, the guitar pickup converts mechanical input signals (string
vibrations) into electrical output signals. This represents the normal case; an operation as
actor – which is also possible – is of interest only in the context of metrology (Chapter 6.5).
According to the idealization used so far, the piezo transducer works as force-to-voltage
converter. It captures in particular the AC-component of the bearing force the strings cause,
and generates a correspondingly proportional voltage. The bearing force does not act directly
onto the piezo crystal, though – a pickup housing, with its masses and springs, represents a
mechanical filter. The effects of this filter will be investigated from a metrology-point-of-
view in the following, using an Ovation pickup (Fig. 6.3) as example.
Fig. 6.3: Two different piezo pickups (Ovation).

6.3 The piezo-transducer as a sensor 6-5
For the measurements, the guitar was laid – within its opened case – on a stone table (with a
weight of 250 kg). An electromechanical shaker (B&K 4810) generated translational
vibrations; force and acceleration could be measured with an impedance-measurement head
(B&K 8001). A small chisel blade was screwed into the impedance head and set onto the
saddle-piece of the bridge, and the load for the pickup during the measurements was 2 MΩ.
Fig. 6.4 shows the measured frequency-dependence of the transmission factor, with the
corresponding ideal curve of a 2nd-order low-pass indicated as a dashed line. The resonance
frequency is located at about 3,9 kHz, the Q-factor is about 18.
Fig. 6.4: Transmission (left) and dynamic stiffness (right) of the piezo pickup (Ovation).
The shown transmission behavior must not be interpreted as the “frequency response” of the
guitar. In the left graph of Fig. 6.4, we see the frequency dependence of the 20-fold logarithm
of the quotient U/F; with U = voltage at the piezo and F = measured force. However, the force
measured in the impedance head does not correspond exactly to the force at the bridge; rather,
a small additional mass is also measured – this mass is due to the mounting plate of the
impedance head towards the load. In other words: during the measurement, there is a small
additional mass of 1,4 gram located on the bridge, and the effect of this mass is also
measured. Together with the stiffness of the bridge, the additional mass generates a resonance
at 3,9 kHz. To confirm this hypothesis, the right-hand graph of Fig. 6.4 shows the quotient
of the measured force F and the measured deflection x (again in the usual dB-scaling i.e. the
20-fold logarithm). For f > 100 Hz, we can nicely recognize the behavior of a mass-spring
system, with its idealization shown in Fig. 6.5. As a simplification, the guitar bridge acts as a
stiffness (about 800 kN/m), and together with the additional mass contributed by the
impedance head, it forms a resonance at 3.9 kHz.
Fig. 6.5: Dynamic stiffness F/x of an ideal

spring-mass system.

6-6 6. Piezo pickup
The resonance at 3.9 kHz results from the cooperation of the stiffness sR of the bridge pieces
and the additional mass m0 generated by the end plate of the impedance head and the chisel
blade. The exact value of m0 can easily be measured applying a no-load condition of the
impedance head: Fig. 6.6 correspondingly shows the magnitude of the complex quotient F/a.
For an ideal mass, a frequency-independent graph would have to result; any deviations are
effects of structural resonances (impedance head and cable).
Abb. 6.6: Magnitude of the co-vibrating complex

mass F / a of the impedance head (incl. chisel).
“LEERER IMPEDANZ-KOPF: Masse” = empty
impedance head: mass; “Frequenz” = frequency
Assuming time-invariance (not reachable to a full 100%), any artifacts of the measurement
head may largely be compensated: first, the force- and acceleration-signals are digitally
recorded with highest possible quality for the empty measurement head. In a second step, the
corresponding analytical signals F and a are generated using a Hilbert transform. The
quotient of F and a, suitable averaged, yields the complex mass m0 = F/a. In order to achieve
the compensation when the head is under load, m0 is simply subtracted from the measured
F/a-quotient. Fig. 6.7 depicts the results achieved with this compensation. As transmission
factor, we get a value of 0,2 V/N (frequency-independent, as a 1st-order approximation), for
the bridge-stiffness, about 800 kN/m result.
Fig. 6.7: Measurements with compensation of the measurement-head artifacts. Transmission factor (left), bridge
stiffness (“Steg-Steifigkeit”, right). “Frequenz” = frequency.
In summary: as expected, the investigated piezo-pickup (Ovation EA-68) operates as a

force→voltage-converter with a transmission coefficient of 0,2 V/N. The guitar bridge acts as
stiffness of about 800 kN/m. For the simple model, these values are frequency-independent;
the 3,9-kHz-resonance is and artifact caused by the drive. A closer check reveals small
structural resonances not reflected via the simple model.

6.3 The piezo-transducer as a sensor 6-7
Fig. 6.8 condenses the results obtained so far into an equivalent circuit diagram. On the
electrical side of the two-port network of the transducer, we see the capacitance of the crystal,
while the mechanical side holds the stiffness of the crystal sK and the stiffness of the bridge
pieces sR. The bridge piece is a plastic piece shaped like a pitched roof; it sits on top of the
crystal and represents the connection to the string. The stiffness of the bridge piece is smaller
than that of the crystal by about three orders of magnitude. Due to this relationship, changes
in the electrical load (e.g. shorts) cannot be measured at the mechanical pickup-input (at F) –
these changes do cause a difference in the input impedance of the transducer (at F2), but they
are completely insignificant relative to sR: when connecting two spring in series, the softer
one dominates. For the same reason, changes in the mechanical loading (e.g. the mass of the
shaker to be connected at the far right in Fig. 6.8) cannot be measured on the electrical side:
the very stiff spring sK dominates since in a parallel arrangement of two springs, the softer
spring has little impact on the overall stiffness.
Abb. 6.8: Equivalent circuit diagram of the piezo pickup (Ovation EA-68).
However, with the measured transmission coefficient TUF = U/F = 0,2 V/N, and with the
pickup capacitance C = 1,45 nF, we have merely two conditions for the three variables sK, α,
and CK at our disposal; only sR is fully defined by the resonance at 3,9 kHz and the mass of
the measuring head. Sind there is no supplementary condition available without invasive (and
therefore undesirable) action, the ratio was arbitrarily taken to be 50%.
Given this, we can nevertheless define input and output impedance, as well as the transfer
function within the framework of the limits of the model – uncertainty remains only when
calculating back to the material parameters. Such an uncertainty, however, exists anyway,
since the distributions in space of the mechanical tensions in the piezo and in the bridge
pieces is unknown.
The following calculations are based on the assumption that there is no single crystal strip
across the whole bridge, but that there is a small, square crystal plate beneath each string,
connected with its neighbor by two contact wires. As a mechanical excitation of one bridge
piece occurs, only one single crystal plate will generate an electrical signal, and the other five
crystal plates have the effect of an electric load. For such a string-specific crystal plate, the
calculation yields:
α = 0,28 N/V, CK = 161 pF, sK = 9,7·108 N/m, sR = 7,6·105 N/m.
Using these data, the model proposed in Fig. 6.8 can explain the following measured
quantities: the electrical impedance (as a pure capacitance), the mechanical impedance (as a
pure spring), the impedance-behavior shown in Fig. 6.5 as a mass is set onto the piezo), and
the frequency-independent transmission behavior. The resonance peaks seen in Fig. 6.7 are
not modeled. We will see how resilient this model is as the direction of the signal flow is
reversed, i.e. as the sensor is turned into an actor when we apply an electrical voltage. For this
case, the above model (using the same parameters) needs to be able to explain the
measurement results (Chapter 6.5).

6-8 6. Piezo pickup
6.4 Reciprocity
The theory of two-port networks describes the relationships between the input- and output-
quantities of linear, time-independent systems using two-port matrices. For the electrical
two-port system, the input signals applied to the input connectors (port 1) are input voltage U1
and input current I1; correspondingly, output voltage U2 and output current I2 are found at the
output connectors. Connecting these four quantities, two-port equations may be defined:
Impedance matrix Z
Admittance matrix Y
Hybrid matrix H
Inverse hybrid matrix G
Fig. 6.9: Electrical two-port with technical

directions of the reference arrows. In the
general case, all quantities are complex.
Besides these four matrices, we may also define the chain matrix (A) and the inverse chain
matrix (B); theses are, however, not required here. Each matrix fully describes the two-port,
and the elements of one matrix may be recalculated from the elements of every other matrix.
In the general case, the four matrix elements are independent of each other. However, for
passive RLCT-two-ports (consisting exclusively of resistors, inductors, capacitors and
transformers), only three degrees of freedom remain [e.g. 7], so that two matrix elements are
dependent on each other in a simple fashion. Such two-port networks are called transmission-
symmetric or reciprocal. Using so-called technical♣ reference arrows [7], the following holds
for reciprocal two-ports:
, , , Conditions of reciprocity
Plugging these conditions into the two-port equations and generating, with U = 0 or I = 0, a
rogue starting state, simple relationships between the operation in the forward and the reverse
directions result.
As an example: The retroaction H12 = U1 / U2 identified for primary open circuit (I1 = 0) corresponds to
the current amplification H21 = I2 / I1 established for secondary short circuit (U2 = 0).
♣
Besides the technical reference-arrow system, applying the symmetric reference-arrow system would be just
as justifiable; in that case, the direction of the output current would be reversed, and in the conditions of
reciprocity, all signs would be reversed. The symmetric reference-arrow system is not used in this book.

6.4 Reciprocity 6-9
In more general terms: knowledge of the transmission behavior in one direction (e.g. from
pole-pair 1 to pole-pair 2, or from port 1 to port 2) enables us to determine the transmission
behavior into the reverse direction (from port 2 to port 1).
A pickup is not a purely electrical two-port, but this does not stand in the way of a
consideration of reciprocity: given the transducer equations and , the
mechanical signal quantities may be recalculated into the electrical signal quantities, and
using Zel = Zmech / α2, the system quantities may be recalculated [3], such that the
electromechanical transducer two-port is changed into a purely electrical reciprocal two-port.
As a rogue connector-loading, v = 0 and F = 0 could theoretically be defined, with only F = 0
being practically significant; the total immobility is not obtainable precisely enough.
Of the four conditions of reciprocity, the equality of the G-parameters is best suitable to
calculate the piezo pickup. Defining the connector pair designated with 1 as electrical side,
and the pair designated with 2 as the mechanical side, the boundary conditions are: from U1 =
0, F = 0 results (no external force, bridge pieces without load), and I2 = 0 implies an electrical
open circuit (the usual piezo-connection). Using the reciprocity-conditions, the
force→voltage-transmission we are looking for can be deduced from the voltage→velocity-
transmission (measureable via Laser-vibrometer).
As an example for the reciprocity conditions, let us examine an ideal lever to which a spring
and a dampening resistance are mounted (Fig. 6.10). The reference system used here is an
option – other reference systems are possible, as well.
ideal spring (Hooke)
ideal friction (Stokes)
small deflections ≈ linear operation.
Fig. 6.10: Mechanical 2-port purely as example;

no direct bearing to the piezo pickup.
The boundary conditions relating to are reformulated (using and

) into a first operational state with mechanical open-circuit (F1 = 0), and into a second
operational state with fixed output (v2 = 0). This yields:
for F1 = 0 and for v2 = 0
In the 1st case the excitation happens at the right-hand connector (index 2), and the velocity-
ratio is determined; in the 2nd case excitation occurs at the left-hand connector, and the force-
ratio is determined. In both cases the same transmission function results – but only for the
specified load conditions, and not for all load-types.
The system specified above can be reformulated as an equivalent electrical system using the
algorithms of electromechanical analogies [e.g. 3]. From the possible choices FI- and FU-
analogies, we select the latter because it corresponds to the physical transducer principles
present in a piezo transducer. For the network-structure, however, we need to consider that

6-10 6. Piezo pickup
not an isomorphic but a duality-network results (the flow-quantity force is mapped to the
potential-quantity voltage [3]). The FU-analogy converts a spring into a capacitance, a
friction-resistance into an electrical resistor, and a lever into an ideal (inductance-free)
transformer that can also transmit DC. For a lever, the ratio of the lever arms (and thus v1/v2)
is stated; a transformer, however, has – corresponding to the duality – a voltage ratio, and thus
the transformation ratio is inverted (Fig. 6.11).
Fig. 6.11: Electrical equivalent circuit diagram derived

via FU-analogy from Fig. 6.10. Equal velocity results in
equal current, i.e. series connection; equal force results in
equal voltage, i.e. parallel connection (dual network)
For the two load conditions U1 = 0 and I2 = 0, a transmission function can be defined for the
above; the same result is obtained with the conversion C = α2/s and R = W/α2 :
for U1 = 0 and for I2 = 0 ◊
Contrary to a purely mechanical system (Fig. 6.10) or a purely electrical system (Fig. 6.11),
the piezo pickup represents an electromechanical system. For such a system, the relations of
reciprocity hold, too. In the ideal transducer (Fig. 6.12), we immediately note that the
transmission factors are the same: for primary open-circuit (F1 = 0), we have I2 / v1 = α, and
for secondary open-circuit, the transducer factor F1 / U2 = α results. This sameness even holds
for every load condition in the ideal transducer – and therefore naturally also for the boundary
conditions of the reciprocity-relations.
Fig. 6.12: Ideal piezo transducer
As opposed to an ideal piezo transducer, the real piezo transducer includes mechanical and
electrical components that are to be connected as a mechanical two-port on the left side and as
an electrical two-port on the right side. The two-ports are of reciprocal character as far as they
merely include masses, springs, friction-resistances and levers, or inductances, capacitances,
resistors and transformers, respectively, and the overall system is then reciprocal, as well. A
corresponding two-port ladder-network is shown in Fig. 6.13 – it may, of course, again be
consolidated into a single two-port. Given the boundary conditions mentioned above, we
obtain, for F1 / U2 and I2 / v1, the same coefficient of proportionality. However, the latter does
not correspond anymore to the transducer constant α, but depends (in a possibly complicated
fashion) on the frequency. This dependency can be measured relatively easily in the I2→v1-
operation, and may be carried over to the F1→U2-operation (that is more difficult to measure).
Fig. 6.13: Real piezo transducer

6.4 Reciprocity 6-11
For the I2→v1-operation, a generator with a low-impedance AC-output is connected to the

electrical input of the pickup. Since the pickup represents (in good approximation) a purely
capacitive electric load independent of its mechanical loading, it is easy to make the
connection from the electrical voltage U2 to the current I2 (I2 = jωCU2). This
current→velocity transmission-coefficient TvI corresponds to the force→voltage transmission-
coefficient TUF for the electrical open-circuit:
TUF for I2 = 0, TvI for F1 = 0
The oscillation-velocity v1 can be determined e.g. with a Laser-vibrometer; due to the small
values to be measured, suitable averaging is mandatory.
6.5 Operation as an actor
Piezo-electric materials convert in both directions: mechanical quantities into electrical ones
(operation as sensor), and electrical quantities into mechanical ones (operation as actor). As
an electric AC-voltage is connected to the electrical connectors of the pickup, the bridge piece
vibrates up and down … a bit. A very small bit, actually: merely a few nanometers. We could
not find out at which voltage the crystal was going to receive irreversible damage, and
therefore the following measurements were carried out with a RMS-voltage of 10 V – no
recognizable damage occurred there. During the measurement, the Ovation guitar was placed
in its case, and the vibration velocity was measured using a laser-vibrometer (Polytec).
Based on the equivalent circuit diagram shown in Fig. 6.8, we would expect, for a mass-free
bridge piece (idealization), a frequency-independent displacement, if a frequency independent
voltage is imprinted. However, the vibrometer – based on the Doppler effect – measures the
vibration velocity as its source-quantity, and therefore the measurement grows more difficult
with decreasing frequency. Nevertheless, using sufficiently narrow-band filters makes
coherent results possible also in the low frequency domain (Fig. 6.14). Both the actor- and the
sensor-measurements show, as a 1st-order approximation, a frequency-independent
transmission factor, although there are smallish frequency peaks – these are mainly caused by
the guitar and not by the measuring process.
Fig. 6.14: Transmission factor GxU as measured with the laser-vibrometer (left). For comparison, the
corresponding sensor-measurement is shown on the right: TxU = C⋅TUF. The correspondence is impressive.
“Frequenz” = frequency; “Aktor” = actor.

The actor-measurement gives us a voltage→deflection transmission coefficient of 0,29 nm/V.

For the selected generator voltage (10 Veff), this implies a deflection of 2,9 nm, and a velocity
of only 1,8 µm/s at 100 Hz, and it also means that the laser-vibrometer generates as little as
0,36 mV (for 0,2 Vs/mm). This small voltage of the signal is clearly below the intrinsic noise
of the laser-vibrometer, and an exact measurement requires deployment of a frequency-
selective tracking-filter (Fig. 6.15). Alternatively to the process applied here (Hilbert-
transform, complex quotient, block-averaging), we could also calculate the complex quotient
of two DFT-spectra – however, the path via the Hilbert-transform is more conducive for the
logarithmic frequency axis we employ.
Fig. 6.15: Measured velocity w/out (---) and with tracking filter (left); filter characteristic at 100 Hz (right).
The comparison of operations as sensor and actor (Fig. 6.14) indicates that the mass-
compensation elaborated in Chapter 6.3 delivers exact results, and that our modeling of the
transducer is suitable as a first approximation. The noise-effects appearing in the low-
frequency range with the laser measurement can easily by reduced by extending the
averaging-window – thus we have at our disposal two equivalent processes for f < 3 kHz. In
the higher frequency range, the two approaches show smallish differences that can be
attributed to an imperfect mass-compensation, and to small differences in the measurement
position.
Many of the high-frequency resonance peaks can be traced to structural resonances of the
roof-shaped string pieces and the strings. Still, the lower side of the pickup merits some
consideration because it forms the reference system relative to which the upper side of the
piezo (oriented towards the string) vibrates. The lower side of the pickup is formed from a u-
shaped aluminum rail laid on top of shims (made from Pertinax) that are placed on the routed-
out guitar body. At least this is the situation with the factory-fitted instrument. For the
measurements described above, strips of corrugated paper replaced the shims; any misgivings
that the absorption would be unduly increased were in fact unfounded. On the contrary, the
original shims resulted in a small dip in the frequency response of the transmission at around
5 kHz that could be attributed to a resonance in the u-rail. It appears the strips of corrugated
paper made for a better contact and therefore a better dampening of this resonance. Note,
though, that axiom “the sound is in the ear of the beholder” does always hold. In its left-hand
section, Fig. 6.16 shows the transmission factor GxU for actor-operation with the original
shims, and on the right the displacement of the u-rail measured for the same drive-level. We
clearly see that the u-rail starts to vibrate more strongly in the range around 5 kHz. The
special shape of these vibrations was not further investigated since the effort would have been
unreasonable.

6.5 Operation as actor 6-13
Fig. 6.16: Transmission factor incl. the Pertinax-shims: bridge pieces (left), u-rail (right). The piezo-pickup was
driven from a low-impedance source for both measurements. We see a resonance of the guitar top around 700
Hz,; around 5 kHz and 15 kHz, a resonance of the u-rail carrying the piezo crystal occurs.
“Frequenz” = frequency; “Aktor” = actor.
Besides the solid-body Viper (EA-68, a rather uncommon design for an Ovation guitar ), a
more typical steel string acoustic (Adamas SMT) was also analyzed. The “mid-depth bowl”-
designated Lyrachord-body carries a laminated top of birch and carbon-fiber layers; a piezo-
pickup is built into the bridge. Fig. 6.17 shows the comparison between sensor- and actor-
operation; again there is good correspondence. Measuring in the actor-operation proved to be
somewhat more difficult than for the Ovation EA-68, because the vibration-happy top was
excited by ambient noise – especially in the range of the 160-Hz-resonance.
Fig. 6.17: Transmission factor of the Adamas SMT. Actor-measurement (left), sensor-measurement (right).
C = 450pF. “Frequenz” = frequency; “Aktor” = actor
Compared to the Viper, two significant differences show up: the capacitance of the SMT-
pickup only amounts to 450 pF (re. 1,45 nF for the Viper), and the transmission behavior of
the SMT includes a boost in the range of the highest octave. Operating the pickup in
conjunction with the FET-preamp built into the respective guitar, a small difference is also
revealed in the low-frequency range: the Viper-piezo sees a load of 500 kΩ resulting in a
high-pass with a cutoff-frequency of 220 Hz. Conversely, the SMT-piezo is connected to a
preamp with an input-impedance of 2 MΩ, yielding a lower cutoff-frequency of 177 Hz. The
above measurements do, however, not show any high-pass behavior: this does not occur in the
actor-operation as a matter of principle, and it was computed out for the sensor-operation.

6.6 The pickup in isolation
Its internal build does not exclusively determine the transmission behavior of the piezo pickup
– the surface it is positioned on also weighs in. Measuring the pickup without an abutting
surface, i.e. without guitar body, therefore represents an obvious step. We must, however, not
expect to find an ideal, frequency-independent transmission factor. Rather, the contrary will
be the case: in isolated condition, strong Eigen-vibrations may possibly happen that only
receive attenuation by mounting the pickup. Fig. 6.18 depicts laser-measurements of the
Viper-pickup (Ovation EA-68) taken out of the guitar and clamped in a bench vise in two
different ways.
Abb. 6.18: Transmission factor of the Viper-pickup taken out of the guitar and clamped in a bench vise.
“Ausgebaut” = de-mounted i.e. in isolation.
This type of mounting is not very meaningful and difficult to reproduce. As an alternative, the
pickup was laid on top of a fist-sized brass-block and firmly braced with 2 to 6 steel wires
(Fig. 6.19). Depending on the contact between u-rail and brass-block, we obtain a big variety
of transmission characteristics – and therefore the measurement of the pickup mounted to the
guitar is indeed the most suitable one.
Fig. 6.19: Transmission factor of the Viper-pickup taken from the guitar and braced to a brass-block.
“ausgebaut” = de-mounted i.e. in isolation. “Frequenz” = frequency
A curiosity on the side: the u-rail had a paper label (type,

serial number) glued to its lower side. Due to this, part of the
rail had no mechanical contact to the base anymore. Genius or
ignorance? Fig. 6.20: Viper-pickup

6.7 Noise 6-15
6.7 Noise
For every piezo pickup cooperating with a battery-powered preamplifier, we encounter

conflicting goals between noise interference and life-time of the battery: the smaller the drain-
current of the involved field-effect transistor, the longer the battery keeps – but the higher also
the noise-voltage density en, and thus the preamp noise. However, as a rule we are not sailing
in very critical waters here, and we can opt in favor of the staying power of the battery.
Fig. 6.21: Schematic and gain of an early Ovation-preamplifier. The adjustable 250-kΩ-resistor serves as tone
control, at the output we find a volume potentiometer of 100-kΩ (not included in the drawing).
Fig. 6.21 shows the schematic of a battery-powered FET-amplifier as it was deployed in the
first-generation Ovation guitars. Given a quiescent current of about 0,1 mA, a battery lifetime
of about 4000 h may be expected (for Alkaline batteries). The two resonance circuits
connected to the source cause an attenuation of the mids that – according to the manufacturer
– influence the sound advantageously. Since in this configuration, the FET reaches a voltage-
gain factor of about 20 and could easily be overdriven, the signal from the piezo needs to be
correspondingly reduced by a voltage divider at the input. In the 2-kHz-frequency-range that
is important to our hearing, we therefore find an effective resistance of 67 kΩ at the FET-
input that causes – with 33 – so much noise that the FET-noise itself may be
disregarded. The multiplication of the noise-voltage density mentioned above with the square-
root of the bandwidth of the 1/3rd -octave at 2 kHz, and with the voltage gain gives us a noise
voltage of 14 µV at the circuit output – that ain’t really much of a low-noise design, but it’s
not that bad either given a maximum obtainable signal voltage of 1 – 2 V.
Later Ovation preamps distinguish themselves from the model discussed above by a
somewhat lower noise, and by distinctly higher current consumption. For example, the
quiescent current of the Ovation Viper described in the preceding chapters amounts to about
1,2 mA, and the SMT even draws as much as 4,6 mA from the 9V-battery. On the other hand,
we should not be silent about the versatile equalizer built into these models that also requires
power. If we want to look into the data of integrated operational amplifiers (OP), we need to
direct our attention first to their voltage supply: typically, an OP is operated with ±15 V;
battery-operation with ±4,5 V or even as low as ±3,5 V is often but not always possible.
Moreover it needs to be noted that some OP-data deteriorate relative to the datasheets if the
supply voltage is reduced.

In order to calculate the amplifier-noise it is conducive to transform all noise sources to the
input of the amplifier and to indicate them as so-called equivalent input-noise sources. In this
respect, datasheets for OPs specify the noise voltage density en and the noise current
density in. A straightforward equivalent-circuit for noise (Fig. 6.22) considers the piezo-
capacitance C (1,5 nF), the input impedance R (1 MΩ) of the amplifier, the noise voltage
density en (42 ) of the amplifier, and the noise current density in (10 ) of the
amplifier. For this first example, the noise densities were taken from the datasheet of a FET-
OP (TL061) that is suitable for battery operation (current drain: 0,25 mA).
Fig. 6.22: Equivalent circuit for noise of a piezo pickup with connected amplifier.
As an approximation, all noise processes are statistically independent, and their effects need
to by added using a Pythagorean summation. Relative to the noise current density iR caused by
a resistor R, the noise current density in of a typical FET-OP is insignificant. In conjunction
with R and iR, the capacitance C forms a low-pass with a cutoff frequency at around 150 Hz.
Therefore, the capacitance of the piezo pickup increasingly shorts the resistor noise toward
higher frequencies. The noise generated in the kHz-range that is significant for our hearing is
predominantly caused by en. For the TL071, the resistor noise (iR ) dominates in the low-
frequency range, while for the TL061, the OP itself makes a considerable contribution, as
well (flicker noise).
Fig. 6.23: 1/3rd-octave spectrum (= “Terzspektrum”) of the piezo-preamp output. 0 dBµ ⇒ 1 µV. Ovation-Viper
(left), Ovation-SMT (right). Voltage gain factor = 0,5. Ovation: measurement; FET-OP: calculation.
Fig. 6.23 depicts measurement and calculation in comparison: the TL061 is well suitable, and
with 0,25 mA supply current, a battery life of 1600 h should be possible. Even at only 7 V
supply voltage, the output can deliver 1,7 Veff. The TL071 requires, at 2 mA, clearly more
current, but it is less noisy in the high-frequency range. Significantly less noise if generated
by the AD743, but this OP draws 10 mA and should be operated from a mains power supply.
The degree of suffering that we encounter here, however, is not that terrible that it would push
us too much towards using especially low-noise amps: when playing loudly, the pickup
delivers about 1 V, and therefore an adequate signal-to-noise ration is reachable even with the
Ovation preamp – in particular since the guitar itself generates interference noise, too.

6.8 Piezotonabnehmer vs. Mikrofon 6-17
6.8 Piezo pickup vs. microphone
An acoustic guitar fitted with a piezo pickup offers two possibilities for recording: airborne
sound via a microphone, and structure-borne sound via the piezo. Since the sound radiated by
the guitar body is not particularly loud, a recoding microphone needs to be placed as close as
possible to the guitar – which a) obliges the guitarist to maintain substantial “positional
discipline”, and b) generates the permanent fear that in a moment of negligence he/she might
ram that one-of-a-kind pre-war Adirondack-fir-top into the mike. Indeed, these shortcomings
led to the development of the piezo pickup in the first place: a pickup enabling the player of
an acoustic to move around. It sounds differently, though. The preceding chapters have shown
that the piezo pickup can deliver the full frequency range of the human hearing with good
quality. Since non-linearities (harmonic distortion) and noise do not show up as quality-
degrading factors, either, the piezo pickup in principle allows for a sound recording of very
high quality. Nevertheless, clearly noticeable differences compared to the radiated airborne-
sound remain – this is due to the lack of the transmission filters (formed by the guitar body
and neck). The piezo pickup converts the alternating force fed (perpendicularly to the guitar
top) to the bridge into an electrical voltage. Conversely, the microphone converts the airborne
sound radiated by the guitar top (and the neck, bottom and sides) into an electrical voltage,
and therefore measures a different quantity. The latter also does have its source in the force at
the bridge (amongst other forces), but depends on it via highly complicated functions.
In order to capture the differences between piezo- and microphone-recording in a practice-

oriented manner, an Ovation guitar (Adamas SMT) was recorded in the anechoic chamber
using a measuring microphone (B&K 4190), and in parallel the piezo-signal was recorded
without any filtering. The free-field-equalized microphone generates an objective reference,
independent of any treble boost as it is common in studio-work. There are myriad possibilities
to position the microphone – not all were tried. For stage-work, often a microphone is pointed
to the rear section of the guitar top, supplemented by a second microphone aiming at the
neck/body-transition. Arrows mark these positions in Fig 6.24; in addition, the
correspondingly measured 1/3rd-octave spectra are shown (1/3rd-octave wide filters, measured
with overlap at 1/6th-octave distance). A chord-playing guitarist activated the guitar-strings. In
both spectra, we recognize a strong emphasis in the frequency range below 500 Hz; this
corresponds to the auditory impression. For music productions, an equalizer would be called
in to perform suitable attenuation, but for the measurements, filtering was dispensed with.
Fig. 6.24: Measured 1/3rd-octave spectra (SPL re. 20 µPa) and microphone positions (at 12 cm distance).
Arbitrarily fingered chords, strings struck with a pick in quick succession. “Mikro“ = microphone (“mike”),
“Frequenz” = frequency.

Fig. 6.25 shows the 1/3rd-octvae spectrum of the piezo signal that is comparable to Fig. 6.24.
It emphasizes the mids but is, in its even curve, well suited for further processing. The right-
hand section of the figure highlights the difference between microphone- and piezo-recording:
between 100 Hz and 500 Hz, resonances of the guitar top generate a strong emphasis in the
microphone-spectrum, and around 700 Hz we find a minimum in the radiation which is,
according to statements by manufacturers, desirable.
Fig. 6.25: 1/3rd-octave spectra of the piezo signal (left); 0 dB arbitrarily set for a curve comparable to the one in
Fig. 6.24. Right: difference “mike-spectrum minus piezo-spectrum” (2 mike positions). “Micro” = mike
Differences in the 1/3rd-octave levels of ±15 dB indicate considerable differences in sound –

listening tests immediately confirm this. However, as soon as we start to filter one of the two
signals with an equalizer such that one spectrum approaches the other, the audible differences
become weaker, and limit themselves to minor effects that might possibly also disappear if we
could only edit the EQ-curve subtly enough. In fact, as long as we model the guitar as a
linear, time-invariant system, this should indeed (better) be the case: the two transmission
functions from source (string) to destination (piezo and microphone) are each describable by a
fractionally rational function that an equalizer can emulate at least approximately. Non-linear
effects are certainly also involved in the generation of sound – but in the investigated
example, their impact obviously was rather small.
Consequently, the signals from piezo and microphone are similar to a high degree, and can be,
respectively, converted into each other via (a not entirely trivial) filtering. This does not deny
that significant portions of the sound energy may be fed to the neck by the nut (or the frets),
and radiated there or after transmission to the guitar body – something the pickup integrated
into the bridge would not sense. Still, both the sound-flux through the nut/frets and through
the bridge have the string vibration as their common source, and are mathematically
connected via a fractionally rational function. Conversely, the direction of the string-vibration
has a more complicated effect: if a fretboard-normal 110-Hz-eigen-oscillation and a fretboard-
parallel 111-Hz-eigen-oscillation jointly act onto the bridge, they might both lead to a
radiation of sound. The piezo in the bridge will mainly capture the fretboard-normal
component, though. In theory, one signal could again be derived from the other by filtering,
but the required Q-factors of the filter stand in the way of a realization. Presumably, one of
the main reasons for remaining differences is found here – but the latter are indeed weak.
In summary: neither the piezo-signal nor the microphone signal would be mixed down in the
studio in their raw version; given suitable filtering, however, both form a good basis for
further processing. Whether one chooses the piezo signal or that from the microphone, or uses
a mix of both – that decision needs to be taken based on artistic reasoning.

6.9 Mikrofonie 6-19
6.9 Microphonics
Piezo pickups do not only react to vibrations caused be the guitar strings (desirable), but also
to vibrations generated by the airborne sound impinging on the guitar (not desired). A sound-
wave arriving at the top of the guitar sets the bridge (and thus also the pickup) in motion.
Since due to its mass- and spring-loading, the small piezo plate will not join in with this
movement in an identical fashion, and a change in thickness generating a corresponding
piezo-voltage will result. If the latter is amplified and reproduced by a loudspeaker positioned
close-by, a loud feedback-howl might happen. The following investigations will target the
quantitative description of this sensitivity to airborne sound (termed “microphonics”).
In systems theory, the term feedback designates a signal path back to the input; this may be
in-phase (positive feedback) or with opposite phase (negative feedback). In the studio- or
stage-environment, the term “feedback” usually indicates that the feedback loop has already
reached self-excitation, and oscillation occurs in the form of howling or whistling noises.
Self-excitation requires a value of equal to or larger than one for the magnitude of the loop
gain, and a phase of 0°. For example: a microphone generates a voltage of 50 mV at an SPL
of 1 Pa, and a loudspeaker fed from this microphone generates, at the location of the
microphone, an SPL of 1 Pa from an operation with 10 V input voltage. If the microphone
voltage is amplified by a factor of 200, feedback howling could start (given a matching phase
shift).
In practice, reaching self-excitation depends on many details: the directivity of microphone

and loudspeaker, the filters, the transmission function of the room. With the Ovation
Adamas-SMT guitar as test-object, the following will exemplarily illustrate the difference in
feedback-sensitivity between operation with a microphone, and operation with the pickup.
Other guitars, other speakers and other rooms will lead to similar but of course not identical
scenarios.
If an acoustic guitar is to be captured and amplified live via a microphone, we will try to
position the microphone as close as possible to the guitar. In this setting, the top of the guitar
will act as a reflector that may, depending on the circumstances, direct the impinging sound in
an unfavorable manner to the microphone. Fig. 6.26 depicts a setup as it was put together in
the anechoic chamber for orientation measurements. A loudspeaker radiated sound to a guitar
that had a measuring microphone (B&K 4190) set up in front of it. The right hand graph
shows the comb-filer effects caused by the reflections.
Fig. 6.26: Experimental setup in the anechoic chamber; comb-filter-effect due to reflections. “Kammfilter” =
comb-filter”, “Frequenz” = frequency.

At low frequencies, the dimensions of the guitar are small compared to the wavelength, and
the reflections are insignificant. Around 200 Hz, the guitar acts as an absorber – and at mid-
range frequencies we see the typical comb-filter peaks. Due to the irregular shape of the
guitar, the diffraction waves have slightly different delay times, with the result being that at
high frequencies we find an even curve. The increase in the two highest octaves is due to the
microphone directionality: nominally, the 4190 is omni-directional, but at high frequencies it
does feature increasing beaming – as do all ½"-microphones. With studio-microphones
having stronger directionality, the corresponding effects will show up already in the mid-
frequency range. Such microphones may be able to attenuate the sound coming from the
loudspeaker, but not the sound reflected by the guitar. This example is meant to highlight that
the guitar as reflector can distinctly deteriorate the resilience against feedback. The degree of
this deterioration (and its frequency-dependency) will be a function of the given microphone-
distance and –directionality, and of the specific guitar- and microphone-geometry.
To arrive at a statement regarding the sensitivity to feedback in the piezo pickup, we first
need to take care that the microphone and the piezo lead to the same sound. For this reason,
the microphone signal was filtered such that the same transmission function was achieved
with both microphone and piezo. In a second step, we introduced a slight mid-cut, combined
with a slight treble boost – adjusted according to artistic criteria. Both via microphone, and
via piezo, the sound of the guitar corresponded to the taste of the guitarist. Now the guitar was
subjected to the sound emitted by the loudspeaker according to Fig. 6.26, and the loop gain
was measured (Fig. 6.27). The origin of the ordinate was assigned to the maximum of the
airborne sound.
Fig. 6.27: loop-gain for microphone-use (---) and for

piezo-use (–––). Both transmission paths equalized
for the same sound. “Frequenz” = frequency
Fig. 6.27 shows a high loop-gain in the bass range when using the microphone pickup – this
will quickly lead to howling feedback as the volume-control is turned up. Clearly, the filtering
used here is unsuitable for live-conditions, and it will be mandatory to attenuate the bass –
and make do with a more “slender” sound. Not that the latter would be unusable; it is just less
full than the sound heard from the unamplified guitar. Compared to using the microphone
with the attenuated bass, the piezo offers – in this example – an advantage of 20 – 25 dB. This
is the degree to which the level generated by the loudspeaker can be increased before
feedback occurs. As already mentioned, this value is for orientation – in every individual
case, many parameters will determine the onset of feedback. In any case, however, the
differences are so pronounced that a piezo bridge-pickup will always drastically reduce the
danger of feedback.

6.10 Unterschiede zum Magnet-Tonabnehmer 6-21
6.10 Differences to the magnetic pickup
Compared to guitars that use one or more magnetic pickups as transducers for the sound,
guitars fitted with a piezo pickup sound different (when operated with an amplifier). There are
mainly two reasons for this: the transmission range of the pickup, and its position on the
guitar. The lower cutoff-frequency of the transmission is sufficiently low for both transducer-
types; the upper limit, however, differs: it is 20 kHz for the piezo pickup but only 2 – 5 kHz
for your regular magnetic pickup. It would not be a particular problem to radically change the
treble response: without any capacitive loading, the magnetic pickup can deliver signals up to
15 kHz as well, and conversely an increase in the mass of the bridge for the piezo pickup can
reduce its upper cutoff frequency. However, the piezo pickup typically will “give more treble”
than the magnetic pickup.
The magnetic pickup is positioned about 3 – 15 cm away from the bridge – mounting it
within the bridge is disadvantageous because here the string-velocity that needs to be captured
is almost zero. The piezo pickup, on the other hand, can only be mounted in the bridge (if we
disregard transducers stuck to the guitar top as they have become less important). According
to the theory of linear time-invariant systems [6], every signal exciting the strings can be
interpreted as the sum of super-positioned impulses. Each of these impulses runs along the
strings as a wave (Chapter 2) and is reflected at the bridge and at the nut (or fret). Within one
round (1 period) it therefore passes the position of the magnetic pickup twice. The delay time
between both passes results from twice the distance between pickup and bridge, divided by
the phase-velocity of the transversal waves♣. Since the transversal wave captured by the
magnetic pickup is reflected at the bridge with opposite phase, the double-sampling acts like a
filtering with sine-magnitude frequency response (comb-filter):
Comb-filter
In this equation, τ stands of the (single) travel time between pickup and bridge. Disregarding
any dispersion-effects, the zeroes of the comb-filter lie at the integer multiples of f0 = fG ⋅M/d,
with fG = fundamental frequency of the string, M = scale, d = distance between pickup and
bridge (Fig. 6.28). Due to the dispersive propagation of the transversal waves, we get a
spreading of the zeroes towards high frequencies (Chapter 1.3).
Fig. 6.28: Normalized comb-filter frequency responses. E2-string (–––), E4-string (----); d = 5 cm, M = 65 cm.
“Frequenz” = frequency.
♣
We could also use twice the distance between pickup and nut/fret here; the results would be different at first
but can be reformulated into an equivalent model..

Because misunderstandings can easily happen when we term the transmission coefficient of a
magnetic pickup as “comb-filter-like”, let’s be precise: the transmission coefficient of
velocity→voltage is of a low-pass characteristic if “velocity” indicates the local string
velocity over the magnet of the pickup. With respect to the velocity of the transversal wave,
the mentioned comb-filter comes into play in addition to the low-pass characteristic (low-pass
and comb-filter serially connected), if one oscillation period is observed as the timeframe. For
the steady state (very long time-window, no dampening), this frequency-continuous
transmission function is to be sampled at the locations of frequencies of the partials
(frequency-discretization).
For the piezo pickup, it is force at the bridge rather than the velocity that forms the input
signal to the transducer – though of course for both the initial source is the wave travelling on
the string. To compare with the magnetic pickup, it is conducive to specify the same input
signal for both transducers, for example the (particle) velocity of the transversal wave. Since
the wave-impedance is real, we have proportionality between the velocity of the propagating
wave and the respective force; and because the force-wave is reflected at the bearing with the
same phase, the bearing-force is also proportional to the velocity of the transversal wave. We
arrive at the following conclusion: starting from the transversal-wave velocity (the particle
velocity of a propagating transversal wave), the piezo pickup mounted in the bridge
practically transmits independently of the frequency. If the electrical load-impedance requires
it, we may need to consider a high-pass with a cutoff-frequency of about 200 Hz in some
circumstances (Chapter 6.5). Conversely, the magnetic pickup will generate string-specific
comb-filtering. The further the pickup is located away from the bridge, the closer the
“prongs” of the comb-filter will be (Fig. 6.29).
Abb. 6.29: Normalized frequency responses of the comb-filter. E2-string (top), E4-string (bottom); M = 65 cm.
d = 5 cm (left), d = 15 cm (right). “Frequenz” = frequency.

Supplement: equations of state 6-23
Supplement to Chapter 6: piezo-electric equations of state
The description of piezo-electric energy conversion uses generally accepted formula symbols
that, however, often have a different implication in other frameworks. The symbol S, for
example, elsewhere often stands for a surface area; for the piezo crystal, however, it designates
the relative deformation (strain). The letter E finds much use as indicator for field strength, but
also represents energy, or the modulus of elasticity. The letter d may serve to designate a
diameter, or a thickness – or the piezo-modulus. The following table lists the formula symbols
required to describe piezo-electric transducers. In order to avoid ambiguities, the definition of
some of these quantities is limited to the present special chapter.
Symbol Unit Designation Symbol Unit Designation

A m2 (Surface-) Area L H Inductance
C F Capacitance m kg Mass
d m/V Piezo-modulus Q As Electrical charge
D As / m2 Dielectric displacement R Ω Electrical resistance
E D 2
e N / Vm Piezo-force constant s ,s m /N Elasticity-coefficient♣
E , ED
E
N / m2 Modulus of elasticity♣ S 1 Relative deformation
E V/m Electric fields-strength t s Time
F N Force T N / m2 Mechanical stress
g Vm / N Piezo-voltage constant U V Electrical voltage
h V/m - W Ws Energy
k 1 Coupling factor α N/V Transducer constant
l m Length ε , εT
S
F/m Dielectric constant♣
In the indexing, the three coordinates in space (x, y, z) are replaced by the numbers 1, 2, 3 –
in agreement with common convention, the direction perpendicular to the vibrating surface of
a thickness-mode oscillating block is indexed with the index 3 (Fig. 6A.1). The thickness-
mode oscillator is the transducer type most often found in pickups; in the following only this
type will be discussed. Designated here with l, the thickness typically amounts 0.2 … 1 mm in
many cases, while the surface A will be about 0,1 … 1 cm2. Except for the thickness-mode
oscillator, there are also flexural oscillators, planary-mode oscillators, shear-mode oscillators,
and more types. The thickness-mode oscillator is sometimes also called longitudinal
oscillator, or thickness-oscillator with longitudinal effect – the terms are not consistent.
Fig. 6A.1: Piezo-crystal with directional definitions.

Top and bottom are metalized and have the surface A
each; between the surfaces the electrical field strength
is formed. The height (thickness) of the crystal is l. For
the thickness-mode oscillator shown here, the
movement (and force) occurs in the vertical direction
indexed with 3. The electrical fields run
correspondingly in parallel.
♣
The superscript letters designate special load conditions – this will be elaborated on later.
© M. Zollner 2006
The piezo-crystal shown in Fig. 6A.1 is subjected to the force F3 in the vertical direction, and
a vertical mechanical stress T3 = F3 / A results. This, however, is the first particular case that
needs to be specially designated: only one single mechanical loading is to be present –
without any electrical loading. The electrical voltage therefore needs to be set to zero via a
short across the electrodes (the metalized connecting surfaces). Because, due to this short, the
electrical field strength is zero, this special condition is designated with a superscript E –
which must not be seen as a mathematical power. Electrically shorted, the crystal reacts as if
the piezo-effect did not exist, i.e. for static loading the crystal merely reacts like a spring:
Hooke’s law
In this equation, S3 designates the relative change in length often termed ε in mechanics.
However, since ε is also the symbol for the dielectric constant (that is required here, as well),
we term – in electro-mechanics – the relative length-change with S. Specifically, we term it
with S3 because it is directed vertically in the direction indexed with 3. The mechanical stress
occurring in the vertical direction is designated with T3 (normally in mechanics the symbol
would be σ). stands for the modulus of elasticity without piezo-effect, i.e. for E = 0. The
double-indexing (33) is required in the framework of general considerations, because
electrical vectors (first number) and mechanical vectors (second number) can occur in
different directions. For the present considerations, however, we have a limitation to the
vertical direction (see figure), so that the indexing could be dispensed with. Even though, we
will keep it in order to maintain conformity to literature. The superscript E is not a
mathematical power but a reference to the boundary condition of the electrical field strength:
E = 0. The formula symbol E (for the modulus of elasticity) can easily be mixed up with an E
designating the field strength; we therefore normally use, rather than the modulus of
elasticity, its inverse s with the same indexing. Thus, does not represent a stiffness here,
but it indicates the elasticity-coefficient in the direction 3 for the field strength set to zero.
While this nomenclature requires getting used to, it is also found in datasheets (e.g. Siemens
piezo-ceramics).
To summarize: if we disable the piezo-effect by electrical shorting, we obtain – for static

mechanical loading – a crystal acting like a spring in the direction 3, having an elasticity
coefficient , the (purely mechanical) behavior of which is described by Hooke’s law.
Fig. 6A.2: The piezo-crystal under static loading.

Contrary to Chapter 6.1, the area-specific and length-
specific quantities are given here. Quadripole arrows in
the technical direction. E = E', S = S'.
We must now account for the fact that the electrical side will of course not always be shorted.
For the second particular case with purely electrical loading that is looked into now, any
contribution from the mechanical side is prevented by the condition S = 0. This mechanical
short circuit, also called “firmly-braked condition”, necessitates complete rigidity on the
mechanical side: the relative change in length (change in thickness) needs to be zero.
Conceptually, this can be achieved with an infinitely stiff crystal.
© M. Zollner 2006
It is customary to indicate this lack of mechanical deformation (S = 0) in the dielectric

constant ε by a superscript S:
Capacitor-equation
In summary: if we disable the piezo-effect via a mechanical short, we obtain – for static
loading in the direction 3 – a capacitor with the dielectric constant ; the (purely electric)
behavior of this capacitor is described by the capacitor equation.
In the third and last step, we drop the particular conditions of the purely mechanical or purely
electrical loading; we now arrive at the general operation by superposition of the two
particular cases. The ideal piezo-electrical transducer-process (as shown in Fig. 6A.2 by the
rectangle) connects electrical and mechanical quantities:
Differential transducer-equations
Using the reference arrows defined in Fig 6A.2, the two node-conditions read:
Node-equations
The two-pole equations of the storage-elements connect flow- and potential-quantities:
Two-pole-equations
From this, we can deduce the system of equations for the general operational case:
The matrix shown on the right maps the two potential-quantities (E, S) onto two flux-
quantities (D, T). It may be interpreted as conductivity-matrix, and transformed to the chain-
matrix; at the same time new piezo-coefficients (d, e, g, h) are defined, as well:
; ; ; ;
The chain-matrix A connects both quadripole input-quantities (E, D) to both output quantities
(S, T). Its determinant [ ] is negative because we have a gyratorical
mapping here: the ideal transducer maps the potential quantities (E', S') onto the flux-
quantities (D', T') [3]. The correspondingly specified signs are in agreement with the four-
pole-theory, but in contradiction to the piezo-parameters normally stated in datasheets – these
parameters are based on old IEEE-recommendations.
© M. Zollner 2006
Using the definition of the algebraic sign as it is customary in datasheets, the determinant of
the chain matrix will be +1, and not -1 (as it would be specific for a gyrator). Now, there are
several possibilities to invert the signs in a series connection of three quadripoles. The
conversion from the technical direction of the arrows to the symmetrical direction is easy to
interpret. This simply corresponds to reversal of the direction of the forces – a direction that
requires special attention anyway. Reversing the direction of the forces also reverses the
reference direction of the perpendicular stress T specific to the surface-area – the reversed
variant of which is termed Tx in the following (Fig. 6A.3).
Fig. 6A.3: The piezo crystal under static stress. Flow-arrow at the output reversed relative to Fig. 6A.2).
The quadripole equivalent-circuit as shown in Fig. 6A.3 uses the symmetric arrow-direction
(as it is common in quadripole theory for the X-, Y-, H- and G-matrix) for the definition of
the chain-matrix. This is not the usual approach but it is a way to arrive at compatibility with
the datasheets. The individual mappings are as follows:
Consolidating the three matrices into an overall matrix via a multiplication, we get:
In this representation, the determinant of the chain-matrix is +1, as is customary in the

datasheets. The piezo-parameters can be converted as follows:
; ; ; ;
The superscript letters in these formulas refer to setting the respective quantity to zero, i.e. for
example = dielectric constant for zero-ed (mechanical) perpendicular stress T. However,
this always refers to external quantities: in Fig. 6A.3, this would be T x = 0 and not T ' x = 0,
and D = 0, not D' = 0. For the other signal quantities, this distinction is not necessary because
of E = E' and S = S'. The subscript indices need to be included if the orientation in space is to
be specified: for example d33 (thickness-mode oscillator), d15 (thickness-shear-mode
oscillator), d25 (surface-shear-mode oscillator).
© M. Zollner 2006
Besides the description of the system with differential quantities (referring to length and area),
there is also an integral (macroscopic, global) representation in which U, I, v, F are used rather
than E, D, S, T. The electrical field strength E is the length-specific electrical voltage U, the
mechanical stress T is the area-specific force F:
For the other two quantities, a time derivative (or an integration, respectively) is required – in
the spectral representation, this corresponds to a multiplication with (or, respectively, a
division by) jω. All signals are complex, as is always the case in general signal theory; the
under-strike often is dispensed with:
The microscopic description using differential quantities looks at the scenario of static stress
(f = 0). Here, a velocity – the vibration- (particle-) velocity – would be of little help, though,
and therefore its integral referring to the length (the relative deformation S) is used.
Correspondingly, the current strength I (which is zero in the case of a static load) is replaced
by its integral referring to the area, i.e. the charge density (displacement density) D. This
static load condition is, however, less relevant for the practical deployment: the electrical
resistances cannot be increased indefinitely, and therefore there is always some current
flowing that leads to recharging processes. For this reason, calculations in practice mostly use
U, I, v, F. Applying integral notation, the equations for the ideal piezo transducer read:
Integral notation
The algebraic signs are oriented towards Fig. 6A.2, and therefore do not correspond to
datasheet-conventions. The apostrophes express that the ideal transducer-effect is referred to –
without any contributions by mechanical or electrical two-poles. A block diagram is shown in
Fig. 6A.4 – contrary to Fig 6A.2 it now includes the integral quantities. The two-pole
parameters change correspondingly:
For the “firmly-braked”, mechanically fixed case (S = 0, or v = 0), the dielectric constant εS
becomes the capacitance Cv, for the electrical short-circuit case (E = 0, or U = 0) the elasticity
constant sE becomes the spring-compliance nU.
Cave!: in this chapter, sE does not designate the stiffness (= force / deflection) but the
elasticity coefficient (= 1 / modulus of elasticity)!
Fig. 6A.4: Block-diagram of the piezo transducer. The

capacitance shown in the figure is the crystal-
v
capacitance from Fig. 6.1 B (C = CK), the shown
compliance is reciprocal to the stiffness of the crystal:
n U = 1 / s K.
© M. Zollner 2006
Empty page
© M. Zollner 2006
7. Neck and Body of the Guitar
“For those who are new to guitars: The neck is the long, thin object to which the rest of the
guitar is attached.” Yes, that is it, in a nutshell – we certainly may agreement with P. Day
here. The rest of the instrument, the body, will be analyzed later. First, let’s talk about this
long, thin object:
7.1 The Guitar Neck
The neck of the guitar has two main functions: with its fretboard, it serves as a platform for
fingering the strings, and it accommodates the tension of the strings. In the case of medium-
gauge strings, a pull of 700 N is generated, and in the case of heavy gauge, it may be up to
850 N. This is about the same weight as that of a person (71 and 87 kg, respectively). To
prevent the neck from warping under this load, it is reinforced by means of a lightly bent steel
rod (known as a truss rod) that runs the length of the neck. On one side, the truss rod ends in
an externally accessible nut that facilitates adjustment of the effective length, and thus also
adjustment of the neck relief. Only those with sufficient experience should undertake
adjustments, as the truss rod may break. Occasionally, there are poorly placed truss rods that
develop a life of their own and start to buzz. The problem is not so much the resulting soft
background noise, but in the fact that vibration energy is lost (dissipated).
The part of the neck that faces the strings is most often formed of a glued-on fretboard of
about 3 – 5 mm thickness. In the case of Fender, the neck is traditionally made of maple,
with the fretboard of rosewood. However, there are also necks without a separate fretboard
(pioneered early on by Fender) – the frets then being embedded directly in this “one-piece
maple neck”), or maple necks with a separate maple fretboard (maple cap). Gibson, the other
big name in guitars, traditionally manufactures the neck out of mahogany, and the fretboard
out of rosewood or ebony. Many manufacturers also produce specially designed guitars
(Custom Designs, Custom Shop Models, Signature Models, Artist Models, etc.), and there are
necks made from alternative materials (e.g. carbon fiber, aluminum, walnut, exotic
hardwoods, and many others), as well.
In the case of necks with a separate fretboard, the truss rod can be inserted from the front –
which is not possible in the case of one-piece necks. For the latter, a groove is milled into the
rear of the neck. After the truss rod has been inserted, the groove is covered over with walnut
or similar wood. Also available are guitar necks with a separate fretboard into which the truss
rod has been inserted in this manner. In addition to the decorative appearance of this option,
the manufacturer may argue that it improves the sound.
Guitar types may be manufactured for many years (decades), but they are not necessarily
always made in the same way. Even the Gibson Les Paul, which was the prototype for
mahogany-necks, was made with a maple neck in 1976 [13]. Moreover the construction of
the neck underwent changes: originally it was made from a single piece of wood, but for some
periods this was switched to a three-part build. The single-piece neck is sometimes said to
produce a better sound, whereas the multi-piece neck is regarded as having better shape
retention. However, such judgments should be used with caution, as there are single piece
necks that retain their shape well, and multi-piece necks producing a good sound.
© M. Zollner & T. Zwicker 2019 Translated by Tilmann Zwicker

7-2 7. Neck and body of the guitar
At the upper end of the neck (the so so-called headstock or peghead), we find 6 tuners (or
machine heads) that the strings are wound on to; the strings then run via the nut to the bridge.
To assure a secure contact between nut and strings, the headstock is recessed a bit to the rear,
or positioned at an angle relative to the neck. This typically makes for a bend-angle of the
strings of 5 – 15°, creating a force pressing the strings to the neck of 9 – 26% of the tension
force.
A bend angle of 15° will certainly be sufficient to guarantee a solid string-to-nut contact – but
5° is relatively little. For this reason, many manufacturers mount small T-shaped string
retainers (“string trees”) between nut and tuners – these increase the bend angle for individual
strings. In this, always a compromise between good contact and little friction needs to be
found. So-called staggered tuners may lend support here: they feature a length of the tuner
shaft that varies depending on the associated string.
At its end opposite to the headstock, the neck is connected to the guitar body. Customary are
bolted (bolt-on neck) or glued (set-neck) connections. Since the neck is excited to vibrate via
nut (or fret) and via bridge/guitar-body, the connection between neck and body must not
dissipate any significant amount of vibration energy. This interface therefore needs to be
given highest attention. In addition, mechanical stability is to be considered. To “defuse” this
system-immanent weak-point, some guitars feature a neck-though construction: the piece of
wood the neck is made from runs the full length of the guitar from headstock to end pin.
“Wings” are glued to this center section to form the actual body. This is a good solution – as
long as the neck does not break. In that unfortunate case, a bolt-on neck would be much
preferable because the neck can be easily exchanged.
More expensive guitars often sport a neck-binding consisting of a narrow ornamental strip
running along the rim of the fretboard. Multi-layer binding is sometimes called “purfling”.
Binding will upgrade the looks of a guitar, any effects on the sound are extremely likely to be
negligible. We did, however, not test this by measurements.
The surface of the neck pointing towards the strings is given a slightly convex shape in the
direction perpendicular to the strings; the backside is of half-round shape (Fig. 7.1). The
profile is V-, U-, D-, or C-shaped – or whatever other designation the marketing departments
come up with. Objective criteria for good or bad neck profiles can only be determined at the
extremes – in the end, every guitarist needs to decide individually what feels good to him or
her, and lets him or her play well. If – as taught for classical playing styles – the thumb is
placed behind the neck, V-profiles are not likely to please: these are more suitable for players
whose thumb traverses the whole neck circumference and sticks out over the fretboard.
Fig. 7.1: Semicircular neck profile with different fretboard radius (7"
and 12"), and differing thickness of the neck (18 and 23 mm). The
depictions below show different shapes of neck profiles
[Fender USA].
For thick necks Custom-made Classical V, Compromise-V Universal shape Modern shape,
Blues, Country for flat necks
Translated by Tilmann Zwicker © M. Zollner & T. Zwicker 2019

7.1 The guitar neck 7-3
The round of the fretboard is defined as the radius of the transverse contact circle. A
stronger round (= smaller radius) matches better with the flexion of the fingers and makes
playing chords easier, while a flatter fretboard facilitates string-bending (in the direction of
the frets). Again, it will be a subjective decision which fretboard-radius is seen as an
optimum. As a standard, guitar manufacturers offer dimensions from 7.25" (18.4 cm, for
example vintage Fender), via 12" (30.5 cm, e.g. Gibson) up to more than 16" (40.6 cm).
Occasionally, fretboards with a “compound radius” are found: the radius changes along the
neck e.g. from 11" – 14".
In combination with the profile, the thickness of the neck is specified at the 1st and the 12th
fret. There are slender necks with a thickness of 1.8 – 2.0 cm, and 'baseball bats” of 2.3 – 2.5
cm (or even more). Our fingers and hands sense already very small differences. The width of
the neck is not standardized, either: it varies (measured at the nut) between 15/8" and 17/8" (4.1
and 4.8 cm). Narrow necks are advantageous for short fingers but require more precise
fingering. Towards the higher frets, the neck usually gets wider: at the 12th fret, we typically
find a width of 5.1 – 5.5 cm. All these measurements hold for 6-string guitars; for 7- and 12-
string guitars, the width of the neck at the 1st fret will be about 4.8 cm.
The graphs in Fig. 7.1 show the round of the fretboard – but this should not be seen as a
purely cylindric round. Rather, there is slight concave curvature also in the direction of the
strings. Pressing the down the string at the same time at the first and last fret will not make the
string touch the frets in the in-between area – a distance of a few tenths of a millimeter will be
retained. Small divergences may be corrected by dressing the frets and/or adjusting the truss-
rod, larger deviations require sanding down (honing) the fretboard and installation of new
frets – a task for the luthier, not for your DIY “Tim-Taylor-Home-Improvement”-approach.
As the interface between musician and instrument, the neck is of central importance for
playing the guitar. A bent neck hampers playing or even renders it impossible. Bent implies
here that the surface of the fretboard diverges from its optimum shape (Fig. 7.2). A twisted
neck (right-hand figure) may lead to laying down the guitar in its final resting place – but in
rare exceptional cases it is the expression of a special art of guitar construction (Lace guitars).
Fig. 7.2: Fretboard (schematically), seen from its headstock-end. The neck shown on the left is in good order, the
other two necks are warped and twisted, respectively.

7.2 The Frets
7.2.1 Position of Frets
The right-handed guitar player changes the length of the vibrating string by ‘fingering’ (i.e.
fretting) with the left hand. (Other fretting and playing techniques are also possible.) With the
finger, he or she presses the string to be sounded towards the neck so that contact is made
with the fret♣. Frets are metal wires with a T-shaped profile (Fig. 7.3). They are set into
grooves in the fretboard cut transversely to the direction of the strings. The upper part of the
fret that protrudes from the fretboard is rounded, and the lower part that is set into the
fretboard is designed as a barbed hook to ensure that the fret remains fixed in place. Frets
make it easier for the guitar player to achieve clean intonation (achieve correct pitch). The
length of the vibrating string is "fretted" at discrete intervals only rather than continuously.
As a first approximation, it does not matter where between the two neighboring frets the
finger is pressed. For the string, the important contact occurs on the fret. Upon closer
inspection, though, we observe that, particularly in the case of tall frets (protruding more), the
strength and position of the fingering can have a small effect on the pitch (see also Fig.7.5).
Fig. 7.3: Longitudinal cut along the

neck. Usually, the finger does not press
the string („Saite“) all the way down to
the fretboard („Griffbrett“).
„Bund“ = Fret.
The open string is supported at bridge and at nut. The distance between the latter two, the
scale, is 24” – 25.5” i.e. 61-65 cm. However, guitars with a longer scale (baritone guitar,
LONG NECK GUITAR) are also in use, as are short-scale guitars (3/4-guitar). Electric
guitars generally have 21-24 frets, not counting bridge and nut. The length of the fretted neck
(i.e. the length of the fretboard) amounts to approximately ¾ of the scale. In some guitars, the
strings do not run directly from fretboard to nut but pass over a zero fret. In that case the
string is always in contact with the same material, regardless of whether it is played open or
fretted. The resulting higher friction has noticeable disadvantages, though: for easy tuning,
the string should be able to longitudinally move over the nut or the zero fret with as little
friction as possible. With too much friction, undesirable hysteresis may be the result.
The difference between simple theory and reality can be found in the distance between frets.
To simplify the calculation, the string and its supports are assumed to be ideal. The guitar is
tuned in equal temperament, with the semi-tone intervals being uniformly approximately 6%.
Semi-tone interval
Already the choice of this approach will invite and define fundamental deviations from just
intonation, amounting e.g. to -0.9% for the minor third, and +0.8% for the major third. Just
intonation is not the preferred ideal (Chapter 8.1): (equally) tempered tuning is the standard
used today. The reciprocity between string length and fundamental frequency results in a
geometric progression for the distances between frets. If the distance between the nut and the
first fret amounts to ∆B, then the distance between the nth fret and (n+1)-th fret is ΔB / IHn.
The distance between frets thus diminishes from nut towards bridge, while at the same time
the neck width (and thus the length of the frets) increases (Fig. 7.4).
♣
Sometimes the area between two fret wires is referred to as “fret”, as well.

7.2 The Frets 7-5
Fig. 7.4: Fretted neck (22 frets) with strings and bridge. E2= low E string (at the thumb), E4 = high E string. The
octave relative to the open string is fretted at the 12th fret – at the mid-point of the string. The full length of the
strings across the nut on the headstock is not shown. (“Sattel” = nut, “Steg” = bridge)
The above calculation does not take into account that the string tension is increased when the
string is pressed down; this results in a further change in the pitch. For instance, if a string is
fingered at the 12th fret, its fundamental frequency should actually be doubled. However,
pressing down on the string causes a minimal lengthening of the string, causing a further
increase in the frequency. The fact that the frequency of the lengthened string is higher rather
than lower is due to the tension change that is dominant here (compared to the change in
length).
In the following calculations, it is important to distinguish between the string length L and the
change in length ∆L. The change in length ∆L is designated the strain ξ. A string that has the
length L in its unfretted state (scale + residual lengths♣ to the tailpiece and to the tuners) is
stretched by the tension Ψ to the new length of L+ξ. The more the string is stretched, the
higher the fundamental frequency ƒG (given a fixed scale M).
Strain ξ
is the mean density (see annex), E is Young’s modulus, κ is the ratio of core diameter to
outer diameter for wound strings (for solid strings, κ = 1) . With the latter, the E-modulus of
the core should be used; κ is between 0.3 and 0.6. The E4 string must be stretched by 5.3 mm
(standard tuning, L= 77 cm), die E2 string by 1.7 mm (κ = 0.42). We observe that the strain
depends on the square of the fundamental frequency fG, and that the fundamental frequency is
proportional to the square root of the strain. The formula for the relative changes is derived
from the differential quotient of the curve.
Relative change in frequency
The relative change in frequency is half the relative change in strain. Note: This is not about
the relative changes in length! If, for example, a string fretted at the 12th fret is extended by
0.02 mm, its length changes by 0.026‰, but this is not what is meant here. The strain
changes by 3.8‰ (E4), or by 11.8‰ (E2) – it is this difference that causes problems. Even if
the E2- and E4-strings are pressed down an equal distance towards the fretboard, the E2-string
goes out of tune by a much greater amount. In practice, however, the E2- string is given an
even greater distance to the fretboard (2 – 3 mm inside width at the 12th fret) than the E4-
string (1 – 1.5 mm). The frequency-increase for the E4-string therefore is negligible in
practical terms, while for the E2-string it is quite large: 0,5⋅11.8‰, corresponding to 10 cent.
♣
Here, the friction that occurs in the nut and bridge needs be considered. Given high friction, L = M applies.

To correct this frequency error, it would be possible to mount the frets in a slanted fashion,
but except for a few exotic constructions, this is not done. Rather, the bridge is positioned
with a slight slant such that the bearing (bridge saddle) of the E4-string is exactly at double the
distance from the nut compared to the 12th fret, but the bearing for the E2-sring is moved back
a few millimeters (= longer string). The exact amount necessary for this correction depends
on the strings, the bearings, and the string action (inner distance of string to fret). For nylon-
string guitars, almost no correction is required due to the smaller Young’s modulus. In steel-
string acoustics we often find around 3 mm (E2); a slant of up to 6 mm (E2) may be necessary
for typical electric guitars.
As shown above, this shift of the bridge does not only depend on the fundamental frequency
but also on the type of string winding. The E2-, A- and D-strings are of the wound type while
the B- und E4-Saite are plain; the G-string may or may not be wound. The individual string
data require a string-specific shift in the bridge. Therefore, many electric guitars feature a
bridge with individual bridge saddles that are adjustable via small screws. After the guitar is
restrung, the natural harmonic of the respective string is played (by very lightly touching the
string – as it is being picked – exactly at its half-way point), and the bridge saddle is adjusted
such that fretting the string at the 12th fret generates the same pitch as that harmonic. In some
cases, two adjacent strings share a common bridge saddle – requiring a compromise in terms
of the intonation.
A special example will show the influence of the overall length of the string: on some
guitars, the string runs a considerable additional length on the other side of the nut and bridge
– up to 25 cm in extreme cases. Conversely, on guitars with a string-clamping system, freely
moveable string length and scale are practically identical. If all other parameters are kept the
same, the string-strains differ by a factor of 88/63 = 1.4 between these two conditions.
However, the (absolute) change in strain due to pressing the string to the fretboard depends
solely on the scale and on the inner distance between (open) string and fret, and not on the
overall length. This means that the longer the string is run outside of nut and bridge, the
smaller is the detuning due to fretting♣. In the example, the relative change in strain (and thus
the detuning) is a factor of 1.4 in the clamped string compared to the unclamped string. This
needs to be considered if a guitar is to be retrofitted with a clamping system.
Fig. 7.5: Relative detuning due to pressing the string to the fret. Scale: M = 0.625m; L = 0.72m.
String-to-fret distance (1. fret / 12. fret) = 0.3 / 1,5mm (E4); 0.4 / 2.0mm (E2); 0,5 / 2,5mm (E2).
Left: core-/outer diameter = κ = 0.38. Right: κ = 0.42. Position of bridge not compensated.
“Bundnummer” = number of fret.
♣
If the string were tensioned via a weight, the tensile force would not change at all when pressing down the
string; the detuning would be negligible (merely a minimum change in length).

7.2 The Frets 7-7
By slanting the bridge, the problems mentioned above can be taken care of – such that the
octave (12th fret) can be played fully in tune. That does not imply, however, that all other frets
offer correct intonation. Fig. 7.5 shows the relative detuning occurring as the string is fretted
– at first without slanted bridge position. The distance of string to fret (the “action”) was
assumed to grow linearly between the 1st and the 12th fret; for the low E-string, two different
cases are calculated. In the left-hand graph, the ratio between core diameter and outer
diameter equals κ = 0.38, on the right it amounts to 0.42. A smaller core diameter results in
smaller detuning but also increases the danger of string-breakage.
For the left-hand section of Fig. 7.6, the bridge received a slant. The change for the E4-string
is very small (0.5mm); for the two E2-cases, 2.5 and 3.9 mm are necessary, respectively. This
already offers a decent solution. A detuning of 1 cent does not really require correction,
anyway. Shifting not only the bridge but also the nut (in the direction towards the bridge), a
further improvement is possible (right-hand graph), although a precision of 0.1 cent
(0.0006%) is merely of theoretical interest. Basis for the calculation was that the string runs in
a straight line from the nut to the tip of the fret, and continues to the bridge from there. Since
the finger fretting the string during play does not actually provide an ideal line-contact but
presses down the string behind♣ behind the fret, an additional strain of the string results and
the required shift of the bridge increases.
Fig. 7.6: Data as in Fig. 7.5 but κ = 0.42 (unchanged). In the graph on the left, only the bridge is shifted; on the
right-hand graph additionally also the nut. For the E4-string, the calculated shift of the bridge by 0.05 mm is not
relevant, for the E2-string, 0.3 and 0.5 mm, respectively, are calculated (line-contact at the tip of the fret only).
“Bundnummer” = number of fret.
In summary, we find the following rule for guitar construction: first, the theoretical scale M
is set, e.g. at 625 mm. Then the calculation of the distance between nut (or zero fret) and the
1st fret results in = M / 17.817. We obtain the distance between the n-th and the
n+1-th fret by dividing the distance between n-th fret and bridge saddle by 17.817.
Alternative: the n-th fret is at a distance of from the bridge. As a next step, the
nut is slightly slanted: its position remains unchanged for the E4-string, while at the E2-string
it is shifted in the direction of the bridge by about 1 mm. Now the bridge saddle is shifted
such that we can play the exact octave at the 12th fret. Given these adjustments, every string
should now be tuned with equal temperament. An additional check using a measurement
device, or our hearing, is advised – possibly small modifications are necessary.
♣
"behind" means: in the direction of the headstock.

Fig. 7.7a: Deviation of the measured position of a fret from the theoretical position. The dashed limit lines show
pitch deviations of ± 1 cent. Measurement tolerance: ± 0.05mm. “Bund-Nummer” = number of fret.

7.2 The Frets 7-9
Relative to the exact equal-temperament tuning (i.e.

HT = 1.05946, k = 17.817), both Gibson guitars
deviate significantly from nominal value in the
higher positions. With the (inaccurate) rule-of-18,
however, (column on the right), the tolerances are
o.k. Could it be that Gibson would still use outdated
rules for manufacture even in the year 2004?
Epiphone, Gibson’s budget line produced in Asia, on
the other hand, shows much higher precision.
Fig. 7.7b: Deviation of the measured position of a fret from the theoretical position. The dashed limit lines show
pitch deviations of ± 1 cent. Measurement tolerance: ± 0.05mm. “Bund-Nummer” = number of fret.
Fig. 7.7 depicts the measured fret positions for a number of guitars. Since frets are rounded-
off metal wires and do not provide sharp delimitations, the exact position is only available in
approximation, with a measurement tolerance of about ± 0,05 mm. We can see exemplary
precision for Taylor and Martin; for the other guitars the deviations are larger but still
acceptable. Only Gibson shows to be the odd one out. Interesting in the two Fender guitars:
Japanese “budget” production by no means shows worse results compared to fabrication in
the US – rather, the contrary is the case. For all graphs, scale and position of the nut were set
for an optimal curve. This is because the actual effective position of the string bearing is
difficult to determine (rounding off of notches, bending stiffness of the strings).

7.2.2 Fret materials
The material most often used for frets is German silver (nickel silver, white copper, argentan,
alpakka§). The precious metal silver is not contained in this material, though – or only in
traces. German silver is an umbrella term for copper/nickel/zinc-alloys that could also be
termed “highly nickel-alloyed special-brass”. The pure copper-zinc alloy is called brass; it is
used as material for frets, as well, albeit rarely. By itself, copper is of a reddish color, adding
zinc results in yellowish hue, and nickel moves the coloring towards the greenish and on into
the silver-white area – thus the name nickel-silver. The addition of nickel and zinc
particularly adds to the hardness – that should be high anyway because the contact to the steel
strings wears down the frets during play. Also, nickel makes the fret more resilient against
tarnish and corrosion. The high nickel content of >30% required for the silver-white color is
not found in guitar frets, though: customary is a nickel additive of 12% (acoustic classical
guitar), or 18 % (electric guitar). More recently, steel has also become popular as a material
used for frets. No problems or issues have been reported – steel seems to be well suited for
this application.
Due to the direct contact to the strings, the frets have a sound-determining function, and thus
we do find a wide range of materials and shapes. Besides German silver, bell brass should be
mentioned – another copper alloy, but with tin as additive rather than zinc. The silver-white
bell brass contains about 77 – 80% copper with the rest being constituted by tin. Relative to
German silver, bell brass tends to corrode more easily but this rarely poses a problem. Despite
the fact that tin is a soft metal, Cu/Sn-alloys reach a similar hardness as German silver.
The dimensions of the frets visible to the guitarist are width and height of the tip of the fret.
Small frets (often found on vintage guitars) have a width of between 1 mm and 1.7 mm, and a
height of 0.6 mm and 0.8 mm. Medium size would be B = 1 – 2.6 mm, and H = 0.7 – 1.1 mm,
while large (jumbo-) frets would measure B = 2.6 – 3 mm, and H = 0,9 – 1,5 mm.
The nut is fabricated from bone, or from a special low-friction plastic, or in some cases from
metal. There is no limit to the imagination of the manufacturers when coming up with
designations: Vintage Bone, Bonoid, Ebonol, Graphite, Graph Tech, TUSQ, to name but a
few. Much attention is paid to the low-friction aspect because the strings need to slip through
the nut free of any hysteresis while they are tuned or bent. Static friction prohibits this slip to
some degree and creates a zone of lacking discrimination. Roller-nuts promise a particularly
low friction; however, they are wider than regular nuts and not necessarily easy to install. The
string should find a firm but still almost frictionless bearing in the notch (or groove) of the
nut. Well suited are v-shaped nut-notches that offer a small seating towards the headstock and
end abruptly towards the side of the fretboard. Shape and depth of the groove of the nut are
carved into a blank nut-piece using a nut-file. Also worth mentioning is the clamping nut that
however requires fine tuners at the other end of the string.
Frets and nut are rounded off at their sides so as not to hurt the hands and fingers of the
guitarist – a fact that is less relevant to guitar-physics but more to accident-prevention
regulations, practices of law, and pathology. Still, a connection to physics may be made: the
dimensions of wood are humidity-dependent while the dimensions of fret-material are not (i.e.
not fretted by it …). If in winter you suddenly feel the ends of the frets on your guitar: file
them off, or see to a higher humidity of the air!
§
inconsistent spelling. N.B.: Alpaka = Lama.

7.2 The Frets 7-11
7.2.3 The Buzz-Feiten-system
In his US-patent no. 6642442 (uspto.gov), Howard B. Feiten describes a system for a
tempered tuning of fretted musical instruments. Supposedly, an extraordinarily pleasing
intonation is achieved by applying small deviations from the traditional tuning. "One very
important aspect of acoustic guitars that has been overlooked is proper intonation" – i.e.
according to Mr. Feiten, luthiers have – until the year 2002 – studiously overlooked that
guitars indeed need to be correctly tuned. Without a doubt, old-world luthiers will beg to
differ, insisting that correct tuning was an objective since long before building guitars became
fashionable west of the Canary Islands. Anyway, not every guitarist is always happy with the
result of his strive for balanced tuning. Enter Howard “Buzz” Feiten.
In the description of his patent, Mr. Feiten explains that guitars have their frets positioned
according to the Pythagorean Scale. One is tempted to object with a “well in that case …”
and to add – depending on ones disposition – a sarcastic “yeah, maybe the axes made
stateside”; but let’s hold our horses for a minute. To begin with, and in order to avoid
misunderstandings, the term “Pythagorean” is explained: "The Pythagorean Scale is based
upon the fourth, the fifth, and the octave interval ratios." Without a doubt: that’s Pythagorean.
However, what’s that got to do with the guitar? In Europe, especially in Old Europe♣, tuning
is accomplished since the 1700’s using equal temperament, not Pythagorean. But let us allow
Mr. F. to continue his explanation: "To determine fret positions, guitar builders use a mathe-
matical formula based on the work of Pythagoras, called the rule of 18 (the number used is
actually 17.817). This is the distance from the nut to the first fret." May the present work be
charmed against that many errors in a single paragraph – that’s what one instinctively thinks
as the author of the book you are reading … Anyway: the rule of 18 generates a geometric
sequence for fret positions: equal temperament; but he latter does not trace back to Pythagoras
whose intonation is based on fifths, fourths and octaves, as Mr. F. elucidates himself. What
Mr. F. seeks to express with “this is the distance” remains shrouded in Greek history, too. The
subsequent explanations in the patent (not cited here) then do reasonably and correctly clarify
what the rule of 18 purports. Let’s note: H. B. F. sees the reason for the inadequate precision
of tuning in the use of the Pythagorean (fifths-) tuning as it is contained in the rule of 18. That
is incorrect but apparently did not phase the patent examiner (the one at the US Patent Office).
To cite Mr. F. some more: "Prior to the mid 1600's, pianos had evolved from a 'just'
intonation to 'equal temperment'; i.e., tuning the instrument so that all the notes were
mathematically equidistant from each other. … It was only partially successful and resulted in
the entire keyboard sounding slightly out of tune, especially in the upper and lower registers.
In the mid-1600's, an enormous breakthrough occurred in piano technology. The 'well
tempered' keyboard was conceived." Let us ask J. M. Barbour to comment about just
intonation: "There is no such thing as just intonation, but rather many different just
intonations; of these, the best is that which comes closest to the Pythagorean tuning". So
indeed: in the Middle Ages there was need for action, and "equal temperament", i.e. an
intonation causing equal beats within the scale and allowing for modulations across the whole
circle of fifths was considerable progress. However, "equal temperament" must not be
confused with the "well tempered intonation" proposed by H. B. Feiten! The latter in fact
distinguishes between "equal temperament" and "well tempered". "Well tempered" is a
specially modified tuning derived from the uniformly-beating equal temperament.
♣
The ethnologist Donald Rumsfeld specified this term (otherwise to be understood more as a geographic
distinction) via his subjective, differential diagnostic observations that were complemented by the philosopher
Joschka Fischer by an evaluation of the origin of European and US-American culture (translator’s note: you
better read this observing a STRONG twinkle in the author’s eye …)

Mr. F. continues: "the universally accepted method for intonating guitars represents a form of
"equal temperment" … a method that was abandoned in the 1600's by piano tuners". Hold it,
Buzz, didn’t you say a moment ago that guitars are tuned pythagoreically? "Based upon the
fourth, the fifth, and the octave interval ratios." Ye gods and little fishes! Hopefully we all
still know after reading the patent how in the end any tuning and intonation is achieved. At
this point, why don’t we digress a little into contemporary literature, to show what kind of
wide appeal a topical – if error-prone – patent can have:
Musik Produktiv, one of the giants in German music commerce, state regarding the matter:
with this modification of the scale, Buzz Feiten achieves a “well-tempered” tuning for the
guitar. Um … the term “well-tempered” is a bit misleading here because Mr. Feiten expressly
seeks to avoid this piano-tuning. Musik Produktiv continue: a piano tuner explained to Buzz
Feiten that an electronic tuner cannot generate a well-tempered tuning. We may disregard the
nitpicky rationale that – as a measuring instrument – an electronic tuner can never by itself
generate a tuning, but we cannot help recognize another considerable discrepancy: the piano
tuner sought to achieve a stretched tuning (according to the Railsback curve). That is – at least
according to conventional terminology – something rather different than a well-tempered
tuning that is directly connected to Bach/Werckmeister (supposedly equally beating).
Musik-Thomann, the huge mail-order shop, writes: For calculating the scale and adjusting
the intonation, people relied on old, traditional formula. These heirlooms were based on a
method that piano tuners developed already in the 16th century: the equal-temperament
tuning. The commonly used formula to position the frets had already been developed by
Pythagoras. There is, however, an error due to the stiffness of the string that generates too
strong a disturbance. In this comment in its German language form, “equal temperment” has
been translated into the German expression for “equal temperament” – not as intended, but
with some good will we can arrive at the intended interpretation. Again, Pythagoras is brought
in. And finally: All over the world, more and more guitarists have their darlings modified by
authorized retrofitters. Right, a lot of offers are said to exist across the Internet. Guitar players
may need that kind of thing. Cave inflammtio!
Here’s what Proguitar (not yet a giant) contributes: The formula for positioning the frets was
already developed by Pythagoras. Mr. P. must have been a very early fan of the Strat.
Maybe we can clarify this jumble a bit: Pythagoras is readily cited with his insight that given
constant string tension, frequency and length of the string are reciprocal (monochord = single
string instrument). Still: already before Pythagoras, the Egyptians, Sumerians, Chinese,
Indians, and presumably many other peoples in the ancient world knew about physical and
mathematical interrelations. However, the Pythagorean school had the greater impact onto
Western civilization, and in particular it left written documents early on (Euklid, Didymos,
Ptolemäus, and many more). This Pythagorean school spawned a tonal system based on fifths
and octaves that to this day is designated the Pythagorean tonal system (Chapter 8.1). It is
applied, in its pure form, by the canons regular up to the 16th century, and in a modified form
by the harmonists [Simbriger/Zehelein, Barbour]. When, from the 16th century, keys with
more and more chromatic signs appeared, the subjectively perceived dissonances of the
Pythagorean system were increasingly felt (or rather heard). Two improvements were devised
as a remedy:
1. Increasing of the number of steps within an octave, and

2. Tempering, i.e. the fine-tuning of individual notes.

7.2 The Frets 7-13
The temperament may feature equal or different beating between notes. Simbringer/Zehelein
date the first introduction of temperament back to 1482: Bartolomeo de Ramis demands that
the difference between the third (5:4 = 1,2500) and the fifth no. 4 (81:64 = 1,2656) be
balanced out by temperament. Barbour assumes the date to be 1496, and lists 17 different
temperaments, designating them “meantone temperament” and “comma temperament”.
Around 1533, Lanfranco lays the foundation for the equal-beating intonation (equal
temperament), and subsequently Vincenzo Galilei and Marine Mersenne (1636) concern
themselves with the question how to calculate the 12th-order-root of 2 (or an approximation as
precise as possible, at least), without having a pocket calculator at hand. Then, the works of
Neidhardt and Werckmeister – carried out around 1700 – become very popular. Almost 200
years later, Alexander Ellis reports that even “the best British piano tuners” could not produce
an acceptable tuning with equal beating, and in 1948 and 1943, Railsback and Schuck/Young
publish in the Journal of the Acoustical Society of America (JASA) the stretched intonation
found in pianos: high notes are tuned slightly sharp, and low notes slightly flat.
At this point, the Feiten-patent picks up: "In the mid-1600's, an enormous breakthrough
occurred in piano technology. The 'well tempered' keyboard was conceived, and with it a new
standard for piano keyboard intonation which we still use today." In the mid-1600's, i.e. in the
17th century, Mersenne & Co. were working on that root-calculation and developed the equal
temperament. Does therefore, in Mr. F.’s book, equal temperament mean “well tempered”?
No, that can’t be because he has (correctly) termed the equal temperament with “equal
tempered". But why would he (with the apparent support of the patent examiner) then write:
"The inventors believe that the reason that guitars still sound out of tune, in spite of 'perfect'
intonation, is that the universally accepted method for intonating guitars represents a form of
'equal temperment' … a method that was abandoned in the 1600's by piano tuners!"? Quite
enigmatic, these Americans! The patent continues: "When a piano tuner intonates a piano, he
uses one string as his 'reference' note, typically, A-440 (or Middle "C"). He then 'stretches'
the intonation of the octaves, plus or minus a very small amount of pitch. These units are
called cents". Ah – here’s the crux of the Buzz-ing matter. Even without further historic ado
(already the ancient Greeks …) we could formulate the idea behind the patent application as
follows: similarly to pianos, guitars should be tuned using a stretched intonation.
That justifies a quick look into JASA: Schuck/Young cite in their publication (JASA 1943)
the stretched intonation found by Railsback. Below E2 and above approximately E6, a
considerable effect is indeed recognizable, with the piano tuning deviating by up to as much
as 30 cent. That is not a wonder: the investigated pianoforte will be challenged to conjure a
whopping bass of 25,6 Hz out of a mere approx. 1 m string-length, and at the top there’s
about 4 kHz tickling out of some tiny 5 cm string-length – in this scenario, dispersion-induced
inharmonicity will definitely play a role. In the guitar … how shall we put this without being
transatlantic-ally un-accommodating … well: it’s not directly possible to coax 27.5 Hz out of
your regularly tuned guitar, and 4.2 kHz on an open string is more of an un-feasible wish, to
put it mildly. That’s not even considering that in the piano, for the medium pitch range, the
different frequencies come from strings of almost equal in gauge but of different lengths,
while on the guitar we have strings of the same length but differing thickness. Schuck/Young
explicitly say: "The sharpening is least in the two octaves below middle C". “Sharpening”
relates to the partials and can be taken to be synonymous with “stretching”. Middle C is on
the E4-string at the 8th fret. The string-pitches of the guitar therefore fall exactly into the range
where the effect is minimal on the piano. Nevertheless: 2 cent per octave may occur according
to Schuck/Young and Railsback, i.e. about 0,12%. For the piano, that is – with its E4-string
being about 1 mm thick. That is about 4 times the thickness of the corresponding guitar
string! And thus on the piano the build of the partials includes much more inharmonicity.

Nevertheless, there must be something to this patent by Mr. Feiten. Doesn’t Californian
guitar-god Larry Carlton say about the Feiten-tuning: “I’ve been playing the guitar since I
was six years old, and finally it is in tune.” One could of course find a number of reasons for
this♣. How about: Larry C.’s nickname is Mr. 335 due to his penchant for Gibson’s thin-line
guitars. If, on the necks of his noble vintage pieces, the frets were placed as inaccurately as on
all Gibson necks this author checked, then reworking of the necks might indeed have made
for audible added value. Whether the application of the Buzz-Feiten-offsets alone really
involves significant advantages … every guitarist needs to find that out for him/herself. Here
are the Feiten-tuning-offsets proposed for electric guitars, to be adjusted after the nut- and
bridge repositioning has been done:
E: 0 / 0 0 / 0
H: +1 / 0 0 / –1
Patent USA-6642442 (Feiten), uspto.gov:
G: –2 / 1 0 / +1 Tuning-offset in cents relative the equal-temperament intonation. The first
D: –2 / 1 0 / +1 number holds for the open string, the second number is for the octave fretted
A: –2 / 0 0 / +1 at the 12th fret. Column on the left to be applied to electric guitars; the
E: –2 / 0 –1 / 0 column on the right is for acoustic guitars.
In terms of the tuning offset, the Feiten-patent only distinguishes between electric guitar, steel
string acoustic, and nylon string acoustic. It ignores the fact that, in wound stings, the ratio of
core- to winding-diameter influences the inharmonicity of the partials. On the other hand,
rather extreme precision is required, as the table for the acoustic guitar shows. The high E-
string is to be tuned exactly to the (pure) octave for the 12th fret: for all other strings, the
octave is detuned by 1 cent. Just as a side-remark: if a string is heated up by a mere 1° C, the
string frequency diminishes (for unchanged mounting) by 9 cent. Therefore, the string may
change its temperature by no more than 0.1 °C in order to maintain the Feiten-tuning! That
fearful sport, father attempt not too oft! [Schiller]. And from the same author and the same
poem (“The Diver”): Let not man to tempt the immortals e'er try, Let him never desire the
thing to see, That with terror and night they veil graciously. And since there is still some
room here: the unforgotten K.-H. Hansen tells us: Easily does the lad talk big about the mil –
he will be an old man by the time he achieves the hundredth part.
In conclusion of this chapter a bit of an anecdote: a much-lauded Californian guitar god (we
shall omit the name … for legal reasons) visits Germany to play a concert. Just before the gig,
he takes his el-cheapo♥ six-string to the local shop for a quickie-bridge-adjustment. The latter,
however, runs into a substantial snag because the bridge is attached so firmly with double-
sided adhesive that to forcedly move the thing is deemed dangerous and inviting real damage.
The guys in the shop don’t dare to do anything, cause let not man to tempt the immortals e'er
try (see above), especially with the gig looming that same night: with that god visiting Old
Europe once every blue moon, you don’t want to botch up his guitar – here in Germany, of all
places. So: bring back the guitar unrepaired. That evening: the god plays god-like, in spite of
the “displaced” bridge. Or was it because of the resulting special intonation? Who knows how
exactly a god ticks?
♣
Larry C. is not really a spring chicken anymore; it’s been a while since he passed the age of 6.
On impulse, also the thought pops up: what further heights might Jimi H. have climbed, had he in time ...
♥
The real (precious) stuff will probably and preferably stay safely at home in CA ...

7.3 Geometry of strings and neck of the guitar 7-15
7.3 Geometry of Strings and Neck of the Guitar
At its bearings (nut and bridge), the string receives a directional bend of 4° – 30°. This bend
angle ensures of the required bearing forces but also contributes undesirable frictional forces.
Moreover, fretting the string results in a further bend angle at a point where tension forces and
playing forces interact. The following two chapters will describe the geometry of these bends
– Chapter 7.4 will then look into the corresponding forces acting on the string.
7.3.1 Angle of headstock and neck
For most electric guitars, the upper surface of the fretboard does not run in parallel with the
upper surface of the guitar body. Rather, the neck is very slightly angled towards the rear.
This way, the point where the string runs across the bridge moves somewhat away from the
guitar, helping to increase the bend angle of the strings across the bridge, and thus to increase
the bearing forces at the bridge. Typically, the neck angle is 0° - 7°. With a set neck, the angle
is fixed; for a bolt-on neck, the angle can be adjusted using screws and/or shims, or by
removing wood from the neck-joint section.
Fig. 7.8: Bend angles of headstock, neck and strings. On the left, the angles between headstock and guitar neck,
and between neck and guitar body are 0°; on the right, the angles are 10° and 5°, respectively. (Customarily, the
sharp angle, e.g. 10° is given, rather than 170°). On the right, the bend angle of the strings is greater, resulting in
a higher bearing force at the nut and bridge, respectively.
The adjustability of neck angle, neck curvature (truss rod) and bridge height (Fig. 7.8)
provides for the possibility to accommodate the playing action as desired by the guitarist.
More extensive adjustments involve the height of the frets but taking these steps remains with
the specialist. With regard to the playing action: whether changing the neck angle or the
bridge height makes no difference – but pickup-to-string distances, and the bend angle at the
bridge are affected. If, when playing the strings on the upper frets (around the 12th fret and
higher) strong string-rattle occurs, either the bridge needs to be raised, or the neck needs to be
angled towards the front. In case the action is now too high at the lower frets, the truss-rod
may be (carefully!) tightened. The manufacturers recommend a slightly concave curvature of
the fretboard (between nut and bridge). Every guitarist needs to find his or her own optimum
setup. Those who prefer a low action for solos may adjust the neck to be completely straight
… convex curvature, however, is to be avoided. More of a concave curvature will be
preferable to those mainly playing chords on the lower frets. Still, the necessary adjustments
will be a matter of a few tenths of a millimeter.
Again, a warning: the truss-rod may break if tightened too much – to replace it, the glued-on
fretboard will probably have to be removed, or the neck will have to be replaced (if bolt-on).
Therefore, beware of overly enthusiastic “home-improvers”, and too zealous “sales experts”!
The effect of the neck angle on the sound is and remains a mystery: in the book “The Gibson”
[13], the neck angle of the 1952 Les Paul is specified with 1°. Over the following years there
are several changes; the reissue towards the end of the 1960s sports a 7°-angle. The 1952
original is attested "very good sustain", the 1953 Les Paul (with 3° neck angle) still receives a
"good sustain" verdict.

When in 1960 the neck angle increases to 5°, "less sustain" is noticed. The book continues:
"in the 1970s it was increased to 7°. A lot of players at that time complained that the guitar
sound became harsh and had less sustain compared to the older models. The reason was of
course the much steeper angle of the neck". Of course? The book gives conjectures without
any reasoning. In the 5-page Les-Paul-historiography, so many changes in the details are
listed that we could surmise that only the number of strings and the name of the guitar were
left alone. Subject to change were: frets, materials of body and neck, neck cross-section,
number of parts the body and the neck is – respectively – made from, the bridge design (at
least seven variants!), the pickups … the neck angle was merely one of many details. In the
absence of exact data, pure assumptions as those given above are eagerly seized and spread
like wildfire in the guitar-world. “Mystery of the old Les Pauls (by now selling for 6-figure
numbers) unraveled: it’s all in the neck angle!”
Let us give two experts a change to speak: "The first Gibson Gold-Tops had a very shallow
neck angle… This design made the sustain suffer. Around 1953, the angle was increased … it
made for improved sustain." So much for Bacon and Day’s comments in the Les-Paul-Book.
Jun Takano writes quite differently in The Gibson: "The angle of the neck joint was 1° when
the Les Paul Model was introduced. The sustain of the guitar was very good because of the
shallow angle of the neck. In 1953, when the bridge was changed to a stud type, the angle of
the neck joint was altered to 3°. However, sustain was still good, because the neck angle re-
mained shallow." While Bacon/Day assume that the sustain improves with increasing neck
angle, Takano surmises the exact opposite: "As the angle of the neck gets shallower, the string
tension♣ gets lower and sustain gets longer." The authors of both books do however also
report about a change of the bridge in 1953 – therefore there are, at least, two potential
reasons for a change in sustain (translator’s note: if indeed there really is such a change).
Changes in the neck-angle can have two effects: they can change the instrument geometry and
thus the instrument’s Eigen-vibrations (“natural” vibrations at natural frequencies – or Eigen-
frequencies – of the instrument), and they can change the bend-angle of the strings at the
bridge. That angle depends on two sections of the string, though! When in 1954, due to the
introduction of the Tune-O-Matic bridge, another change in the bridge happened, the bend
angle at the bridge became variable, and now was not at the neck angle’s mercy anymore.
Because we do not want to rip the neck from a 1954 Les Paul only to glue it back at a
different angle, we only have this – rather poor – lesson: there are many speculations about
the effect of the neck angle on the sound; we should refrain from adding yet another one.
Rather, we can again recall specialist literature:
"The sound is particularly wonderful if the neck angle is 3.5°. One could say that at 3.5°, the
tone sings; at 4° there is a fatter bass – but it won’t sing as well.” That’s luthier Thomas
Kortmann speaking, at Gitarrist.net. Wouldn’t it have been wonderful to know what the
sound at 3.6° is? "The sound of an electric guitar is predominantly determined by the bend
angle of the strings.” E-Gitarren, page 89. That will make all those happy who own
substandard pickups. "We often find reports about Gibson having increased the neck-angle
when the new McCarty-bridge was introduced. That is a myth and not true. A Les Paul
sporting a stud features the same neck angle as an earlier version with trapeze. Only in 1955
the neck-angle was substantially increased… and even better sustain resulted."
Gitarre&Bass, Gibson special edition, 2002, page 15. Whether that result was due to the neck
angle, or due to other factors introduced at the time is not specified.
♣
This speculation about the string tension certainly will not find any scientific basis: given the same string, the
pitch depends solely on the string tension - the latter therefore remains unchanged with constant pitch.

7.3 Geometry of strings and neck of the guitar 7-17
7.3.2 String tree
The strings of a guitar rest on two small ridges: the nut and the bridge. In order to obtain a
force-fit without slack, the string needs to experience a change in direction. The
corresponding bend angle is specific to the guitar and varies between 5° and 15°. For
achieving the bend at the upper string end, the classical method has the headstock angled back
relatively strongly towards the rear. In a Martin D-45, for example, the resulting bend angle is
15°. Gibson’s electrics reach this value, too – or even exceed it. With Fender, the situation is
different. Always frugal with the available materials, Leo Fender does not waste any wood
and refrains from angling the headstock backwards. The result is a very small bend angle of
the strings (Fig. 7.9), and string buzz occurs when playing open strings. Remedy is found in
the form of hook-like string guides that deflect in particular the B-and E4-strings, pulling them
towards the headstock
Fig.7.9: Side view of the headstock of a Stratocaster. Without string tree, the bend angle of the E4-string
amounts to only 2°; with string tree, it increases to 6° (dashed). For the E2-string (thin line), the bend angle is
sufficient without string tree.
String trees are offered in different shapes: as “butterfly” vintage original (a stamped strip of
sheet metal), as roller bearing (roller string tree), as washer, or a thin oblique pin. They do
increase the bend angle at the nut – but also generate an additional frictional force in terms of
any longitudinal movement of the strings. This friction is considered undesirable. The wrap-
around angle at the string tree of a Fender guitar can amount to as much as 7° – generating a
frictional force that is even higher than that occurring at the nut. On the other hand, a Gibson-
typical bend angle of the strings is, at e.g. 15°, more than twice that on the Fender-ish
competition – and yet in guitaristic circles, Gibsons are not actually known to be unplayable.
That does not mean that friction would generally be no problem at all: there are sharp-edged
string trees of the butterfly-type that wound strings more or less clamp themselves to. Let’s
not enter an expert discussion here why, in the first place, a wound string would have any
business interacting with a string tree on one of Leo’s guitars … anyway: corrective action is
easily possible via a delicate file or a practically invisible strip of Teflon. That aids the
mechanics and does not hurt the look. Oil, Vaseline or machine grease would be suitable to
reduce friction, as well.
Does a string tree change the sound? No, here we do not mean the targeted improvement via
the increase of the bearing force of the string at the nut – but would there be any additional,
possibly undesirable effects? Very theoretically, the Eigen-frequencies of the headstock could
retune themselves due to the additional mass of the string tree, but that has no practical
relevance. The same holds for the Eigen-frequencies of the strings running across the
headstock: plucking an open string and damping these remaining string sections with the
other hand will not cause any changes in the “electrical sound”… N.B.: the acoustical sound
of an electrical guitar is insignificant, anyway (Chapter 8).

7.4 Dynamics of the Strings
With dynamic, what is meant here is not a contrast to static, but a reference to the forces
acting on the string. In Chapters 1 and 2, this topic was already addressed, albeit in a rather
theoretical fashion, and without connection to a specific string bearing. In Chapter 7.4.1 now
follows an analysis of the real playing situation: the fingers of the fretting hand need to exert a
playing force F on the string to press it downward or to “bend” it (push and pull the string
parallel to the frets). Since the string is deflected to the fret, two themes join up: the string
dynamic and the guitar geometry. In Chapter 7.4.2, the forces transmitted by the string to its
bearings (nut and bridge) are investigated both in static condition, and for the vibrating string.
7.4.1 Playing forces
When pressing down the strings, the fingers need to muster the force F that runs transversely
to the direction of the strings. If we assume, for example, that a string is to be deflected at the
12th fret transversely by a distance η, then the transverse force F required for this amounts to:
Ψ = tension force, M = scale
From the differential equation of the oscillation, the tension force Ψ results in:
A = overall diameter, = mean density
From this, we can calculate the force necessary to fret the string when playing:
Force required to press down the string
The playing force is proportional to the scale length: electric guitars with a scale of 25,5"
(648 mm) require forces higher by 6,25% compared to guitars with a 24"-scale (610 mm).
The playing force is also proportional to the action of the guitar: if the distance between
string and fret is increased by 10%, the necessary playing force also rises by 10%. Moreover,
the playing force is proportional to the square of the string diameter: using a 10-mil-E-string
rather than a 9-mil-E-string increases the required playing force by 23%. The playing force is
of the same value for all strings if the string-diameter relates inversely to the fundamental
frequency. For wound strings, we need to consider that their effective density is smaller by
about 10% compared to the fully solid string (see appendix A1). Light strings require playing
forces between 0.5 and 1.5 N; for heavy strings, the playing forces are about double.
When executing string bending, the string is not only pressed against the fret but
simultaneously stretched in the transversal direction – this leads to a clearly noticeable pitch
increase. Applying a fast transversal movement creates a vibrato, slower movements result in
glissandi that often amount to a pitch-change by a whole step. Since the string-tension force
Ψ is proportional to the square of the frequency, a pitch increase by a whole step implies an
increase in the string-tension force by 26%. The tension force is proportional to the elongation
(strain) ξ that depends on the overall string length – i.e. also the (residual) sections of the
string that extend beyond the nut and the bridge (that is assuming there is only small friction
at the nut and at the bridge).

7.4 Dynamics of the strings 7-19
From the strain ξ we can deduce the required transversal displacement η. The latter may be
recalculated into the required transversal tension force FZ:
Force for string-bending (12th fret, +1 whole step)
The “string-bending force” depends on the scale length with a power of three! Changing from
a 24”-guitar to a 25.5”-guitar increases the required pulling/pushing force by 20%. The
remaining residual string R (i.e. the part of the string behind the bridge and behind the nut)
enters the section of the equation below the square root as relative difference. On a Gretsch
‘Tennessean’, for example, the G-string extends by 16 cm from the zero-fret to the machine
head, and another 11 cm add themselves behind the bridge towards the Bigsby-vibrato. This
extends the 24.5”-scale (M = 62 cm) by R = 27 cm to an overall length of 89 cm – 44% more
than what would correspond to the actual scale length. The string-bending force rises by 20%
due to this setup. In the bending force, the ratio κ of core-diameter to overall diameter of the
string makes itself felt, as well; κ is 1 for solid strings, and 0.3 … 0.6 for wound stings.
If small string-bending forces are desired, the string should end as close as possible behind
nut and bridge. Since with the gauges as they are typically included in string-sets, the bending
force for the E4-string is about 50 – 60% higher than that for the B-string, a short remaining
string length would be particularly desirable for the E4-string. The Stratocaster (and most
other Fender guitars), however, feature an E4-string that is the longest of all … well, Leo was
not a guitar player (Translator’s note: also, in the early 1950’s, bending strings was only
starting to become fashionable). Those who would like to experiment can restring a left-hand
guitar to be played right handed – bending will be easier on it. Or, conversely, a right-hand
guitar may be restrung for left–hand use … oh – hi there, Jimi! A clamping nut (Floyd Rose,
Schaller, Steinberger) will also bring improvements.
Bending by a whole step at the 12th fret will typically require string-bending forces in the
range of 5 – 10 N for light strings. Since there is a square dependence on the string diameter,
heavier strings will easily demand (up to) double the bending forces, requiring quite strong
muscles in the hand and lower arms.
Easier on the muscles is changing the pitch via the vibrato arm (tremolo). The latter engages
at the spring-loaded tailpiece and ensures a comfortable lever-transmission for changing the
string tension and thus the pitch. In this construction, the tailpiece is not rigidly mounted to
the guitar body but remains moveable by via a rotatable shaft, or a knife-edge bearing. One to
five springs counteract the pull of the six strings. The effective spring stiffness related to
operating the vibrato results from the sum of the stiffnesses of strings and tailpiece-spring.
Soft springs are necessary if the vibrato arm is to be operated with little force. Such a setup,
however, will increase the forces required for pushing/pulling the strings. A simple thought
experiment may elucidate this: let us assume that the guitar is bolted down, and the string
tension is provided by a weight – e.g. 6 kg for the G-string. (a) Pulling that string will, despite
the exertion of force, not change the pitch since the force of the weight is not changed (given
that the bearings have no effect), after all. (b) If now all 6 strings receive their tension
combined from a single weight (e.g. 60 kg for a 009-string-set), the stiffness of the bearing of
the G-string is given by the sum of the stiffnesses of the remaining 5 strings. (c) If the tension
force is not generated by a weight but by a tension spring, we get, for the bearing stiffness of
the G-string, the sum from 5 times the stiffnesses of the strings plus that of the tension spring.
(d) If the tailpiece is fixed to the guitar body in a non-moveable fashion, the stiffness of the
string bearing is infinite. Only this latter case provides for easiest string bending.
Case (a) is unusable for string bending: despite applying a bending force we do not get any
pitch change. The last case (d) is ideal: due to the immobile tailpiece, the whole of the string
bending force is used to change the tension force (via the force parallelogram). The practical
situation with a vibrato-tailpiece remains in between: the stiffer the vibrato-strings, the easier
the string bending gets. That’s why some guitar players opt for the stiffest of all variants, and
block the vibrato mechanism with a piece of wood.
Most vibrato-systems operate based on a simple principle: as the vibrato arm is pressed down,
the pitch is lowered, if it is pulled up, the pitch increases. As a rule, the different strings are
not equally de-tuned, and therefore the notes in a sounded chord will not remain relatively in
tune when using the vibrato arm.
In the Bigsby vibrato, the tailpiece (string retainer) is constituted by a metal cylinder around
which the strings are run. As the vibrato arm is pressed down, the cylinder rotates: all strings
are shortened by the corresponding (equal) distance. To obtain equal de-tuning for all strings,
not the absolute strain ξ, but the relative strain Δξ/ξ, would have to change by the same
amount. Since the elongations for correct tuning amounts to, for example, 1.2 mm (E2) and
4,8 mm (E4), respectively, an absolute constant change of the elongation will give 4 times the
detuning for the E2-string relative to the E4-string!
For the Stratocaster, the situation is mildly better: the bridge/tailpiece-contraption as a whole
tilts around a straight line, with this line – defined by 6 (or 2) set screws – located just ahead
of the string-anchoring points. Since the bearing point of the E2-string is set back (Chapter
7.2), the effect of the operation of the arm is a bit weaker for this string, and the relative
detuning is not as strong compared to the Bigsby vibrato. Not that much is gained, though:
instead of detuning the E2-string by a factor of 4 relative to the E4-string, the detuning is by a
factor of 3 in the Strat. If indeed depressing the vibrato arm was to detune all strings by the
same interval, an individual lever-transmission would be required for each string, and it
would need to depend on the string data (κ). Luthiers did apparently not see much of a need
for this, and neither did most guitarists.
As a typical example, we measured the force required on the vibrato arm of a Stratocaster to
change the pitch by a ¼-note (2.93%): E2: 2,9 N, A: 4,9 N, D: 5,2 N, G: 3,3 N, H: 3,9 N,
E4: 8,5 N; (009 string set, 3 vibrato springs installed). The guitar was fitted with all 6 strings
and the vibrato arm was depressed until a detuning of ¼-note was reached for the respective
string. A rough estimate shows that the main part of the spring stiffness results from the
strings; the vibrato only provides about 1/3rd; it is relatively soft. However, the guitar player
wishing for higher stiffness may increase the number of the vibrato springs up to 5 ... with
corresponding provisions and suitable space already provided by the manufacturer.
A vibrato system is detrimental to the tuning process: as one string is tuned to pitch and the
next string is tensioned, the spring-loaded tailpiece slackens, and the string already tuned up
will be out of tune. Tuning up all 6 strings thus becomes an iterative process. If a string breaks
during playing, the pitch of the remaining strings rises because the pull previously provided
by the (now) broken string is lost. Last not least, worse tuning stability should be noted: due
to unavoidable bearing friction, hysteresis appears. Depending on whether the vibrato arm is
let go from the push- or the pull-position, different tuning will result. Resulting frequency
deviations are inaudible only with high-grade vibrato systems. The sound may be influenced,
as well: the spring-loading of the mass of the tailpiece may lead to low-frequency resonances,
and spring vibrations may be transformed into electrical signals by magnetic pickups.

7.4.2 Bearing forces at the nut
Unless fretted, the middle, freely vibrating section of the string is delimited by two
"supporting bars". The upper support bar (at the headstock) is called nut, while the lower one
is termed bridge (sometimes specified by constructional details such as roller bridge, bar
bridge, etc.). For the following considerations of the bearing forces, nut and bridge are not
actually distinguished, and for this reason both shall be called saddle here.
In classic guitar design, the string experiences a (sharp) bend at each saddle, changing
direction by the bend angle α (Fig. 7.10). Most electric guitars have adopted this bearing-
principle. The clamping nut would represent an alternative – but this will not be considered
here. The change in direction of the string leads to a spatial (vectorial) force-decomposition:
besides the tension-force Ψ of the string, there is also the saddle-force FS that rises as the
bend angle increases. For our first considerations, let’s assume that the string can slip across
the saddle without any friction: the tension forces on both sides of the saddle are therefore
equal in strength. The saddle force can be calculated from:
Saddle bearing-force
Given a tension force of 850 N (013 set of strings, all 6 strings tuned to pitch), and a bend
angle of 10°, a saddle force of 148 N results – corresponding to the weight of a mass of 15 kg.
Back in the day when guitars did not have electric pickups, the guitar top needed to be made
of vibration-happy wood as thin as possible – this in order to achieve a decent sound volume.
The bridge (-saddle) therefore could not absorb strong forces, and the bend angle thus had to
be small. For the above example, small bend angles involve a proportionality between bend
angle and saddle-force: halving α to 5° will halve the saddle-force, as well – to 74 N. A high
saddle-force is nevertheless desirable for the string to solidly lie on top of the saddle and not
start any secondary motions that would kill off vibration-energy. Guitars with thin tops
require a compromise between stability and sound: a large bend angle guarantees safe and
solid bearing, but the resulting strenuous loading of the top may lead to fracture. Solid-body
guitars do not have that problem; any bend angle is possible. However, for large bend angles,
friction effects increasingly need to be considered. If the tension-force changes on one side of
the saddle, the string slides length-wise across the saddle; in the case of considerable
frictional forces, the string may be out-of-tune. Specifying the bend angle of the strings
therefore is an important step in the design of a guitar.
Fig. 7.10: Force-decomposition at the saddle. Fig. 7.11: Static and dynamic saddle-forces.
Ψ = tension-force of the string, FS = bearing-force. FT = tangential force, FN = normal force.

At rest, two static forces act onto the string: the tension force and the saddle force. However,
as soon as the string is plucked, a dynamic♣ load is added. In the general case, the string
vibrates with a combination of a fretboard-parallel and a fretboard-normal vibration: the
normal force FN, with its direction perpendicular to fretboard and saddle, changes the bearing
force (Fig. 7.11). If the normal forces get larger than the saddle-force, the string lifts off the
saddle – it is imperative to avoid this. Undesirable string movements may however occur
already when the normal force is not quite as big as the saddle force: the fretboard-parallel
tangential force FT acts against the static friction force FR. As soon as FT becomes larger
than FR, lateral movement of the string is possible if the notches in the saddle are too wide. If
this movement happens, vibration energy is irreversibly transformed into caloric energy
(heat), and the duration of the string vibration is shortened. Additionally, interfering noise
may become audible.
As an example we can check out the B-string of a 010-string-set (13-mil). Its tension force Ψ
amounts to 70 N; for a bend angle of 10°, the bearing force at the saddle will be 12.3 N. Let
us further assume that, at a distance of 9 cm from the saddle, the string is moved away from
the fretboard using a perpendicular force that rises from 0 to 5 N. As this plucking force
increases, the saddle mounting force decreases at the same time by 4.3 N from 12.8 to 8 N
(Fig. 7.12). At the moment when the plucking force jumps back to zero, wave movements
start: given a dispersion-free model, the saddle force jumps back and forth between 8 and 13
N; if there is dispersion, additional oscillations occur (Chapter 1.3.1). In this example, we do
not see a negative bearing force, though: the flow of force is not interrupted. However, if we
recalculate using a bend angle of merely 4° (as it is the case for the Gretsch Tennessean), the
static bearing force is merely 4.9 N – now a 5-N-plucking-force will already lead to short
occurrences of lift-off of the string, resulting in a buzzing, less-than-clean sound.
Fig. 7.12: Time function (modeled) of the saddle bearing

force (“Sattel-Auflagekraft”). At t < 0, the string is pulled
away from the guitar; the saddle force thus decreases to 8 N.
At t = 0, the tension force jumps to 0; it takes 1/14 period
(9cm / 126 cm) until this change arrives at the saddle.
If the direction of the plucking force is not perpendicular but parallel to the fretboard, shifts of
the string to the side can occur. Given bearing at the rim of the saddle (Fig. 7.11), a tangential
movement starts as soon as the static friction force is surmounted. In the individual case, the
static friction coefficient depends on the specific pairing of materials – for a rough estimate
we may assume µ = 0,1 ... 0,5. Ideal would be a small friction along the direction of the
string, and a high friction in the perpendicular direction – but this only works if a groove is
employed. Choosing µ = 0,1 and a bearing force of 12.3 N (as we calculated it in the above
example), it shows that already a tangential force of merely 0.1 x 12.3 N = 1.2 N acting along
the rim of the saddle may push the string back and forth on the saddle. A precisely machined
V-shaped notch may prevent this back-and-forth sliding – however not every string bearing
acts as a V-shaped notch!
♣
“Dynamic” is used here in contrast to “static”, and not in the sense of “dynamis = force”.

In the V-shaped bearing (Fig 7.13), the string rests in a defined fashion at two points. The
sideways sliding is prohibited that way, however the static friction in the direction of the
string is about double of what it would be using a rim bearing. The bearing in a threaded
groove – as it is found in old Fender guitars – may result in two different states depending on
whether the radius of the string is larger or smaller than the rounding-off of the groove of the
thread. A bearing within a slot either has a clamping effect or a clearance – there are no
special fits. A rounded-off bearing (fabricated using a round file) is similar to the bearing in a
wide threaded groove – for heavy strings, there may however be clamping effects, as well.
Fig. 7.13: String bearing: V-shape, threaded groove (w/light & heavy string), notch bearing, rounded bearing.
The ideal bearing should prohibit the string to move sideways. At the same time, it should
have the least amount of friction in the direction of the string in order to allow for an
unambiguous tension force. The tension force changes in particular during tuning of the
guitar, and when bending strings. In case a vibrato-system is installed, operating it will also
provide a change of the tension force. To illustrate the effects of bearing friction (in the
direction of the string), we investigate a string with two fixed end-points and a saddle (Fig.
7.14); the main section of the string has a scale length M.
Fig. 7.14: String with scale length M and remaining

section R. “Sattel” = saddle.
The maximum force F that can be applied to the saddle in horizontal direction♣ is the static
frictional force FR (= friction at the bearing). If this limit-value is surpassed, the string will
slide across the saddle. Assuming that we are, force-wise, just below this occurrence, FR acts
onto two spring stiffnesses working in parallel: the longitudinal stiffness of the main section
of the string (with the length M), and that of the residual R. Given smallish bend angles, we
obtain (with S = area of the cross-section) the effective overall stiffness s:
Effective longitudinal stiffness
The change in length of the main section of the string (due to the frictional force) amounts to
. As we move the saddle to the right (in the figure) such that no string slippage
across the saddle occurs, the main section of the string is elongated: its frequency rises.
Moving the saddle to the left, the frequency drops. However, the relative frequency change
does not depend on the relative change in length but on the relative change in strain. A string
of the length M needs to be stretched by x to generate the frequency fG. The strain for a B-
string (247 Hz), for example, is 2.6 mm. Since the frequency is proportional to the square root
of the strain, the relative frequency change corresponds to half the change in relative strain
(differential for small changes).
♣
parallel to the longitudinal axis of the string.

Given strain x, Young’s modulus E, scale length M, and density ρ, the fundamental frequency
fG is:
; . Dependency of the frequency on the strain
The change in strain Δx results from the frictional force♣ that in turn depends, via the frictional
coefficient µ, on the tension force Ψ . The relative changes are:
; Relative changes
To stick to our B-string example: given M = 64 cm, R = 12 cm and µ = 0,15, we calculate a

relative detuning of ± 1,2 %, corresponding to ± 21 cent. Therefore, the actual fundamental
frequency of the string is in fact indeterminate for a saddle having friction, with as the
frequency incertitude. There are three ways to decrease this incertitude: either the friction is
made so strong that changes in the tension force will always remain smaller than the frictional
force – this results in the clamping saddle (clamping nut, clamping bridge) in which the
string cannot budge at all. Or the friction is reduced as far as possible by using low-friction
saddle materials and/or small bend angles – this will increase the danger of relative
movements between saddle and string, though. Or, the remaining string section between
saddle and tailpiece is shortened to just a few centimeters (e.g. Les Paul with Stopbar
tailpiece), and the longitudinal stiffness is thus increased. Besides these approaches, there
have been experiments making the saddle (or the bridge as a whole) moveable in its bearings:
examples are the Fender Jazzmaster and Jaguar … not very successful, getting little love, and
resurfacing more as a cult object. For further data see Table 7.1.
At this point, however, we need to remind ourselves of a fundamental law of sound

generation: individuality is imperfection. Whether a bridge has little attenuation effect on the
string vibration (i.e. long sustain), or much attenuation (i.e. percussive sound) – in the end that
remains a matter of taste. The sitar bridge that can be retrofitted to Telecasters may serve as
particularly succinct example: for it, a special string bearing intentionally generates
“interfering noise”, kind of a “boiiiinnnggg” … not to be mistaken with a “bonk” … you hear
me, Mr. Clinton?
Bend angle at the nut Bend angle at the bridge

Stratocaster 7° – 9° up to 90°
Telecaster 7° – 9° 9° (top loading), 34° (through-body)
Jazzmaster 7° – 9° 6° – 7°
Gretsch Tennessean 5° – 15° 4°
Rickenbacker 335 7° – 12° 4,5°
Gibson ES 335 15° 9° – 10°
Les Paul '59 reissue 17° 19° – 26°
Taylor PS-54-CE 13° 12° / 27°

Ovation SMT 11° 25° – 30°
Ovation EA68 11° 33°
Martin D45V 15° 25° – 30°
Table 7.1: Typical bend angles of strings
♣
The formula for wrap-around friction exp(µα) lends itself for a more precise calculation.

7.5 Reflection and absorption at bridge and nut 7-25
7.5 Reflection and Absorption at Bridge and Nut
Nut (or fret) and bridge act as bearing for the string, defining the limits of its vibrating part.
Their effect onto the sound generation may be looked at twofold: considering the reflection of
a wave propagating along the string (time domain) on the one hand, and on the other hand
regarding the steady state (spectral domain). For both cases only approximations are possible
since the actual processes are indefinitely complex. The bridge is the separating point between
a light, easily moved medium (the string), and a heavy, almost immovable (the guitar body).
While in Chapter 2 we had mainly shown the vibration processes of the string, we shall now
more deeply investigate the string bearings and their effect onto the reflection process.
A technical/mathematical analysis of the reflection process shows that the degree of reflection
amounts to almost 100 % – in sharp contrast to portrayals in (some circles of) popular science
that surmise almost complete matching (i.e. no reflection): "The largest portion of the string
vibration should be transmitted to the body. Indeed, if the latter is supplied with uninhibited
vibration energy, a maximum in tone and sustain develops.[Gitarre&Bass 12/05]" – with this
assumption not being directed to acoustic guitars but to electric ones. However, if indeed the
largest portion of the vibration energy arriving at the bridge would be transmitted to the guitar
body (i.e. the energy would be absorbed in it), then only the small remaining portion could be
reflected … and this would render a vibration of any duration (i.e. any sustain, even a short
one) impossible. This reasoning based on the principle of energy conservation is often met
with the argument that the string bearing would be “re-implying” – meaning that the energy
fed to the guitar body would come back to the string. While that is not entirely wrong, the
wording is somewhat unfamiliar for the acoustician. A bit closer to home would be the
separation into active energy and reactive energy: the reactances (the masses and springs)
store reactive energy that they can release again in its entirety without losses. To compress a
spring, energy is necessary; as the spring is relaxed again, this energy is released. The spring
acts as storage for potential energy. Conversely, all friction resistances absorb active energy
that is immediately and irreversibly transformed into caloric energy: heat. N.B.: even if the
guitar were further heated up, this heat energy could not be re-converted into vibration energy
(translator’s note: just needs to be mentioned with regard to the ill-devised “re-implying”
idea of the guitar body giving back vibration energy as mentioned above). In reality, all
springs are in contact with friction-resistances, and therefore every deformation of material
results in a loss of vibration energy. Thus every body that is deformed by vibration (i.e. that
“resonates”, to use the popular term) will convert vibration energy into heat … not a good
basis for long sustain.
If long sustain is desired in a plucked string, then that largest portion of the vibration energy
must exactly NOT be transmitted to the guitar body – neither as active energy (immediately to
be converted into heat energy), nor as reactive energy. The latter would necessarily lead to a
deformation of spring elements, and since those never are loss-free, at least some conversion
into active energy would happen. It is an entirely different question whether all partials
(harmonics) of a string vibration should have the longest sustain possible. Because not all
guitars are supposed to sound the same, be differences in the spectra of the partials may
actually be welcome – meaning both guitar-specific amplitudes of the partials, and guitar
specific decay constants. In the following we will describe first the reflection process, and
then follow with the analysis of the decay of the partials.

7.5.1 Reflection- and absorption-parameters
The descriptive parameters of waves, and associated reflection and absorption, were already
elaborated in Chapters 1 and 2 – however, assuming that not every reader seeks to struggle
through those chapters, a short summary is given in the following.
Via the plucking/picking of the string, it is deflected with the force F in the transversal
direction. This plucking force delivers the energy that – after release – makes the string
vibrate (oscillate). The unit for the force is the Newton (N) with 1 N = 1 kg ⋅ m / s2 in the mks-
system. To characterize the motional magnitude of the oscillation, the (particle-) velocity is
used (besides displacement and acceleration) – it is not to be confused with the propagation
speed. The product of force and particle-velocity calculates the power P given in units of
Watt (W); 1 W = 1 Nm/s. Temporal integration of power yields the energy E, with the
associated unit Ws = Nm. The typical excitation energy of a string amounts to a few milli-
watt-seconds (mWs). Caution should be exercised with regard to the letter m, since it
represents the abbreviation for both the unit of length meter, and the prefix milli (1/1000th).
The quotient of force and particle-velocity of a propagating wave forms the wave-
impedance: ZW = F / v. However, ZW does not stand for the quotient of the force acting on a
little piece of string and the velocity of this piece of string, since these magnitudes could have
resulted from superposition of several waves. For example, the velocity in a vibration node is
always zero – but that does not mean at all that the wave impedance is zero at that location.
Rather, the node generated in standing waves is the result of two waves running in opposite
directions. The wave impedance is defined only for the individual wave. In steel strings, the
wave impedance is about 1 Ns/m; more exact data may be found in the appendix.
As a propagating wave hits an obstacle, part of the wave power is reflected. An obstacle is
given by a location in the medium the wave travels in where the wave impedance differs from
that of the string, as is the case in particular for the string bearings (bridge, and nut or fret).
The degree of reflection denotes the portion of the wave power that is reflected – the non-
reflected portion is absorbed. Besides the degree of reflection there is also the reflection factor
that is required for calculations involving force and velocity. If, for example, 25% of the wave
power is reflected, this implies that the force of the backward-running (reflected) wave is 50%
of that of the forward-running wave. In this case, the particle-velocity of the backward-
running wave is also 50% of that of the forward-running wave, since both waves are
connected via the wave impedance (see above). From this we get: the degree of reflection is
the square of the reflection factor. The degree of reflection and the degree of absorption add
up to 1. Thus, 25% degree of reflection pertains to 75% degree of absorption
The formula representations for reflection factor and degree of reflection are not always
handled uniformly: often r is used for both. To avoid mix-ups here, we will use r for the
reflection factor, and r2 for the degree of reflection. Alternatively, the terms “degree of energy
reflection” rE and “degree of power reflection” rP are made use of – the magnitudes of both
quantities are equal.
Due to the law of energy conservation, the degree of reflection cannot be larger than 1. In
guitars, values of just shy of 1 (e.g. 99%) are typical. In some frequency ranges, however,
there may be much absorption – the vibration energy is efficiently transmitted into the
bearing, and it either is converted directly into heat, or (partially) radiated as sound.

7.5.2 Analysis of reflections
In the simple model of the string (Chapter 1), two transversal waves propagate along the
string after it is being plucked (Fig. 7.15, left). Let us take as stimulus a short impulse (at A)
that – in the ideal case – is reflected at both bearings with opposite phase. The corresponding
idealized time function at the observation point (B) is seen on the right. In reality, we get
large deviations from this idealized model: the propagation of transversal waves is dispersive
(Chapter 1.3), and at the bearings there is both a loss of vibration energy and an exchange of
energy between transversal waves and longitudinal waves. This exchange is the subject of the
following elaborations based on measurements that targeted the development of a more
precise reflection model.
Fig. 7.15: Simple model for transversal waves: impulse excitation at A (left); time function at B (right).
If the time function appearing at B (in Fig. 7.15) were as simple as depicted in the graph, it
would be easy to distinguish between the wave running forwards and the wave running
backwards, and to determine a reflection factor. However, due to the frequency dependency
of the group delay (dispersion), short impulses get broadened already after a few centimeters
into a chirp such that we get a mix of the two waves running to and from (Fig. 7.16).
Fig. 7.16: Forwards- and backwards-running waves at point B, with dispersion (left); superposition at B (right).
If the requirements with regard to accuracy are not too high, it is possible to de-convolve one
of the two waves, and in the end to obtain a separation with respect to time, after all. To do
this, the string is interpreted as a linear, time-invariant system (Chapters 1 and 2), the output
signal of which is a convolution of input signal and impulse response. For small
displacements, assuming linearity provides a very reasonable approximation. Time-invariance
can be obtained only for short durations of time due to unavoidable temperature changes.
From output signal and impulse response, the de-convolution (as the inverse operation of
convolution) yields the input signal. The latter could be a short impulse (at A), while the
output signal is the measurement result determined at B (e.g. the particle velocity), and the
impulse response can be calculated on the basis to the string data (Chapter 1). However, as
elegant as this approach may be in theory, in practice we quickly recognize new problems: the
signal energy delivered via a short impulse is small, and therefore a lot of noise but only little
useful signal arrive at the measuring point. The de-convolution is correspondingly impaired.

Fig. 7.17: Summation signal at B (left), de-convolution (right). Both curves are calculated, not measured.
“Anregung” = excitation; “reflektierte Welle” = reflected wave.
Fig. 7.17 indicates how all this would work in theory: via de-convolution of the output signal
at B (left section of the figure) with the inverse impulse response that maps from A to B, the
(positive) output impulse is retrieved. The reflection is added in – it is not fully de-convoluted
because it has run a longer distance, after all. After separating the signal, though, the
reflection could also be de-convoluted back right up to the excitation impulse.
However, as has already been mentioned, excitation with a short impulse has its problems: the
useable signal dynamic is small. We get better results if we excite the string at point A using a
special chirp, the mapping of which onto B results in a short impulse. The excitation signal A
thus needs to be configured exactly such that its dispersive mapping onto B yields a short
impulse that can easily be separated from the subsequent reflections (Fig. 7.18).
Fig. 7.18: Excitation chirp at A (left), output signal at B (right); model calculation.
In contrast to the model calculations elaborated so far, we now turn to measurements with an
experimental setup. A string a length of 13 m length (translators note: no kidding!) made of
steel wire at 0.7 mm diameter was tensioned to yield a fundamental frequency of 6 Hz. It was
deflected transversally with a chirp using a shaker (B&K 4810). Since the transmission
characteristic of the shaker is frequency-dependent, the chirp depicted in Fig 7.18 needed to
be pre-filtered once more, such that the forward-running wave shaped up to a half-sine
impulse at a distance of 3 m. This can only be achieved as an approximation because a time-
limited impulse would require a infinitely broad spectrum; bandwidth-limiting to technically
feasible ranges leads to a broadening of impulses – to spurious oscillations, specifically. After
some lengthy optimization work, a usable compromise could be found that delivered reliable
measurement results, after all.
The experimental setup is shown in Fig. 7.19: the 13.3-m-long string runs across a knife-edge
bearing and to its end-fixture 10 cm away. The bend angle at the knife-edge is adjustable. Via
a chirp generated by a B&K 4810 shaker, the string receives excitation to oscillate
transversally. At about 20 cm distance from the knife-edge, the transversal velocity of the
string is measured using a laser vibrometer. The chirp is calculated such that the transversal
wave running towards the right generates a velocity impulse of half-sine shape.
Fig. 7.19: Reflection-measurement setup. Length of the residual string (“Saitenreststück”): 10 cm; distance
between knife-edge bearing (“Schneidenlager”) and access point for laser beam (“Laserstrahl”): about 20 cm;
distance between measuring point and shaker: about 3 m. The string (“Saite”; ∅ 0.7 mm, overall length 13.3 m)
is tuned to a fundamental of about 6 Hz. Right: time function of the velocity measured via the laser vibrometer.
The scaling in Fig. 7.19 (right-hand section) is such that the (negative) excitation impulse
happens around 0 s on the time axis, accompanied by some minor spurious oscillations. From
about 1 ms, the dispersively broadened reflection becomes visible, and from about 3 ms,
vibrations caused by the residual part of the string become apparent. Only for a string without
any bending-stiffness would the two sections of the string remain completely decoupled via
the motion-free bearing. Given a string with bending-stiffness, a coupling of the vibration
does happen with such a bearing, as well (Chapter 2.7). The bearing is able to take on
transversal forces but cannot absorb the bending moments that also occur in a flexural wave –
these moments are transmitted across the bearing to the respective other section of the string.
The length of the residual string (98 mm) corresponds to a fundamental frequency of 920 Hz;
at this frequency and its “multiples”, energy is withdrawn from the reflection. The term
“multiple” should not be taken entirely literally here because the Eigen-frequencies are spread
out dispersively. In Fig. 7.20, the de-convolved impulses running back and forth are shown,
and also the spectra for the two functions. Apart from the many ripples representing analysis
artifacts, two peculiarities can clearly be recognized in the spectrum of the reflection: the
unexpectedly high damping (attenuation) of the reflection (on average about 1 dB), and
selective minima at the Eigen-frequencies of the residual part of the string.
Fig. 7.20: De-convolved reflectogram (left): excitation spectrum (----), reflection spectrum (––––). On the right,
the velocity-spectrum of the vibration of the residual string is represented as the thin line.

Previous experiments had already shown that a bending-coupling was happening across the
knife-edge bearing, and that vibration energy was withdrawn from the string selectively
(Chapter 2). It is, however, surprising that the energy of the wave reflected by the bearing is
20% less than the energy of the wave running towards the bearing. If indeed the wave would
loose 20% of its energy at each reflection, a 500-Hz-vibration would be attenuated by 48 dB
already after 0.1 s – this contradicts any real experience. Justifiably, the spectral distribution
of the reflected energy may be questioned because every short-term spectrum will include
artifacts. The time-function of the velocity measured via the laser (Fig. 7.19) is much more
reliable, though. If the degree of energy-reflection remains so clearly below the expected
value (assumed at possibly 98%), then there are two possibilities for errors: either the
measurement of the excitation energy yields is too high a value, or that of the reflection yields
one too low. An error regarding the excitation energy can by and large be excluded: the
summation of the squared signal happens across merely one 1 ms – at that point no reflection
has happened. For the reflected energy, however, the situation is different: theoretically it
would be possible that the excitation is not concentrated in the impulse shown in the figure,
but includes an additional component that is in opposite phase to the subsequent reflection.
This highly unlikely scenario could be excluded, as well, by modifying the parameters: all
measurements gave a degree of energy-reflection of 80%, calculated from the time function.
Changing the bend angle of the strings (across the knife edge) shows us where the
seemingly lost energy goes. For the first measurements this angle (that indicates the change of
direction of the string at the knife edge) amounted to α = 7°. In a second series of
experiments, α was increased to 23°, resulting in a further reduction of the degree of
reflection to only 68%. However, for these measurements, 5 ms after the first reflection a
further, smaller reflection appears, and this provides the key to understanding. The overall
length of the string used for the measurements was 13.3 m: with vD = 5200 m/s, an expansion
wave (longitudinal wave, dilatational wave) needs 5.1 ms to run back and forth two times
(without dispersion!). Evidently, the transversal wave arriving at the knife-edge bearing is
transformed (to a non-negligible degree) into a longitudinal wave that runs quickly and
dispersion-free to the other end of the string. There, it is reflected partially as longitudinal
wave, and partially as transversal wave. Crucial in the present context is the reflected
longitudinal wave: it returns after 5.1 ms and releases a secondary transversal wave at the
knife-edge bearing. With the laser only detecting transversal waves, the longitudinal wave
remains invisible. That a longitudinal wave is in fact generated is easily verified using a force
sensor measuring the longitudinal force at the end of the string: indeed, exactly between the
two reflections depicted in Fig. 7.22, we get an impulse of longitudinal force (not included in
the drawing, see Fig. 7.29).
Fig. 7.21: As Fig. 7.20 but with the bend-angle of the string increased to α = 23°.

The results so far can be summarized as the following insight: the stronger the string changes
direction at the bearing (i.e. the larger the bend angle is), the more efficiently the
transformation from transversal to longitudinal vibration energy will be at the bearing. More
concisely: the more pronounced the bend, the stronger the mode-coupling. Two more
experiments are to deliver support for this hypothesis: in Fig. 7. 22, the bend angle was
enlarged to 45° – a value typically found in a Stratocaster. The degree of reflection decreased
to 59%, the 5-ms-reflection becomes even stronger.
Fig. 7.22: As Fig. 7.20 but with the bend-angle of the string increased to α = 45°.
As the other extreme, the measurement doing away with any bend angle concludes the
experiments – the string simply ends in a heavy brass block without previously crossing a
bridge (Fig. 7.22). The degree of reflection calculated from the measurement is 99%, with a
5-ms-relection not identifiable. All these analyses allow but one conclusion: for the vast
majority of electric guitars, the wave reflection at the bridge does not bother about any
textbook teaching! The function of the guitar bridge, ideally assumed to be a stiff bearing or,
more realistically, modeled as transverse impedance (or admittance), is more complex than
assumed so far. On top of its dispersive characteristics that have not been examined yet in the
present context, it acts as interface between transversal and longitudinal waves – with a
degree of energy coupling up to 40% … not a negligible order of magnitude anymore. What
follows for the sound of the guitar is this: the string vibrates not only with the Eigen-
frequencies (natural frequencies) for the transversal wave but also with those for the
longitudinal wave, and in addition mixed modes are possible, as well (e.g. a wave running in
one direction as transversal wave and backwards as longitudinal wave). In the spectrum, we
therefore expect deviations relative to the Eigen-frequencies of the rigid string.
Fig. 7.23: As Fig. 7.20 but without any bend angle of the string.

Fig. 7.24 once more compiles all four of the measured reflectograms. Established scientific
consensus is that reflection needs to happen with opposite phase – this is fully confirmed.
However, the dependency of the amplitude of the reflected impulse on the bend angle of the
string requires an extension of the usual models by a bearing-specific mode-coupling.
Fig. 7.24: Compilation of all reflectograms. α = bend-angle of the string; Saite, rE = degree or energy-reflection.
Checking out popular guitar model shows that the manufacturers take advantage of the
possibilities of varying the bend angle of the string: the range goes from about 4° to 45°. For
the purely acoustic archtop guitars with their convex tops, choosing the bend angle was kind-
of walking the tightrope between string buzz and a collapsing top: the more pronounced the
bend angle, the bigger the forces acting onto the top. Given an (overall) string pull of 1000 N
it is clear that the bend angle must not be too big. Some electric guitars adopted the basic
build of the arched, unsupported top (Byrdland, ES-330, Tennessean, Casino), and imported
the shallow bend angle, as well. With the introduction of the reinforced top (e.g. the “sustain
block” in the ES-335), stability problems should in fact have been part of the past – but during
the 1960s and 1970s, the ES-335 sported the trapeze tailpiece (a string fixture extending to the
endpin, similar to that of the ES-330). Probably, this was for cosmetic reasons. For solid-body
guitars such as the Jazzmaster and the Jaguar, the reason for the shallow bend angle was the
vibrato system that was supposed to be free of detuning (with the help of the point-support
bearing of the bridge). At Fender, they could have calculated that this construction would not
be the best solution – but they left the market to decide. Which the marked did – without
much mercy, even though these guitars were Fender’s top-of-the-line models. In its first
edition, the Telecaster had a steep bend angle (string-through-body), but in 1959 this was
changed towards a smaller angle (top-loading). In 1960, both versions were available and
after that, the steep angle was back again exclusively. One of the most pronounced bend
angles is found on the Stratocaster and its countless copies.

Fig. 7.25 shows some examples of typical bridge designs. The Stratocaster bridge is available
with bridge saddles of bent sheet-metal, or with solid bridge saddles, and also with vibrato or
without it (“hard tail”), and in other variations not shown here. Gibson’s mainstay, the Tune-
O-Matic bridge is found in two versions: with the heads of the adjustment screws pointing
towards the headstock, or towards the tailpiece. For the latter configuration (pictured on the
left), the residual strings can make contact to the screw-heads! The bridge variant shown for
the 1952 Les Paul is the one found in Lester Polfuss’s patent application; the production
model, however, wrapped the string under the bridge. The bridge of a 1952 Les Paul is solidly
anchored into the guitar body; it could be almost seen as a solidly fixed bearing. However,
since the strings could not individually be adjusted in length, Gibson introduced their Tune-O-
Matic bridge in 1954 for almost all their guitars.
Fig. 7.25: Cross-sections of some selected guitar bridges.

We can see from Fig. 7.25 that the bend angle of the strings turns out rather differently
between the various guitar types. Moreover, the length of the residual section of the string,
i.e. the section between bridge and tailpiece, needs to be considered – it does make for
selective absorptions. The latter are not particularly conspicuous in themselves but need to be
mentioned for the sake of completeness. From the relationship between the length of the
remaining string and the scale, it would be easy to estimate the resonance frequencies of the
residual strings: at these frequencies, energy is withdrawn from the main section of the string
– due to the bending-stiffness coupling across the bridge. The mentioned relationship is
specific to the guitar type (Fig. 7.26), and sometimes even specific to the year of manufacture.
If the length-relation is e.g. 10%, the fundamental of the resonance provided by the residual
part of the E2-string will be 824 Hz. In practice, there will be small deviations, because, for
one, the bending stiffness makes for a stronger detuning for shorter strings, and also because
the mounting of the strings in the tailpiece is not ideally rigid. Fig. 7.26 depicts, as a typical
example, the trapeze tailpiece of the Gibson ES-335; the table on the right gives an
impression of the common length relationships.
Length or residual part of string vs. scale length

Gibson ES 140 6%
Gibson ES 330 10 %
Gibson ES 335 17 %
Fender Jazzmaster 19 %
Gibson Byrdland 20 %
Fig. 7.26: Typical bridge with a long residual part of the string (Gibson ES 335 w/trapeze-tailpiece).
Already the few examples shown above indicate how different the build of guitar bridges may
be, with corresponding differences in their influence on the string vibration. The
measurements document unequivocally that a significant coupling of modes happens at the
bridge. In the following we shall investigate in depth why this happens. In particular, the bend
in the string requires a more exact documentation. The bend angle merely specifies the
tangential (asymptotical) directions but not how the bend evolves locally across the bridge
saddle. There are two extreme cases: either the string is smoothly bent across the bridge in a
round just as much as is required to achieve the bend angle, or a sharp bend is introduced to
the string at the bridge. The former leads to a reversible deformation in the case of not too
big a bend, the latter brings an irreversible kink (plastic deformation) that remains even after
removing any tension. Fig. 7.27 represents both these cases for a 45° bend angle.
Fig. 7.27: Rounded-off and sharply-

bent shape of the string for identical
bend angles (α = 45°).
For the two resulting shapes of the string, the reflection behavior is different: the rounded off
shape generates a stronger coupling of modes. Apparently it is not so much the change in
direction of the string but the curvature that is of significance. In the right-hand section of the
figure, a strong curvature is present only directly at the bridge – but here the vibration
amplitude is practically zero. In the left section of the figure, however, the curvature extends
across a longer section of the string, albeit in a weaker fashion.

Reflectograms corresponding to the string geometry in Fig. 7.27 are shown in Fig. 7.28. The
primary reflection (at 1 ms) is, compared to the more softly bent string, more pronounced for
the sharply bent string. Assuming that only little energy is dissipated within the bearing for
any single reflection, a weaker longitudinal wave is thus generated for the sharply bent string.
The secondary reflection occurring after 6 ms qualitatively confirms this, although it has gone
through an additional longitudinal/transversal coupling after the transversal/longitudinal
coupling, and therefore is not a direct measure for the dilatational wave.
Fig. 7.28: Reflectogram for two different string curvatures; bend angle α = 45°.
The simplest way to measure the dilatational wave generated at the bearing via the mode-
coupling is using a force sensor that gauges the longitudinal force at the other end of the
string. What needs to be considered, though, is that at a fixed bearing the motional
magnitudes become zero while the forces are doubled (Chapter 2.2). Therefore, Fig. 7.29
indicates all bearing forces only with 50% of the measured value (hoping, of course, that the
bearing indeed is ideal). In order to exclude the influence of any residual part of the string,
this was eliminated by anchoring the string rigidly in a rotatable cylinder. Twisting the
cylinder enabled the creation of a bending moment resulting in the curvature of the string. The
measurement point of the laser was located at a distance of 38 cm from the cylinder, and at
another 3 m distance the shaker was positioned. Force measurements were carried out at the
other end of the string where the bearing caused no curvature.
Fig. 7.29: Transversal velocity (top) and longitudinal force (bottom) for three different string curvatures.
Note: the polarity of the force signal depends on the direction of the curvature.

The measurements shown in Fig. 7.29 are impressive evidence that the mode-coupling is not
the result of the presence of the residual string, but the effect of the string curvature alone.
Without curvature, there is next to no longitudinal wave. If a curvature close to the bearing
exists, a unipolar half-wave impulse generates a bipolar impulse that is reminiscent of a
temporal differentiation. “Force = mass x differentiated velocity” would appear to be an
obvious train of thought – however, the explanation proposed here points into a different
direction: it’s not the infinitesimal differentiation that brings us to our goal but the difference
– more specifically: the superposition of the time-shifted, opposite phase signals. The
velocity-wave reflected away from the bearing is in opposite phase to the wave running
towards the bearing such that directly at the bearing the sum of the two is zero – the ideal
bearing indeed is still. Just ahead of the bearing, however, the two waves are shifted relative
to each other by a small delay time. The superposition results in a signal approximately
corresponding to the differential. In Fig. 7.30, a correspondingly calculated superposition is
juxtaposed to the measured longitudinal force. The basic shape is nicely present. It is
understandable that not all small peaks are modeled: for one, band-limiting and de-
convolution make for artifacts; more importantly, though, the calculation was only done for a
single point: at 1 cm ahead of the bearing. In a real string, however, the mode-coupling does
not only happen in a single point but across a range.
Fig. 7.30: Measured longitudinal force (––––), compared to a simple model calculation (-----).
On the right, the coupling between longitudinal and transversal movement is depicted.
The principle mechanism of the mode-coupling is represented by the picture on the right:
pulling the string to the left at the same time results in a downward movement.
Correspondingly, a downward transversal displacement makes for an additional displacement
towards the left. The dependency on the sign was already shown in Fig. 7.29 – model and
calculations agree well in this respect.
Fig. 7.30 shows an amplitude of the longitudinal force (force-amplitude of the dilatational
wave) of 0.3 N. The particle-velocity of the dilatational wave is connected to this via the
impedance of the dilatational wave ZW = 16 Ns/m, yielding a velocity amplitude of the
dilatational wave of 19 mm/s; that is about 7% of the velocity amplitude of the transversal
wave. The energy of a wave depends on the square of the velocity but also on the wave
impedance: E = ZW ⋅ v2 ⋅ t. The wave impedances for longitudinal and transverse waves differ
by the factor of 30 in this case – the relation between the velocities mentioned above would
therefore indicate that the energy of the longitudinal wave is about 30% of that of the
transversal wave (double impulse, factor of 2!). The energy-reflection reflection, at rE = 65%,
is a little smaller in this example – the missing 5% could have been “lost” when the
longitudinal wave ran across the shaker. Given the multitude of possible artifacts, this degree
of precision is deemed as good.
As a summary: the more the string is curved, the higher the coupling transversal waves ⇔
longitudinal waves is. The factor of the energy coupling may be as big as 40% or more. The
time functions of the waves generated by the coupling are, approximately, the derivatives of
the time functions of the primary waves.
Fig. 7.31 summarizes the time-processes. A transversal velocity wave of half-sine shape is
(partly) reflected as a sine-shaped longitudinal velocity wave. The latter is again (partly)
reflected as a transversal velocity wave with 1.5 sine-periods. Using the derivative as an
approximation becomes increasingly inaccurate as more steps in the mode-conversion are
simulated. The spectral operation matching the time-derivative is a multiplication with jω, i.e.
a bass-attenuation. That is why the mode-coupling has no impact for the real string in the low-
frequency domain – only above 1 kHz, effects become noticeable.
Fig. 7.31: Primary transversal wave (left), dilatational wave generated by it (middle), transversal wave generated
in turn by the latter (right). All graphs are measurement results with a solid 0.7-mm-string.
For the following analyses (concentrating on spectral effects) the scale of the string was
shortened to guitar-typical values; this was in order to get into the range of guitar-typical
resonances. To enhance the effects, the string was first clamped down in a fixed manner on
both ends, but then alternatively run across a knife-edge at both ends (i.e. it received a bend).
The measurement with the laser-vibrometer was done at a distance of 7 mm from one of the
bearings. Close to the other bearing, the string was excited with a short transversal impulse.
The spectra (Fig. 7.32) will therefore show a comb-filter-like shape that depends on both the
shape of the excitation impulse, and on the distance from bearing to measurement point. The
high degree of correspondence between measurements and transversal-wave model indicate
that for the clamped-down string, barely any dilatational waves have been generated. The
spreading of the partials is the regular one, i.e. it corresponds to the theory introduced in
Chapter 1.3.
Fig. 7.32: Level-spectrum of a clamped-down 0.7-mm-string; measurement (“Messung”, left) and model
calculation (“Rechnung”, right). The measured frequencies of the partials perfectly agree with the calculation
(dots in the left-hand graph).

For the string run across the knife-edge bearings, the frequency-deviations relative to the
regular spreading are evident (Fig. 7.33) – this is due to a stronger coupling of the mode. Both
a detuning of the partials, and additional partials (e.g. around 3 kHz), can be seen.
Fig. 7.33: Measured spectrum of the clamped-down string (left), and of knife-edge-supported string (right). The
irregularities of the frequencies of the partials are due to the coupling of modes.
Fig. 7.34 depicts two sections from the spectrum of the knife-edge-supported string. In
contrast to the clamped-down string (where the frequencies of the partials agreed with the
calculated values in good approximation, Fig. 7.32), considerable deviations between
calculation (dots) and measurement (line) are evident.
Fig. 7.34: Sections from the spectrum shown in Fig. 7.33 (right), for the knife-edge-supported string.
Of course, the question whether such effects are audible is of particular importance.
Orientating listening experiments showed that all recorded velocity data sets sounded
differently – however this was not due to the different string bearing but due to the non-
identical excitation. Already minute differences in the plucking/picking of the string changed
the spectral envelope to such an extent that differences could be heard. Therefore two sounds
were synthesized that had identical spectral envelopes but different frequencies of the partials
as given by Fig, 7.33. The result was that the differences in the inharmonicity are rather
insignificant to the sound. Of course, it will depend on the individual case whether special
beats in the partials shape the auditory impression in a particular fashion. However, from a
general point-of-view, the way the string is plucked very clearly takes precedence over the
inharmonicity. (See Chapter 8.2.5 for the audibility of inharmonicities in the partials).

7.5.3 The mechanical bridge-impedance
Interpreting string and saddle (bearing of the string)♣ as parts of a mechanical system, two
different paths offer themselves towards a mathematical description: either the space- and
time-dependent analysis including differential- and wave-equations, or the spectral
presentation of the post-transient, settled condition. The previous chapter had analyzed the
processes in terms of their evolution over time, below we now get into the frequency domain.
For the simple model of the string, the saddle is motionless and reflects ideally, with
reflection-factor r emerging to be +1 (force) or -1 (motion), respectively. From the real wave-
impedance ZW of the string, and the complex saddle-impedance (bearing impedance) ZL, r had
been calculated in Chapter 2.5. To keep the relations manageable, we shall look at a single
wave-type (e.g. the transversal wave), and a loss-free bearing. The saddle-impedance ZL = F/v
is now a two-pole function of the reactance. The reactance is the imaginary part of a
complex impedance. A reactance two-pole includes an imaginary impedance only, and
therefore only masses and springs are allowed as elements – damping resistances are not.
Given the complex frequency p [e.g. 6], the mass impedance is calculated as pm, and the mass
reactance as ωm, respectively; the spring impedance is s/p, and the spring reactance is –s/ω. If
a mechanical reactance two-pole is comprised of masses and springs (loss-free, there will be
no other elements), its impedance is given by a reactance two-pole function of the form:
Reactance-two-pole function
Herein, zi are the coefficients of the numerator, and ni those of the denominator. The largest
power in the numerator-polynomial (ν), is either larger by 1 than that of the largest
denominator-polynomial (µ), or smaller by 1. The larger value of ν and µ corresponds to the
order n of the system – in canonical systems it would be the number of free memories.
Example: a system containing but one single spring is a 1st-order system. It would already be
possible to describe an ideal bearing that way, but the spring-stiffness would then need to be
infinite (a spring of infinite stiffness is unyielding). Given finite stiffness, an approximation is
possible as long as the bearing reactance remains large relative to the wave-impedance. With
s = 106 N/m we get, at 1 kHz, a bearing reactance of –160 Ns/m; this may be seen (in terms of
the absolute value) already as large compared to the wave impedance of a steel string (0,1 ... 1
Ns/m). A bearing with this kind of suspension would reflect low-frequency waves almost
exactly as a rigid bearing would. Mounting a small mass ahead of this bearing spring, and
another spring ahead of this small mass, a 3rd-order system would emerge. At low
frequencies, this new system would behave in a spring-inhibited manner, and at high
frequencies the same. Between the two resonance frequencies, however, it would be mass-
inhibited (inert). The frequency dependency of the bearing-reactance has two effects: a de-
tuning of the frequencies of the partials, and the generation of additional partials that would
not exist with an ideal string bearing (Chapter 2.5.2). The higher the order of the bearing
impedance, the more additional partials are created.
What is the typical order of magnitude of a bearing-impedance? It should be infinite, since

nut, bridge and body are continua – but in practice it is finite, after all, because we regard
merely a finite frequency range.
♣
As already noted on page 7-21, the term „saddle“ is used here generally for the bearing of the string, i.e. for
both nut and bridge – and for the respective fret, as well, in case of a fretted string.

The theorems about reactance-two-pole functions [e.g. 7] say that along the jω-axis, poles and
zeroes alternate. Between the poles and zeroes, the mechanical reactance-two-pole (i.e. the
“reactive” string bearing) behaves either like a spring or like a mass; the partials of the string
are correspondingly detuned higher or lower (Chapter 2.5.2). One pole-zero-pairing each (on
the positive imaginary axis) generates an additional partial; given n = 8 we therefore get
already 4 additional partials. Since all parts of the guitar are force-fitted to each other, we
would in theory have to account for a whole lot of vibration-happy sectional masses and
springs, and consequently would have to deal with a high number of additional partials. That
is for the loss-free bearing, though! As soon as we grant resistive elements to the bearing, the
situation changes from the ground up: only those resonances that are extremely weakly
damped can change the phase by 2π, and generate additional tones. All other resonances only
result in small frequency shifts.
The mechanical impedance of a lossy bearing is not a reactance-two-pole function but a two-
pole function, i.e. a real, rational and positive function of p. All poles and zeroes of the
bearing impedance Z(p) are located left of the jω-axis. Mapping Z(p) onto the complex
reflection factor r(p), we do not obtain a pure all-pass function – rather, phase and damping
are frequency dependent.
Complex v-reflection-factor
The bearing impedance can be expressed as a rational function Q / V; from the n poles and
zeroes, the mapping generates n new poles and zeroes – the order n is retained (for real wave-
impedances). From W⋅V + Q, we obtain via zeroing the poles of the reflection factor – all
positioned left of the jω-axis (stable system). From W⋅V – Q, we obtain the zeroes of the
reflection factor, but these may now be distributed across the whole of the p-plane! If r zeroes
are located on the jω-axis, matching will occur at the corresponding frequency: the
reflection-factor is zero – the bearing absorbs the whole of the wave energy. If r zeroes are
located right of the jω-axis, the reflection factor contains (inter alia) an all-pass. If r zeroes
are located left of the jω-axis, the reflection factor is without an all-pass (= of minimal
phase).
Slightly simplifying: each resonance of the bearing results in a pole/zero-pair of the reflection
factor. The poles of the reflection factor are always located within the left p-half-plane; the
zeroes of the reflection factor may be located left or right. A zero on the left merely causes the
detuning of a partial, while a zero located on the right will generate also an additional partial.
A resonance circuit (i.e. a spring/mass/damper-system) within the bearing will make for a
narrow-band absorption of vibration energy; it decreases the magnitude of the reflection
factor from 1 to values just shy of 1. If the resonance circuit is of minimum phase, there will
only be a small phase shift: below the resonance, the phase of the reflection will be slightly
negative while above the resonance, it will be slightly positive. With increasing distance from
the resonance frequency, the phase shift decreases towards zero. However, in case the
resonance circuit contains an all-pass (i.e. it is not of minimum phase), it will shift the phase
with increasing frequency by -2π, generating an additional partial. The basics for this
approach to looking at things are found in systems theory [6, 7]; the reflection process can be
understood as mapping characteristic of a linear and time-invariant system.

Let us take a little detour to explain the process of a reflection for an electric transmission line
– subsequently we shall then look at the corresponding analogy for the mechanical line. An
electric transmission line with a wave-impedance of W = 50Ω [5] is terminated with an
RLC-series-circuit, for example R = 25Ω, L = 0.1H, C = 1µF. A reflection factor r follows:
The zeroes of the numerator-polynomial are located in the right-hand p-half-plane, i.e. the
reflection is not of minimum phase but includes an all-pass. If we modify the three elements
of the termination-impedance to be R = 100Ω, L = 0.2H, C = 0.5µF, this will not change the
magnitude of the reflection factor. The phase, however, does change: a minimum-phase (i.e.
all-pass-free) system results. Fig. 7.35 depicts the corresponding magnitude, phase, and locus
of the reflection factor. For minimum-phase, the r-locus runs left of the coordinate-origin,
while for the solution comprising an all-pass, the solution encircles the origin.
Fig. 7.35: Reflection factor (“Reflektionsfaktor”): frequency response of magnitude (“Betrag”) and phase, and
locus (“Ortskurve”); ---- = minimum-phase.
Now on to the mechanical transmission line: the string. Given a dispersion-free situation, the
wave impedance is again real, and the termination (bearing) impedance is a two-pole-
impedance. As an example, a 5th-order bearing consists mainly of a stiff spring but also
includes two small masses, two springs and four dampers (Fig. 7.36). It is not directly
obvious whether or not the mechanical system generates an all-pass-free reflection; both
characteristics may result from the same structure – merely the component-values differ.
Fig. 7.36: Magnitude and phase of the reflection factor (“Reflektionsfaktor”) for two sets of component values.
The frequ. responses of the magnitude (“Betrag”) are identical, those of the phase differ (---- = minimum-phase).
V s m W s0 V S m W
3,98 Ns/m 2859 N/m 0,36 g 0,1 Ns/m 5087 N/m 19,1 Ns/m 2054 N/m 1,43 g 0,17 Ns/m
25,6 Ns/m 7498 N/m 0,33 g 0,005 Ns/m 2221 N/m 34,9 Ns/m 281 N/m 0,244 g 0,048 Ns/m
Table: Component values. The last line results in a reflection comprising an all-pass.

From the point of view of function-theory, mapping the complex Z-plane onto the complex r-
plane is a conformal mapping. Normalizing the bearing impedance Z relative to the wave-
impedance W, we obtain – with z = Z/W – the following normalized conformal mapping:
Fig. 7.37: Conformal mapping of the normalized Z-plane onto the r-plane.
“Minimalphasige” = minimum phase; “Allpasshaltige” = comprising an all-pass; “Reflexion” = reflection;
The unit-circle of the r-plane is a mapping of the imaginary axis of the z-plane – the left-hand
z-plane is mapped to the exterior of the unit circle. Because all real bearing-impedances are
two-pole functions with a non-negative real part, the magnitude of the reflection factor cannot
become larger than 1 (this is mandatory from an energy point-of-view, as well). The band of
the z-plane hatched in grey is mapped to the grey area of the r-plane; the range of Re(z) > 1 is
mapped onto the pale-ish/thin circle. For Re(z) > 1, Re(z – 1) > 0 holds, i.e. it is a two-pole
function including zeroes in the left-hand r-plane (thus of minimum phase). In the case of the
electrical transmission line (the above example), z is the straight line z = 0.5 for R = 25 Ω. It is
located in the grey area – the reflection therefore comprises an all-pass. For the mechanical
line (string across a bearing), Fig. 7.38 shows the loci – including one peculiarity:
Fig. 7.38: Loci for the mechanical line; compare to Fig. 7.36. All loci are run through clock-wise with increasing
frequency. “Normierte” = normalized; “Ortskurve” = locus; “Impedanz” = impedance; “Minimalphasige” =
minimum phase; “Allpasshaltige” = comprising an all-pass; “Reflexion” = reflection; “Admittanz” = admittance.

Between 100 Hz and 650 Hz, the normalized impedance-locus runs to the right of the dashed
delimitation line; the corresponding reflection factor lies within the small circle. The two
maxima of the reflection damping (at 160 Hz and 400 Hz) cause two small loops in the r-
locus – they are located within the small circle and therefore have minimum-phase
characteristic. Globally seen, though, the phase of the v-reflection-factor changes from π to 0,
which is a characteristic of every spring-type bearing (s0). For frequencies approaching zero,
the system shown in Fig. 7.36 acts spring-like; the impedance thus shows a pole at p = 0 (and
a zero at p = ∞). If we wanted to avoid this peculiarity, the bearing would have to be designed
to have an essentially resistive characteristic (i.e. it would have to be a damper); however, this
setup would not enable the system to absorb any pre-load force. Therefore we have a spring-
type bearing, and consider all-pass characteristics only within the relevant frequency range.
Fig. 7.38 also contains the locus of the normalized admittance Y = G + jB = 1/Z. The real
part G of the admittance is termed conductance, the imaginary part B is called susceptance.
Whether you will want to work with the impedance (and its components resistance and
reactance), or with the admittance (and its components conductance and susceptance) is a
matter of taste; the conversion from one world to the other is simple. When calculating the
absorption in a bearing, the admittance yields the shorter formula – that’s why it will be used
in the following. The power absorbed in the bearing is not available anymore to the reflected
wave – every bearing will cause, besides phase shifts, an absorption (i.e. damping). The
degree of absorption a2 tells us the relative portion of the effective power irreversibly
absorbed in the bearing:
2 2
a = degree of power-absorption, r = degree of power reflection.
Both these magnitudes depend on the corresponding factor with a square-relationship: if, for
example, the reflection factor is r = 50%, then 25% of the power gets reflected, and 75% gets
absorbed. The effective power PW absorbed by a two-pole may be represented in four ways:
Effective power
For the string bearing, the degree of absorption in the bearing computes as:
Degree of power-absorption in the bearing
Assuming a relatively stiff bearing with small conductance and small susceptance, the power
absorption is proportional to the conductance. Measurements carried out by Fleischer [2006]
show that at least the conductance mostly remains below 0,01 s/kg – which is small compared
to the inverse of the wave-impedance of customary strings (1 – 10 s/kg). Fleischer does not
explicitly specify measurements regarding the susceptance, but the order of magnitude is
comparable for circle-shaped loci. In the example presented in Fig. 7.38, the conductance in
the minimum-phase system reaches values just above 0,15 s/kg; however, the absorption is
chosen to be untypically large in order to be able to depict the curves purposefully. For the
reflection comprising an all-pass, however, a rather different scenario emerges: here we get
conductance-values that are larger than the inverse of the wave-impedance. This is due to the
longitudinal (dilatational) waves already mentioned above – almost half of the power of the
transversal waves arriving at the bearing they can be converted into them (Chapter 7.5).

The (at least theoretically) high importance of the presence of an all-pass is also shown by the
following measurement that had already been indicated in Fig. 7.34. In Fig. 7.39, we see a
segment from the spectrum of a string tuned to 152 Hz. The vertical grid-lines are matched to
the calculated frequencies of the partials as they would be present in a rigid string clamped
fixedly at its ends. Thus the spreading of partials caused by the bending stiffness is considered
here – the correspondence remains rather poor, though: the frequencies of 8 of the partials
clearly miss the calculated values, und there are 9 additional lines. Both the deviations and the
generation of additional partials are the result of the phase of the reflection-factor:
minimum-phase zeroes cause detuning, all-pass-behavior generates additional tones.
Abb. 7.39: Spectrum of a plucked string running

across a bearing saddle via a 45° bend angle.
We must, however, not imagine the sound-effect of the additional tone as an inharmonic
interference next to the actual guitar sound. If the level of such an additional tone is small, it
remains totally inaudible. Quantitatively, it is difficult to state anything here because the
psycho-acoustic masking mechanisms are highly complicated for complex stimuli. A
qualitative statement is easy to formulate: in every guitar sound, there are partials that are
visible in the spectrum but still remain inaudible. If they do become audible (given sufficient
level), they come across not as interference but as a change in sound color. For example, an
additional 3416-Hz-tone appearing next to a 3406-Hz-tone may cause a beating effect in this
frequency range. However, since inharmonic (spread out) spectra sound with a slight beating
effect anyway, the addition of an extra tone will at most make for a marginal change (as long
as the additional tone stays within certain limits). It is not possible to quantify this statement
further for your typical situation on stage (or in the studio) because there are too many
unknown influences: filters, amplifiers, loudspeakers, room resonances, etc.
Is it then purposeful at all to measure guitar-vibrations, since at the end there is (yet) no way
to give quantitative statements about the sound? Of course, measurements can only be
supplementary to listening experiments, and not a replacement. Measurement of vibrations
support (or refute) assumptions about models – and they therefore deliver building blocks for
a psycho-acoustical model about sound. This model at present only exists in rudimentary form
but takes shape as the findings progress. From a theoretical point of view, the exact
understanding of reflection processes naturally is indeed important: it allows for defining
reasons for abnormalities even if the latter do not become audible in every case. It is
reassuring to be able to in fact attribute an unexpected result to the investigated object, rather
than fearing that the equipment is at fault. We could call such a fault a “system-immanent
artifact”, but it would be disturbing just the same. Spectral analysis – as the basis for every
determination of partials – includes many artifacts that could massively influence the final
result. The spectrum calculated according to the classic Fourier-integral does not exist at all
for real tones: not many people are willing to wait the formally required infinite period of
time. Weighting windows create approximations within a finite time, but they do this at the
expense of un-ambiguity. Chapter 7.6 will address, in depth, the instrumentation analytics, but
first let us take a look at conductance- and absorption measurements for real guitars – given
all the above theory it will be good to now show some quantitative measurement results, as
well.

7.5.4 Measurement results
Measuring the mechanical parameters of a string saddle is complicated because on the one
hand mode-coupling and dispersion lead to a large variety of parameters, and on the other
hand very high measurement accuracy is required. It does make a difference whether 99.8%
or 99.9% of the incoming energy is reflected. Given as little as 0.5% measurement error, a
degree of reflection of in excess of 100% could result – which of course is nonsense. The
analysis of the decay (of string vibration) would seem to offer a welcome alternative –
however, this allows only for statements relating to both string-bearings such that a
differentiation of nut/fret and bridge is not possible. Moreover, measuring the mechanical
impedance or admittance of the saddle shows only part of the picture since it captures neither
bending coupling (Chapter 2.7) nor the excitation of dilatational waves. The measurement
results given in the following therefore are a first step towards an analytical description of the
reflection process.
The measurement results for a Les Paul Standard are shown in Fig. 7.40. All 6 strings were in
place; the guitar rested on a stone table (with a mouse pad serving as a cushion). The
impedance measurement was done using a B&K-4810 shaker and a B&K-8001 impedance
head. The tracer pin of the impedance head was placed onto the bridge saddle of the A-string
in such a way that the impedance perpendicular to the fretboard could be measured.
Representation on paper with logarithmic scaling on both axes shows many resonance
maxima that are not all of interest in detail. The degree of absorption therefore is depicted
with linear ordinate scaling – to guide the focus to the essential.
2
Fig. 7.40: Conductance G and degree of absorptions a , measured at the A-bridge-saddle of a Gibson Les Paul
Standard. “Konduktanz” = conductance; “Leistungs-Absorptionsgrad” = degree of power absorption.
Essential is: below about 1 kHz, the absorption is very small, above 1 kHz several selective
maxima of the absorption show up. The degree of absorption is calculated for a wave
impedance of 0,7 Ns/m, approximately corresponding to that of an A-string. If we assume that
the nut has a similar absorption behavior as the bridge♣, there would be twice the absorption
loss per period of the fundamental (i.e. per 9 ms); with a2 = 9.5% this would give us an
increase in damping of 95 dB/s. On the other hand, e.g. a2 = 0,1% would yield as little as 1
dB/s. This example indicates the range of the absorption: 1 dB/s would for normal guitar
playing have the effect of almost non-existent damping (“endless sustain”), while 95 dB/s
would mean immediate complete loss of the tone. In reality, however, we cannot assume the
same absorption at nut and bridge, and therefore additional measurements are necessary at the
nut (see Fig. 7.41).
♣
This assumption does not correspond to reality, though – see below.

Showing results for the nut of the Les Paul, Fig. 7.41 supplements the measurements of the
absorption behavior. Here, the highest absorption shows up in the low-frequency range – of
course we always need to consider how the selective absorption maxima correspond to the
frequencies of the partial of the strings [compare to Fleischer 2001].
2
Abb. 7.41: Conductance G and degree of absorption a , measured at the nut (A-string) of a Gibson Les Paul
Standard. “Konduktanz” = conductance; “Leistungs-Absorptionsgrad” = degree of power absorption.
Consolidating the degrees of absorption at both nut and bridge via computation, we get the
graphs depicted on Fig. 7.42. What causes these extreme maxima in the absorption?
Summarizing/simplifying a bit: the low-frequency absorptions result from resonances of
the neck, the high-frequency absorptions stem for bridge resonances. Fleischer has
clearly shown in several of his publications that it is not possible to manufacture a resonance-
free guitar neck. At first glace, it may be surprising that even Gibson’s much-lauded Tune-O-
Matic bridge successfully operates as a vibration-killer at some frequencies – but in the end
that is a concession to the adjustability: many parts – many resonances.
Abb. 7.42: Overall degree of absorption for one period of the fundamental (A-string). Left: string supported by
nut and bridge. Right: string supported by 12th fret and bridge. Gibson Les Paul Standard.
“Leistungs-Absorptionsgrad” = degree of power absorption.
In his analyses of a Les Paul, Fleischer observed bending Eigen-shapes of the neck at 208 Hz
and 445 Hz – this is a good match to the absorption spectra shown above. For 208 Hz, a node
exists at the bridge and at the 10th fret. For 445 Hz, 3 nodes show: one at the bridge, one at the
12th fret and one at the 2nd fret. That there is no exact match between the measurement results
is not a surprise: first, it was not the same specimen of guitar, and second, the bearing of the
guitar was different. There is, however, a simple procedure to unambiguously identify the
guitar neck as reason for the absorptions: detuning its resonances by an additional mass.

To accomplish this, a 250-g-vise was clamped to the headstock of the Les Paul – which
indeed re-tuned the low-frequency resonances (Fig. 7.43). There was, however, little
influence of this additional mass on the higher-frequency absorption maxima – the latter are
not caused by neck resonances but by resonances in the bridge. This was clarified via
measurements for which a metal clamp was mounted to the bridge (Fig. 7.43, right-hand part).
What was said above is again supported here: below 1 kHz neck resonances form selective
vibration absorbers, above 1 kHz the corresponding effect is the result of bridge resonances.
Fig. 7.43: Gibson Les Paul Standard, A-string. Degree of absorption calculated from the conductance
measurement. Left: degree of absorption at the nut, without (red –––) and with (black –––) vise clamped to the
headstock. Right: degree of absorption at the bridge, without (red –––) and with (black –––) a small clamp
mounted to the bridge. “Leistungs-Absorptionsgrad” = degree of power absorption.
At this point we do not seek to carry out any detailed modal analysis, but rather to outline the
principle of the absorption behavior at nut and bridge. The exact shape of the absorption
spectra depends on all involved masses, springs and dampers – it is specific to the individual
guitar, and string and fretting. We may assume the same mass for each of the bridge saddles
on the Gibson Tune-O-Matic bridge, but already their position on the adjustment screw, and
the area and condition of their seating (surface area!) is specific for each string. The bridge
itself, i.e. the part in which the bridge saddles are held, vibrates in the higher-frequency range
according to Eigen-modes, but these cannot be excited to the same degree from every point:
given a node, the admittance is small, and only little excitation happens. The conductance of
the A-bridge-piece will thus in detail be different from the conductance of the D-bridge-piece.
Fig. 7.44: Gibson Les Paul Standard: spectra of the degree of absorption for the individual bridge saddles.

As common ground of all 6 measurements in Fig. 7.44, we recognize merely minute

absorption at low frequencies – only above about 1 kHz, individual maxima in the absorption
show. The height of these maxima depends on the measured conductance and the wave
impedance, for which in all graphs 0.7 Ns/m was taken as a basis. This is typical for the A-
string – for all other strings, a different wave impedance should in fact have been used.
However, the decay behavior of the string depends not only on the degree of absorption but
also on the fundamental frequency. Since fundamental frequency of the string and wave
impedance are approximately reciprocal to each other, a string-specific consideration is not
imperative in this first step.
In view of an individual fit and position of every bridge saddle of the Tune-O-Matic bridge it
is, however, easily comprehensible that the absorption spectrum looks different for every
bridge saddle. On the other hand, we would not expect such differences for the nut, because
all 6 strings run over the same strip of plastic. Still, Fig. 7.45 shows that there are differences
here, was well: in the middle of the neck (for the D- and G-strings), the absorption is smaller
in the higher frequency range when compared to the edges of the neck (E2- and E4-strings).
Presumably, the distal strings (in contrast to the mesial♣ strings) can more efficiently excite
torsion-vibrations of the neck [Fleischer 2001].
Fig. 7.45: Gibson Les Paul Standard: spectra for the degree of absorption measured for the nut.
When analyzing the saddle-absorptions, we must not forget one significant absorber: the
guitarist. To determine the above absorption spectra, the guitar was laid on a stone table
aiming for a low-attenuation fashion; in the following, external absorbers will also be
considered.
♣
Mesial: located towards the middle; distal: located towards the edge.

There are mainly two external absorbers that act on a guitar: part of the back of the guitar
body is in contact with the body of the guitarist, and moreover the fretting hand dampens the
back of the neck. To approximately model these absorbers, the guitar was laid on the stone
table such that a large area of the cranial half of its the rear body rested on a soft mouse-pad,
and moreover a hand clasped the rear of the guitar neck at the 5th fret (without touching a
string). The effects of this additional damping are shown in Fig. 7.46.
Fig. 7.46: Gibson Les Paul Standard: spectra of degree of absorption measured at the nut (A2-slot).
Left: low-damping guitar bearing. Right: including typical external absorbers. Some frequencies of partials are
marked with dots (fG = 110Hz). “Leistungs-Absorptionsgrad” = degree of power absorption.
The additional absorbers reduce the height of the maxima in the spectrum of the degree of
absorption, and the peaks get broader. We may, however, not derive from these absorption-
maxima how these absorbers influence the decay of the string vibration – rather, the crucial
value is the degree of absorption at the frequencies of the partials (marked by the dots). It is
easy to see in particular for the guitar positioned on the low-attenuation support, that already a
minor de-tuning of the string may result in a considerable change in the degree of absorption.
Of course, the same holds for modification of body- and neck-parameters.
Abb. 7.47: Les Paul Std.: spectra of the degree of absorption measured for the bridge (A2-bridge-saddle).
Left: guitar on a low-attenuation support. Right: with typical external absorbers.
However, it is almost impossible for the guitarist to influence the absorption behavior of the
bridge (Fig. 7.47) because the bridge is rarely touched when playing the guitar (translator’s
remark: in fact, this usually happens only when strong string damping is sought, anyway). In
contrast, some absorption maxima change if the bridge is shifted back and forth within the
slack resulting from manufacturing tolerances (compare to Fig. 7.40).

Via combining the nut- and bridge-absorption, we arrive at the overall degree of absorption of
a string, i.e. at the magnitude that represents the energy loss per period of the fundamental
oscillation (Fig. 7.48).
Abb. 7.48: Gibson Les Paul Standard: overall degree of absorption of the A-string, without (left) and with (right)
damping by the fretting hand. Frequency dependence calculated from conductance measurements.
“Gesamter Leistungs-Absorptionsgrad” = overall degree of power absorption.
Fig. 7.48 shows the overall degree of absorption for two cases: for the freely vibrating guitar
neck, and for the neck damped by the fretting hand. The A-string has a fundamental
frequency of 110 Hz, i.e. a basic period of 9.1 ms. If the string were to lose 8% of its vibration
energy per basic period, its oscillation level would drop by 40 dB per second – that would be
a strong damping. For 1% loss we would get a 4.8-dB-drop per second, and for 0.1%, 0.5
dB/s loss would remain. Two other processes need to be considered here, though: the degree
of absorption is only of significance at frequencies where the string offers Eigen-oscillations
(partials), and there are other absorption-mechanisms besides the absorption at the bearings
(Chapter 7.7).
The un-damped neck of the guitar investigated here shows a pronounced maximum of the
conductance (or the damping) at 200 Hz – this is close to the 2nd partial (220 Hz) of the A-
string. If this resonance frequency (or the frequency of the partial – e.g. when tuning down) is
detuned by as little as a few percent, the absorption for this partial changes significantly. All
maxima seen in the figure are of a relatively narrow-band characteristic, and therefore the
damping of the partial that occurs in the end depends strongly on minute de-tuning effects.
As the fretting hand touches the backside of the guitar neck (not something entirely unheard
of when playing a guitar), the low-frequency peaks become wider and the extreme frequency-
dependency decreases somewhat. The damping of the first 5 partials is, however, increased.
Last, it should be mentioned that for wound strings, the exact frequencies of the partials of the
strings depend on both the string-diameter and the ratio of core-diameter to overall-diameter.
The inharmonicity-parameter (b in Fig. 1.7) determines the spreading of the spectrum, and
thus the exact position of the individual partials. The damping of a certain partial therefore
is a highly fragile quantity that depends on many parameters and may not be seen as a
guitar-specific constant.
Chapters 7.7 and 7.12 will investigate the individual damping mechanisms in detail.

7.6 Instrumentation for vibration-measurements 7-51
7.6 Instrumentation for Vibration-Measurements
Every theoretical description has to face up to an evaluation via practical application. Of

course, an equation representing a vibration will never completely cover the motion of a real
string – but it does not have to, either, because that would make everything infinitely
complicated. Rather, mathematics offer theoretical models, and it is the job of practical
application to recognize the limitations of these models. The following chapters are to assist,
and avoid that this practical evaluation does not itself become a source of uncertainty.
7.6.1 Impedance- / admittance-measurements
The mechanical impedance Z = F / v is calculated as the quotient of force and particle-

velocity; the admittance is the reciprocal. The impedance-head is a typical sensor for
measuring the impedance; in the case of the B&K 8001 we have a thimble-sized cylinder that
contains, in its interior, two piezo-crystals. These crystals measure force and acceleration,
with the latter yielding the velocity via integration over time. Now, any measurement will
affect the value to be measured: sometimes almost not at all (for example radiation pressure in
contact-free laser measurements), but sometimes significantly, as it happens e.g. in impedance
measurements. This is due to the fact that the force sensor is not located directly at the
measuring point but within the impedance head. From the force sensor, the connection to the
external world is made via a rubber-cushioned nut. Since there is no mass-free nut, about 1 g
of parasitic mass m0 needs to be considered. If measurements at higher frequencies are the
objective, further artifacts join the list. Fig. 7.49 shows a measurement with the impedance
head in the no-load condition:
Fig. 7.49: Left: magnitude of the co-vibrating
complex mass F / a of the impedance head.
Below: Force-flow-diagram of the equivalent
systems. “Leerer Impedanzkopf” = impedance head
by itself; “Dynamische Masse” = dynamic mass.
Depicted is the magnitude of the complex mass of just shy of 1 g. A tracer pin screwed into
the nut increases the mass further so that in total 1.3 g of parasitic mass show up. This mass
appears to act ahead of the measuring object (between impedance head and measuring
object). In the force-flow-diagram, m0 is connected in parallel to the device under test,
because the flow-quantity force [3] is divided into two paths: the inertia-force for m0, and the
force FM towards the measuring object: F = a⋅m0 + FM. The impedance ZM of the device under
test is increased that way: Z = ZM + jωm0. The manufacturer is aware of this issue and
therefore offers the mass-compensation-unit B&K 5565. Using the latter is not without its
pitfalls, but we shall not go into more detail here: from today’s point-of-view, the 5565 is out-
dated. Also, the whimsicalities of polarity (the B&K 2625 inverts, the B&K 2623 does not)
shall be mentioned here only in passing in this one sentence, although its disregard can cost
you half your leave days …

In Chapter 6, an up-to-date variant of the mass-compensation has already been introduced,

and therefore we will look here merely at the effects of an uncompensated parasitic mass.
Dividing the impedance Z into its real part R and its imaginary part X, we see that ωm0
merely increases the reactance X. Therefore, if only the resistance R is of interest, we can do
entirely without any mass compensation. For the admittance Y, however, ωm0 has effects on
the susceptance B and the conductance G:
; with .
The conductance G = Re(YM) of the device under test becomes a complicated term that
corresponds to G merely in exceptional cases, and even then only in approximation
(denumerator → 1). Generally, the discrepancy continues to decrease the smaller the parasitic
mass m0 is, and the lower the frequency becomes. Fig. 7.50 shows the effects a 2-g-mass has
on the calculation of conductance. Since it is difficult to predict how large the measurement
errors will become without mass compensation, it is preferable to measure admittances
generally only with mass-compensation.
Fig. 7.50: Conductance without (––) und with (----)

2 g parasitic additional mass. At 1.7 kHz, the
interaction of bearing suspension and additional
mass results in an additional resonance peak.
For all analyses presented here, the force- and acceleration-signals of the impedance head
were recorded using a Cortex CF-100 workstation; the mass-compensation was calculated
using a Hilbert-transform. The typical noise-spectrum of the setup is shown in Fig. 7.51:
clearly the intrinsic noise is negligible compared to the noise generated by the sensor. With
400 mV/N and 35 mV/g, the impedance head we used was sufficiently sensitive to capture
small signals, as well – a dynamic-optimization adapted to the respective problem was
nevertheless required for the corresponding measurements.
Fig. 7.51: 1/3rd-oct. analysis of the system noise:

impedance head with charge amplifier (–––);
analyzer Cortex CF-100 (----). 0 dBµ ↔ 1 µV.

Optimizing dynamics means adapting the dynamic range of the signal to the dynamic range
of the system: on one hand we prevent weak signals from disappearing into noise while on the
other hand avoiding overdrive in the signal paths. Usually, this works well for measurements
taken at merely one frequency, but it may become an issue for broadband measurements. Fig.
7.52 shows force and acceleration for the impedance head without load; both magnitudes
change only by a factor of 1:20 across the frequency range – this is well manageable.
Fig. 7.52: Mechanical no-load condition: force and acceleration (left); dynamic mass (right); UShaker = 2 Veff.
“Impedanzkopf mit Taststift = impedance head including tracer pin; “Dynamische Masse” = dynamic mass
An impedance head in the no-load condition (open circuit, load impedance = 0) represents
one extreme; the other extreme would be the firmly fixed head (short ckt, load imped. = ∞).
The parasitic mass shows up for the open circuit, and the stiffness of the tracer pin can be
seen for the short circuit – although only in approximation because a counter-bearing at
complete rest is impossible to realize. To measure the stiffness, the tracer pin was set against a
stone table of 200 kg; the result is shown in Fig. 7.53: we find a stiffness of s = 6 MN/m that
– in conjunction with the parasitic mass (1.3 gram) – results in a resonance at 11 kHz. The
force varied by a factor of 1000 in this measurement – given the dynamic range of the
analyzer (100 dB), this should not be a problem. However, calculating the voltage generated
by the sensor (400 mV/N) for a force of 1 mN, we arrive at merely 400 µV: for the broadband
measurement, this is below the noise floor of the older charge amplifier (2625) used for this
measurement. Still, the coherent averaging following the Hilbert transform manages to find a
resonance at 11 kHz. The force minimum is not correctly identified in the broadband
measurement, though: merely noise is detected (Fig. 7.53; selective meas’mnt → Fig. 7.56a).
Abb. 7.53: Mechanical short circuit: force and acceleration (left, broadband), magnitude of the impedance (right,
frequency-selective); UShaker = 0,1 Veff. “Impedanzkopf” = impedance head; “Kurzschluss” = short circuit;
“Impedanz-Betrag” = magnitude of impedance

It is difficult to foresee from which level of excitation-amplitude a structure will generate

significant non-linearities. In any case, the drive-level limit given in the datasheet of the
shaker must never be taken as a guide value for linear behavior. The maximum admissible
supply voltage of 7 Veff (B&K 4810) is a thermal limit that would have led to completely
useless results of the present measurements. Even at merely 0,25 Veff, pronounced non-
linearities showed up in the vicinity of the 1.9-kHz-peak, and only at 0.1 Veff the nonlinear
distortion was sufficiently small – but now the noise was too strong. This led to the question
which control signal would be optimal.
The two classical excitation signals for measurements of the frequency response are sweep
and pseudo-noise. For the sweep, the frequency of a sine tone increases over time, with the
amplitude remaining constant. Pseudo noise is a special noise repeating after a period T. The
density spectrum of the pseudo noise is constant (white), and its periodicity corresponds to the
block length of the DFT-analysis (e.g. at 48 kHz sampling rate and N = 4096 ⇒ T = 85,3 ms).
Due to the strong time-variance of the short-term spectrum, time-weighing windows have to
(!) be dispensed with – however, this does not pose any disadvantage because due to the
identical periodicity there is no leakage. Also suitable is true stochastic noise (normal- or
equal-distributed); however, this signal requires windowing and averaging.
Fundamentally, the frequency response of signal- and system-quantities may be explored via
three different approaches: first, selective excitation and broadband measurement; second,
broadband excitation and selective measurement, and third, selective excitation and selective
measurement. Your typical sweep-measurement belongs to the first group, your typical noise-
excitation to the second. Both approaches have disadvantages in case the system shows
substantial non-linearities. A broadband measurement with sweep excitation may preclude
capturing selective minima, as seen around 11 kHz for the force measurement in Fig. 7.53.
Here, excitation via noise paired with selective analysis will deliver better results – but it may
lead to arriving at the wrong conclusions in case of distortions: any signal limiting occurring
at 1 kHz will change the results of the analysis at other frequencies. As a example, an arctan-
function is inserted as a non-linearity into a band-pass system: the transfer function
determined via DFT is bent more and more with increasing distortion – however, this happens
not predominantly at 1 kHz, but at 3 kHz (k3) and at low frequencies (difference tones). If this
non-linearity would be connected in the signal flow ahead of the band pass, the measurement
results would be useless – in their entirety, even.
Fig. 7.54: System analysis with pseudo-noise and DFT: no/weak/strong non-linearity (left to right).
If mechanical systems are to be analyzed across several frequency decades, the measured
quantity may vary by 105; a harmonic distortion of a “mere” 0.1% will become a serious issue
here. The analyzer may have a harmonic distortion of 0.001% but the sensor is not likely to
be up to this. Moreover, the distortion in the actor may even exceed 1%.

If limiting is introduced to the sweep-sine-signal by an arctan-function (non-linearity ahead

of the system resonance), we get a different result: for frequencies around 333 Hz, the
resonance amplifies the 3rd harmonic that is created by the non-linearity, while the higher
frequency range is void of any selective errors (Fig. 7.55). If the non-linear limiting happens
post-system-resonance, there is a tendency to represent the maxima with too small a value.
Fig. 7.55: System analysis using sweep: no/weak/stronger non-linearity ahead of the resonance (left to right).
The sweep measurement is least sensitive to non-linearities if the output signal of the system
to be measured is filtered with a narrow-band tracking filter (selective excitation, selective
measurement). For particularly pronounced signal dynamics, the frequency dependence of the
sweep amplitude may additionally be matched to the system. Of course, corresponding
filtering is possible for pseudo-noise, as well, since the periodicity does not change due to
this. Fig. 7.56a compares sweep measurements with and without tracking filter – the result
speaks for itself.
Fig. 7.56a: Force measurement; probe tip directed Fig. 7.56 b: Acceleration measurement, as given in
against the stone table. Without (----) and with (––––) Fig. 7.6.8a. Without (----) and with (––––) tracking
tracking filter. The force changes by 1:70000 across filter. At 600Hz and 900Hz, non-linearities excite the
the frequency range; such a large dynamic range is 1800-Hz-resonance.
only manageable with a selective measurement. “Impedanzkopf” = impedance head;
“Impedanzkopf “ = impedance head, “Kraft” = force “Beschleunigung” = acceleration
During mechanical impedance measurements, interruptions in the force flux are a particular
problem. The probe pin of the shaker cannot be welded to the guitar bridge (or fret) but it is
pressed to the object using a constant load offset. For the B&K 4810, the largest admissible
force is about 6 N. In this situation the shaker is already deflected to its limit, though,
operating with strong non-linearity. A 3-N-offset would be the optimal theoretical value, with
an alternating force of at the most 2 Neff superimposed. Reducing the offset carries the risk
that the probe pin lifts off, while increasing the offset will generate single sided force
limitations (i.e. non-linearities). Sometimes there is the possibility to generate a load offset via
supplementary spring; the force of this spring should not be included in the measurement.

7.6.2 The spectrum of decaying partials
An ideal, un-damped vibration of a string has a harmonic spectrum and may be represented
by a Fourier-series without much effort. In a real string, however, several damping
mechanisms are at work at the same time: the string itself radiates sound, heat is generated in
its interior, and in the bearings, active power is withdrawn from the vibration of the string.
The amplitudes of the partials (constants in the Fourier-series) become time-dependent, and
the string vibration looses its periodicity. The standard tool for analyzing non-periodic
vibrations is the Fourier integral comprising a special window-weighing. Choosing the
parameters of the analysis, though, we run into the classic conflict of goals that cannot always
be resolved satisfactorily: using a short window duration, the leakage effect broadens the
frequency lines, but a long window duration will deteriorate the time resolution too much. If
all partials were regularly spread out (Chapter 1.3), we would possibly be able to find an
acceptable compromise, but the allpass-driven generation of additional tones requires an
analysis with bands as narrow as possible (compare to Fig. 7.39).
As its amplitude becomes time-variant, the spectral line of a continuous tone (represented as a
Dirac in the density spectrum) turns into an infinitely broad, continuous spectrum [6]. For the
exponential decay process, the Dirac line needs to be convolved with the Fourier-transform of
the e-function; the parameter of the latter is the time constant τ:
Fig. 7.57: Time function and spectrum of an exponentially decaying cosine-tone; f = 100 Hz, τ = 200ms.
The faster the tone decays (i.e. the smaller τ is), the broader the spectrum will be in the
relevant level range. Fig. 7.57 depicts time function and spectrum of a decaying 100-Hz-tone.
In support of a purposeful presentation, the time constant is – at 200 ms – chosen to be
relatively short; in guitar strings, values of in excess of 5 s are possible. The widening of the
spectrum in Fig. 5.57 must not be confused with your typical DFT-leakage; rather, it results
purely from the decrease of the amplitude over time. If we cut a DFT-frame (block) starting
at t = 0 from the infinitely long time function, and transform it into the spectral domain (short-
term spectrum), this weighing over time results in an additional convolution in the frequency
domain. This is the leakage – an additional diffluence of the spectrum particularly noticeable
in the area of the peak. The shorter the DFT-frame, the stronger the spectral diffluence is –
with the shape of the weighing function over time (window function) to be considered as a
further parameter.

Instead of convolving the spectral line twice, it is usually easier to interpret the DFT as a
filtering analysis. The transfer function of the analysis filter is the Fourier-transform of the
window function. Fig. 7.58 presents the spectrum of a decaying two-tone-signal, and the
frequency response caused by the filter-effect of a Kaiser-window♣: on the left for a 4-k-DFT,
on the right for a 16-k-DFT. We can clearly see that the 4-k-DFT will not be able to separate
the closely adjoining lines, and even a 16-k-DFT will not be able to deliver the perfect result.
There are two reasons for this: the window-lobe is still relatively wide, and moreover the
spectra of the two decaying tones will overlap. The smaller the distance in frequency, and the
faster the decay process, the more difficult the analysis becomes. From the figure, we can also
clearly see a further issue specifically present for the Kaiser window used here: the side lobes.
While the latter may be reduced in height by choosing a larger window-parameter (β), this
change will, however, further widen the window-lobe even more.
Fig. 7.58: Two-tone signal (100 Hz, 104 Hz, τ = 0.5s), Kaiser-window (β = 5), sampling frequency 48 kHz.
A decaying partial may be described by four parameters: frequency, level, phase and time
constant. The phase of real guitar tones is of minor importance for the sound (compare to
Chapter 8.2.5) – however, all three other parameters should be identifiable with high accuracy
via a spectral analysis. If it nevertheless does not work out that way, the fault does usually not
lie with the analyzer but with the measuring principle.
Each DFT shows special level errors that may well amount to 1.4 dB for a Hanning window:
this is known and for the most part acceptable. If not: there are many alternatives with a level
error smaller than 1 dB. This is for continuous tones, though! Because for all other signals
much larger level errors may result – otherwise we would work exclusively with the flat-top
window. Every window has its specific advantages and disadvantages, the awareness of
which singles out the expert. Tried and tested are the Blackman-Harris windows, and the
Kaiser windows – the parameter of the latter is not consistently specified, though. Not all
windows have such a parameter. If it exists, it is useful to search for a compromise between
strong attenuation of the side lobes, and small bandwidth of the widow. Enlarging the
parameter increases the dynamic range of the measurements but deteriorates the spectral
selectivity. The parameter should therefore be chosen such that a dynamic range of about 40 –
60 dB is obtained.
The following table offers a short overview regarding some important window parameters.
More extensive details may be found in specialist literature.
♣
Parameter-definition as customary in MATLAB

Table: Data of common-place DFT-windows [from M. Zollner: Signalverarbeitung, HS-Regensburg, 2010].
SLA MLW Ripple BW FM PM

dB lines dB dB
Rectangle 13.26 1.62 3.92 0 1.000 1.000
Triangle 26.52 3.24 1.82 1.25 0.500 0.333
Exponential α = 1 12.6 1.72 3.65 0.34 0.632 0.432
2 10.8 2.16 3.03 1.18 0.432 0.246
3 − − 2.35 2.19 0.317 0.166
4 − − 1.77 3.17 0.246 0.125
Hanning 31.47 3.37 1.42 1.76 0.500 0.375
Hamming 42.68 3.83 1.75 1.34 0.540 0.397
Rosenfeld 48.4 5.78 0.90 2.81 0.381 0.277
Gauss α = 2.50 (40) (5.9) 1.58 1.60 0.495 0.354
3.16 60 7.1 1.06 2.53 0.396 0.281
3.76 80 10.4 0.76 3.27 0.333 0.235
4.32 100 13.9 0.57 3.87 0.290 0.205
Blackman exact 68.24 5.87 1.15 2.29 0.427 0.308
" approx 58.11 5.64 1.10 2.37 0.420 0.305
Blackm.-H. 3/62 62.05 5.38 1.27 2.07 0.450 0.326
3/71 70.83 5.91 1.13 2.33 0.423 0.306
4/74 74.39 6.43 1.03 2.54 0.402 0.290
4/92 92.01 7.88 0.83 3.02 0.359 0.258
Nuttall 3/47 46.74 5.78 0.86 2.89 0.375 0.273
3/64 64.19 5.88 1.05 2.49 0.409 0.296
4/61 60.95 7.79 0.62 3.64 0.313 0.226
4/83 82.60 7.88 0.73 3.27 0.339 0.244
4/93 93.22 7.92 0.81 3.06 0.356 0.256
4/98 98.17 7.33 0.85 2.96 0.364 0.261
Kaiser-Bessel 1.74 40 3.84 1.63 1.49 0.533 0.385
2.60 60 5.45 1.16 2.26 0.431 0.313
3.42 80 7.03 0.91 2.81 0.378 0.272
4.22 100 8.60 0.75 3.25 0.340 0.245
Flat-Top 40 40 5.34 0.05 4.44 0.299 0.247
60 60 7.01 0.05 4.97 0.260 0.212
80 80 8.78 0.05 5.40 0.233 0.188
100 100 10.29 0.05 5.69 0.216 0.173
Dolph-Tsch. 2.4 40 3.80 1.78 1.55 0.537 0.412
3.4 60 5.26 1.29 2.07 0.450 0.326
4.4 80 6.68 1.03 2.61 0.395 0.285
5.4 100 8.06 0.88 3.04 0.356 0.256
Barcilon-T. 2.21 40 3.71 1.74 1.38 0.536 0.395
3.26 60 5.18 1.27 2.12 0.446 0.324
4.27 80 6.60 1.01 2.65 0.391 0.282
5.30 100 7.98 0.85 3.08 0.353 0.254
SLA = Sidelobe Attenuation BW = Power bandwidth in dB

MLW = Mainlobe Width FM = window mean value
Ripple = Level error PM = power mean value

The window parameters may be calculated either from the time function of the window ,
or from the spectral window function . The time functions of the window are always
amplitude-normalized, i.e. their maximum value is 1. The zero-point for time is located in
the middle of the window for all symmetric windows; for unsymmetrical windows it is at the
beginning of the window. The term polynomial window characterizes symmetrical windows
the time function of which may be described as a superposition of cosine functions:
Formulae for polynomial window

⇓
Mean value of window FM:
FM is the linear mean value. Max(w)/FM is termed coherent gain in English language literature
Power mean value PM:
PM is the mean value across the squares of the weighing function.
Effective bandwidth :
Often, is referenced to the line-width Δf : BWL = Beff / Δf; BWdB = 10lg(BWL)dB.
Effective duration :
[for Max(w)=1]
Often, is referenced to the duration T of the window: T% = Teff ⋅100% / T.

Side lobe attenuation SLA:
For the Kaiser-Bessel-window shown here, SLA is 60dB; for the Flat-Top-window it is 91 dB.
Main lobe width MLW:
MLW is specified as multiple of the line distance; in the example: MLW = 2⋅1.87 = 3.74
Level error (Ripple, Scallop loss):
The level error is determined at half the line-distance; in the example: ΔL = 1,4 dB
Further reading:
Brigham E.: FFT - Schnelle Fourier Transformation, Oldenbourg 1985
Gade S.: Use of weighting functions in DFT-Analysis, B&K T. Rev. 387
Harris F.: Use of windows for harmonic analysis, Proc. IEEE, Vol.66, 1/1978Δ
Papoulis A.: The Fourier Integral and its Applications, McGraw-Hill, 1962
Zollner M.: Signalverarbeitung, Hochschule Regensburg, 2010
Zollner M.: Frequenzanalyse, Hochschule Regensburg, 2010.
Δ
This (actually very good) publication contains a number of typing and drawing errors!

If you trust the DFT, you will expect the frequency of a partial to coincide with the local
maxima of the magnitude spectrum. Fig. 7.59 clarifies that this is not necessarily the case: the
spectrum of the decaying two-tone signal indeed contains two maxima but these are not
located exactly at the frequencies of the two partials. Nor is it evident that, at t = 0, both
partials have the same level. Looking at the theory, it does of course work out: since the
higher-frequency partial decays faster, its energy within the DFT-frame is lower. However,
this cannot be gathered from the figure without knowledge about the signal – and as a rule
you will want to analyze unknown signals.
Fig. 7.59: Decaying two-tone signal: f1 = 3000 Hz, τ1 = 0.4 s, f2 = 3006 Hz, τ2 = 0.1 s; u = u1 + u2. On the left,
the time-envelope is shown; on the right we see the level-spectrum. On the right, two vertical lines mark the
frequencies of the two partials.
It is not possible to gather the decay time-constant from one single spectrum although it must
somehow be contained in the latter. Since that is not the case in an obvious way, we put
together a spectral array in which the level is registered as a function of time and frequency.
To assemble the array, the window is shifted along the time axis (possibly with overlaps) with
the shift yielding the abscissa-value for the representation of the level decay. We do hit one
snag, though: a purposeful spectrum cannot be obtained for a point in time (the spectrum of a
Dirac is of “white” characteristic) – spectral analysis is meaningful only for a time range (the
frame). To derive the decay of a partial from the DFT, we therefore need to know whether the
zero on the time axis is allocated to the start, or the middle, or the end of the frame. In the
following, we always use the middle of the window as a reference; consequently the level
decay cannot be represented from t = 0. Given a sampling frequency of 48 kHz, 8192 sample
points result a window duration of already 170 ms, and the first spectrum therefore needs to
be assigned to t = 85 ms. This exposes a fundamental issue of every short-term DFT: in order
to obtain a high resolution with respect to time, the frame needs to be short (e.g. N = 256), but
this leaves the spectral resolution lacking. A longer frame (for example a 32-k-DFT) yields
good spectral resolution … but now the time-resolution leaves much to be desired.
So as not to dwell too extensively on synthetic signals, we shall now analyze recordings of a
real guitar string (∅ = 0,7 mm, M = 68 cm, f = 152 Hz). It was stretched across a stone table
with steel cylinders (∅ = 3 mm) serving as bearings (i.e. representing nut and bridge). The
bend angle was 17°. A laser-vibrometer took measurements of the transversal vibration at a
position of 9 mm from the “nut”. The string was struck in impulse-fashion very close to the
“bridge”. String, strike direction, and laser beam all were in the same plane. The laser signal
was sampled and recorded with 48 kHz; the subsequent evaluations were done using
MATLAB. The low partial showed regular decay – irregularities start from about 3 kHz.

Fig. 7.60 shows, on the upper left, an excerpt from a level spectrum, and next to it the decay
of the 3633-Hz-partial derived from the DFT-array. This spectrum alone would not
compellingly reveal any irregular decay of the tone; the short-term DFT, however, shows
intense beating. If we change the DFT parameters (only the number of the points shall be
varied here), entirely different decay curves result. These have again a different shape if the
type of window is changed. With N = 8192, we see not one partial at 3633 Hz, but two – the
corresponding levels do still not decay in a linear fashion, though. They show a beating, and
thus more partials must exist. The latter can, however, not be isolated even with N = 32768.
The window duration amounted already to 0.68 s, so that not much room remained for any
further time-shifting within the signal duration of 1 s. By the way: it should not bother us that
the minima cannot be reproduced with precision: they result from an interference that is
highly sensitive to variations in attenuation. Moreover, it is actually a characteristic of a
window to attenuate partials.
Fig. 7.60: Spectrum (left). Level decay (right) of individual DFT lines. Kaiser-window (MATLAB).

Now, which spectrum is the correct one? Of course, they are all correct – there is an infinite
number of (correct) short-term spectra. Which one is purposeful? That question is much more
difficult to answer. Counter-question: purposeful for what? Investigating causes, or effects, or
mechanics, or psychoacoustics? To determine Eigen-frequencies, a DFT with a high number
of points is usually chosen; for the perceived sound, however, the exact frequency of a tone is
only of indirect importance. Two tones of small frequency distance are not perceived as
detuned but as beating i.e. as a temporal rather than a spectral effect. Therefore, the structure
in time will be more interesting for investigations into sound, and short DFT-lengths will be
preferable – knowing full well that psychoacoustics still have a hard time with complex
sounds. On the market for psychoacoustics-analyzers we see devices competing with each
other that have very different analysis filters, we see gamma-tone filters put next to 1/3rd-
octave-like critical-band filters as if they were equivalent, we see specific loudness calculated
via 6-pole reference filters or via true critical-band filters, we see no importance given to the
filter phase at all. A rough indication may be determined that way – but not subtle differences
in sound. The cochlea is a time-variant, non-linear system the transmission characteristic of
which (i.e. frequency response of phase and amplitude) is influenced by the sound signal. In
contrast, customary analyzers use time-invariant filters, and if they at all calculate the non-
linear fanning out of the upper masking flank (as it is found in the hearing system), they do so
after the fact as a correction into the signal flow. This approach works for relatively simple
signals but gives merely an orientation for complex sounds.
Since evidently it is not possible to determine all partials of a guitar tone metrologically, it is
not recommended to try and achieve an ever better frequency resolution via an ever increasing
number of DFT-points. Rather, we could consider whether it is at all purposeful to seek to
expand a signal that consists – in the model – of a series of decaying inharmonic partials, into
a series of non-decaying partials! In fact, this is what the DFT-algorithm will do: the process
of the Fourier-integral seeks to find steady tones, and it determines the corresponding
amplitudes, frequencies and phases. There is no mention of any decay constant. For causal
time-functions – the Fourier-transform of which does not include any poles on the jω-axis –
the Laplace-transform may be specified as alternative to the Fourier-transform. This
theoretically opens up the possibility to search for residuals, and derived from that a
description via poles in the complex p-plane. Among other aspects, MATLAB offers – with the
Prony-algorithm – a possibility to determine from a signal directly the poles and zeroes of an
ARMA-Model (IIR/FIR-filter), and thus to find the Eigen-frequencies and attenuation factors
of individual partials. In order not to stress this algorithm too much, it appears purposeful to
feed the signal through a band-pass filter first of all, such that only few partials (difficult to
separate) remain. Still, it must not be expected that now all signals can be analyzed as desired:
each process includes system-immanent artifacts, and with increasing complexity of the
signal, these artifacts become more complex, as well♣.
In order not to drown completely in the mist of the speculative, here are two
recommendations: for the analysis of low-frequency partials, the DFT is well suited. It may be
deployed for the bass strings up to about 1 kHz and for the treble strings up to about 2 kHz. In
the higher frequency ranges, it may be purposeful to additionally carry out a 1/3rd-octave
analysis that is better adaptable to time effects. Not of any purpose is a loudness analysis of
the pickup signal: it never reaches the hearing system in this form!
♣
This highly general statement even leaves room for a seemingly philosophical question: is a square signal very
complex because it is composed from an infinite number of sine tones, as is generally known, or is the sine-tone
the more complex signal because it may be summed up from an infinite number of square signals?

The bandwidths of 1/3rd-octave filters (23%) approximately correspond to the bandwidths of

the filters found in the hearing system for frequencies above 500 Hz. Therefore, combining
neighboring partials into one analysis channel rests on a similar basis. Also, a 1/3rd-octave
filter will share with the critical-band filter the characteristic that a very low priority is given
to phase responses.
The Volagramm♣ gives a clear (yet somewhat arbitrary) representation. It shows the decay of
individual partials (in fact: DFT-lines) as a difference spectrum: L(f, t + Δt) – L(f, t). Fig. 7.61
conveys an idea of this approach: level differences were calculated for 4 DFT-spectra
(determined for 0 / 170 / 340 / 510 ms) and outlined as polylines. On the left, we see a rather
regular decay of the partials, as time passes, the polylines fan out downwards because higher-
frequency partials decay more quickly that the lower-frequency ones. On the right, more
pronounced irregularities can be seen – caused by fluctuating envelopes of the partials. This
representation is not unambiguous because both the type of window and the time-spacing are
chosen arbitrarily – but it does provide a quick orientation across frequency ranges of interest.
Fig. 7.61: Volagrams: string mounted on the stone table; ends of the string clamped (left) and supported (right).
0,7-mm-string, fG = 150 Hz, Δt = 170 / 340 / 510 ms, N = 4096, Matlab-Kaiser-window (β = 12).
As ordinate, the attenuation is shown; as time progresses, the polyline fans out downward.
7.6.3 The decay time T30
There are several possibilities to quantitatively describe the attenuation in a resonance circuit:
degree of damping, time constant, loss factor, logarithmic decrement, measure of decay, or Q-
factor. The vibration of a spring-mass-system damped by Stokes-friction will decay
exponentially after excitation by an impulse:
; τ = time-constant
The full designation for the time-constant used in this formula is amplitude-time-constant
because it describes the decay of the amplitude. The power also decays exponentially for this
vibration, but because power has a square-dependency on the amplitude, the time constant for
the power decay will be different: this so-called power-time-constant is half the other time-
constant. Standardized sound measurements use e.g. a power-time-constant of 125 ms in the
“Fast”-mode of averaging; the corresponding amplitude-time-constant is 250 ms. A time-
constant specifies the period of time during which the quantity characterized with it decays to
1/e = 0.368. Alternatively, a decay to other specific values may be given – such as is practice
e.g. with the reverberation-time TN used in room acoustics. During TN, the signal level drops
by 60 dB (i.e. the amplitude drops to 1/1000). Since such a drastic drop is lacking in practical
relevance for musical tones, Fleischer [9] has proposed 30 dB as decay-time T30.
♣
volare = to disappear, to be volatile, to decay (latin); graphein = to draw (greek).

The decay-time must, however, not be understood such that we pluck the string and then wait
until the level has dropped off by 30 dB. Rather, we have to form a smoothed straight line in
the L(t)-diagram, the gradient of which results in the decay-time. On the left in Fig. 7.62, we
see a perfectly linear level decay. With an exponential decay of the amplitude over time, the
level (i.e. the logarithm of the amplitude) will decrease linearly over time. The decay-rate –
the negative gradient of the curve – is 8.7 dB/s in this example; the time-constant is 1 s and
the decay-time is 3.45 s.
Fig. 7.62: Various decay processes.
The centre curve shows the level decay of a beating signal: after 0.31 s, the level has dropped
(relative to the initial value) by 30 dB for the first time. This is, however, not the decay-time –
that amounts to 3.45 s just as in the example on the left and is calculated via the (dashed)
envelope. Such beats occur if two partials of the same initial amplitude and the same
damping, but with slightly different frequencies, decay jointly. In this example it is not
difficult to find an envelope for the maxima of the curve, and to determine its gradient. The
process become more difficult if the periodicity of the beat is much longer, e.g. if the first
minimum is only reached after 5 s. It may be impossible to determine the level values of
subsequent maxima because the signal has already become too small and disappears in the
ever-present noise.
The analysis of the decay becomes even more problematic if partials of very different time-
constants decay (Fig. 7.62, right). We could determine T30 from the initial slope (as it would
be done in room acoustics for the early-decay-time), or from the final slope, or we could –
after all – take the point in time when L passes through -30 dB. In the case of a combination
of beats and different time-constants this could easily lead to an unusable T30-value, though.
In most cases, the decay-time is a highly useful measure to describe decay processes or
attenuation. Still, in some special scenarios it may not be purposeful. Therefore, caution is
advised especially when using programs that automatically calculate d T30.

7.7 Absorption of String Oscillations
When plucked, a string registers an input of a few mW´s of potential energy that will be
converted into heat while the string remains oscillating. This dissipation is based on several
mechanisms, some of which have their origin in the string itself and some in its immediate
surroundings. While Chapters 1 and 2 encompassed the un-damped wave propagation, we
will now focus on individual damping mechanisms in more detail. According to the
predominant opinion in musicians’ circles, it is the body wood that causes the damping of
string oscillations. Highly desirable is long sustain, i.e. a long lasting decay process of a
plucked string; however, allegedly not all woods will cooperate with the musician as desired.
Whether indeed the wood itself represents the main cause of the damping of string oscillations
(and therefore also shapes the sound) will be the subject of the following chapters.
7.7.1 Attenuation by radiation
The oscillation energy of the string is reduced, among other factors, by the fact that a
frictional resistance must be overcome when moving in air – if the string were to oscillate in a
vacuum this resistance would not be present (i.e. nil). This effective resistance can be seen as
the real part of the complex radiation impedance – its imaginary part, a tiny mass, may be
ignored. The real part dampens the string oscillation; it therefore is termed air damping.
Given a damped oscillation with an exponentially decreasing amplitude, the decay speed can
be defined via the time constant τ, or its reciprocal, the decay coefficient δ = 1/τ . These
terms contain a constant term (D0) and a frequency-dependent term ( ). According to
Stokes (summary in [1]), the following holds:
Herein ρair and νair are, respectively, the density and the kinematic viscosity of air; ρ is the
density of the string and D is the string diameter. Fletcher/Rossing [1] combine both
attenuation terms into one formula, therein specifying the decay time constant of the energy.
In order to avoid confusion, only the decay time♣ T30 = 3.45/δ shall be used in the following.
Given 7.9 g /cm3 (not an unusual value for the density of steel) for solid strings and 7.1 g /
cm3 for wound strings, the frequency-dependencies of the decay time T30 of the partials are
obtained as shown in Fig. 7.63. According to the above formula, T30 approximately depends
on the reciprocal of the square root of the frequency, and on the reciprocal of the string
diameter – sets of heavier strings give a longer sustain (in this respect!). We arrive at a decay
time of about 80 s for the fundamental of the E2-string (0.046"), and of about 6.8 s for the
fundamental of the E4-string (0.009"). These results (from using the model) will in the
following serve merely as orientation values; we will not further investigate whether the
radiation impedance of the oscillating cylinder should not be modified, after all – given that
reflectors (guitar body and fingerboard act as such) are positioned in direct vicinity. But even
avoiding an escalating of theory: the almost unending sustain that some wonder guitars are
imputed with … that is impossible solely as a result of the attenuation by radiation (which
colloquially could be called “air damping”) alone.
♣
During T30, the level decreases by 30 dB [Fleischer 2000].
Translated by Volker Eichhorst, Gabriel Mallory & Tilmann Zwicker © M. Zollner & T. Zwicker 2019
Fig. 7.63: Decay time T30 caused by attenuation by radiation (= air damping) for guitar strings (hybrid, 9/46).
The right-hand diagram shows the decay time measured with a steel string (∅ = 0.7 mm) mounted on a stone
table as well as the corresponding calculated attenuation by radiation. Only the low-frequency decay behavior
only can approximately be explained that way. “Strahlungsdämpfung” = attenuation by radiation.
The attenuation by radiation can explain the decay behavior only when measuring in the low
and middle frequency regions (right-hand diagram), and even then only if the bearing
attenuation is very small. In the region of higher frequencies, additionally a loss mechanism
taking place inside the string does have an effect, as will be discussed in the following.
7.7.2 Internal damping
When oscillating, the string changes its shape, i.e. its curvature and length, and energy is
correspondingly required. For the main part, this is reactive energy temporarily stored as
potential energy within the resilient string, but there is also active energy, causing minimal
warming of the string. The active energy is lost to the oscillation process, and therefore such
attenuation (damping) mechanisms are also termed loss mechanisms. If the losses occur
within the string they are designated internal losses. In engineering mechanics, loss
coefficients are defined as the imaginary part of the complex spring impedance (or
admittance), which is in marked contrast to electrical engineering, where real loss resistance
is assigned, for instance to an inductance. Both paths will lead up the mountain because in
both cases, orthogonality is ensured.
In machinery acoustics and materials engineering, internal losses are commonly described
using the loss factor d, with d interconnecting the imaginary part E2 and the real part E1 of the
complex Young’s modulus E: E = E1 + j⋅E2, d = E2 / E1. However, it is very difficult to find
reliable statements concerning d. This may be due to the fact that the split of E into merely
two components is just a very simple model, but also due to the fact that e.g. steel appears in
different types, not all of which can be assigned the same loss factor. The loss factor and the
dissipation model based on it are therefore adequate as a first approximation only. Fleischer
[2000] sets d = 0.001§, with the cautionary remark "tentatively estimated", and a few years
later reduces this value to 0.0004 [Fleischer 2006]. Lieber♣ specifies d = 0.00017, Kollmann♥
d = 0.0001, and Cremer/Heckl [11] offer 0.2 – 3⋅10-4.
§
Fleischer designates the loss factor with η, as usual in the older literature.
♣
Lieber, E.: Vibration of stretched strings, acta acustica 1996 Suppl. Vol. 82, p.187.
♥
Kollmann F. G.: Maschinenakustik, Springer 1993.
© M. Zollner & T. Zwicker 2019 Translated by Volker Eichhorst, Gabriel Mallory & Tilmann Zwicker
Cuesta/Valette♣ extend the above-mentioned formula of the decay coefficient by another two
terms, thereby also taking into account dislocation processes in the crystalline structure as
well as heat conduction (stretching has a cooling effect, compression a warming effect):
Herein E stands for Young´s modulus (modulus of elasticity) and σ for the normal stress in
the string. Using d = 0.7⋅10-4 for the decay coefficient in these equations leads to the curve
shown on the left in Fig. 7.64. Measured were the decay times T30 of the partials for a
vibrating steel wire (∅ = 0.7 mm) stretched between two bearings on a heavy stone table.
Fig. 7.64: Left: comparison of measurement and model calculation. Right: orientation lines (10/13/16 plain).
“G3” = G-string, “H3” = B-string, “E4” = high E-string (E4).
The decay times calculated with the model may certainly be longer than the measured times
because besides radiation damping and internal damping there are further damping
mechanisms that shorten the decay time (Chapter 7.7.3). It is beyond of the aim of the present
work to attribute the individual components of the oscillation damping to material-specific
causes. The matter is a complex one, as already acknowledged by a more authoritative source:
The physical processes that cause the internal damping of metals are very complex and have
not yet been completely investigated. Moreover, it is not that simple to measure the often very
small loss factors, and therefore some of the values found in the literature do not actually
describe the losses within the examined material, but rather tell us about losses within the
measurement equipment, or about losses due to sound radiation [11]. Therefore, very
pragmatically, lines of orientation (Fig. 7.64, right-hand section) are given in the following.
These lines provide a basis to classify and assess decay times measured with guitars. As a
working hypothesis, we assume that the decay behavior specified in the lines of orientation is
primarily determined by radiation-damping and internal damping. As additional findings
become available, the curves may be moved further upwards. For the treble strings (G-B-E4)
the orientation lines provide a good working basis; for wound bass strings (E2-A-D),
however, bigger discrepancies are to be expected: to calculate internal losses, the model of a
solid steel cylinder cannot be used. Rather, three damping mechanisms need to be taken into
account; damping in the core wire (steel), damping within the winding (nickel or steel), and
gap damping at the contact surfaces. All this would be time-variant – of course …
♣
Cuesta C., Valette C.: Evolution temporelle de la vibration des chordes de clavecin, Acustica Vol. 66, 1988.
7.7.3 Winding attenuation
When considering internal damping of a string, its structure must be taken in to account. Solid
strings (most often E4, B3, G3) are made of solid wire of spring steel whereas bass strings
(E2, A2, D3) are wound. The G-string may be plain or wound. Strictly speaking, the term
"internal damping“ only identifies the dissipation losses in the metal. Within the wound
string, however, there are also the contact surfaces between core wire and winding wire, and
here (as well as between the turns of the winding) friction is generated and consequently
energy dissipation (= damping). Measurements of the decay behavior of new strings indicate
that the decay times of partials in solid strings are close to the values calculated on the basis
of the model. Wound strings, however, may be described using the formula given in Chapter
7.7.2. If the outer diameter is used (as it is with solid strings), the calculation for high-
frequencies results in much too short decay times, but taking a reduced "effective“ diameter
will yield an arbitrary result.
That a simple formula for wound strings cannot exist, is shown by the following experiment:
on a US-Stratocaster, two brand-new D-strings from two different manufacturers were
measured (Fig. 7.65). Although the analysis for the two strings was done one right after the
other, on the same guitar, and using the same slot in the nut, the decay times differ very
significantly. This can only be explained by significant differences exhibited in the windings
of the two strings. It is far beyond of the aim of the present work to investigate the material-
specific and structure-specific reasons of these differences; instead, the empirically found
best-case measured values are given in the following diagrams, facilitating orientation and
assessment of the results. The majority of the measured decay times were shorter than the
given orientation values, but in individual cases they were slightly longer.
Fig. 7.65: Decay times of partials, Stratocaster; comparison of two D-strings (0.026" each, wound, brand-new).
The grey area (“orientation line”) estimates an upper limit of T30 due to radiation and internal damping.
In Fig. 7.66, the decay times of the partials of the open strings are given for all 6 strings of a
10/46-string-set (10, 13, 17, 26, 36, 46). As we will have to explain later in Chapter 7.7.4.1, it
is appropriate to derive the damping of the string oscillation from the energy-related sum of
the oscillations normal and parallel to the fretboard – the curves in Fig. 7.66 were established
that way. For a number of selective minima, the cause is known; these minima were removed
(not considered) – they will be analyzed in depth in the following. The causes for the global
decay processes shown in Fig. 7.66 essentially are attenuation by radiation, and internal
dissipation in the case of solid strings; in the case of wound strings, damping due to the
winding weighs in, as well. Damping due to bridge, guitar neck and guitar body leads to
small, frequency-selective peaks – a separate section will be dedicated to them below.
Fig. 7.66: Decay times of partials, US-Stratocaster, best-case. Factory-fresh strings (10, 13, 17, 26, 36, 46). The
region marked in grey (estimate of upper limit due to radiation and internal damping of the string, "orientation
line") was seldom if ever reached during any measurement.
In order to prevent misunderstandings, we need to remind ourselves again that Fig. 7.66 does
not show spectra but decay times. During the decay time T30 the level of the respective partial
is reduced by 30 dB. In the case of the low E-string (E2) it takes 25 s until the level of the
fundamental (82.4 Hz) is reduced by 30 dB. It takes nearly the same time until the level of the
4th partial (330 Hz) decreases by 30 dB – for the 12th partial (1 kHz), however, the same drop
takes a mere 7 s. The graph does not tell us how loud the partials are – or, rather, which level
they start to ring with. The level of a partial is easily changed e.g. by filters, but the decay
time is not – as such the decay times represent a much more guitar-specific parameter than the
spectrum. Mind you: it’s guitar-specific, but also highly string-specific. The global tendency
of the high-frequency drop-off clearly is string-specific, a relation to specifics of the guitar
only can be attributed to the small variations in the progression of the curve. In fact, exactly
this is the subject of the next Chapter 7.7.4.
7.7.4 Bearing absorption
In Chapter 2, the discussion has focused in detail on describing the string as a mechanical line
along which waves are running. The reflection process occurring at both bearings (bridge, and
nut or fret) is defined by the characteristic wave impedance of the string, and by the
respective particular bearing impedance (or admittance). Typically, the bearings are rigid -
thus having a very high mechanical impedance – so that nearly the whole wave energy will be
reflected. However, a small percentage will be absorbed at the bearing, and this is where the
designs of bridge and nut/fret come in, as well as the materials used for these components.
The guitar neck and its resonances [Fleischer] need to be looked into at some point, and
subsequently, at the very last, one may also wonder about the wood of the guitar body. First,
however, term "bearing absorption" must be clarified – because a simple punctiform
impedance is not good enough. Instead, we can isolate several absorption processes, each of
them to be discussed in their own subchapters.
7.7.4.1 Coupling of transversal waves

The magnetic pickup customarily deployed on electric guitars transforms into an electrical
voltage predominantly those string oscillations that occur perpendicular to the fretboard
(Chapter 5). Therefore, it is obvious when performing measurements to pluck the string
normal to the fretboard, and to measure the fretboard-perpendicular string-oscillation
component e.g. using a laser vibrometer. In the simple model, an exponential decay of the
velocity of the partial is assumed:
τ = amplitude-time-constant
Because the instantaneous power is proportional to the square of the velocity, its decay needs
to be described by a power-time-constant – that is half as big as the amplitude-time-constant.
Thus, if we talk merely about a "time constant", there is a risk of confusion. However, the
specification of the decay time T30 (during which the level is reduced by 30 dB) is clear; it
will be applied in the following. The decay time T30 is 3.45 times the amplitude-time-constant
or 6.9 times the power-time-constant. However, not all analyses of partials show a purely
exponential decay. In Fig. 7.67, the measured decay of the 4th partial of a B-string of a
Stratocaster is shown. An analysis encompassing 2 s shows a progressively decreasing curve
to which a single gradient can only hardly be related – both inserted approximation lines
mightily reek of being arbitrary. Enlightenment in the truest sense of the word is provided by
a second laser-vibrometer that upgraded our lab-setup to a 2D-measuring-station. The
fretboard-normal and the fretboard-parallel string oscillations perfectly complemented each
other to sum up (in terms of the energy) to an exponential decay that would do justice to any
textbook, and featuring a decay time (5.7 s) significant longer than the one initially expected.
Fig. 7.67: Decay of the 4th partial of the B-string (Stratocaster). On the right, the level of the fretboard-parallel
oscillation is plotted in addition, and also the level of the sum (----).
The interpretation of these measuring results may vary depending on the question. The
fretboard-normal oscillation the pickup senses – only this being relevant to the sound –
reaches a minimum after 2.5 s. The related loss in level of 22 dB must not be considered as an
energy drop of 99.4%, because a part of the energy is not yet “lost” but stored temporarily in
the orthogonal oscillation mode. After another 2.5 s, the level therefore has not decreased by
44 dB in total but only by 26 dB. However, this does not at all help the guitar player who
wants to play a tone that lasts 2.5 s – he simply feels the sustain of this particular partial as
being all too short. Let’s assume that in particular this partial is of eminent importance, and let
us hold fast onto this: the decay time measured in one oscillation plane must not just blindly
be converted into dissipation parameters.
Conspicuously, the decay analyses of the investigated American Standard Stratocaster

showed that in particular the B-string featured strong beats of partials. Now, of course every
‘in-the-know’ guitar player is aware that these beats, this ‘chorus-like warble’, belongs to the
specific charm of the Strat, and – being privy to it all – our man knows the (supposed) cause:
it’s the magnets! These conniving guys sneakily exert a vicious pull on the strings and ‘hinder
them to decay freely’. We do not know the originator of the moderately intelligent term
'Stratitis' for this ‘illness’ of the Strat … but that’s probably for the best. In Chapter 4.11, we
had already explained that pickup magnets in fact may change the decay characteristic of
individual partials – however, this mainly affects the fundamental. To be on the safe side, the
pickups had been lowered as much as possible before the measurement specified in Fig. 7.67
was taken – in other words: it’s not the magnets, they are not responsible for this beating.
Fig. 7.68 shows further levels of partials of this B-string – all fraught with various beats. If
one does not have unlimited possibilities for modal analysis (one does not: the Free State of
B. in the south of the country G. needs cut back and saving money after the latest banking
disaster), only simple approaches remain for such studies. In the present case: we lift the B-
string out of its groove in the nut, move it sideways by a millimeter, re-tune, and repeat the
measurements. And behold: the beats were yesterday. If only all analyses were that easy.
In its original state (Fig. 7.68, left-hand section), the B-string of the investigated Stratocaster
generates audible beats that one may love or hate. Still: this characteristic definitively must
not be attributed to the specially selected and long stored wood of this American dream – the
mundane source is in the nut. No, don’t even go there and say that this nut has been filed
down with love and given brilliant workmanship exactly in such way that these beats result,
because only they would generate that authentic ´Strat-sound´. Once the measurements had
been carried out, the B-string was allowed back into its original groove and was re-tuned …
and there they were back again: the beats. However, they were not the same anymore – a
closer look showed deviations in frequency and amplitude of the beating. Thus, this sound
characteristic has to be seen as accidental and fragile – a result of a naturally always
tolerance-affected manufacturing. In the case of the investigated Stratocaster, only the B-
string showed such strong beats, all other strings behaved completely inconspicuously. It is,
however, to be expected that among the many Strats manufactured to this day today there are
more than a few that feature more than one string generating stronger beats, and perhaps these
are in fact exactly those holy cows a lot of money is shelled out for. The top nut, stupid ...
No, of course the nut is not the only reason for certain sound characteristics, it is essentially
involved in sound shaping, though. At the beginning of the 21st century, aficionados still
commemorate those fair maidens (or ladies) who – by hand! – wound Fenders' first pickups
(hail oh Mary, Gloria, Abigail!); however, that kind of honor and appreciation is denied to
that master nut-slotter (translator’ question: would that be a nutter, then?). By Leo, he would
have deserved it, too.
Fig. 7.68: Level of partials, B-string, Stratocaster. Left: string in the saddle groove. Right: string beside the
groove. Bold line: fretboard-normal string oscillation. Thin line: fretboard-parallel string oscillation. ---- = Sum.
Before thinking about how the wood of the guitar body could affect the string oscillation, we
should first consider those components that are in direct contact with the strings. These are in
fact the nut (or fret) and the bridge saddles – but not any pieces of ash or alder. If the string
does not rest on a line that is perpendicular to its longitudinal axis, a coupling of the
oscillation planes may result. The same might happen if the compliance of the support is
direction-dependent. The coupling of the transversal oscillations as it is caused at the string
bearing is shown in Fig. 7.69 as an orbit-diagram (abscissa = fretboard-parallel oscillation,
ordinate = fretboard-normal oscillation). In the upper-left diagram we can see how the string
first begins to oscillate vertically, but then subsequently shifts the oscillation plane first to the
left, and then to the right. After about 370 ms, the vertical oscillation has nearly decayed to
zero, and the oscillation energy has mainly been transferred to the orthogonal component.
This is completely different for the B-string when positioned beside the groove of the
headstock saddle: it substantially keeps its oscillation plane, because the coupling between
both oscillation modes is much smaller (bottom images).
Fig. 7.69: Orbit-diagrams (vertical vs. horizontal movement). B-string, Top: bearing in the nut-groove, bottom:
bearing beside the nut-groove. Stratocaster, 10th partial of the B-string (2476 Hz). The analysis had been run
with signals that were similar but not identical to those used for Fig. 7.68.
When investigating damping (dissipation processes), we need to analyze both oscillation

planes. If merely the voltage generated by the pickup is of interest, only the fretboard-normal
oscillation-component is essential. That common magnetic pickups can pick up not only
transversal oscillations but also longitudinal oscillations is explained in Chapter 2.9, while the
directional characteristic of these pickups is looked into in Chapter 5.11.
The mode coupling at the headstock saddle (nut) of the B-string found in the above example
is, of course, only relevant as long as the open B-string is plucked. As soon as the string is
pressed down on the fretboard by a finger, the fret that is next to it takes over the bearing
function. Furthermore, corresponding coupling may just as well occur at the bridge saddle -
and this will have effects also when the string is fretted. The bridge construction of most
electric guitars encourages the assumption that the designers did not worry about mode
coupling, but predominantly considered as their task the adjustability of the action, and lowest
possible production costs. On the Jazzmaster (planned to be Fender’s top model), Leo Fender
guided the strings at the bridge by means of screw threads. However, he did not use screws
with six different threads – no, three different threads had to be enough. As generally known,
the strings have six different diameters, and therefore the fit for the strings will turn out to be
very different from string to string … What? Fit?? On the Tune-O-Matic bridge, Gibson
guides the strings by means of bridge saddles looking fishily similar – all six of them! The
guys at Rickenbacker lay the strings into small rollers, probably hoping that the gap damping
won’t become all that pronounced. And surely: there are six identical rollers! Obviously, not
all builders of electric guitars were aware to the same degree of the function of the guitar
bridge in terms of vibration technology.
More details regarding bridge constructions are compiled in Chapter 7.10.
7.7.4.2 Damping of longitudinal waves

Chapter 7.7.4.1 had shown that a coupling of transversal string-vibrations occurs at the bridge
and at the nut (or fret). In addition, transversal and longitudinal oscillations exchange part of
their oscillation energy, as well (Chapters 1.4 and 7.5.2). The dilatational waves induced that
way showed high loss factors in the decay measurements: individual partials decay rapidly,
i.e. they exhibit short decay times. For the following vibration measurements, a Fender US-
Standard Stratocaster was used with its tremolo (aka vibrato) genre-typically adjusted to be
floating. The investigated string was plucked fretboard-normally close to the nut; an
oscillation analysis was made close to the bridge using a laser vibrometer.
Fig. 7.70: Decay (left) and time function of the fretboard-normal velocity. Dilatational wave period = 0.42 ms.
For the fretboard normal velocity, the left-hand image in Fig. 7.70 shows the decay time of
the D3-partials. Damping maxima – i.e. T30 minima – can be identified at 2.36, at 4.7 and at
7.1 kHz; resonances of dilatational waves can be assumed to be the cause. In the time
function we can see that - even before the transversal wave arrives at the measurement point –
small impulses with a periodicity of 0.42 ms occur. Although the laser vibrometer (which is
sensitive to lateral string oscillations) cannot itself detect the dilatational waves, it does
capture their secondary waves (Chapter 1.4). Apparently, dilatational waves are absorbed
efficiently in the wound D-string, and a selective damping arises at a frequency of 2.36 kHz
(and its multiples).
Depending on how well the resonance frequency of the dilatational wave matches the
frequency of the partials, this dilatational-wave damping can be more or less pronounced. The
measurements done until now let us assume that especially the fretboard-normal oscillation
can transfer its energy to the dilatational wave; the cause could be the curvature of the string
at the bearing (Chapter 7.5.2). In Fig. 7.71, the level drops of the partials of the D-string are
represented: the fretboard-normal oscillation decays very fast at 2364 Hz, while the fretboard-
parallel oscillation exhibits a decay time as it is found with the adjacent tones.
Fig. 7.71: Level drop of partials; bold = fretboard-normal oscillation, thin = fretboard-parallel oscillation
7.7.4.3 Residual damping

Generally, the string does not end at the bridge or the nut but passes over it to its actual
mounting point. In certain circumstances, these remaining sections of the strings (residual
strings) located beyond the main section of the string may form an effective absorber that can
deprive the main section of the string of oscillation energy. This is termed residual damping.
If the string would exhibit a pure transversal movement, it could not transfer energy to the
residual string across the fixed support bearing. However, as was already explained in
Chapter 2.7, the string is also subject to a bending stress, and the related bending moment
acts across the bearing and excites the residual string. Also, the longitudinal forces occurring
within the string (→ dilatational wave) may at least partially act across the bearing –
especially for small bend angles, the string may relatively easily slide across the contact area.
Fig. 7.72: Modification of the decay of partials (at the specified frequ.) due to mass loading by the residual string
at the bridge; Gibson ES-335 TD; “H-Saite” = B-string, “G-Saite” = G-string, “mit Klemme” = with clamp.
To quantify the effects of this residual damping via two examples, a string of an ES-335 was
plucked fretboard-normally near the nut; measurement of the fretboard-normal velocity was
done near the bridge saddle using a laser vibrometer. As a modification, a small clamp was
attached to the residual string near the bridge (Fig. 7.72, arrow). The measurements were
carried out for the B- and G-string, with always the plucked string being measured and
modified. For many partials, no considerable effect resulted – but in some cases the decay
was indeed influenced. This happened in different ways: for the partial of the B-string shown
in the left picture, the additional mass improves sustain and level, while in the other example,
the additional mass chokes off the oscillation rather rapidly♣.
It is difficult to formulate these damping mechanisms analytically because two transversal

modes and one dilatational wave occur in combination – in fact on both sides of the bridge!
Therefore, these examples only serve to show that the effect of the residual strings must not
be generally neglected. However, because the decay of only a few partials will vary, the sonic
impacts remain fairly low. With the investigated ES-335, no audible difference in the
"electrical sound" could be found when damping the residual strings during playing with the
heel of the hand.
It is obvious that such a damping mechanism cannot be found with measurements at an empty
bridge (bridge without string). On the other hand, the saddle conductance (Chapter 7.7.4.4)
can only be determined without the string because the location of the string bearing can only
be allotted once to one single taker. Already the ancient philosophers knew: where there
already is something, nothing else may be.
♣
For the sake of completeness it is noted that even between the individual strings und their partials, vibration
coupling and thus damping may occur – this effect will not be further investigated here, though.
7.7.4.4 Bearing conductance

All damping mechanisms considered so far had their origins in the string or in the
surrounding air – the guitar and especially its noble tone-wood were not investigated as being
involved in the sound shaping. However, they of course also affect the string oscillation, and
therefore now a more detailed analysis of the mechanical properties of both string bearings
will follow. Consideration of the string as a waveguide (Chapter 2) shows reflection processes
that can approximately be described by the wave impedance of the string, and by the bearing
impedance. The wave impedance is a string-specific quantity (Chapter A.5), the bearing
impedance is formed by the nut and the bridge saddle. However, not only these play a role but
also their substructures, i.e. bridge base, and neck and body of the guitar. The bearing
impedance is the mechanical impedance Z = F/v found at the bearing by a wave running
along the string. An immobile, rigid bearing features a velocity of v = 0, and therefore the
bearing impedance of an ideal bearing is infinite. Such a perfectly loss-free bearing would
show perfect (i.e. loss-free, total) reflection – but this only occurs in the ideal model. Every
real bearing absorbs a small part of the incoming wave energy (e.g. 1%) so that e.g. only 99%
will be reflected. The more often per second this absorption occurs, the faster the string
oscillation decays. Assuming 1% of energy loss at each bearing for a string oscillating with
100 Hz, a wave reflected 200 times per second at each bearing will have only 0.99200 = 13%
of its initial energy after 1 s. The corresponding level-decrease would be 8.7 dB; for a string
oscillating at 200 Hz, the energy would have decreased to 1.8% after 1 s (i.e. by 17.4 dB).
The bearing absorption may be described by the bearing conductance G. This is the real part
of the bearing admittance (admittance = 1 / impedance, for more detail see Chapter 7.5.3).
The higher the conductance, the more the bearing absorbs, and the shorter the “sustain”. On
the one hand, the power absorption factor of a bearing is proportional to the wave impedance
of the string, and on the other hand it is proportional to the bearing conductance. With the
wave impedance of each string being proportional to its diameter squared, we get: the heavier
the string, the more the bearing damping affects the string oscillation. In Fig. 7.73, the power
absorption factor is given percentage-wise for three string sets, with G = 0.001 s/kg.
W = wave impedance
D = diameter of the string
ρ = density of steel
f1 = fundamental frequency of the string
a2 = degree of power absorption
2
Fig. 7.73: Degree of power absorption a for three different string sets.
A transversal wave running along the E2-string will, depending on the string thickness, lose
0.22% – 0.44% of its power at a bearing which has a conductance of 0.001 s/kg. For the E4-
string this would only amount to 0.04% – 0.1%. For comparison: given these conditions, a
power loss of about 1% would result for the E1-string on an electric bass! It must be borne in
mind, though, that the wave propagation speed decreases with decreasing frequency, as well –
on the E1-string of an electric bass, the transversal wave arrives at the absorbing bearing
significantly less frequently (sic!) than on an E4 string of an electric guitar. Therefore, two
processes working in opposite directions dominate the frequency-dependency of the decay
time: the process of decreasing absorption from the bass strings to the treble strings, and the
increasing frequency that the absorption happens with.
The bearing absorption caused by constant conductance (e.g. G = 0.001 s/kg) is, in a simple
model, of the same value for all partials of a string: here, both the wave impedance and the
conductance are constant. And because in the simple model (i.e. leaving aside dispersion) all
partials of a string propagate with the same wave velocity, the decay time correspondingly
caused does not show any frequency dependency, either. Thus, given an overall consideration
of various absorption mechanisms, the frequency-independent bearing absorption defined for
a constant G will mainly have an effect in frequency ranges where other absorption
mechanisms are weak, i.e. in the low-frequency range, and for the bass strings. For real string
bearings the conductance is not constant, though, but rather frequency dependent. Fig. 7.74
shows related measurement values gathered within the nut groove of the E4 string of a Les
Paul Historic (with the string taken off). Eigen-oscillations of the open string are possible
only at positions marked by dots, and only here the measured conductance values have any
impact on for the decaying oscillation of the E4-string.
Fig. 7.74: Les Paul, E4-string: conductance (“Konduktanz”) at the nut (left), calculated decay times (right).
The right-hand diagram shows calculated decay times for the partials of the E4-string
considering the attenuation by radiation, the internal dissipation, and the bearing absorption.
One bearing absorption only - because the bridge saddle had not been considered yet. In
general, this calculated curve stands up nicely to measurement curves. Not that this is all that
surprising – T30 is, in the end, predominantly determined by the attenuation by radiation and
the internal dissipation. The bearing absorption dominates only if a conductance maximum
happens to be near the frequency of a partial frequency, and in that case a selective absorption
maximum results (i.e. a selective minimum in T30). For the fundamental of the E4-string (at
330 Hz) this is nearly the case: if one would merely tune the E4-string down by approximately
a semitone, the decay time of the fundamental would be reduced to half (2.2 s). On the other
hand, the decay time of that fundamental may also be extended up to more than 7 s, for
example if the guitar is laid in a different way onto the measuring table for the conductance
measurement (Fig. 7.75). However, only the damping of the fundamental will change in this
case, all other T30 minima remain practically unchanged.
Fleischer [8] has published a variety of different impedance plots for various guitars,
measuring not only at the nut or bridge saddle, but at each fret, as well. These and further
investigations [Fleischer 2001, 2006] indicate bending and torsional vibrations of the guitar
neck – causing low-frequency bearing absorptions. If the string bearing happens to be at a
node of the neck oscillation (in consideration of the frequency relations), small conductance
and thus long sustain result, bearing at an anti-node position yields high conductance and
"dead spots". Once again, it is shown that a noticeably resonating guitar neck may delight the
sense of touch – but it is likely to be detrimental long sustain in one way or another.
Fig. 7.75 shows how the conductance at the nut can be changed without permanently
damaging the guitar. For both measurements shown in the left-hand image, the guitar (again
the Les Paul Historic) was placed on a stone table, supported underneath the neck/body-
interface by a soft mouse pad. The other bearing – the edge of the body near the rear belt pin
– was placed directly onto the stone table for one of the measurements. For the other
measurement, a second mouse pad served as a cushion (and as damper). As a result, we see
pronounced resonance shifts below 400 Hz, but there is practically no change in the frequency
range above. On the one hand, this indicates a good reproducibility; on the other hand it
shows that low-frequency modes of the neck vibration depend on the bearing of the guitar – to
the vibration engineer, that’s not actually a highly unexpected behavior.
Fig. 7.75: Les Paul, conductance (“Konduktanz”) at the nut: E4 (left), E2 (right). Mechanical modifications.
In the right-hand diagram, the differences are caused by a vise mounted to the headstock. This
now is an approach that tackles the situation in close proximity of the string bearing – the
effects therefore are bigger than those in the left-hand diagram. Neither result can be
interpreted as improvement, or as deterioration: both have an impact on all strings. Even
though the decay time of one partial may be extended according to Fig. 7.75, it is to be feared
that, at the same time, the decay time of another partial is reduced.
Fig. 7.76: Les Paul, bridge saddle conductance (“Konduktanz”), E4. Modifications = clamp mounted to the
bridge.
The bridge saddle conductance of the Les Paul Historic, measured at the E4-bridge-piece, is
shown in Fig. 7.76. From his oscillation measurements, Fleischer concludes that the neck of a
solid-body guitar is relatively flexible whereas the bridge remains relatively immobile.
Our own measurements confirm this for the frequency range up to 700 Hz (the range
investigated by Fleischer). However, for higher frequencies, and depending on its design, the
bridge absolutely may show some veritable Eigen-oscillations, and thus may become an
efficient absorber. In Fig. 7.76 three measurement curves are shown in either diagram: one for
the guitar in its original condition, and two more for the bridge modified via fixing a clamp on
it. In particular the conductance maxima – important for the string damping – react to these
modifications, leading us to the assumption that these maxima are bridge resonances. This
hypothesis found support via measurements using a laser vibrometer: significant bridge
oscillations showed up in critical frequency ranges. At low frequencies, the bridge is nearly
immobile, and thus an attached additional mass attached will not bother it. However, there
are strong bridge resonances between 1 and 1.5 kHz, as well at around 4 kHz, and those will
change when attaching an additional mass.
Supplementary findings regarding the effect of the bridge design on the decay of partials of
the string-vibration were provided by measurements with a non-trem Strat. Two variants are
common as bridge saddle: on earlier Strats, the string was fed through an S-shaped sheet
metal – the vintage bridge saddle – that could be adjusted with three adjusting screws. In late
1971, the design was changed to the solid die-cast (injection-molded) bridge saddles still
customary today [Duchossoir]. For both bridge-piece designs, the decay of the partials of a
0.013" B-string was analyzed. Fig. 7.77 (left-hand image) shows corresponding decay times.
Disregarding – for the moment – the smaller variations in the curve, we find the following:
the string supported by the injection-molded bridge saddle (continuous line) shows a behavior
nicely approaching the orientation line given by radiation attenuation and internal damping.
Conversely, the decay time of the string supported by the vintage bridge saddle is only about
half as long at high frequencies. The explanation is simple: The sheet-metal bridge saddles
bend easily, and thus absorb more than the solid design. So: do upload the graph to the
Internet – and we have one more ineradicable rumor.
Fig. 7.77: Decay times of the B3-string (= “H3”) of a non-trem Stratocaster. Left: solid (–––) or vintage (---)
bridge saddle. Right: solid bridge saddle (–––), other specimen of vintage bridge saddle (---).
To re-check, the solid bridge saddle was mounted to the guitar again: the measured curve
(right-hand graph) is quite comparable. Then it was sheet-metal saddle’s turn again; however,
a different specimen was used: different results show. Fig. 7.77 unambiguously indicates that
the bridge saddle affects the decaying oscillation of the string to a not inconsiderable extent. It
therefore participates essentially in the shaping the sound. Obviously, there are non-negligible
manufacturing tolerances in the bridge saddles – not surprising when taking a closer look at
the particular construction. As Kollmann [1993] notes very persuasively: the gap absorption
is the most important damping mechanism in machine acoustics.
There is a generous helping of gaps within the construction of the Stratocaster bridge, e.g.
between the saddle and its three screws, between the screws and the support plate, and of
course between the string and the saddle. The whole contraption does not seem to be expert-
optimized in terms of its damping properties; therefore it may actually be even expected that
each bridge develops an individual life of its own, and its individual damping character.
To clarify this once again: given such pronounced inter-individual scatter we cannot maintain
that the vintage saddle will absorb significantly differently compared to the solid saddle.
Instead, we only may conclude that even identically constructed saddles may differ in their
damping properties.
The damping processes presented so far shall in the following be summarized in an example.
The measurements were carried out on a Gibson ES-335 equipped with new strings (9/46).
The A2-string was plucked fretboard-normally near the nut; its oscillations were detected two-
dimensionally with two laser vibrometers. The left-hand section of Fig. 7.78 shows the
evaluation of the decay times of the partials. Up to about 1 kHz, the minima can be attributed
to neck resonances, the two dips between 1.5 – 2 kHz are related to dilatational wave
resonances and to bridge resonances, respectively.
Fig. 7.78: Decay times of partials of the open A2-string of a Gibson ES-335; different bridge positions.
The bridge of the ES-335 is of the famous "Tune-O-Matic” type. As it often happens with
celebrities, there is an obvious tendency towards lability. In particular, the bridge is given
height-adjustment – and it can move laterally because some excessive clearance has been built
into it. The right-hand section in Fig 7.78 shows a family of curves that results from the
bridge being moved laterally. The overall trend remains while differences appear in the
details. For a Les Paul (Fig. 7.76), it already has been demonstrated how the string damping
caused by the bridge can be modified by mounting a small clamp. Fig. 7.79 now gives
additional proof. In the left-hand section of the figure, the decay times for the A2-string are
shown: once for the guitar in its original condition, and once more for a modification (a clamp
on the residual string at the bridge). Especially around 1 kHz the decay of the partials changes
– suggesting the combination bridge/residual-string to be a possible source of attenuation. The
right-hand section of Fig. 7.79 shows a velocity spectrum. It is gathered with a laser
vibrometer, the beam of which was focused directly beside the A2-saddle onto the bridge
below it. To measure, the A2-string was plucked fretboard-normally near the nut. An
oscillation maximum can be seen between 1.5 and 2 kHz – obviously there must be a bridge
resonance here. And once again we get confirmation on what guitar magazines have a hard
time to grasp: bridge oscillation = string damping.
Fig. 7.79: Decay times of the partials of the empty A2-string on a Gibson ES-335. Left: original condition (---),
small clamp on residual string at the bridge (---). Right: velocity spectrum of the bridge, next to the A2-saddle.
The fundamental frequency of the resonances of the A2-dilatational-wave is at 1.8 kHz – these
resonances can contribute to the attenuation, as well (3.6 kHz). Compared to the area offset in
grey and marking the global shape of the string-damping curve, the decay time of the ES-335
shows characteristic deviations. These are even more striking if we do not evaluate the string
vibration two-dimensionally, but analyze only the fingerboard-normal string vibration (just as
the pickup would). The corresponding decay times are shown in Fig. 7.80. Differences
between the two types of analysis can be attributed to non-exponential decay (Chapter 7.6.3);
beats or salient curves lead to ambiguities. Differences between the results for the ES-335 and
the Stratocaster analyzed in Fig. 7.80 need to be discussed with regard to two focal points: Up
to approximately 1 kHz, neck resonances determine the string damping, and in the frequency
range above there are mainly bridge- and string-specific processes. The drop of the ES-335
between 1.5 and 2 kHz clearly has its cause in a bridge resonance, possibly amplified by a
dilatational-wave resonance. The latter are also highly likely to be the cause for the minima at
3.7 and 5.4 kHz. Not looking at these specifics, only small differences remain in the range
above 1 kHz. These small differences moreover change in many details as minor shifts are
made to the respective bridge saddle. Therefore: although the two guitars differ considerably
in construction (Strat = solid-body, ES-335 = thinline), the treble range of the string vibrations
is determined by the string and its bearings only. There is practically no influence by the
wood. Below about 1 kHz, neck resonances (very selectively) determine the string damping,
and only here does the wood have an impact. The wood of the neck, that is! Although the
body as a bearing for the neck is also involved, the bending- and torsion-resonances of the
neck are the decisive factor.
Fig. 7.80: Decay times of partials in a Gibson ES-335 (left), and a Fender Stratocaster (right). The thick line
refers to the fingerboard-normal string vibration; the thin line refers to the two-dimensional analysis.
Fig. 7.81 shows a similar comparison, but now for the E4-string. In the 2D-analysis, there are
only small differences; these may in part be due to the fact that the string diameters were
different. In the Stratocaster, some partials decay with a beat, this leads to the already
discussed discrepancies. In direct comparison they are just about audible, but do not have
their cause in either the pickup magnets (completely lowered for this measurement), nor in the
body wood, but exclusively in the string bearings. The guitar body certainly has considerable
impact on the radiated airborne sound, but for the voltage generated by the pickups, it is
insignificant as long as typical design rules are not grossly violated.
Fig. 7.81: Decay times of the partials of an ES-335 (E4, 0.009", left), and a Stratocaster (E4, 0.010", right). The
bold line refers to fingerboard-normal string vibration; the thin line refers to the two-dimensional analysis.
The T30-differences found so far shall be discussed again with consideration of musical
requirements. How relevant is the difference between, e.g., T30 = 3.0 s and 2.5 s? For a tone
duration of 0.5 s (a quarter note at 120 bpm), a level drop of 5.0 dB occurs at T30 = 3.0 s, and
6.0 dB at T30 = 2.5 s. By contrast, the level of a partial may change by 10 dB (or much more)
when the string is plucked an inch or so closer to the bridge! This is not to say that a short
decay time can generally be compensated with a higher level. These are entirely independent
quantities to start with – they do now receive a special joint assessment by the hearing
system. Defining "Attack" as the first section of approximately 100 ms of the guitar tone, we
can choose a time span that corresponds to the integration time of the ear [12]. During this
time-span, psychoacoustic “trading” between initial level and decay time is actually possible.
However, the change in the location where the string is plucked has a much greater effect on
the sound than e.g. the differences shown in Fig. 7.81. Listening tests confirm this: you can
almost always hear differences, but in most cases these are due to slight differences in the
picking location or in the way the plectrum is held. There is no denying that substantial
physical differences exist between T30 = 1.5 s and T30 = 0.4 s (Fig. 7.80) – however, if these
differences occur at 4 kHz, their auditory relevance is very low. In fact, the ear combines into
a joint processing about 7 partials in the corresponding critical band (the hearing-related
frequency-range division); thus the level of one single partial does not play a significant role.
Also, we must not forget that the decay times shown so far have all been measured with
brand-new strings - just a few minutes of (more or less) virtuoso playing will deposit skin, oil,
and fat particles on the wound strings – significantly reducing the decay time, and thus even
more significantly reducing the influence of any parameters of the guitar body (translator’s
note: if there are is such an influence at all). So again: it's in the fingers, in every respect …
7.7.5 Damping by finger, hand and capo
Placing the guitar on a heavy stone table in order to do measurements usually guarantees good
reproducibility, because no trembling hand moves the object to be examined. However, such a
bearing is not very close to practical reality, because of course the strings are touched while
playing. The fretting hand (mostly) rests against the back of the guitar neck, and a finger
presses the string against the fret. From the point of view of vibration technology, the neck of
the guitar is a cantilever and/or torsion beam; despite being of a stiff structure, it can
nevertheless be bent and twisted – not by much, but to a significant degree. Contact with the
hand and/or finger alters the mechanical vibration parameters of the neck, and influences
resonance frequencies and their damping. The bendable and rotatable guitar neck resonates as
soon as a string vibrates, sending vibration energy into the hand and the finger(s). The more
vibration-energy our hand and fingers pick up, the stronger the string will be damped. The
exact differences between active and reactive energy shall not be discussed in more detail
here; the measurement results are self-explanatory.
Fig. 7.82: Decay times of partials of the string vibration, Stratocaster. The dashed curves were measured with a
capodastro placed on the 3rd fret for measuring of the B-string (“H3”), and on the 4th fret for the G-string.
In order to keep the guitar as still as possible, a Shubb capo was mounted to the neck of a
Stratocaster, replacing the gripping hand (Fig. 7.82). As with the other analyses, the string to
be measured was struck fingerboard-normally next to the capo, the measurement of the string
velocity was taken two-dimensionally near the bridge. The exact capo position and its contact
force probably play a role, but detailed investigations were not planned at this point. Instead,
the principal effect was to be demonstrated; this is successfully done with Fig. 7.82: the capo
acts mainly as an additional damper, and it reduces the decay times especially in the high-
frequency range. Similar results were found in experiments with a hand placed on the back of
the guitar neck; depending on circumstances, the damping caused by the hand may be even
more pronounced than that produced by the capo. Since attaching the capo will change the
pitch of the string and the position of its bearing, other pairings between frequencies of
partials and frequencies of neck resonances occur – this will bring other selective resonance
dips into play. In view of such grave effects, it does not make sense to pay very close
attention to small dents in the T30- curve. The sound changes caused by the plucking/picking
hand and fretting hand will dominate compared to most damping mechanisms due to bearing
of the string. Only a few resonances of bridge, neck and/or string will be able to shorten the
decay time of individual partials considerably – corresponding listening tests are summarized
in a separate section.
7.7.6 Ageing of the strings
The decay diagrams shown to this point in Chapter 7.7 have all been measured using fresh
strings – a condition that never lasts long. Three processes predominantly change the
mechanical properties of a string: the surface corrodes, the cross-section changes, and
deposits of grease and skin-particles appear on the string. Solid strings (usually G3, H3, E4) are
always made of spring steel; the magnetic and mechanical requirements leave little room to
maneuver for manufacturers here. The core of wound strings (E2, A2, D3) also consists of
spring steel, with the wrap spinning made of nickel wire or steel wire. Brass or bronze wound
strings are only found on the acoustic guitar, since these materials are of a non-magnetic
nature and generate too weak a voltage in the magnetic pickup.
When playing, the string is pressed against the frets, and moved back and forth with (possibly
virtuoso) finger vibrato. Even if the string is made of hard steel: over time, small grooves are
ground into its lower side, and these affect the propagation of the waves generated during
plucking. Each groove is a discontinuity (a local change in the wave impedance) that leads to
unwanted reflections, detuning of partials, and vibration-damping. The most serious cause of
the change in the string parameters, however, is the accumulation of skin-, grease- and dust-
residue, especially if this gets into and between the layers of the wrapping. These deposits are
efficient absorbers that can extract much more vibration energy from the string than all
previously presented absorption processes. In Fig. 7.83, the left-hand picture shows the decay
time of partials of the A-string of a Les Paul Classic. This was a "not entirely fresh anymore"
A-string the history of which could not be determined more precisely. Compared to a fresh A-
string, the decay time has decreased by 50 – 80%; in particular the treble fades away much
faster. Compared to this very efficient damping mechanism, only two vibration absorbers can
compete: a neck resonance at 220 Hz, and the known bridge resonance around 2 kHz. After
just a few minutes of typical guitar playing, grease and skin deposits already can lead to
measurable effects – therefore comparisons must only and always be done with completely
new strings!
Fig. 7.83: Decay times of partials, on a Les Paul Classic (left), and on a stone table (right). Note the extended
ordinate range!
The graph on the right shows the decay times of the partials for a very old E2-string stretched
across a stone table. The high-frequency signal components (that are still present at the
moment of plucking) lose their energy instantly; the T30 treble-dropoff occurs with a slope of
f to the power of 3. Nevertheless, such a string is not completely useless: bring in a distortion
device, and the string rises to new heights … well, the treble is kind of revitalized, anyway.
7.7.7 Flatwound strings
The bass strings of the electric guitar are wound to reduce flexural rigidity and inharmonicity
(Chapter 1.2). If the fingers glide over the grooves of the wrap spinning during play,
characteristic scratching or sliding noises are generated. These noises may be perceived as
typical, or as annoying, or both. To offer a choice, flatwound strings are available that are not
wound with round wire, but with wire of the cross-sectional shape of a flat band, this allowing
for an almost smooth string surface. The winding produced in this way is often two-ply: the
inner wrapping is done with round wire, the outer with the band-shape (Fig. 7.84).
Fig. 7.84: Flatwound (left), roundwound (right).
The decay time measured with a flatwound string set is shown in Fig. 7.85. Except for the G-
string (wound with merely one layer of round wire in this set), a stronger treble damping is
evident. This increased damping is due to larger internal friction losses within the string build.
In addition to the reduction of the gripping noise, the flat winding therefore also causes a loss
of brilliance, as it would be found in a similar way in old roundwound strings.
Fig. 7.85: USA Stratocaster, fresh flatwound set (22/30/40/50). G3 = round-wound.
7.8 Specialist journalism 7-87
7.8 Specialist Journalism
"Only write about subject matter that you are familiar with!" Any reference book on
journalism will teach that – or at least it should teach it. So, why would a renowned
neurobiologist write: "Presumably, due to the homogeneous and generally similar structures
in the neo-cortex ..."? If he comprehends how neuro-plasticity works, he wouldn’t have to
assume anything, would he? And if he does not have sufficient comprehension, he shouldn’t
write a book about it, now, should he!? Well, if only the matter were really exactly known! In
the fringe areas of research, at the borderline to terra incognita, conjecture is indeed quite
permissible. And thus, a scientist may publish a new idea in the hope that soon proof will
appear, and that – given that he has more of these ideas – he will eventually be regarded as the
genius forward thinker. Even though one day it transpired that more than the six planets
hypothesized by Kopernikus are revolving around the earth, the book published by him still
remains one of the biggest steps ever taken in the recognition of natural processes♣.
Is the above sufficient to legitimize the conjecture found in many a specialist magazine for
guitar and bass: that the vibration of the string should be conducted as comprehensively as
possible into the instrument body? In corresponding elaborations, the restriction
“presumably” does not even appear anymore … apparently there is a common consensus
about the “fact”. Or maybe it is common nonsense – the corresponding circles are, after all,
known for a notoriously carefree handling of loanwords and technical terms. Popular mix-ups
are between inductance and induction⊗, Oersted and Gauß, stiffness and compressive strength,
bifilar winding and criss-cross winding. An alleged “expert” in circuit design introduced
himself with the kind words: myself, I often can’t distinguish between when currents “flow”
and when a voltage “is on”. Doesn’t matter, the guy is allowed – in spite of such profound
moxie – to write up a monthly column about circuitry details in tube amps, praise the brash
vibrations of the guitar “that you feel in your gut” as a quality attribute, and discuss relaxation
effects in guitar cables as soon as he doesn’t confuse the spelling of the term anymore. Are all
these areas in fact fringe areas of research? Has nobody discovered the law of conservation of
energy, the permeability, the Young’s modulus, or the susceptibility? Of course they have –
and quite some time ago, too! If you still surmise that tone wood will only by deformed
permanently if the weight of 9 Leopard tanks presses down on the area of a 1-Euro-coin, you
expose yourself to ridicule; you’re not a specialist journalist but a universal dilettante then.
A journalist should impart knowledge, not stupidity. How can somebody be allowed to roam
free while seriously claiming that the pickup-voltage of a Les Paul would change
‘significantly’ if its paper-thin layer of varnish has cracked in a few places? Or that the
coloring (!) of the wire-insulation has effects on the sound? You may have such thoughts …
but only in private. The reader does not have to know what inductance means in a pickup, and
what induction brings to the table. Not knowing indeed often is motivation to read up. The
author, however, really does need to know – otherwise he should dispense with publishing.
One myth rattling around in guitar literature is the topic of the next two chapters: allegedly,
the body of an electric solid body guitar has a “primary tone” that shapes the sound to the
extent that the pickup can only add a few nuances. It’s ok to speculate about that, but it may
not be corroborated with ludicrous assertions as it happened in the article “Stratone” (see
Chapter 7.8.3). Sorry, dear author, but you have crossed the Rubicon. If you're looking for
trouble, you’ve come to the right place. There’s no more mere rolling of the dice from hereon.
♣
Der große Herder, Volume 7
⊗
That’s about as small a difference as between astronomy and gastronomy.

7.8.1 The legend of the primary tone
A citation from Gitarre&Bass 11/05: Let us first distinguish between two categories of
parameters that decisively influence the sound (primary tone) of a guitar. On the one hand,
there is the basic construction that is determined by the selection of wood, the geometry, and
the craftsmanship of the builder. The second category describes the (flexible) tuning of the
hardware and the process of playing-in the guitar. Type of wood, geometry, and processing –
without doubt these are important elements of the electrical guitar. The expensive but not very
solid balsa-wood is as unsuitable for a guitar neck as a 5mm-thick plank from spruce would
be as the sole constituent for the guitar body. Glue joints of a width of a millimeter are not
evidence of great craftsmanship – even if found in expensive instruments [G&B 1/09].
However, what significance do differences have that occur within the well-trodden paths of
decades-old tradition? Ash vs. alder, polyester vs. nitro lacquer, 3.1 vs. 4 kg, brass vs.
aluminum? Corresponding opinions range from completely irrelevant to fundamental, and the
rationales are as inconsistent as the proclaimed dogmas. The following statements are given
by one and the same author: the so-called old-wood tone of 50’s Les Pauls is of such
excellence that these icons cost a fortune today [G&B 10/08]. Personally though, I do think
that there is not much to the legend of the old wood [G&B 2/07]. A real 59 is, after all, the
holy grail that bewitches our ears. Apparently, there is something to that. [G&B 3/08]. The
so-called vintage market is a first-grade web of prejudices [G&B 4/06]. The sound depends
on the selection of the wood first [G&B 3/06].
Of course: the wood: The electrified solid-body guitar predominantly is an acoustic

instrument. The woods determine the character of the sound; the pickups contribute only very
little [G&B 02/00]. It should follow that if we change the wood, the sound would also change.
That is the primary tone, i.e. the sound in the air as radiated by the guitar itself, and as a
consequence allegedly also the “electric sound” generated by the loudspeaker. It is taken to
such an extent as self-evident that the electric sound is coupled inextricably to the sound
coming from the body, that test reports often not even clarify anymore what is designated
“smack”, “throaty”, or “funky” in this area. But is it really true that the guitar played “dry”
already reveals “it all”? Scientific theory vehemently objects – but the true old-wood guy will
not be ruffled in his opinion by something as academic as vibration theory … especially if he
has just spent a small car’s worth of dough on selected premium wood. Well then – let’s
experiment. Since it is practically (!) impossible to play an electric guitar with sufficient
reproducibility the same exact way again after exchanging the body, we choose a different
approach. A Stratocaster (American Standard) was played in the anechoic room in turn with
and without solid contact to an (open) wood cabinet. With contact indicates that the lower
bout of the guitar was placed on the cabinet; without contact means that the guitar was at a
distance of a few centimeters from the enclosure. Using this method it was possible to enlarge
the guitar body by a resonator of a volume of 75x39x25 cm3 and to more than double its
weight. The airborne sound was recorded using a measurement microphone (B&K 4190)
positioned at 10 cm in front of the neck pickup; in parallel the voltages generated by the neck
pickup was also captured. Just like in the “Stratone”-report [G&B 5/07], an experienced
guitarist continued to play an E chord in the lowest position. As was to be expected, the
change from “without contact” to “with contact” was clearly perceived when listening to the
airborne sound: the guitar sound stronger and fuller with contact – the drastically enlarged
body converted the Strat into a kind of semi-acoustic guitar. So, with the purely acoustic
sound telling us the whole story about the primary tone of this modified Strat, the electric
sound would have to show a similar difference. Getting a bit uneasy now, dear wood-freaks?
A small difference maybe? A slight trend, at least? Mind you, the body of the Strat has
increased by half a square meter. Let’s listen, and let’s take some measurements.

.Pickup. .Pickup.
Airborne Airborne
Fig. 7.86: Stratocaster. Left column: body with (–––) and without (----) contact to the wooden box.
Right column: with/without contact to the neck. Middle row: pickup output, bottom row: airborne sound
In its right-hand column, Fig. 7.86 shows 1/3rd-octave spectra of the pickup voltage and of the
airborne sound. The standard deviations of the 1/3rd-octave levels across 20 E-chords with and
without contact amounts to about 1 dB; therefore the differences visible in the graph for the
pickup-spectrum are definitely insignificant while the differences in the airborne sound
certainly are. However, we wanted to avoid getting too theoretical – so let’s move on to the
listening test: the microphone recordings reveal every single contact change, while from the
pickup signal not a single change can be identified. That’s 100% against 0% - it does not get
any clearer than that. Increasing the Strat body changes the airborne sound (the primary tone)
so dramatically that even the layperson will recognized the difference. For the pickup
signal, even the expert cannot hear any difference.

Just to be safe, the experiment was repeated the next day, this time using 2x32 E-chords. The
result was the same; changes in contact were very clearly audible in the airborne sound while
there was no way to hear from the pickup signal whether the guitar had any contact to the
wooden box or not. There is a simple explanation: the retroactive effect of the guitar body on
the string is so minute that the pickup voltage changes only insignificantly. From a very
theoretical point of view, the voltage must change a tiny little bit, after all – but we were not
going into much theory here: that was considered extensively in Chapter 1.6. Or by Fleischer
[8], who puts the differences between acoustic and electric guitar very nicely in a nutshell:
while in an electric guitar the strings are to be kept as immobile as possible at the bridge,
they have to rest on a moveable bearing in the acoustic guitar. There is no other way to
convert the string signal into sound in a purely mechanical manner without electro-acoustic
means. In the acoustic guitar with its efficient coupling between string vibration and
airborne sound, the material of the top merits highest attention because the top needs to
vibrate and radiate the main share of the airborne sound. Compared to this situation,
vibrations in the body of a solid electric guitar are of subordinate significance. This does not
mean that just any material would be suitable for the electric guitar, and that its body has no
impact at all: rubber would indeed be not conducive. However, if contact to half a square-
meter of blockboard has no audible effects onto the electric sound, the issue of whether the
body is made of ash or of alder moves rather far into the background.
Besides the body, there is another resonator that is the object of any evaluation of a guitar: the
guitar neck. Due to its elongated, relatively thin shape, it is predestined to exhibit Eigen-
oscillations (natural oscillations), and its material is subject to extensive speculations, as well.
Therefore, in a second experiment, the headstock of the Stratocaster was brought into contact
with the wooden box, or not – so again it’s with and without contact. The results can be seen
in the right-hand column of Fig. 7.86, and this time we can isolate a small effect: at two
places in the pickup-spectrum, small differences show up – they may just be noticeable. In
theory, that is – practically, again no difference was audible in the listening tests. In contrast,
the microphone signal did sound different. It seems that the two small changes in the pickup
signal do not weigh in sufficiently compared to the many 1/3rd-octave levels that remain (on
average) the same. The acoustician is not surprised about any slightly more pronounced
differences in the pickup spectrum that appear as the neck is brought into direct contact with
the wooden box: the neck could operate as transformer (lever) and improve the matching to
the wooden box. To verify this hypothesis, supplementary vibration measurements would be
required; but these were dispensed with because the effects are too small.
And so we arrive at our first conclusion: those who enthusiastically record the airborne sound
of a solid-body guitar, those who may even consider such activity as the main purpose of such
guitars – they certainly will be wise to pay attention to the material and build of the guitar
body. We other guitarists who plug the guitar into an amp that we then turn up, we should
only think about the wood when it comes to ergonomics or cosmetics; the luthier will not
have used insulation board for the instrument i.e. the material will be ok. Back in the day, Leo
Fender cut up wood that was affordable and in reach to him – neither ash nor alder are
classical tone woods, and they don’t have to be, either. For the acoustic guitar, things are very
different: who would ever bolt a steel-sheet casing of the dimensions of an external 2.5” hard
drive to his or her pre-war Martin? Fender does that for their VG-Stratocaster, and nobody is
bothered; the guitar fully meets all expectations [G&B 7/07]. A hole for the battery
compartment is also milled into the body – great guitar, still! And why not, the thing will
work as long as neck and bridge can be solidly fixed to the body. With the Martin (to stay
with the example), that situation would be very different, but that guitar is something else
entirely, compared to Fender’s VG.

Would all the energy fed to the guitar via the plectrum be (acoustically) converted into sound,
an SPL of more than 90 dB at the ear of the guitarist would result. In reality, we find only 60
dB, corresponding to an efficiency of a mere 0,1%! In fact, the energy in the airborne sound is
a part of the original oscillation energy of the strings, and thus any changes in the airborne
sound do indicate changes in the string damping – but this effect is most insignificant. If we
assume the hearing threshold to be at an SPL of 10 dB (as a simplification), and a decay time
of the tone of 6 s, the guitar sound having 40 dB SPL initially would become inaudible after 6
s. A cavity (resonator) coupled to the guitar improves the sound radiation to 50 dB SPL, and
we hear the guitar sound for 8 s – the decay time is the same. If we amplify the pickup
voltage, and have a loudspeaker generate an SPL of 70 dB, we hear the string ringing for 12 s.
There are several simplifications in this scenario, but the basic principle remains clear: the
time we hear the strings of the unamplified guitar ring says nothing about the mechanical
damping of the strings, nor does it indicate anything about the sustain of the guitar played
through an amp.
Drilling holes into the guitar body that later are again covered up by a panel (the pickguard)
will have a significant impact on the vibration behavior of the body. In a Stratocaster (Fig.
7.86), the pickguard covers almost half of the upper side of the guitar body and will
necessarily influence the sound radiation. Of course, apart from the size, also the local
distribution of the vibration velocity is important for the sound generation of such a panel, and
here we run into difficulties. There is no doubt that the pickguard vibrates, but how does it
vibrate? Over the years, the thickness and the material have changed; for some periods there
was an aluminum sheet between wood and pickguard; then again the sheet was replaced by a
mere foil, and even the number and placement of screws fastening the pickguard varied. In
later years, the cutout in the wood was enlarged so that humbuckers could be accommodated,
up to the point where a huge section of wood between neck and bridge was removed. All this
will supposedly have no impact on the sound, as long as ash (or alder) is used for the wood of
the body? Only the wood determines the sound? An allegation that is highly questionable.
Fig. 7.87: Transmission from excitation force to SPL (bold line: w/pickguard, thin line: w/out pickguard).
Right level differences; positive values = pickguard amplifies, negative values = pickguard attenuates.
A simple experiment demonstrates the effect of the pickguard: via a shaker (B&K 4810)
located close to the bridge, a Stratocaster was subjected to vibration in the anechoic chamber,
and a measuring microphone (B&K 4890) recorded the radiated airborne sound in front of the
guitar body. Fig. 7.87 shows the transmission from excitation force to airborne sound: with
pickguard, and without. Clearly, the pickguard changes the radiation by more than ±10 dB –
in fact that’s just the behavior that may be expected given a bolted-on cover. As we change
the torque used to fasten the mounting screws, the radiation changes, as well; the same
happens if we sand down the surface the cover rests on. The voltage of the pickups is not at
all affected by any of this in any way. It may be relevant for the “dry test” (i.e. the purely
acoustic sound of the solid-body electric guitar). It’s just that the “dry test” itself is entirely
irrelevant (for the electric sound of the solid-body electric guitar).

7.8.2 "Stratone"
A citation from the German magazine “Gitarre & Bass”, issue 05/2007: For the sound,
density, elasticity and hardness are decisive. Compared to ash, alder features a 10 – 20 %
smaller density, i.e. a smaller weight relative to the volume – resulting in a faster response.
The oscillation excited by the guitar string needs to push less mass. Furthermore, the smaller
mass of alder results in a higher Eigen-resonance of the body. Resonances in particular
absorb much oscillation energy, and so these frequencies disappear first. Ash, however,
sounds brighter, and richer in harmonics.
The faster response of alder mentioned above is further supported by the material’s higher
elasticity. While the density describes the amount of mass per cm3, the modulus of elasticity
describes the maximum pressure that wood can mount against an external force without any
permanent deformation.
Alder’s higher elasticity (relative to ash) has the effect of a cushioning of the vibration and
thus of extraction of energy from the string, resulting in a shorter response time (i.e. harder
attack), but also in a shorter sustain.
From the point-of-view of sound, the hardness of wood is considered in conjunction with the
density. Harder woods, especially those with a high density, react substantially more
sluggishly to vibrations than softer woods, and extract less energy from the string. The string
shows a slower transient response but holds the vibration for a longer time. In the present
case, ash is categorized as medium-hard to hard (i.e. with longer sustain), and alder as soft.
Compared to a Stratocaster made of alder, ash sounds harmonically richer and has longer
sustain. On the flip-side, the response of alder is more direct, the guitar reacts more
dynamically.
So much for the first part of the article, titled Body. Stating: Alder has a smaller density
relative to ash – is that already incorrect? No, that’s o.k. as such, especially in a magazine
article with the required reasonable length. Of course, there is not “the” alder nor “the” ash –
we find black, white, green and red alder, and black, white and green ash (the latter also
termed red or swamp ash), and of course the climatic conditions under which the trees grow
will vary resulting in different physical parameters. So, only a radical simplification is the
way out, if you do not want to “go down for the third time” already in the first paragraph,
with special literature describing not just three but “65 different types of ash”. Given this
simplification, alder has a 10 – 20% smaller density: 0.55 g/cm3 versus 0.69 g/cm3. We most
humbly add that according to datasheets there is also heavy alder at 0.86 g/cm3, and light ash
at 0.41 g/cm3. In dried condition, that is, because humidity will also influence the density.
However, whether a smaller density will result in a faster response is unfounded speculation.
The oscillation excited by the guitar string needs to push less mass. O.k., so what? A mass
alone will not define any attack-time. And most of all: it is the vibration of the string that the
pickup of the Stratocaster investigated here samples, and not the vibration of the body.
Furthermore, the smaller mass of alder results in a higher Eigen-resonance of the body. Huh?
How does that work? A resonance frequency depends on both mass and stiffness, and nothing
has been said about the latter so far. Body resonances exist – no contest there. But they exist
already from a few hundred Hertz, and their impact on the string vibration remains purely
speculative in the article to begin with.

The faster response of alder mentioned above is further supported by the material’s higher
elasticity. While the density describes the amount of mass per cm3, the modulus of elasticity
describes the maximum pressure that wood can mount against an external force without any
permanent deformation. We are now leaving the area of journalistic freedom and encounter
the first grave error that unfortunately exposes the author as rather ignorant and apparently
missing crucial knowledge. To confuse elasticity and mechanical strength – that must not
happen. The E-modulus – abbreviated from “modulus of elasticity” (or Young’s modulus) – is
a characteristic variable in the area of elasto-mechanics. Elasticity is the characteristic of a
solid body to resist a deformation caused by external forces given linear, reversible behavior.
Now, the modulus of elasticity is NOT the limit of the linear behavior but specifies the
behavior far below the limit of linearity (i.e. at small loads). Indeed, while this modulus of
elasticity does have, with N/mm2, the same unit as the pressure, it does in no way specify a
maximum allowable pressure. Dear young friend who writes about limit values in such an
easygoing way, have you at all considered that this would-be limit-pressure for ash would be
approximately reached if the mass of a car (1300 kg) pushes down on 1 mm2!? Or if 9 combat
tanks weigh in on the surface of one Euro (translator’s remark: that’s about as many tanks
loaded onto the surface of one quarter US-$)!? No way, not even high-grade steel could
withstand that. The E-modulus is a kind of specific stiffness: for alder about 9000 N/mm2, for
ash about 13000 N/mm2. Alder’s higher elasticity (relative to ash) has the effect of a
cushioning of the vibration and thus extraction of energy from the string, resulting in a
shorter response time (i.e. harder attack), but also in a shorter sustain. Indeed, alder features
a smaller E-modulus, i.e. a smaller stiffness and thus a higher flexibility (which we could call
elasticity) – but besides that, things already go awry again: extraction of energy, i.e.
dissipation, will happen only in resistive elements (friction resistances) and not in springs.
With the E-modulus, a parameter for spring stiffness was chosen, and not one for losses. How
fast the vibration energy of the string is converted into heat depends on several parameters
(see Chapter 7.7), the E-modulus alone does not help us here.
Since we have at our disposal now a specific stiffness (= E-modulus) on top of the volume-
specific mass (= density), let’s have another look at the resonance frequency. Assuming a
piece each of ash and alder with the same dimensions, the mass of the piece of ash will be
larger than that of the piece of alder – at least given the simplifications discussed above.
However, not only is the mass of the ash larger but the stiffness of the material is also higher,
and since the resonance frequency is dependent on the quotient of stiffness over mass, the
resonance remains the same in a first-order approximation. No further speculations are
allowable because both density and E-modulus vary – the piece of alder will therefore not
universally have the higher resonance frequency.
From the point-of-view of sound, the hardness of wood is considered in conjunction with the
density. Harder woods, especially those with a high density, react substantially more
sluggishly to vibrations than softer woods, and extract less energy from the string. The string
shows a slower transient response but holds the vibration for a longer time. In the present
case, ash is categorized as medium-hard to hard (i.e. with longer sustain), and alder as soft.
Ash is harder that alder, that much is correct. Any connection to transient processes (that is
the term systems-theory has for “attack” and “decay”) is totally speculative und unfounded. A
transient process cannot be explained by a single material parameter. Which system should be
in transient, anyway: the string or the body? If the body reacts sluggishly, the string should be
able to respond quickly, shouldn’t it?

Compared to a Stratocaster made of alder, ash sounds harmonically richer and has longer
sustain. On the flip-side, the response of alder is more direct, the guitar reacts more
dynamically. At last, here we have a statement that does not ride on any pseudo-scientific
reasoning, and, as a subjective opinion, it does not make itself very vulnerable. It the author
perceives it that way, he certainly may write it down. It is, however, clear from what follows
that the unamplified sound radiated from the solid body is meant – but that sound is so
unimportant that it does not take long to deliberate whether the descriptions are correct.
So what remains as a first appraisal before we turn to the neck of the guitar? There are some
reasonably correct statements regarding density, stiffness, and hardness, we read some
unfounded or even incorrectly reasoned assumptions regarding resonances and transient
processes, and we find speculations about the holiest of all cows – the sustain – without a
single word about material-specific damping parameters. But let’s see how things progress:
From 1959, a rosewood fretboard was used because of its higher durability. Taken by itself,
the higher density and hardness of the rosewood would point to a more pronounced content of
harmonics. However, the overall construction with the glued-on fretboard results in an
additional disruption of the sound propagation within the wood – this makes for a softer and
slower string attack in the rosewood-fitted neck compared to a solid maple neck. Are you sure
about that? Is it the string that is excited by the guitar body? It is almost as if the guitar player
hits the guitar, the body of which needs to start vibrating in order to then make the strings
vibrate. Just to be clear: the guitarist deflects the string with the pick, or fingernail, etc., and as
soon as he lets the string go, it commences to vibrate. The latter happens very, very quickly,
and completely independently of the guitar body during the first few milliseconds. That a
string will start to vibrate more slowly and mellow – that is nonsense. From 1959 – 1962, the
interface between the maple neck and the rosewood fretboard was flat (slab-board). The
sound becomes particularly meaty and fat, and gives an enormous depth to the characteristic
mids. In the book by Day/Rebellius, that reads rather differently: the "slab-board" is one of
the secrets of the renowned, old, crystal-clear vintage sound. And then we find other verdicts,
as well: the direct A/B-comparison between a poplar-Strat with one-piece maple neck and an
ash-Strat with maple/rosewood neck indeed reveals only minute differences (Gitarre & Bass,
Fender special edition). Even more radical is the statement by Lemme: a one-piece maple
neck and a neck with (extra) fretboard sound identical [Lemme, 2003].
And on we go to the “playing-in”: scientifically, playing a stringed instrument for a long time
implies, first of all, that the instrument is subject to vibrations for a longer time. Hard to
believe: that is actually totally correct! But then: the effects are almost impossible to capture
analytically because a piece of wood excited by the corresponding vibrations would have to
be compared to an identical piece of wood that has been merely stored and not played. Or as
alternative: we would have to construct a setup that allows for a reproducible picking of the
string both before and after the “playing-in”. That would not be impossible – but it is not
entirely trivial, either. Let us remember, though, that the energy transferred from the player to
the string is very small (typically a few mWs per struck string). If we would take the above
conjectures about non-reversible deformations seriously (granted – that sounds a bit polemic
here), then permanent deformations would occur only at more than 13000 N/mm2. So: no
worries, mate, in reality only about 0.1 N/mm2 weigh down on the wood, and that is even less
than the compressive strength specified in datasheets (around 50 N/mm2). We do not want to
dispute generally that a guitarist may perceive outrageous improvements in the sound after a
period of “playing-in”, but the reasons for that can be highly diverse.

So much for the first part of this “specialist article” about the Strat. Given what we have
established so far, it ends with an outright threat: in the following issues I will report on the
development of the mechanical components and the electronics of the Stratocaster. This then
reads as follows: until the beginning of the 1970’s, bent steel was used for the bridge saddles.
The elaborate manufacturing process resulted in particularly dense material. Afterwards, the
bridge saddles were first made of brass, then of coated, cast zinc. Relevant to the sound is the
density of the materials – it dropped with each successive version of the bridge saddles. The
densest material then is the steel. Zinc is even lighter than brass. According to generally valid
material science, a less dense material absorbs less energy than a denser one. Excuse me?!
The density of brass is, according to generally valid books on material science, 8.1 – 8.6,
while that of steel is 7.7 – 8.0, and that of cast zinc is about 6.7, each with the unit kg/dm3.
How much energy a material absorbs (i.e. converts into heat) depends not primarily on its
density but on its internal damping parameters. The latter are, however, nowhere specified in
the text – rather there is speculation about frequency dependencies: less density and mass
result in fewer harmonics. In other words: higher density supposedly will give more
harmonics. A few lines on, however, we read: a Strat with bridge saddles made from steel
that sounds too twangy and sharp can sound milder and more balanced with bridge saddles
made of brass. How can that be? Brass is, in this group of materials, the one with the highest
density! It is hard to avoid the impression that the term “density” has been misunderstood.
What happens if you compress a material? It becomes denser! And what were the bridge
saddles of old Strats made from (according to Duchossoir)? From “pressed steel”! Well then
… pressed steel, that’s compressed i.e. mightily dense, isn’t it? No Sir, it ain’t – you failed to
understand what the term actually means. Pressed steel means: the part is made of punched-
out steel bent into shape. That is what the bridge saddles of old Strats were made of, and how
they were made – in sharp contrast to the block-shaped pieces introduced later that – simply
due to larger volume – featured more mass. The latter aspect is, however, totally ignored, just
like the unavoidable friction occurring in the gaps between the parts of the bridge assembly.
It does get still worse, though: Compared to a block of cast steel as it has been used since the
1970’s, the earlier, cut-out block contains less oxygen and therefore has more mass. Oxygen
within a block of steel: now that’s not something the metallurgist likes – at all. From way
back, our memory switches on a red warning lamp when oxygen and iron show up in
combination: RUST! The generally valid material science comments: the oxygen bonded to
the iron atoms is present as FeO-slag after solidification of the molten mass, and can partially
be released to other metals (e.g. Al) as desoxidisation happens. In any case, the share of
oxygen remaining in steel is so small that it cannot have any substantial effects on the density.
That’s what material science says. The Stratone-author, however, says: predominantly, the
cut steel-block makes itself felt via additional harmonics and stronger attack. Rather on the
side, we are informed that the cast-iron block is thinner by 2 mm compared to the cut block.
That could also have an effect on the mass, couldn’t it? Nobody denies that the tremolo block
can influence the sound, but ludicrous conjecture (the behavior is similar for metal and wood)
does not help to get to the bottom of that. Quite amusing: another G&B-expert states in G&B
7/2005 that titanium would be the best material for the trem-block. Titanium, however, has –
at 4.5 kg/dm3 – an even smaller density than cast zinc, and therefore there should cause a
treble loss, according to the first author. Far from it, though: due to the titanium block, the
sound is richer in harmonics. Despite the fact that the titanium block – precisely weighed – is
roughly 120 g lighter than the original. Isn’t that strange? What does hold here: less mass =
additional harmonics (G&B 7/2005), or less mass = less harmonics (G&B 6/2007)? In any
case, we get: less mass = more money, because titanium was never cheap – 330 Euro, to be
precise. That’s just for the trem-block, not for the guitar, and including stainless-steel screws
… for titanium screws would have cost another 40 Euro extra.

Before we call in frequency spectra to corroborate this G&B-mess of loosely collected

conjecture, let’s digress a little into vibration engineering. We read: the nut is supposed to
transfer the vibration energy as completely as possible into the neck. Sure: the neck should
vibrate tremendously, and the string should transfer its vibration energy as completely as
possible to the neck, and consequently stop vibrating … This would follow from the well-
know physics-law of conservation of energy. Because when the string has transferred all its
vibration energy to the neck, it has no vibration energy anymore itself. Too bad, we would
have gladly granted it that extra sustain, the holiest of cows. But more about that later.
Now, the Stratone-author does not limit himself to conjecture about the scientific reasons for
differences in sound, but he procures 7 different Strats, and analyses their sound: for the
recording of the unamplified sound of the guitars, a Rhøde NT2 condenser microphone was
positioned at 10 cm distance pointing to a spot between neck pickup and heel of the neck.
Then analysis was done using short-term DFT. The spectra depicted in the magazine are
without scaling on the ordinate and can therefore not purposefully be evaluated. However, the
sound files were also available at www.gitarrebass.de, and with these a scaled analysis could
be carried out. We will not right now go into whether it is meaningful at all to analyze the
purely acoustic sound of these solid body guitars; let’s just look how the measurements and
the G&B-statements line up.
Subject to analysis are alder-Strats built in 1959, 1962, 1972, and 1974, as well as ash-Strats
built in 1972 and 2005. The 1995 alder-Strat was not evaluated – its file differed too much
from the others. In the following analyses, ash-Strats are designated with an S and alder-
Strats with an L. Under scrutiny is the G&B-statement: compared to the alder-Stratocaster,
the ash-Stratocaster sounds richer in its harmonics and has a longer sustain. In Fig. 7.88, the
analyses of the first 4.5 s of sound are shown. And here we already run into the first problems:
the sounds result from an E-major chord played across all 6 strings – but the author was not
aware that he should strike all 6 strings as similarly as possible for all test sounds. And so the
plectrum gets caught a bit in this string or that string, or it audibly strikes the pickguard. Well,
we have to live with these inadequacies – no other recordings have been published. Let us
regard the first statement: compared to the alder-Strat, the ash-Strat sounds richer in its
harmonics. The 1/3rd-octave spectra averages over the first 4.5 s do not confirm this
assumption: it’s the 1972-alder-Strat that featured the strongest treble. In the summation level,
the assumption regarding the sustain cannot be confirmed, either: an alder-Start is ahead only
between 0.5 and 1.2 s, from then on there is no difference remaining between 2 ash- and 2
alder-Strats. Since for all guitars the higher-frequency partials decay more quickly than the
lower-frequency partials, a faster decay of the overall level is to be expected for more trebly
sounds (slightly simplifying things): the more the higher partials define the overall sound, the
faster the latter decays.
Of course, one may object to these analyses that neither the summation level nor the
averaging over 4.5 s is very meaningful. Narrow-band level measurements, encounter other
problems, however: the levels of individual partials decay only in exceptional cases according
to a simple exponential function, and beats often occur due to circular wave polarization and
due to bearing impedances dependent on the direction of the oscillation (Chapter 1.6).
Moreover, there are interactions between the partials of individual strings that can lead to
pronounced beating. Fig. 7.89 shows the 1/3rd-octave levels for the individual guitars. The
1/3rd-octave level at 80 Hz approximately captures the E2-fundamental that may decay both
for the alder- and the ash-Strats with or without strong beating – no confirmation is found in
these measurements that ash-Strats would have a longer sustain.

The comparison at 125 Hz is different in that with one single exception, all 1/3rd-octave levels
decay with practically the same speed, but again there is no longer sustain apparent for ash.
Yet different again is the situation for the 160-Hz-level: all 6 measurement-curves differ
significantly – as it is the case at 500 Hz, as well. Given such strong level-fluctuations, a
general statement in the sense of ash-Strats have a longer sustain has no foundation.
Fig. 7.88: 1/3rd-octave spectra (left) und overall level (right); sound-files acc. to G&B 5/2007 p.212, normalized.
Fig. 7.89: Decay of individual 1/3rd-octave levels. Since in these curves only the decay (or the slope) is of
interest, they were vertically shifted for best possible evaluation and comparison.
Fig. 7.90: Decay of individual 1/3rd-octave levels, compare to Fig. 7.89.

It does not make any sense to measure the beating-parameters, because with a minimal
detuning of one or more strings, the beating will change. Such beats also show up in the
waterfall-spectra published in the G&B-article albeit there is no scaling. As we move towards
higher frequencies (Fig. 7.90), the beating decreases in strength (many partials per 1/3rd-
octave), but again the statement regarding the difference in sustain cannot be confirmed. At
1.25 kHz, the ash-Strats correspond to two alder-Strats, and at 2.8 kHz already the two ash-
Strats (red/magenta) differ significantly from each other. Conclusion: the level measurements
cannot support any significant difference in sustain between ash and alder. Even before we
advance to the core question of whether parameters in the airborne sound have any relevance
for the pickup signal, we have to recognize that already the statements about the parameters of
the airborne sound fail to bear objective scrutiny. Therefore, dear Strat-analysts: don’t forget
to always put a scale to the ordinate – then you will see this result yourselves, too.
While it is highly commendable that statements regarding the wood are deduced from results
of experiments, the framework for these experiments still needs to fit, and the investigated
guitars must exclusively differ in the wood. In the present experiments, they do not – as the
G&B-author attests. Most important for the decay process of the string vibration is the
distance of the string to the frets, and the initial displacement of the string. No information at
all is given about the condition of the frets, and we can only surmise that the guitars were not
refretted before the experiments. Not even regarding the action (distance from string to frets
or to fretboard) there is any information, and the author is silent about the age of the strings,
as well. So what’s the point here? If we carry out such experiments, it is mandatory to restring
all guitars with the same kind of strings, and the action needs to be adjusted to be as similar as
possible. The strings must be reproducibly picked with a suitable device, and even the support
for the guitar is significant: already lightly gripping the guitar neck with thumb and 1st finger
(without even touching the strings) changes the decay behavior quite substantially (compare
to Chapter 7.7). However: even with perfect conditions, what actually is the connection of the
airborne sound recorded at a distance of 10 cm from the guitar to the voltage generated at the
output jack? That is the central question here … but the answer shall be put on the backburner
because there is a lot of text regarding the guitar electrics still to be looked at. The signal runs
through a capacitor that presently has a value of 0.022 µF. In combination with the coil it
forms a band-pass. Close, but topologically this is a low-pass (series-L and parallel-C). Our
comparisons show that the more massive build (of the capacitor) promotes a more musical
effect due to fewer frequency cancellations in the pass-band. The sonic image of the larger
capacitor seems fuller and denser. What rubbish – here a blind man judges colors. The pass-
band is in fact characterized by passing signals, not cancelling them. Also, what is actually the
pass-band of this band-pass? What does a more massive build indicate? Smaller, i.e. less
(geometric) volume? Ceramics rather than foil? Or more weight? Very puzzling, this …
Regarding the pickup: sonically relevant is not only the main resonance that can be
calculated mathematically, but also the countless ancillary resonances and cancellations. The
winding is mainly responsible for this … back in the day it was customary to guide the wire in
such a way that overlaps would result i.e. that not all turns were exactly in parallel. This
method is called biphilar winding. Okay … phew … after we’ve all managed to compose
ourselves again, and have not incurred any permanent damage by this blow, let’s get this
straight: a bifilar winding is set up if induction is not desired – in short. The term is bifilar
(not biphilar) because two threads (Latin: filum = thread) are wound. Using today’s
terminology we would say: two parallel wires with connected beginning are wound. The two
ends then make for the connection poles. A coil with two opposed windings will result i.e.
one without inductance (idealized).

Had the pickup a bifilar winding, no voltage could be induced. What the author means to
address is “wild winding” or “cross-winding”, in contrast to “winding in layers”. Differences
exist between how old and new pickups were wound – no contest there – but there was never
ever a bifilar-ly wound pickup. Still, a clear sonic tendency of the biphilar winding can be
recognized: our investigations show that the magnetic field takes on more homogenous
characteristics compared to a machine-wound coil. Certain level-values of the resonances are
simply not exceeded. Level-values of resonances? Does the man mean the Q-factor of the
resonance? Why doesn’t he then just use that term? And what are those resonances that
allegedly occur in countless numbers? They may not be countable as we progress towards
infinity – with a pickup, though, 10 kHz is the utmost limit. Even if we think 20 kHz is
required: there are not countless resonances. It is indeed not possible to model every pickup as
a 2nd-order-system (i.e. with one single resonance), but with a 4th-order-model we get
extremely close. However, apparently, something else is meant: the result of the machine-
winding is a frequency graph with very narrow and very loud level peaks … Moreover the
coil is more loosely wound by hand, resulting in more resonance frequencies in the treble
range. Here we can’t help but suspect that when regarding spectra he has not really
understood, the author interprets maxima generated by the string-partials as pickup
resonances. Or does he imply that the hand-wound coil has a lower winding-capacitance
resulting in a main resonance of higher frequency? The term “winding capacitance” doesn’t
appear anyhere, though – but we do find: from a higher inductance, an upward-shift of the
main resonance results. Wrong again: the resonance frequency drops with rising inductance.
Why is it actually absolutely necessary for a person to write a “specialist article” in a so-
called “specialist magazine” if that person is not at all, in any way, a specialist in the given
specialist area?
And we get to the wire: given a diameter of 0,0030", the wire was, 46 years ago (i.e. 1961,
thicker by 0,0004" compared to today. Combined with the fact that today 400 more turns are
included, a smaller inductance results, i.e. a lower output voltage of the old pickup. Whether
a diameter with or without insulation is meant remains unclear. Duchossoir opines that from
the 1950’s to the 1990’s, 42-AWG was used always, i.e. 63,5 µm Cu … may the better man
win. The number of turns varies so strongly over the years (according to Duchossoir: from
7600 to 9000) that “400” should be interpreted rather generously. And at last the insulation:
not only the thickness of the coating has an effect on the sound, but the material, as well,
because the material surrounding the copper within the coil has, acting as a dielectric, a
direct influence on the magnetic field. Nope – again close but no cigar: dielectrics act in a
polarizing fashion in the electric field, while in the magnetic field, the permeability is the
quantity with direct influence. The Formvar coating consisting of a resin composite makes for
a more open and lively sound than the chemical Polysol layer. Of course: the chemical stuff
doesn’t sound right! What does the chemical scientist comment regarding Formvar, though?
Formvar lacquers contain polyvinyl-acetal to which phenolic resin is added. And phenolic
resin is counted as a … chemical synthetic.
A person testing a guitar is certainly at liberty to write as a conclusion of his labors: I like the
1962 Strat the best. However, as soon as this subjective evaluation is being substantiated with
misunderstood scientific principles, the dulling of the reader’s mind begins. Mistakes happen,
of course. The admitance, the cahtodyne, the E-modul – even biphilar would not be worth a
single line if it were just a spelling mistake. Specialist magazines with a good reputation have
an editorial office and proof reading where most of the smaller errors are caught and ironed
out. They also have a reviewer who will point out subject-specific deficiencies.

7.8.3 “Flachjournalismus” – Where’s the bottom in that barrel of specialist journalists?
Translator’s note: I choose not to translate this very satirical sub-chapter since much of its effect
relies on idiosyncrasies of the German language and on specifics of the music scene in Germany
(especially the Munich scene into which the author has some serious personal insights) and
Austria. To those with some knowledge of German and looking for a laugh, reading of the chapter
is highly recommended – although they should not expect much scientific gain from it. I might still
try to do a translation in the future, but it will be a real labor of love to match this pun on rock
concerts, music magazines and specialist journalism to American and/or British culture.
Thorbi hatte Ungeheuerliches entdeckt. Eigentlich war er zwar völlig privat unterwegs, aber was heißt schon
privat – ein Starfotograf ist praktisch immer im Dienst. Eigentlich ... eigentlich sollte um 21:00 der
Liedermacher Wolf Amboss im Zirkus Krone eines seiner zahlreichen Comebacks performen, doch nun war's
schon 21:30, und das Volk wurde langsam unruhig. Also schlenderte Thorbi, der diesen Act der Nachwelt ganz
privat in Bild (legal) und Ton (nun ja) erhalten wollte, schlenderte also Richtung Bühne, Backstage. So ganz
dicht kam er zwar nicht ran, konnte sich aber (man kennt sich ...) einen relativ guten Platz direkt bei der Inneren
Security erobern, von dem aus er Amboss fast sehen konnte. Weil's aber eben nur fast war (wozu man auch
sagen hätte können, es war überhaupt nichts zu sehen), scannte sein Gehör die unmittelbare Umgebung, und da
wurde er Zeuge eines journalistischen Komplotts, das er seiner GuitarLicks&Tricks-Redaktion unbedingt
mitteilen musste.
Direkt neben ihm überlegten nämlich zwei arbeitslose Chemie-Abbrecher, wie sie zu Geld kommen könnten.
Das wäre zwar im Grunde so alltäglich, dass kein Mensch hingehört hätte, doch im Gedränge hatte sich Thorbis
Recorder eingeschaltet, ein Segen, wie sich alsbald herausstellen sollte. "... geben wir eine neue Zeitschrift
heraus ... Zange&Tupfer .... so mit Medizinberichten und Tablettentests und so ..." Thorbi rückte näher, um
besser hören zu können. "Nee, der Titel ist Scheiße, besser was mit Anspruch, Health&Care vielleicht?" "Das
können wir ja noch später, wenn wir die ersten Entwürfe fertig haben. Aufs Cover kommt immer ein Foto der
getesteten Tabletten, dazu ein Interview mit einem Chefarzt, eine Kolumne 'Das haut rein', weil der Piepenbrink
ja auch noch keine Stelle hat, dazu viel Pharma-Reklame, und zweimal im Jahr einen Bericht von der EITA.
Geil, oder?" Ehe eine Bewertung erfolgen konnte, öffnete sich eine Backstagetüre, und eine dickliche, schwarz
kostümierte Blonde lief heraus, etwas angewidert einen Lappen von sich weghaltend. Ganz automatisch brachte
Thorbi die Minikamera in Position, und vergaß für einen Augenblick das Tabletten-Komplott, doch – wie
erwähnt – der Recorder recordete sowieso schon. Durch die geöffnete Türe konnte man sehen, dass Amboss
nicht etwa, wie der Hallensprecher mehrfach durchgesagt hatte, im Berufsverkehr festsaß, nein, er war schon da.
Teilweise, zumindest, das Physische jedenfalls lag in voller Größe am Boden. Ein Unfall? Thorbi musste mehr
wissen, und stupfte Moski, den nahe im stehenden Security-Boliden, mit einem fragenden Blick an. "Hat was
Falsch's gessn" war die wenig ergiebige Antwort, doch die begleitende Handbewegung lies keine Zweifel offen.
Später erfuhr Thorbi, dass WA eine zufällig rumstehende Flasche mit einem Mikrofon verwechselt hatte,
probehalber reinsingen wollte, von der dabei rauslaufenden Flüssigkeit so überrascht wurde, dass ihm ein kleines
... äh ... Missgeschick passierte, auf dem er dummerweise ausrutschte und der Länge nach hinfiel. Als
Profimusiker sollte man wirklich mehr drauf achten, was man vor dem Gig isst ...
Im Raum, dessen Türe immer noch offen stand, liefen mehrere wichtige Leute hin und her, und ein Oberwichtl
rief mehrmals "so könn ma den net rauslassn, der Siggl soll sofort kommen". Der kam postwendend, sein
Gesicht hinter einem Bass versteckend, (Fender, Shortscale), und verschwand flugs in der Türe, doch Thorbis
Nachbarin hatte ihn schon erspäht. Ihr "mei is der kloa" war etwas deplaziert, und ihr "is die Nikoll aa do?"
attestierte ihr eine gewisse Ignoranz, die dem Münchner Opernpublikum aber auch nachgesagt wird und
stadttypisch ist. Denn der Eine hatte mit dem Anderen rein gar nichts zu tun, hier erhielt gerade eine im
Stadtwesten bekannte Boygroup die Chance, groß rauszukommen, doch das ist eine andere Story. Von
Backstage war nur mehr "zwoa extra starke Kaffee für'n Barny" zu hören, dann wurde die Tür zugeworfen, und
Thorbi hatte wieder Ohren für seine Nachbarn.

"... hab ich schon Vorarbeiten für einen MAO-Hemmer-Test durchgeführt, etwa in der Art: Auf dem Zettel steht
zwar, mit etwas Flüssigkeit nehmen, aber ich mache zuerst immer einen Trockentes t. Schon dabei fühlt man eine
Art Vibrieren, das den ganzen Körper durchdringt und selbst im Bauch spürbar ist. Nimmt man beide Tabletten
zusammen, entsteht so eine Art glockiges, glasiges Gefühl, man durchlebt alle Höhen auf einmal, während die
Einzeldosis, (und zwar die Tablette, die am Rand der Cartridge sitzt), mehr ein erdiges, die Gefühle
verzerrendes Erlebnis bringt. Unsere Messung ergab, dass diese Tablette etwas schwerer war als ihre Kollegin:
3 gegen 2 Gramm. Dass die Anzeige der Küchenwaage in beiden Fällen zwischen 2 und 3 hin- und hersprang,
müssen wir ja nicht dazuschreiben, oder?" "Und das geht so einfach, ich meine, so ganz ohne großes
Drumrum?" "Natürlich, und am Ende schreiben wir noch eine plus/minus-Bewertung drunter ... und das einmal
pro Monat." "Aber wenn sich nun jemand auf unsere Bewertung verlässt, und das Zeugs kauft, ich meine – wir
sind doch noch keine Profi-Pharmakologen, wenn da ein Fehler drin ist, kann's da keine Schadensersatz-
forderungen geben?" "Tja, Kohle will ich keine rausrücken, dann nehmen wir halt ein anderes Sujet ..." er sah
sich prüfend um, bis sein Blick an einem Musiker hängen blieb, der, leicht schwankend, 'wo's mei Ka...ffee'
lallte, und dabei Halt an einer Gitarre suchte (Fender, Strat, weiß, relic). "Eine alte Stratocaster, das ist eine noch
bessere Idee, wir könnten doch auch ein Fachmagazin für Gitarren herausgeben, oder?" "Aber so richtig
Ahnung davon..." "Das macht nix, die anderen haben doch auch keine Ahnung. Ich schreib meinen Tabletten-
Test über die MAO-Hemmer leicht um, Tonabnehmer statt Tablette, Henry statt Gramm, und du ..." er
unterbrach einen Moment, weil die Türe aufging, und eine Gestalt heraushuschte, hinter einem Fender-
Shortscale Schutz suchend, "du machst die Bass-Testberichte. Vielleicht kriegen wir ja noch ein Interview mit'm
Amboss. Wenn der schon wieder einen Auftritt verkackt, kann er froh sein, wenn sich noch irgendjemand für ihn
interessiert." "Ja, aber besser erst morgen, wenn sich sein Zustand wieder normalisiert hat." "Lieber keine
Vermutungen, was bei dem normal ist. Wie heißt eigentlich die Ersatzband, die jetzt gerade einläuft?"
"Irgendeine regionale Boygroup, irgendein Murphy, der Sänger wahrscheinlich." Und weil aus dem Hintergrund
ein 'Basedow' zu hören war, kam's zu der viel beachteten Headline: AMBOSS WIEDER INDISPONIERT, MURPHY
BASEDOW RETTET DEN ABEND MIT BAYERISCHEM POP. Das mit dem gründlichen Recherchieren werden sie
schon noch lernen, den Umgang mit den Anwälten auch. Letztlich waren sie dann froh, dass (dank Thorbis
dezenter Vermittlung) bei GuitarLicks&Tricks zwei neue Stellen geschaffen wurden, und so begannen zwei neue
journalistische Karrieren ...
Amboss kam übrigens doch noch auf die Bühne, gerade als "Murphy Basedow" seine erste Zugabe spielte, war
aber leider nur zu sehen, und nicht zu hören, weil er in eine zufällig rumliegende Flasche sang, die er für ein
Mikrofon hielt. Umstehende sagten später, derartige Verwechslungen hätten auch ihr Gutes, aber das waren
vermutlich fanatische Murphinisten. Die verstehen ja auch nicht, warum zwei Tage später in ebay ein
Putzlumpen ("nicht neu, mit Gebrauchsspuren, backstage im Beisein des Künstlers versiegelt") für 18 Euro
versteigert wurde. Ja gut, man hatte sich mehr erhofft, aber dank prophetischer Weissagungen war man vor-
gewarnt gewesen: ... es is scho oos und du hast glaabt es fangt erst oon ...

7.9 The Wood Determines the Sound?
Mahogany! Maple! Rosewood! Men oft believe, if only they hear wordy pother, that there
must surely be in it some thought or other [Goethe]. And the usual thinking is: “the electric
guitar is a musical instrument made of wood. In all musical instruments made of wood, the
wood determines the sound. The more noble the wood, the more noble the sound.” Goethe’s
witch’s kitchen – a suitable location for deception and magic – holds more such articles of
faith, but let us keep some distance from alchemy, and give physics the priority here: how
does the body of the guitar vibrate, and in what way will the vibration of this body influence
the sound? In the material-science course, every luthier learns about different tonewoods and
their sound-determining material-parameters: “the denser the wood, the more brilliant, treble-
rich the sound; the higher the stiffness, the longer the sustain (P. Day).” As if that were self-
evident, this statement and similar ones are based on the assumption that the findings that are
valid for violins and acoustic guitars apply to electric guitars, as well. If we now add that
board of experts who listen to an electric guitar first of all without amplification, we quickly
arrive at a conglomerate of teachings that, between them, could not be more contradictory. All
the while two simple principles would really help us:
1) Compared to the acoustic guitar, the electric guitar functions very differently. Findings
derived from the one type of guitar may not be sight-unseen applied to the other type.
2) There is a connection between the vibration of the strings and the (airborne) sound directly
radiated by the electrical guitar. There is also a connection between the vibration of the strings
and the sound radiated by the loudspeaker – but this latter connection is very different from
the former.
The fundamental differences between acoustic and electric guitar become evident when we
look at the energy flow: being plucked, the guitar string receives energy that is in part
converted in to sound energy, and in part into caloric energy (heat). A – not untypical –
excitation energy of E = 3.6 mWs corresponds to the billionth part of one kilowatt-hour
(kWh); that’s really very little compared to household appliances but still enough to generate
a sound that is clearly heard. With an acoustic guitar, this energy can generate an SPL of
about 94 dB at the ear of the player; a Les Paul only reaches about 64 dB. A level difference
of 30 dB translate onto a power relationship of 1000 to 1, which confirms quantitatively what
was qualitatively already known: the electric (solid-body) guitar is a very inefficient sound
source – at least as far as the directly radiated primary sound is concerned. However, the
electric guitar is of course not intended to generate primary sound – it is there to generate
electrical voltage. The big difference between the two modes of operation: in the acoustic
guitar, the sound energy needs to travel “through” the body i.e. “through” the wood, while in
the electric guitar the part of the sound energy that is “reflected from the wood to the string”
is captured. Any conjecture that, in the electric guitar, the vibration energy needs to be also
fed to the guitar body as much as possible, is wrong. ”The biggest part of the string vibration
should be conducted into the body. If the latter is fed with unrestrained vibration energy, a
maximum of tone and sustain develops [G&B 12/05]." How should the string ring for a long
time (i.e. have a lot of sustain), if its vibration energy has gone into the guitar body? The law
of energy conservation dictates that energy cannot appear out of nowhere. The excitation
energy is present only once; the part of it that is fed to the guitar body is missing to keep the
string ringing. The banjo is a good example for an instrument that withdraws a lot of energy
from the string within a short time. However the sound of a banjo (and in particular its
sustain!) is not much like that of an electric guitar.

7.9 The wood determines the sound? 7-103
From a systems-theory point-of-view, the string represents a mechanical waveguide on

which waves propagate. As these waves impinge on the bridge and the nut (or the fret where
the string is fretted), one part of the energy in the wave is reflected, the other part is absorbed
by the bridge/nut/fret (and adjacent structures). Again, the law of conservation of energy
holds: the sum of the reflected and of the absorbed energy corresponds to the energy in the
wave impinging on the bridge/nut. We get a high rate of absorption if the wave impedance
and the impedance of bridge/nut/fret have comparable values. The wave impedance of the
string (see Chapter 2) depends on the diameter and on the material: typical would be 0.2 Ns/m
(E4-string) to 1 Ns/m (E2-string). These are very small values compared to typical bridge
impedances (100 – 1000 Ns/m). The situation is comparable to an airborne wave that hits onto
a concrete wall: because the wave impedances again differ by several orders of magnitude,
almost all of the sound energy is reflected. The same happens with the string: the vibration of
the string is, for the most part, not fed to the guitar body but it is reflected. In the solid-body
guitar, a degree of reflection of 99.9% for low-frequency partials is not untypical: of the
vibration energy arriving from the direction of the nut, 99.9% are reflected and only 0.1% are
absorbed. There is no other way a vibration could remain for any extended periods of time: if
for the E2-string 50% of the energy would be absorbed at each reflection, only 0.1% of the
initial energy would remain after only 10 reflections – and 10 reflections have happened after
a mere 60 ms for the E2-string! Given a 99.9%-reflection, 37% of the initial energy will
remain after 1000 reflections (that’s 6 s)♣. Therefore, a simple connection exists between the
decay time (the sustain) and the degree of absorption: the higher the degree of absorption, the
shorter the sustain. And here we arrive at an explanation that is not so easy to refute: if the
sound depends on the sustain, and the sustain depends on the absorption, and the absorption
depends on the bridge/nut/fret, then the wood of the guitar body will determine how the guitar
sounds, won’t it, after all?!
Given the intense and controversial discussions about the “tonewood”-topic, let us make a bit
of room for some fundamental considerations: if a string is struck once, its vibration energy
decreases over time. The main reasons for this decay are: sound radiation directly from the
string, internal absorption within the string, and absorption at the string bearings. The first
effect is so small that it is normally neglected. The second effect is significant in the middle to
high frequency range for unwound strings; this is elaborated in Chapter 7.7. The third effect is
the only one that can be connected to body-parameters. If we neglect the first two effects, the
string vibration – and thus a component of the sound – indeed is completely determined by
the guitar body. That is defining the “body” very extensively, though: it would have to
include everything that abuts to the string, in particular the bridge that for example consists of
18 individual components in the case of the Gibson ABR-1 bridge. There is much wailing all
over the place that the super-rare tonewoods of the early Les Pauls are not available anymore,
and thus the sound of these originals will never be duplicated. Interestingly though, the
question rarely asked is to which extent the individual pieces of the ABR-1 bridge were
deburred, and how clean the force fit between the movable bridge saddles and the base is. The
bearing impedances at the bridge and at the nut (or respective fret on the neck) strongly
influence the decay of the individual partials of the sound. Before the vibration energy arrives
in the body, it needs to traverse the bridge/nut/fret; the stronger these elements reflect the
vibration, the less important the material of the guitar body is.
All this is, however, valid for the acoustic guitar, as well – so what is basically different in its
sound generation compared to the electric guitar?
♣
We have neglected other mechanisms of absorption in this example.
© M. Zollner & T. Zwicker 2019 Translation into English by Tilmann Zwicker

In the acoustic guitar, the sound to be radiated needs to first get from the bridge, via the
body, to the radiating surface; therefore the build of the body has an effect on the sound from
the very first millisecond. The top of the acoustic guitar, its bracing, its shape, the location of
the bridge – all this influences the radiated airborne sound from the first moment on. The
retroactive effect of these details onto the degree of reflection is, however, rather small. In
fact, it needs to be small so that a vibration can happen in the first place. It is exactly at this
point where the experienced master-builder is required: the optimization of that wooden
transmission-filter requires much specialist knowledge and – no contest – special materials.
While the guitar body shapes radiated sound from the very first moment, what happens to the
string is quite different: its vibration is at first not much influenced by the body – only with
time, the absorption at the bearings takes an effect. That is why two electric guitars fitted with
the same magnetic pickup and the same strings, and plucked in an identical manner, will
sound very similarly at the first moment. That’s irrespective of what wood they are made of♣.
The may differ in their acoustical sound because the mechano-acoustical filter may differ
drastically depending on the circumstances, but the retroactive effect of this filter onto the
string vibration is rather small in typical electric guitars. It is not conducive to cite that
famous rubber-guitar that supposedly had a terrible sound (if it existed at all in reality):
presumably its bridge impedance was not several orders of magnitude above the wave-
impedance, presumably its degree of absorption was bigger that 0.1% … presumably that
guitar made from rubber is pure fiction.
The fundamental differences between the “electric” and the “acoustic” sound in electric
guitars may be explained by an example: two electric guitars reflect the wave energy in the
same way at 300 Hz, while at 600 Hz, one of the two (Git1) reflects 99.9%, and the other
(Git2) reflects 99.6%. Idealized, the energy lost by the string is completely radiated as
airborne sound. Given identical string excitation, these two guitars will radiate the same
sound energy at 300 Hz, while at 600 Hz, the radiated sound energy will differ by a factor of
4. Git2 radiates the latter range more loudly; a four-fold higher energy at 600 Hz corresponds
to a level difference of 6 dB. Apart from the differences in the radiated airborne sound, the
differing absorption will also result in a difference of how quickly the string vibration decays:
Git1 still features 95% of the original vibration energy after 50 reflections, while in Git2, only
82% remain. Expressed in levels, the 600-Hz-level drops by 0.22 dB during the first 50
reflections in Git1, and by 0.87 dB in Git2. The airborne sound between the two guitars
therefore differs by 6 dB from the first instant, while the electrical sound is identical at first
and changes by 0.6 dB by the 50th reflection. If we now drop the idealizing assumption that all
absorbed energy is converted to airborne sound, and if we allow dissipation (the absorbed
energy is partially converted into heat), then larger as well as smaller level differences could
be generated in the airborne sound. To carry things to extremes: both guitars are picked in
identical manner by a small actuator, but one of the guitars is located in a guitar case (the lid
of which does not touch the strings). How would now the electrical sound differ? And how
the acoustic, airborne sound?
The conclusion of these considerations can therefore only be: the geometry and the material
of the guitar body do shape the radiated airborne sound from the first moment on – but
regarding the attack of the “electrical sound” that is highly important for the perception of the
sound, there is only minor influence. The airborne sound radiated by an electric guitar does
correlate with the pickup voltage, but in a highly individual manner.
♣
Provided that the string can vibrate freely and does not hit the frets.

The following example will show how much the spectra of the airborne sound can change
while the pickup signal remains identical: for a Squier Super-Sonic (similar to a Strat), the
pickup signal and two microphone signals were recorded at the same time (Fig. 7.91). One
microphone was at the position where the ear of the guitarist is usually located, the other
microphone recorded the airborne sound in front of the guitar at 50 cm distance. The two
airborne sounds differ significantly because the guitar operates as a dipole in several
frequency ranges, and destructive interferences (cancellations) happen in the plane of the
guitar body. These differences in (airborne) sound become also audible if the guitar is rotated
slightly around its longitudinal axis while playing: the sound immediately changes. That the
electrical sound does not change should be clear even to the most ardent skeptic. And a last
example: the airborne sound of the Squier changes as well when its body is set onto a
tabletop, because the radiating surface is enlarged. It would also be possible to say that the
body is enlarged. This change does, however, not have any audible effect on the electrical
sound.
Fig. 7.91: 1/3rd-octave spectra of a Squier Super-Sonic: airborne sound (––––), pickup voltage (----).
Microphone located in the plane of the guitar body (left), microphone located in front of the guitar (center),
level difference of the spectra of the airborne sound (right).
If the wood of the guitar body had a significant influence on the “electrical sound” of an
electric guitar, we would find a clear mapping between type of wood and sound attributes in
the corresponding specialist literature. Such mappings do exists but they show an astonishing
variation from source to source. For example, the sound attributes for alder read: sweet;
mellow; warm; many harmonics; restrained share of treble; gentle; fat bass; rather subtle
share of bass; strong mids; well-rounded share of mids; much sustain; accentuated; squishy;
good presence; undifferentiated; balanced; full sound; thinner in its sound compared to
basswood. How can a type of wood generate both an accentuated and a squishy sound? How
can it support a fat bass with a rather subtle share of bass? Sure, the above terms have not
originated with the same author, that’s a cross-section through many specialist articles. There
are several explanations for these clearly contradictory evaluations: it is not elaborated
whether the electrical or the acoustical sound is referred to, because (allegedly) everybody
knows that there is no big difference between the two: the electrified plank-guitar primarily is
an acoustic instrument. The wood makes for the character of the sound; the pickups only have
a small share. And thus a humbucker cannot exorcise the characteristic sound- and attack-
evolution from the Strat with an alder- or ash-body (G&B, 2/2000). The experts may borrow
approaches found in violin-making, because: what holds for the violin cannot be wrong for
the guitar, can it? Of course, the number of strings does differ slightly, and size and weight
are admittedly not the same, either. And, well … Stradivari did not actually build electrified
plank-violins, and there are no frets on a violin, either. But: both are made from wood! Still:
the apparently valid formula that old wood always is suitable wood is only correct in part. We
need to look more closely – which leads us directly to the Italian or alpine violin builders. …
Only so-called tonewood gives us, after processing, in the end those clean, vocal tones, a
dynamic and prompt response, and this hauntingly beautiful scope, or power of self-assertion.

Scope! In an electric guitar! Indeed, the pictures to which the cited text relates show Les Pauls
and Stratocasters. The dogma of the sound of tonewood is deep-seated – so deep-seated, that
many an author will do a true backwards somersault, and vote for and against at the same
time: every piece of wood has its own sound, we read in a book about electric guitars. A few
pages on, the author opines (in the same book): the sound of the electric guitar depends
largely on the pickup, but announces in the next edition: the body has – in the solid guitar, as
well – a decisive influence on the sound. Six pages on, we read in the same book: the
different sound of electric guitars is, to a large extent, due to the pickups. It gets even more
extreme in a different book: solid-body guitars may be manufactured in nearly all sizes and
shapes – significant effects on the sound should not be expected. The same author states 65
pages on: the sound characteristic of the electrical guitar is significantly determined by the
selection of the wood. Pickups and amplifier support the sound of the guitar but rarely
influence or characterize it fundamentally. In test reports, the contributors seem to be caught
in this corset, too. On the one hand, we find: Of course, the wood of the guitar body decisively
characterizes the Fender-sound. Ash sounds brighter and with more harmonics compared to
alder, and it features longer sustain. On the other hand, referring to ash-Strat vs. alder-Strat:
there are only minute differences in sound. Alder-Strat vs. poplar-Strat: they differ only in the
finest degree. Mahogany-Squier vs. basswood-Squier: almost identical sound. (citation from
reviews published in Gitarre & Bass).
How far the wood of the guitar body in fact determines the (electric) sound of the electric
guitar shall be investigated first given the boundary condition that the string can decay freely
(i.e. it does not hit the frets). Corresponding measurements were taken with a Les Paul ’59
(Historic Collection) that was fitted with a solid 26-mil-string as a D-string (fundamental
frequency = 200 Hz). This string was excited, next to the nut, with a short impulse; the
fretboard-normal velocity was measured next to the bridge saddle using a laser-vibrometer.
Fig. 7.92 shows the spectra of the first 21 ms of these velocity signals. The short length of the
analysis-interval results in a relatively broad leakage (i.e. a broadening of the spectral lines).
The excitation impulse approximately corresponds to half a sine wave; the spectral envelope
can be described as superposition of two si-functions. The lower line in the figure depicts a
calculation according to the correspondingly simplified model. There is relatively good
correspondence; the measurement results deviate only at a few places – and the following
elaborations focus on these discrepancies.
Fig. 7.92: Spectra of the first 21 ms after plucking the string. Two different excitations (left/right).
The lower row shows the spectra calculated according to a simplified model. “Messung” = Measurement

All measurements confirm the hypothesis that the attack-spectrum is predominantly

determined by the string excitation. The difference between the spectra shown left and right in
Fig. 7.92 mainly consists in that the place of excitation was shifted by a few millimeters; as a
result the impulse-length and –envelope were changed. There are two places in the spectrum
(3.7 kHz, 7.4 kHz) where the measurements deviate from the model envelope in a two-fold
fashion: both the frequency of the partials and the level of the partials are not as calculated,
and in addition a partial at 3.8 kHz becomes visible that does not fit into the frequency grid.
All these deviations are clearly affiliated with the bearing of the string – but not necessarily
with the wood of the guitar body. Deviations in the frequency of the partials have already
been discussed in Chapter 2.5: a spring-like bearing that bounces up and down extends the
effective string length and decreases the vibration frequency, while a bearing characterized by
a mass effectively shortens the string and increases the vibration frequency. Additional
partials also have already been deduced in Chapters 2.5 and 7.5: as result of the bearing
impedance containing an all-pass characteristic. The bearing impedance is, after all, not
infinite but depends in a complicated fashion on the frequency. Its frequency-dependent
imaginary part renders the effective string-length frequency dependent; this leads to detuning
of the partials. The frequency-dependent real part results in frequency dependent decay-
constants of the partials. All these aspects are string-specific effects of the bearings – the
exact source of which is to be documented in the following decay-analyses.
Fig. 7.93: Two superposed spectra, measured at a time interval of 100 ms. Frame = 21 ms, Kaiser-Window.
In the right-hand picture, only the spectrum measured after 100 ms is shown. * = areas of high damping.
Fig. 7.93 shows the measurement results of the string-decay analysis. A signal section of a
length of 21 ms recorded immediately after the plucking of the string was transformed into
the frequency domain via DFT-analysis; a second section recorded 100 ms later was treated
the same way. Comparing the two spectra (white vs. grey), we recognize the particularly fast
decaying partials: the strongest damping is found at 4.4 kHz – it shall be looked at in the
following. The second-strongest damping happens at about 3.8 kHz. Its cause becomes clear
as we look at the first few milliseconds of the signal (Fig. 7.84). Even before the relatively
slow flexural (transversal) wave reaches the measuring point (this happens at about 1 ms), a
faster longitudinal wave has already been reflected multiple times. Its effect is visible to the
laser-vibrometer only as an evoked transversal wave; this was already extensively elaborated
in Chapter 7.5.2. Since the impedance for longitudinal waves is about 20 times that of the
impedance for transversal waves [see appendix], the former wave-type finds much more
favorable matching at the bearing i.e. it is much more strongly damped.
Fig. 7.94: Measured lateral (particle-) velocity of

the string vibration. The circles mark the period of
an oscillation at 3800 Hz.

The simple formula for the fundamental frequency of the longitudinal wave confirms the
approximate location of the frequency; but it gives, compared to the measurement, somewhat
too large a value: for a string length of 63 cm and a propagation velocity of 5 km/s we should
arrive at about 4 kHz, and not 3.8 kHz. However, on the one hand we do not know the exact
density and E-modulus of the string. On the other hand, the impedance of the bearing
determines the phase of the reflection for longitudinal waves, and with it the exact oscillation
frequency. To re-check, the length of the string was shortened by 6% via application of a
capo, which increased the frequency of the observed irregularity from 3.8 kHz to 4 kHz. Very
generally, the following holds for this and all following interpretations: the investigated
irregularities do not result from definite, isolated effects, but from an interaction of many
components. Mono-causality must not be expected here.
Let us now look at the extreme damping of the partial at 4.4 kHz, the level of which drops
by 50 dB during the first 100 ms. Cause of this attenuation is the resonance of the transversal
wave carried on the remainder of the string on the other side of the bridge. At the bridge of
the ‘59 Les Paul, the strings form a sharp bend as they run across an adjustable bridge saddle
that is shaped like a mono-pitch roof. The remaining piece of string (residual string) ends after
about 3 cm at the stop-tailpiece. Since, as a 1st-order approximation, we can assume the bridge
to be immobile with respect to lateral movement, any flexural wave should in fact be reflected
at the bridge. However, due to the non-negligible bending stiffness of the string, there will be
a bending-coupling of the two sections of the string, as discussed at length in Chapter 2. It
can be easily verified that the fundamental frequency of the residual string amounts to 4.4
kHz by directing the laser vibrometer to it. Further confirmation is given by a small metal
clamp that is set onto the residual string, detuning its resonances – indeed the damping effect
shifts from 4.4 kHz to 4.6 kHz (Fig 7.95).
Fig. 7.95: Change in damping of the partial via an additional mass on the residual string. ∗ = 4.44 kHz.
Next, the attenuation of the 12th partial (2.44 kHz) catches our eye. At 10 dB /100 ms, it is not
as pronounced as the damping experienced by the partial discussed above, but still clearly
stronger than for most other partials. Deploying the metal clamp on the residual string has no
effects on this partial … the cause for this damping is an Eigen-resonance of the famous
Gibson-bridge (ABR-1). This resonance can again be shown with a small clamp that is this
time fitted to the bridge (Fig. 7.96). Adding such extra masses is a simple and quick
alternative to high-effort scanning analyses. While it does not enable us to determine the
shape of the vibration of the bridge, we can easily verify its involvement in the damping of
the partial. Before this measurement, the bridge-piece had been slightly readjusted to optimize
the intonation – already that had effects on the decay of several partials. Attaching a small
clamp to the ABR-1-bridge detuned the bridge resonances and led to a further change in the
decay of the partials. With this modification of the bridge, both the damping at 2.44 kHz and
at 1.82 kHz can be traced to resonances of the bridge – although these resonances always
needs to be seen in their connection to the residual string and the tailpiece.

Fig. 7.96: Changes in the attenuation of partials by attaching an add’l mass to the bridge. ∗ = 1.82 / 2.44 kHz.
As has already often been expressed: damping in musical instruments is not necessarily a bad
thing. The character of the instrument expresses itself in the damping of individual partials:
it’s the imperfection that results in individuality. Simpler said: the picking and the pickup
yield a spectral envelope, and the string bearing determines the decay of the partials. Not the
analyzer but only the listener decides what’s good or bad; and what is actually audible, as
well. Not all effects visible in an FFT-spectrum are audible, too.
Whether a partial is at all audible to begin with depends on so many parameters that just on
this subject, whole volumes are compiled [12]. If the 2.66-kHz partial decays faster than the
2.44-kHz partial, this is just about audible under laboratory conditions. A fundamental good-
guitar/bad-guitar discussion must never be started on such a basis. To bear in mind the
heading of this chapter: none of these effects is due to the wood of the guitar body – these are
artifacts due to strings and bridge, and they are quite substantial just looking at the physical
parameters. From the viewpoint of subjective perception, they are “almost insignificant”,
however. The resonances of the Gibson bridge attenuate the partials slightly above and
slightly below 2 kHz, and lead to a coloring that speech-scientists would attest a trend either
upwards to the “i” or downwards to the “a”, because the 2nd formants of these vowels lie
above and below 2 kHz, respectively (Fig 8.44). Attenuation at higher frequencies quickly
looses any significance for a guitar fitted with humbuckers, because the transmission range of
these pickups does not extend much beyond 2,5 kHz. The main effect therefore lies with the
bridge resonances, and of course with the resonances of the guitar neck as the measurements
in Chapter 7.7.4.4 have shown. It is impossible to build a resonance-free neck: density and E-
modulus result in masses and springs, and from this inevitably resonances. Fig. 7.97 shows
three Eigen-shapes of a beam clamped at one end. Transferred to the guitar, we would have
the body on the left side and the headstock on the right side. Real neck-resonances deviate
somewhat from this idealized picture, because the body does not represent an entirely
immobile clamp for the neck, because the cross-section of the neck is location dependent, and
because on top of the bending movement depicted here, there is also torsion of the neck
[Fleischer 2006].
Fig. 7.97: Shapes of Eigen-oscillations of a beam clamped at one end.
Not considering extra long baritone guitars and short 24”- or 22.5” guitars, the neck-lengths
of most guitars are very similar, and consequently we find similar Eigen-resonances. Not
identical, but similar. The neck-width may vary by 5%, the thickness of the neck by 10% –
these are not dramatic variations. The material and thickness of the fretboard will also modify
the neck-resonances somewhat, as will size and (a-) symmetry of the headstock.

Comparing different guitars, Fleischer [2001] identifies the asymmetry of the headstock as the
source of torsion-resonances: below the lowest guitar tone (83 Hz), a bending vibration exists
that has no direct effect regarding the tone generation. Towards higher frequencies, two
characteristic bending-vibration shapes have been established that have frequencies of about
200 Hz and 450 Hz for guitars with substantially symmetric build. Due to additional torsion-
vibrations, the former resonance splits up into two variants in guitars with asymmetric
headstock. The frequencies of the two variants can be up to 50 Hz lower or higher than those
of the “homogeneous” main vibration. A splitting-up of the second of the mentioned
resonances (450 Hz) was only found in one extremely asymmetric guitar (Gibson Explorer).
At its Eigen-frequency, the neck can be made to co-resonate particularly easily – if the
excitation does not happen at the location of a node of vibration. If, however, the place of
excitation (the bearing of the string) is located at an anti-node, a lot of vibration energy can be
transmitted from the string to the neck – which strongly dampens the vibration of the string.
For the open A-string, this case does happen for the second partial (220 Hz): the decay
analyses depicted in Chapter 7.7 show a relatively fast decay for this partial, the main cause
being the neck resonance. However, not only the nut also the fretting hand can act as an
absorber if it touches the rear of the neck. The same holds even for the guitarist’s belly – it
will always somehow touch the guitar body. Has anyone compared the belly-admittance of a
gaunt teenager with that of, say, a portly elder bluesman? No? But you did compare the wood
of the ’61 Strat with that if the ’64, didn’t you? The true connaisseur can hear entirely
different characteristics in a ’61 Strat compared to a ’64 [G&B 3/06]. (Translator’s remark: in
German, the English term anti-node translates into what would literally translate as “oscillation-belly” – which makes for a
great pun here in the German version that makes no sense in its translation (“excitation of an oscillation-belly”) … and does
end with an apology to the great B.B. …
Let’s summarize: the guitar body gives support to neck and bridge, and therefore is not
entirely uninvolved. However, the body represents a practically immobile base for the bridge,
as long as we deal with solid-body guitars (as they were considered here). Towards the neck,
the body is not totally static, and therefore the exact resonance of the neck depends on the
body, as well. However, before you run along to speculate about ash/alder differences, do not
forget to take a look at how the neck is mounted: remains of lacquer, shims placed in
between, uneven contact surfaces, and loose screws are potential sources for problems, just
like bridges and bridge saddles resting on hollows, or bridge saddles with bad notches. A
cheap plank of wood sourced from your local hardware store can be the basis for a great
guitar, while AAAAA-wood seasoned for 80 years may lead to disaster if there is only one
single mistake made in a joint somewhere. Of course, 80-year-seasoned wood, combined with
error-free, master luthier-y … that creates space for that 5-figure-stuff, and why not?!
The thinner the string, the less it is affected at all by the resistive component of the bearing-
admittance (or -impedance). This implies that the thicker the string, the more likely are
selective drops in the decay times due to the bearing. A set of 12s on an acoustic guitar will
be more strongly influenced in its vibration behavior by the guitar body than the set of 009s
on a Strat. Thus, if you chop up your SJ-200 to mount a Strat-pickup, sound differences to the
original Strat are easily conceivable. Within the group of solid-body electrics, however, the
wood the body is made of plays a highly subordinate role for the electric sound – here it is
(besides the guitarist) indeed the pickup that determines the sound. On the following pages,
the citations from literature that were already introduced in the introduction are again listed. If
the wood were actually and clearly a determining factor for the sound, these opinions should
not diverge so strongly.

Thicker neck = advantageous regarding the sound (G&B 8/02).

Extremely thin neck = round, fat primary tone (Jimmy-Page-Modell, G&B 10/05).
A thin neck does not feature any acceptable vibration behavior (G&B 3/97).
The Ibanez JEM 777 has an extremely thin neck-construction: the basic sound character is
vigorous and earthy (Fachblatt, 6/88).
Thin necks do not sound right. A nicely vibrating mass in the neck makes for more than 30%
of the tone. Go for a build that’s as light as possible, because the best sound is generated close
to the breaking point of the materials (LuK-Guitars, G&B 1/06).
Something that’s not true at all is that thick necks sound better than thin ones. I have built the
same guitar with a thick and a thin neck, and could not find any difference. Luthier Thomas
Kortmann (gitarrist.net).
Thin neck: the smaller the mass that needs to be moved, the more directly and quickly
articulation and tonal expression get off the starting blocks (G&B 3/05).
Zappy and direct in its response, every note takes off in a quick and lively manner, despite the
immense neck-bulk (that needs to be set in motion to start with) (G&B 9/05).
It is of sonic advantage that the neck weighs in with a lot of mass (G&B, Fender special
edition).
Bolt-on neck = shortening of the tone (Meinel).

A bolt-on neck can yield long sustain, too (Lemme).
Indeed, glued-in and bolt-on necks feature equal decay times (G&B 3/97).
Generally, maple necks are known to give the instruments a percussive touch (G&B 4/06).
The "Slab-Board" (rosewood fretboard) is one of the secrets of the highly praised, crystal
clear vintage sound especially of Fender guitars (Day/Rebellius).
The neck fitted with a rosewood fretboard has a fuller sound than a one-piece maple neck
(G&B 5/07).
The sound of the slab-boards is particularly fat; mids of enormous depths (G&B 5/07).
A one-piece maple-neck sounds just like a neck fitted with a fretboard (Lemme).
I like fretboards made from maple much better than the ones made from rosewood since the
former have a much tighter, stronger sound (Eric Johnson, G&B special Fender-edition).
The maple fretboard results in a clearer sound, the rosewood fretboards sounds "meatier"
(Duchossoir, Strat).
Without doubt, using Brazilian Rosewood for the neck decisively contributes to the sound of
the PRS-513 (G&B, 2/05).
It is certainly not exaggerated that Rio-rosewood generates a “full octave of additional
harmonics” (Day et al.).
Rio-rosewood is much harder and quicker in the response compared to East-Indian types
(G&B 4/09).
But – that’s all horseshit, isn’t it? Old Indian rosewood sounds just as nice as Rio-rosewood,
after all (G&B 5/06).
It appears that the material used for the neck in fact exerts even more influence on the
primary sound than the wood of the body (G&B 4/08).
Solid-body guitars, however, may be built in almost all shapes and sizes – we should not
expect significant effects on the sound from this (Day et al. p.140).
Looking at the process of the sound generation, it quickly becomes clear that the condition,
and the type of wood used, exerts an influence on the sound of the instrument just as massive
as its design. (the same author, the same book, p. 206).

Wood does not influence the sound (May, p. 144).

Wood influences the sound (May, p. 145).
High-grade wood is unnecessary (May, p.86).
The influence of the wood on the sound should not be underestimated (G&B 3/97).
The sound of an electric guitar depends mostly on the pickup (Lemme).
To a relatively strong degree, the sound of an electric guitar depends on the wood (Meinel).
The experts agree that the sound of a solid-body guitar is predominantly determined by the
electronics (Carlos Juan, Fachblatt Musikmagazin, 1996).
The sound does not mainly depend on the pickup – rather, the wood generates the basis;
therefore you should listen to an electric guitar without amp first (Jimmy Koerting, Fachblatt).
Pickups transform the vibrations they come upon into sound, and are not generating sound
themselves (G&B 5/06).
p.205: the type of construction has a massive influence on the sound. p.140: All sizes and
shapes in solid-body guitars, without significant effects on the sound. (Both: E-Gitarren).
Wood not only determines the color of the sound but mainly the information of the string
vibration. (G&B 02/00).
It’s probably known that light tonewoods feature particularly good vibration- and sound-
characteristics – this does not hold universally, though. Many a 4½-kilogramm-guitar has
turned out to be extremely resonant (G&B 2/06).
The denser the wood, the more brilliant, treble-rich the sound; the higher the stiffness, the
longer the sustain (P. Day).
The older the wood, the drier it becomes. The lack of liquid makes for more vibration, this is
to be equaled with more sound (Marc Ford, G&B 8/07).
Besides, I actually think that the component wood is, in general, overrated (Ulrich Teuffel,
G&B 5/04).
Bob Benedetto, whom many (practically all) take to be the best luthier alive, states: “popular
opinion demands wood that has slowly grown (slow growth shows in narrow tree rings).
According to my knowledge, that is a myth. … some of my best guitars are made from spruce
that some would take as substandard. Check out the old masterpieces from Stradivari or
Guaneri – they are made from wood with wide tree rings, as well. Maybe we have fallen, for
years, for the advertisement in the brochures of a few companies that promote wood with
narrow grain. … Once I went to a wood supplier in Pennsylvania and bought the worst wood I
could find. I built a guitar from it that sounds excellent – after all, Scott Chinery bought it.”
(G&B 9/02).
A connection between the width of the tree rings and the acoustically important
characteristics of resonance woods cannot be specified (D. Holz, IfM Zwota).
The latest investigations in the Institute for Musical Instrument Making essentially confirm
this (G. Ziegenhals, IfM Zwota).
Bob Taylor is said to have stated that his 300-series beginner guitars offer 90% of the sound
of the 900-series premium guitars at not even 1/3rd of the cost. Such a comment clarifies that
it is predominantly the design and the build of the guitar that characterize the sound to a much
greater extent than the woods used (Gerken et al.).
Taylor builds good guitars because we now how to do it. To prove that, we have built an
acoustic guitar from an old, rotten pallet we found in the garbage. The top was from a
scrapped plank of which we could not really determine the wood. We so elaborately glued
together the top from 6 slats that it is hard to even detect that, and the holes from the nails …
were highlighted with small aluminum discs. This pallet-guitar was one of the most noticed
guitars at the winter-NAMM-show (Bob Taylor, ISBN 3-932275-80-2).

The Platinum Beast sounds powerful, warm and balanced, with velvety brilliance and delicate
harmonics; the Evil Edge Mockingbird sounds somehow feeble, deprived of mids, with
somewhat more succinct bass, but instead much more brilliant and harmonically richer.
Thanks to the hot humbuckers, this all sounds entirely different when connected to the
amp, because – hard to believe – both instruments now sounded almost identical (G&B 8/06).
Comparison: Gibson New Century X-Plorer vs. V-Factor: surprisingly, the differences in
sound we found in the dry-test showed up much less when connected to the amp (G&B 7/06).
Ash-Strat vs. poplar-Strat: only “minute differences” (G&B Fender special edition).
Alder-Strat vs. poplar-Strat: differ only in the “finest nuances” (G&B 10/04).
Squier-Stratocaster: comparison: mahogany body vs. basswood body: using the middle or
neck pickup, the two guitars sound almost identical (G&B 5/06).
"The 94-Amber (pickup) indeed transports a pronounced Strat-tone – and it does that as a
full-size humbucker and implanted into a Les Paul of 4 kg and typical mahogany/maple-
combination. … In particular the neck pickup reminds us – in its tonal color – of an ultra-fat,
Texas-Blues-heavy Stratocaster – an awesome sound that we would have never connected to
a Les Paul” (G&B 11/07).
A Strat will never become a Les Paul, even with a humbucker (G&B 2/00).
By far the “Strat-iest” Gibson sound that I have every heard. Nighthawk (G&B 5/09).
Still, the PRS EG surprises with incredibly authentic Strat-sounds; mahogany neck,
rosewood fretboard, mahogany body (G&B 9/05).
"The purely acoustical comparison gives opposite insights compared to the earlier comparison
of the Mexico Classics. Now the 50s-version delivers the more balanced, open and zappy
sonic picture while the 60’s-version sounds more mid-focused, warmer and somehow more
well-behaved.” (G&B 2/02). About the cited Mexico Classic, we read: “The 50’s Strat
generates a strong, mid-focused sound picture, defined by crisp, concise bass, delicate
harmonics and a certain warmth. More brilliance, a more lively harmonic spectrum, more
open mids, and a somewhat gentler bass is what the 60’s Strat offers”. However, there is also:
“the A/B-test indeed reveals only tiny differences.” (G&B, Fender special edition, Mexico-
Classics comparison). In both comparisons, the 50’s-Strat features a one-piece maple neck the
upper surface of which forms the fretboard, while the 60’s-Strat has a one-piece maple-neck
with a glued-on rosewood fretboard.
Hairline cracks are of the highest importance for the sonic results (G&B 2/07).
We were able to borrow a ‘56 and a ’58 Les Paul Standard, and fabricated exact templates
of the original shapes and contours. In the process we realized that the Historic-Collection-
model had slight differences to the two originals. … Since it was not possible to change
anything about the Silhouette (the Historic-Collection-model is meant here), at least the
contour of the top was to be matched. Using a violin-maker’s device, we took the exact
contour of the old Les Pauls and shaped an exact model from wood. From this model, we then
shaped the new contour. This was an elaborate procedure because work had to be done using
the smallest wood planes and card scrapers. ... (Pipper, G&B 12/06).
... they have made them a molten calf, and have worshipped it, and have sacrificed thereunto,
and said, these be thy gods, ..., (The Bible, Exodus 32.8).

Alder: silky, mellow, warm, tender, many harmonics, restrained share of treble, fat bass,
rather subdued share of bass, strong mids, round share of mids, much sustain, succinct,
squishy, good presence, undifferentiated, balanced, full sound, a sound thinner than that of
basswood, faster response than basswood.
Basswood: mellow & low mids, squishy, good response, undifferentiated, somewhat mid-
laden, similar to alder, relatively little sustain, warm sound that lacks zappy-ness, unobtrusive,
forceful, rather dull-sounding.
Poplar: clear treble, more airy than basswood, unobtrusive, round sound, like basswood but
thinner, the tonal characteristics correspond to those of alder but lack warmth and brilliance,
more crisp than basswood.
Maple: rich in attack, brilliant, rich in harmonics, lively, much sustain, not warm, warm bass,
lacking warmth, mid-emphasizing sound, hard sound, singing tone.
Ash: mellow, rocking, soft, bass-y, brilliant, no pronounced share of mids, balanced, lively,
powerful, tight, warm bass, long sustain, dry, airy, hard-wood-y, rich in attack, strong
assertiveness (because ash is of stiff structure), responds considerably faster than alder,
brighter and richer in the harmonics than alder.
Swamp ash: balanced, perfect balance of brilliance and warmth
Mahogany: mellow, low-mid emphasis, very bass-y, good sustain, delicate brilliance, silky,
warm sound, warm mids.
Rosewood: powerful, harmonic sound, airy basic character, loose and full bass range,
sparkling treble, Rio-rosewood generates a full additional octave of harmonics.
Neil Young: I am convinced that very note ever played on a guitar somehow remains in it.
While it does leave the guitar body as sound, it still is within the wood. Everything that
happens on a guitar remains in it and sums up to an overall experience (G&B 12/05). Chris
Rea: it’s funny – often the cheapest guitars sound the very best. … the Epiphone Byrdland is
4000 pounds cheaper than the Gibson Byrdland, and I cannot feel any difference – apart from
the logo on the headstock (G&B 12/05). Richie Sambora re. the topic of “sound”: “You still
hear, however, that Hendrix went directly through the amp. It’s his fingers. The same with
Jeff Beck: you may use his rig and his guitar, but you will never sound the same. It’s in the
fingers.” (G&B 11/02) Van Halen: it’s not a question of equipment – it’s the fingers (G&B
7/04). Eric Johnson: the source of more than 75% of the sound is in the fingers (G&B 5/01).
Jeff Beck: no shenanigans, no mumbo-jumbo – just the fingers (G&B 3/07). Jaco Pastorius:
piss off the amp and piss off the instrument. It's all in your hands (G&B 1/06). Victor Bailey:
once I had the opportunity to play Jaco Pastorius' Jazz-Bass; you cannot imagine how
terrible it was: lousy action, didn’t sing at all. I was thoroughly disappointed. Jaco noticed
that, grabbed the bass and played. It sounded gorgeous: the bass sang and growled (G&B
1/06). Snowy White: Peter Green sold his Les Paul to Gary Moore. I jammed with Gary
once and it sounded o.k. But since it left the hands of Peter, it’s just an ordinary guitar –
nothing special anymore. A guitar is fabulous only as long as somebody fabulous plays it
(G&B 11/07). Jan Akkerman: it all comes down to your hands (G&B 1/07).

The largest part of the string vibration is to be transmitted to the guitar body. If the latter is
supplied with unrestrained vibration energy, a maximum of tone and sustain develops. (G&B
12/05). Because the nut should transmit the vibration energy as fully as possible to the neck
(G&B 6/07).
The design shows considerable resonance characteristics; after every picking attack it vibrates
intensely and clearly noticeable. (G&B 9/06).
From a vibration engineering point-of-view, the MTM1 ranks at the highest level, because
the whole design resonates intensely to the last wood fiber after each picking attack, and a
slowly and steadily decaying sustain results (G&B 8/06).
Although the Lag vibrates with marked intensity, and lively after each picking of a string,
only a somewhat anemic sound reaches the ear … The bridge pickup, for example, pushes
through to the ear in a powerful and assertive manner. … The single coil at the neck pushes
the low end and the lower mids with much power (G&B 12/06).
Picking up the Pensa-Suhr-guitar and playing it unamplified, a reasonably trained ear
immediately hears that this is gonna be good. … Both standing up and sitting down, you feel
already in your belly the fantastic vibration behavior of the excellently matched woods
(Fachblatt, 6/88).
Since a relatively large body mass (3,9 kg) needs to be excited to vibrate, the response seems
a bit ponderous, and the tones do not get off the starting blocks as quickly (G&B 7/06).
The guitar vibrates intensely, responds directly and dynamically, every chord and every tone
unfolds zappily and lively. Weight: 4,15 kg (G&B 8/06).
Less mass can more easily be made to vibrate (Kortmann, gitarrist.net).
Despite the enormous wood-mass (3,85 kg), almost every tone responds zappily and
dynamically, and unfolds very swiftly (G&B 7/06).
Thinner guitar body = less bass (G&B 4/04).
Sparingly varnished guitar body = rounder, more succinct tone (G&B 7/05).
A more slender guitar body makes for a more slender tone, too (G&B 7/02).
The tone of a guitar with a fully hollow body is fragile and has an enormous momentum
(G&B 8/06).
Guitar with hollow body = more mellow sound (May).
Brian Setzer is known for his extremely powerful, in fact brash sound that only archtops with
suitable pickups can offer (G&B 8/06).
Semiacoustic guitars sound brighter, more transparent, more brilliant (E-Gitarren).
335-Sound: a warm, fat sound that is highlighted, due to the semiacoustic build, particularly
in attack and response (G&B 1/07).
In the hands of Alex Conti, the 335 sounds not much different than his Les Paul. The fingers
make much more of a difference than one would think (Richie Arndt G&B 9/07).
Danelectro: hollowbody, decent sustain, probably thanks to the maple neck with the luscious
rosewood fretboard (G&B 12/06).
Cavities (in the solid-body guitar) have no impact on the sound (Lemme). To improve the
body's resonance, the core body is drilled with eleven 1,5"∅ cavities". (Duchossoir, Tele).
The cavities in the Les Paul have no effect on the sound-characteristic of that model – we
have tested this (Henry Juskiewicz, president of Gibson; Les Paul Book). The Les Paul
Custom Classic receives an additional percussive and crisp touch from the milling in the
wood. The Gibson Custom Shop now offers some models as so-called chambered variants.
What was introduced simply as a means to save weight back in the day now receives an
entirely new, tonal significance (G&B 8/07).

The electrified plank-guitar is predominantly an acoustic instrument. The wood makes for the
character of the sound; the pickups contribute only a very small part. And so a humbucker
cannot exorcise the characteristic unfolding of sound and attack from a Strat with an alder or
ash body (G&B 02/00). Edward van Halen: the boys in the band didn’t like the sound of the
Stratocaster because it is naturally so thin. So I mounted a humbucker (G&B 9/02).
Gary Moore: some people think they hear a Stratocaster on “Ain’t Nobody” – however, in
reality that’s my own signature Les Paul, (G&B 7/06).
Jimmy Page recorded the entire first Led-Zeppeling album using a Telecaster; the guitar
sound on that album is exactly like that of a Les Paul (G&B Fender special issue).
Mark Knopfler: if I want a thicker sound, I use my Les Paul – that’s not to say, though, that I
couldn’t do the same thing with a Stratocaster. Even if B.B. King plays a Fender, it still
sounds like a Gibson Lucille (G&B 9/06).
Les Paul Custom: one-part mahogany body (The Gibson).

Around 1952, the Gibson designers produced prototypes of their first solid-body guitar, the
Les Paul, completely made of mahogany. This design did not satisfy them tonally, though, but
rather motivated them to carry out further experiments with other types of wood. The result
was a mahogany body with a maple top (Day et al.).
Les Paul: back then my idea was to build the whole guitar – i.e. headstock, neck and body –
from one and the same piece of wood. They didn’t do it. When I asked the president of
Gibson why not, he said: “because it is more inexpensive this way” (G&B 9/05).
Gibson Les Paul: "The rims of the electronics compartment and switch chamber again reveal
appalling workmanship: they are partially downright frayed, and the impression rises that the
wood to be removed was blasted away. ... Just about tolerable to me is the pickup switch that
due to the curvature of the top lives in its chamber in a totally crooked manner, touching the
milled wall …” (G&B 12/06). Only the price seems to be on target: 2655,-- Euro.
Lester Polfus answering the question whether he had ever imagined that the Les Paul could
be such a successful guitar: “Of course. I believed in this guitar from the very start” (G&B
Gibson special edition). But then, he also says: "Never ever. I would not have thought that
this guitar could be that popular 60 years on” (G&B 9/05).
The image of old Les Pauls was forged systematically by the pertinent dealers; they simply
imputed the vintages 1958/1959 with a legendary sound (Carlos Juan, vintage dealer, in
Fachblatt Musik-Magazin, 1996).
Investigating the term “vintage” more closely, it turns out to be substantially an empty
catchword that frequently serves to sell questionable product at inflated prices (Lemme).
Most vintage instruments are not suitable for serious stage work in their original condition,
and as they are being made workable, they are not vintage anymore. The opinion that
everything becomes wonderful or improves because it is 50 years old or carries a spaghetti-
logo, is itself long in need of repair (Carlos Juan, vintage dealer, in Fachblatt Musik-Magazin,
1996).
Kevin Walker: I would never buy a Gibson built later than 1972. … only the vintage stuff
has the good sound (G&B 5/06).
Well … it’s a piece of wood with 6 strings on it – that must not be overrated. Pat Metheny on
his guitar (G&B 6/08).
Certain is that nothing is certain, and therefore I am wary – just to be safe (loosely translated,
after the Bavarian poet Karl Valentin).

7.10 Guitar bridges 7-117
7.10 Special Bridge Designs
The guitar string is supported by two bearings: at the nut (or the fret) and at the bridge. The
nuts of acoustic and electric guitars may show variations, but normally they are of similar
build. The bridge, however, will be different for these two categories of guitars (save for some
exceptions). In acoustic guitars, the bridge often consists of a light (!) saddle made of bone or
plastic while the strings of an electric guitar will rest on a solid, massive, adjustable
contraption made of steel. Electric guitars often feature individual bridge saddles shaped
cylindrically or like a mono-pitched roof, and adjustably resting on the bridge base.
The bridge of an acoustic guitar needs to be light so that any vibration energy worth
mentioning can be transmitted to the top of the guitar. In contrast, the bridge of an electric
guitar (where the body is not supposed to vibrate) may be of a very solid build. A few
adjustment screws will obviously not get in the way here, because otherwise they would not
have been included with much enthusiasm. This adjustment possibility is not entirely useless,
either (as was elaborated in Chapters 1 and 2), because: in order to achieve correct tuning, the
steel strings require corrections in length that may amount to up to 5 mm. We therefore have
adjustable bridge saddles und adjustment screws. Leo Fender had still been mightily thrifty
when designing his first electric, the “Broadcaster”; he positioned two strings each on a steel
cylinder (later a brass cylinder). Apparently, it was attractive to make each string individually
adjustable because the successor, the “Stratocaster”, featured string-individual bridge saddles
made out of pressed steel, and adjustable both in length and height. Although no classical
guitarist will ever demand this from his Ramirez, it seems almost indispensable for the
electric guitar to have the action fully adjustable – best in three dimensions: height, length,
and distance between the individual strings.
Inevitable, this possibility of adjustment entailed a diversification in the components: what

had successfully been achieved with just a board and some strips of bone, suddenly required
screws, springs, straps, wires, curled nuts, rollers … a hodgepodge of 18 parts or more. Do
these vibrate, then? Hold on – these are the ‘50s, and this question was not on the table, yet.
Sets of 012 strings were standard; these could not easily be pushed out of the groove in the
bridge saddle – something that would become a potential problem with the later 009, or even
008 string sets. "The 'Floating Bridge' consists of a master bridge adjustable to varying
heights. On it rest the six individual bridges each adjustable for string length and height,
making possible extremely accurate adjustments for perfect intonation and custom playing
action [Fender Jazzmaster, 1968]." This is the way a faulty design was advertised back then,
and room was created for "retrofitters" who could earn their money with accessories for
correcting the mistakes. Chief issue: to be adjustable.
Admittedly, it was not easy to design a bridge that was solid and at the same time adjustable.
In particular, many a guitar was now treated to a vibrato-system: a spring-mounted string-
retainer that offered variation of string length – and thus pitch – via a lever (whammy-bar).
However, with varying length, the strings needed to slide across the bridge somehow – or the
bridge could be made movable in the direction of the string, and move with it. The latter
approach resulted in the 'floating bridge' of the Jazzmaster (and other guitars). That bridge
could develop, with thin strings, an undreamt-of potential to float around. Under these
circumstances, Leo Fender’s unceremonious renaming of the vibrato-effect into tremolo was
no help, either: this wobbly-jelly did irritate more than just a few guitarists. Everybody else
was of course highly enthusiastic: "Careful design and outstanding playing characteristics of
the Jazzmaster have made it one of the favorites of guitarists around the world [Fender
1968]." Cheers, then!

7.10.1 Simple equivalent systems
The guitar bridge mechanically interconnects string and guitar body. As a system of
mechanical vibration, it is an object of mechanical systems theory, the latter analytically
representing movements and forces. The small masses, stiffnesses and resistances
differentially distributed over a continuum can, however, not be described with complete
accuracy – only given the limitation to a finite effort the simplified representation via an
equivalent system is possible. In contrast to the continuum, the equivalent system consists of
a few, discrete elements that vibrate in one dimension only (a further simplification).
Mass (Newton), stiffness (Hooke), and frictional resistance (Stokes) are the fundamental
elements of mechanical systems. While a mass can relatively easily be specified as a
multiplication of density and volume, the analytical description of stiffness, and in particular
of resistance, is difficult. As an example, Fig. 7.89 shown a cylindrical pin made of metal, the
rounded lower side of which sits on a flat surface. The mass of the pin is easily calculated, as
is the stiffness of an axially loaded cylinder (sZ = ES / l).
Fig. 7.98: contact stiffness

F = axial force, E = E-modulus,
R = radius of round
Given an elasticity modulus E = 2.1⋅1011 Pa, the axial stiffness of a steel cylinder of 4 mm
length and a diameter of 2 mm is calculated, resulting in sZ = 165 MN/m. However, the
largest deformation does not happen in the cylindrical part of the pin but at the contact point.
Assuming a spherical round, the axial contact pressure force leads to a circular contact
surface. The radius r of the latter depends on the contact pressure. The stiffness occurring at
the contact point is force-dependent, as well: with increasing force, the stiffness increases,
too. With the keywords contact problem and Hertzian stress, specialist literature [z.B. Szabó]
offers approximations for the deformation from which the contact stiffness sK can be
calculated. There are several contact points in a guitar bridge, and therefore several
stiffnesses. The magnitude of the latter depends on two variables: on the radius R of the
round, and on the force. Both the pressure force perpendicular to the guitar top, and the
traction force in parallel to the top depend on the force of the string tension that amounts to
between 47 and 135 N (for a set of 010 strings, the benchmark is 80 N). In the Fender bridge,
the axial force acting on the height-adjustment screws moreover depends on the bend angle of
the strings as they run across the bridge; for e.g. the Jazzmaster this would be only about 6°.
Given a string tension of 80 N, a pressure force of 8.4 N results, and since two screws support
each string, the force is 4.2 N per screw. The calculation results in a contact stiffness of just
under 5 MN/m, with a radius R = 1 mm. For bridges with a higher bend angle (e.g. the
Stratocaster) the contact stiffness mounts and can reach, for thick strings, up to 15 MN/m.
This is still much smaller than the axial stiffness estimated above, so that the conclusion for
the aforementioned cylindrical pin is: in terms of its effect, the contact stiffness is the
dominant one of the two stiffnesses.
Besides the contact pressure force, the radius of the round R is also found under the square-
root in the above formula – and here things become complicated: this radius may vary
depending on the deployed screw and the manufacturing quality, and therefore the resonance
frequencies dependent on R may vary, as well! Similar issues appear for all other joints where
two components lie on top of each other: depending on the surface roughness, and on the
more or less protruding drilling burrs, an undefined bearing results that may undergo further
variations when the strings are changed.

The first joint occurs between string and bridge saddle. The approximation using a spherical
surface certainly is inappropriate here; due to the string flexion, a roller-shaped interface
surface (like in a roller bearing) does not correspond to reality, either. The geometry of the
string is predetermined, and cannot really be changed due to its high hardness and stiffness.
Unknown, however, is the geometry of the bearing surface (the bridge saddle). The high E-
string (E4, ∅ = e.g. 0.25 mm) rests on a probably 10-µm-wide strip; in this scenario, a
production tolerance of 1 µm would be advisable – not something that “every manufacturer is
likely to achieve”. Thus, the mechanical data of this joint can be estimated only very roughly.
The second joint is located between bridge saddle and adjustment screw (there may be up to
three of the latter per bridge saddle). Where the contact surfaces actually occur, and what the
corresponding stiffness is, remains completely undefined, just as the resulting friction. In case
of higher age, the degree of rust and corresponding mechanical parameters are also undefined.
The third joint occurs between screw and bridge base (or directly between bridge saddle and
bridge base). Adjustment screws (so-called setscrews or headless screws) come with 4
different end-surfaces: level, spherical (convex), tapered, or concave (Fig. 7.99). These
screws are mass-produced and not optimized with regard to any requirement of vibration
mechanics, and therefore the contact stiffness varies from one screw to the next. The contact
stiffness also changes as the screw is turned (an action that must be considered a regular fate
of any screw).
The fourth joint is found between bridge and guitar body, and again, what has been stated
above holds (as it is the case for any further joints): the stiffness and the resistance of/at the
joint are largely undefined, as are their effects on the resonances of the bridge.
Fig. 7.99: Typical setscrews of a guitar bridge.
Can the guitar then function at all? Sure it can – some kind of stiffness and resistance will
always develop; the term “undefined” used above merely means that the corresponding values
may vary from one guitar to the next. Some of the involved variations may be without any big
effect on the sound, but some will lead to audible inter-individual differences. Because it is
very difficult to determine the joint-parameters of a given guitar, a different approach shall be
applied now: in a model, we will assemble some basic elements (Fig. 7.100), and for these –
and some modifications – the frequency responses of the conductances will be determined.
Fig. 7.100: Simple equivalent system for a guitar placed on a stone table.

Fig. 7.101 shows the results of the calculations corresponding to Fig. 7.100. Between string
and bridge saddle, the very simple Kelvin-Voigt model consisting of a lossy spring was used.
Even though the values of the latter are unknown and can only hypothetically be assumed: it
only has an effect in the highest frequency range that is relatively unimportant when magnetic
pickups are used. This finding holds even if the actual stiffness were only 1/10th of the
assumed value. As an orientation, the conductance calculated for the E2-string is shown in
grey in the figure (correspondingly see also Chapter 7.7.2. & 7.7.3); the more the bridge
conductance is below this grey line, the less it bears any importance to the overall damping.
Somewhat more important is the (lossy) spring (s2, W2) located between bridge saddle and
bridge base. It influences the high-frequency resonance that is found at 7.5 kHz for the above
values. Again, we need to bear in mind that there are no measurements as basis for these
values, and thus it is possible that the grey curve is crossed (e.g. for a smaller resistance W2).
In this model, particularly important is the bridge-base resonance formed (in approximation)
by the mass of the bridge (50 g) and the spring stiffness (s3 = 4 MN/m). Measurements with
Gibson bridges show similar resonance behavior and high string damping (Chapter 7.10.2).
The next spring in this model is found between guitar body and stone table (s4, W4) – it
influences mainly low-frequency resonances. The stone table with a mass of 250 kg vibrating
aperiodically damped with 2 Hz forms the conclusion: it is insignificant for the current
measurements.
Fig. 7.101: Input conductance of the system according to Fig. 7.100; variation of the system parameters. The
grey line is the “orientation curve” recalculated from Fig. 7.66 (E2; esp. radiation damping and inner damping).
Fig. 7.102: Conductance measurements (thin line); left: Les Paul body, right: Les Paul bridge.

For comparison, Fig. 7.102 shows the related measurements. While the multitude of highly
different resonances can of course not be modeled with such a simple equivalent-circuit
approach, the order of magnitude fits well. It is in any case conceivable how the addition of
further resonators enables the model to also represent narrow peaks in the frequency response
of the conductance (Fig. 7.103).
Fig. 7.103: Additional resonance circuit (Zx), effect on the frequency response of the conductance.
The equivalent circuit shown in Fig. 7.100 is no complete model of a guitar – it does not even
begin to represent the multitude of body- and bridge-resonances. These resonances may not
only occur as one-dimensional vibrations (as contained in the model), but can take on the
shape of three-dimensional flexural vibrations, in combination with torsion-vibrations. Still,
the model allows for estimating the approximate orders of magnitude of the stiffnesses and
frictional resistances, and of the approximate effect the latter two have on the bridge
resonance. It is obvious that this resonance can influence the decay of all partials und thus the
sustain. However, if the bridge conductance is small (e.g. 10-4 s/kg), then the bridge and
everything that is mounted to it (including the guitar body) has practically no effect on the
sustain! The measurement curve shown in the right section of Fig. 7.102 crosses the grey line
only twice: between 100 Hz and 200 Hz, and at 1.8 kHz. If a partial falls into one of these
ranges, then the absorption at the bridge does influence the decay process. The other
resonance peaks may be attributed a theoretical influence (“everything depends on
everything”), but they have no practical relevance.
The considerations related to the contact problem have shown that stiffness and resistance
strongly depend on the contact pressure force and the contact surface. Both change if the
guitar bridge is shifted within the clearances given by manufacturing tolerances. The bridge
saddles of a Stratocaster may have contact to each other – or not. The bridge saddles of an
ABR-1 may have a burr on their lower surface, they may have contact on one side or on both
sides, or they may be clamped down by the set-screw with a force fit. The distribution of
forces (and therefore the stiffness) between the 6 screws holding the old Stratocaster bridge is
undefined and depends on the smallest of manufacturing tolerances – or on the tear and wear,
which does not make things any simpler. Changes in the contact parameters do not
necessarily lead to changes in the sound but they are potential sources of damping that need
consideration due to their closeness to the string.
The varnish of a solid body guitar, on the other hand, is far removed from the string, and its
mass is small. Still, for completeness sake a few citations: However, practical use has taught
us in the past that very sparingly varnished instruments have generated a rounder, more
succinct tone [G&B 7/05]. Hairline cracks (in the varnish) lead to an un-damping of the
resonating body [G&B 1/06]. The varnish can constrict an instrument and thus dampen it, or
it can adapt itself to the natural resonance characteristics and co-resonate [G&B 1/07].
Actually, any beer-belly will do the same ...

7.10.2 Bridges without Vibrato
7.10.2.1 Gibson’s ABR-1-Bridge

Orville Gibson was a guitarist and a luthier – characteristics that were not originally natural to
Leo Fender, who of course was a builder of guitars, too – but not in the traditional sense. As
Tom Mulhern notes in his book on Gibson guitars [Rittor 1996], Orville’s ideas often
originated from violin making, and it is therefore not surprising that the famous Style O guitar
features an arched top that carries merely the bridge but no tailpiece. The strings are anchored
in a trapeze tailpiece that is mounted to the end-side of the guitar body. This separation of
bridge and tailpiece resurfaces half a century later in the ES-335, although that guitar is based
on a very different principle of construction. The top of Orville’s acoustic guitars needed to be
thin in order to radiate sound. While it was still possible to anchor gut strings in a combined
bridge/tailpiece that was glued to the guitar top, this approach became a problem with the
steel strings increasingly demanded by musicians: their higher pull (parallel to the top) could
warp the top, or rip off the glued-on bridge and destroy the thin top. Conversely, with a
tailpiece mounted to the side at the end of the guitar, the top was subjected merely to a
perpendicular force it was able to withstand due to its curvature similar to the arch of a bridge.
The bridge of the Gibson Style O is of a single piece and not adjustable – again similar to that
of a violin. However, 2-piece bridges soon found their way to the Gibson acoustics,
presumably so that the action that increased with age could be compensated for. The 2-piece
bridge includes a base and an upper part both made from wood; the 2 sections can be spread
apart via screw and curled nut. Starting out from this construction, it is not all that far to
Gibson’s patented Tune-O-Matic bridge (US patent 2,740,313, filed in 1952), the top part of
which carries six individually adjustable bridge saddles. 6 bridge saddles, 6 adjustment
screws, one bridge base, 2 post screws, 2 curled nuts, and the fastening wire that arrived later:
all in all that’s 18 individual pieces. With this bulwark between the string and the guitar body,
it is no wonder then that the latter has so little influence on the string vibration. Worse,
though: the joints occurring between string and guitar body are undefined to a high degree!
The T-shaped bridge saddles are positioned within a groove to which they have contact in
some kind of way. The contact between bridge and the curled nuts is not defined, either, and
consequently it is no surprise that the mechanical characteristics change as we lightly press
against the bridge. Still, the contraption does work – in fact some masterful guitar playing
happens using it. A word, however, to all you Gibsophiles ecstatically dancing around every
golden calf-o’-1956: before pondering about the woods, you should target the bridge,
beginning with the question which way round the bridge should be mounted. On most of the
guitars shown in the Gibson book, the heads of the setscrews point to the pickups, but for
quite a few, they point to the tailpiece. Indeed, the screwdriver access is easier in the latter
case, but now the strings run across the screw heads! These are the residual strings between
bridge and tailpiece; they may contribute to the vibration absorption, as shown in Chapter
7.7.4. Thus: it may not be the hairline cracks in the varnish that “are of highest significance to
the resulting sound [G&B 2/07]" – rather, the bridge may contribute much more.
The strings excite the bridge saddles (T-shaped when seen from the tailpiece, and of
monopitch-roof shape seen from the side) to vibrate – the saddles should resist this excitation
so as to keep the vibration energy within the string as much as possible. The force fed from
string to bridge saddle splits up into an inertia force (to accelerate the mass), and a remaining
force that is conducted on to the bridge base. Between bridge base and bridge saddle there are
several joints the mechanical impedance of which is of significance to the string vibration.
Therefore, requirements regarding the manufacturing tolerances of these components would
be very high. That formulation should be agreeable even to laywers, shouldn’t it?

If there were any burrs on the Gibson bridge (with “were” expressing purely hypothetically a
possibility), the setscrew in its end position would lever the bridge saddle halfway out of its
embedment, and the bridge-base/bridge-piece impedance would drastically change. Where in
fact does the line of force-flux between string and guitar body run for this bridge? Fig. 7.104
shows several views of the Gibson ABR-1. A bridge saddle (T-shaped or of the shape of a
monopitch roof depending on the view angle) is movable within a groove via a setscrew, with
the string (secured in a small groove) resting on the saddle. On what does the latter rest?
Fig. 7.104: String bearing

in a Gibson ABR-1 bridge.
Since the middle section of the “T” does not reach to the bottom of the groove, we could
surmise that the bridge saddle rests directly on the sidewalls of the bridge. That, however, is
not the case – not for the Historic Les Paul under scrutiny, anyway, nor for the ES-335 from
the 1960’s. Every introductory course for mechanical engineering includes the lesson that
objects not supposed to move need to be fixated with regard to three translational and three
rotational movements. Translational movements are longitudinal movements (in the
direction of the string-axis z), lateral shifts (x) and changes in height (y). In the z-direction,
only the setscrew can absorb any forces – but it does so with some slack. Pressing the bridge
saddle to the right (in the figure), the conical screw-termination has contact, pressing it to the
left, it is the chamfered collar that stops the movement. Possibly, the whole setup was at some
point meant to remain under tension and therefore be without slack – the implementation
ain’t, though. For the y-direction, it is immediately clear that either the screw, or the lower
side of the T-piece can transfer any pressure force from the strings, but not both (dividing the
force would be at random and fragile). If the bridge saddle rests on the bridge, the setscrew
has slack, and if the setscrew absorbs the force, the bridge saddle has slack. Purely
theoretically, we could consider of shift-fitting or pressure-fitting – but only those without any
experience in production of mechanical elements will go there. No, that T-shaped saddle has
slack, resting somewhere on something, depending on production tolerances (Fig. 7.105).
Fig. 7.105: Changes in position.
Trying to push a piece of paper in between the T-shaped bridge saddle and the bridge base is
met with success, and proves that the two do not rest slack-free on each other. This test does
not work everywhere, but at several places. In the worst case, this instability leads to torsion
movements – then the string vibration is completely done for. Dear Mr. McCarty (rip), how
was that supposed to work? String movements in parallel to the fretboard imply torsion
excitation – anyone disagreeing? The sales speak for themselves? Ah – sorry, that explains
everything. We can hope that the whole contraption somehow gets wedged or rusts shut (and
for many guitars that will in fact happen) – a planned force-fit looks different, though.

For Fig. 7-100, the interface between bridge saddle and bridge base had been modeled with
Z2, and Fig. 7.101 shows the effects of variations. That changes in stiffness and damping
occur only in the high-frequency range might explain why this bridge in not entirely
impractical. As long as the bridge-piece-T rests (or is in firm contact) somewhere, and as long
as it is not subject to any torsion movement, any audible effects keep within reasonable limits.
However, if any gaps resulting from manufacturing tolerances allow for a twisting motion, the
bridge-T will wipe out string vibrations. This "epoch-making bridge" "was not designed for
solid body guitars like the Les Paul models [Berger / Perrius]", but for acoustic (!) arch-top
guitars – which renders the whole thing even more problematic, because these guitars should
be able to reproduce high-frequency partials as well, shouldn’t they? Anyway, around 1954,
this contraption-of-the-century found its way onto the Les Paul Custom, and from then on
there was no holding back any more – in the true sense of the term.
The force introduced by the string to the bridge saddle (as far as it does not serve to move
mass) is handed over via undefined paths to the bridge base that, true to its name, crosses the
distance from one “shore” to the other. To the mechanical engineer, the arrangement is a
cantilever supported on both ends. The lowest Eigen-frequency of this cantilever can be
calculated using area moment of inertia, geometry of the cantilever, E-modulus and density; it
turns out to be 1.6 kHz. Although the bridge has immobile support at both its ends (bearings),
it can still vibrate in between; and if it receives excitation to do that, it will dampen the string
vibration. Unfortunately, it does receive this excitation, and it is just that string delivering it
that should be given an immobile bearing by the bridge. In order to not just theoretically
calculate this friendliness towards vibration, measurements were taken with an ABR-1 bridge
positioned on a stone table (Fig. 7.106). At 1.6 kHz, the conductance rises to almost 0.2 s/kg,
reducing the degree of reflection for the E2-string to below 60%. This means that more than
40% of the vibration energy is absorbed for each reflection! We should not universally
condemn such a behavior because only the effect of absorption will enable the luthier to
individually influence the string vibration; however, if this was supposed to emerge as
McCarty-specific, then this absorption would have to be of a fixed value for all guitars of the
same build. That, however, is not the case, since already a slight shift of the bridge (which on
top of everything is adjustable in height, as well) changes its damping parameters.
Fig. 7.106: Conductance (left) and reflection coefficient (power) of the ABR-1 positioned on the stone table.
The measurements related to Fig. 7.106 were done with an ABR-1 bridge that was positioned
via two knurled nuts directly on a stone table. On the Les Paul, the two disc-shaped knurled
nuts conduct the bridge-force to two threaded posts that protrude perpendicularly out of the
guitar body. The mechanical input impedance (Z3 in Fig. 7.100♣) between bridge and knurled
nut depends on the surface quality (burrs!), and on the angle between bridge and post; when
adjusting the bridge height, the contact surface changes, and with it stiffness and damping,
and thus also frequency and Q-factor of the resonance.
♣
Fig. 7.100 models the bridge as a discrete mass and does not (yet) consider any Eigen-modes.

As an example for differences related to manufacturing tolerances, the decay times for partials
were analyzed for a Gibson ES-335 (new strings, Fig. 7.107). This semi-hollowbody guitar
dating from 1968 sports the bridge shown in Fig. 7.104, across which the strings run to a
trapeze tailpiece. The guitar must not be blamed for showing a few minima in the decay times
– a bit of individuality certainly is okay. That the decay times change as the bridge is shifted
by one or two 10th‘s of a millimeter – that’s owed to the adjustability. The decay curves will
again change somewhat as one seeks to correct a slightly-off intonation with the bridge-piece
screws. Guitar gurus, near and far, you who claim to hear with your golden ears the smallest
detail in the wood-composition, you who even put your guitars in the freezer to get a few
more cracks in the varnish so that the guitar body at last is “freed to vibrate” and the sound
“blossoms”, do see the signs: the one who, in “specialist” magazines, every month propagates
bullshit, may finally drown in the same.
Fig. 7.107: Decay times of partials, ES-335; bridge shifted within the scope of manufacturing clearances.
The grey area indicates the theoretical maximum T30-decay due to internal & radiation damping (Chapter 7.7.2).
Since Gibson’s ABR-1 bridge is height adjustable, it has some horizontal clearance, as well.
Shifting the bridge within the scope of this bearing clearance changes bearing and damping
parameters. Still, there was not much attention paid to this sensitive contact surface: in the
Gibson bridge shown in Fig. 7.108, two burrs influence the surface between bridge and the
knurled nut below it; these burrs co-determine the contact stiffness. To vindicate Gibson, it
may be noted that not all bridges show such a dismal production quality. However, even a
specimen bought for much money in 2010 had not ever come into any contact with a
deburrer. That can only lead to the assumption that Gibson does not attribute much
significance to this contact point. That, however, makes the little damping peaks that the
guitar body itself generates finally loose all relevance. The holy wood, with all the entwined
myths – can it be nothing but hype? Yes, it may and must be seen that way, because why
would Gibson manufacture the most important link between string and the “holy grail” so
sloppily, so wobbly, so unreliably, if the guitar body were important at all? What remains is
the insight that while all those little peaks can be measured, they barely have any influence on
the sound. That’s irrespective of whether they result from the sloppily manufactured burr, or
from the wood seasoned for decades. Goof rules.
Fig. 7.108: Gibson ABR-1 bridge with burr.

The belief in the Holy Grail is deep-seated – deep enough that instructions for a conversion
are found in G&B 6/2010 that are supposed to turn a “halfway” Holy Grail into a “real” one: a
‘52 Goldtop into a ‘59 Les Paul. Too bad, though, that due the rather special neck angle, the
ABR-1 bridge does not fit. Therefore a band sander (Fig. 7.109) is called into action … which
is a physically correct step if the focus is merely on the playability.
Fig. 7.109: ABR-1. Left the model from G&B 6/2010, center/right the copy pre/post milling.
Playability, however, is not the only aim of the conversion: in the end the golden one is
supposed to sound like a 1959 Les Paul. The same sound – although its sanctum, the link
between strings and wood, has been brutally abused with a belt sander? Let us take a look at
the conductance of an original ABR-1 bridge (Fig. 7.110) – not the specimen analyzed in Fig.
7.110 but the one bought in 2010. Since professional tooling was available, the bridge was not
“minimized” with a belt-sander, but via a high-tech milling machine. The rigorous thinning-
out reduces the mass and in particular the stiffness such that the main resonance finally comes
down from 1400 to 850 Hz. This gives rise to the question whether a guitar fitted with this
bridge can ever sound like one fitted with a bridge developing its maximum genuine
absorption at about 1500 Hz?
Fig. 7.110: conductance of the ABR-1 bridge; original condition (left), after milling (right). Stone table; knurled
screws; bridge without bridge saddles; measuring point between the D- and G-guide-slots.
The alterations described in G&B were performed by a well-known and well-respected luthier
having many years of experience under his belt, and therefore we are not willing to simply
discard his evaluation (“same sound”) as unqualified – there must be something to it. And so
the only conclusion can be: if such big differences in the absorption have next to no effect on
the subjectively perceived sound, then the body wood (with much lower conductance values)
has even much less influence. It may be emphasized again and again by some that this Holy
Grail cannot be topped in view of the length of time it has had for being played-in (56 or 57
years), but the reasons for that must reside rather more in the metaphysical realm. Which is
where a grail is best kept, anyway.
Fig. 7.110 documents the changes in conductance caused by the milling. It is obvious that the
decay times of the partials will be influenced, as well, but proper proof is still required. Both
variants of the bridge were therefore tested on a Les Paul (R9), with the decay times of the
partials of the D- and G-string being analyzed for the original bridge and the milled bridge.

Fig. 7.111 shows the corresponding results. Since the bridges are not positioned on a stone
table anymore but reside, as intended, on a guitar, the resonance peaks measured in Fig. 7.110
shift a bit. Moreover, the manufacturing tolerances of the bridge saddles come into play now,
and these are (benevolently speaking) of an abysmal quality. Three of the bridge saddles fall
off if the retaining wire is removed while the other three are so stuck that they can hardly be
moved at all via the adjustment screws. This was not an effect of the milling process but the
original condition of the bridge. Whatever, after the measurements are concluded, this piece
of junk will be discarded anyway. What remains as a result: the main resonance – having
shifted to 850 Hz – will clearly cause shorter decay times of the partials in that frequency
range for both strings … which apparently does not harm the rating “same sound”, though.
This implies that the instructions to modify the guitar hold a contradiction: if the minimal
influence that the “holy” wood has on the string vibrations is taken to be crucial, then the
differences caused by the modifications in the bridge needs to be held as existential feat, and
there should be no talk at all that here the ‘59-sound has been created (simply put, this is the
Holy Grail, G&B 8/2010). If, on the other hand, the bridge resonances are taken to be of
insignificant effect, then the microscopic effect of the wood should be assumed to be
irrelevant. In that case, however, any plank from the DIY-store would have sufficed, too ...
Fig. 7.111: Les Paul R9, decay times of the partials. Original bridge (---), milled bridge (–––).
Of course, a recurring thought has been “it’s all in the fingers”, and so we can indeed just use
the guitar according to its purpose: to be played. Yes, that is possible on most guitars,
irrespective of where exactly the bridge is positioned.
After so much theory, how about a bit of some more “specialist press”? Here we go (all
statements made by one and the same author): Kluson machine heads are also very
lightweight. The small mass can be easily excited to vibrate but it decays all the more quickly.
Theoretically, that would mean that Kluson machine heads give less sustain [G&B 8/05].
Small changes in the height of the stop-tailpiece in part drastically change the sound (of the
Les Paul) [G&B 7/05]. Machine heads (about $ 1200) and stop-tailpiece (about $ 2000) had
only very little influence on the sound (of the Les Paul)[G&B 2/07]. Of course, build and
material (of the stop-tailpiece) have an important influence on the vibration transmission to
the body [G&B 7/05]. Hard to believe that simply swapping the machine heads (on the Les
Paul) could lead to such (sound-) changes. [G&B 8/05]. Sometimes, I find it inappropriate
how self-proclaimed equipment-missionaries roam about seeking to convert everyone to the
true belief [G&B 8/07].

7.10.2.2 Leo Fender’s Telecaster

The Telecaster was Leo Fender’s first “true” electric guitar. To start with, it was designated
Esquire, then Broadcaster, and finally Telecaster [Duchossoir]. According to Fender’s patent
application US 2,573,254, the string length was to be individually adjustable – but that is only
possible in pairs of two: two strings have to share a cylindrical bridge saddle (Fig. 7.112).
Compared to the non-adjustable bridges customary until then, that definitely represented an
improvement although it still was a compromise. Fender however already points to a further
development: the bridge saddles are drilled through at an angle. Each of the three bridge
saddles may be adjusted in height using two setscrews, and a long tensioning screw takes care
of the intonation adjustment. A thick steel plate anchored with 4 large bolts in the guitar body
serves as a base for the setscrews and the tensioning screws. The strings run across the bridge
saddles through the guitar body to fastening bushings mounted from the rear of the body.
Fig. 7.112: The Telecaster bridge. [Duchossoir]
What has been said in Chapter 7.10.1 holds for the setscrews and tensioning screws – their
transmission stiffness depends on contact surfaces and forces. In principle, this bridge does
work. It may, however, develop an idiosyncrasy that helps this guitar to achieve a special
status: the steel plate rests in an undefined manner on the guitar body, and its resonances
(Eigen-modes) may re-act on the strings – not necessarily, but possibly. Using a hard
nonmagnetic item to knock on the upward-bent flanks of the plate, we hear a clicking noise
coming out of the amp/speaker. The sheet metal is not comprehensively damped by the body
wood below it, but can resonate with its natural frequencies at a high Q-factor. Mechanical
reactions from sheet metal to bridge saddles are possible, and – given steel as material – also
inductive coupling to the bridge pickup. Generally, the sheet metal is electrically conductive
and thus a place where eddy currents circling the pickup may roam (Chapter 9.5).
The necessity to make the string action adjustable was not only connected to the drive of all
guitar players to make each new guitar “playable” according to one’s own ideas. It was also
unavoidable in view of the separation of guitar body and (bolt-on) neck into two individual
production entities each subject to manufacturing tolerances. From Duchossior’s close-up
pictures it can be seen that these adjustment possibilities were indeed put to use, and that the
bend-angles that the strings form as they run across the bridge saddles are specific for each
individual guitar (they are string-specific in any case). However, this means that the vibration
characteristics of the bridge are specific to each individual guitar, too.
Fig. 7.113 shows the decomposition of forces at the bridge saddle. The string-tension force Ψ
is almost the same on both sides of the bearing cylinder (bridge saddle), and the frictional
force may be neglected in a first approximation. The force F acting towards the lower left has
two components. The setscrew just resting on the surface below can only take on the vertical
component Fy; the horizontal component Fx is taken care of by the tension screw.
Nevertheless, Fender’s patent application shows a set screw mounted at an angle (Fig. 7.113)
Fig. 7.113: String with bearing cylinder (left),
force resulting from the string forces (center),
decomposition of the resulting force in
horizontal and vertical component (right).

We shouldn’t expect too much here: "Leo Fender was an ingenious, resourceful technician,
but – as it is frequently reported – he had not even had formal training as an engineer, and he
certainly was not a guitarist – couldn’t even tune the instrument" [G&B Fender special
edition]. Duchossoir’s citation is even more merciless: "Leo had a knack of thinking slowly
and consecutively – no flashes of genius – a merciless unstoppable slow degree of thinking
[Tavares]." At some point, the angled screws raised some eyebrows, and one day they stood
upright (Fig. 7.112). More specifically: in the patent application [USPTO.gov], the angle
between setscrew and tension screw amounts to 70°, in later bridges it is 90°. What is better?
The mechanical engineer would probably prefer the setscrew perpendicularly positioned on
the base plate, because it can transmit only vertical forces in any case (horizontally, only
small frictional forces remain). With the screw positioned at an angle, a bending moment
results that loads cylinder and tension screw flexurally, while with a perpendicular setscrew
there is merely a tensile force acting on the tension screw. What in fact prohibits sideways
motion of the latter? This would be a motion within a borehole in which – according to the
patent publication – the screw should be borne "sufficiently loosely"! An additional brace
could make for more stability but the effect would probably not be very dramatic. Also, an
axial force applied to a setscrew could possibly readjust the screw over time – therefore the
perpendicularly oriented screw may offer slight advantages. These are, however, untried
speculations for which no additional experiments were done. With
, , ,
we can see the angle dependency of the forces; the bend-angle of the strings α amounts to
about 25° to 50°. The tension screw has the sizeable length of 32 mm – apparently indeed
necessary to allow for a sufficient adjustment range. With Ψ = 50 N (certainly possible for
thin strings), Fx amounts to a minimum of 4.7 N, and Fy amounts to a minimum of 21 N♣.
For heavy strings, Fx = 50 N and Fy = 100 N are possible, as well. That is quite a respectable
degree of variability in the compression force, and correspondingly large will be the
differences in the contact-stiffnesses and –impedances. Which tilting, rotating or wobbling
motions the bridge saddle will be subjected to under real deployment conditions cannot be
anticipated with a general consideration – the conditions vary too much. The offset-force
acting on the setscrew presumably is so strong that this screw can a priori not be suspected as
a “vibration killer”. A longitudinal force of merely 4.7 N is scant, but then there are 2 strings
pulling at one screw. Within the string, however, also longitudinal vibrations appear
(dilatational waves) that could excite the bridge saddle to rotational vibrations. In that case,
too, much slack between the screw and the bridge saddle would be counterproductive.
Fig. 7.114: Forces on the guitar body
The interface from the base plate to the guitar body is shown in Fig. 7.114. The sum of the
two string forces generates a resulting force pointing towards the lower left, just missing the
wood screw and thus resulting in a torque around the bearing point (circle). The main share of
the retention force occurs at the screw; in order to compensate for the torque, a supplementary
force F is necessary (here sketched in arbitrarily).
♣
A mass of 1 kg generates a weight force of 9.8 N (1 N = 1 Newton → 102 Gramm).

Where exactly which forces act cannot be specified, because how the bridge rests on the body
wood remains undefined: this depends i.a. on the curvature of both components and materials.
The torque designated with M rises with increasing string diameter and increasing string
height (distance between string and base plate). If there is (at F in Fig. 7.114) a tiny gap left of
the pickup, half of the base plate is suspended in mid-air … opening un-dreamt of possibilities
of vibration.
Also undefined is which of the four wood screws bears the main load – they just somehow
share the retention forces. If the guitar body as a vibrating system were to be coupled in a
defined manner to the bridge (or the bridge to the body), a completely different design would
be required. No, this ain’t no sound-design – it was simply a matter of bolting a base plate to a
wooden board – over and done! As Fig. 7.115 shows, the arrangement can in fact work pretty
decently: here we see the decay times of the partials compared to the situation with a
Stratocaster. Both guitars were measured with brand-new strings, although the diameters were
not completely the same (Tele: 009 – 046, Strat 010 – 046). For the Tele, the decay is slightly
faster, and it depends a bit more on the frequency. Before anyone starts to derive the general
verdict that a Tele would have a shorter sustain than a Strat, let’s be reminded that what we
have here are individual results, measured merely with one single representative of its
species♣.
Note: in Fig. 7.116, the grey area indicates the theoretical maximum T30-decay due to internal
& radiation damping (Chapter 7.7.2).
If we would want to extract Telecaster-typical characteristics, we would first have to define

what a typical Telecaster in fact is: over the decades, Fender changed the headstock, the neck,
the body, the pickups, the bridge – it was only the body shape that approximately remained
the same: consequently, there is not “the” Telecaster. For most variants, the bridge does have
the base plate of about 85x74 mm2, but differences start already with the bearing-cylinders:
thick, thin, made from brass, or from steel, with/without groove, with/without thread. From
the 1970s on there is also a version with small or large Strat-like individual bridge saddles, or
even a pure-bred Stratocaster bridge. Telecaster-typical remains apparently merely the body
shape but that has next to no influence on the sound. Even if we limit ourselves to the single-
coil-fitted original type, we find a multitude of different variants: 250-kΩ- or 1-MΩ-pot,
bridge pickup impedances between 5.5 – 11 kΩ, (complete) solid body or (half the weight)
Thinline body, bolt-on neck, tilt-neck, set neck [more info in Duchossoir]. If the pickup cover
is the secret of the neck pickup, why then does Fender include a different pickup in the
Thinline-Telecaster (2nd version), the Tele Plus, the Elite Telecaster, the Telecaster Deluxe
and the Custom-II? Why are there also Lace and Seymour Duncan variants on top of the
Fender version? Presumably, that is so that each guitarist can realize his/her personal idea of
the Telecaster sound.
In http://www.tdpri.com/forum/telecaster-discussion-forum/77808-new-body-material-build-w-sound-clip.html,
Terry Downs presents his new guitar, and lets the congregations of fans guess which material
the body is made from. Everybody enthuses about the sound, and conjecture includes: oak,
masonite, teak, cork, semi-hollow-body, synthetic counter top material, soy, hedge apple tree,
and others – most guesses meant seriously. In fact, it was three medium density fiberboards
that were bolted on top of each other – that’s it. Result: Sounds like a Tele – what else.
♣
The multitude of limitations in the framework of university operations unfortunately does not make more
comprehensive investigations possible.

Fig. 7.115: Decay times of partials, Tele (009/046 set, left), Strat (010/046 set, right). “H3” = B(-string).

7.10.3 Bridges with vibrato
A precursor to the electric (spanish) guitar was the steel guitar (lap steel). Its notes can be
generated at any arbitrary pitch without the discretization forced due to frets. This feature was
attractive to the country musicians (and/or their audience), and it was offered by Bigsby,
Kaufmann, Fender, and Co., by mounting vibrato- (or tremolo-) systems on their first electric
guitars. These gentlemen could not have had any inkling that years later someone would use
the device to interpret the “Star Spangled Banner” in quite a peculiar way, and even less of a
foreboding that somebody else would at some point measure bridge conductances – they were
fully absorbed in inventing a bridge that at the same time was steadfast and moveable …
steadfast regarding the string vibrations, and moveable to achieve the “tremolo♣".
To change the string pitch in a continuous manner, the tension force Ψ of the string needs to
be changed; this is done via changing the strain (i.e. the length). Accordingly, the tailpiece
has a moveable, resilient bearing: pressing (or raising) the vibrato lever changes the string
bearing and thus the strain (Fig. 7.117.
Fig. 7.117: Vibrato-system

The basic problem of all vibrato systems is the tuning instability caused by inevitable
frictional forces. Pulling the vibrato level upwards and releasing it leads to a different tuning
compared to pushing it down and releasing it. The friction forces are not particularly strong,
but to achieve a pitch error of less than 5 cents, the frequency would have to be correct by
0.3%. In vibration engineering, we like to work with friction forces that are proportional to
the particle velocity, because they allow for setting up linear systems. However, reality has in
store also the Coulomb friction, and that is of non-linear character. For the Coulomb friction,
the friction force depends solely on the normal force and the friction coefficient µ, but not on
the particle velocity. There is, however, a distinction in the friction coefficient between static
friction and dynamic friction; as such the coefficient is movement-dependent, after all – but
rather in a non-linear fashion.
If the string runs around a fixed cylinder with an encirclement-angle α, the two tensile forces
differ by . Pulling to the right (in the figure), the right-hand force is (at
the max) larger by this value; pulling to the left, the left-hand force is. In conjunction with the
radius of the cylinder, this force difference generates the friction torque M = ΔF⋅R, which is
absorbed by the bridge. The friction is only small if the cylinder can rotate – but easily
rotatable, loose rollers do not make for an ideal guitar bridge. As an alternative, bridges with a
knife-edge or point bearing have been invented, but these can also only work properly if all
strings have the same distance to the axis. That, however, is not the case if the bridge is set
on top of the guitar body. It is the case approximately, if the pivot is moved into the guitar
body. If the residual string (from bridge to the tailpiece) is long, and if the bend-angle is small
(such as it is on the Jazzmaster), again other problems result – it’s simply not an easy job. In
the end, some creative thinking indeed led to usable results, as long as the involved guitarist
limited him/herself to moderate pitch changes. For those operating with brute force, further
developments came later, such as the clamped-string approach.
♣
Kauffman and Fender designated the frequency vibrato with the (not really correct) term "Tremolo"

7.10.3.1 Fender’s Stratocaster vibrato (aka tremolo)

On August 19, 1929 – when few people were thinking about electric guitars – Clayton
Kauffman filed for a patent under the title "apparatus for producing tremolo effects" (US
1,839,395). According to it, a spring-loaded, movable tailpiece enabled the change in pitch,
"so as to produce a tremolo effect". Indeed, it was this "Doc" Kauffman who later was Leo
Fender’s business partner for a short time in the jointly operated K&F company, before
Fender started his "Fender Electric Instrument Company" in 1946 [Duchossoir]. The latter’s
first electric guitar, the Esquire, successfully entered the market around 1952, and then had
weathered the metamorphosis into the Telecaster. Time was right for the release of a further
guitar: "We didn't invent the tremolo thing. It had been used on many other instruments, but
we wanted it because it seemed to be very saleable [Tavares]". On August 30, 1954, Leo
Fender filed for a patent for the Stratocaster (US 2,741,146), an electric guitar with a
"synchronized tremolo". Duchossoir describes the first experiments: "the first vibrato
designed by Leo Fender was, by all accounts, fairly similar to the unit later installed on the
Jazzmaster guitar released in June 1958. It allowed some string length between the bridge
and the tailpiece, were the strings were anchored. This early version was fitted with
individual roller bearings, meant to facilitate return to pitch, but in fact they were damping
the string sustain because of too much lateral vibration. It would also appear that the steel
rod used as a tailpiece did not anchor the strings firmly enough and their energy was
dissipating to the detriment of tone and sustain." Leo Fender comments: "We had to chunk the
whole thing and completely retool". And: "With a string, you can't have vibration in any
direction at the bridge, it's got to be as solid as the Rock of Gibraltar". This is stated by Leo
Fender (bookkeeper by education), and darn is he on target. It’s a different story that as late as
2005, the “experts” at Gitarre & Bass opine that the largest part of the string vibration should
be fed to the body.
In order to keep bridge and tailpiece from developing too much of a life of their own, Fender
combines both into a single unit supported on knife edges – that was the groundbreaking idea.
Why he deviates again from it in the Jazzmaster remains Fender’s secret. Fig. 7.118 shows a
cross-section through the Stratocaster vibrato. The strings run across adjustable bridge saddles
to a so-called “sustain block” fitted with tension springs at its lower side that provide the
counter-traction. The L-shaped base plate is held in place by 6 wood screws that are not fully
bolted down such that the base plate can easily be tilted upwards. The rotational axis is
located between wood screw and slightly countersunk hole in the base plate. The traction
force Ψ generated by the strings (at the time about 730 N) causes a torque at the short lever
(about 9 mm) that is compensated by 5 tension springs at the long lever (about 42 mm).
Today, lighter strings are customary and often only 3 springs are used. Their exact traction
force may be adjusted via two tension screws (not shown in the figure). The pronounced bend
angle with which the strings run across the bridge saddles causes relatively high contact
pressure forces, and any residual damping due to the short residual string section (Chapter
7.7.4.3) is weak. Nothing is perfect, now even this design, but it works well enough that to
date Fender has only introduced small changes.
Fig. 7.118: The Stratocaster vibrato.

One of these changes concerned the mounting screws: how does a force distribute itself across
6 screws? In an undefined manner! And so the 6 mounting screws were reduced to two in
1987 for the American Standard Stratocaster, which resulted in a reasonably unambiguous
knife-edge bearing, at last. The second change concerned the bridge saddles: originally
shaped from sheet metal, they became die-cast cuboids in the 1970’s. Not on all models,
though: some were still produced with sheet-metal bridge saddles. Both versions do work –
however, they have their special manufacturing tolerances. Depending on circumstances,
every bridge saddle is one of a kind with the 3 screws at each end giving it highly individual
contact-stiffnesses and -damping.
Fig. 7.119: Flux of force in the Stratocaster bridge.
The flux of force in a bridge saddle is shown in Fig. 7.119: the bent string exerts a force
pointing downward to the left at an angle onto the bridge saddle (compare to Fig. 7.113).
Since string diameter, and position and angle of the bridge saddle vary, the amount and the
direction of this force vary, as well: its x-component can range from 5 to 35 N, its y-
component from 21 to 73 N. The y-component is absorbed by the lower side of the vertical
adjustment screw1, and the x-component is absorbed by the horizontal tension screw. Since,
however, the correspondingly parallel forces do not run through the same point, two torques
will result – designated Mx and My here. As a rule, these torques will not be of the same
magnitude which is why the small vertical force FM needs to additionally act on the tension
screws. Given the usual geometry, this force will be directed downward (in the figure) and
finds its counterforce (not indicated) at the vertical adjustment screw. The larger FM is, the
more the horizontal tension screw braces itself into the thread of the bridge saddle, and the
more this connection becomes solid. Thus: the smaller FM is, more wobbly the arrangement.
FM becomes small if the string runs across the bridge saddle at a small bend angle. This is, at
the same time, the scenario in which the other forces become small and in which only small
relative movements – which would remove vibration energy from the string – are possible.
Now, the users of Strats are not exactly know for constantly complaining about un-playability
and lack of sustain – for the majority of these guitars, the adjustability of the bridge saddles
does not need to be exploited to the limit, and most bridge saddles offer a secure footing to
the string. If the bridge saddle is moved back so far that the string experiences another bend at
the oblong hole, adequate retention forces can be expected also for light strings. Problems can
result only for guitars with such an unfavorable neck fitting that the bridge saddle needs to be
positioned at the furthest front end (i.e. the beginning) of the tension screw. Still, when
comparing this to the jiggle existing on the Jazzdesaster (Chapter 7.10.3.2), even Fx = 21 N
could still be called rock-steady.
When dealing with a vibrato system, the main questions always are: how stable is the tuning,
and how large is the possible detuning? In this respect, the Stratocaster vibrato offers an
acceptable performance, with some potential for improvement. The effect of the vibrato is,
however, not limited to the above main functions, and therefore we will in passing look at
some side-effects: the tension spring located within the guitar body vibrate close to the bridge
pickup and induce electrical voltages, and moreover the sustain block with all the springs
constitutes a resonance system.
1
Friction forces are disregarded for his simplified consideration.

6 steel strings are positioned above the bridge pickup of the Stratocaster, und 5 steel springs
below it (today, often there may be merely 3 of them). Normally, the steel springs are
concealed but that does not keep them from having an inductive effect – and one that is only
bearable because they are further away from the pickup than the strings. Each of the springs
can adopt longitudinal, transverse, and rotational vibrations, and will do so, too, as soon as
strings and/or guitar body are set in motion. Apparently, this latent life of its own is not
entirely undesirable but is seen as a kind of Strat-typical reverb system (although there are
also guitars with the vibrato springs wrapped in a soft cloth to reduce just that effect). A
reverb in the usual sense must, however, not be expected because this system features merely
a few pronounced resonances. The investigated Strat-specimen (010-gage string set, 3
springs) showed a 47-Hz-resonance that also prominently manifested itself as a line in the
pickup spectrum. This is the Eigen-frequency (natural frequency) of the vibrato arrangement,
composed of the stiffness of strings and springs, and (mainly) the mass of the steel block.
Eigen-vibrations of the springs appear around 140 Hz, and at harmonics thereof. The resilient
string bearing makes itself felt as selective absorption in the bridge conductance at a
frequency range around 500 Hz – however, this effect is not very pronounced.
The following table shows orientation values for string tension, string strain, and longitudinal
string stiffness, for a 009-set, and for a 010-set of strings. As the vibrato lever is operated, it
needs to act against the sum of all string stiffnesses plus the spring stiffnesses.
Diameter 9 11 16 24 32 42 mil
Tension force 59 50 66 75 75 72 N
Strain 4.8 2.7 1.7 3.8 2.4 1.6 mm
Stiffness 12.3 18.5 39 20 31 45 N/mm
Diameter 12 16 24 32 42 53 mil
Tension force 105 105 133 133 130 116 N
Strain 4.8 2.7 1.7 3.8 2.4 1.6 mm
Stiffness 22 39 78 35 54 73 N/mm
Table: String diameter, string tension force, string strain, and longitudinal stiffness of string.

7.10.3.2 Fender’s Jazzmaster vibrato (aka. tremolo)

He did give it another try … according to Duchossoir, Leo Fender had already sought the
separation of bridge and tailpiece in the Stratocaster, but it did not work out in that first
attempt. Once more into the breach, then: in 1958, the Jazzmaster was presented, offering a
"floating tremolo with a floating bridge" based on a tailpiece-bearing on a knife edge, and a
bridge set onto two pins (Fig. 7.120). The 6 bridge saddles (short, threaded rods) sat in a u-
shaped rail that itself was positioned on two pointed posts. As we operate the vibrato lever
(we do call it that, dear Leo, because it is – after all – not a tremolo that we achieve) the
strings do not need to slide (with much friction) across bridge saddles, but rather the whole
bridge tilts back and forth on the very-low-friction steel points. The inner diameter of the
bushing is slightly larger than the diameter of the posts and allows for a shift of the bridge of
about ±1 mm. That is enough for moderate pitch changes – they do primarily not depend on
the length variation of the string but on the strain variation!
Fig. 7.120: Vibrato system of the Jazzmaster.
The main issue with the Jazzmaster vibrato system is that the strings bend across the bridge
saddles with a very the shallow angle (6 – 7°). As late as 1968, 10 years after the introduction
of the Jazzmaster, the Fender catalog specifies 012-strings as factory fit; and it was
presumably this string gage with which Leo Fender optimized his guitars. For a set of 012-
strings, the tension force of the E4-string amounts to 105 N, for a 009-set it is 59 N, and for a
008-set it is a mere 47 N. This results in a string pressure at the bridge of Fy = 5.2 – 12 N, and
a force at each of the two vertical adjustment screws of 2.6 – 6 N (the thinner the string, the
smaller the forces become). The longitudinal force resulting from the bend amounts to only Fx
= 0.3 – 0.7 N i.e. it is barely existent at all. This force should not be pronounced, too, because
it can only be absorbed via the string friction as the bridge “floats”. To keep the bridge
saddles from longitudinally resting on the bridge in a totally undefined manner, Leo Fender
fitted them each with a coil spring – but this generated only a weak tension in the case of the
treble strings. For the bass strings, the coil springs got in the way of perfect intonation plus
they had to be shortened, presumably killing off many a precision wire cutter.
Maybe this guitar (just like the Jaguar fitted with the same bridge) was reasonably playable
with 012-strings, but with the increasingly popular light gauge strings, problems mounted,
and the success on the market failed to materialize. Jazz players did not want to change, and
all others already had the Stratocaster and the Telecaster if they opted for buying a Fender.
Dutifully, the promo-department had exaggerated: Fender's famous Jaguar guitar is the
standard of solid body excellence on today's musical market. This exceptional instrument
incorporates Fender features offering playing versatility unmatched by any other. Well…
Hendrix did not burn his Strat at Monterey out of frustration, only to change over to the
mentioned “standard” with flying colors, did he? Some sources say that he was seen with a
Jazzmaster initially … but only for a short time, and from 1966, the Strat was it for him.

7.10.3.3 Paul Bigsby’s vibrato

As some kind of Renaissance man about town, Paul Bigsby repaired and invented devices of
all kinds. Around 1947, he also built a few electric guitars (e.g. for Merle Travis). His real
claim to fame, however, was his vibrato system that was deployed on many early guitars. The
strings were hooked into a rotatable shaft, with the counter-torque being delivered by a
spring-loaded lever. Allegedly, it was a spring taken form a Harley – an obvious choice for
motorcycle mechanic Bigsby. The vibrato system shown left in Fig. 7.121 is one from a
Gretsch Tennessean built around 1960. Here, the bridge merely consists of a solid metal
cylinder that can be adjusted in height via screw and threaded nut – there were however also
other bridge designs (aluminum wedge, roller-bridge).
Fig. 7.121: Left: side view of the Bigsby vibrato. Right: different
variants [Rockinger Guitars].
The Gretsch Tennessean is a hollow guitar without any sustain block; its thin top cannot take
any large forces. Maybe the bend angle of the strings must in fact not be more than 4° (as it
showed up on the investigated guitar), maybe more could be allowed … we cannot find out
using a non-destructive approach. At least the strings do rest on a solid steel cylinder and not
on jittery bridge saddles. For those who like to use thin strings and can do without the rather
instable vibrato system: replace the vibrato shaft by a cylinder, drill 6 holes through it and
insert the string through the holes. This increases the bend angle, and the bearing forces reach
about the value they had with the factory-supplied strings. All that is at your own risk, of
course.
For guitars that are able to withstand larger forces on their tops, the Bigsby was (or is) also
available with an additional pinch roller increasing the bearing forces but also the disruptive
frictional forces (shown on the right of the figure).
The bridge in the form of a cylinder (of a diameter of originally 13 mm, later 9.5 mm) acts as
non-linear bearing because the string experiences a shortening as it vibrates towards the guitar
body. This effect is, however, not strong; compared to a sitar, the cylinder radius is small
[Burridge et al. 1982: The sitar string, SIAM J. Appl. Math. 42, 1231 – 1251].

7.10.3.4 The Rickenbacker vibrato

According to GRUHN'S GUIDE TO VINTAGE GUITARS, as early as 1932 an electric
Rickenbacker guitar was built with a Kauffmann vibrato – that’s 20 years ahead of the
Stratocaster, after all. The version described here is, however, not this archetypical guitar but
a later variant from the golden 1960’s, when the Byrds, the Beatles and the Who helped to
create a short period of blossoming of the Rickenbacker tulips. To be specific: it’s a model
Nr. 335 from 1966. The bridge consists of a u-shaped rail open to the top in which standing
“forks” can be shifted back and forth via adjustment screws. In a recess, the forks carry a
small roller on which the string rests. The whole thing is tightened up in such a remarkably
rigid fashion (at least it is on the investigated guitar) that even the rollers cannot be moved
(anymore?). So, is this the perfect bridge? Well, there are 4 screws inserted through the U-
shaped rail; they rest on a metal plate (Fig. 7.122). With 3 screws, we would achieve a
defined bearing but with 4 screws the situation remains undefined. The height of the bridge
needs to be very carefully adjusted so that all 4 screws transmit approximately the same force
– and then we need to hope that this adjustment never changes again. If we moreover mount
heavy strings and take the vibrato arm off …
Fig. 7.122: Side view of the Rickenbacker vibrato (1960’s vintage).
The spring-loaded tailpiece rivals Fender’s ideas when it comes to ingenious simplicity: a u-
shaped sheet metal into which 2 further sheets are hooked – done. The vibrato lever serves to
bend the u more closed or more open, and changes the string tension that way. Once the
strings have been inserted into the tailpiece, the latter for starters won’t cause any problems.
The latter may, however, occur at the bridge: first because the bearing there is undefined, and
second because the bend angle of the string is, at 5°, even smaller than that on the Jazzmaster.
It should be noted when considering these numbers that they are measurements on individual
guitars; any production tolerances from the 1960’s were not looked into.
The Rickenbacker 335 is not a solid body guitar but has a hollow body with a 4 mm strong,
vibration-happy top. Compared to a Les Paul, this “semi-acoustic” build leads to higher
conductance values and thus to a stronger damping of partials (Chapter 7.11). However, much
faith in a well thought out vibration design is not coming our way: the top is stabilized on its
lower side with a rather archaic cross-bracing, but then a ½“-cutter was used to mill slots into
the top for the pickups – with the cutter taking no prisoners and clearing its way through part
of the bracing, as well. Of course: pickups have first priority in the electric 6-string. What’s in
the way gets removed.

7.11 Solid-body vs. hollow-body 7-139
7.11 Solid Body vs. Hollow Body
In order to be able to radiate sound as well as possible, the archetypical guitar featured a
hollow body sporting a thin top. Early protagonists of electrification tried to sense the
vibration of the top using pickups of record players, but Adolph Rickenbacker, Paul Bigsby,
Les Paul, and Leo Fender (to name but a few) soon realized that the sound-amplifying effect
of a hollow body can be dispensed with as soon as the loudspeaker takes over. Enter the solid
body guitar. Its body consisted of a solid board (or several boards glued together) of an
overall thickness of about 5 cm, and it was not hollow anymore but solid (hence the name).
However, not all electric guitars operate according to this principle, there have been (and still
are) several variants:
• The “electrified” acoustic guitar, that received pickups merely as an add-on,
• The hollow semi-acoustic guitar,
• The semi-acoustic guitar fitted with a sustain block (semi-solid guitar),
• The solid guitar (solid body).
The electrified acoustic guitar (having a “full resonance”) has a hollow body of about 12 cm
thickness and includes 1 to 3 magnetic pickups. N.B.: alternatively, it may feature a piezo
pickup stuck to the top; after Charlie Kaman took care of associated groundwork, this pickup
has been banished into the (Ovation-) bridge. Besides these big matrons that are often
lovingly cradled in the arms of Jazz guitarists, we find (heavy) solid-bodies (e.g. Les Paul or
Stratocaster), and in between the more or less hollow ones: semi-solid (e.g. ES-335) and
semi-acoustic (e.g. ES-330).
Fig. 7.123: The four basic types; acoustic, semi-acoustic, semi-solid, and solid-body guitar.
On top of the basic models shown in Fig. 7.123, there are some more intermediate variants
such as the solid body into which more or less extensive cavities have been milled, or the
more or less braced semi-acoustic; and all these with or without sound- (or F-) holes (real or
just painted-on). The bridge finds very solid (sic!) footing with little damping on the body of a
solid-body guitar such that the vibration of the string is determined for the largest part by
attenuation due to air, internal damping, and damping due to the neck♣. Give a freely
vibrating top, things are very different: the bridge placed there is not as immobile as it is on
the solid-body, it yields somewhat to the string excitation and in turn dampens the decaying
oscillation of the string. The determining magnitude here is not just the bridge mass because
any stiffness acting on the bridge will reduce the reactive share of the mass. As a formula:
Spring/mass/damper-system
Combining this equation with the condition for resonance ω2 = s/m, the imaginary (reactive)
parts compensate each other, and only the damping resistance W remains. The active share of
the bridge-admittance Y = 1/Z, i.e, the conductance introduced in Chapter 7.7.4.4, reaches
values of such magnitude in acoustic and semi-acoustic guitars that it becomes significant
relative to other damping mechanisms. Only in guitars of such build has the wood of the body
a more-than-marginal influence on the “electric sound” – only for such guitars is it worth to
investigate the construction of the body more closely.
♣
Other mechanisms of damping are summarized in Chapter 7.7.

Fig. 7.124 juxtaposes the conductance and the decay times of the partials in several guitars,
including the Fender Stratocaster as archetypical solid-body, the Gibson ES-335 as sustain-
block-reinforced semi-solid, the Rickenbacker Nr. 335 as semi-acoustic with a strong top, the
Gretsch Tennessian with a thin top, and the Martin D-45V as purely acoustic guitar. The
measurements were taken with some time lag; smallish differences between the decay curves
and the conductance curves may therefore be possible.
Fig. 7.124: Bridge conductance and decay times of partials (open E2-string).

7.11 Solid-body vs. hollow-body 7-141
Fig. 7.124 shows the Stratocaster-bridge to be relatively immobile; only between 500 Hz and
800 Hz we find any significant maxima in the conductance – these are due to the special
design (spring bearing). The ES-335-bridge, too, is rather on the immobile side, with the
exception of the bending vibration between 1.5 kHz and 2 kHz. The Rickenbacker-bridge
fails to find a truly solid base on the freely vibrating top (only reinforced by a simple X-
bracing), and the quite significant conductance results in a reduction of the decay times up to
about 2 kHz. The Tennessean exhibits an even larger conductance with the thin top vibrating
strongly (and absorbing correspondingly). The Martin D-45V is a pure-bred acoustic guitar
(without pickup), and its top has pronounced low-frequency resonances.
The conductance at the bridge is not the only source of string-damping, but it may become its
main component. However, if the conductance drops to insignificant values (such is the case
for the Stratocaster above 800 Hz), the bridge and the body below it do not contribute much to
the sound (in this range) anymore at all. It has already been elaborated in Chapter 7.7 that the
string moves in circular and longitudinal vibrations, that inner damping and neck-damping
also contribute to the overall damping, and that pick and the attack performed by the player
have a big influence on the sound. Generally evaluating the decay time of the Stratocaster, we
identify three ranges: below 300 Hz there’s neck-absorption, between 500 Hz and 800 Hz,
there’s bridge/spring-absorption, and above 800 Hz, the treble dissipation due to inner
damping occurs. Addressing all those who seek to give extra value to the many individual
peaks: be cautious, since these small peaks change permanently if we merely press the little
finger against the bridge. It has also repeatedly been noted that the decay time shown in Fig.
7.124 can only be measured if brand-new strings are being used. After only 30 min of stage
work, the decay times for the E2 at middle and high frequencies may have dropped off to 1/3rd
or 1/4th!
Compared to the Stratocaster, the other guitars analyzed in Fig. 7.124 reveal shorter decay
times in some ranges – this is i.a. due to the respective bridge- or top-construction. How much
(or how little) the bridge conductance depends on the given bridge of one and the same guitar,
is shown in Fig. 7.125: the Gretsch Tennessean (made in the 1960’s) was fitted with an
aluminum bridge (Rocker bridge), but the cylindrical bridge (straight bar bridge) could be
found in the case, as well. Up to about 1 kHz, the bridge conductances differ only slightly –
for this guitar, the influence of the top dominates in this frequency range. At higher
frequencies, the differences in conductance between the two bridges are more pronounced,
because the cylindrical bridge is a bit less happy to vibrate and therefore dampens slightly
less. However, these conductances have little influence on the string movement because in
this (middle and high) frequency range, the internal damping of the string already dominates.
Fig. 7.125: Gretsch Tennessean, bridge conductance for two different bridges (E2-string);
“Zylinder-Steg” = cylindrical straight-bar bridge; “Alu-Steg” = aluminum Rocker-bridge.

7.12 Vibration – Sound – Tone
Presumably, most guitar players seek to make music rather than solve differential equations
for vibrations. That’s the right spirit – despite all that physics stuff we should remind
ourselves of that main purpose of the electric guitar! Okay, the instrument may also be set on
fire (Monterey et al.), but that brings the service life down to inacceptable levels. How
marvelous when we use it as tone generator: notes and sounds, rapture and ecstasy,
consonance and dissonance, emotion and tinnitus. However, just asking about the tone
generation, we’re back in physics: vibration generates electricity generates airborne sound
generates perceived tone.
The guitar string will carry out vibrations – if you let it. The previous chapters have shown
that, for the magnetic pickup, the vibration velocity of the section of string above the
magnetic pole is significant. The pickup transforms the vibration velocity (the particle
velocity) into an electrical voltage that, if sufficiently amplified, will power the loudspeaker
membrane. The latter in turn excites air particles into vibrations, and these propagate as waves
in the medium of air, and form the sound field. As these sound waves reach the ear, they are
converted into membrane- and lever-vibrations, in the end generating impulses on nerve
fibers and auditory perceptions: tones, in plain language. The term tone is used in a number of
ways: in signal theory, it may designate the sum of many individual harmonic partials♣, while
in auditory psychology, it may indicate any perceived sound. Outside of the realm of
psychoacoustics, however, the tone simply is what science designates as “tone color” or
“sound color”: the guitar has a “throaty”, “chunky”, “singing”, or simply a “hot” tone. How
does it do that?
7.12.1 Linear string vibrations
The source for the pickup voltage is the section of string vibrating over the magnet. Plucking
the string feeds energy to it that then is lost again during the decay process. Friction against
the air and internal friction within the string convert part of the energy into heat while the
remaining part wanders off: via the string bearing (bridge, and nut or fret) into guitar body
and neck. And no, despite what many guitar and bass magazines’ continuously circulate: the
vibration energy should not as much as possible transferred to the guitar body, it should nicely
stay within the string. It has proven to be conducive to expand the string vibration into a
harmonic series (Chapter 8.2.4) i.e. to interpret it as the sum of individual partial sine-tones.
The previous chapters have shown that these partials decay quickly or slowly, depending on
the partial-specific damping mechanism. The tone results from frequency, level and decay
behavior of all partials – that’s easily said but much harder exactly described, because e.g. for
the E2-string, we would need to analyze more than 60 partials that do not simply decay
exponentially. Due to this vast variety of parameters, one may arrive at very different
strategies: we could mistrust “any theory whatsoever”, and plug different guitars into various
amps to conduct listening tests, or we could extract typical parameters from vibration
measurements to synthesize artificial tones. Both approaches have their merits as long as the
experimentation methodology does not contain any grave errors. Unfortunately, many of
those seeking a “practical” approach are of the opinion that one cannot go wrong with
performing listening tests. Rest assured, you can …
♣
The sum of sine-tones of only whole-number frequency relationships may also be termed complex tone.

7.12 Vibration – Sound – Tone 7-143
Here’s a typical “listening experiment”: at a concert, you get to hear Draco Deathbringer
playing his black Gothic Special (the one with the real-blood position markers). The next day,
you visit your large local store down by the river – they advertise just that model (along with
1400 others). You check out the guitar on display and are disappointed beyond measure. The
sales guy has an insiders’ tip: the production model sports only dabs of red varnish, it doesn’t
have the real thing … ah, in that case …! Exaggerated? How about this: you grab a Strat with
an alder body, and one with an ash body, play both extensively, hear differences – you have
discovered the influence of the wood on the tone! That cannot be an exaggeration because
something like it happens daily in editorial offices around the globe. Please listen up, dear
specialist editors: if you want to fathom one single factor of influence, you may only change
one single influencing factor. The same type of string needs to be installed on both guitars to
be compared, yep - brand new ones. Action and pickup-positions need to be identical, and of
course you need to mount one and the same pickup (have fun repeatedly de-installing and re-
mounting it). Because: if you do not use the same (specimen of) pickup, you risk evaluating
differences in the pickups and not in the bodies. And while we are at it: the guitar body
normally ends where the neck starts. So: go ahead and swap the necks, as well, otherwise you
will assess the neck differences. That doesn’t work when comparing the LP Standard to the
LP Studio? Don’t loose heart – let’s consult that compendium about glues over there in the
corner. Seriously, though, it is here where the limits of this experimental methodology
become visible, long before we arrive at the recommendation that the strings need to be
picked to the millimeter at the same position, and that we need to carry out blind tests, of
course, and … and … and …
Such “listening experiments” often degenerate into euphoric racketing (you don’t get to play a
’58 Strat very often, do you!), followed by the insight that the ’58 sounds more authentic than
the relic’d copy. This may happily be corroborated with the rationale that the old woods are
just so much more inclined to vibrate along, and most of all, they have been “played-in” for
decades. However, maybe it’s only that nut, rock-solidly glued-in by some previous owner so
that it cannot be changed anymore without damage? Or it’s that loose vibrato fit? Or the
worn-down frets on the ’58 that must not be changed? Or the metal pickguard; you would
never imagine it to throw in a damping by eddy-currents, maybe because you have never
heard of them? Or the cables of different lengths that are being used to plug the two guitars
into that home-made switch box? Or the coat of varnish that hampers the guitar body to
“vibrate freely”? That lost pickguard-screw? The Leonids? For real, the latter actually exist,
turning up each November – probably to help prepare Fender products for the Winter-NAMM
(what “The Emissary” and “The Orbs” are for the people of Bajor, the Leonids are for the
Fenderides). The multitude of possibilities that may influence the sound of a guitar is
staggering, and herein lies the problem of such listening experiments: it is simply impossible
to separate the manifold causes, or to attribute exactly one single cause to one single effect. It
is here where the opportunity of artificial sounds lies: because we know exactly how they are
generated, we can change every signal parameter arbitrarily, and check for its audibility or
relevance. Nothing is perfect, though, and we run into other difficulties: how authentic is the
artificial sound – have we considered all significant parameters – doesn’t this all sound very
technical, still? Most of all: what does the (in-) audibility of the 15th partial tell us about the
ash/alder-issue? It all remains difficult … many paths lead up the mountain; not in the
otherwise customary disjunction, though, but rather in a unifying conjunction. Investigations
into materials are more the domain of the manufacturer because other folks can hardly screw,
one after the other, 10 necks to a body just like that. These would be necks for which it is
certain that really only the fretboard differs, and not the bearing of the truss-rod.
Investigations into parameters, however, may well be carried out in a university lab, and the
following pages will be dedicated to them.

It is conducive to divide the string vibration, with regard to time, into a “forced” part and a
“free” part. Plucking is the forced movement (it is not really a forced vibration in the strict
sense), because the string is forced to follow the pick (or fingernail, or such). After the string
and the pick have lost contact, the string may still come into (possibly frequent multiple)
contact with the frets – this will be elaborated on extensively in the following chapter. If the
string has no further contact to the frets (except the one against which it may be pressed), the
“free” vibration (also called the decay process) sets in.
The quantity relevant to the hearing system is the short-term spectrum of the velocity of the
string; specifically of the part of the string located above the pickup magnet (aperture, see
Chapter 5.4.4). The pickup maps the velocity to an electrical voltage that – after amplification
– is converted into a sound wave by the loudspeaker. Given the usual parameter setting♣, the
short-term spectrum (also called spectrogram) shows the level of the partials over time; the
parameters are fundamental frequency, inharmonicity, attack- and decay-spectrum, T30-
spectrum.
• Fundamental frequency (e.g. E2 → 82.4 Hz) and inharmonicity (e.g. b = 1/8000)

were explained extensively in Chapters 1 and 2.
• The attack-spectrum is the magnitude- or level-spectrum of the plucking/picking
process.
• The decay-spectrum is the magnitude- or level-spectrum at the start of the regular
decay process.
• The T30-spectrum indicates the decay time of the partial as a function of frequency.
All magnitudes mentioned above are simplifications: in particular for light strings and strong
picking attack, the fundamental frequency is time-dependent. The inharmonicity does not
describe the irregularities caused by all-passes (Chapter 2.5.2), and the attack may not be
describable with a single spectrum. Moreover, the T30-spectrum may consider beat-effects too
little or too much. Still, it is appropriate to start with a simplified consideration that may be
extended in special cases to a more complicated model. Especially for weakly plucked strings,
the pickup voltage can be described adequately well with the above model parameters; non-
linear behavior will be examined in Chapter 7.12.2.
Fig. 7.126: Spectrograms (0 – 650 ms, 0 – 5 kHz). E3 plucked weakly (left) and strongly (right) on the A-string.
Color dynamic (blue ... red) = 60 dB. Fender Telecaster, fresh strings, bridge pickup.
♣
Window length 20 – 40 ms, Chapter 7.6.2, Chapter 8.6.

Fig. 7.126 shows two spectrograms of the output voltage of a Telecaster. The A-string was
fretted at the 7th fret and plucked at a distance of 12 cm from the bridge; the guitar was
connected to a high-impedance instrumentation amplifier input via a regular cable (580 pF). A
suitable normalization compensates for the lower voltage generated by the less strong
plucking; there are, however, still differences in the attack which forms the very short signal
portion immediately following the plucking. During the attack, the spectral lines form; the
duration of this onset of the tone can only be approximately determined: in the picture on the
left, it is about 20 ms, while in the picture on the right, it is about 60 ms long. If the strings
buzz audibly, the attack phase may take even longer – this will be discussed in Chapter
7.12.2. For the lightly picked A-string (picture on the left), the decay spectrum establishes
itself after about 20 ms; the levels of its partials decay approximately linearly during the
period following the attack. We shall investigate later why some partials contravene this
approximation, decaying with a beating effect. As a first-order approximation, it is assumed
that the decay process is comprehensively described by the decay- and the T30-spectrum
(Chapter 7.6.3).
In Chapter 1, the plucking of the string was interpreted as a step-excitation of a linear system,
supplemented by recognizing that the step is not ideal but “rounded-off”. From the positions
of the plucking point and of the pickup, two interference filters result (Chapter 2.8), and the
pickup acts as a treble-attenuating low-pass filter (Chapter 5.9.3). In the transmission model,
the excitation step passes through the mentioned filters; the latter map the step onto the
voltage. So, we now have: step, pick-filter (for the “rounding off”), plucking-interference-
filter, pickup-interference-filter, pickup low-pass, and output voltage. The two interference
filters have a particularly strong influence – their effect is shown in Fig.7.127. Just shifting
the plucking position by as little as 5 mm already substantially changes the interference filter
(and thus the spectrum; upper right and lower left). The same happens as the pickup is moved
by 3 mm (lower right). Those who see the pickup as immovable may consider that it is the
distance between pickup magnet and bridge saddle that counts: the latter certainly can (and
may need to) be shifted. N.B.: it’s mere millimeters that are crucial here!
Fig. 7.127: Interference filters for the Telecaster. Upper left: A-string plucked at the 7th fret, bridge pickup.
Upper right: plucking position changed by 5 mm (added to upper left graph). Lower left: line spectrum for the
upper right picture. Lower right: as lower left but with the pickup position changed by 3 mm.

Fig. 7.128 shows to which extent model and reality agree. On the left, the decay-spectrum♣ is
depicted; it was on the one hand derived from a measurement (Fig. 7.126), and from the
above filter model on the other hand. In view of the large differences that already show up as
the plucking position is changed by a millimeter, the correspondence is to be seen as very
good. The T30-spectrum (the decay times of the partials) is shown in the picture on the right,
with the grey area indicating an estimate for the upper limit valid for the open A-string (due to
radiation and internal damping of the string, see Chapter 7.7.2, “orientation line”).
Fig. 7.128: Left: decay-spectrum (measurement –––, model: ----); right T30-spectrum (measurement). The region
marked in grey estimates the upper limit of T30 due to radiation/internal damping of the string, Chapter 7.2.2).
From the dataset shown in Fig. 7.128, we synthesized an artificial guitar tone (fG = 165 Hz
and b = 1/6060). The spectral analysis (Fig. 7.129) indicates a good correspondence – merely
the beating is (deliberately) not included. In turn played back via a guitar amplifier, both
signals sound identical as long as the duration is kept short (about 0.5 s). Only when
extending the duration to several seconds, minimal differences in the strength of the beating
become apparent. However, since any halfway gifted guitarist would almost always play a
note held for 3 s with finger-vibrato, this effect was ignored. If it were considered to be
relevant, after all, it would be very simple to add some beating to the artificial signal.
Fig. 7.129: Spectrograms. Left: real Telecaster-signal, right: synthetic signal. 0 – 5 kHz, 0 – 2.5 s.
We now arrive at a first conclusion: the time-variant short-term spectrum is a powerful tool to
analyze the voltages generated by weakly plucked strings, and the associated analysis of
partials is well suitable to generate artificial guitar tones. The decay spectrum results from the
data of pick, string and pickup, and from the plucking- and pickup-position. There is
practically no dependence on the remaining guitar parameters (in particular not on the wood).
The T30-spectrum, i.e. the speed of the decay of the partials, is defined by the remaining guitar
parameters.
♣
Here and in the following the results are not shown anymore as discrete frequency lines but as a polyline.

In Fig. 7.128, the decay time reveals the already known decrease towards high frequencies, as
well as smaller frequency-selective variations. In the following we shall investigate how far
these small T30-peaks (e.g. around 2 kHz) have an effect on the audible sound. In the
following listening experiments, the tone synthesized according to Fig. 7.128 was the
standard sound. Starting from it, the decay times of individual partials were changed as
modifications. In these experiments, we could confirm very quickly the masking models [12],
according to which partials with small levels contribute practically nothing to the perceived
sound. It consequently is not important for the auditory impression how fast the 11th partial
(1.8 kHz) decays, as long as the modifications stay within the regular range. Even reducing
the decay time of the 1.8-kHz-partial to 0.5 s, or extending it to 2.5 s (Fig. 7.130) does not
change the aural impression. For the same reason, the particularly clear level difference
between model and measurement (Fig. 7.128, decay spectrum) for specifically this partial at
1.8 kHz is insignificant: the partials with low levels are masked and do not contribute
anything to the aural impression.
Fig. 7.130: Modification of the decay times of individual partials. The changes affected in the left-hand graph
remain inaudible; the extension (red) in the right hand graph is audible, the shortening (blue) remains inaudible.
The region marked in grey estimates the upper limit of T30 due to radiation/internal damping of the string).
A different situation is found for the 13th partial (2170 Hz): extending its decay time to 2.5 s is
audible, while the shortening to 0.5 s remains inaudible. However, small changes in the decay
time are caused already by minor shifts in the position of the fretting hand (Fig. 7.131 left) –
again, this has very little bearing on the sound. Only when playing the notes for several
seconds and when directing ones concentration specifically to the fundamental, miniscule
sound differences become apparent – but these have practically no importance. Yet another
situation emerges at the 2nd fret of the D-string: although the same note (E3) is being played,
both decay- and T30-spectrum are different. Despite the same plucking- and pickup-positions,
the two interference filters change their frequency response – due to the changed relation: the
A-string is plucked at the relative position 12/44, the D-string at 12/58.
Fig. 7.131: T30-spectrum. Left: E3 played at the 7th fret the A-string, left hand held in different positions.
Right: E3 played at the 2nd fret of the D-string. Fender Telecaster, fresh strings (009 to 046).

The changes in the decay time (Fig.7.131 right) are due to a strong location dependency of
the neck-conductance on one hand, and on the other hand due to the length-dependency of
the internal string damping. For the synthesized tone it is now very simple to keep the decay
spectrum (as shown in Fig. 7.128), and to change at the same time the T30-spectrum according
to the right-hand graph in Fig. 131. Does the sound change audibly due to this? It’s the same
fundamental frequency, almost the same inharmonicity, the same spectrum at the beginning,
but a different decay of the partials – indeed, that sounds different. Not yet for very short
durations (250 ms = 1/8th-note at 120 BPM), but already from a duration of 500 ms. The
longer the tones last, the more muffled the A-string note sounds compared to the D-string
note. This difference cannot be compensated for by the tone control – cranking up the treble-
knob does not change the decay speed of the partials!
These audible differences between E3‘s played on the A-string and on the D string can hardly
be attributed to the body wood, because that is the same for both notes. To once more
summarize the causes for the differences: even when keeping the location of plucking the
string and the location of the pickup constant, the relative distances still change, and so do the
two interference filters. This is, for “normal guitar-playing” the main difference between the
A-string E3 and the D-string E3. If both notes are given the same decay spectrum (which is
only possible for synthetic notes), we notice a progressive treble-loss for the A-string E3: the
string sounds increasingly duller. The D-string sounds progressively brighter in comparison.
Now on to the Gretsch Tennessean, a true semi-acoustic guitar. Its hollow body promises
peculiarities in the decay behavior – but the listening tests do not show this. Of course, the
Tennessean sounds different – but that is mainly due to the different pickups and their
different position (compared to a Telecaster). The scale is different, as well – and therefore
Telecaster and Tennessean form different interference filters, even if the same note is played
on the same string. However, if – for the same decay spectrum – we change only the T30-
spectrum (i.e. the spectrum of the partials), we cannot hear any difference between Telecaster
and Tennessean for short notes. Only as the duration of a note increases to about 500 ms,
differences start to become noticeable – and these are minute differences!
Fig. 7.132: Gretsch Tennessean, E3 played on the A-string (left) and on the D-string (right). The solid lines are
the result of fresh strings (009 – 046), the dashed lines are the result of “broken in” strings (009 – 046).
The differences caused by the aging of strings are much larger (Fig. 7.132). Only for
completely fresh strings, any frequency-selective peculiarities can be detected at all –
for “broken in” strings, string-internal loss mechanisms dominate. Still, there is no generally
applicable rule about the loss of brilliance as a function of time, because the individual
parameters (dust, skin-fat/oil and -abrasions, bending-grooves, rust, fret- and string-material)
are too diverse.

Fig. 7.133: Decay time of partials. Comparison Fender Telecaster (----) vs. Gretsch Tennessean (–––).
Fresh strings (009 - 046), E3 on A-string (left) and on D-string (right), plucked with a pick.
The direct comparison between Tennessean and Telecaster is shown in Fig. 7.133. In some
ranges, shorter decay times can be seen for the Tennessean – an effect of the hollow, flexible
(and thus absorbing) body. For shorter notes, however, these differences are certainly put into
perspective: a T30-difference of 1.9 s relative to 3.0 s translates into a level difference of 1.5
dB for a note of 0.25 s duration (an 8th-note at 120 BPM). Such a difference may just be
noticeable under conducive laboratory conditions, but it is of not much significance in
everyday life on stage or in the studio. We cannot often enough remind ourselves of this: it’s
in the fingers. And also in the pickup, and in its position. How the spectrum is shaped by the
plucking- and pickup-positions, that is subject of the following investigations.
As with every spectral analysis, we need to find a compromise between high spectral and
high time-related resolution (compare to Chapter 8.6). In order to keep leakage-effects at a
bearable level, the time-function needs to be subjected to a “window”. In Fig. 7.134, the left
hand part depicts the pickup voltage of a plucked E3, starting with a positive peak.
Multiplication with a window-function yields the right-hand graph – and it is here where the
multitude of parameters catches up with us, because duration as well as type and parameters
of the chosen window define the spectrum.
Fig. 7.134: Weighing over time of the pickup voltage (E3) by a Kaiser-Bessel-window (N = 2048).
To spectra for the above depictions are shown in Fig. 7.135: on the left using a 2048-point
window, on the right with a 1024-point window. We could live with both representations, but
due to the clearer line-structure the following analyses use the 2048-point window.
Fig. 7.135: DFT-spectrum for Fig. 7.134, Kaiser-Bessel-window, N = 2048 (left), N = 1024 (right).

Fig. 7.136: Spectra and time-functions (subjected to the window) of the pickup voltage. E3 on D-string,
Telecaster. The calculated model-envelopes (----) fit perfectly to the measured line spectra.
In Fig. 7.136 we see the spectra corresponding to different time-excerpts. During the first
milliseconds, the pickup position, and pick- and pickup-filter determine the shape of the
spectral envelope (dashed in red). As the pick leaves the string, two step-response waves run
off in two directions. The wave running towards the bridge crosses the bridge pickup first and
is compensated shortly afterwards by the opposite-phase bridge-reflections – this results in a
short positive impulse. The other step-response wave is reflected by the 2nd fret and reaches
the pickup somewhat later – only now the picking-interference filter (on top of the pickup
interference filter) takes an effect (dashed in blue).

The impact of these two interference filters depends on the string length, and on the picking-
and pickup-positions. In this example, the pickup itself approximately acts as a 2nd-order low-
pass (fx = 4 kHz, Chapter 5). As shown by the example, the aperture of the magnet and the
rounding-off of the excitation step may very nicely be modeled by a further 4.3-kHz low-pass.
The body wood? That can exert any influence on the spectrum and the sound of the guitar
only via the damping of the reflections, and here the following holds: the shorter the
observation period, the fewer reflections we get, and the more insignificant the body! Even at
150 ms (lowermost graph), we still recognize the dominance of the two interference filters in
the decaying spectrum. Only the 5.3-kHz-partial behaves differently: it decays significantly
faster than its colleagues. Still, as already elaborated extensively in Chapter 7.7: first we need
to consider the damping characteristics due to strings and bearings. The influence of the body
wood comes last.
Fig. 7.136 justifies a distinction into an attack- and a decay-spectrum. During the first
milliseconds, the spectral envelope (dashed in red) depends only on the pickup-interference
filter (besides the pick- and pickup-low-pass). Only from about 10 ms, the picking-
interference filter gains in significance. For the auditory perception, the decay spectrum
formed by all filters is decisive; its envelope is included dashed in blue in Fig. 7.136 for the
30-ms-spectrum.
A different situation only appears as we move to the non-linear system, but before we concern
ourselves with its idiosyncrasies in the next chapter, let’s first look at some fringe-effects.
Besides the guitar, also amplifier, loudspeaker, and listening room will of course influence the
sound arriving at the ear. Boosting a partial originally weak in level by a frequency selective
filter (EQ), this partial may increase in significance and become audible. A similar
development may occur if the decay time of a partial is changed to the extreme (e.g. from 1 s
to 5 s). Any statement regarding the audibility (or non-audibility) is therefore never of general
validity but should be taken in the framework of normal stage- and studio-technology.
Moreover, the results found for one note cannot be directly carried over to all other tones – a
guitar has more than one string, more than one fret, and more than one partial. Only the
transmission-filter of the pickup (Chapter 5.9.3) may be seen as reasonably string-unspecific.
Pick- and aperture-filter are string-specific, and on top the two interference filters are also
strongly position-specific.
Last, it should be mentioned that we were not out to obtain the absolute threshold of
perceived differences. In basic research it may be justified to exactly determine the difference
between a real guitar note and a correspondingly synthesized note using a representative
group of subjects. Or to e.g. find out that beat-differences were recognizable from a note-
duration of about 0.23 s. However, if the fretting hand grabs the neck a bit more strongly for
the repeat measurements, this duration would change, and the same would happen if the angle
of attack of the pick would change minimally, of if the guitar is pressed a bit more tightly to
the belly. This threshold of perceived differences is not unimportant – but it is connected to an
overwhelming variety of parameters. The functional model including pick filter, picking-
interference filter, pickup-interference, and pickup-transmission filter explains the
spectrogram in a simple manner; the data-sets of decay-spectrum and T30-spectrum are the
most important ones for this spectrogram. If we additionally supplement fundamental
frequency and inharmonicity, weakly plucked notes of an electric guitar may be synthesized
with good quality, as long as their duration is not too long. Based on this model, parameter
variations may be checked, with the result being the assessment of the relevance of individual
components. Indeed, this is much better than the pure hunch that alder would give a shorter
sustain due to its higher elasticity [G&B and others, see Chapter 7.8].

7.12.2 Non-linear string oscillations
Now, we will have to dive into the thicket of complicated matter. That’s because
communication engineering teaches us that for non-linear systems there is neither
superposition nor proportionality, and neither transfer function nor step response. Of course,
we may drive also a non-linear system with a step excitation, but the response (reaction) is not
a signal-independent system function, but it depends on the excitation signal – and as a result
there is not “the one” step response but there is an infinite number of such step responses.
Strictly speaking, every technical system is non-linear, but often this is to such a small degree
that the transmission characteristics may be simplified towards linearity. Chapter 5.8 had dealt
with the non-linearities occurring in pickups – the effects are far from insignificant yet they
are far outweighed by the non-linearities possible in the string vibrations. The latter become
non-linear if, after being plucked, the string hits the frets. In this case, the step-waves
generated by the plucking are not only reflected by the bridge and the nut (or the fretted fret)
but also at the (other) frets. This process is dependent on the plucking-strength and therefore it
is non-linear.
Fig. 7.137: Spectrogram of a plucked D-string (E3, 0 – 800 ms, 0 – 10 kHz, dynamic in the graph = 30dB).
Left = lightly plucked, right = strongly plucked. Fender Telecaster, strings 009 - 046, bridge pickup.
The analyses were scaled to the same maximum drive level.
Fig. 7.137 shows spectrograms of a D-string plucked with different strengths. Depicted are
auditory spectrograms (Cortex VIPER) the analysis-parameters of which are adapted to the
characteristics of our hearing system. It is hard to believe that both analyses were obtained
with the same guitar, the same string and identical plucking positions – only the plucking
strength varied. The lightly plucked string clearly reveals the interference filter, with the
spectral emphasis being formed by the first three partials. The outcome for the strongly
plucked string is very different: the first two partials (fundamental and second harmonic) have
only a weak level – their vibrations cannot unfold due to the amplitude limitation. Even if a
simple model would attribute the same displacement-amplitude to each partial, the pickup
voltage – corresponding to the velocity – would increase with increasing order of the partial
in this model. In a real plucked string, the partials do not have the same displacement
amplitude: the plucking-interference-filter causes gaps (e.g. for the 5th and the 10th partial).
However, already the first string/fret-contact starts to fill in these gaps. If we interpret the
plucking of the string as a step-excitation, the string hitting the fret could be seen as a kind of
impulse-excitation, albeit quite a special one. This is because while the plucking action feeds
vibration energy to the string, hitting the frets can only cause an energy loss. How big this
loss is depends on the surface qualities (among other factors): little loss for a fresh string and
a clean fret but more loss, if an in-between layer of dust/grease/talc acts as an absorber.

The (normalized) time function shown in Fig. 7.138 underlines the differences between the
lightly and the strongly plucked string: during the first few periods, only the dispersion has an
impulse-changing effect in the left-hand graph, while for the strongly plucked string,
reflections are clearly visible already within the first period. These are reflections that can
only stem from the obstacle located closest – and that is the last (22nd) fret. There is a
significant chance that a strongly plucked (thin) string comes in contact with the last fret
already in the plucking process (compare to Chapter 1.5.3), but the exact evolution over time
of this and other fret-contacts are dependent on the individual plucking – this is in fact why
the system behavior is non-linear. A model can therefore emulate the fret-bounce either only
for the individual case, or simulate – as a stochastic model – a generic average event. Which
is the problem: we will not get far with one model alone, because the well-versed guitarist is
able to generate a multitude of fret-bouncing “snap-sounds”.
Fig. 7.138: Time function of the pickup voltage; left = lightly plucked string, right = strongly plucked string.
Fig. 7.139 depicts in which unexpected variants the decay of a strongly plucked string can
occur. Again, the D-string of a Telecaster fretted at the 2nd fret is shown, strongly plucked at
12 cm distance from the bridge (as in Fig. 7.137). As opposed to the above analysis, the string
was not pushed downward at a slant, but lifted up and then let go. At 0.6 s, the spectrum (and
the sound) change unexpectedly: the 5th partial literally cuts out, while other partials only
come to life at that point in time. These changes are not connected to the fretting hand but are
the work of the string alone – in cooperation with the frets.
Fig. 7.139: Spectrogram of a plucked D-string (E3, 0 – 1500 ms, 0 – 10 kHz, dynamic = 30dB).
Right: red = 5th partial, blue = 4th partial, black = level of fundamental; all as a function of time.
The string, with a vibration that is at first almost perpendicular to the fret-board, hits the frets
which transfers part of the vibration energy into a mode parallel to the fret-board. Over time,
however, the plane of vibration changes back again, as easily visible from the level of the
fundamental (----). Around 0.6 s, the increasingly fret-board-normal vibrating string
approaches the frets again such that a further crash occurs. This crash considerably disrupts
the 5th-order vibration, but at the same time retriggers and amplifies the 4th-order vibration. A
model describing vibrations in only a single plane would not succeed for such a behavior,
even if that model would allow for non-linear amplitude limiting.

Fig. 7.140 shows further spectrograms; again only the strength of the plucking was varied. It
is characteristic that the partials decay neither exponentially nor according to a simple
beating-model, but suddenly change their decay behavior. Even an increase in level is
possible (albeit one only for a limited time) – this can be attributed to a slowly rotating
polarization plane. Contact between string and fret may be limited to the first 0.1 s, but may
also still occur after 1 s. The evolution of the level and the sound color over time is
correspondingly rich in variation. These figures highlight that the neck (or rather the frets)
enjoy an elementary significance: a fret minimally projecting over the other frets will generate
other bounce-contacts than one that is worn down.
Fig. 7.140: Auditory spectrograms and levels of partials of a plucked D-string (compare to Fig 7.139).
Level-normalized scaling; dynamic in the graph = 30 dB. Telecaster, bridge pickup, fresh strings (009 – 046).

It is not difficult to corroborate speculations about string/fret-contacts by measurements: for

this, all 22 frets of a Telecaster were electrically connected to a 22-channel analyzer, and all
contacts occurring during the decay were stored. The representation in a tactigram (bounce
chart) shows characteristic patterns that agree well with a line model (Fig. 7.141). The D-
string of the Telecaster is pushed down such that it comes into contact with the last (22nd) fret
– this happens often when light strings are used. As the string looses the contact to the pick,
waves propagate in both directions and are reflected at the last fret and at the bridge, and lift
the string off the fretboard, as shown in Fig. 7.142.
Fig. 7.141: Bounce chart. Telecaster, D-string fretted at the 2nd fret, pressed down strongly 12 cm away from the
bridge and then released. Top: measurement; bottom; model-calculation. Dots = string/fret contacts.
After half a vibration period, a maximum in the displacement has formed above the 6th fret
(Graph #8 in Fig. 7.142); it breaks down again during the further continuation of the
vibration. Immediately afterwards, the string hits the fretboard, with curvature of the neck and
condition of the frets deciding where exactly the string/fret contact happens. The angle with
which the string is pressed down also plays a role: it makes for a difference whether the string
is pushed down exactly perpendicular to the fretboard or with a slant relative to the fretboard.
This is because the orientation of the string excitation determines the share of the fretboard-
parallel vibration. During the decay process, the plane of vibration rotates (even specifically
to each partial), and it is in particular the fretboard-normal share of the vibration that is
clipped by bounce-processes. The fretboard-parallel vibration-mode is a kind of energy-
storage that only slowly feeds its vibration energy to the fretboard-normal vibration. The latter
(being important for the pickup signal) can therefore repeatedly generate further string/fret
contacts. Note that in the model calculation shown in Fig. 7.142 only one plane of vibration
was considered.

Fig. 7.142: String displacement at various points in time. Parameters as in Fig. 7.141.

Fig. 7.143 shows how big the spectral differences between strongly and lightly plucked
strings can be: in the left column we see, during the first milliseconds, the attack-spectrum
already known from Fig. 7.136 (red envelope, interference gaps dependent on the pickup
position); it transitions into the decay spectrum (blue envelope). In the column on the right, a
sinc-shaped envelope cannot form at the beginning due to the string/fret contact supplying
additional impulses. Only later, an influence of the pickup position establishes itself as an
outline, while the plucking-position is not evident anymore at all as interference filter.
Fig. 7.143: Spectra of the pickup voltage (subjected to a window). E3 on D-string, Telecaster (cf. Fig. 7.136).
Left = lightly plucked string, right = strongly plucked string. Level-normalized representation.

An entirely different vibration happens if the string is not plucked pressing down, but is lifted
and then released. This plucking technique is found especially with guitarists playing finger-
style (i.e. without a pick) – but even with a pick, an at least similar behavior can be achieved.
Fig. 7.145 shows snapshots for a string pulled up far enough so that it bounces on the frets
after its release. In this example, first contact happens at the last (highest) fret, followed by a
series of contacts running along the fretboard towards the lower frets. Then (Graph #7), the
string looses contact only to touch all frets in quick succession (Graphs #13 – 14). Or at least
almost all frets – in details this of course again depends on minute differences in the heights
of the frets.
Fig. 7.146 once more compares contact-measurements with model calculations. Considering
the complexity of the matter, the correspondence is very good at the start – as they progress
along the time-axis, the two representations differ more considerably. This is because
dispersion was not modeled, because the polarization was only calculated for one plane (and
not circularly), and because the fret-heights were idealized in the model (in the investigated
Telecaster specimen, the fret were already slightly worn).
Fig. 7.147 indicates that string/fret contact is not necessarily limited to the attack-phase. In
this example, the string repeatedly bounces off the 3rd fret – however this happens so lightly
that no annoying buzz but merely slight brightening of the sound (a mixing-in of treble)
occurs.
We see from Fig. 7.144 how strongly even tiny differences in the height of the frets can make
themselves heard. Here, we first calculated the string velocity over the pickups using the non-
linear string model, and then derived the spectrum from it. This was done for two different
fret-boards on which the 18th fret differed in height by 0.2 mm.
Fig. 7.144: Calculated spectra of the D-string bouncing off the frets. The only difference between the two graphs
is that the height of the 18th fret differs by 0.2 mm.
These results give an indication of what can happen when comparison tests are run by a
magazine checking out the “holy grail” – i.e. if, for example, a original 1950’s Les Paul is
compared to a more recent reproduction. Of course, the frets of the priceless♣ vintage guitar
are worn, maybe so strongly that it causes the celebrating tester to grimace a lot, and of course
the trained ear will hear all kinds of differences. Too bad: as soon as this “grail” is put in a
playable condition, its $-value takes a nosedive. Thus do note: on every grail rests a curse of
some kind.
♣
Not to be taken all that literally: that’s from about € 200.000; quite nicely done fakes may be acquired.

Fig. 7.145: String displacement at various points in time. Parameters as in Fig. 7.141.

Fig. 7.146: Bounce chart. Telecaster, D-string fretted at the 2nd fret, pulled up at 12 cm from the bridge and
released. Top: measurement, middle and bottom: model calculations. Dots = string/fret contacts.
For simulation 2 (bottom), the bridge was raised by 0,3 mm relative to the setting for simulation 1.
Fig. 7.147: D-string fretted at the 2nd fret: even after 0.5 s there are string/fret-contacts

In a nutshell, we have the following situation (for further elaborations see Chapter 7.12.3):
In electric guitars with heavy strings and a high action, string/fret contacts (other than the
actual fretting action) are rather rare, the low partials can develop nicely, and the electric
sound is quite full. The string vibration can approximately be modeled in a linear fashion.
Given light strings and the correspondingly connected light playing forces (Chapter 7.4.1)
each individual note may be accompanied by string/fret contacts (especially for strong
plucking), resulting in a more percussive sound with more treble. The low-frequent partials
are less distinct because they lack the required amplitude. Of course, the individual plucking
process always is essential: with brute force, it is possible to make a heavy string bounce onto
the frets, as well, and a light string may be plucked so gently that it does not come into
contact with the frets. That’s what this statement is based on: it’s all in the fingers, man!
For short notes, the guitar body has next to no influence on the electric sound, and for solid
body guitars no influence is felt for longer sustained notes, either. With hollow-body
instruments, in particular two effects are found: since especially the low-frequency notes are
(acoustically) radiated better, the corresponding decay times are shorter, and for the same
reason these instrument tend to feed back more quickly.
In terms of influencing the sound, the way/style of playing comes first, and strings and
pickups are next (in high quality guitars). We then get to the mechanical characteristics of the
bridge, and then to the frets (even the higher-most, possibly “never used” ones). That the
acoustical sound radiated by an electric guitar would give “complete” testimony about the
electric sound is a fairytale – albeit one that apparently cannot be silenced. Already Leo
Fender and Les Paul fully understood that the vibration-energy needs to remain in the string
as long as at all possible – as little as possible should be transferred into the body. Any
acoustic sound needs to be channeled through the body (to use layman’s terms) – so the
material it is made of is relevant, but – alas! – only for the acoustic sound. The guitar body
can influence the electric sound, but only in terms of absorption. Since it seems that every
guitar player demands a sustain as long as possible, the absorption needs to be as low as
possible. In that case, however, the influence of the body wood on the electric sound has to be
as small as possible, too. Knowing that, it is not surprising that an electric guitar build from
undefined, knotty platform-wood can fill the guitar player with enthusiasm due of its sound
(G&B 7/10) … because of its electric sound, that is, of course.
7.12.3 The roots of the electric sound
Of course, the pickup voltage does not yet yield a “sound” – for that, amp and speaker are
required, and – diving into philosophy – a listener, as well. Wouldn’t it make for a great
debate to ask whether airborne vibrations that are not heard by anybody merit the term
“sound”? But that would be the realm of those physicists who – good heavens! – seek to
become a DPhil rather than a DSc, i.e. move into a world completely foreign to the Doctor of
Engineering. In short: without amplification, the electric guitar generates an acoustic sound,
amplified it generates the electric sound. Only the latter is addressed in the following, as is the
analysis and description of its origin.
Step-excitation and pick-filter

From a systems-theory point-of-view, plucking a string represents an impressed force-step –
however not one in the form of an ideal step-function but modified by the pick-filter (Chapter
1.5.2). Due to mode-coupling in the bearings (bridge, frets) and magnetic pull-forces, the
string vibration does not remain in one plane but starts a wobbling motion in space (circular

polarization, Chapter 7.7.4). Using angle of pick-attack, plucking-strength and -direction, and
via further finger/string contacts, the guitarist shapes the electric sound to a significant degree.
String vibration
Starting from the plucking location, transversal waves and dilatational waves run in both
directions. They are reflected at the bearings and (approximately) result in standing waves.
Dilatational waves that propagate dispersion-free are of lesser significance but they may lead
to frequency-selective absorption losses (Chapter 7.7.4.2). Transversal waves (the important
wave-type) propagate with dispersion; their propagation velocity increases towards higher
frequencies, leading to an inharmonic spreading of the line spectrum (Chapter 1.3). This
inharmonicity (dependent on the string diameter) is quite desirable: it livens up the tone,
especially in case of non-linear distortion in the amplification chain (difference tones, see
Chapter 10.8.5).
String material
The (manufacturer-specific) relation of core- and winding-thread (Chapter 1.2) is – right
behind the overall diameter – the other important parameter influencing the inharmonicity of
the partials. A further influential factor could be how tightly the outer thread is wound onto
the core; but compared to ageing processes (skin oils, corrosion), it takes a backseat.
Plucking (picking) position

The plucking-position separates the string into two sections, the length-ratio of which
determines the zeros of the plucking-interference-filter. The closer the plucking happens to
the bridge, the further apart the filter-zeros are, and the harder and more trebly the sound gets
(Chapter 2.8). The plucking-interference-filter operates with an individual characteristic for
each string and cannot be simulated with a simple effects device.
Pickup position
Just like the plucking-interference-filter, the pickup-interference-filter is a comb-filter; its
zeros are, however, determined by the pickup position and not by the plucking position
(Chapter 2.8). For the single-coil pickup, one comb-filter is active, for a humbucker there are
two. If there is a difference between the two humbucker-circuits, further degrees of freedom
in the signal filtering result. Again, the pickup-interference-filter acts string-specific, and its
effect is dependent on the fretted pitch.
Magnetic aperture, non-linearity

The aperture-filter is a string-specific low-pass (Chapter 5.4.4) that is defined by the width of
the magnetic window. Decreasing the distance between magnet and string, and increasing the
magnetic strength increases the cutoff frequency. The filter is string-specific. For strong
picking attack, the magneto-electric transfer (Chapter 5.8) based on the law of induction
shows a non-linearity that should not be neglected. This non-linearity is string-specific and
therefore must not be mixed up with amplifier distortion.
Pickup directionality
If a pole-piece of a pickup is positioned exactly underneath the string, the pickup will sample
almost exclusively the fretboard-normal string-vibration (Chapter 5.11). This implies that
pickups offset to the side will to some extent tap into the fretboard-parallel vibration, as well
– this may be of significance for fret-bounce processes (Chapter 7.12.2)

String damping
The string vibration is dampened by several mechanisms (Chapter 7.7); it is in particular the
internal damping, and the damping at the bearings, that stand in the way of “endless sustain”.
The internal damping is generated by micro-friction as the string is deformed; for wound
strings, the damping occurring between core-wire and winding may be included here, too.
Also, already minor residue from skin or talcum in the winding leads to dramatic reduction of
the decay time (Chapter 7.7.6). The damping due to bearings happens at the bridge and at the
frets (or at the nut) because a small part of the vibration energy is not reflected but drains into
the bearings. This is the only mechanism with which the guitar body can have any influence
on the electric sound at all♣, but because this damping effect is supposed to be as small as
possible in order to support long sustain, the influence of the body is very small. In particular
in solid-body guitars, the body inflicts little absorption. The bridge – located between string
and guitar body – may exert a comparably larger influence on the string vibration.
Neck, action, frets

Forceful picking and/or low action will have the effect that the string often bounces off the
frets (Chapter 7.12.2). The percussive sound caused by this depends largely on the height of
the individual fret – and the “never touched” uppermost frets are relevant here, as well.
Therefore, if a musician notices the sound of a guitar changing over time, this is not because –
as Neil Young opines in G&B 12/05 – every played note somehow stays in the guitar , but
very probably because of fret-wear . Which would also explain why that vintage guitar
acquired for a 5-number sum does suddenly not sound “vintage” anymore at all after the
urgently required re-fretting job has been performed.
Finger- and hand-damping

As soon as the fretting hand touches the guitar neck, it acts as an absorber and potentially
reduces sustain, and a similar effect is caused by the finger pressing the string against the fret.
We may find pertinent frequency-dependencies with open-played, brand-new strings –
however, these dependencies quickly loose their significance after having played for half an
hour, and when analyzing not only open strings.
Pickup transmission
The transfer-function of a magnetic pickup is predominantly determined by the inductance of
the winding, and the capacitance of the cable (Chapter 5). Together, the two form a low-pass
the cutoff frequency of which may lie below 2 kHz, or above 5 kHz – thus, the pickup plays a
decisive role for the electric sound. The transmission coefficient that may easily vary by
+300% contributes significantly to the sound in case the input stage of the amplifier is
overdriven. Consequently, there can be no serious statement along the lines that the pickup
would just add a few “nuances” to the “sound of the wood”. Apart from the LC-lowpass, the
pickup may contain further frequency-determining components, such as metal sheets causing
dampening of eddy-currents, or guides for the magnetic field that result in a spatially more
spread-out sampling of the string vibration. In humbuckers, inductive and capacitive coupling
processes may cause complex filtering. The parameters of pickups of seemingly the same
build can have considerable scatter: in particular in old pickups, the number of turns in the
winding, the thickness of wire and varnish, the magnet material, and the fittings can vary
strongly, and even magnets mounted the wrong way ‘round may occur. Moreover, old
pickups may have shorts in the winding, and therefore there is not “the” Strat-pickup, nor is
there “the” 1958-Strat-pickup.
♣
Regarding body- and neck-resonances, see Chapter 7.7.4.4.

Electrical circuit in the guitar

The electrical components (potentiometers, capacitors, possibly also coils) included in the
guitar form an electrical network the filtering effect of which may be described without much
effort. The “holy aura” attributed to old components can scientifically not be substantiated,
and in particular horrendous markups are not justifiable, even if corresponding myths are
eagerly celebrated by some failed HiFi-authors. On the other hand, the coaxial cable
connected to the pickup may spring a surprise due to a possible peculiar, humidity-dependent
capacitance. Also: amplifier, loudspeaker and room must not be forgotten (Chapter 10 & 11).
The insignificant
Of course, given the right equipment and putting in many hours of effort, even minute
changes in the decay behavior can be measured, e.g. when machine heads (tuners) are
exchanged. The same may be possible if varnish is stripped off the guitar body, or if it is
replaced by another type of varnish. However, all these changes are so tiny compared to the
variations effected by the fretting hand that they simply bear no significance whatsoever.
Kaput: the broken, busted, worn out and dead

And then there are of course all those more or less broken, in fact unplayable guitars that
“feature” unacceptably uneven frets, loose necks, rattling truss rods, pickups with shorts in the
winding, scratching pots, bridges that shift from one rest-position to another at the slightest
touch, or a “custom job” done by Mr. Knowitall. May the Eternal Shredder graciously accept
their souls ….
You others, though, who in your hands an unbroken guitar you hold:
Do search not for new gimmicks, but to play learn – everything else come to you it will.
7.12.4 There’s nothing there, or is there??
That we tried, in this chapter, to trace the tiniest measurement artifact, and to capture
conductances with, if possible, no less than 80 dB dynamic range – that does not imply that
all the little peaks we could eventually measure are at all audible. Just as the executive
authority needs to be separated from the judiciary authority, we need to distinguish
psychoacoustics from instrumentation when doing an analysis of sound. The better the
analytics, the safer it is to attribute a measured effect to the object to be measured, rather than
running the danger that the measurement device fooled us. Indeed, it is a great result, as well,
if a bridge conductance measured with much effort proves to be so small that its irrelevance is
now safely established. And even if an audible effect shows up: not every difference in sound
points to the source of the purportedly never-again-reproducible vintage-tone (whatever that
may be) … not every fart renders the planet inhospitable.

7.13 Neck curvature and fret height 7-165
7.13 Neck Curvature (relief) and Fret Height
The stronger a string is plucked, the more easily it can bounce onto the frets (those other than
the fingered fret, Chapter 7.12.2) – the exact position of these “other” frets is therefore a
decisive factor in the sound. On the unstressed guitar neck, all frets will have the same
height; however, the string tension bends the neck and changes the distances between the frets
and the string. According to the manufacturers, a slightly concave neck curvature is ideal;
Fender recommends 0.25 mm clear width at the 8th fret, with the E2-string pressed down
simultaneously onto the 1st and last frets. Lemme allows for 0.3 – 0,5 mm, and Gibson
proposes a different procedure: press down the string on the 12th fret and measure the clear
width at the 7th fret – it should not read more than 0.4 mm. Ovation, on the other hand,
reckons that the E2-string should be pressed down at the 1st and the 13th fret, and the clear
width at the 5th fret should read 0.13 – 0.38 mm. All these procedures merely allow the global
neck curvature (also called neck relief) to be measured – the individual fret height cannot be
checked this way. The latter is, however, of course just as important: if the 9th fret protrudes
as little as 0.1 mm, the string fretted at the 8th fret does not sound right.
The neck relief depends on the solidity of the neck and the load forces acting on it: the string
pull, and the counteractive force exerted by the truss rod. As the simplest model for an
analytical description of a concave neck shape, mathematics offers us the 2nd-order parabola
(Fig. 7.148). For the graphs, the height of the first and the last frets was defined as equal; on
the left, three different neck curvatures are shown. However, a 2nd-order parabola can only
serve to realize even (axially symmetric) functions – a scenario that cannot generally be
expected for a guitar neck with a tapering cross-section. Help is on the way in the form of a
3rd-order parabola (right-hand graph); it can represent skewed curvatures, as well. To which
extent a guitar neck shows a skewed curvature depends on the progression of the cross-section
of the neck, and on the truss-rod. The progression of the cross-section therefore does not only
influence how the neck “feels” but also how and where the strings bounce onto the frets.
Fig. 7.148: Neck relief; 2nd-order parabola (left), 3rd-order parabola (right).
The graph as shown in Fig. 7.148 offers a good view of the neck curvature, but the relation to
the string is still missing. If the latter is not fingered, it is supported by nut and bridge. The
curvature is adjusted via the truss rod, and the so-called “action” (the distance between neck
and string) is adjusted via the bridge height. As a benchmark, we find the general
recommendation to adjust the action to about 1.5 – 2.5 mm at the last fret. This distance is
difficult to measure with a ruler; it is more conducive to use a set of drills with a 1/10-mm-
gradation between them. The drill-shank is pushed between string and last fret, and a check
which drill fits just shy of lifting the string reveals the action.

Fig. 7.149: Two different fret heights relative to the string (“Saite”);
2nd-order parabola (–––) / 3rd-order parabola (----).
Fig. 7.149 shows two differently curved necks and the string above them (dotted line). The
width of the graph corresponds to the scale (about 65 cm) i.e. it covers a multiple of the height
of the graph: the curvature therefore is strongly exaggerated. There is no “optimum” neck
relief, but we frequently read that a convex curvature is not conducive, and too strong a
concave one is not, either. Theoretical essays about the optimum neck relief most often
assume sinusoidal string movements (often even with a fret-independent displacement-
amplitude), and fail to consider the (in fact rather complicated) real shape of the string-
oscillation. Generally, sections of shallow incline tend to be unfavorable, and therefore each
of the (established) neck-profiles includes advantageous and disadvantageous sections. In the
end, the choice is a matter of taste.
Contrary to the global neck curvature, the differential fret-height has to satisfy objective
criteria: the height of each fret needs to precision-fit within a few hundredths of a millimeter
relative to a regression curve (representing the average curve)! This precision can, however,
not be achieved if the frets are simply hammered into the neck – further dressing (sanding) of
the frets is mandatory. On the following pages, measurements are depicted that were taken on
a measurement table using a dial gauge. Control measurements with a straightedge showed a
measurement error of about 1/100th of a mm – accurate enough for such measurements.
All measurements were taken between the D-string and the G-string with the guitar laid on
the measurement table; the guitar body was (horizontally) pressed down onto the table. The
neck was not supported. The electric guitars were strung with a 009 – 046 string set (Ernie
Ball) and tuned to 440 Hz (regular tuning). The setup for the acoustic guitars included a 012-
string-set and also regular tuning. The graphs need to be seen as showing a randomly picked
sample; they are not necessarily typical for the respective type of guitar. Still, we can obtain
from this reference values about common curvatures and errors in the individual frets.
The USA-Standard-Stratocaster had been played little; it still featured the original frets.
The small warping at the 7th fret is found between the E-and the A-string, as well – it is
therefore not due to wear. The four graphs represent four adjustments of the truss-rod. The
Yamaha was brand new.

The neck of the Jazzmaster (built in 1962) had been re-fretted once (ca. 1969); the guitar had
been rarely played afterwards. The truss-rod was slackened, and with the 009 – 046 string-
set, no sizeable neck-curvature could be adjusted. When the guitar was built, a 012-string-set
was still standard. Note should be taken of the position of the apex: it was not found in the
middle of the neck (like in the other Fender necks) but around the 6th fret.
The Squier Super-Sonic is worn; its fretboard would need a dressing. The 100-Euro-Squier
is new and shows that an acceptable neck can be realized at very low cost (the loathsome
machine-heads being a different matter).
The Gibson Les Paul still has its original frets and (hopefully) a long life ahead. The ES-335
(built in 1968) received a new fret-job sometime and was played little afterwards.
The Duesenberg Starplayer TV is new; its frets were dressed on a CNC-sander. They are
perfect.
The Gretsch Tennessean (built around 1964) has seen a lot of action; it shares its fate with
the above Jazzmaster of similar vintage: the truss-rod is fully slackened (lower curve). The
guitar has presumable been re-fretted at some point – that job was done with poor quality.
The Ovation Viper EA-68 already stood out in Fig. 7.7; it does not ignite much enthusiasm
regarding the height of the (practically untouched) frets, either. The SMT, on the other hand,
had received a makeover by the distributor of Ovation, and is perfect within the framework of
the type-specific neck-shape. This guitar is not likely to be played very often on the highest
frets.
The Collings is of Texan nobility and under no circumstance wants to be confused with a
Collins. It (the Collings) is perfect – as is its price.
More than double the cost of the Collings (and still almost brand new) is the blue Personal-
Taylor, with its 13th fret marching to a different drummer (no wonder given the number). So
what – you won’t want to press down 12 strings up there anyway. In Fig. 7.7, this was the
landmark guitar.
The almost new Martin D-45V affords itself a swerve at the 16th fret – that is certainly not
the result of excessive use. At the given price, something like that should not occur … but it
doesn’t really get in the way, either.

Fig. 7.150a: Fret-heights of various guitars.

Fig. 7.150b: Fret-heights of various guitars.

7.14 Vibration-undamping
Does the sound of an electric guitar improve if it is subjected to loud music for a day?
Apparently so: more than a few people use this process to shorten an extended “playing-in”
process. Others send their guitar off to Emil Weiss (G&B 12/08), who subjects it to a high
vibration load (not without charging 770 €) – and voilà: the body can resonate much more
freely. The basis of this shaking of guitars is a process that Gerhard A. von Reumont had
patented in 1980 (German patent no. 27 12 268): Process and apparatus for improving the
resonant behavior of resonance bodies in musical instruments. Mr. Weiss does not really
explain in detail what he does to the guitars, and he does not mention Mr. von Reumont as the
originator, either. However, the pictures published in G&B do have a striking resemblance to
those in von Reumont’s book [ISBN 3-87710-173-9]. The latter does cite a publication co-
authored by E. Weiss and A. von Reumont – thus it’s probably the same process. The
underlying patent for the process “has by now expired”, as von Reumont states in his book
published in 1996. Correct: the patent office confirms that the protection became void in 1984
due to lack of payment of the fees. Which is by no means an indicator for bankruptcy but
could have been connected to the annually rising patent fees.
Von Reumont writes that the transfer of the string vibration to radiated airborne sound is
improved by the vibration-undamping – he does not mention electric guitars or basses. His
investigations relate to upright bass, cello, acoustic guitar and piano, i.e. instruments the
efficiency of which depends on resonances of the instrument body. We learn that wood would
be strongly deformed during the construction of the respective instrument, and that this
deformation hampers free vibration. "The effect of the vibration-undamping is based on
relaxation processes caused by frequently repeated over-stressing." The over-stressing is
taken care of by an electric motor fitted with a heavy spot (i.e. artificially unbalanced) and
mounted to the bridge of the instrument such that the latter is given a good shake for hours on
end. Does that actually work? Apparently so – the evaluation by musicians was “entirely
positive”. Given such overwhelming evidence we certainly can grant, for once, an advance in
terms of trust – but we still kindly ask for some quantitative data. These are found, as well:
e.g. for an acoustic guitar, "14.5% decrease in damping at 85 Hz" are noted. The damping,
that’s the resistive component in a resonating system – in textbooks we also find the damping
constant, or the degree of damping, or the damping coefficient; mechanical engineers have
created different terms, and on top of them the electrical engineers, as well, and the physicists,
too … it’s almost a little Tower of Babel. For the term damping, the Schalltechnisches
Taschenbuch (Pocketbook of Sound Engineering) references “irreversible processes that
transform part of the motion energy into heat (-losses)”. That is not something you want in a
musical instrument, and therefore the damping is reduced via vibration – by exactly 14.5% for
the example mentioned above. How does one measure this value with such accuracy? By
measuring the power consumed by the electric motor, advises von Reumont – before and after
the treatment. If, after a 2-hour shake, the power consumption falls by 14.5%, the instrument
has been un-dampened by 14.5% (at this frequency). Good riddance, then? Not quite, there is
some truth in this – a tiny truth, but still …
Electric motors are electro-mechanic transducers that transform electrical energy into
mechanical energy. Not exclusively, because heat also is generated – but mainly it is
mechanical energy. The latter is needed to overcome friction in bearings, swirl the
surrounding air, or – if there is a load – to create torque at a connected shaft. All these
mechanical energies do not appear out of thin air but are drawn from an electrical source
(battery or mains power supply), and thus the following holds: if the mechanical load
changes, so does concurrently the electrical power consumption.

7.14 Vibration-undamping 7-171
From this energy balance, von Reumont derives the assumption that the reduction of power
consumption of the electric motor observed over hours can only be due to the decrease of the
power drain, and that this can only be due to the reduction of the damping in the instrument
material. So, if the power consumption falls by 14.5%, the damping in the instrument must
also have reduced by just this 14.5%. Enter: the devil’s advocate.
1) The power consumption also depends on the resistance of the motor winding (copper wire);
this resistance changes as the motor heats up – by +5% for a temperature rise of a mere 13°C.
2) The friction losses in the motor bearing are temperature-dependent, as well, and anyway
also dependent on wear – given the unbalanced loading not an unimportant aspect. Von
Reumont reports frequent motor changes, and service lives of only 150 hours.
3) Suppose the mechanical power does actually decrease by 14.5% – why would then only the
undesirable losses (the damping) have dropped? It could also be that the desirable power – the
radiated sound-power – has decreased, couldn’t it?
To prove the effectiveness of his process, von Reumont cites an investigation by the PTB
(Physikalisch Technische Bundesanstalt – a governmental technical/physical authority in
Germany), in which an upright bass was analyzed before and after treatment. Result: in
several narrow frequency bands, there is a rise in level by about 3 dB, and at 2.5 kHz it’s even
5 dB. Unfortunately, the PTB does not measure the power fed to the bass, but merely operates
the LDS-shaker from a constant voltage source. That is too bad: in 1978, the Type 8001
impedance head by B&K had been on the market for years – it would have been a hassle-free
measurement. Still: there’s more output – now officially confirmed.
Because of the extremely strong vibrations remove the tensions dwelling within the
instrument, “the wood now sounds as if it had been seasoned for a long time.” Whether the
bridge has been slightly moved as the eccentric tappet was mounted, and whether the sound
post within the instrument has shifted a bit – no, that is unfortunately not checked. But let us
by all means insinuate that there are indeed changes of some kind in the wood. Five hundred
enthusiastic musicians can’t be wrong. Also, the errors that von Reumont made in terms of
the physical magnitudes are not the end of the world: on a global scale they are more a
petitesse (3 dB does not indicate doubling of the SPL, and energy is not current x voltage, and
Watt is not the unit for energy, either).
However, measurement accuracy (or rater in-accuracy) occurring in his setup requires a
close look. After all, we do learn in the basic course for instrumentation: if a result is given
with three digits, the input quantities need to be similarly accurate. These input quantities are
electrical voltage and current at the motor. "The accuracy of the reading is 0.2 V for the
voltage and 10 mA for the current. This is adequate for normal treatment." The datasheet for
a treated acoustic guitar shows voltages between 0,9 and 2,8 V, and currents between 480 and
1180 mA. Specifically: at 85 Hz the voltage decreases according to the measurement log
(after 120 min) from 1,6 to 1,5 V; the current increases from 860 to 780 mA. Ergo: 14,5%
power decrease, and thus 14.5% un-damping. However, unfortunately there is also a
measuring error range of 14.7% – which puts things a bit into perspective. Von Reumont
repeatedly notes that his instrumentation equipment is of hobbyist-grade. In itself a laudable
approach: every well-versed hobbyist should be put in the position to assemble an un-
damping setup. Even a source for the motors is given (at volume less than 1 Deutschmark),
and the schematic for the power supply is included, as well. If the motor does not run
smoothly, you reverse it, or give it a quick spin at high revs, if necessary.

The clutch is constituted by a valve rubber (no, that hopefully will not have a damping
effect…), and if worse comes to worst, the motor is exchanged. And then you carry on
measuring, don’t you?
The data for the following graphs were taken from the tables published by von Reumont. For
the guitar, we find a vibration-undamping of 5,1% at 10600 min-1 (= 177 Hz). However, given
the established measurement tolerance, an un-damping of 24% could have been explained just
as well, or … no effect at all.
Fig. 7.151: Un-damping measured by von Reumont on an acoustic guitar (left) and a cello (right). Dashed line:
corresponding maximum measurement tolerance established according to von Reumont.
As we enter the world of chance and coincidence, we of course also have to also concede that
the biggest possible measurement error will not happen for every measurement. Indeed, for
larger instruments and the corresponding higher motor voltages, the reading tolerances
suddenly drop to the extent that a viable significance appears: yes, the power consumption
does drop over time, for whatever reason. Weren’t there actually not any more precise
voltmeters available back then (around 1989)? There were, but on the digital display, “the
numbers were passing through so quickly that a readout could barely be taken, or even not at
all.” That is because the motor shaft wobbles a bit, and the passing resistance of the carbon
brushes is not time-invariant, either. And yes, here’s another fact: the power consumption will
also depend on the latter issue. And on the damping-losses of the bridge adapter made from
small boards (with sticky velvet-foils). And on the foam material clamped below the strings.
True: it is not easy to convince a skeptic. It seems it is easier to convince the upright-bass
players, “since they are rarely happy with their instruments.” And who knows, maybe the
advertised un-damping process does work, after all♣.
For hollow-bodied instruments, that is – the wallings of which need to vibrate! Given that,
why does Mr. Weiss include, in his reference list, Stratocasters vibrated by unbalanced
electric motors? … Strats, the bodies of which are not supposed to vibrate at all (as noted by
the wise Mr. Fender – and, for once, here he is correct). Only heaven knows … where, by the
way, L.F. is assumed to reside according to popular belief. Santo subito – for believers. For
astronomers, though: L.F.’s accommodation may be in the Leonids, rather.
Bottom line for the electric guitar: much noisy ado about nothing? Wrong: it will set you back
€ 770.-
♣
Once you’ve shelled out 770 Euro ...

8 Psychoacoustics
Musical notes are both sound events and auditory events – at least from the point of view of
the perceptional psychologist. It is not denied here that these musical sounds may be – in the
holistic-philosophical sense – even more than that. The sound event: that is the musical
sound from the physical perspective. It is characterized by its physical parameters such as e.g.
frequency, level, spectrum, or envelope. The investigation of individual physical parameters
in isolation will, however, not give any information about the auditory perception: a tone of
40 dB level is audible under normal conditions if its frequency is 1 kHz, but at 50 kHz is will
be inaudible. The perceived sound volume (psychoacousticians us the term loudness)
therefore is not equivalent to the sound level. In the present context, the second syllable –ness
in the term loudness is intended to indicate that not a physical quantity is meant but one that is
connected to the auditory event. The science of psychoacoustics describes the functional
connections between the parameters of the sound event and those of the auditory event. In
other words, this science seeks to e.g. find out how the loudness may be calculated from the
physical parameters. Besides the loudness, there are many more parameters (features,
attributes) of the auditory event, examples being timbre, subjective duration, or pitch. The
latter term in particular is often erroneously seen as equal to the frequency. A closer look
shows that this is not tenable: the tone of 50 kHz can be precisely defined in terms of
frequency, but it cannot be assigned any pitch at all because it is inaudible. The 1-kHz-tone,
on the other hand, does generate a pitch, but the latter will be – despite constant frequency –
dependent on the sound level, i.e. it is not constant.
Not every audible sound may be assigned a pitch: a voiceless f, for example, is perceived as a
broadband noise lacking any pitch. Guitar sounds, on the other hand, are characterized by
strongly pronounced pitch (although there are exceptions here, too). In a simplified
consideration, the pitch of the guitar tone is matched to the fundamental of the string. This
may lead to the following definition: the pitch of the A-string amounts to 110 Hz. However,
the unit Hz is for frequencies and not for pitch. So how can we quantify the (subjectively)
perceived pitch with sufficient accuracy? A frequently used method would be the comparison
with a pure sine tone. The test person (in such experiments usually termed subject) alternately
listens to the tone to be assessed and to a sine tone, and adjusts the frequency of the sine tone
such that both sounds generate the same pitch. The frequency of the sine tone, given in Hz,
may now be used as measure for the pitch of the tone to be evaluated. This methodology is
sufficiently accurate for the following observations; in scientific explorations, more elaborate
procedures are applied, as well. If we have the pitch of the A-string evaluated with the above
method, we indeed obtain a frequency of the comparison tone of 110 Hz – but there are small
yet significant differences. One reason for the deviations is the dispersion appearing in string
oscillations, and the resulting inharmonicity of the partials. Moreover, the interaction of the
partials in the perception process also plays a role.
While the first chapters in this book were dedicated to the physical principles of the sound
generation, we will now focus on the auditory event. Actually, the guitar is not just a sound
generator, but indeed a musical instrument. For extensive presentation and derivation of the
fundamentals of psychoacoustics, reference is made to the literature cited in the annex.
© M. Zollner & T. Zwicker 2004 – 2020 Translation by Tilmann Zwicker

8-2 8. Psychoacoustics
8.1 Tonal systems
The strings of the guitar are tuned to E-A-D-G-B-E (for standard tuning). These notes are
(inter alia) elements of a musical scale that itself is an element of a tonal system (other
expressions used are pitch space, or system of tonality). The latter is understood as the
(theoretically unlimited) set of all ordered notes, and includes defining the individual
distances between pitches. In Western music, the tonal system with 12 steps is predominant,
with the musical scale formed of 12 notes. Deviating from this system are, for example, the
pentatonic system (based on merely 5 notes), and the diatonic system. The distance between
the notes (the frequency relationships) can be derived from the rules of the tonal system, and
from this we obtain the design rules of the guitar, and the tuning rules of the individual
strings. In this context, basic knowledge of vibration engineering proves to be helpful.
It is mainly transversal waves that propagate on the guitar string; they are reflected at the
termination of the free string (nut, bridge). A single-frequency excitation of the string leads to
particularly strong vibration patterns at specific frequencies (Eigen-modes at the Eigen-
frequencies i.e. natural modes at the natural frequencies). The lowest frequency at which such
an Eigen-mode occurs is the fundamental frequency of the string. In a simplified view, all
higher Eigen-frequencies are integer multiples of this fundamental frequency; a more detailed
analysis shows a slight spreading of the frequencies (see Chapter 1).
Fig. 8.1 shows the first three Eigen-modes of an ideal string vibrating in a single-frequency
fashion. If the excitation of the string is not with a single frequency but with a plurality of
frequencies (e.g. via an impulse), the superposition of many of these Eigen-modes may lead
to the formation of a complex vibration-pattern. Each one of the Eigen-modes (in theory
there is a an infinite number of them) is characterized by four individual parameters: its
Eigen-frequency that for the n-th Eigen-mode corresponds (in the dispersion-free string) to
the n-fold fundamental frequency (n being an integer number); its amplitude and phase, and
its direction of vibration. Of these 4 mode-specific quantities, only the frequency shall be
considered in the following. Arbitrarily choosing 100 Hz as the fundamental frequency, the
frequencies of the higher-order partials (n > 1) are 200 Hz, 300 Hz, 400 Hz, etc. Halving the
length of the string while maintaining an equal tension-force yields twice the fundamental
frequency, with the frequencies of the partials now 200 Hz, 400 Hz, 600 Hz, 800 Hz, etc.
Fig. 8.1: The first three Eigen-modes of an ideal string.

Left: fundamental (1st partial), center: first overtone (2nd harmonic); right: 2nd overtone (3rd partial).
The individual partials do generate individual auditory perceptions in the sense that a
multitude of tones becomes audible as a single string is plucked. Rather, the pitches of the
partials (perceived on a largely subconscious processing plane) blend to form a single pitch of
the string, with only this pitch being perceived consciously – given favorable conditions. The
pitch of the plucked string corresponds approximately to the pitch generated by the
fundamental vibration, but it is not identical♣. There are small deviations between the two –
but for our first basic considerations the deviations shall not be regarded.
♣
The higher-order partials (overtones, n > 1) change the pitch of the string only to a minor degree, but they do
contribute substantially to the timbre – which is not considered here.
Translation by Tilmann Zwicker © M. Zollner & T.Zwicker 2004 – 2020

8.1 Tonal systems 8-3
The string mentioned in the example having a fundamental frequency of 100 Hz, and the
string shortened by half (fundamental frequency 200 Hz) each generate a tone designated
T100 and T200, respectively. Played one after the other in direct comparison, T100 and T200
sound very similar – this is not actually surprising since the frequencies of the partials
contained in T200 represent a subset of those contained in T100. This example may be
extended by subjecting the halved string (T200) to another halving (T400). The resulting
frequencies of the partials (400 Hz, 800 Hz, 1200 Hz, etc.) are again a subset of the
frequencies of the partials contained in T100 and T200. Further halving of the string length
gives corresponding results. All notes generated by such halving (or doubling) sound very
similar, although their pitches differ markedly. Since the frequency relation generated by
halving and doubling of the string lengths (2:1 and 1:2, respectively) are designated octaves
in the musical context, the resulting notes are called octave-related. The high degree of
auditory relationship between two notes distanced by an octave has led to designating such
notes with the same letter. For example, the reference note used for tuning to standard
(“concert”) pitch is internationally as a rule designated A4, with the note one octave above
being designated A5. However, depending on the national context there are also variations to
this system of designations, e.g. a1 (or a'), and a2 (or a''), respectively.
8.1.1 The Pythagorean tonal system
Continued halving of the string-length is a first step towards generating related notes of
differing fundamental frequency. Following this approach, we find notes with corresponding
frequencies of partials also when reducing the string-length to one third. The partials of the
resulting note (designated T300) are located at 300 Hz, 600 Hz, 900 Hz, 1200Hz, etc.
However, compared to T200 now only the frequencies of every other (even-numbered) partial
is in correspondence, namely 600 Hz, 1200 Hz, etc. (Fig. 8.2). The fundamental frequency of
the string reduced to 1/3rd in length relates to the fundamental frequency of the halved string,
as would 3:2; this frequency relation (frequency interval) is called, in musical terms, a fifth.
For the associated notes, the concept of fifth-relationship is derived from this. Compared to
the octave-relationship, the fifth-relationship is less pronounced.
Fig. 8.2: Spectra of partials of

strings with the relative lengths:
L1 = 1, L2 = 1/2, L3 = 1/3.
Abscissa: normalized frequency;
ordinate: amplitudes (arbitrary)
Applying jumps of fifths and octaves in combination allows for the generation of a multitude
of notes that all are more or less related. Already in the ancient world a tonal system (among
many others) was constructed from octave- and fifth- intervals; after its protagonist
Pythagoras (ca. 530 B.C.), it is named the Pythagorean tonal system. In theory, an infinite
number of different notes could be generated with it. However, in practice we arrive at a
prominent end point after 12 jumps of one fifth each: after 12 subsequent intervals of one fifth
each, the resulting frequency relationship is 1,512 = 129,746. This brings it close to the 7th
octave, the frequency relationship of which amounts to 27 = 128.

The small difference between these two values of 129,746 / 128 = 1,0136 is called the
Pythagorean comma in music theory. From the sequencing of fifths, and from octave shifts,
all notes of Western music can be generated. In this approach, the frequencies of the notes
positioned at a distance of a fifth are shifted by a number of octaves until all frequencies are
located within one base octave. Starting from the arbitrarily chosen initial frequency 100 %,
the following rounded (!) frequencies result (in order to be able to more easily interpret the
frequencies, they are given in % to begin with; the corresponding frequencies are listed in
Chapter 8.1.3):
100 – 150 – 225 – 338 – 506 – 759 – 1139 – 1709 – 2563 – 3844 – 5767 – 8650 – 12975 %.
100 150 113 169 127 190 142 107 160 120 180 135 203 %.
C G D A E H F# C# G# D# A# E# B#
The first line in this table holds the ascending frequencies of fifths, the second line includes
the corresponding frequencies in the base octave. The designation of the notes is given in the
third line (# stands for ‘sharp’). For example, 2563 % needs to be shifted (towards lower
frequencies) by four octaves in order to arrive at 160%: 2563 / 24 = 160. Rearranging the
frequencies in the second line in monotonously ascending order, the sequence of frequencies
of a scale results (values rounded off):
100 – 107 – 113 – 120 – 127 – 135 – 142 – 150 – 160 – 169 – 180 – 190 – 203 frequency / %
C C# D D# E E# F# G G# A A# B B# note-designation
Besides the ascending sequence of fifths, the descending sequence of fifths may also be
generated: again neighboring notes are fifth-related. In correspondence to the example above,
the initial frequency 100% would have to be repeatedly divided by 3/2: 67 %, 44 %, etc. With
suitable octave shifts (towards higher frequencies), again a scale results – with calculated
frequencies that slightly differ from the ones given above, though.
In the classical Pythagorean tonal system, not all of the notes calculated above were
employed. Starting from the keynote C, users made do with 5 ascending fifths (C-G-D-A-E-
B) and one descending fifth (F). They were able to form a scale that way:
1 Q2/2 Q4/4 Q-1⋅2 Q Q3/2 Q5/4 2

C D E F G A B C'
1\1 8\9 64\81 3\4 2\3 16\27 128\243 1\2
In this table, Q represents the interval of the fifth♣ (frequency ratio 2\3); the corresponding
exponent indicates the number of the required jumps of a fifth each. From the denominator,
we can take the number of the additionally required octave shifts. Q5/4 indicates 5 fifth-jumps
towards higher frequencies, and subsequently 2 octave-shifts (22 = 4) towards lower
frequencies. The third line yields, referenced to the keynote, the frequency relation as a
fraction. The notes of the scale given above, and their frequency relation (interval), is
designated according to their place number:
C = prime, D = second, E = third, F = fourth, G = fifth, A = sixth, H = seventh, C' = octave.
♣
To specify the frequency relations in an interval-designation, two different styles are customary: for the fifth
e.g. 2:3 but also 3:2. Both relations are self-explanatory, while the letter-designation (C-G) does not
unambiguously identify which one of the two is the lower note. In the following, the lower note is always
positioned first (to the left) as is usual for axis-scaling. However, following through with this train of thought
would result in fractions that are smaller than 1, such as e.g. fC1 : fG1 = 2:3 = 0,666... While this representation is
in itself correct, it is in contradiction with the practice of indicating intervals with number that are larger than 1.
This contradiction is resolved in the following via using the back-slash (as used in Matlab): fC1 \ fG1 = 2\3 = 1,5.

The terms are related to numeration in Latin: primus, sekundus, tertius, quartus, etc. In its
precise meaning according to the theory of harmony, these expressions designate the
distances between two notes (inter-vallum = space between palisade beams), but in everyday
use they also represent the names of notes: the fourth on the C-scale is an F. Distance in the
above sense means to indicate the distance to the root note i.e. the ratio of the frequency of the
note in question (e.g. an F) to the frequency of the keynote; in this example it is 3\4,
corresponding to a fourth. It is also possible to form the ratio of two notes directly
neighboring on the scale; this yields:
fC \ fD = 8\9; fD \ fE = 8\9; fE \ fF = HT; fF \ fG = 8\9; fG \ fA = 8\9; fA \ fB = 8\9; fB \ fC' = HT;
Of these 7 frequency ratios, 5 correspond to a so-called whole-step (‘whole note’, ‘whole

tone’), specifically C-D, D-E, F-G, G-A, A-H. The remaining two intervals of neighboring
notes are half-steps (HT, ‘half notes’, ‘semi-tones’, ‘half-tones’). In Pythagorean tuning, the
frequency ratio in a whole step amounts to 8\9 = 1,125, and the one in a half step (E-F, B-C)
HT = 243\256 = 1,0535. The resulting scale is called diatonic scale because it is comprised of
two different steps (namely whole-step and half-step). As supplemental information,
‘Pythagorean tuning’ should be indicated – there are many different tunings, after all.
N.B.: with respect to the note that is internationally designated B, there is a particular
idiosyncrasy when the German language is used: there, this note is designated H. Originally
(in fact: obviously), letters (starting with A) formed the names of the notes in the scale: A-B-
C-D-E-F-G. However, medieval hexachord theory required (on top of the B as mentioned
above) a second note half a step lower. In order to distinguish between the two, the
designations B-quadratum (B-durum) and B-rotundum (B-molle) were introduced – derived
from the angular (hard, durum) and round (soft, molle) writing styles of the letter b. The
angular b mutated to an h … and now musicians in Germany, Austria, and the German
speaking part of Switzerland found themselves with a peculiarity that continues to lead to
(sometimes serious) complications when communicating internationally.
The diatonic scale as introduced above consists of 5 whole-steps and 2 half-steps. Each one of
the whole-steps can pythagoreically be subdivided into two half-steps – however this may be
done in two different ways. In the international note designations, half a step upwards is
indicated with adding the syllable “sharp” to the note, and half a step downwards by adding
the syllable “flat”. The diminished D is called D-flat (Db, with the b standing for
‘diminished’), the augmented C is C-sharp (C#). It has already been shown that all notes can
be generated by using upwards-fifths and downwards-octaves in the Pythagorean sense:
C–G–D–A–E–B–F#–C#–G#–D#–A#–E#–B#.
However, all notes may just as well be generated via downward-fifths and upward-octaves:
C–F–Bb–Eb–Ab–Db–Gb–Cb–Fb–Bbb–Ebb–Abb–Dbb.
The notes Bbb, Ebb, Abb and Dbb result from diminishing B, E, A, D by two half-steps,
respectively.
Fig. 8.3 shows the keynote frequencies of these two Pythagorean-chromatic scales. Due to the
Pythagorean comma, no frequencies in a pair in the sequence of upwards-fifths and
downwards-fifths are the same (except for the starting pair). If we limit ourselves to
diminishing by a single half-step, a scale of 21 steps results: each of the 7 diatonic steps C-D-
E-F-G-A-B is allocated a lower and a higher half-step. This 21-note tonal system was actually
the basis for keyboard instruments – however it was deemed too complex.

Many musicians therefore simplified the scale by enharmonically equating similar notes. The
resulting 12-step Pythagorean-chromatic scale is indicated on the top of Fig. 8.3 via
squares. Only a single half-step is introduced each between all whole-steps, but the half-tone
distances are of different size, as is clearly visible ( –– ).
C C# D Eb E F F# G G# A Bb B C’
C D E F G A B C’
Fig. 8.3: Fundamental frequencies of the Pythagorean-chromatic scale, shown on a logarithmic frequency axis.
Δ = deduced from the first 6 upwards-fifths jumps; ∇ = deduced from the first 6 downwards-fifths jumps;
, = remaining 7 fifths jumps;  = used in medieval times as chromatic scale.
The scale with equal temperament developed around 1700 is indicated with dashed vertical lines (8.1.3).
The different half-step distances complicate changing keys: the second (C-D) based on the
keynote C has a larger frequency difference that the one based on C# (C#-Eb), and other
intervals (e.g. C-E, G#-C) meet a similar fate. Depending on the specific case, the flawed
consonance when two notes are played simultaneously may be another problem. The
fundamental thought behind the Pythagorean tuning was the note-relationship based on fifths
and derived from the sequence of partials. Well meant that is – but you know how things are
with relatives: as the distance grows, the similarities wane. Fig. 8.4 schematically shows the
frequencies of the partials for the prime (C) and the third (E). If, in simultaneous playing of
the two notes, individual partials get to lie (frequency-wise) in immediate vicinity, beats may
become audible. An example would be the 5th partial of the prime (C) and the 4th partial of the
Pythagorean third (Ep).
Fig. 8.4: Spectrum of partials of the notes C (prime) and E (third). Beats are generated between the 5th partial of
the prime and the 4th partial of the Pythagorean third (EP), due to the small frequency difference. For the pure
third (ER) the corresponding frequencies of the partial are identical. Abscissa: normalized frequency of the
partials of the prime.
Beating happens when two mono-frequent notes of equal amplitude and similar frequency are
played at the same time (i.e. they are added). Every note from a guitar consists of a multitude
of (mono-frequent) partials, each of which is, individually considered, sine-shaped (a cosine-
oscillation has the shape of a sine, as well). The 5th partial (= 4th overtone) of an ideal string
vibrating at 100 Hz has the frequency of 100 Hz x 5 = 500 Hz, the 4th partial of the third
according to Pythagorean tuning is at 126 Hz x 4 = 504 Hz. The frequency difference of the
two partials is 4 Hz.

If we now regard merely the oscillation of the sum of the two partials, a figure similar to Fig.
8.5 (center) results. The phase difference of the two partials fluctuates with the rhythm of the
difference frequency, and amplification and cancellation alternate with the same rhythm.
Given sufficient levels, one single partial with rhythmically fluctuating (i.e. beating) loudness
is heard rather than two partials of almost equal pitch.
Interpreting the summation-curve (middle section of Fig. 8.5) is facilitated by reformulation

towards a multiplicative operation:
In this product-representation, fΣ stands for the frequency of a cosine-oscillation with its

amplitude changing “with the rhythm of the difference frequency fΔ”. The above example has
fΣ = 502 Hz, thus it lies exactly in between the primary frequencies f1 und f2. The term
“difference frequency” should be used with care: it is calculated as fΔ = 2 Hz, this is half the
frequency distance between f1 und f2. However, the maximum of the beat-envelope appears
(amount!) with double this frequency i.e. twice per fΔ-period. The above beating with 500 Hz
and 504 Hz as primary frequencies may therefore be seen as a tone at 502 Hz featuring 4
envelope maxima and 4 envelope minima per second. It therefore becomes louder and softer 4
times per second. The auditory effect of a beating of partials is difficult to predict – it may
even be inaudible (despite its physical presence) due to masking by neighboring frequency
components. If it indeed is audible, it may sound pleasant or displeasing. During many
centuries the opinion was held that any beating of partials is undesirable, resulting in the beat-
free just intonation (Chapter 8.1.2).
Fig. 8.5: Two cosine oscillations (top, bottom) slightly different (5%) in frequency, and their sum (middle). The
curves are of equal phase at the left and right boundaries of the figure, and in opposite phase in the middle.
Same-phase addition results in doubling of the amplitude (constructive interference), opposite-phase addition
leads to cancellation (destructive interference). Abscissa: time.
8.1.2 Just intonation
In this context, harmonic and natural also stand as synonyms for just – the rationale being
that nature herself allegedly had shown the way in the form of integer frequency ratios of the
partials. The term divine tuning therefore is not far off, creating work for philosophers and
esoterics, but mainly for mathematicians ... who not necessarily were musicians.

Just intonation – the term rings of teachings of justice & purity, and expressions such as
fairness, correctness, or well justified come to mind – the opposite of unjust, wrong, or
unjustified, and thus anything not conforming to the just intonation could in any case only be
heresy. It is easily imaginable how hordes of mathematicians have deduced justifications for
this or for that intonation … generating tables with an accuracy up to 12 figures! Or, rather:
tables with 12 decimals, since the actual accuracy may have been a bit of an issue [Barbour].
Irrespective of any (not infrequently occurring) calculation- and rounding-errors: given a 1-m-
long monochord string, 12-decimal-accuracy implies a length-tolerance of no more than 0,001
nm. Just to compare: the wavelength of visible light amounts to around 600 nm. Specifying
pitch deviations with an “accuracy” of 1/10000000th of a cent is similarly nonsensical.
The just intonation may be traced back to ancient times. Two doctrines of thought emerged
from the Pythagorean school (that originated around 530 BC): the canons regular (canon =
rule, law) advocated the conservative opinion, while the harmonists gave priority to
euphony, even if that required modification of mathematical laws of nature. The canonical
doctrine regarded the frequency ratio 6 : 8 : 9 : 12 as “holy matrimony” between the forth and
the fifth (Fig. 8.6) with the major second (full step F-G) being the result. Simbriger/Zehelein
give an astounding assessment for this approach: we have already met this grouping of notes
in primitive music; with the Pythagoreans, we find that same basic occurrence substantiated
and sanctioned with the background of advanced civilization. There you have it: if – as a
musician or listener – you recognize certain intervals as harmonic/consonant, then that’s
primitive ado. However, if you smudge some divine-cosmic-mystical mumbo-jumbo around
that finding, it takes its place in high culture.
Fifth 2:3 Fourth 3:4 Fig. 8.6:

Fourth 3:4 Fifth 2:3 The "holy matrimony"
Still: despite some massive mystical sanctioning it was not possible to hide that the use of
Pythagorean intonation made some chords sound less than pleasant. Young J.-apprentice: "oh
honorable master Y.: them chords, they will not sound – try as I might! Those fifths and
thirds, they fail to soothe us.” Y.: “Do or do not: there is not try … but quiet now be, young
one; in a special realm here taken we are. Let be it, for divine this is – of The Force” . Many
will have conformed to this sage advice from a long time ago and a galaxy far, far away …
but some went public. In the olden days, on this planet, that could well lead to premature
termination under artificially elevated ambient temperature – or it could open the door to
eternal fame and glory. Or both. Didymos (Didymus) and Ptolemy, Alexandrinian savants by
trade (and, to begin with, both by all means proponents of the Pythagorean third), evidently
found the silver bullet (at the time probably the silver arrow). They replaced the Pythagorean
third (based on the divine fifth) by an at-least-as-divine relation of whole steps: the major
third – in Pythagorean intonation the frequency interval 64\81 = 1,2656 – was shifted to 4\5 =
1,2500 in the so-called Alexandrinian system. Didymos borrowed the minor third (27\32 =
1,1852) from the Pythagorean system, and Ptolemy modified it to 5\6 = 1,2000. In principle,
anyway. Looking closer, we find [e.g. in Barbour] two didymian intonations, and no less than
7 ptolemyan intonations. Nevertheless, the foundation block for the just intonation was laid.
Studying literature, it is easy to come to the impression that (as mentioned above) something
divine is connected to the just intonation. However, as confusion grows, the realization does
manifest itself that it must in fact be a kind of polytheism. Barbour defines just intonation as:
based on octave (1\2), fifth (2\3) and major third (4\5); the intervals themselves are designated
just (or pure), as well.

Elsewhere, however, Barbour extends the term just intonation to: based on octave (1\2), fifth
(2\3), fourth (3\4), major third (4\5), and minor third (5\6). Other authors even designate as
pure intervals all intervals the frequency ratios of which correspond to the whole-numbered
ratios of the frequencies of the first 16 partials. All intervals? Well, almost … those ratios that
fit to some degree, anyway. But not the 7th , 11th , 13th , and 14th partials! Of course not.
Valentin substantiates: the miraculous, natural, and therefore not worked-out order of the
whole system stems from the sequence of the composition of just intervals contained in these
notes that – with a suitable octave transposition – yield our whole scale system. The 7th, 11th,
13th, and 14th partials are the “black sheep”; nature finds space for something like that, too.
Only for C-F# (or C-Gb) no fitting frequency ratio at all could be found in the natural order.
Therefore, the devil had to be called in as the usual suspect – only he/she could have
smuggled in such an inconvenient, devilish interval (Tritonus, Diabolus in Musica). The
question: “how could God allow this …” again created many workplaces for philosophers
(compare Theodizee), but this would go beyond the scope of scientific considerations.
The just intonation derives its rationale from the whole-numbered frequency ratios of the first
16 partials. But why exactly 16 partials? That’s because the 16th partial is exactly 4 octaves
above the fundamental. But why then not just 1 or 2 or 3 octaves? That would be because that
way you could not yet generate a chromatic scale. Moreover, wind instruments can just about
reproduce the 16 “natural tones” (Eigen-tones, partials). The peculiarity of the tritone with its
45\64-ratio was justified on the basis of this fact that about 16 but not those 64 Eigen-tones
could be generated. Fig. 8.7 shows the frequency ratios of a just-intoned scale. Besides the
devil’s interval, there indeed is nothing fishy in there: numerators and denominators are
integers between 1 and 16. The major third C-E that would with the Pythagorean intonation
carry beats – it now is beat-free (compare to Fig. 8.4).
1 15\16 8\9 5\6 4\5 3\4 45\64 2\3 5\8 3\5 5\9 8\15 1\2
C – Db – D – Eb – E – F – Gb – G – Ab – A – Bb – B – C'
C Db D Eb E F Gb G Ab A Bb B C’
Fig. 8.7: Just intonation (Mersenne’s lute tuning Nr. 2). The tritone was given also as F# with 32\45, for the Bb
also 9\16 are found instead of 5\9.
Besides C-E, the combinations F-A and G-B (with 4\5) also make for a beat-free major-third
interval. For the minor thirds, however, differences already appear now: E-G, A-C, and B-D
yield 5\6, but D-F yields 27\32. Looking at the fifth-intervals: C-G, E-B, F-C, G-D and A-E
yield 2\3, but D-A → 27\40. The whole-step intervals are at 8\9 or 9\10; the half-step inter-
vals on the C-major scale are at 15\16, with the remaining (chromatic) half-step intervals at
24\25, 25\27 or 128\135. Despite the legitimization by nature herself, this gave opportunities
for mockers: are you still learning, or do you play with a special intonation system?
It wasn’t that these dissonances remained hidden to the working musicians. The latter knew
about them, limited their music-making to a few keys, and tried to give a wide berth to the
howling-wolf intervals. Alternatively, instruments could be built that divided every octave
into 21 in-between notes. And if that didn’t suffice: J. M. Barbour lists a plethora of other
divisions, for example: the 31-division (Fibonacci-sequence), the 53-division (Bosanquet-
harmonium), and don’t you forget that the 118-division has both fifths and thirds that are
superlative (0,5 cent flat and 0,2 cent sharp, respectively).
There we are! … and there we go. On top of all that, other just (!) intonations were developed
for the twelve-section octave, as well – which makes Barbour infer: the just intonation does
in fact not exist; rather, there are many different just intonations, with the best being
the one that comes closest to the Pythagorean intonation.
As desirable as “just” (or “pure”) intervals may be in polychoral play: for intervals succeeding
each other errors do cumulate. Take, for example (and see Fig. 8.8), Jimi Hendrix’ “Hey Joe”
(to solidly arrive back again in more modern times): the accompaniment first climbs down a
minor third from E to C (or that could be interpreted as climbing up a minor sixth) and then
runs through 4 jumps of fourths: E → C – G – D – A – E. Given just intervals, one revolution
gets us this:
; C ↓ E = 5\4, fifth = 2\3, octave jump = 2\1.
C-major --------------------- G-major -------------------- D-major -------------------- A-major ------------------ E-major
Fig. 8.8: Jimi Hendrix / Noel Redding: bass chromatic in "Hey Joe".
On the basis of just intervals, the full revolution of a cadence (lasting about 12 seconds in the
original tempo) would lead to a frequency increase of 1,25% – after one minute, that would
already make for no less that half a step. To execute every revolution at exactly equal pitch,
e.g. the step from D to A (fifth) would have to be performed with the deviating ratio of 27\40.
That, however, would mean a conflict with the pure (just) school of the first 16 natural notes.
Another “law of nature” (one that chalked up some success in architecture) is the golden
section (or ratio). However, for Barbour the “golden tonal system of theoretical acoustics” is
worth only a few lines. His bottom line: a jack-o’-lantern (ignis fatuus).
8.1.3 Tempered tunings
In music, the term temperament is used synonymously with the term tuning. Tempered
tuning is no pleonasm, though. It is the technical term for tunings that, on a small scale and in
a targeted manner, deviate from global tuning rules. Early versions of the tempered tuning
may be traced back to Giovanni Maria Lanfranco (1533); starting from just intonation, he
proposed to slightly down-tune the fifths, and to up-tune the thirds just as much as was
tolerable (in terms of the perceived sound). During subsequent eras, there were countless
attempts to define this advice more precisely. Starting from empirical results (the fifth should
cause one beat per second), via graphical designs, nomograms, scary formula, and versatile
tables, the path led to the equal-temperament tuning that dominates today: the octave is
divided into 12 equidistant half-note-steps – and that’s it. That this seemingly simple rule has
not been in practice much longer – that is probably due to its demanding a readiness to
compromise. It does require, after all, detuning those just and highly consonant intervals (such
as the fifth). Not all musicians show a corresponding capacity for suffering: the cellist Pablo
Casals speaks of the brainwashing of the tempered tuning, and the violinist Carl Flesch
allegedly was unable to play together with a piano (in tempered tuning) subsequent to a
rehearsal with a string quartet.

Well, with direct access to the string and thus the pitch as a continuum, a violinist has the
advantage of a completely free-wheeling intonation. The piano does not offer this possibility.
If a growth of the number of keys into the infinite is to be avoided, the only remaining
solutions are a highly key-specific temperament, or a universal equal-beating (equal
tempered) tuning.
Intervals between notes are characterized by the corresponding frequency ratios. Using the
equal temperament, 12 similar half steps succeed one another within an octave, with a
geometric frequency sequence resulting: 1, HS, HS⋅HS, HS⋅HS⋅HS, etc. Here, HS indicates the
half-step interval, the 12-fold repetition of which yields the just (pure) octave: HS12 = 2. With
this, the frequency relation of directly neighboring notes (at half a step distance each)
calculates as:
Half-step interval in the equal temperament
The 12th root of 2 … that is an irrational number. In the actual sense of the word it is a number
opposing reason. That may also be why a queasy feeling crept up on many a music-theorist.
3\4 for the just-intonated fourth is specified by nature herself; the counterpart in equal
temperament, on the other hand, defies – with HS5 – all sanity. And yet the numerical
differences are not all that big: 3\4 = 1,33333... , HT5 = 1,33484... , that’s a gap of no more
than merely 0,1%. However when principles are at stake, the gods themselves fight in vain.
And sure: the differences may indeed be larger for other intervals. The following table lists all
notes and frequency ratios in the equal-temperament scale. Other notes are not defined i.e.
there is no distinction between C#/Db, E#/F, Abb/G, B/Cb, and so on.
C C#=Db D D#=Eb E F F#=Gb G G#=Ab A Bb B C'

0 1 2 3 4 5 6 7 8 9 10 11 12
1 1,0595 1,1225 1,1892 1,2599 1,3348 1,4142 1,4983 1,5874 1,6818 1,7818 1,8877 2
Table: Notes and frequency ratios in equal-tempered tuning. The second line yields the half-note steps, the third
yields the frequency ratios rounded to 4 decimal places. Reference = C.
In German-speaking lands, the term gleichschwebend (= with equal beating) could be

misinterpreted such that all intervals would cause similar beating. This is not the case. The
English designation EQUAL TEMPERAMENT is not self-explanatory, either. It is the half-note
steps that are equal (in terms of the frequency ratios), and not the beats. Also equal (in the
sense of relatively equal) is the distribution of the Pythagorean comma into all 12 jumps of
fifths. Occasionally, well-tempered is found as a synonym for equal tempered; this can
probably be traced back to J. S. Bach’s preludes and fugues that he published under the title
“The Well-Tempered Clavier”. However, presumably Bach’s instruments were not intonated
with equal temperament (equal beats), but according to Werckmeister. Andreas
Werckmeister (Musikalische Temperatur, 1691) had developed a tuning that comes close to
the equal-temperament tuning but is not identical. Already one century earlier (around 1596),
Simon Stevin had built a monochord the half-step frequency ratio of which corresponded to
the 12th root of 2 (i.e. 1,059…). Presumably this was the first such instrument in Europe
[Barbour]. Almost at the same time (around 1636), Marin Mersenne♣ carried out
comprehensive theoretical groundwork.
♣
1492 Franchinus Gafurius: Theorica musicae 1511 Arnolt Schlick: Book on organ-building
1533 Giovanni Lanfranco: Scintille di Musica 1544 Michael Stifel: Arithmetica integra, z.B. log
1596 Simon Stevin: Monochord mit HT = 21/12 1636 Marin Mersenne: Harmonie universelle
1691 Andreas Werckmeister, Musicalische Temperatur 1706 Johann Neidhardt: Gleichschweb. Temp.

In his chapter Equal Temperament, Barbour lists no less than overall 41 different tempered
tunings: eventual success had many parents that presumably had to fight vehemently for
recognition. Even today, bitter adversaries turn up who are bothered by beating, “unnatural”
intervals, while proponents of equal temperament revel in unlimited modulations. Guitar
players should better make sure they run with the latter group because their instrument is
manufactured using equal-temperament tuning.
In order to unambiguously define the whole relational range, it is also necessary to specify an
absolute value besides just the frequency relations of the notes on a scale. As the long-
standing reference (concert pitch), a1 – the so-called middle A (also designated a' or A4) – is
in service. Today, the standard tuning frequency is 440 Hz while in past centuries there were
significant deviations in the range between 337 Hz and 567 Hz. In Germany, the reference
was fixed to 422 Hz in Berlin in 1752. The year 1858 saw a proposal for international
standardization on the conference on concert pitch in Paris, followed – on the corresponding
conference in Vienna in 1885 – by the adoption of 435 Hz. On the ISA-conference in London
in 1939, this value was increased to 440 Hz, and confirmed in 1971 by an ISO-resolution
(ISO = International Standard Organization). In conjunction with the standardization, it was
suggested to use the reference pitch for interval signals in radio and television, and as dial
tone for the telephone. This was not a successful marketing idea: for the telephone, check
measurements in 2004 showed a 6% deviation. The following table gives some fundamental
frequencies for notes tuned to equal temperament; reference for A4 is 440 Hz.
C C#=Db D D#=Eb E F F#=Gb G G#=Ab A Bb B

523,25 554,37 587,33 622,25 659,26 698,46 739,99 783,99 830,61 880 932,33 987,77
261,63 277,18 293,66 311,13 329,63 349,23 369,99 392,00 415,30 440 466,16 493,88
130,81 138,59 146,83 155,56 164,81 174,61 185,00 196,00 207,65 220 233,08 246,94
- - - - 82,41 87,31 92,50 98,00 103,83 110 116,54 123,47
Table: Frequencies of tones tuned to the equal-temperament scale, referenced to A4 = 440 Hz; rounded to two
decimal places. The open strings on the guitar E2, A2, D3, G3, B3, E4 are in bold.
In order to obtain convenient specifications of small deviations from correct tuning,

Alexander John Ellis defined (in 1885) the cent as the (supposed) pitch-atom:
Interval =
1 cent amounts to 1/100th of a half-step, or to the 1200th part of an octave. The frequencies
2000 Hz and 2001,155 Hz differ by 0,058% i.e. by 1 cent. Simbriger/Zehelein cite Preyer
with the insight (questionable from a present-day perspective) that the hearing system was
able to distinguish 1200 pitch steps between 500 Hz and 1000 Hz. Presumably, many a
teacher scared away their pupils by demanding that the latter should be able to discern
intonation errors of a 100th of a half-step. Chapter 8.2.2 has more on this topic.

8.1.4 Intervals in the equal temperament
The interval (inter vallum = space in between) is the distance of two notes; expressed
numerically by the relation (ratio) of the frequencies of the corresponding tones. The names
of the intervals are derived from the place numbers within the scale – for the C-major-scale,
this implies: C = prime, D = second, E = third, F = fourth, G = fifth, A = sixth, B = seventh,
C' = octave. Between the 3rd and 4th notes, and between the 7th and 8th notes, we find a half-
step, all other notes are a whole-step apart each. In the equal-temperament tuning, a whole-
step consists of two equal-size half-step (HS). All intervals can be represented by multiples
of a HS:
Distance between notes (intervals) in the diatonic scale, represented by half-steps:

C-C = 0, C-D = 2, C-E = 4, C-F = 5, C-G = 7, C-A = 9, C-B = 11, C-C' = 12.
Intervals are not just definable as HS-multiples in their relation to the root note C of the C-
scale, but also between all notes: e.g. D-E = 2 HS, G-H = 4 HS, F-A = 4 HS.
By the subdivision of the whole-step into two half-steps, new notes are obtained; they are
designated by the chromatic sign relative to their neighbors: C# = C-augmented-by-one-HS,
and (in the equal-temperament tuning) identical to the Db = D-diminished-by-one-HS.
Corresponding: D# = Eb, F# = Gb, G# = Ab, A# = Bb. Equating the diminished notes and the
augmented notes (e.g. C# = Db) is called the enharmonic equivalent (or enharmonic
ambiguity). Out of experience, it appears that guitar players are more familiar with the
augment-sign (#) than with the diminish-sign (b), and therefore we will give the former
priority in the following. From the 7-step diatonic scale (C-D-E-F-G-A-B), a 12-step
chromatic scale emerged:
C – C# – D – D# – E – F – F# – G – G# – A – A# – B chromatic scale
Each hyphen in this sequence represents a HS; the size of an interval can therefore be easily
accounted for as HS-multiples. The regular numerals (second, third, fourth, fifth, etc.) are,
however, already used (up) for the 7-step major (diatonic) scale, and this led to a somewhat
confusing nomenclature: unison (0 HS, also called keynote or root), fourth (5 HS), fifth (7 HS)
and octave (12 HS) are designated as “perfect” intervals, even if their tuning is not “pure”
and free of beats! Caution is advised: C-G, for example, is designated a “perfect fifth” even in
equal-temperament tuning. All other intervals within the major scale are “major” and thus: C-
C = (perfect) unison, C-D= major second, C-E = major third, C-F = perfect fourth, C-G =
perfect fifth, C-A = major sixth, C-H = major seventh, C-C' = perfect octave.
Reducing a large (major) interval by a HS results in a small (minor) interval. To get there,
two possibilities exist: either the higher note is pushed down by a HS, or the lower note is
pushed up by a HS: C-Db = C#-D = minor second, C-Eb = C#-E = minor third, C-Ab = C#-
A = minor sixth, C-B = C#-H = minor seventh. If a perfect (or major) interval is enlarged by
a HS we have an augmented interval; if a perfect (or major) interval is reduced by a HS we
have a diminished interval. This results in two schemes:
diminished – minor – major – augmented (second, third, sixth, seventh)

diminished – perfect – augmented (unison, fourth, fifth, octave)

C-D# therefore represents an augmented second; in the sense of the enharmonic equivalent
within the equal-temperament tuning, however, it also corresponds to the minor third C-Eb.
Purists turn away in horror, but the pragmatist just deals with it in everyday life: "C-D# is a
minor third." Indeed, it is without purpose to ponder the differences between C# and Db when
working with equal-temperament tuning. Of course, singers or violinists (as an example) will
tend to intonate the augmented notes (#) slightly higher and the diminished notes (b) slightly
lower, but that is then outside of equal-temperament tuning. When playing chords, the guitar
player (and we are concerned with the associated instrument here, after all) has hardly any
possibility to modify individual notes within the chord in their pitch. When playing single-
note melody, higher-order knowledge of harmony could be put to use – unless the keyboard
player in the band with his/her equal-temperament tuning shoots that down.
The following list gives an overview for all intervals, in this case referenced to C; with these
representations: p = perfect, d = diminished, mi = minor, ma = major, a = augmented:
d-octave: C-C'b p-octave: C-C' a-octave: C-C'#

d-seventh: C-Bbb mi-seventh: C-Bb ma-seventh: C-B a-seventh: C-B#
d-sixth: C-Abb mi-sixth: C-Ab ma-sixth: C-A a-sixth: C-A#
d-fifth: C-Gb p-fifth: C-G a-fifth: C-G#
d-fourth: C-Fb p-fourth: C-F a-fourth: C-F#
d-third: C-Ebb mi-third: C-Eb ma-third: C-E a-third: C-E#
d-second: C-Dbb mi-second: C-Db ma-second: C-D a-second: C-D#
d-unison: C-Cb p-unison: C-C a-unison: C-C#
This way, and given the enharmonic equivalent, every tone of the chromatic scale may exist
in two different interval relationships to the keynote (in this case C):
C perfect octave 12 augmented seventh octave

B major seventh 11 diminished octave major-7th
Bb minor seventh 10 augmented sixth seventh (mixo)
A major sixth 9 diminished seventh sixth (dorian)
G# minor sixth 8 augmented fifth #5
G perfect fifth 7 diminished sixth fifth
F# augmented fourth 6 diminished fifth tritone (lydian)
F perfect fourth 5 augmented third fourth
E major third 4 diminished fourth major 3rd
D# minor third 3 augmented second minor 3rd
D major second 2 diminished third whole step
C# minor second 1 augmented unison half step (phrygian)
C perfect unison 0 diminished second root
The first column in this table holds the designations of the note, the second column the
preferred interval designations. The third column represents the half-note intervals relative to
the keynote, and the fourth column represents the alternate designations. In the fifth column,
some abbreviations customarily used by musicians are found (there may be others, of course).
Again: this is based on equal-temperament tuning including enharmonic equivalents.
Classical harmony theory finds reasons for a further differentiation; however, this is beyond
the aim of the present elaborations [see secondary literature].

The below table indicates the numerical differences between just tuning and equal-
temperament tuning. The deviation is just tuning vs. equal temperament tuning.
Interval name no. of HS notes frequency relation cents deviation
Perfect octave 12 C-C' 1\2 1200,00 0,00

Major seventh 11 C-B 8\15 1088,27 -11,73
Minor seventh 10 C-Bb 9\16 996,09 -3,91
Major sixth 9 C-A 3\5 884,36 -15,64
Minor sixth 8 C-G# 5\8 813,69 +13,69
Perfect fifth 7 C-G 2\3 701,96 +1,96
Tritone 6 C-# 32\45 590,22 -9,78
Perfect fourth 5 C-F 3\4 498,05 -1,95
Major third 4 C-E 4\5 386,31 -13,69
Minor third 3 C-D# 5\6 315,64 +15,64
Major second 2 C-D 8\9 203,91 +3,91
Minor second 1 C-C# 15\16 111,73 +11,73
Perfect unison 0 C-C 1\1 0,00 0,00
cent
Table: Frequency relations of octave-internal intervals for just tuning. The deviations refer to the corresponding
interval in equal-temperament tuning. Specifications in 1/100th cents should be in practice rounded off to whole
cent-values. Compared to the major third in equal-temperament tuning, the major third in just tuning is to low by
14 cents. The other way round: compared to the just-intonated major third, the major third in equal-temperament
tuning is too high by 14 cents. A deviation of 1 cent corresponds to a frequency difference of 0,058%.
We can see the frequency relations for different tunings in the following Fig. 8.9. The
abscissa is a logarithmically divided frequency axis.
B
.
Fig. 8.9: Pythagorean (), just-intonated (), und perfect (o) intervals.
Since the half-step intervals are all equal in equal-temperament tuning, changing key (i.e.
moving to a scale with a different keynote) does not represent a problem. For example,
referencing to E results in the following intervals: E-F# = 2 HS = major second, E-A = 5 HS =
perfect fourth. The reference to a particular key may now be omitted, because every interval is
unambiguously defined by the number of its half-steps (HS).
Further interval designations exist beyond the octave space, as well: minor ninth (13 HS),
major ninth (14 HS), minor tenth (15 HS), major tenth (16 HS), perfect eleventh (17 HS),
augmented eleventh = diminished twelfth (18 HS), perfect twelfth (19 HS), minor thirteenth
(20 HS), major thirteenth (21 HS); the half-step distances are given in brackets.

8.1.5 Typical detuning in guitars
Every guitarist will have experienced days when his/her guitar would just not tune properly. It
typically gets really bad when we try to re-tune individual notes within chords. Even with a
perfectly fretted neck and premium strings, this problem may occur – the most likely reason
for which is the difference between just intonation and equal-temperament intonation. While
the fifth tuned according to the latter is, with a deviation of 2 cents, really close to the perfect
fifth tuned with just intonation, we find a much larger deviation for the third: that would be
+13,7 cents for the major third, and as much as -15,6 cents for the minor third! Such detuning
is already well audible, and the guitarist simply has to live with it. Trying to chord-
specifically retune individual strings (towards just intonation) may easily generate deviations
of as much as 29 cents for notes in other chords – and with that it now gets really shoddy. As
an example:
An A-major chord (played without barré) consists of the notes [e-a-e-a-c#-e]. Given that all
notes are tuned to equal temperament, it is in particular the C# played on the B-string that
creates problems: it is sharp by 14 cents compared to a justly intonated C#. If we now retune
the B-string by -14 cents (7,9 ‰), this A-chord will sound perfect. However: if, with the same
retuning, e.g. an E-chord [e-h-e-g#-b-e] is played, the resulting sound is atrociously off. What
happens is that the down-tuned B-string sounds a flat fifth – while the major third in that E-
chord (the G# played on the neighboring G-string) is sharp. The interval between these two
strings (3 half-steps G#-B) is too small by 29 cents! Changing from that re-adjusted A-major
chord to a D-major chord creates a similar disaster: the down-tuned B-string now sounds too
flat a D. The major third (F#) played on the neighboring E-string is already anyway too sharp
by 13,7 cents and now sounds doubly out-of-tune relative to the tonic (that is lowered by 13,7
cents).
There may always be special cases when – given a limited selection of chords – a special
detuning creates advantages. For example, it does not sound bad at all to slightly lower the
tuning of the G-string for E and A7. E-major has [e-b-e-g#-b-e], and A7 has [e-a-e-g-c#-e]. In
the E-major chord, the third profits, and in the A7 chord the diminished seventh – both are
sharp in equal temperament relative to just tuning so that this detuning makes sense. For the
same reason, the same detuning works well with the B7-chord [f#-b-d#-a-b-f#]. But don’t you
dare now changing to C or G … Thus, for universal deployment it is the equal-temperament
tuning (executed as perfectly as possible) that remains a workable solution.
8.1.6 Stretched tuning
Piano tuners are known to tune not exactly according to equal temperament but in a slightly
stretched-out fashion. In particular, in the very high and very low ranges, deviations of up to
30 cent can result. A spreading-out of partials, and in addition a narrowing of the pitch
perception, are given as justification. In the guitar-relevant pitch range, however, the effect
(merely 2 cents per octave) is rather weak, and the (compared to guitar strings) much heavier
piano strings are no adequate equivalent. “Buzz” Feiten has obtained a US-patent for the
stretched tuning – see Chapter 7.2.3). Fender, on the other hand, recommends adjusting the
octave at the 12th fret with no more than 1 cent error – no spreading. To each his/her own …

8.2 Frequency and pitch
Frequency is a quantity from the realm of physics, while pitch – as a sensory perception-
quantity – belongs with the auditory event. Usually, frequency is measured in Hz,
representing the numbers of oscillations per second. The unit Hertz (abbreviated Hz) is named
after the physicist Heinrich Hertz. The inverse of the frequency is the period (short for
duration of periodicity of a cycle). A period of T = 2 ms corresponds to a frequency f = 500
Hz. The pitch may either be determined via self-experiment (introspection), or indirectly via
evaluation of the reaction of a test-person (a “subject”). Although the pitch is a subjectively
rated quantity, it can be measured numerically. Measuring means in this context to allocate
numbers to an object-set according to predetermined rules, with these numbers being
reproducible within purposeful error margins. Now, what one considers purposeful – that
again is a rather subjective decision♣. Most psychometric experiments yield intra- and inter-
individual scatter: one and the same test person may give different evaluations when carrying
out the same experiment a number of times (intra-individual scatter), and the assessments of
different test persons may vary when an experiment is presented once for each person (inter-
individual scatter).
8.2.1 Frequency measurement
Simple measurement devices for frequency count the number of oscillations occurring per
time-interval: 5 oscillations per 0,1 s yields 50 Hz, for example. ‘Oscillation’ always implies a
whole period. For a string, this means: swinging from the rest-position in one direction,
reversal at the crest (apex), swinging (across the rest-position) fully to the other apex, reversal
at the latter, and swinging back to the rest-position. Given an ideal oscillation, terms such as
frequency or period are thus easily definable – real oscillations are, however, not ideal. Signal
theory defines a periodic process as infinitely repeated in identical form. Thus, a sine tone is
periodic and has one single frequency. A sound composed of a 100-Hz-tone and a 200-Hz-
tone (in music this would be called a note) would be periodic, as well (Fig. 8.10). However,
since more than one frequency appears here (i.e. 100 Hz and 200 Hz), we need to distinguish
between frequency of the partial and the fundamental frequency. Now, the fundamental
frequency is not necessarily that of the lowest partial, but the reciprocal of the period. The
oscillation-pattern of a sound comprised from sine components at 200 Hz, 300 Hz, and 400
Hz repeats after 10 ms; the fundamental frequency therefore is 100 Hz although there is no
actual partial found at 100 Hz within that sound. Generally speaking, the fundamental
frequency is the largest common denominator of the frequencies of the partials, and the period
is the least common multiple of all periods of the partials.
Fig. 8.10: Sine tone (100Hz), two-tone sound (100|200Hz), three-tone sound (200Hz|300Hz|400Hz);
0–50 ms each.
♣
A driver of a vehicle that has just reflected a high-frequency radar-beam may possibly demand a larger margin
of error than what the municipal administration profiting from motoring fines would see as appropriate.

Evidently, a tone does not need to be of mono-frequent characteristic to feature one frequency
(more exactly: one single fundamental frequency). In theory, there even may be an infinite
number of partials (as is the case for an ideal square-wave sound). However, the partials have
to be harmonic i.e. their frequencies need to be integer multiples of the fundamental
frequency. This condition cannot be met e.g. for irrational numbers such as und . In
practice, though, frequencies can be specified only to a finite number of decimals, e.g. 1,414
Hz, or 1,732 Hz. If these examples would be rounded-off roots, a specification of “the
fundamental frequency is 0,001 Hz” would be very arbitrary. Nor would it be within the
meaning of the largest common denominator; 0,002 Hz, at least, would be a common
denominator. It should be noted that the issue with the irrational numbers is of a less
academic nature than one would think. This is because string vibrations are never of an
exactly harmonic nature. The decay process gives every “period” different amplitudes, and
the partials are not actually in a strictly harmonic relationship (i.e. they are un-harmonic),
due to bending stiffness, and to the dispersive wave propagation connected to it. Let us
assume that the decay process is so slow that its effects on the spectral purity may be
disregarded. Let us further assume that the analysis of a guitar tone has yielded four
components (partials) at the frequencies of 100 Hz, 201 Hz, 302 Hz, and 404 Hz. What would
be the frequency of this tone? It makes no sense to specify 1 Hz as the fundamental
frequency, and to call the partials the 100th, 201st, 302nd, and 404th harmonic, respectively.
What remains is the sobering insight that a guitar tone has no fundamental frequency. It
does, however, have a pitch! Determining that pitch shall be reserved for a later paragraph –
first we still have to clarify what a tuning device is in fact doing given the above finding, and
why a string may be tuned – despite all this.
Fig. 8.11: 4-partials sound, f1 = 100Hz, f2 = 201Hz, f3 = 302Hz, f4 = 404Hz. 1/f –envelope; t = 0 – 0,5 s.
Fig. 8.11 depicts the first 0,5 seconds of a sound comprised of the frequencies mentioned
above. How many periods appear during that time interval? Trying to count the maxima, we
get into a bit of trouble approaching the mid-section of the figure, but we can make it to the
right-hand end with the finding that there will be about 49 and 3/4th periods. But what is that
in this case: a “period”? To deliver a visual evaluation, our optical sense seeks to – as well as
possible – perform visual smoothing (i.e. filtering!) and locally limited auto-correlations.
What else could a visual system do in the first place. But will that be helpful in the context of
an acoustical signal? What could an exact algorithm be? Simple measurement devices
determine the upward (or downward) zero-crossing. Given the above signal, that will make
for considerable problems between 0,15 and 0,2 s, and between 0,35 and 0,45 s. Of course,
smoothing (i.e. low-pass filtering) is a solution, but with it the frequency of the filtered signal
will be determined. In the extreme case, the filtering will pass on merely the 100-Hz-
oscillation – with that, the frequency-measurement certainly is most straightforward.
Presumably most tuning devices (electronic tuners) have a built-in low-pass filter that filters
string-specifically, or at least instrument-specifically. Also, they will accept small deviations
from the nominal value. It may still happen that the display sways back and forth between
correct and incorrect. The well-versed guitar player will then turn down the tone control (low-
pass filter) or relinquish any high demands on accuracy. Some may celebrate an act of the
gripping drama: “Guitarists never stop tuning, guitars eternally refuse to be correctly tuned”.

The frequencies on which Fig. 8.11 is based show the fundamental problem but do exaggerate
the situation. The spreading of the partials found with electric guitars amounts to about 0,2%
for the E2-string at 500 Hz, and to about 1% at 1 kHz. Still: if the 12th partial of the low E-
string is represented with substantial level in the overall signal, a possibly annoying
discrepancy of about 9 Hz between ideal ( ) and real (997,7 Hz) may
result. Such an error would be inacceptable for precise tuning. However, the amplitudes of the
higher partials usually decay much faster than those of the lower partials, and thus most
electronic tuners achieve an acceptable reading, especially since the guitarist will pluck the
string rather softly so as not to emphasize the harmonics too much. For the lower partials of
the low E-string, the inharmonicity will then be rather unproblematic with 0,02% for the third
harmonic, and 0,05% for the fourth. There will be even less of an issue for the higher guitar
strings: due to the smaller string diameter, the bending stiffness plays not as big a role, and
the number of the possibly interfering harmonics decreases due to the low-pass character of
the pickup.
As a summary, we may therefore note: even though the string vibration is comprised of
inharmonic partials and therefore in theory has no fundamental frequency, electronic tuners
will in practice detect the frequency of a “practical” fundamental, or a value that is very close
to it. Whether our hearing system arrives at the same conclusion is, however, an entirely
different question (see Chapter 8.2.3).
8.2.2 Accuracy of frequency and pitch
Following a chapter on frequency measurements, it would seem natural to explain pitch

determination in more detail. First, however, desired accuracy and measurement errors shall
be looked into. This way it will be easier to assess the properties of the hearing system that
will be the focus in the subsequent chapter.
The frequency of a strictly periodic tone can be measured with an accuracy that is more than
adequate for musicians. Precision frequency counters feature relative measurement errors in
the range of 10-5, and 10-6 is not impossible, either. In a watch, for example, an error of 10-5
leads to an inaccuracy of 1 second / day. The problem does not lie in the underlying reference
(oven-stabilized quartz generators are extremely accurate) but in the signal to be measured.
Measuring does become tricky if this signal does not have exactly identical periods. Given a
known shape of the signal, frequency measurement is simple and quick: three points on a sine
curve (excluding a few special points such as the zero crossing) suffice to determine the three
degrees of freedom: amplitude, frequency, and phase. In theory, the three points may succeed
one another very quickly, and thus achieving both high measurement precision and a short
measuring time is not a contradiction. These highly theoretical findings based on function
analysis do not help for measuring the frequency, though. This is because the shape of the
signal is not known, and with that the rule holds that the duration of the measurement and
the accuracy of the measurement are reciprocal to each other. If the frequency measurement
is based on counting periods of the signal, a measurement of a length of 10 s is required in
order to achieve an accuracy of 0,1 Hz. Interactive tuning would be impossible given such
long durations. Frequency-doubling or half-period-measurements could be advantageous, but
requires that the duty-factor of the signal is known – which is not the case with sounds of
musical instruments. What remains is to determine the frequency of individual partials.
Presumably, most tuning devices will indentify the frequency of the fundamental, and – in the
case of the guitar – will indicate that as the frequency of the string.

It is not only the measurement process that requires us to consider the measurement duration,
but also the fact that the signal to be measured is time-variant. The amplitudes of the partials
decay with different speed as a function of time, and moreover the frequencies of the
partials will change. This is connected to the string being elongated and thus stretched more
as it moves from its rest-position: the larger the vibration-amplitude, the higher the frequency.
Further, it needs to be considered that real string oscillations are never limited to remain in
exactly one single plane. During the decay process, the plane of oscillation rotates; this can
be seen as the superposition of two orthogonal vibrations. Due to direction-dependent
bearing-impedances, these two vibrations may differ slightly in their frequencies, and
consequently there will be changes in amplitude and frequency over time.
A (non-representative) field experiment shall give some indications of how accurate the
frequencies of strings can be measured despite all these issues. From the many digital
electronic tuners on the market, three were selected and checked using a sine generator and a
precision frequency counter. The ranges within which the devices registered a “correct
tuning” measurement were ±1,6‰, ±2,0‰, and ±2,3‰, i.e. on average ±2‰. This
corresponds to ±3,5 cent. To be clear: “correct tuning” in this context means that, for
example, the device under test evaluated all frequencies between 439,4 Hz and 440,7 Hz as
correctly tuned to A. The width of that tolerance interval is a compromise between high
precision (possibly never achievable due to the aforementioned issues) on the one hand, and
more easily achievable “kind-of-in-tune” state (that may not be accepted due to audible
deviation from the ideal value) on the other hand.
Fig. 8.12 shows the progress over time of such a measurement. Using a tuning device (Korg
GT-2), the tuning of two guitars was assessed; depicted are the deviations of the value
indicated by the tuner from the reference value (during 8 seconds of a measuring time; for
each string). The string was plucked with regular strength at t = 0; all non-involved strings
were damped in order to avoid interferences. For the measurement with the Gretsch
Tennessian, the stronger decrease of the pitch during the first seconds stands out. This effect
was not further investigated; a cause may be found in the relatively thin strings: their average
tensile stress is increased with strong vibration. Towards the end of the shown measuring
time, the deviations increase; this is due to the decreasing signal level. The Ovation (with the
signal of the piezo pickup measured) also caused some fluctuations during the measuring
time; the causes for these were looked into in more detail.
Fig. 8.12: Pitch measurement with the electronic tuner Korg GT-2. Tennessean (left), Ovation SMT (right).

In Fig. 8.13, the measured pitch is compared to the level of the fundamental over time. The
signal generator is in both cases the plucked B-string of the Ovation SMT. At 3,5 s we see a
minimum of the level of the fundamental. Assuming a time lag of around 0,5 s due to the
processing, pitch-fluctuations at about 4 s can be explained; the other fluctuations cannot be
attributed to anything specific with any certainty.
Fig. 8.13: Measured pitch-deviation, level of fundamental; Ovation SMT, B-string plucked at t = 0.
The measurements show that – despite alleged digital precision – considerable fluctuations in
the display value are to be expected. Since the electronic tuning shows a highly accurate
display without any noteworthy fluctuation when a precision generator serves as input, only
the guitar tone itself can be the reason. The more “lively” this tone is, the larger the
fluctuations in the measurement result will be, and the larger the variations in pitch.
At this point, a short digression into thermodynamics makes sense. The linear thermal
expansion coefficient describes how dimensions change dependent on temperature. If the
dimensions are “imprinted” (forced), the mechanical stress will vary as the temperature
changes. This implies for steel strings: the un-tensioned string will experience an elongation
by a factor of 16x10-6 for a temperature increase of +1°C. While this appears insignificant
compared to the 2‰ mentioned above, we need to consider that for the change of the string
frequency, the relative change in stress is the influential factor. Typically, an E2-string needs
to experience an elongation (strain) of about 1,5 mm for correct intonation. It is this 1,5-mm-
strain that needs to be seen in connection with the change in length caused by the temperature
change. The relative frequency change corresponds to half the relative change in strain
(square-root in action here!). For our example, this means: with 1°C temperature change,
the string frequency changes by 5,3‰. Here we assumed that the dimensions of the neck
and body of the guitar remain constant; given the highly different thermal time constants over
short time-periods, this is justified. Confirmation was provided by an experiment: taking a
correctly tuned guitar (Gibson ES-335) from a room to the outside (cooler than in the room by
a few degrees) raised the frequency of the E2-string within a few seconds by 12‰.
Conversely, it follows: if you seek to keep the tuning of a guitar constant within 1‰, you
need to demand that short-term temperature fluctuations remain within 0,2°C .
We have saved the most important question for last: how precise actually is the hearing
system? In the terminology of psychoacoustics: how large is the threshold of pitch
discrimination? You will find quite different answers – it depends on the experimental
methodology. Fundamentally, we need to distinguish between a successive pair (2 tones
follow each other in time) and the dyad (two-tone complex; two tones are sounded at the
same time)

When concurrently presenting two tones, the smallest of differences between frequencies may
be noticeable – depending of the circumstances. For example, if two 1-kHz-sine-tones are
detuned by 0,1 Hz with respect to each other, a beating results: i.e. a tone is gradually getting
louder and becoming softer again, with its amplitude reaching its maximum value every 10
seconds. The latter duration is short compared to the average life expectancy, and also small
relative to the tolerance-span of the test persons (subjects) – therefore it is well observable.
For the same reasons, a periodicity of 0,01 Hz would still be observable – but with 0,001 Hz
the limited patience of the subject might become a problem. Relative to 1000 Hz, 0,001 Hz
already represents is factor of 10-6. However, to conclude that the frequency resolution of the
auditory system would always be 0,001‰ – that would be nonsense. The result is only usable
in the given experimental context.
Clearly, a large part of music consists of sounds comprising two or more tones – so: what
gives? The answer will necessarily remain unsatisfactory because music is diverse, but there
are rough guidelines. A first borderline is defined by the duration of sounds. If a sound
consisting of two tones lasts only for a second, a frequency deviation between the two tones
of 0,1 Hz will not be detected. Sounds of longer duration generally facilitate recognizing
frequency differences. Still: long sustained notes are often played with vibrato (for the
terminology see Chapter 10.8.2), and in this case a small detuning will be noticed less. Pitch
vibrato, however, cannot be generated on every type of instrument – but then a polychoral
design will make for audible modulations already in single notes. On the piano, for example,
most notes are generated by two or three very closely tuned strings; beats will be inherent in
the system here. Even when trying to tune all strings of one piano note to exactly equal pitch,
the overcritical coupling of the strings will result in beating. Besides the beats audible in the
single note, additional beating between different notes may be audible as a separate
characteristic – but this will depend on too many factors to make an analysis with simple
algorithms feasible. Looking at the distribution of how often musical notes of certain
durations occur, and considering the auditory fluctuation assessment, we may cautiously
estimate the following: upwards of an envelope-period of about 1 s, beats loose their sensory
significance. This corresponds to a frequency resolution of about 1 Hz.
Given a sequential presentation of tones, beating is excluded. Or so many

psychoacousticians believe. However, of significance is not which sounds are generated, but
which sounds actually arrive at the ears of the subjects. Presentations of sounds in a room are
always accompanied by reflections – if these occur in great numbers, they are called reverb.
If the pause between sequentially presented tones is too short, there may still occur a short
beating at the transition, and this beating may be perceived depending on the circumstances.
Such experiments should therefore exclusively be carried out using headphones. A room as a
transmission system has other issues, as well: due to the superposition to interleaved
reflections, the impulse response is lengthened. The Fourier transform (the transmission
function) obtains selective minima and maxima, and between these includes steep flanks. A
frequency change of 1 Hz that is inaudible as such may now receive a change in level of
several dB. This will be audible – however, although the original cause is a frequency change,
it is the threshold of the hearing system for amplitude discrimination that is decisive for the
detection.
For sine tones of a duration of no less that 0,2 s (sequentially presented via headphones), the
threshold for frequency discrimination is about 1 Hz in the frequency range below 500 Hz.
Above 500 Hz, this threshold is not constant anymore, but about ca. 2‰ of the given
frequency. With shorter duration (< 0,2 s), the discrimination threshold deteriorates. These
data are averages from a large number of psychoacoustical experiments.

For a sine tone, it is easy to assess whether it ties in with the 1-Hz-criterion, or with the 2‰-
critrerion: the limit is at 500 Hz, with a transition from one limit value to the other♣. For
sounds comprising several partials, this decision is not so simple anymore. Given an E2-string,
the first 6 partials are below 500 Hz, all further partials are above that limit. In such cases the
following holds: frequency changes become audible if for at least one (audible) partial the
threshold or frequency discrimination is surpassed. For the E2-string it thus is not the 1 Hz /
82,4 Hz 12‰ criterion that forms the basis for the decision but the 2‰-harmonics-criterion.
This is a good match to the tolerance range we found in electronic tuners. With the conversion
into the unit cent that is customary among musicians, the tolerance range is 3 – 5 cent (with 1
cent = 1/100th semi-tone interval 0,58‰). The 1-cent-accurracy that is sometimes
demanded is exaggerated: on the guitar, the temperature of the strings would have to be kept
constant within 0,1°C (which may be difficult when playing your hot grooves, as cool as they
may feel). If the guitar can be tuned with an accuracy of ±2‰, we are on the safe side. This
does not mean, though, that every larger deviation will immediately sound out-of-tune. Our
hearing system can be quite forgiving and ready to generously compromise in certain
individual situations.
8.2.3 Pitch perception
It has already been noted above that pitch and frequency are different quantities. Our auditory
system determines the pitch according to complex algorithms – an associated comprehensive
discussion would go beyond the scope of this book (specialist literature exists for this). A first
important processing step is the frequency/place transformation in the inner ear (cochlea): a
travelling wave runs within the helical cochlea, with the wave-maximum depending on
amplitude and frequency of the sound wave. Tiny sensory cells react to the movement of this
travelling wave; they transmit nerve impulses among various nerve fibers to the brain. The
latter performs further advanced processing. A regularly plucked guitar sound consists of a
multitude of almost harmonic partials. Round about the first 6 – 8 of these partials result in
distinguishable local travelling-wave maxima, the higher partials are processed grouped
together.
Normally, we cannot hear the individual partials when a string is plucked. Rather, we hear a
complex tone with one single pitch. With a little effort, however, these individual partials may
be heard, after all. To do this, we first suppress a given partial using a notch-filter, and then
switch off the filter-effect so that the original signal is reproduced. From the moment the filter
is switched off, the partial in question will be audible for a few seconds, and then merge again
with its colleagues to form the integral sound experience that was originally audible. A
sufficient level of the partial is a requirement; the partial may not be masked to such an extent
by its spectral neighbors that it does not contribute at all to the aural impression. How the
single elements are grouped and combined together – that has long been a topic of research
for the Gestalt-psychologists. This topic resulted first of all in the Gestalt laws for the visual
system (see Chapter 8.2.4). In particular, it is the “principle of common fate” that also plays a
role in the auditory system if the issue is to group the individual partials of a complex sound
event, attributed them to sound sources, and to assign to the latter characteristics such as e.g. a
pitch.
♣
Both “1 Hz” and “2‰” are to be taken as approximate values that are subject to individual variations.

As a rule, the pitch recognition works rather well for complex tones with exactly harmonic
partials – especially if there are lots of partials. However, just like in the visual system with its
optical illusions, we know in the auditory area of special sounds that lead to seemingly
paradox perceptions. If the partials are not harmonic – as it is the case e.g. for bells – the pitch
algorithm develops estimates based on probabilities. Results can be that a subject (test person)
cannot decide between two pitches, or that two subjects allocate entirely different pitches to
the one and the same sound. Sounds of strings are, however, only mildly in-harmonic, and
merely octave confusions are conceivable in the worst case. As a rule, for the pitch of a
string tone a value is determined that is close to the fundamental but not identical to it. In a
first step, the auditory system allocates to all non-masked partials their spectral pitch, and on
that basis calculates a spectral rating curve that has a flat maximum at around 700 Hz ♣ –
this is the virtual pitch. Higher-frequency and lower-frequency partials therefore contribute
less to the pitch than middle-frequency components. Experiments carried out by Plomp♥ show
that it is – in particular – not the fundamental that defines the perceived pitch. In a piano tone,
the frequency of the fundamental was decreased by 3%, while all other partials were increased
by 3 %; the result being that the perceived pitch went up by 3%. While the fundamental can
have a big influence on the tone color, it is rather insignificant for the pitch as long as there
are sufficient higher harmonics available.
Now, in the guitar, the harmonics are progressively shifted towards higher frequencies (at 1
kHz easily by + 15 cents). If we calculate back the pitch from this, we arrive at a value that is
higher than the reading on an electronic tuner (measuring merely the fundamental). We
should still not retune to make the tuner display 15 cent more – things are more complex. The
perceived pitch of the fundamental (or its frequency) is not simply the n-th fraction of the
frequency of the n-th partial: Fastl/Zwicker [12] report of hearing experiments with
harmonically complex tones with a perceived pitch lower that the objective fundamental
frequency. The error of the mentioned electronic tuner would thus tend in the same direction
as processing in the hearing system. Moreover, it needs to be considered that the pitch
(despite constant frequencies of the partials) is dependent on the sound level: as the level
increases, the pitch decreases by as much as 5 cents per 10 dB. Even larger effects can be
created by additional sounds that are superimposed on the guitar sound: literature [12]
knows of pitch shifts that can be as large as a semi-tone in extreme cases! Such shifts may not
be part of everyday guitar playing, but all in all there is a wide field leaving much space for
fundamental research. What also transpires: cent-exact tuning is not actually possible. Even
though frequencies of individual partials may be measured and adjusted with high precision –
it’s the hearing system that decides whether the tuning is “correct” … and it will use
complicated, situation dependent and even individual criteria. That laboratory experiments
indicate that pitch differences of 3 – 5 cent are recognized does not imply that this accuracy
needs to be always observed. It is impossible to specify a mandatory limit for tones hat would
be audibly out-of-tune, because too many parameters determine the individual case – but in
practice the following rule-of-thumb has proven itself: a tuning error of no more than 5
cent is desirable, with 10 cents often being acceptable. Those listeners who have privilege to
experience sound through “golden ears” may happily halve these numbers.
♣
Terhardt E.: Pitch, Consonance, and Harmony. JASA 55 (1974), 1061–1069.
♥
Plomp R.: Pitch of complex tones. JASA 41 (1967), 1526–1533.

8.2.4 Grouping of partials
Customarily, string vibrations are described as a sum of differently decaying partials. This
“expansion according to harmonic members of the series” is not imperative, but it is the
standard tool of spectral analysis – and in fact it derives at least some of its justification from
the hydromechanics in the cochlea♣. Even though it is, after all, a model: the tone of a guitar
does “consist of” partials. Upon plucking of a string we do, however, not hear a multitude of
tones but only one tone – so there are grouping mechanisms in auditory perception that form
groups of connected partials from the spectral pitches (of the non-masked partials), the latter
having been gathered on a low processing level. The brain (the human CPU) receives
information from the sensory receptors and evaluates it, i.e. reduces this immense flood of
data by categorization- and decision-processes. Just as an example: 1,4 million bits of
information are contained in just one second of music from a CD! Whether it’s 50 bits (per
second) that reach our awareness or a few bits more or less: the major portion of the arriving
information needs to be jettisoned. But which portion would that be?
Fig. 8.14:
Examples for visual
grouping.
On the basis of experiments relating to visual perception, Gestalt-psychologists such as e.g.

Max Wertheimer have formulated the Gestalt laws that are applicable also in auditory
perception. Presumably, the recognition mechanism includes a reduction of the multitude of
data delivered by the receptors according to grouping-processes and -patterns already stored
in memory. The already-known-and-plausible is given a higher priority compared to the
unknown and illogical. The arrangement of two logs of wood shown in the middle section of
Fig. 8.14 can be interpreted three-dimensionally at first glance, even though the drawing
plane has merely two dimensions. Some small changes (graph on the right) make the spatial
impression all but (or completely) go away. It would go too far to elucidate in detail the
principles of closeness, similarity, smooth flow, coherence, and of common fate – the
reference to literature in perception psychology must suffice here. As an example that circles
back to acoustics, Fig 8.14 shows on the left the word “pitch” represented via an incomplete
outline-font. Despite considerable deficits in the picture as such, our visual sensory system
succeeds without problem in completing the given lines in a sensible manner, and in giving
them a meaning. “Pitch” is captured as a word, and not as a bunch of lines. Perceiving the
latter is also possible, though – our visual system is more flexible in this respect compared to
our hearing. While it is visually possible to deliberately separate the lines or a grouped object,
this is very difficult or even impossible in auditory perception: compared to “pitch” in the
figure, it is not at all as simple to switch back and forth between the individual object (the
partial) and the grouping (guitar tone). Plucking the string, we hear one (musical) tone but
find it difficult to pick out individual partials. It may not be entirely impossible but we have
serious difficulty doing it compared to separating a read word into its letters and their lines
and curves. Insofar there exists a difference between the visual and the auditory processing,
but there are also shared characteristics, such as the ability to group, or the hierarchical
structure. According to the pitch model by Terhardt, spectral pitches are determined first (in
the cochlea) and from these the virtual pitches (on a higher processing level).
♣
Frequency-place-transformation [12] chapter 3.

The processing step on the lowest (peripheral) level of this hierarchy is similar to a short-term
Fourier analysis (although with very special parameters). Already on this processing level,
partials are sorted out – those, the energy of which is so small that “you wouldn’t recognize if
they were missing”. This is because not every partial contributes to the aural impression: if its
level is too small compared to its spectral neighbors, it is suppressed (this effect being termed
masking in psychoacoustics). The partials that are not or only partially masked are given a
corresponding spectral pitch each. This pitch will be subject to weighing in the higher
processing levels, and synthesized into a virtual pitch. It is no issue in this process if the
fundamental of a harmonically complex tone is entirely missing. For example, the telephone –
with its band-limitation to 300 – 3400 Hz – is not even able to transmit the first two partials of
a male voice (fG = 120 Hz), but the pitch of the fundamental can still be reconstructed when
listening. The perception of a speaking child never appears.
Fig. 8.15: Spectrograms of two tone-sequences: on the right, the descending sequence is frequency-modulated.
One grouping-rule (of several) says that concurrently starting sinusoidal tones with an integer
frequency relation are likely to stem from the same sound source and should be grouped
together into one object. Natural sound sources (and only those were available for training the
ear during its evolution) almost never generate pure tones. Even if that would occur, it would
be extremely improbable that at the same instant several of such sound sources would start to
emit sound, and even less likely for them to have an integer frequency relation. If such a
harmonically complex sound is heard, it can therefore only come from one source. Given this,
it is purposeful in the sense of information reduction to combine the corresponding spectral
lines, just as (optically) the two lines of the letters L, V or T (respectively) are seen as
belonging together. The visual signal processing can separate two superimposed
@ letters, and
the hearing system can follow one speaker – even in the presence of a second concurrent
speaker. That does not function perfectly, but still astonishingly well: Chuck’s “long distance
information” is clearly intelligible, despite the competing accompanying instruments, and
similarly fare “O sole mio” or “We’re singin’ in the rain”. More or less, that is – depending on
orchestra/band and singer. The latter may have to push himself quite a bit (or instruct/bribe
the sound man conducively) to make sure that the audience (if they listen that closely at all)
will not with surprise take cognizance of the fact that “there’s a wino down the road” ,
rather than that Mssrs. Plant, Page, Jones & Bonham, jr. “wind on down the road” (if they
ever play the tune in question again together). Indeed, the grouping of harmonics (and thus
their decoding) does not always succeed flawlessly. Fig. 8.15 gives an idea of difficulties that
may occur: on the left we see the spectrogram of a little two-part melody: it is not easy to say
which lines belong together. In the figure’s middle section with its larger frequency-span, a
formation rule starts to emerge – but only on the right we get some clarity: given different line
width and a frequency modulated top voice, the separation becomes easy. The hearing system
(especially the musically schooled one) will separate the two voices already without vibrato
into an ascending and a descending one; with vibrato it comes even more naturally. That
would be one reason why singers and soloists often use vibrato: they can be identified more
easily among the multitude of accompanying tones. Since the modulation in the soloist-
generated sound will run similarly for all partials, the hearing gets help for grouping.

The perceptional psychologist uses the term law of common fate in this context: everything
that starts concurrently and ends that way, too, “presumably” belongs together. In order to
further facilitate the recognition (or the grouping), the soloist chooses a modulation frequency
of about 4 – 7 Hz; this is because the hearing system is particularly sensitive for such
modulations (fluctuation strength [12]). Accompanying musicians (in the choir or orchestra)
also often use vibrato: in part because they just can’t help it anymore, but in particular
because that way messy beatings can be avoided that would otherwise automatically arise
from playing with several voices. From the "orchestra hacks", however, some restraint is
required with respect to vibrato – unless some serious bedevilment is actually called for.
How vibrato will influence the grouping of partials is shown also on the left of Fig. 8.16: first,
a 100-Hz-tone sounds that is comprised of its 1st,2nd, 3rd, 4th, 6th, 7th, 8th, and 9th harmonics.
From half the shown time interval, an additional tone comes into play in a fifth-relationship
(strictly speaking it’s the twelfth) because the 3rd, 6th and 9th harmonics are slightly modulated
– the latter now form in a new grouping the 1st, 2nd, and 3rd harmonic of the additional 300-
Hz-tone.
Fig. 8.16: Partial with/of a common fate are grouped to objects.
In the middle section of Fig. 8.16, some partials are started with a delay: first, a 100-Hz-tone
sounds, followed by a 300-Hz-tone. However, this happens only if the delay is long enough
(e.g. 100 ms). With a delay of about 30 – 50 ms, a sort of initial accent results, with the
delayed partials only audible for a short time, as a sort of “livening-up” of the 100-Hz-tone.
For an even shorter delay (e.g. 5 ms) this accent looses significance and we hear only one
single tone. Despite the objective delay, a subjective commonality results that is assigned one
single common cause.
In the right-hand section of Fig. 8.16 the level of the 3rd, 6th, and 9th harmonic is abruptly
changed – indicated by the darker lines. We hear a 100-Hz-tone, and an additional 300-Hz-
tone in the time interval between 0,2 – 0,4 s. However, if the levels of the 3rd, 6th, and 9th
harmonics are changed continuously, we hear only one single tone with a changing tone color.
Our experience teaches us that an abrupt change can only stem from a newly introduced
object, while slow changes may be attributed to single objects, as well.
The discovery and understanding of the auditory grouping algorithms (here only outlined via
a few examples) is not only of interest to musicians and psychoacousticians, but increasingly
also to neuro-scientists. Those who seek to immerse themselves into cortical hard- and
software find a profound supplement in Manfred Spitzer’s book "Musik im Kopf" [ISBN 3-
7945-2427-6] (translator’s note: this book is apparently only available in German, the translation of the title
would be: "Music in the Head".)

8.2.5 Inharmonicity of partials
Due to the dispersive transversal-wave-propagation, the partials of guitar tones are not strictly
harmonic♣, but spread-out spectrally: the frequency of the ith partial is not , but a bit
higher. The analytical connection between bending stiffness and spreading-out of partials has
been already discussed in detail in Chapter 1.3 – we will now look at the connected effects on
the perceived sound.
In the following analyses, a real guitar signal will be juxtaposed to several synthetic signals.
The real signal was picked up (without any sound filtering) from the piezo-pickup of an
Ovation Viper EA-68 guitar; it was stored in computer memory. For these recordings, the
open E2-string (D'Addario EJ-26, 0.052") was plucked with a plectrum right next to the bridge
in fretboard-normal fashion; the first second of decay was used for the psychoacoustic
experiments (listening tests). Exponentially decaying sinusoidal oscillations were
superimposed and saved as a WAV-file for the synthetic signal.
The DFT-analysis of the real signal yielded (with very good precision) the spreading-
parameter of b = 1/8000; given this, the frequencies fi of the partials are calculated as:
fi = frequency of the partial; fG = frequency of the fundamental.
Fig. 8.17 shows the percentage of frequency-spreading for the spread-out partials; fi is the
abscissa – and not . On the upper right, the levels of the partials are depicted; on the
lower right, we see the time-constants of their decay. With many partials we find in good
approximation exponential decay; some partials, however, show strong fluctuations in their
envelopes. For the first experiments, these beats were ignored – they were approximated
(replaced) via exponential decay.
Fig. 8.17: Percentage of spreading-out of partials (left); levels and decay-constants of partials (right).
The data for levels and decay of the partials of the real signals formed the basis for generating
the different synthetic signals.
; synthetic signal
♣
Harmonic spectrum: the frequencies of the partials are all in integer ratios relative to each other.

In the formula, A is for the amplitude, τ is for the decay time-constant, fi is for the spread-out
frequency, and ϕ is for the phase; all these parameters are functions of the order i of the
partials. The phases of the partials had not been measured – contrary to the level-spectra,
phase-spectra require considerable post-processing in order to obtain graphs that can be
reasonably well interpreted.
For a first listening experiment, a synthetic signal was generated that consisted of partials
with amplitudes and decay time-constants corresponding to those of the real signal. All
phases of the partials were set to zero, though, and the frequencies of the partials were integer
multiples of the fundamental frequency (i.e. they were not spread-out). A signal synthesized
that way sounds different compared to the real signal. In view of the frequency shifts shown
in Fig. 8.17, one might spontaneously consider a difference in pitch – this was in fact indeed
noticed during the first listening test. However, the “exact” fundamental frequency of the real
signal can – at a signal-duration of 1 s – not be determined with sufficient accuracy; it
moreover also changes during the decay (mechanics of the string). Therefore, the synthetic
signal was tuned by ear to fG = 81,9 Hz; the pitch was sufficiently well matched that way.
Subsequently, the essential difference in sound could be determined via the listening
experiment: the synthetic sound was described as “clearer, more buzzing, spatially smaller”,
while the real sound received the attributes of “more rusteling, more metallic, spatially
larger”. When presenting the sounds using loudspeakers (broadband speakers, normally
reflecting room), an interesting effect with respect to distance could be observed: as the
distance to the loudspeaker increased, real and synthetic signals became more and more
similar.
The hearing system has no receptor that would analyze the sound pressure arriving in the ear
canal with respect to time. Rather, the sound signal is first broken down into spectral bands
(called critical bands in this specific context) with a hydro-mechanical filter [12], and is only
subsequently recoded into the electrical nerve impulses (action potentials). It is nevertheless
purposeful to take a look at the time-functions of the sound signals – at least as long as we do
not loose sight of the band-pass-filtering included in the hearing system. Fig. 8.18 depicts the
time-functions of the real signal and of the synthetic signal – they differ considerably.
Fig. 8.18: Time-functions of the real signal and of the synthetic signals; E2-string.

The synthetic signal shown in Fig. 18.8 is periodic while the real signal is not. However, the
main difference between the two signals is not found in the periodicity but in the crest-factor
(ratio of peak value to RMS value). The considerable content of impulses in the synthetic
signal also shows up in a hearing-related spectral analysis (Fig. 8.19) as it is generated e.g. in
the CORTEX-Software "VIPER": here, we see time represented along the abscissa, and along
the ordinate the critical band rate (a non-linear mapping of the frequency as it occurs in the
auditory system [12]), scaled in the unit Bark. Coded via the color is a specific excitation
quantity derived from the signal filtering as it occurs in the inner ear (i.e. in the cochlea).
While the synthetic signal excites the hearing system across the whole signal bandwidth, this
synchronicity appears only in the low-frequency range for the real signal. Looking at the
pictures it becomes clear why the synthetic signal would be designated “buzzing”, while the
attribute “rusteling” is used for the real signal. We can also surmise why the distance between
loudspeaker and listener has such a big influence on the sound: given a larger distance, the
gaps between the impulses in the synthetic signal are filled with echoes, and it comes closer to
the real signal. Evidently, it is not the inharmonicity per se that is so special about the real
signal, but the lack of a strictly time-periodic structure featuring a high content of impulses.
Fig. 8.19: Auditory spectrogram (CORTEX-VIPER), real signal (left), Synth-1 (right).
There is a simple way to check the hypothesis related to impulse-content (or hamonicity): not
setting all phases of the partials to zero but having them statistically uniformly distributed
yields a so-called pseudo-noise-signal. Due to the strictly harmonic structure of the partials,
this signal is periodic, but the wave-shape within one period (in this case amounting to about
12 ms) is of random nature. Fig. 8.20 shows the auditory spectrogram, and Fig. 8.21 depicts
the time-function. Although this signal (like the Synth-1-signal) does not include the
frequency spreading of the real signal, it sounds almost exactly like it. Some test persons with
a trained hearing will still detect small differences; in particular in the attack, the signal
Synth-2 does not sound as precise.
Fig. 8.20: Auditory spectrogram (CORTEX-VIPER), Synth-2 (left), Synth-3 (right).

Still, the difference in sound between the real signal and Synth-1 is much larger that the
difference between the real signal and Synth-2. The rusteling heard in the real signal is
present in Synth-2, as well, but the latter lacks the buzzing that is characteristic of Synth-1.
Highly discriminating subjects may even hear “a tad too much rusteling” in Synth-2, but most
test persons will perceive no difference at all compared to the real signal. An alternative to the
equal-distribution phase would be a phase frequency-response suggested by M. R. Schröder♣
that will again guarantee a small crest-factor. The signal designated Synth-3 comprises a
harmonic spectrum (i.e. non-spread-out), with the phases of the partials defined according to
the following formula:
; Schröder-phase
Hearing them for the first time, real signal, Synth-2, and Synth-3 differ little; Synth-1,
however, sounds distinctly different. Given headphone presentation, a trained ear will notice
differences between all four signals, but with presentation via loudspeaker at close distance
only Synth-1 sounds different, and for bigger loudspeaker distances, all four signals sound
practically the same.
Fig. 8.21: Time functions of the real signal and of the three synthetic signals.
Since all three synthetic signals have identical amplitude spectra but still sound partly similar
and partly different, the frequency resolution of the hearing system cannot be of significance
in this respect. Exclusive basis for the differences in sound is the difference in the phases – it
is only in this parameter that the formulas used for the synthesis distinguish themselves from
each other. If one of the signals is transmitted via loudspeaker, the frequencies of the partials
do not change, but the phases of the partials do. This bold statement may not be entirely
correct from the point of view of signal theory (because a decaying partial is not described by
a single frequency but via a continuous spectrum that may well be changed via loudspeaker
and room), but it is quite usable as an approximation. The direct evaluation of frequency
responses of the phase is, however, of no help: the auditory system does not include a receptor
that would a priori determine the phase. Rather, small sensory (hair-) cells within the organ of
Corti sense the frequency selective vibration of the basilar membrane. The vibration-envelope
of the latter delivers the basis for the auditory sensations of sound-fluctuations and -roughness
[12]. The attribute of buzz given to the signal Synth-1 is typical for a “rough” sound. Classical
psychoacoustics defines roughness as the sensation belonging to a fast signal modulation.
“Fast” modulations are those with a modulation frequency of between 20 and 200 Hz.
♣
M. R. Schroeder, IEEE Trans. Inf. Theory, 16 (1970), 85-89.

At 82,3 Hz, the frequency distance of the spectral lines of all three synthetic signals is very
close to 70 Hz (i.e. the reference frequency for roughness-scaling). However, besides the
modulation frequency we need to also evaluate the time-functions of the excitations on
adjacent ranges of the basilar membrane: their cross-correlation functions are a kind of
weighing-function for the overall roughness♣ that is generated from the sectional
roughnesses. In Synth-1, all frequency bands are active concurrently – shown in Fig. 8.19 by
the fact that the red ranges lie on top of each other (for the same t-values). Concurrence is a
required condition for roughness. In Synth-2 (Fig. 8.20) the red ranges are dispersed; they
appear in the individual frequency bands at different times. This is the reason why the
resulting sound is not a buzzing one – but rather one of a rusteling character.
Besides assessing the roughness of the signals, the subjects also judged the perceived size of
the sound source. This is a typical phenomenon in perception psychology: while the objective
size of the sound event (the dimensions of the loudspeaker) remains unchanged, the size of
the auditory event varies with the changes in (relative) phase. Synth-1 appears to arrive
punctiformly from the middle of the loudspeaker membrane, while Synth-2 seems to be
radiated from a range in space. The latter does not appear very big (maybe 10 cm by 10 cm)
but is still not punctiform. And something else attracts attention: all sounds except Synth-1
seem to originate from behind the loudspeaker; they have more spatial depth. This impression
is created in particular if first Synth-1 is listened to, and then one of the other synthetic
signals. An explanation could be that the hearing system is not able to detect any echoes in
Synth-1, and interprets the other two synthetic sounds as similar but containing very early
echoes. Echoes do lend spaciousness and size, even when arriving from the same direction as
the primary sound.
In summary: the frequencies of the partials of a real signal are spread out, but this spreading-
out is merely of secondary influence on the pitch. If we compare the real signal with a
synthetic one that carries the same partial levels as the real signal but has the partials set
harmonically (i.e. not spread out), a very similar aural impression results as long as the phases
of the partials are chosen such that the crest-factor does not become too high. If, however, all
phases of the partials are set to zero, a different, more buzzing sound results that seems to
originate form a point in space (for loudspeaker presentation), while all other sounds are
perceived to originate from a range in space.
Next, the synthesis is modified such that the frequencies of the partials are defined via the
spreading formula given above (b = 1/8000). Synth-4 is a synthetic signal with the
frequencies and the level-progressions of the partials corresponding to those of the real signal.
Differences exist in the phase of the partials (in Synth-4 these are all at zero), and in the
details of the progression of the levels of the partials. As already noted, the partials decaying
with beats are replaced in all synthetic signals by exponentially decaying partials. Right off
the bat, the inharmonic synthesis is convincing: Synth-4 is barely distinguishable from the
real signal even given headphone presentation. And yet, the two time-functions and
spectrograms show differences (Fig. 8.21) … but this was to be expected: the synthesis is
limited to merely 45 partials (f < 4,1 kHz) that all decay with a precisely exponential
characteristic.
♣
W. Aures: Ein Berechnungsverfahren der Rauhigkeit, Acustica 58 (1985), 268-281.

Fig. 8.22: Synthetic signal with spread-out spectrum (Synth-4).
The spreading of the partials leads to a progressing loss of synchronization in the time
domain. At the instant of plucking (attack), all partials need to cooperate with equal phase in
order to effect the abrupt change in signal. In Synth-1, the attack is repeated in identical
shape: the maxima appear at the same times (pan-cochlear synchronicity); the tone buzzes. In
Synth-2 and Synth-3, this pan-cochlear synchronicity is by and large destroyed, but the period
of excitation remains constant in all critical bands. In Synth-4, the period of excitation
decreases with increasing frequency, and the cross-correlation function (that the formation of
roughness is based on) becomes time-variant. It is no issue that – due to the intra-cochlear
time delay of 6 ms (max.) – a true pan-cochlear synchronicity does not actually appear: the
hearing system is used to that. All impulses suffer the same fate … and still remain one
object.
It is not a matter of course that changes in the phase spectrum become audible. If we would
repeat the above experiment with a fundamental frequency of 500 Hz, the mentioned phase
shifts would still change the time function, but they would not be perceived. It has proven to
be purposeful to assume the time-resolution of the auditory system to be about 2 ms: at a
fundamental frequency of 82 Hz, the hearing can still “listen into the gaps” but not anymore
at 500 Hz. However, apparently a particular sensitivity towards how of the critical-band-
specific loudness evolves over time does not exist: Synth-1 is clearly recognized as being
different, while Synth-2 and Synth-3 sound very similar despite different cross-correlation
functions. It should be noted that this similarity is subject to inter-individual scatter: it may
happen that a special sound is perceived as tuned too low. Changing the fundamental
frequency (e.g. from 81,9 Hz to 82,3 Hz) removes this discrepancy … now we are in tune.
Perfectly, even. A few minutes later, however, the same tone is suddenly too high – and needs
to be retuned down to e.g. 81,9 Hz. In the best case, our hearing may notice a frequency
difference of 0,2% [12]. It may – doesn’t have to. The listening experiments convey the
impression as if the attention of the test-person works selectively: sometimes, more attention
is paid to pitch, other times roughness is in focus – or other attributes that go beyond the
scope of generally understandable adjectives for sound such as “steely”, “wiry”, “metallic”,
“rolling”, or “sizzling”, “swirly”, “brown”. We seek to describe the remaining difference in
the color of the sound somehow, but semantics do let us down here. And then: lets hope that a
translation into another language is never needed. Who would think that "kinzokuseino"
means metallic? Or that "hakuryokunoaru" means strong? What does "namerakadadenai"
sound like? Can you hear “roughness” in there? Or “r-aow-hig-ka-it" (to try – and fail – to
represent the German word Rauhigkeit for this attribute)?

Most partials of the real guitar signal decay in good approximation with an exponential
characteristic, but with some we observe a beating. The reasons for this shall not be
investigated here – we are looking into auditory perception at this point. Already the second
partial gives rise to the conjecture that a beating minimum would occur shortly after the end
of the recording (duration 2 s), i.e. a beat-periodicity of about 5 s. Within the duration of the
listening experiments (1 s), this can still be nicely approximated by an exponential decay, but
in the 17th partial there are two beats in combination: a slower one with 1,6 Hz beat-
frequency, and a faster one with 18,4 Hz (Fig. 8.23). This partial has, however, a low level (in
particular relative to the 15th partial), resulting in this beating being practically unperceived –
it is masked [12]. For the 27th partial, we find an again different scenario: if features a
classical beating with a periodicity of 950 ms. At first glance there seems to be no strong
masking: all neighboring partials have similar levels – but they all decay relatively smoothly
such that the overall critical-band-level (that is formed from the levels of 4 partials) features
almost no fluctuation. The levels of the partials obtained via narrow-band DFT-analysis
deliver objective signal parameters but do not allow for any conclusion about the audibility of
special sound attributes. Psychoacoustical calculation methods such as roughness- or
fluctuation-analysis also are to be taken with a pinch of salt: our knowledge about the
interaction in inharmonic sounds is still too limited. Listening experiments yield the best
results about the audibility of beats in partials – no surprise there, of course. In the case of the
above guitar tone, they lead to the clear statement: despite inharmonic partials, beating is
practically inaudible.
Fig. 8.23: Decay curves of individual levels of partials; Ovation-guitar, piezo pickup.
Still, we must not conclude from the fact that no beats are perceived in the guitar tone
presented here that beats are inaudible in general. They are present, and they will be audible if
the levels of their partials stand out sufficiently from their spectral neighborhood. Cause for
the beats may be found in magnetic fields of pickups (Chapter 4.11), or coupling of modes
within the string bearings (Chapter 1.6.2). The inharmonicities of partials, however, can
(regarded by themselves) generate only minor fluctuations. Beats within octaves [Plomp,
JASA 1967] or time-variant cross-correlations [Aures, Acustica 1985] explain only very
subtle fluctuations – partials creating a clearly audible beating require two spectral lines that
are in close vicinity, and of similar level. Such lines cannot be generated merely by
inharmonicity, though. “In the LTI-system”, we are temped to add in order to have really
thought of everything … and we suddenly realize that in particular this limitation is not
fulfilled in many cases for guitar amplifiers. Spectral inharmonicity can certainly generate
neighboring tones if non-linearities are allowed!

In guitar amplifiers, non-linear distortions appear to various degrees. While the acoustic
guitar amplified via piezo-pickup will usually not be given audible distortion, the contrary
might be the case for the electric guitar with its magnetic pickup (depending on musical
style). A non-linearity – or, to put it simply, a curved transmission characteristic – enriches
the spectrum by additional tones. A mixture of two primary tones (at the input of the
nonlinearity)
is mapped onto the output signal y(t) via the nonlinear transfer function (in a series expansion)
For purely 2nd- or purely 3rd-order distortion, the spectrum belonging to y(t) may be easily
calculated [e.g. 3]. For distortion of any order, the above input signal will create a distortion
spectrum that is harmonic relative to the new fundamental frequency ggt(f1, f2). The operation
ggt(f1, f2) determines the largest common denominator of the two frequencies f1 and f2. Given
e.g. f1 = 500 Hz and f2 = 600 Hz, a distortion spectrum with spectral lines at the integer
multiples of 100 Hz results, while for e.g. f1 = 510 Hz and f2 = 610 Hz, a distortion spectrum
at integer multiples of 10 Hz is created.
If we generalize the two-tone signal x(t) to an n-tone signal, then the distortion spectrum of
the latter will be harmonic relative to a fundamental frequency corresponding to the largest
common denominator of all n frequencies of the participating primary tones. If x(t) is a time-
periodic signal with the periodicity of T, then its spectrum will be harmonic, i.e. all
frequencies of the partials are an integer multiple of fG = 1/T. The largest common
denominator of all frequencies of the partials is also fG, and therefore a non-linearity does not
change the harmonicity (or the time-periodicity). However, given a spread-out spectrum, a
vast variety of new frequencies is created (the root-function is irrational), and these create a
noise-like or crepitating additional sound. Fig. 8.24 depicts the spectrum resulting from a
time-periodic signal (Synth-1), and a synthetic signal (similar to Synth-1 but with b = 1/3000),
both being fed to a point-symmetric distortion characteristic. In this conglomerate of
superimposed primary tones and distortion tones, everything is possible – including beats.
Fig. 8.24: Spectra of signals subjected to non-linear distortion. Left: harmonic primary signal; right: spread-out
primary signal. Cf. Chapter 10.8.5.

Conclusion: due to their bending stiffness, strings do not have a harmonic spectrum but a
spread-out spectrum; therefore the corresponding time-function is not periodic. If we compare
the inharmonic sound with a harmonic sound (of the same fundamental frequency) that
features levels of partials at least approximately corresponding to those of the inharmonic
sound, we realize that the phase of the partials is significant. A harmonic sound carrying
partials that all have a phase of zero (or π) sounds buzzing and clearly different from a real
guitar sound. However, given a suitable phase function that creates a small crest-factor,
harmonic tones that can be synthesized that differ only marginally from a real guitar sound.
Using headphones, the trained ear may still recognize differences, but with loudspeaker
presentation, the sounds are practically identical. The inharmonicity is clearly noticed only if
the spreading parameter b is set significantly above about 1/5000 (this would not be typical
for guitar strings). For example, at b = 1/500 a dark chime like that of a wall clock results,
while with b = 1/100 synthesizer-like sounds are created. However, if a strongly non-linear
system (such as a distortion box) is connected into the signal path, even weakly inharmonic
signals may drastically change their spectrum (including additional frequencies) and thus their
sound. In such a configuration, harmonic signals experience a change in amplitude and phase
only – they remain harmonic.
These statements should be interpreted as results of a small series of experiments and not be generalized to every
instrument sound. The aim of these investigations was not to find the absolute threshold for perception of
inharmonicity but to demonstrate the rather small significance of guitar-typical inharmonicities. If the decay of
higher-order partials is different, inharmonicities based on a much smaller inharmonicity parameter may well be
noticed (Järveläinen, JASA 2001).
Compilation of formulas:
Synth-1
The function of the angle was formulated as a sine in order to not make the crest-factor even larger.
Synth-2
The phase angles ϕ(i) are equally distributed within the interval [0...200°].
Synth-3
2
The phase angles ϕ(i) are calculated (according to Schröder) as ϕ(i) = 0,04⋅π⋅i . This corresponds to a group-
delay linearly increasing with frequency.
Synth-4
The frequencies of the partials are inharmonically spread out.
; fG = 81,9 Hz; b = 1/8000; i = 1:45;

8.3 The character of the musical keys 8-37
8.3 The character of the musical keys
"And indeed it appeared that for Beethoven, certain keys had certain characters that made
them useful for corresponding moods and content♣". This statement and similar ones have led
to attribute to a musical key an absolute character, in the sense of Eb-major = heroic, C-major
= impersonal, E-major = solemn. This may have different reasons:
a) Musical keys do have a character. Brilliant musicians (such as e.g. Beethoven) have
recognized this, and have oriented their compositions accordingly.
b) For whatever reason, brilliant musicians have believed in a character of the keys. Admirers
of their music have internalized this, learned from it, imitated it, passed it on … and thus a
self-fulfilling prophecy came into being: because Eb-major sounds heroic, heroic music is
composed using Eb-major … and so Eb-major sounds heroic.
c) The whole shebang is nothing but coincidence.
Mies reports from an experiment that points to the existence of absolute character. He played
Schubert’s Impromptu to about 20 pupils: once in G-major and Gb-major each. He asked
which of the two was the original key. 3 pupils voted for G-major and the rest for Gb-major.
They reasoned that the key they chose fitted better to the mood of the piece. Mies knew about
the limited validity of such a single experiment and started systematic investigations with a
multitude of piano pieces. At first glance, his results are contradictory: on the one hand he
does arrive at a character correlation (see table below), but on the other hand he summarizes:
“and here the investigations are clear proof that there is no general character of the keys
across ages and composers, meters and time, rhythms and melodies. A general character of
a key that would be independent of composer, time, listener, etc., does not exist.” In fact,
this summary does not actually contradict the table because the latter answers the question of
which matching the investigated composers have preferred. If D-minor feels passionate to
Brahms, this is of no more significance than the statement that Eb-major was Beethoven's
favorite key. Who would deny any great composer a subjective preference? A pars-pro-toto
principle is, however, not justified by this.
C-major: Objective, superficial, impersonal. Key of truth. For thanks and salute.
C#-major: Glimmering, sparkling, lively, virtuosic.
Db-major: Soft, gentle, emotional.
D-major: Key for marches, fanfare, cheerfulness, joy, festive splendor, scenes of revenge.
Eb-major: Serious, grave, deep love, tormenting lovesickness.
E-major: Solemn, serious to gloomy, belongs to exalted and otherworldly moments.
F-major: Friendly, natural, moderate.
F#-major: Passionate, ardent love.
G-major: Simple, uncomplicated, cheerful.
Ab-major: Quiet, emotional, longing. Sinister scenes.
A-major: Manifold, lovely, serenade-like. Key of happy people. Expression of splendor.
Bb-major: Cheerful, playful, gently. Cordial sentiment.
B-major: No general character.
Table: The character of the keys. Paul Mies, 1948. Strongly abbreviated representation.
♣
P. Mies, Der Charakter der Tonarten, Staufen, Köln 1948

In his account, Mies explicitly points to the enharmonic identities and refers to the piano
tuned to equal temperament. If the keys had an absolute character, Db-major could not appear
soft and gentle while C#-major is sparkling and lively. That C#-major appears virtuosic – that
is conceivable to the musician given its no less than 7 accidentals! If we look long enough, we
do find contradictions: Gb-major = sad; F#-major = passionate. On the other hand, a lot can
indeed be made to fit: the CD tries to make us believe that "Roll Over Beethoven" (with
apologies, dear & highly esteemed Ludwig van) is played in Eb-major. Tormenting
lovesickness? Maybe not … presumably heroic – after all, Chuck B. is the true R’n’R hero to
many. However, what does the songbook tell us (the authentic transcriptions with notes and
tablature)? D-major!! That would be “joy”, then! That’s gotta be it: the tape machines ran a
few percent slow back in the 1950’s, and Chuck must have certainly played in D-major.
Cheerfulness, joy, festive splendor – that’s really like him! However, Mies also lists “scenes
of revenge” related to D-major, and lo and behold: “Don’t you step on my blue suede shoes" –
this little sideswipe is put in proper perspective, too. Or, how about something from the
Stones’ songbook: "Let's Spend The Night Together" of course is in Ab-major. That fits him
to a T: longing and emotional chap that he is, our Mick. Did somebody say “sinister scenes”?
Rather, a certain “sensitivity and delicacy of feeling”, as Riemann elaborates. “Of practice-
character”, Mies complements, and is bang on target (you need some guiding, baby). Typical
for Ab-major are also “the mid-tempos taking up the most extensive space” (I'm in no hurry I
can take my time) and "medium and slow tempi with frequently on-going movement”: an
excellent match for what the Stones’ front-man stands for. Not to forget: “sweet, romantic
melancholy and longing” (now I need you more than ever), as well as Stephani’s "soft-solemn
seriousness” (Oh my, da da da da da da da da). And finally: “ the movement often
perceivable in the tempered pieces is also felt in the accompaniment” (around and around, oh
my, my, yeah) – perceivable movement in the background vocals, indeed. Much could be
added here, for example the 19th Nervous Breakdown (E-major, otherworldly moments) or
Street Fighting Man (F-major, friendly, natural, moderate). And many more …
Still, we do see criticism, as well: “in view of all these statements, an absolute character of
Eb-major certainly cannot be observed.” Or: “indeed, literature does not agree about Ab-
major”. Or: “would it not be possible that Beethoven’s quotation (not actually from his own
notes) was not correctly handed down in its relation to the keys; no support can be found in
his own works.” These are all Mies’ citations. It would also be possible that everything is one
big misunderstanding.
In the course of the last centuries, highly diverging opinions can be found about the absolute
tuning of an instrument: the chamber pitch (concert pitch, standard pitch), i.e. the middle a (a',
A4), varied in its frequency by as much as 337 – 567 Hz! Even going back only to the 18th
century (checking in with Beethoven or Brahms), we still find a scatter of just shy of a
semitone). That could turn A-major (“key of happy people”) into Ab-major and thus call for
“sinister scenes”. Nightmarish: the A-major scherzo played too low by a hair becoming the
hotbed for sinister Hans-in-Lucks and happy gloom-o-philes – a cut set of joy and sorrow?
"Die then, die now, die! Haha! Hahaaaa! Hahahaha! Die! Die!" "Welcome oh blissful woe –
continue, go on.” Without doubt: A-major, a quarter-step too low?
Nay, psychoacoustics does not know of a “tonal character” based solely on the frequency
position. It is still conceivable that Schubert’s Impromptu sounds more authentic in Gb than
in G. There are no known recordings of Mies’ experiment – we can therefore only speculate:
Mies had presumably rehearsed the piece in Gb, with the transposition to G requiring
different finger movements and possibly resulting in a different sound character just because
of that. The experienced subject can detect timing-differences as small as 5 ms (Chapter 8.5)!

8.3 The character of the musical keys 8-39
Also possible: the special tuning of the piano used resulted in characteristic beats that of
course are key-dependent. Specific resonances of the individual piano may have played a role,
changing individual notes/passages/chords in a key-specific manner. And finally it cannot be
excluded that Mies (who knew when he was playing in the original key) was himself not
convinced of the use of G-major, and therefore played with inferior expression in that key.
For a double-blind test this was NOT.
Could we repeat the Mies-experiment as a double-blind test? We would require a very good
pianist who practices both the G- and the Gb-version with the same dedication. That would
seem doable. Presumably, however, this pianist would (wittingly or unwittingly) prefer one of
the two versions, and thus would not be able to play both with the same expression. In this
case the listeners would assess the way of playing and not primarily the key. We would
therefore have to directly ask the pianist (or several pianists) but this would put the general
validity of the experiment. Given modern options, a purely electronic transposition would be
feasible: the piece is recorded e.g. in Gb-major and reproduced with a 6% higher speed (or
sampling frequency). But then not just the key changes but also the timing: the G-major
version is faster by 6% compared to the Gb-version. That’s not optimal, either. Using
harmonizers or pitch-shifters (special equipment used in recording studios) that change the
pitch without influencing the reproduction speed calls for skepticism, as well, because with
them the subject may judge the quality of the signal-processing algorithms and not just the
character of the key.
Conceivable would be the following approach: the pianist plays the piece in the original key,
and the key movements are electronically recorded (via MIDI or something better). From the
stored data, artificial piano sounds can be created – both in the original key and in a
transposed version. This ‘electronic music’ may now be judged with respect to the character
of the key♣. Today, psychoacoustics assumes that such music has no inherent key-specific
character, i.e. that aggressiveness, passion, or sorrow need to be expressed by means of
harmony and rhythm.
However, this does not imply that the character of a piece of music accompanied by the guitar
cannot change if the piece is transposed form G-major to A-major. If the guitarist plays a G-
major chord without barring strings (g-b-d-g-b-g), and changes to the ‘open’ A-major chord
(e-a-e-a-c#-e), the color of the sound will change significantly. However, this is not due to the
different key, but results from the different chord composition. In the G-major chord the fifth
appears only once, but three times in the A-major chord. Conversely, an A-major chord
played in the 5th position (barré on the 5th fret) has only two fifths. Thus the simple conclusion
is: when changing keys, the character of the sound can change – however this is not according
to a generally applicable scale but specific to the respective interpretation and instrument.
♣
Similar experiments had already been carried out by Terhardt and Seewann – however with sounds that
differed from those of the acoustic piano. The objective of these tests related to perfect pitch (absolute pitch) and
not to the character of the musical key [Aural key identification and its relationship to absolute pitch. Music
Percept. 1, 1983].

8.4 Consonance und dissonance
Although every musician thinks he/she knows what a dissonant chord is, a scientific
description proves to be difficult. The Pythagoreans considered the octave, the fifth, and the
fourth to be consonant (symphonical), and all other intervals as dissonant (diophonic). There
is one peculiarity: to the astonishment of the westerner, now and then the major second is also
designated as being consonant. This has its basis in the totally different concept of melodic
consonance: consonant is that which is easy to pitch (Simbriger/Zehelein). Another
explanation would be: the major second must not stand too far apart since it is the fruit of the
“holy matrimony” (Fig. 8.6), and thus sanctioned via the insights of a “advanced civilization”.
Apparently there is more than one type of consonance. As a synonym, we often find euphony,
coalescence, serenity, relaxation. Playing in the middle range on the piano two notes at a
distance of a fifth, both melt into one harmonic sound. The two notes “like each other”, they
sound well together, and that is exactly what con-sono means. Very different are two notes at
a distance of a half-step: the esthete downright hears the fight they slug out, while the signal-
theoretician detects beats, the psychoacoustician notices roughness – and the musician
perceives dissonance.
Already early on, the nominal attribute became a ordinal attribute: for dyads, not only a
statement was sought that they harmonize well, but also an assertion about how well they
harmonize (concord) – in the sense of a ranking. Franco von Köln put together a five-step
scale in the 13th century (C. = consonance, D. = dissonance):
Complete C. Medium C. Incomplete C. Incomplete D. Complete D.

Prime Fifth Major third Minor third Second
Octave Fourth Minor sixth Major sixth Seventh
The high consonance of the fifth is already evident from Fig. 8.2: in the spectrum of the
partials there is a close relationship. The 3rd, 6th, 9th, etc. partials of the lower note have the
same frequency as the 2nd, 4th, 6th, etc. partial of the higher note – given perfect tuning and
dispersion-free wave propagation. What could be closer than to derive, from the similarity of
two notes, rules for the generation of consonance and dissonance? For example:
• The more shared partials, the higher the consonance. Or:

• The simpler the frequency ration, the higher the consonance.
However, there were also cautious rearguard actions: “essential are only the odd-numbered
partials”. Or: “the 7th, 11th, 13th, 14th, and 17th partials are excluded”. Or: “the fourth is a
perceptional dissonance”. Or: “there are dissonant chords of highly consonant sound”. Or “In
context, a consonant chord very often is bestowed a dissonant purpose”. And rather recent
from Haunschild’s ‘New Theory of Harmony’ (1998): "In general we can note that the human
understanding of consonance and dissonance more and more shifts away from consonance, in
favor of dissonance. This means that more and more intervals and chords that were surely
classified as dissonant back in the day, are today rated as consonant. It is only the intervals
with a so-called semi-tone-friction (minor second and augmented seventh) that are truly
assessed as dissonant.”

8.4 Consonance und dissonance 8-41
Let us give the philosophers some space, as well: "corresponding to the relations of the
natural degrees of consonance it is possible to say that every entity, every form of being is the
more complete, and thus the more in harmony with its physical and social environment, the
closer it is positioned to its origin. The principle of consonance is the connecting within the
differing – it therefore corresponds to the harmony, the organic integrated-ness in higher
unity, in other words: love" (found in Simbriger/Zehelein). So then Schönberg possibly was a
love-less person? He opined: “today we have already gone so far as to not make a difference
anymore between consonance and dissonance.” Rossi similarly (but not quite as radically)
says: "consonance and dissonance greatly depend on each individual's musical experience,
and, more broadly speaking, musical culture".
It shall not be disputed that beats, roughness, fluctuations, frictions, or anything else you
would want to call the envelope variations of the partials, represent a cause for the
perception of dissonance or consonance. However, perception psychology increasingly
distances itself form the so-called unbiased scaling, i.e. an absolute, purely signal-dependent
scaling. At the 8th Oldenburg Symposium, Viktor Sarris elaborates: "Whereas classical
sensory psychophysics relies mainly on the (illusory) assumption of absolute, i.e. invariant
stimulus-response laws, the relation-theory in psychophysics is based on the general premise
that, on principle, one and the same stimulus may be perceived and judged very differently as
a function of the variables implied by the total 'contextual' situation at hand. … Contextual
effects in psychophysics are of major importance since virtually all kinds of sensory-
perceptual-cognitive judgments, whether in direct or indirect scaling resp. in discrimination
and postdiscrimination-testing, are contextual." The insight that evaluations happen in
relation to the given situation also concerns judgments of consonance – in particular if these
are delivered by persons with musical experience or education.
We may use as an example a dyad with the two tones forming a major sixth – i.e. for
example B-G#. Let us imagine two guitar players: one of them continuously plays an E-major
chord, the other frets the B on the G-string, and alternatingly (e.g. with a 6/8th rhythm) the G#
on the high E string. Both B and G# are included in the E-major triad; the two guitars play in
harmony and the result is a tension-free sound. Now the “man of the 6th“ shifts his fretting
hand upwards by 3 semi-tones, i.e. he frets the D on the G-string and the B on the E-string.
After one bar he shifts upwards by another 3 semi-tones and plays F-D (Fig. 8.25). All the
while the accompanying guitarist continues playing the E-major chord. The second sixth is –
with D-B – still close to E-major; the D (representing the minor seventh) does already build
some slight tension, though (E 7-chord). However, only the third sixth brings some serious
dramatics to the game: the D can again be taken as the minor seventh, but the F – representing
the minor ninth – is dissonant to a high degree (E 7/ b9-chord). Every player of the electric
guitar with some classical education (i.e. Beatles-Beck-Blackmore) knows this skewed chord
from Lennon/McCartney’s I want you. The interesting thing in this example is: even if no
accompanying guitar is playing along, the experienced payer of sixths still hears these
mounting dramatics! The latter may be relaxed (resolved) e.g. via a concluding augmented
sixth to E-C#. Again: a guitarist plays (now without accompaniment) the augmented sixths:
B-G#, D-B, F-D, E-C#, and he/she hears an arc of suspense – although always an equal (not
one and the same!) interval is being played.
Fig. 8.25: Augmented Sixths.

Requirement for the changing musical tension is a reference carried along in memory –
which needs to be available to every musician. Otherwise there would be no way to play an
improvisation that is guided by accompanying chords. Now, the well-versed musician will (in
contrast to the beginner) not need any audible accompaniment at all – he/she will generate it
“within”, using the “internal ear”. The whole thing is less esoteric that one might fear. The
reader could, as an example, begin to speak but stop at the last moment – intending to say “a”
but keep the vocal chords shut. Dutifully, the tongue will already have moved into position,
and a well-formed notion of how the vowel will sound (had only been allowed to do so) has
emerged. The “internal ear” has already heard the “a” although the latter has physically not
manifested itself at all. A vocalist could in addition also already set the vocal chords to the
appropriate tension in order to produce a targeted pitch; however, already this will not work
as well anymore without vocal training. The reason is that the internal ear requires
connections between the motor-control areas and the sensory areas in the brain. Strangely,
when it comes to hearing, the sensors not only comprise the 8th brain-nerve (N. acusticus). If a
layperson-singer (i.e. in this case a person that wants to sing but lacks any skill) is played a
note and then asked to sing it, a more or less horrible control process♣ starts: the vocal chords
generate a tone but only as the latter is physically existent can the hearing recognize the pitch
and make the vocal chords change their tension. An expert singer, however, is expected to
immediately produce the correct pitch without any interfering control processes. This he/she
can do, too, because he/she has learned to pre-tension the vocal chords correctly already
before the tone sounds (“muscular tone-memory”).
Magnetic resonance imaging has enabled us to “watch the brain thinking”, and we have
started to understand how the individual brain regions cooperate. Or rather: we have a certain
conjecture, because an actual comprehensive grasp has yet to be established. Some interesting
connections have been observed in pianists: if a pianist listens to piano music, regions in
his/her brain that are assigned to the fingers become also active. Presumably, the brain already
practices how the fingers would have to be moved in order to play what is heard – even
though the pianist merely listens and does not actually play. This works the other way round,
as well: playing on a keyboard that does not sound any audible notes still activates brain
regions related to auditory perception – that is the “internal ear”. With beginners of the piano,
these senso-motoric connections have, by the way, not been observed. Rookies need to first
configure the hardware by practicing.
But back to our topic of consonance: at least the well-versed musician supplements the
sounds aurally recorded by fundamental and accompanying notes that exist only in his/her
imagination. The supplement may be more or less consonant, and therefore consonance is
describable by physical signal parameters alone. The major sixths mentioned in the above
example will generate an increasing tension only if the E, or the E-major chord, are retained.
If the listener thinks of a concurrently changing fundamental note (in the example i.e. E – G –
Bb), then the tension is not changed. Setting the respective current reference is an individual
process that will follow some roughly predefined rules, but it will not run a predetermined
course in the individual case. Rather, musical training as a general criterion, and musical
context in particular, are significant. It is easily imaginable that probabilities related to the
given choice of the fundamental are set up and evaluated, and that relations within the
partials, as well as chord relationships, play an important role. After all, the brain is most
powerful in supplementing missing sections in visual impressions – there should be
similarities in the auditory system.
♣
We are familiar with this from casting shows that have spawned frog-like superstars (Kermit on dope), the
skillfulness of which with regard to intonation have called for critical voices to speak up even within Lower
Bavaria (!).

The third sixth-dyad (F-D) described in the above example may be seen as part of a tetrad in a
third relationship. Two tetrads are third-related if three of the four notes in the two chords
correspond, and if moreover the root notes are located a third apart. In the example, the first
two sixth-dyads form – with B-G# and D-B – the basis for an E 7-chord. The third sixth-dyad
is part of a diminished seventh-chord (G#07). Third-related to E 7, it forms an E 7/ b9-chord with
the latter. This rule of formation is not compulsory; alternative reference systems may be
imagined. Indeed it is specifically the possibility of multiple reference systems that renders
the degree of consonance not unambiguously definable.
Fig. 8.26: Left: third-related

tetrads; right: arc of suspense
with resolution in A-major.
“Gis” = G#
Fig. 8.26 clearly shows the third-relation mentioned above: E-G#-B-D forms an E 7-chord that
has three notes in common with G#-B-D-F (G# o7). The latter supplement the E 7 to an E 7/ b9-
chord. The right-hand section of the figure depicts the first sixth-dyad (open note-symbols),
and the mentally supplied root note E (filled symbol). This pattern is stored in memory, and
the next sixth-dyad is added, resulting in the E 7-chord. The latter is memorized as well (filled
symbols) and supplemented by the third sixth-dyad … and there we have our dissonance.
Actually played are merely the notes given by the open symbols; all other notes exist only in
memory. In case the guitarist plays, in conclusion, also the major sixth E-C#, a nice resolution
(relaxation) in A happens; this works in particular if he/she imagines E-A-C# in addition.
The above example was intended to show how the consonance of a major sixth can turn
dissonant – if the imagination (the internal ear) plays along. Of course, not only the imagined,
but also the notes existing in reality influence the perceived dissonance. In general, the major
seventh (e.g. E-D#) is considered to be dissonant. However, if it is generated using two sine-
tones, “actual” beats do not happen (in contrast to the minor second E-F), but octave beating
(so-called 2nd-order beats) results. Experiments tapping the electrical potentials of the cortical
nerve give rise to the assumption that our hearing system performs some sort of half-wave
rectification within the analysis of the vibration of the basilar membrane♣. The patterns seen
in the action-potentials on the nerve fibers change their shape in the same rhythm as the
difference frequency (in this example defined by T = 1 / (f2 –2f1)). Fig. 8.27 depicts such a
signal; the shown section corresponds to just this beating-periodicity. To compare: in Fig. 8.5,
a 1st-order beating was shown. 2nd-order beats act in a more subdued fashion compared to 1st-
order beats [Plomp, JASA 1967].
Fig. 8.27: Octave beating. Sum of two primary tones of the same level. The frequency of the higher tone is
larger by 2,5% than twice the frequency of the lower tone: f2 = 2,05⋅f1.
♣
At least in the frequency range below 1,5 kHz.

Tones from instruments are almost never composed of only one single partial, though. In the
guitar, we will normally have to deal with several partials – and in this case 1st-order beats do
determine the sound, as the following example will show. Playing the just mentioned major
seventh on the guitar (e.g. the E on the D-string and the D# on the B-string) indeed yields a
sound that most would call dissonant. However, as soon as we supplement additional tones to
these two tones to form a complete E maj 7-chord E-B-E-G#-D#-G#, the dissonance is gone♣.
Causes may be found in the many consonant intervals that this chord features, or in the
destruction of the strong envelope fluctuations by the additional partials. What is interesting
in this context: in the chord sheets of e.g. the book “Rock Gitarre” (Bechtermünz publishers),
a different E maj 7-chord appears: E-B-D#-G#-B-E. These are the same note-designations as
above, but the root position has changed. The chord rumbles a bit and does not ring with the
same beautiful melancholy as the chord mentioned first above. But again this is a subjective
assessment. In fact, there can be no wrong chords – only wrong expectations.
Fig. 8.28 shows both E maj 7-chords in comparison. The spectra are based on equal-
temperament tuning; all partials have (arbitrarily) the same amplitude. In the second chord,
two partials with only 9 Hz distance appear at 160 Hz – they generate a fast beating that
sounds rather unpleasant. The neighboring partials at 415 Hz have a distance of 3 Hz: they
beat, as well, but slowly and more in the sense of a vibrato i.e. less annoying. What’s
happening at 311 Hz / 330 Hz? Here we have the intended dissonance of the major seventh
that showed up already in the first chord – given by the E- and D#-partials.
maj 7
Fig. 8.28: Amplitude spectra and musical score of the E -chords elaborated in the text.
The closer two partials are spectrally located, the slower the resulting beats. Very small
distances of partials (e.g. 1 Hz) happen in single notes, as well – due to slight detuning of the
circular string-polarization, due to progressive spreading of partials, or because the instrument
is polychoral (e.g. the piano). Somewhat faster beats (e.g. 4 Hz) may also appear for single
notes, for example if the tone is generated using vibrato or tremolo. Even faster beats that are
in part perceived as fluctuation strength [12] and in part as roughness, are typically only
generated as several tones are played simultaneously. The borderline between fluctuation
strength and roughness lies at a modulation frequency of about 20 Hz. Tones modulated that
way – whether rough or fluctuating – can diminish the euphony and sound dissonant. As the
modulation frequency further grows, the impression of dissonance decreases again –
otherwise already the (harmonically complex) 100-Hz-tone would be dissonant (which it
isn’t). It may be deemed rough, but not dissonant. It cannot be specified by a single number at
which distance between the partials a maximum dissonance occurs; the terms consonance and
dissonance are too complex, and the sounds are too diverse.
♣
Again, this is naturally a matter of the approach taken, and may be subjectively judged differently in the
individual case.

Psychoacousticians like the sensory to separate consonance and the musical consonance, or
similar (often historically established) terms. Sensory consonance is represented in the
absolute scaling, the "unbiased Scaling" that psychologists will readily put into question. So:
put on the headphones, don’t think of anything bad (and of course not of anything good,
either), and evaluate the consonance of the two sine-tones presented. Just to avoid any
misunderstanding: that is not pointless – from this we obtain elementary basic knowledge that
may at some point form the fundamentals for a comprehensive theory on dissonance.
However, it is still a long way from the dissonance of two sine-tones to the dissonance of a
E maj 7-chord. That is true not only because here musical context, musical experience, and
culture need to be involved (all elements of the musical consonance, also termed tonality), but
also because already the purely psychoacoustical analytic poses considerable problems. Issues
easily dealt with given an AM- or FM-tone turn voyage-into-the-unknown for a chord. As
nice as the formulas about frequency- and level-dependencies of roughness and fluctuation
strength are – they are of no help when dealing with signals containing complicated, time-
variant partials. That E maj 7-chord has neither a modulation frequency nor a modulation index
– just like the car engine the roughness of which has kept generations of acousticians busy.
The synthesis of specific roughnesses proposed by Aures [Acustica, 58, 1985] shows a way
but also reveals problems: we need to know not only the level of every partial (this could be
measured) but also determine the phases of the partials because a cross-correlation is required
across the specific roughnesses of neighboring frequency bands. That implies the time-
functions, and therefore the phase is of importance. Your customary analyzer will, however,
model only the damping function of the hearing-specific critical-band filters with reasonable
accuracy. Not much is known yet about the (level-dependent!) phase response of these filters;
and even if we would have that information, we would still have only captured one single
dimension. Because: One and the same stimulus may be perceived and judged very differently
as a function of the variables implied by the total 'contextual' situation at hand [Sarris].
If we don’t pitch (sic!) our expectations that high and content ourselves with qualitative rules
– then we can actually explain quite a few phenomena. Such as: if on a guitar the low E (open
E-string) and the D# on the B-string (4th fret) are plucked with the fingernail, a dissonant dyad
is sounded. The dissonance is significantly diminished if the fingertip is used for plucking.
Explanation: dissonant beats may occur between the fundamental of the D# and the 4th
harmonic of the low E. This will only happen, though, if this 4th harmonic is present with a
sufficient level. Plucking with the fingernail or the plectrum will emphasize harmonics and
generate a sufficiently strong 4th harmonic if the strings are not too old. Plucking with the
fingertip, however, makes for a weaker excitation of the 4th harmonic – the dissonance thus is
less pronounced. The markedness of the dissonance in this example is influenced by the
playing technique (the interpretation) and cannot be determined merely on the basis of the
interval. Of course, it is highly important how well the guitar radiates these neighboring
partials, and how quickly they decay – and how the room transmits them … and whether
further strings are plucked such that individual partials are masked. Roederer♣ describes a
supplementary example: if e.g. a clarinet and a violin play a major third with the clarinet
playing the lower note, this interval sounds “smooth”. If the clarinet plays the upper note,
though, the interval sounds “rough”. The reasoning again lies in the instrument-specific
structure of the harmonics which may not only be influenced by the mechanisms in the
generator itself but also by the musician, the room and the setup in it, and of course by all
other sources that may concurrently sound. In the end, a subjective assessment happens on the
basis of the knowledge of the listener in relation to the musical context. The result is a highly
subjective degree of dissonance the may certainly not merely be calculated just based on an
interval relationship.

A tremendous range is covered from the older books on harmony1 that attested to the major
sixth a general dissonance, to more modern books2 seeking to attribute this feature only to the
minor second, from divine perfection and imperfect devil’s notes via Helmholtz-ian tone-
relations, all the way to multiple regressions3. They all share the search for rules, because:
music is played according to rules … rather complicated ones at that, though. Auditory
processing of acoustical signals also follows rules – and again the latter are all but simple …
and they are subject to inter-individual as well as intra-individual scatter.
There are good reasons to assume that auditory perceptions emerge on the basis of audible
partials. Audible are partials only if they surmount both the hearing threshold in quiet and
masking thresholds caused by other tones. ‘Audible’ in this sense does not mean, though, that
the partial would necessarily be audible as individual tone. To that effect, a partial is audible
(i.e. it contributes to the overall hearing sensation) if the aural perception changes when the
partial is filtered out. If the perception does not change, the corresponding partial is not
audible. If we regard the interaction of individual partials as the source of the perception of
dissonance, the (so defined) audibility of these partials is prerequisite. With this, however,
dissonance becomes dependent on the individual sound spectrum and can by no means be
calculated “from the score”. If, conversely, the basis is the sound spectrum arriving at the ear,
then orientating calculations are possible – albeit right now only with considerably reduced
general validity. Daniel’s3 conclusion may serve to obtain three insights: roughness and
sensory euphony are (negatively) correlated; roughness and unpleasantness are (positively)
correlated, but: sensory euphony and unpleasantness are not correlated. Daniel moreover
states: "this points to a significant difference between the opposite pairs pleasant – unpleasant
and euphonious – dissonant”. Daniel does not further delve into the subjects of pleasant
dissonances or unpleasant consonances. It is now difficult, though, to repress the question of:
what actually do subjects judge when asked about the consonance of a musical chord? Is it the
pleasantness … or the euphony?
It is not a wonder that already 50 years ago Michael Dachs4 arrived at this rationale: in
context, a consonating interval often gains a dissonant meaning. Around the same time,
Simbriger/Zehelein opine: There is barely a second problem that would be as controversial in
modern acoustics as that of consonance and dissonance. While 50 years of supplemental
research have considerably widened the body of knowledge available back then, an algorithm
for calculating consonance that is at the same time manageable to the musician could still not
be made available. Which is not necessarily a disadvantage: if you can feel it, you can play it.
Oh yeah: those musicians, always having a solution at hand. And if you can't make it, fake it.
♣ Roederer J.: Physikalische und psychoakustische Grundlagen der Musik, Springer 1999.
1
Z.B. H. Grabner, Handbuch der funktionellen Harmonielehre, Max Hesses, Berlin 1950.
2
Z.B. F. Haunschild, Die neue Harmonielehre, AMA, Brühl 1998.
3
Z.B. P. Daniel: Berechnung und kategoriale Beurteilung der Rauhigkeit und Unangenehmheit von
synthetischen und natürlichen Schallen, Universität Oldenburg, 1995.
4
M. Dachs: Harmonielehre, Kösel 1948.

8.5 Timing and rhythm
It’s not only (apparently) world-weary John Rowles who lamented "If I only had time" – this
is also the cri de coeur of every bad drummer. To stay in time is difficult, and even just to
define time is not an easy feat. Usually, one refers to Augustine (a monk, not a drummer); his
contemplations about time are widespread, and they are readily abbreviated to: “when nobody
asks me what time it is, I do know it; as I seek to explain it to an asker, I do not.” It is
something like that when it comes to rhythm, groove, and timing. Even modern books on
rhythmics♣ make do entirely without defining rhythm – and, yes, it is a challenge. If indeed
someone tries, it reads something like this: rhythm is the regular (periodic) repetition of
accents that are pooled together.
A pattern arises out of grouped accents – with an accent being a distinctive feature, i.e. for
example the beats on a bass drum. Already now we can think of examples where this does not
fit … in any case: if everyone can criticize this definition because he/she knows anyway what
rhythm is, then an extensive definition is indeed superfluous. At least it is in the present
context where the focus is on auditory perception, and not on teaching rhythmics. Fig. 8.29
gives a brief outline on the hierarchical processing of continuous time, its discretization into
basic beats, and the latter’s grouping and accentuation. Based on this ordering scheme is the
individual pattern that repeats within one bar in this example. In a two-bar pattern, two
different patterns would alternate (Bossa) – but, again, the emphasis here is on the hearing
system and not on the music.
Rhythm pattern
Binary subdivision
Emphasis: the meter
Grouping: the beat
Discretized time Ternary subdivision
Time as a continuum
Fig. 8.29: Beat, meter, rhythm (left), binary and ternary subdivision of the beat (right).
In Fig. 8.29, the dots mark the start of individual notes – and the layperson believes that a
musician with good “timing” needs to reproduce these starting points as precisely as possibly
in order to receive the “playing like a machine”-distinction. Checking whether this is actually
true shall be postponed to the next page; first, the focus is on the analysis: what is the
accuracy that the hearing system can muster to analyze fluctuations in rhythm? Practical
recording-studio experience provides the barely contested orientation value of 10 – 20 ms.
Timing errors of less than that quickly become meaningless. So, after all: the pro should hit
his/her notes with a precision of about 1/100th of a second, and hard- and software in music
needs to react very quickly in order not to make the ever-present signal-processing delays
subjectively noticeable. Also: it must not be overlooked that an effects device with a basic
delay of 7 ms is uncritical but 4 of them in series are not tolerable. For this reason some
processors offer two settings: little time-lag (low latency) for playing live, and more time-lag
(high latency) – but also much effect – for off-line processing.
♣
Marron E.: Die Rhythmik-Lehre. AMA 1991.

But now on to the actual topic of this chapter: depending on circumstances, a good
musician may need to do an objectively inaccurate performance so that it sounds correct
subjectively. The listener does the subjective assessment; the analyzer delivers the objective
data. A good musician will not generally play his notes at the time as written but with a slight
offset that is in part deliberate (determined) and in part unintentional (stochastic). Both (!)
offsets are desirable and as such generic. This is why the sentence in bold above is not good
as an excuse for the beginner – objectively wrong playing may indeed simply sound very
wrong. But then: what is would be "correctly wrong"? Not in the sense of an aggravation or
emphasis of "wrong", but rather: which deviation from the objective click of the metronome
leads to a rhythm subjectively perceived as "good"?
Let’s look at the determined (deliberate) deviations first. We have three main aspects here: the
tone generation (transient, onset, attack), the tone perception (the auditory event), and the
interpretation. Regarding the tone generation: 20 ms may have passed by the end of the tone
onset (attack phase) of a wind instrument – or as much as 100 ms for low-volume-notes. Of
course, it is not the moment when the player’s lips open that counts as the onset of the tone,
but a somewhat later point in time that can only be defined via the perception. Therefore the
wind-instrument player needs to the start blowing before the tone is supposed to sound. When
exactly the played note is considered to be existent – that is a decision made by the hearing
system i.e. it is an act of the tone perception. In his book on psychoacoustics [12], Fastl lists
eight examples for tones with different time-envelopes (TE), and determines the
corresponding subjective start of the note. Only if a note has an abrupt onset of tone and
immediately decays again (decaying TE) do objective and subjective starting points
practically coincide. Given an increasing TE, the subjective start of the tone is up to 60 ms
later than the objective one. For these special sounds, such numbers are of course dependent
on the special experiment. However, even if all notes have a steep attack, we find astonishing
differences in terms of seemingly equally long pauses: for the hearing system to assess a tone-
duration as equally long as a duration of a pause, the objective duration of the tone needs to
be considerably shorter relative to the objective duration of the pause! In Fastl’s example, first
an allegretto eighth-note and then an eighth-pause (both of 240 ms duration) are to be played.
In order for tone and pause to sound equally long subjectively, the tone must not be played for
240 ms but for a shorter 100 ms, while the pause needs to be lengthened to 380 ms! For
quarter-note and quarter-pause (nominal length 480 ms), the ratio is not quite as dramatic:
tone duration = 260 ms, pause duration = 700 ms. The explanation of these indeed
considerable discrepancies is found in the auditory transient processes (attack and decay) that
extend the subjective length of a note (relative to the objective length) and thus shorten the
subjective length of a pause. On top of these discrepancies (caused by the processing), the
individual interpretation of the musician also needs to be considered. For example, given
special stylistics, the “one” (1st quarter) will deliberately be played a bit early, or the “three”
might arrive a tad later that it nominally should. This is not for a lack of exercise but to
demonstrate individual virtuosity. If that weren’t the case, all those 1000s of “Elises”
celebrated on the pianoforte would have to sound identical.
It is in fact exactly this deviation from absolute rules that marks the virtuoso, the person “in
the know” who understands not only where to deviate from the strict formula, but also how
much. This knowledge often exists only implicitly i.e. without explicit awareness of it. If a
virtuoso is asked to play the same passages, we will recognize always (almost) the same
deviation. It is an expression of the personal style and not random at all. However, if we ask
at which points he or she has shifted the “one”, the artist will have difficulties supplying a
complete list, and if we inquire about the degree of shifts, the answer is likely to be: just as I
feel it, I don’t check the clock.

In their work on swing-rhythmic, Lindsay/Nordquist♣ investigate an example regarding

maintaining rhythm: "Fever" by Ray Charles (2004). The piece is dominated by a finger-
snapping that puts the emphasis on the elsewhere often less accentuated even-numbered
quarter notes (backbeat, more or less: iambus rather than trochee). These snaps are incredibly
precise, as shown in Fig. 8.30: here, the envelope over the time reference is depicted (894 ms,
corresponding to 67 bpm), and reference is hit with merely ±2 ms deviation. There is no
further info but we can only surmise that some kind of assistance system was involved – it is
hardly imaginable that a freely playing musician can after 3 minutes still be within 2 ms of the
original time. That “Fever” – despite this machine-like precision – still never gets boring is to
the credit of the continuously changing pattern played by other instruments. The bass
basically plays half-notes, but already the first 4 bars reveal some of the bass notes locking
into a ternary grid (splitting the fourths into thirds). The congas, as well, cannot do without
the ternary splitting in their “da-dub-da”. If merely accents on the four basic beats were
allowed, the whole charm of the piece would be gone; it would be life-less and without that
“swing”.
Fig. 8.30: R. Charles / N. Cole: Fever. Left: timing-analysis; right: four bars at the beginning of the piece.
"Hit the Road Jack" is another piece of Ray Charles’ that provides aid to answering the
question how precise the pro keeps time. In this version, Ray C. plays the first 1,5 minutes
without accompaniment, and presumably also without click-track. He cranks up the tempo
from 91 bpm to more than 95 bpm (beats per minute) – in a dance contest with rigid tempo-
specifications (often as little as ±2%) that would be quite borderline. But hey, this is Brother
Ray, and it ain’t no dance contest, either: that piece needs to be exactly how he plays it.
Fig. 8.31: Ray Charles 1981, Hit the Road Jack. Absolute (left) and differential-tempo (right) deviations. On the
left, the (absolute) deviations are referenced to the smoothed tempo-model (---).
If we subtract out the slow tempo changes (as presented on the right of the figure), short-term
fluctuations of a maximum of ±50 ms remain (there is some arbitrariness in that – of course
other deviations will result when choosing another bpm-curve). A maximum of 50 ms in a bar
of 2,5 seconds – that’s all right … especially considering that the larger deviations are a good
match to the form of the tune. The first entry of the choir is preceded by a minimal delay that
in no way sounds off, but rather expresses the individual interpretation. Two hands on the
piano and the voice generate a very lively rhythm that subjectively is perceived as correct –
irrespective of what any timepiece says.
♣
Lindsay K.A., Nordquist P.R.: A technical look at swing rhythm in music; http://www.tlafx.com/jasa06_1g.pdf

Fig. 8.32 indicates that "Hit the Road Jack" may also be interpreted in a different manner.
Same tune and same singer, but recorded 18 years earlier. Relative to the reference defined at
90,8 bpm, the tempo first minimally lags (-0,4 bpm), then catches up, and drops again towards
the end. The deviations rarely cross the 15-ms-mark (it needs to be considered here that in
contrast to Fig. 8.31 now quarter-notes are the reference). This is possible because the drums
play along from the beginning, and their accents are easier to more precisely measure
compared to the start of a piano chord. Neither video footage nor sound documentation
reveals whether a metronome was put into service during the recording. Thus hypothetically:
a metronome clicks along, the drummer realizes after about 18 s (towards the end of the
second chorus) a slight time lag, counteracts and achieves perfect time again at the end of the
2nd chorus. We find larger deviations in the middle of each chorus – which fits the structure
because each chorus consists of two halves: this would be a justification for that little swerve
in their middle. It’s a wrap – next take.
Abb. 8.32: Ray Charles 1963, Hit the Road Jack. Absolute (left) and differential (right) quarter-deviations.
Now on to the Eagles, the members of which are taken to be musicians that have a lot
experience in the studio. At the start, their "Heartache tonight" exhibits a pronounced
backbeat, i.e. an emphasis on the even-numbered quarter notes, generated by handclaps. In
this tune, the handclap is not always present, therefore the snare drum was analyzed: Fig. 8.33
shows only very small deviations in the quarter notes following each other at a distance of one
second, with the extremes correlating with the structure. Here it can be assumed that a click
track was used: both to achieve high precision but also because by now this is usual studio
practice (post-processing simply becomes that much easier). This assumption does not at all
seek to deny that Mr. Henley does precision work. Indeed, the presence of a click is no
guarantee for a precise rhythm.
Fig. 8.33: Eagles, Heartache Tonight. Absolute (left) and differential quarter-deviations.
“Heartache Tonight” was analyzed as a 4/4-beat: the quarter notes are emphasized, and the
counter-beats of the snare drum (offbeats) are on the even-numbered quarter notes. The
subdivision of the quarter notes is ternary – it’s a shuffle. Each 1st and 3rd quarter is given a
grace note that sounds not on the eighth note before but is (due to the shuffle-partitioning)
slightly delayed. In triplets notation, the second quarter would be divided in three parts; the
grace note would then be located on the last third (eighth triplet). The same would be done
ahead of the start of the bar.

Moving to the Rolling Stones, the "biggest Live-Band ever". Active since about 1962, and
still rocking the scene♣. Somehow, anyway. When they recorded the live version of their
"Honky Tonk Women" in 1969, they did have studio experience as well – but somehow in a
different way compared to the Eagles1. The piece starts at about 103 bpm, speeds up mightily,
and ends with about 120 bpm. Not that the guitar had accidentally started off too slowly – no,
it’s all supposed to be that way. Taking off with the heavy, earthy guitar riff, it ramps up and
reaches operating temperature with the first chorus. Towards the second verse, the tempo
eases off a bit, and then it’s pedal-to-the-metal to the conclusion. Is that wrong? No way – it
does groove. The deviations that, after all, are much larger compared to Fig. 8.33 do not stand
out much, because the live recording offers a charming “togetherness” in particular in the
chorus – it sounds like live recordings from that time simply were. Lively – but not wrong.
Fig. 8.34: Rolling Stones, Honky Tonk Women ("Get Yer Ya-Ya’s Out!"-LP, 1969). Bpm-tempo (upper left),
and absolute half-note deviations (upper right), referenced to the smoothed tempo model (---). The differential
half-note deviations (lower left) also relate to the tempo model. Lower right: rhythm pattern (intro).
The world famous intro-riff is a nice example for the ability of the listener to detect the basic
beat even if that is not even played at all. Only at the very beginning is the “one” emphasized
in the intro, from then on the power chord is tied to the “four-and” and across the measure
line to the following quarter note. The “three”, elsewhere the other accent in the standard 4/4-
beat, is not emphasized, either. Due to the accent on the “two” that immediately is interpreted
as offbeat, the beat is found without any effort. In this example, the quarter notes with their
time-distance of half a second are interpreted as basic beat; it is this rhythm that the listener’s
head synchronizes to as he/she “grooves along”. To nod with the head two times per second is
a very atypical movement. It would be possible to perceive the eighth note as basic beat but
the corresponding head movement would already be too fast. However, the eighth-note tempo
is a great match for drumming along with your fingers. You probably would not want to
shake your whole human body in this tempo - but then that’s a quite subjective decision.
Customarily, 120 bpm is the best “groove along” tempo, which is why it is found often in
dance music (moderato – allegretto). The dimensions and masses of the members of grown-up
people specify – in conjunction with the spring stiffnesses – the natural frequencies of this
“body”-system. And again we find: the tendency to oscillate is especially strong at resonance.
♣
Two musicians talk: „I’ve read that cockroaches presumably could survive an atomic war“ – „Maybe – but
Keith Richards would, in any case.“
1
(Translator’s note: about “live”: the Eagles’ live-version of “Heartache Tonight” is amazingly similar to the
studio version – the point where one might ask how live the live-recording actually was …)

Not everything was entirely in time with the Rolling Stones, as shown by an early recording
that (embarrassment-city!!) actually is called "Time is on my side". The tempo remains at a
rather constant 67,4 bpm during the first half of the song, and then suddenly drops to 65.4 at
the start of the guitar solo (Fig. 8.35). That would not be a tragedy because we can see a
motivation for this change. The short-term fluctuations that were analyzed using the beat of
the snare drum are not really problematic, either (graph on the upper right). What is really off
is the tambourine♣ that does get out of (time-) line again and again.
Fig. 8.35: Rolling Stones, Time is on my side, 6/8-beat; (Rolling Stones No. 2, 1964).
In the graph on the lower left, the time lag between tambourine and snare drum is depicted; it
often surpasses 50 ms, and several times 100 ms – and that is audibly wrong. If the (still)
tambourine is not hit with the (one) hand but the (other) hand is used to hit the (one) hand
with the tambourine, then two sounds are created: one when accelerating the tambourine, and
the other when it impacts/decelerates. Apparently, the latter playing mode was employed
because often two tambourine hits are heard quickly following each other. The first one
(shown in the graph on the left) is almost always early while the second one floats around the
reference time (shown in the graph on the right as a bold line). So: is time on my (or rather the
Stones’) side? Not really, that was "out of time" even as early as 1964. Occasionally the two
tambourine hits follow each other so quickly that they “melt together”, and at other instances
it will in fact have been only one single hit. Frequently, the time distance is more than 50 ms,
and that’s where (as an orientation value) the limit of the appearance of echoes lies. From a
time distance of about 50 ms, single echoes become audible; repetitions with a smaller delay
are pooled together by our hearing system into one single event. The haphazard change from
single- to double-impact playing and the strongly fluctuating timing in this example – it is not
perceived as stimulation anymore, but only and simply as timing errors.
A slightly delayed tambourine might be still acceptable because it is at least partially masked
by the snare drum (accessory masking). However, the early tambourine beat (in the range
above about 4 kHz mostly the only event) attracts much attention. This all the more because
its percussive character marks a point in time – contrary to the slightly open hi-hat, the sizzle
of which marks a range in time. Having the hi-hat ahead by a sixteenth (150 ms) – yes: that
might also have worked. But then it should have been consistently ahead, and not: now on the
“three”, then on the “four” (together with the snare), and then again on the “three-and-three-
quarters”. Charley would surely have known all that – someone else must have Jaggered this.
♣
A head-less tambourine is what is meant – the sound of the jingles is discussed here.

As a résumé, let us jot down a few numbers: the limit of perception regarding group-delay
distortion is around 2 ms; for your usual music performance, this value is of no significance.
From about 5 ms, timing errors may be noticeable in individual cases – but only from about
10 ms, the delay-range relevant for work in the studio starts. To avoid any misunderstanding:
we are talking tone/note-onset here, and not a time-shifted superposition! Phaser, flanger, and
similar effects devices (Chapter 10.8.3) do generate audible effects already at very short delay
times, but these effects are noticeable because of the variations in the spectral envelope, and
not as time-related in the first place. Given short repetition-periodicity, the perception
threshold for timing drift may be as small as 10 ms (depending on the circumstances), but for
rhythms that are not as fast it will extend to about 20 ms. And that’s it, then – even greater
delay may indeed become problematic and sound wrong. Still: it may – but doesn’t have to.
If the rules of timing could be summarized in a few lines, there would be great-sounding
drum-machines only. Exceptions start already in a simple ¾-beat: playing all three quarter-
notes exactly on the regular time will sound wrong. Playing the second quarter note just a tad
early creates a fit, making the listeners feel that all quarter notes have the same time-distance.
A corresponding shift of as little as about 1/6th of a quarter note will suffice, depending on the
tempo. However, even with this shift of the second quarter note, the rhythm will start to
become monotonous – for example if it’s always the same drum pattern that is repeated. A
drummer would indeed never repeat exactly the same beat; the drumstick will hit the skin at
different positions leading to similar beats that still differ in the details. He/she would also
introduce small variations: on top of the sound-color, he/she would vary the volume, and –
yes – the timing, as well, in order to optimally support the musical piece. All this is alien
territory for a drum-machine on its simplest programming level, and so it sounds just a
primitive apparatus. Which it in fact is.
The shuffle, that galloping groove onomatopoetically described with “dumm-da-dumm-da-

dumm”, is an example for large time-shifts. Here, every eighth-note (if a 4/4-beat can be
taken as a basis) is played later than the binary notation would require it. How much later –
that is left to the artistic interpretation, and it determines whether the musicians in the
ensemble play with or against each other. The musician ostentatiously playing a shuffle
written “evenly” (i.e. binarily) just as evenly will kill the whole number although he considers
him/herself as doing the right thing. Switching to ternary notation (triplets) helps only to some
degree because this does not express, either, how strong the “shuffeling” actually is. That is
decided by the “feeling”, the experience, and expression of the musicians, and by their ability
to empathize. That’s empathizing with the interpretation of the co-musicians, and with the
given piece of music. As a band “grows together” by frequent rehearsals, each musician
acquires experience about how the others interpret the music, and in the end everyone
“shuffles” so that they all match. That does not necessarily mean that they all play with
exactly the same deviations.
Table: tempi (bpm)

Largo 40 – 60 Larghetto 60 – 66 Adagio 66 – 76
Andante 76 – 108 Andantino 100 – 108 Moderato 108 – 120
Allegretto 120 – 132 Allegro 120 – 168
Presto 168 – 200 Prestissimo from 200
Table: meter
Iambus: –o–o–o–o Trochee: o–o–o–o–
Daktyl: o––o––o––o–– Anapaest: ––o––o––o––o

8.6 Loudness & timbre
Compared to an acoustic guitar, the electric guitar sounds different even if both play the same
note. Pitch and tone duration may be identical, but given “species-appropriate” play, sound
(timbre) and loudness will differ. This is due to the different way the partials evolve during
the respectively sounding note, i.e. due to the individual, instrument-typical attack- and
decay-behavior of the particular partials.
In simple models, each partial is assigned a frequency that is more or less an integer multiple
relative to the fundamental (dispersion, Chapter 1.3). Per partial an initial level is specified,
and also a decay time-constant that has the effect of an exponential decay of the amplitude
over time, or of a linear decay of the level. However, more exact analyses show that most
levels of the partials decrease according to more complicated functions; thus we have the
basis for developing more complex models that define every (primary) partial as a sum of
secondary partials. Other approaches are also possible – let’s remind ourselves here that the
quantity of mathematically equivalent models is in fact not limited.
As we reduce the sound pressure level (SPL) of all partials by the same dB-amount, the
volume drops; as we turn down the level of the higher partials by turning the treble control
counter-clockwise, the sound gets dull – that’s well-known. It is much more difficult to
answer the question how volume and sound depend on physical sound-parameters, and what
would be the characteristics of a good or a bad sound to begin with. The volume of a tone
(termed loudness in the following) is a monotonous function of the SPL-level (termed merely
level in the following). Since the level is dependent on the power, “more power = more
loudness” holds. Of course, we need to define this simple dependency in much more detail,
because otherwise there’s the danger that the result of the consideration would quickly read: a
100-W-amp is louder than a 50-W-amp … however this cannot be the general statement.
First, we need to distinguish between amplifier power and sound power (or acoustic power).
The amplifier power (that would strictly speaking have to be sub-classified into effective – or
active, or wattful – power, reactive – or wattless – power, and apparent power) is the power
that the amplifier delivers to the loudspeaker: e.g. 10 Watts (10 W). The largest part of this
power is converted into heat by the loudspeaker (Chapter 11); only about 1 – 10% are
converted into sound. For example, a highly efficient guitar speaker would convert 9 W of the
10 W electrical power fed to it into heat, and radiate 1 W as sound. In the immediate vicinity
of the speaker, this acoustic power is concentrated onto a small spherical surface and
generates a high intensity (the intensity is the power per area [3]). Assuming that the
loudspeaker generates a short sound impulse, this results in an imagined spherical wave that
propagates around the speaker and increases its radius (and thus its surface) with increasing
time. Since the surface grows with the square of the radius, the intensity drops with the square
of the distance. This is in the free, unperturbed sound field that we now focus on first.
Because the intensity is in a square-relationship with the sound pressure, the simple 1/r-law
(one-over-r-law) is applicable: doubling the distance to the loudspeaker reduces the sound
pressure by half, or as equivalent: the SPL drops by 6dB (more details in [3]). As an example:
an efficient guitar loudspeaker generates an SPL of 110 dB at 1 m distance given an input of
10 W amplifier power. At a distance of 2 m the SPL is therefore 104 dB, and at 10 m distance
it is 90 dB. If the objective is to generate not 90 dB at 10 m distance but 100 dB, the
amplifier power needs to be upped to 100 W, and for 110 db it would have to be 1000 W. So,
already here we notice the limits of this model that may remain linear only with regard to the
sound wave – for the loudspeaker, load-limits need to be respected, the efficiency is of course
power-dependent, and the speaker will die on us when overloaded.

8.6 Loudness and Sound-Color 8-55
In the open, given unperturbed sound propagation, the level decreases by 6 dB per doubling
of the distance. This fact is usually noticed with horror by the guitarist playing an open-air
concert for the first time: that amp that was way too loud every time back in the club now is
hopelessly drowned out all of a sudden. In the open, the reflections from the walls and the
ceiling are missing – they lead to the sound reaching the listener not just once but (as echo)
repeatedly. In the room, a superposition of free sound field and diffuse sound field is
generated, with the free sound field dominating close to the loudspeaker, and the diffuse
sound field dominating further away. The border between the two sound fields is represented
by the diffuse-field-distance (also called reverberation radius). It amounts to a few meters in
regular rooms (more precise information is found in [3]). Beyond the reverberation radius, the
SPL stays independent of the location within the room; or so says simple theory. For the
above example, this would imply: if the reverberation radius were 5 m, we would get (for 10
W input, and calculated starting from the speaker) at this distance a decrease in SPL down to
96 dB. In the remaining room (r > 5 m) the SPL would be 96 dB independent of the location.
Of course, additional factors such as beaming effects, the actual geometry of the room, and
the distribution of reflectors and absorbers would have to be considered – but this would go
beyond the scope intended here. This example is to show that – before we start thinking about
sound volume – sound source and room need to be looked into: which electrical power do we
have, what is the efficiency of the loudspeaker, into what kind of room does the speaker
radiate, and at last: where is the listener located? The SPL developing at the ear of the listener
is the result of all these parameters, and from it – not just from the power of the amplifier –
we can obtain indications for the generated loudness.
Psychoacoustics investigates the connection between SPL and loudness. Nowadays there is a
standard for that – which is not undisputed. How loud you perceive a sound to be is a highly
personal matter that is still interesting to science. And so we inquire with test persons
(subjects) about their impression of loudness, we have them give categorical assessments
(soft, loud, very loud), we make them perform magnitude estimates (double as loud as the
reference sound), and let them determine thresholds (now the sound becomes audible). It is to
be expected that not all human beings hear exactly the same thing, and neither that one and
the same person will give the exact same response when asked again. This insight, however,
will not be of much help – the psychoacoustician will want to know by how many dBs the
level needs to be increased in order to make the subject perceive double the loudness. It is
right here where the problems start: in fact, there is a multitude of experiments targeted to
find out exactly that – but unfortunately there is also a multitude of answers or resulting
models, not all of which generally correspond. Estimating the doubling or halving of loudness
is a frequently practiced experiment from which the whole scale from inaudible up to too loud
is assembled. Hellbrück [1993] has addressed this topic extensively and describes both the
pros and the cons of the standardized loudness model of Stevens/Zwicker: power law, or
exponential function? Stevens and his sidekicks had the subjects judge loudness relationships,
and therefrom derived the loudness power law – it teaches that loudness depends on SPL
according to a power law. In order to double the loudness of a 1-kHz-tone (in the level range
> 40 dB), the level needs to be increased by 10 dB according to this law. Accordingly,
upping the level by 20 dB corresponds to quadrupling the loudness, and +30 dB will match
eight-fold the loudness. Recalculating this in terms of amplifier power: to double the loudness
(and given linearity), the amplifier power needs to be increased by factor of 10 ten! Thus,
compared to a 10-W-amp, only a 100-W-amp will be double as loud, and not a 20-W-amp.
Still, a lot needs to be added here. To start with, the above law is applicable a priori only to a
1-kHz-tone. Then we find in Hellbrück’s book the lovely but unsettling citation: the
possibility should be considered that the whole of the sone-scale is a pure artifact from
psychometric methods that have been applied inappropriately and mindlessly.

Sone, that’s the unit for loudness. Mindlessly investigated? Let’s not go there – psychologists
and engineers will probably continue to bandy that ball for further decades. If we don’t want
to abort everything with the quite unsatisfactory insight that, due to the individual scatter,
establishing an exact functional correspondence will not be possible, then what remains is
forming statistical mean values. The difficulty is shown by an example from the beginnings
of calculating loudness: during some auditory experiments it was noticed that broadband
noise is much louder than a 1-kHz-tone although both have the same SPL value. Apparently,
the SPL-value is unsuitable as a measure for the perceived loudness, leading to this question:
by how many dB the two sounds will be different if both are adjusted to the same loudness?
For the experiment described in [12], a special noise is used, the so-called uniform exciting
noise (UEN) that may be imagined approximately as pink noise (kind of similar to a spoken
long, slightly dark “sh”). One possibility to estimate the loudness is to present the 1-kHz-
sinetone (e.g. at 80 dB) and ask the subject to adjust the level of the noise such that both
sounds (presented alternately, not concurrently) are perceived equally loud. The reverse
approach would also be possible; the noise is presented and the 1-kHz-tone is adjusted to the
same loudness. Surprisingly, different values result from the two approaches even if the
unavoidable small scatter is averaged out. There clearly is a systematic deviation (on top of
the stochastic one): the adjustable magnitude is adjusted too high. For a presented 79-dB-
noise, an adjustable tone is set to 90 dB, but for a presented 90-dB-tone, the noise is adjusted
to 78 dB to be equally loud.
The measurements shown in Fig. 8.36 give three results:

- For the two sounds to be subjectively of the same loudness, the level of the 1-kHz-tone
needs to be in part more than 20 dB above the level of the noise.
- The results are dependent on the measurement procedures.
- The scatter is considerable.
In Fig. 8.36, the scatter is indicated as interquartile ranges; these represent 50% of the
measurement values, with the values “above” and “below” discarded. As an example: 50 % of
the subjects (the “middle” half) adjust the level of a 1-kHz-tone to an SPL of 83 … 97 dB for
equal loudness with a 70-dB-noise, 25% of the subjects set the level to smaller than 83 dB,
and the remaining 25% adjust the level to more than 97 dB. Additionally, the median value is
given as a dot. We can unequivocally take from this experiment that noise is perceived louder
that a 1-kHz-tone of the same level; however, the quantitative evaluation is subject to
considerable scatter, and the latter moreover is dependent on the adjustment method.
Psychoacoustics factors this in by defining two different loudnesses: a standard loudness
level, and an object loudness level (that of the test sound). N.B.: the loudness comparison
with the 1-kHz-tone historically was the first method to determine the loudness of any sounds,
i.e. objects, via using a standard, i.e. the 1-kHz-tone).
Fig. 8.36: Loudness comparison noise/sinetone (left),

loudness with/without background noise (above),
[12]. “GAR” = UEN, “Terz” = third octave band.

Keeping constant the level of the standard and making variable the level of the object (i.e. in
this case the noise-level) yields the object loudness. Conversely, making variable the level of
the 1-kHz-tone we obtain the standard loudness. The value interpolated between the two
curves (the grey line in the figure) is called interpolated loudness level in older literature.
The tem loudness level was introduced in order not to always have to talk about the “level of
the equally-loud 1-kHz-tone” – rather, the loudness level with the unit phon is specified; the
numeric value is that of the level of the 1-kHz-tone of the same loudness. Thus, if noise is
perceived as equally loud compared to a 90-dB-tone (of 1 kHz), then this noise has a loudness
level of 90 phon. This now makes the loudness sensation quantifiable – with numeric values
that are difficult to interpret, though: 80 phon are not double as loud as 40 phon but 16 times
that loudness. For this reason, additionally the loudness (measured in sone) was introduced.
A 1-kHz-tone of 40 dB level serves as a reference point delivering the loudness of N = 1 sone.
Since a level increase of 10 dB has the effect of doubling the loudness, 2 sone match 50 dB, 4
sone match 60 dB, 8 sone match 70 dB, and so on. Below the level of 40 dB, this
correspondence is not valid anymore: in this range already smaller level changes have the
effect to doubling the loudness.
The upper line in the right-hand section of Fig. 8.36 shows the relation between the level of
the 1-kHz-tone (abscissa) and loudness (ordinate). Again: this is only for 1-kHz-tones – other
spectral compositions necessitate other curves. Another prerequisite is that the 1-kHz-tone is
presented by itself i.e. without other sounds being present. If the latter are presented
concurrently, the loudness of the 1-kHz-tone may be partially masked i.e. reduced. The
lower line in the right hand graph of Fig. 8.36 shows such a scenario: besides the 1-kHz-tone,
a pink noise with the third-octave level of 60 dB is presented at the same time. If the 1-kHz
tone has a high level (e.g. 90 dB), the two curves barely differ – the noise has little influence
on the loudness of the tone. However, as the level of the tone is reduced (e.g. to levels below
57 dB), the tone becomes altogether inaudible because it is “masked” by the noise. Thus,
when there is a masking sound present, the loudness grows more strongly with the level
compared to the situation without masking noise.
For the practical musical performance situations we can learn from these relations that small
variations in the sound power (e.g. +10%) are insignificant for the loudness perception. If the
power of an amplifier is increased from 40 W to 44 W (and given a proportional change in
sound power), we will – as a rule – not perceive a change in loudness. According to common
practice, the just noticeable difference for amplifier power is estimated at about +50%. The
difference between a 40-W-amp and a 60-W-amp is just about noticed – while doubling the
power is clearly perceivable. Any musician deliberating whether to buy a 50-W-amp, or “for
good measure” rather a 60-W-amp should be particularly weary of the efficiency of the
loudspeaker. That is because, for example, a Celestion G-12-M is rated in the datasheet at 100
dB/1m while the G-12-M Greenback is rated at 97 dB/1m. Purely in terms of figures, the
greenback requires double the power in order to generate the same SPL as the G-12-H. How
these datasheets were established, is of course an entirely different story, and that (besides the
loudness) the color of the sound (the timbre) plays a pivotal role – well, that opens yet another
can of worms. It would go too far here to elaborate on all parameters that weigh in when
determining loudness and timbre; those interested are recommended to read up in Fastl’s book
"Psychoacoustics" [12] – on 462 pages, it represents a comprehensive overview of the most
important basics and models. The literature list in the appendix gives further info on related
books.

The color of sound (timbre, sound- or tone-color) is the last sound parameter that we visit
here. For many readers, it will be the most important one – but unfortunately it is also the
most complex one. The sound-color – “the sound” – is being evaluated according to highly
individual criteria, and trying to establish a model to calculate it always leads to failure. Of
course, the sound-color depends on the sound spectrum, but already the metrological
determination of the latter will be unsuccessful unless very simple sounds are analyzed.
Harmonically complex tones are one thing, but a guitar solo played against a full
accompaniment is another. Seeking to attribute roughness or fluctuation strength (based on
modulation-indices and -frequencies) to a sound is futile because this cannot be determined in
the guitar solo. Every spectral analysis may optionally be interpreted as a spectral weighing
with the complex transmission function of a bank of band filters, or a convolution in time
with the impulse responses of these filters. Bandwidth and impulse response cannot both be
limited to a rectangular range, though, and thus every spectral analysis will lead to spectral
and time-related leakage. The term spectral leakage intends to express that even the spectrum
of a sine-tone is not measured discretely at on point of the frequency scale but as a
continuously distributed spectral density. A Fourier series expansion is only possible in
special cases (e.g. when the signal period is known), but this is meaningless in practice.
Because the spectrum of the pure tone is presented in a broadened (‘smeared’) fashion, it is
difficult to separate closely adjacent notes. Since spectral and time-related blur are reciprocal
to each other, it would be possible to extend the duration of the analysis and thus to decrease
the spectral leakage – but then the time-related leakage (describing the broadening –
‘smearing’ – along the time axis) increases. In concrete terms: if 1 Hz separation is desired in
the frequency domain, the blur in the time domain is 1 s. The exact relation between the two
quantities does not need to be deduced here♣, for orientation Δt ⋅ Δf = 1 suffices. If the
analysis-blur along the time-axis is to be reduced to 10 ms, the spectral blur increases to 100
Hz. If we seek to, for example, extract from a musical piece the partials of the lead guitar, and
therefore subject the wav-file to a DFT-analysis, it will be very difficult to decide which of
the lines belong to the guitar, and which should be traced to other instruments. It may be
possible in some cases, but fail in others.
Particular significance needs to be assigned to the “attack” (the onset of the tone). Many
instruments can correctly be identified only via the structure of their attack; suppressing the
first 100 ms tampers greatly with the sound. A good time- and frequency-resolution is
desirable in this time range if the structure of the partials is to be meaningfully detected. The
spectral and time-related leakage effects cannot be seen as errors per se; rather, they are kind
of analysis-immanent artifacts. A Blackman-Harris-window is not more wrong or more right
than a Kaiser-Bessel-window – it is just different. That, however, also means that one window
modifies the structure of the partials differently compared to another window. If guitar tones
were composed of harmonic partials of infinite duration, the analysis would be relatively
simple. But they’re not: the frequency relations of the partials are not integer multiples but
they are spread out, and in addition they are slightly shifted (due to the frequency-dependent
bearing impedances, Chapter 2). The amplitudes of the partials are not constant over time, and
they do not decay according to simple functions, either. Moreover, the almost always present
other instruments weigh in, as well, because pure solo-playing of any length of time does not
occur much. Spectral analyses can certainly help to establish orienting impressions: are only
odd-numbered partials dominant, how strong is the fundamental, do strong partials already
stop at 1 kHz or do they extend up to 5 kHz? However, already with the evolution with time,
with the fluctuations of the partials, it does get complicated, and the results of the analyses
become dependent on the parameters of the analysis filters to a large extent.
♣
See e.g.: Zollner, M., Frequenzanalyse, Hochschule Regensburg, 2009;
or: Zollner M., Signalverarbeitung, Hochschule Regensburg, 2009.

Starting not with the spectral analysis of a whole ensemble, but recording and analyzing the
sound of a single instrument played in the anechoic chamber, will usually result in spectra like
those depicted in Fig. 8.37. They give the insight that e.g. a clarinet generates predominantly
odd-numbered partials – this even being in good agreement with the wave mechanics of this
aero-acoustic resonator (open on one side, i.e. “gedackt” pipe). The graphs on the left and in
the middle stem from different books – both are supposed to show the spectrum of a clarinet.
The graph on the right shows the spectrum of a cello. That the two clarinet spectra differ so
much is not necessarily the result of grave measurement errors but easily due to the variability
of this sound. Indeed, there is not “the” tone of a clarinet, and just as little is there “the”
spectrum of a clarinet. We may be able to recognize characteristic differences in the cello-
spectrum in Fig. 8.37 compared to the clarinet-spectrum, but these become meaningless in
view of the spectral differences between the clarinets. Conclusion: single spectra hold little
validity.
Fig. 8.37: Sound-spectra of some instruments: clarinet, clarinet, cello.
Regarding the sound of the violin, Dickreiter♣ elucidates: the build of the partials of the violin
is relatively irregular, i.e. it changes from note to note. The reason is found in the
complicated resonance properties of the resonance body that strongly influence the material
characteristics and the construction. Thus, the spectrum of a d1 may look entirely different
from that of a g1, and of course the relative position of violin and microphone plays a role
since the radiation happens with a frequency dependent directionality. The first “electronic
organs” sought to imitate the sound of specific instruments by generating periodic tones with
a spectrum that had a supposedly instrument-typical envelope – such as e.g. the cello-
spectrum from Fig. 8.37. It was more or less accepted that the resulting sound was only very
remotely reminiscent of a cello; to sound “kind-of-electronic” was probably o.k. The main
criticism was: the sound of simple organs is too “sterile”; it does not live – the instrument-
typical beats are missing. The latter then were subsequently included via amplitude- and
frequency-modulators (vibrato, tremolo), but again the result sounded artificial again because
the effect was not relative to the partials but global. Only with the emergence of the sampling
keyboards and the availability of huge solid-state memories could instrument sounds with an
acceptable degree of naturalness be synthesized.
It’s not that the spectral representation would be entirely unsuitable to visualize instrument-
typical characteristics – spectra can fully describe signals. It’s just that the information
included in a single spectrum is too limited to already extract the instrument-typical from it.
Typical is e.g. the presence of accompanying sounds that nevertheless contribute to the
recognition of an instrument. The hammer-noise of the piano (“plock”), the blowing-noise of
the flute, the scraping noise of the violin bow, the “squeak” in the horn attack, the impact of
the strings of a bass on the fingerboard (drastically emphasized in the slap-bass style) – these
are examples for such additional sounds, and there are many more. Typical are spectral
maxima (formants) that are at a fixed frequency, or move along dependent on the
fundamental; typical are time-related fluctuations of partials.
♣
Dickreiter M.: Handbuch der Tonstudiotechnik, Saur 1979.

All these characteristics aid the hearing system to categorize sound-colors, and to eventually
allocate them to specific instruments. This then is done on the basis of learned knowledge –
those who never have consciously heard an oboe will not recognize it, and only hear a strange
nasal tone. Even those who in fact know how an oboe sounds will find recognizing the
instrument difficult if one period is cut out from the oboe-tone and periodically repeated
(looped). An oboe-typical spectrum is created – but it’s out of typical context. In the auditory
signal analysis (i.e. when we listen) the arriving sounds are automatically compared with
known patterns stored in our memory. If the presently heard sound and the memorized one
more or less match, the decision is made: sounds like an oboe, and/or like a musical
instrument, and/or nasal, and/or dangerous, or whatever else could be found in the match. We
can imagine the sound-color identification as a multi-stage process: in a first hierarchical
stage, the inner ear determines the time-variant spectrum of the non-masked partials, i.e. the
momentary sound-color – customarily described by one single spectrum. However, since (as
taught be signal theory) a spectrum cannot be ascertained for a point in time but only for a
time-range, the term momentary must not be taken too narrow a view on. The speech analytic
evaluates sections of about 10 – 30 ms length, and it indeed is a powerful tool; as it is applied,
it is often underlined that for the evolution of the hearing system, analyzing speech was even
more important than analyzing music. That does sound convincing – but it does not mean that
each and every musical analysis has to comply. For percussive sound, shorter durations of
analysis may be purposeful, and for very low bass-notes longer ones, as well (because it
allows for a finer frequency resolution). Still, an analysis-duration of 20 ms is quite workable
as an orientation value; this means 50 spectra per second. These of course are not all identical
but time-variant. On the basis of this spectral ensemble, the next-higher sound-color
determination can happen which already yields more than just a “sound kinda like aaa”. It
could e.g. yield “sounds like a trumpet”. In order for this already rather complex analysis to
be successful, typical patterns about tone-onset, fluctuations, duration and decay need to be
memorized. If the deviations are too big, the recognition algorithm fails. Cutting off the first
100 ms of a note will substantially lower the recognition rate; apparently already this short
section includes important instrument-specific information that is not available in the later
parts of the evolution of the note. Alternatively (and this is something we must not overlook),
the cut sound will not be matched to the correct instrument because nothing about it has been
learned yet (i.e. no corresponding patterns have been memorized).
In the processing stage still higher up, the evaluation steps can start that lead to the verdict:
“sounds like Josh Redman”, or “That be Hendrix on the Strat”. Such judgments are, however,
not part of the present reflections … so let us return to the color of sound, the timbre, and its
signal-theoretical basis. We have already known for some time what the color of sound is
NOT, and from this the following exclusion-definition originated: color of sound is that which
remains if loudness and pitch are abstracted from. Alternatively, according to an old
Acoustical-Society-of-America-definition: color of sound is the perception attribute that still
distinguishes two sounds although loudness and pitch are equal. Somehow that feels like a
trash-can-esque definition into which we can throw everything that cannot be defined
precisely. Borrowing from optics helps to move along a bit: like we can objectively define
visually perceived colors on the basis of spectral intensity distributions, the color of sound in
auditory perception can be ascribed to the envelope of the sound spectrum. Like a picture
consists of strung-together locally distributed color spots, the tone of an instrument consists of
momentary timbres strung-together sequentially in time. We need to allow for the fact that
this comparison will arrive rather quickly at its maximum load and hit a wall – the two
sensory channels do, after all, exhibit strong differences besides some similarities.

In order to explain the possibilities and limits of the spectrum-based analysis of tone color, a
dyad shall serve: two added-up sine-tones (300 Hz, 312 Hz) of equal level that are abruptly
switched on at t = 100 ms (Fig. 8.39). The time-function would therefore be:
Beating
Already this simple example exemplifies that there is more than one possibility of
representation for every signal: the dyad may either be seen as the sum of two tones, or as the
product of two other (!) tones. Instead of adding a 300-Hz-tone and a 312-Hz-tone, it is also
possible to multiply a 306-Hz-tone by a 6-Hz-tone. A spectral analysis merely and always
disassembles the signal into its additive components, and not into its multiplicative
components, showing one 300-Hz-line and one 312-Hz-line in the spectrum. The 6-Hz-
envelope that is so nicely revealed in the time function (Fig. 8.39, upper left) remains hidden
in the spectral analysis. Even the 300/312-Hz-pair-of-lines will only be represented as two
separate lines for suitable analysis parameters – and since there is an infinite number of
parameter-variants, there will be an infinite number of spectra.
The long-term spectrum identified for -∞ < t < ∞ is pointless; rather, the spectrogram
obtained by shifting a short window-section is required (Fig. 8.38). In the left-hand graph, a
rectangular evaluation-window is shown; it is slid across the signal as a multiplicative
weighing (over time). From the signal weighed this way (shown at b), the DFT-spectrum is
calculated as a function of the time-shift. Since undesirable jumps occur at the window-
borders for this type of window, the rectangular window is not applied in practice; windows
with a rounded-off shape are customary.
Fig. 8.38: Time-function of a sine-tone; with two different weighing windows.
However, a fundamental problem still remains with the window-weighing as shown on the
right (here: Kaiser-Bessel window): the spectrum is determined based on the windowed (i.e.
modified) signal. Fig. 8.39 shows – for the two-tone signal mentioned above – spectrograms
derived with different windows. The signal was identical for each spectrum; the differences
stem exclusively from the different analysis-parameters. The window-length is specified by
the point-number N, a frame-length of 46 ms belongs to N = 2048. The time specified as
abscissa in the color-spectrum marks the beginning of the window. Since the width of the
latter is not 0 but e.g. 46 ms, we understand why the analysis pushes the start of the dyad
ahead e.g. to the 54-ms-point – although both sine-tones are switched on only at the 100-ms-
point! At exactly this time shift, the start of the signal falls into the rectangular window, and
therefore the corresponding spectrum also starts from 54 ms. Increasing the number of points
to 4096, the window-length grows to 92 ms, and the spectrogram (linked to the rectangular
window) starts at 8 ms.

Rectangular window, N = 2048 Hanning-window, N = 2048
Flat-Top-window, N = 2048 Kaiser-Bessel-window, N = 2048
Kaiser-Bessel-window, N = 4096 Kaiser-Bessel-window, N = 1024
Fig. 8.39a: DFT-Spectrograms of an abruptly switched-on beating (300 Hz / 312 Hz), ΔL = 90 dB.

Rectangular window, N = 2048 Hanning-window, N = 2048
Flat-Top-window, N = 2048 Kaiser-Bessel-window, N = 2048
Kaiser-Bessel-window, N = 8192 Kaiser-Bessel-window, N = 8192, time-domain
Fig. 8.39b: DFT-level graphs (at 306 Hz) of an abruptly switched-on beating (300 Hz / 312 Hz).

It would be possible to scale the time axis such that t = 0 specifies the end of the window; in
that case the corresponding shifts would show up at the end of the signal. Just to be clear: it
again needs to be emphasized that this is not a software error of the analysis program, but a
system-immanent artifact of all spectral analyses. Depending on the window-length (= on the
impulse response of the filter), the analyzed signal becomes longer. Moreover, changes result
in the direction of the ordinate, as well: the switching-on click as vertical streak, and the
spectral leakage as vertical broadening of the spectral lines. In fact, from 100 ms there should
be two lines running in parallel towards the right, as shown in the top-right graph; instead one
single streak is shown. The simple reason: for N = 2048, the analysis bandwidth is too small,
and the two lines cannot be represented separately. If we take the bandwidth as the reciprocal
of the window-width, we obtain the bandwidth of Δf = 22 Hz – that is too broad for a line
distance of only 12 Hz. For the Kaiser-Bessel-window (Fig. 8.40) used in the following, we
moreover need to consider that the effective duration is only about 1/4th of the frame length;
and that the effective bandwidth therefore will be about four times that of the rectangular
window♣.
Fig. 8.40: Time-function

(left) and spectral function of
the Kaiser-Bessel-window,
N = 2048,
sampling frequency:
fa = 44.1 kHz.
If indeed time-related and spectral leakage have effects in every spectral analysis, it stands to
reason to ask whether the like would not appear also within the hearing process – after all, the
signal is broken up into its spectral components there, as well. And sure, leakage will of
course be present, too. However, because the auditory filters are adaptive and non-linear, we
cannot specify one bandwidth and one attack time – things are more complicated. Too
complicated for the present explanations that are merely intended as an overview, and
therefore reference is made to specialist literature, e.g. Fastl’s "Psychoakustik" [12, available
also in English language]. The hearing system processes two tones of large frequency distance
in separate channels, while tones close in frequency are jointly processed. The two-tone signal
mentioned above cannot be separated into its two components by the auditory system, and
one tone of quickly fluctuating loudness is heard – i.e. as product, not as sum. We hear
something that does not actually exist in the spectrum: a 306-Hz-tone! Already this simple
example proves how difficult it can be to extrapolate from a spectrum to the auditory
perception. It is not entirely impossible; the parameters of the analysis can be adapted, after
all. Therefore Fig. 8.39 includes different analyses, with varying window-types and -lengths.
All show the switching-click, to start with. The longer the window, the longer the switching
click. It has to be that way: if, during the shifting of the window, the signal-start just about
falls into the window, it is only an impulse of very short duration that is analyzed – the
spectrum of which is necessarily broad-band. The more the window is shifted beyond the
signal-start, the longer the signal to be analyzed (windowed), and the more narrow-band the
spectrum. Is the switching click audible? No! In any case not as the figures would let us
assume. It therefore is purposeful not to show the color-spectrum with a dynamic of 90 dB (as
is the case in Fig. 8,39) but with only 40 dB: visual and auditory impressions are a better
match that way.
♣
We will not investigate in detail here what is to be understood by the term „effective“.
More details may be obtained from: M. Zollner, Signalverarbeitung, Hochschule Regensburg, 2009,
as well as from: M. Zollner, Frequenzanalyse, Hochschule Regensburg, 2009.

We now take a look at the fast fluctuations that can be clearly seen in the time-function. They
also appear in a time-section of the spectrum, in the so-called slice (level over time with fixed
frequency, Fig. 8,39b). Forming the logarithm of the envelope yields the curve shown in the
graph at the upper right, and the evaluation of the DT-analysis yields the graphs below. Again
it is clear that the time-related leakage has the effect of very differently shaped level curves –
depending on the window-type and -duration. Thus we retain: the DFT-analysis delivers a
multitude of different spectra that – to begin with – allow for only few conclusions regarding
the perception of the sound. Supplementary algorithms enable modeling of hearing-typical
assessments (auditory critical-band filters, contouring algorithms, spectral and time-related
masking), but the scientific investigations have yet to arrive at a true breakthrough.
The two-tone signal analyzed in Fig. 8.39 already revealed the fundamental issues found in
any spectral analysis. Yet, it is a very simple signal – instrument tones are of considerably
more complex build, not to mention chords or tutti-sections. Compared to the latter, the triad
analyzed in Fig. 8.41 is still rather simple: three added-up sine-tones of equal level but
switched on at different times. The 300-Hz-tone and the 312-Hz-tone are switched on at t =
100 ms, and the 400-Hz-tone comes in at t = 134 ms. Analysis is again done using the Kaiser-
Bessel-window, the level dynamic in the figure is, however, reduced from 90 dB to 50 dB
(compared to Fig. 8.39).
Fig. 8.41: DFT-spectrograms of a triad (300 Hz / 312 Hz / 400 Hz). ΔL = 50 dB.
The tone onset blurs as the window length increases, but the spectral separation improves in
turn. The latter does not need be all that great, though – because with this triad, again only
one single tone is heard. Not a sine-tone but rather a lively bubbling tone-mixture – with one
single pitch. Only when listening repeatedly, one could also tend to hear an oscillation
between two pitches … but certainly not anything like what the analysis done with N = 8192
would suggest.

As powerful PCs became available, the desire developed for sound analyses to – at last –
depict “the correct” spectrum, meaning not just 512 lines but 4096, or even 16384, for good
measure. The latter number implies, however, that the sampling window (at a sampling
frequency of 44.1 kHz) has a length of 372 ms, which is too long compared to the hearing
system even when applying the (shorter) effective length. For sound analysis, N = 4096
represents a tried and tested compromise that offers the basis for supplementary DFT-analyses
and post-processing. The latter is urgently required: the 2k-analysis shown in Fig. 8.41 gives
the impression of two sound-parts starting at different times. Objectively seen, this is indeed
correct: a beat from 100 ms, and a sine-tone from 134 ms. Our hearing system, however, does
not care: it perceives one single tone-onset and not two. Even when the two partial sounds
start with a delay of 70 ms between them, they are not heard separately in time. The simple
reason is that the beating in the dyad impedes the recognition of the time-structure. Only from
an offset of about 100 ms, the additional tone coming in with the delay in this example (!) is
recognized as such (compare to Chapter 8.5, though).
Not to stick exclusively to synthetic tones, let us now turn to a real guitar tone: Clapton’s
intro to “Stepping Out”. The guitar plays by itself a number of times – this facilitates the
spectral analysis a lot. Fig. 8.42 shows spectra and time-functions: in the upper two lines of
graphs those of a G3, and below for a C4. That’s four times that “same” G3, but with
considerable differences! Clapton’s sound may not be described with one single spectrum,
after all – and that is the same for J.H., R.B., G.M. and all the other big names: virtuosity
implies change, and that holds for the spectra, as well.
Still, we of course can wring a few commonalities from the G3-spectra: they all feature a gap
between 1 and 1.5 kHz, and a spectral maximum between 1.5 und 2 kHz. This is the range
where the (second) formants of the vowels “ø” and “y” (using the definitions of the
international phonetic alphabet, IPA) reside, so these tones can be attested an ø- and y-like
timbre. Moreover, the strength of the low partials is notable: there are neither exclusively
even-numbered, nor exclusively odd-numbered partials. And finally: the brilliance of a single-
coil-guitar (which would feature a resonance of 3 – 4 kHz) is not achieved; rather we have a
strong, mid-range-y, trumpet-y sound … or a saxophone sound, or a cello-sound with flute-
like harmonics? Journal-literature – (rightfully) praising this phase of Clapton’s as pure
genius – has found, and still finds, many comparisons. It seems strange that to describe a
guitar sound, one would have to borrow from the realm of wind instruments, or strings – but
maybe in the far distant future, a trumpet instructor will shout at his pupil: blow with more
emphasis on the mids; more like Clapton’s guitar sound!
Irrespective of whether trumpet- or cello-like, what does determine that sound and its variance
that appears even for the same notes? First, let’s look at the second part of that question which
is easier to answer: even when fretting the same string at the same fret, the sound depends on
the location of picking, and on the movement of the plectrum. And on the plectrum itself –
although that was certainly not swapped during one take of the recording. The angle of the
plectrum (parallel or slanted relative to the string), the basic movement (up- or down-stroke),
the angle of the movement (relative to the fretboard), place of picking (closer to or further
from the bridge – these are all sound-determining parameters. Then there is how the left hand
is at work: even slight bends can make partials vanish into interference-gaps. That is why the
four analyzed G3’s are not identical, and that is why there is no “one” G3-spectrum, and not
“the” C4-spectrum, either, and least of all “the” Clapton-spectrum. Not to forget: guitar, cable,
amp, room, and recording technique of course also influence the sound – but these would be
time-invariant per recording … presumably, EC will not have jumped back and forth between
amp and mike. But then, come to think of … one could surmise that some musicians ……

Fig. 8.42a: Individual spectra for the spectrograms in Fig. 8.43a. Kaiser-Bessel-window, N = 2048.
Fig. 8.42b: Time-functions for the spectra in Fig. 8.41a. Time grid = period of fundamental.
But where exactly do we now have the analytical proof for Clapton’s “unique” or at least
“groundbreaking” Bluesbreaker-sound; what is so special about these notes and their spectra?
Ultimately: nothing at all! Listening to them in isolation, cut out of the intro, they sound
plenty unspectacular. Maybe like a trumpet, or like a cello, or even synthetic. It gets
interesting only as a group of notes is sounded, as soon as every note is presented with its
attack- and decay-processes only existing in full context. However, it is exactly those
processes that elude any spectral analysis that could reasonably be interpreted.

Somewhat easy to detect is the “lingering” of individual (in fact already terminated) notes
which is due to the strong amplification. The color-spectrogram in Fig. 8.43a shows this,
and it aurally creates the impression of a mighty, fat, powerful sound that can be reigned in
only with difficulty. However, the short attack-noises limited to 20 – 40 ms duration yield all
those problems that have already been described in the context of the dyad- and triad-signals.
Of course it is possible to calculate corresponding spectra, but they will be highly parameter-
dependent. Dear PhD-students who are just now in the process trying to cut another facet into
the diamond that is psychoacoustics: don’t let yourselves be discouraged by that! EC needed
21 years to produce these sounds – you don’t have to have them analyzed within 2 days. Sure,
it is not impossible, but simply applying a bank of Gammatone filters with contouring-
algorithm – that ain’t enough. Here’s a hot tip: do synthesize the sound using the supposed
partials, and listen. This approach very quickly reveals, which formation-rules are verifiable
but not relevant to the auditory system, and what might constitute a “groundbreaking” sound.
And speaking to the gear-heads: you won’t do anything wrong bringing out that original ’58
(or was it a ’59, after all?), but absolutely necessary it is not. Required are the right fingers,
the right micro-timing, the right bends. “Clapton is God” was the writing, not “Paula is
goddess”. This is illustrated by many EC-epigones appearing on Youtube, covering Stepping
Out with at times remarkable equipment (but at times showing dismal timing, too). It becomes
quite clear that the finger-work is much more essential than the question of “R8 or R9?”.
It is time to come back to the starting point of this chapter: to the timbre (or tone color). The
latter may without doubt be determined on the basis of a spectrogram – but in infinite
variations, because there are infinite possibilities to parametrization of spectrograms. If we do
not want to test all of then, then an overlapping 4k-DFT with Kaiser-Bessel-window for the
steady-state part of the guitar-tone will deliver some first orientation values. The onset of tone
(attack) is more difficult to analyze because here the spectrum can change as much as 20 dB
within 10 ms – a typical case of conflict between time-domain-resolution and frequency-
domain-resolution. If several instruments sound at the same time, the analysis becomes
particularly difficult. For the graph in Fig. 8.43a, only a single guitar plays, and the behavior
of individual partials can clearly be observed. This behavior is, however, difficult to measure
since these partials rarely maintain their frequency, not even in the seemingly steady-state part
of a note. We find subtle up-bends (at around 1000 ms), down-bends (also called pre-bends,
around 1900 ms), and half-step bends (around 1600 ms). Thus, it is not sufficient to set the
cursor on the 180th DFT-line and to analyze how its level evolves. This would again be
merely the behavior of the level of this DFT-line but not that of a special partial – the
frequency of the latter is changing, after all (e.g. from the 180th line to the 191st DFT-line).
Contouring- and pitch-follower-algorithms (which one indeed is that “closest neighbor”?) are
applied to assist in this scenario, which is another reason for the multitude of parameters.
Once these problems have been solved (it is, after all, not impossible to track partials), new
challenges present themselves: the partials not only change their frequency but also their
level! And not too little or too slowly, at that: we see e.g. 6 dB / 10 ms. Mind you, the attack-
and decay processes of the DFT-analysis may run with the same speeds. Thus, if we change
the DFT-parameters, the level fluctuations also may change. This multi-variant analysis (or
optimization) would go far beyond the scope intended here, and so what can remain is merely
the qualitative statement: the partials change their amplitude and frequency even within one
single played note. At least the frequency shift is a global one (all partials change their
frequency by the same percentage), but the amplitude shifts are partial-specific. Not all
frequency- and amplitude changes are audible; there are absolute thresholds, masked
thresholds depending on neighboring tones, and pre- and post-masking in time. Only that
which is above threshold is fed to the final post-processor that then forms – among other
things – the timbre, the tone color.

Fig. 8.43a: Excerpt from Stepping Out (Mayall / Clapton), guitar-intro. ΔL = 40dB.
Fig. 8.43b: Excerpt from Stepping Out (Mayall / Clapton), guitar note with finger vibrato (7 Hz). ΔL = 40dB.
Fig. 8.43c: Excerpt from Stepping Out (Mayall / Clapton), fast eight-note triplets. ΔL = 40dB.

All in all: not a trivial analysis. This is not supposed to sound too discouragingly, therefore
let’s quickly look at Fig. 8.43b. In this graph, the visual analysis is facilitated by a gestalt-law
that helps in auditory tone-recognition, as well: the low of common fate (see also Chapter
8.2.4). All lines that move back and forth in synchrony are partials of a guitar note, in
between the horns provide (vertical) accents, and the electric bass lays the foundation below.
It may be added for the Strat-purists: no, you do not need a whammy bar for that; this is done
with a left-hand finger. To bend a note by ± ¼-step with a modulation frequency of 7 Hz –
that is Clapton at his best. In Fig. 8.43c, things get more hairy again. This is one of the
passages with faster playing, and vibrato is not really possible with note-durations of as small
as 100 ms. In this section, already the pitch-tracking is a true challenge, not to mention an
automatic timbre-analysis.
(Translator’s note: the following paragraph only makes sense and works for German speech sounds and words.
It was impossible to find suitable correspondences in English without a complete re-write/re-draw. I have tried
to make sense nonetheless, using again the International Phonetic Alphabet – IPA – where necessary …)
If we do not want to wait until research offers reliable algorithm, we can only resort to
onomatopoeia as it has been practice for centuries. This is an effort of pattern matching
between the spectral maxima of the guitar tone to be described, and those of a speech sound
(formant = frequency of a spectral envelope-maximum). From this, it suddenly becomes
understandable that a “flute-like” (“flöten-artig” in German) guitar sound does not need to
unconditionally sound like a flute. Maybe that guitar sounds merely like a spoken “ø” (as in the
German “flöte”), it “fløøøøøtes” without being that instrument. The corresponding (second) ø-
formant is at 1500 Hz. It may be a bit higher up, if a female speaker is assumed (N.B.: it’s she
the Paula, after all). It wouldn’t be counterproductive, either, that the famed blue Cøløstiøn-
speakers have a maximum in their transmission curve around that frequency.
Fig. 8.44: Formant-frequencies of the German language, f/m speaker; (Sendlmeier/Seebode, TU Berlin).
In short: timbre (the tone color) depends on everything involved in tone generation. Not only
that: the subjective assessment criteria of the listener play a role. To objectively visualize the
sound that generates a timbre, the SPL-time-function is a complete but rather unsuitable and
abstract quantity. The hearing system does not directly process the time-function, but a short-
term spectrum determined according to complex rules. The phase is of secondary importance
in this short-term spectrum; the behavior over time of the spectral amplitudes yields the
primary hearing-relevant data-set. From the latter, and with suppression of masked (inaudible)
ranges, a secondary above-threshold data-set is derived. Contouring-algorithms (maximum-
detection), curve-following- and grouping-algorithms join what belongs together, and enable
– on the basis of memorized knowledge – recognition of instrument-typical characteristics:
timbre, pitch, and loudness, among others. There is a good deal of arbitrariness involved here:
whether strongly modulated tones are attested a fixed pitch with a special modulation timbre,
or a variable pitch with a fixed timbre: that is under the sovereignty of the listening “subject” .

8.7 Auditory experiments 8-71
8.7 Auditory experiments
The predominant part of this book discusses the function of the electric guitar by way of
physical laws, documented via formulae and measurement protocols. This enables us to
explain e.g. wave propagation, induction and signal filtering – but not the actual effect on the
listener. The verdict of the latter is only made available in auditory experiments. Therefore,
the following seeks to give a short summary of methods towards controlled sound appraisal.
8.7.1 Psychometrics
Psychophysics forms an interdisciplinary scientific area bridging psychology (= the science

of sensory perception, among others), and physics (= the science of natural processes); it
researches and describes the connection between physical stimuli on the one hand, and the
sensations and perceptions caused by these stimuli on the other hand. Psychoacoustics
narrows the wide area of physics down to sound phenomena, and connects the “science of
sound” with the “science of hearing”. Psychometrics is a sub-area of psychology that has
specialized in the (in particular quantitative) measurement of sensations. Electrical voltage is
measured with a voltmeter, temperature is measured with a thermometer – but how can we
measure the sensation of sound resulting from listening to a guitar? This can work only if the
human being is both measurement object and measurement device, with all connected
problems. The human being is the measurement object because his/her sound-perceptions are
to be determined; and he/she is the measurement device because he/she needs to describe
these perceptions. Since measurement object and measurement device cannot be separated,
errors are possible. The statement “I do not hear any tone” can mean that the measurement
object (the “subject”) indeed does not hear anything and responds truthfully. However, it
could also mean that the subject lies and does actually hear something. It could also indicate,
though, that the subject thinks that what he/she hears is not a tone but e.g. a noise – in this
case the response “not … any tone” would be truthful from his/her perspective. In order to
avoid such misunderstandings, and to obtain the subject’s assessment in the most unaffected
and most reproducible manner, psychometrics has elaborated guidelines for the execution of
experiments and their evaluation.
Reproducibility of the sound-presentation constitutes a particularly essential aspect. The

reason that a guitar sounds – compared to the studio – different on stage is found more in the
(physical) room acoustics, and not primarily in perception psychology, although the
assessment criteria (measurement device!) can be situation-dependent, as well. In order to
guarantee reproducibility in the presentation, many experimenters used specially equalized
headphones. While this is an improvement over exposing the test person to a totally
undefined sound field, it does not warrant an exact sound exposure, either. The position of the
headphone (relative to the external ear), and the individual shape of the earlobe and the ear
canal do influence the sound level.♣ Another problem is the fact that an entirely unnatural
sound field is created that turns with the head. Using precise instructions, mechanical fixation,
probe microphones, and figurative presentations, these uncertainties can be reduced to the
point that they are seen as “bearable” in daily research routine – this is then simply is as good
as it gets. Sound presentation via one or two loudspeakers would be the alternative – not
small PC-monitors, though, but calibrated premium studio monitors. Indispensable is again
documentation: room acoustics, transfer functions, impulse responses, best supplemented by
dummy-head recordings. The more is documented, the easier the decision after an experiment
series whether an effect is due to the hearing system or due to experimental methodology.
♣
Zollner M.: Interindividuelle und intraindividuelle Unterschiede bei Kopfhörerdarbietungen, Cortex 1994.

It may be that not a stored (or artificially generated) sound is to be assessed, but a sound
source, i.e. an acoustic guitar or a guitar loudspeaker. In this case the question should be
considered whether a recording via microphone or dummy head is made (and the recording
then is listened to as mentioned above) or whether live-presentation is preferred (incl.
documentation like the one involving loudspeaker presentation). A curtain hung in front of a
picture will change the visual perception, and similarly the room between sound source and
listener will influence the auditory perception. If the filtering by the room is ignored, the
assessment is unusable. Only after the sound presentation is fully optimized and documented,
the assessment of the sound may be started.
Auditory experiments may be of simple or complex, of fundamental or special character.

Threshold measurements are easy to do for the subjects. Psychometrics distinguishes
between (absolute) thresholds of stimulation, and just noticeable differences. The threshold
of stimulation tackles merely the issue of whether something is heard. Not every tone with a
sound power different from zero is audible – to be heard, the tone-level needs to be above the
threshold-level. The threshold of stimulation that is determined for tones that are presented in
quiet is called the threshold in quiet. If another (interfering) sound is present besides the
sound to be assessed, the term used is masked threshold, e.g. “threshold masked by pink
noise”. When determining just noticeable differences, the question is from which degree of
signal change a subjective difference is noticeable. For example: which change in frequency
is necessary so that a change in pitch is perceived? The subject’s task becomes more difficult
if the question is not just whether a change is heard but also how big this change is. This
magnitude estimation targeting the numerical assessment of perceived difference can lead to
significant scatter up to the point that it is actually impossible for some experiments. We can
“force” assessments, but it is hardly measurable whether something sounds better by a factor
of two or three. Psychoacoustics states that it is measurable whether a sound has double the
loudness of another sound. Yeah, kind of – but with a scatter of ±6 dB, gripe the critics.
Scatter of measurement results is not at all limited to psychometric experiments – all
measurements will include variance. It’s just that in psychometrics, the variances are
particularly pronounced and therefore need to be looked into with particular scrutiny.
No subject will increase the level always by exactly 10 dB when asked to adjust to double the
loudness. That is why the experimenter will average the intra-individually varying values,
deriving a subject-specific mean value. One subject would represent an unsuitably small
sample, and thus e.g. 24 further subjects need to do this adjustment-experiment, leading to 25
different mean values that show inter-individual differences. Again, an average is taken, and
finally we get the result that will e.g. express that “on average” the subjects will increase the
level by 10 dB to achieve double the loudness. That this mean is not valid for each and every
human being – that is often pushed to the back of our minds. So let’s play devil’s advocate:
literature reports scatter between 5 – 17 dB, and even 4 – 30 dB is found [Hellbrück 1993].
Even so: here the center of the distribution was in the class of 8,6 – 9,8 dB. Well then … that
is almost 10 dB. To conclude from the variance that the whole shebang is one giant hokum –
that would show some uncalled-for ignorance, after all. Insofar as experimenter and subject
are aware of what they evaluate, averaging methods offer the only possibility to reduce
clusters of dots to functions. Whether the assessments of fluctuation strength include a scatter
of factor 4 or 8 – they still clearly feature a band-pass characteristic with a maximum at a
modulation frequency of 4 Hz. We simply have to avoid the mistake to declare such results –
with a three-digit precision – as universally valid; average values do have a limited accuracy,
too.

Of course, experiment and averaging become questionable if experimenter and subject have
different attributes in mind. A strongly exaggerated example would be the following: the
experimenter distributes written instructions regarding the scaling of the sonority of drums.
Questions are not allowed so as not to influence the subject. And off we go – judging away on
a scale from 0 to 10. Not wanting – as a spoor student – to forgo those hourly €15.-, one tags
along. Either according to the best of one’s knowledge (or rather: perception), or according to
the Monte-Carlo-method: everything’s coming ‘round again, and even this hour will pass. The
PC generates some averages, and we have a result. The concept what “sonority” is supposed
to be – that should be shared by experimenter and subject … otherwise it all really is one big
hokum. And nobody say that a good result proves that this term “sonority” is self-explanatory.
A less construed example from the Süddeutsche Zeitung (an internationally read German
newspaper) published on 24.09.2009: positioned within an MRI scanner, a subject is shown
various photographs. Depending on the motif, the MRI scanner establishes different brain
activities. Exceptional here: the subject is a fish. And even much more exceptional: the fish is
dead. In spite of this, the evaluating computer manages to arrive at a significant mean result.
In this case, the experimenter is not a charlatan but an honorable scientist seeking to show
how much nonsense is often practiced in modern experimental brain research. N.B.: having
many subjects at hand and using modern (“Russian”) averaging algorithms won’t guarantee
solid data … or, in other words: garbage in – garbage out.
Modern psychology, and in particular psychometrics, increasingly employs statistical

evaluation methods; that may be pesky, but it’s unavoidable. The most wonderful experiment
is no good if the results are erroneously evaluated. Just as nonsensical is to continue to
(without experimental experience) process mindless data until a convenient result is obtained.
Consider that, in a source-recognition experiment, all guitars are given the numeral 1, all
trombones the number 2, and all basses the numeral 3. If the subject has now recognized four
times the 1, twice the 2, and four times the 3, then we may not average arithmetically and
state that as a mean value a trombone is recognized. These assessments or nominal judgments,
after all, and there is no mean value. It would be similarly absurd to calculate a “mean postal
area code”. That would be possible, yes, but not interpretable.
A nominal judgment groups according to names and thus congregates elements of equal
attributes into groups. Only with an ordinal judgment, a ranking is created – however
without any metric. In metrology, class-0 is more precise than class-1, and the latter is again
more precise than class-2. Class-0, however, does not necessarily feature double the precision
of class-1, and if that were the case, class-1 could well be 3 times as precise as class-2. More
mathematically: an ordinal scale is determined via inequations but not via intervals of equal
size. The latter comes into play only with interval scales, they allow for additivity based on
equidistance. What is not required is that the property of the element with the value “0”
disappears. 0°C does not imply “no temperature” but rather is an arbitrarily fixed neutral
point, and that is also why 20°C is not double as warm as 10°C. At the end of this list we have
the relational scale in which the relations of the numbers mirror the relation of the degree of
manifestation of the assessed characteristics. The sone-scale is such a relational scale: if two
loudnesses have the relation 2:1, the same ratio is also found in the corresponding sone-
numbers (8 sone is double the loudness of 4 sone). Conversely, the phon-scale is not a
relational scale: 60 phon is not double the loudness of 30 phon.
The following table summarizes scales, properties and operations. Nominal scaling only
offers equal or unequal, ordinal scaling adds in larger than and smaller than, additivity comes
in with the interval scale, and product/division is only there from the relational scale.

The median (numerical value) of a nominally scaled set cannot be determined because for this
all elements need to be brought into a ranking – which does not exist in nominal scaling. Only
the modus, the maximum rate of occurrence, may be identified. “Most letters were transported
for postal code 93057” makes sense, but “the median is postal code 93057” does not. As a
rule, to use ratios of levels is pointless – although there may be exceptions here and there,
insofar as “0 dB” indeed is mean to imply “nothing”. In terms of the SPL, level ratios are
usually without meaning – using an equalizer, however, a boost of 8 dB may be double the
boost of 4 dB.
Scale Nominal Ordinal Interval Relational

Synonyms Topologic scale Metric scale, cardinal scale
Allowable Absolute and Cumulative rate of Arithm. mean value, Geometric
statistical relative rate occurrence, variance, mean value
measures of occurrence, median, percentile standard deviation
modus
Operations ,< > , < >, + – , < >, + –, × ÷
Features Nominal feature, Ordinal feature, Cardinal feature,
categorical or ranking feature, quantitative or metric feature
qualitative feature comparative feature
Table: Scales, features, allowable operations. In addition to the statistical measures in each column, all measures
on the left of these are, correspondingly, also allowed.
Once we now have perfected the sound to be presented, and once the feature-scale to be found
is determined, the subjects (test persons) may arrive. From now on it’s: no influencing, and
reproducible instructions. With a statement given right at the start of the sort that EC’s
“Brownie” is to be assessed, an opinion like “sounds a bit thin” is not likely to be voiced –
that guitar will simply sound “killer”. In order to prevent such bias, the desired objective is
the blind test, although that is not always doable. It would be possible to assess two guitar
amps without prejudice if the amps are hidden behind an opaque curtain (a rotary table takes
care of positioning problems); however, the immediate difference between a Gibson Les Paul
and an ES-335 may only be hidden from the guitarist if rather elaborate precautions are taken.
The differences between different scale lengths (e.g. 24" vs. 25,5") are always recognized –
blind tests are impossible here. Written instructions for all subjects ensure that everyone is
told the same, and they also facilitate checking the instructions a year later. If we realize in the
course of an investigation that the subject have difficulties doing an assessment, we must not
change the instructions until the ”correct” result turns up and average subsequently over all
experiments. Out of the question is also something like averaging only over the last five
subjects (because only they have heard the difference). Difficult question: should one single
out unsuitable subjects? To assess drumsticks, you would not ask harpists to give a verdict;
the sound of a guitar amplifier can, however, certainly be judged by a non-musician, as well.
Because there are no set rules here, documentation is particularly essential (questionnaires
handed out to all subjects). If we want to do a true service to science, we measure the hearing
threshold in quiet (audiogram) of the subjects ahead of the start of the experiments. This is
because many a musician (and other people spending any length of time in noisy
environments) have generated (and have been subject to) so much sound energy in the course
of their lives that their auditory system has experienced considerable damage. Corresponding
judgments may therefore not be typical for those of normal hearing. Wouldn’t you concur
with that, dear Mr. Townshend? Mr. Townshend, sir? Mr. Peter Townshend – HELLO there??
MR. TOWNSHEND!!!

Last, we have to consider according to which method the subject is going to deliver the
judgment. That is, “last” in the framework of this short overview, because the rules of
professional psychometrics♣ are more extensive and go beyond the presently set scope.
Methods of acquiring judgments differ (among other aspects) according to the degree of
involvement of the subject. Is the latter merely supposed to give a verbal assessment (“I don’t
hear anything”), or does he/she need to twist a knob such that a tone becomes just audible (or
inaudible)? Is a scale of the assessment presented, or can the subject make one him/herself? Is
the verdict “no difference” allowed, or is a preference forced (forced choice)? Is the response
of the subject considered when new test sounds are selected? May the subject compare test
sounds as long as he/she likes, or is a decision called for after two repetitions? For decades,
psychologists have never grown tired of preaching that all these details in the experiments are
vital to the results, and so we engineers cannot but believe it, and promote it. All the while
hoping that – vice versa – the advantages of correct level-measurements find a similarly
strong lobby in the psychologist-camp.
Scientific auditory experiments are more than just calling in three pals to in order to verify the
hypothesis that the new Fender is another milestone in rock history. The last trap is found in
the formulation of the results. The statement “the Makkashitta VR-6 has some mighty
sustain” is o.k.; however, declaring “due to its maple neck, the Makkashitta VR-6 has some
mighty sustain" is, most probably, rubbish. Unfortunately, it is everyday practice in test
reports: the tester hears something (which is his god-given right), and connects without any
prove what he has heard to some kind of material characteristic (which is stultification of the
reader). Often, evident associations (i.e. from visual domain) are dragged into the arena in
order to substantiate “ear-sounding” connections (i.e. in the auditory domain). Does a silver
trumpet ring more “silvery” than a “warm-sounding golden trumpet? Science says: no, it’s all
but imagination, or influencing the player. If the latter has to play under yellow lights and
cannot distinguish the metals, he/she plays the same, and then the sound is the same, too –
despite different metals (and given equal geometry). Does that big loudspeaker have less
treble because its heavy membrane is set in motion more slowly? Mechanics say: no, you are
mistaking cutoff-frequency with efficiency. Are the sound pressures arriving at the two ear
canals indeed the only excitation quantities for the auditory sense? Well, with the answer “of
course not”, the examinee would have most likely failed the psychoacoustics exam in 1979.
But since then, much has progressed; we do learn all the time. The visual impressions play an
important role in the auditory perceptions, and thus the perceived loudness is dependent on
the distance at which we see the sound source. It’s also why the red express train is perceived
to be louder than the green one, despite equal SPL [Fastl]; and it is the reason why we may
hear “behind us” although the sound source is in front. It’s a wide field, and – for the most
part – still an only sketchily examined one.
♣
e.g.: Kompendium Hörversuche in Wissenschaft und industrieller Praxis, www.dega-akustik.de

8.7.2 The sound of the un-amplified guitar
How does the expert test an electric guitar? He first listens to it without amplification (i.e.
“dry”). “It is certain that – contrary to common opinion – the desired sound of electric guitars
and electric basses is not mainly dependent on the pickups. Rather, the wood forms the basis.
If a customer travels to see me in the ‘Guitar Garage’ in Bremen and seeks to discuss pickups,
I first listen to the instrument without amplifier.” (Jimmy Koerting, Fachblatt Musikmagazin).
Or: “For a first evaluation of the sound quality we do not need amplifier towers nor distortion
boxes – a small combo entirely suffices. It would of course be even better to test the sonic
behavior in a quiet corner, dry, purely acoustically, regarding response, balance, and sustain.”
(G&B 3/97). But why then are two guitars that sound differently dry, not able to feature these
differences anymore when played through an amp? “Surprisingly, the differences in sound
show up – compared to the dry-test – much less when connected to an amp”. G&B 7/06,
comparison: Gibson New Century X-Plorer vs. V-Factor. Or, from a different comparison:
"The Platinum Beast sounds dry powerful, warm and balanced, with velvety brilliance and
tender harmonics, while the Evil Edge Mockingbird is sonically somehow feeble, poor in the
mids, with somewhat more pronounced bass, but also clearly more brilliant and harmonically
richer. Connected to an amp, and thanks to the hot humbuckers, everything is different
though: hard to believe, but the two instruments now sound almost identical.” G&B 8/06.
Extreme examples will not serve to help in this case. Plywood (or even rubber!) is used as
material for the (solid) guitar body in order to justify significance and necessity of high-grade
body-woods. That is one extreme: using a totally unsuitable (absorbing) body, a good guitar
cannot be built; ergo-1: the wood is more important than the pickups. The other extreme: a
brilliant (“under-wound”) Strat-pickup is swapped for a muffled, treble-eating Tele-neck-
pickup with a cover made of thick brass, and the result is the statement ergo-2: the pickup is
more important than the wood. Both approaches are too lopsided.
From the point of view of system theory, the vibrating string is a generator that on the one
hand excites guitar body and neck to vibrate and thus to radiate airborne sound. On the other
hand, the relative movement between string and pickup induces a voltage. Airborne sound
and voltage are therefore correlated because of the excitation from the same source. If the
string-vibration dies down already after a few seconds, the pickup cannot generate a gigantic
sustain. Or can it? Within certain limits, indeed, it could – in cooperation with a suitable
amplifier (+ loudspeaker). The decay behavior is changed if the signal experiences limiting
via the amplifier (overdrive, crunch, distortion). This is the decay behavior audible via the
loudspeaker, because the decay of the string vibration is not changed. Or is it? Things begin
to become unfathomable, and exactly for this reasons we find such contradictory opinions in
guitar literature. If guitar and loudspeaker are positioned closely together, feedback may
certainly influence the string vibration, as well. Maybe this is where the expert advice comes
from: first listen to the guitar without amp. However: hardly any guitar player will buy an
electric guitar to play it un-plugged forever. Sooner of later he will plug it in, and then the
forecasts from the dry-test are supposed to prove to be true. The likelihood of a fortunate
result of that experiment is indeed not entirely zero: electrical sound and acoustical sound are
somehow related (correlated) – but in which way exactly is unclear to begin with.

Let’s imagine a simple experiment: the pickups of a Stratocaster are screwed directly into the
wood so that they have a clearly defined position. Will already that change the sound?
Anyway, let us assume this special sound to be the reference. Guitar, pickups, and now on to
the peculiarity: once with pickguard, and once without. That’s a pickguard made purely from
plastic so that no metal layer may generate any eddy-current damping. Now, do we hear a
difference in sound if that guitar is played with pickguard compared to being played without?
In the acoustic sound: definitely yes – in the electric sound: definitely no. Via the body, the
pickguard – if present – is made to vibrate. It has weakly damped natural modes (eigenmodes)
and is able to radiate audible sound in several frequency ranges. Do these pickguard-
vibrations act back to the string? Theoretically: yes, because “all things are connected” (as
already reportedly pointed out to the US Government by Chief Seattle as early as 1854/5).
Practically no, because between string and pickguard we find the body that weighs in with
many times over of the mass of the pickguard. The string vibrations are changed by the
pickguard only in such an insignificant degree that the electrical sound does not change
audibly. However, the radiated airborne sound does. Or another example: singers perform in
a concert hall, and listener A listens in that hall while listener B listens from an adjoining
room via the open door. Now, the door is closed – what changes? For listener B, a lot – but
for listener A, almost nothing. Very theoretically, we can again call in Chief Seattle and
demand a correction factor for the wall absorption that the closed door has modified, but in
practice not all of such lemmas have been rewarding, as the in the chief’s case rather
unfortunate history has shown.
What is the connection between the singer and the above electric guitar? In both cases there
are two different transmission paths that modify the sound they carry in different ways.
Knowing about one transmission path does not allow – in the general case – for any
conclusions on the other transmission path. The listener in the concert hall cannot even be
certain that the other listener (The Man Outside …) hears anything. This implies for guitars:
what use is the great acoustical sound if the pickup winding is broken. Caution, though: we
are again entering territory of extreme positions. Thus, not assuming a complete sound
insulation for listener B, the latter will be able to make some statements: when singing is
going on, when it is paused. Maybe, listener B can even recognize which one of the three
sound sources is trying to get to that high C: the little one, the pretty one, or Fat Lucy (also
called the stage-panzer). Any problems with intonation are perceived through the closed door,
as well, as long as the latter in not totally soundproof – and if such problems are present
within the expectations of the listener in the first place.
The thing with the expectations can be observed with guitars, also: it is astonishing how some
guitar tester become victims of their own convictions. Irrefutable credo: “Of course, the
original Les-Paul-mix of rosewood fretboard and mahogany body fitted with a thick maple
cap – that gives us the unique Les-Paul-sound”. That’s just how it needs to be written – in this
case in a comparison test (G&B 7/02). And then a copy with an alder body (stigmatized with
a “!” in the test) dares to sound good. It even commands the tester’s respect. “... it can, in any
case – be it alder or mahogany – convince with a first-class clean sound…” Well, well – let’s
not exaggerate here! Don’t forget, its alder!! And lo and behold: “… overall somewhat
subdued and a bit shy.” There we are: typical alder. However, oh great Polfuss, what happens
only a column further, with the Fame LP-IV also included in the test? "Those who go for a
typical forceful Les-Paul-sound without frills should check out the Fame LP-IV. Indeed, it
sounds the most authentic. In all areas, its sound is very similar to that of the original”.
Question: according to the test, the Fame LP-IV sports a maple neck, an oak fretboard, an
alder body, and a mahogany cap – did I get anything wrong here?

Let us postpone the discussion on materials to later, though, and return to the question of how
far the conclusion from the “dry” test to the electrical sound is admissible. Apparently there
are “robust” signal parameters that win through on every transmission path, and “fragile”
parameters that change on their way through the transmission medium. The pitch is quite
robust: whether a guitar is in tune is audible both “dry” and amplified. Not to the last cent, as
psychoacousticians know, but with a precision adequate for some first considerations. The
sonic balance between treble and bass, however, depends on the tone settings of the connected
amp – that much is as uncontested as it is trivial. The “dry” sound can make every effort: it
can never hold its own against a fully turned-up bass control. “Anyway, that’s not what we
mean”, the expert will object, “in the dry-test I can hear the fundamentals of the sound, and
the soul of the wood.” Please, dear scientists and dear psychologists – no malice now … it’s
o.k. to state something like that here, as a guitar tester who does neither have to understand
much about physics nor of psychology. However, the soul of the wood does reveal itself to
the seeker not a prima vista; it does require many séances in which the spirit penetrates the
matter; much knocking on wood needs to happen, and a tuning fork must to be pressed
against the solid body of a Stratocaster (in the Fender ads, anyway), and many years of ear
training are necessary. At least for this last point we should be able to reach a consensus,
shouldn’t we? This is not supposed to be about the guitar-o-phobe agnostic with progressive
dysacusis, but about the more or less pronounced aficionado of the instrument. Those who –
with their more or less extensive listening experience – indeed hear details in the sound not
accessible to the layperson.
Problem: how do you describe such sound-details? This is the classic conceptual formulation
and task of psychophysics and psychometrics that frequently leads to similarly classical
misunderstandings. A verbal description (dead, woolly sound) is rejected at the physical
docking-port as much too ambiguous and imprecise, just like the exact physical description
(8,43% degree of amplitude-modulation at 944 Hz with fmod = 6,33 Hz) is objected to by the
artistic/mystical faction as pipe-dream-y and too abstract. Logically, any proposals of
compromise trying to bridge the two realms are dismissed by both sides. Well then: rather
than the wood’s soul, often a dead or a lively sound is mentioned. What distinguishes live
from dead matter? The matter that is alive – it moves! And already we have the first
objections, because that would define the pen dropping from the table as alive? O.k., so we
turn to a fundamental philosophical contemplation of life in particular, and of the universe
and everything in general … NOT! No, really not. What is alive does move. Period.
Conferred to the guitar sound: an artificial tone with its strictly harmonic partials all decaying
with the same time constant, sounds dead. However, if the partials decay with different speeds
and with different beats, the impression is one of movement and life. In this, the term
“movement” may indeed be seen in its original meaning as change in location: when a sound
source changes its position in a (sound-reflecting) room, time-variant comb-filters vary the
signal spectrum – the movement in space has the effect of a “movement” in the sound. Way
back in prehistoric times it was presumably in support of survival if moving sound sources
were given a higher priority than static sources; at the same time early researchers in
communication discovered that speech sounds will only convey information if they include
variations. Without pushing too far into foreign territory: there would be enough reasons why
the human auditory system continuously hunts for spectral changes. Even though electric
guitars are younger that roaring tigers and vandals screaming “arrrghh!”, our hearing has its
capability to analyze, and it takes advantage of it. A lively tone rich in beats sounds more
interesting than a dead sound – at least as long as instrument-typical parameters are
maintained.

Similarly to the pitch of the string, the beats between partials can be rather robust relative to
the transmission parameters, and therefore it is imaginable that the expert may be able to
deduce criteria for the electrical sound from the “dry” test. On what does the robustness of the
signal parameters depend? Frequency-dependent signal parameters – such as the spectrum –
loose their individuality if the corresponding frequency-dependent system parameter (the
transfer function) has a similar shape. Three examples follow:
1) Psychoacoustics [12] describes the balance between high and low spectral components as
“sharpness”: treble-emphasizing sounds have a high sharpness; turning down the treble
control reduces the sharpness. Spectral details are not as essential for the calculation of
sharpness as the basic (smoothed) run of the spectral envelope. To be more precise: the
sharpness is taken from the weighted loudness/critical-band-rate diagram which has a mere 20
sampling points in the frequency range important for electric guitars. (Transmission-)
frequency-responses of guitar amplifiers can be represented with the same increments (Fig.
8.45), and from the kinship of the two data-sets we can conclude that the sharpness of the
“dry” guitar sound in general does not correspond to the sharpness of the amplified sound. In
other words: changing the tone controls on the amplifier allows for changing the sharpness –
from this point of view, sharpness is not a robust signal parameter.
Fig. 8.45: Tone control of a Fender amplifier (transmission factor). The dots on the top mark the critical-band
grid (discretization of the abscissa in order to calculate sharpness).
2) Beats between partials can in the time domain be described as amplitude fluctuations, and
in the frequency domain as the sum of closely neighboring partials. For example, two partials
of equal level but slightly differing frequencies (e.g. 997 Hz, and 1003 Hz) lead to the
auditory perception of one single 1000-Hz-tone fluctuating in loudness with 6 Hz [3]. In order
to change this beating, a highly frequency selective operation is necessary. Such an operation
is untypical for tone controls in amplifiers. From this point of view beats in partials are robust
relative to simple tone-control networks.
3) The spectrum of a quickly decaying sine-tone (Fig. 8.46) is largely limited to a narrow
frequency range. Any changes in the decay behavior need to be done using highly frequency-
selective methods, too. In other words, a linear, guitar-amp-typical tone control network will
practically not change the decay-behavior of individual partials – the decay behavior is robust
in this respect.

Fig. 8.46: Decaying sine-oscillation, f = 1000 Hz, time-constant τ = 0,3 s.
These simplified representations do, however, require supplements in some points. It is not
only the transfer factor of the guitar amplifier that changes the spectrum of the string
oscillation. The loudspeaker (including its enclosure) acts as filter, as well; in the detail, its
transmission curve is of stronger frequency dependence than the tone-control network is. Still,
a loudspeaker membrane does not reach the high (resonance) Q-factors of decaying guitar
partials – it would have to generate clearly audible natural tones for that, and this it does
exactly NOT do. The last filter in the transmission path is the room with its reflective borders.
Its effect cannot be neglected even in the “dry” test; when playing connected to the
amp/speaker, the distance to the speaker needs to be added in as variable, as well. As long as
one remains within the near-field of the loudspeaker, the effect of the room can be regarded as
equal for both playing situation in a first-order approximation.
Special consideration is required for effects that achieve more than what simple tone control
does. Adding artificial reverb can extend decay processes and feign life that is not included in
the original in that form. Chorus/phaser/flanger are time-variant filters with high (resonance)
Q-factors – their use always aims at changing the fine structure of the partials. Single band,
and in particular multi-band, compressors change the decay time constant of individual
groups of partials. Overdrive has similar effects but adds in additional partials. It is therefore
very well possible to also influence the signal parameters designated as robust above.
However, even without deploying radical effects one may – within certain limits – extrapolate
from the sound of the unamplified guitar to the sound of the amplified guitar. Which of the
many beat- and decay-parameters are crucial to the ‘good’ sound, though, is at the most
implicitly appraisable … and we have not even touched the wide field of frequency- and time-
related masking [12]. Therefore, only this principle can hold: the unamplified sound of an
electric guitar should basically not be evaluated. Only for the expert, and in consideration
of his/her special knowledge and listening-experience accumulated over decades the
exception, this rule allows the exception that the “dry” test reveals “everything”, after all, in
the individual case. Experts who may claim this exception for themselves are: testers of all
guitar magazines, all guitar sales-personnel, all guitar players who have had of who have
wanted to have a guitar for more than a year, and all listeners to CD’s who still have the
sound of Jeff Beck’s signature guitar ringing in their ears (see Chapter 7). And please, dear
experts who have now received so much legitimization for your obviously indispensable
“dry” tests: that the assessment of tactile vibrations is nonsensical – that should by now be
o.k. for a consensus, should it not?

Concluding the topic of guitar tests, a few citations in the following:
Yamaha Pacifica-guitars (maple neck, alder body) in a comparison test: "The acoustically quite
comparable basic characteristics of the Pacificas differentiate themselves rather clearly according to
their pickups, after all. (G&B 6/04)."
Gibson Les Paul Faded Double-Cutaway: "Already with the very first striking of the string it becomes
clear that the economy-varnishing curbs the resonance properties of the woods to a lesser degree. The
guitar vibrates from the feet (strap-knob) to the tips of the hair (machine heads) so intensely that you
can feel this even in your own body; (G&B 6/04)."
Ibanez IC400BK: "The slight underexposure of the E6-string as it appears in the dry test has suddenly
disappeared with the support of the pickups; (G&B 6/04)."
Squier-Stratocaster, comparison: mahogany-body vs. basswood-body: using the neck- or the middle-
pickup, both guitars sound almost identical (G&B 5/06).
"Picking up the Pensa-Suhr-guitar and playing it un-amplified, the reasonably learned ear
immediately hears where this is at. … both standing up and sitting down, you feel already in your
belly the fantastic vibration-behavior of the outstandingly matched woods." (Fachblatt, 6/88).
"Despite using humbuckers, a Strat will never turn into a Les Paul"; G&B 2/00. Ozzy Osbourne on
Joe Holmes: "In fact, I normally don’t like Fender guitars. But Joe gets this fulminant Gibson sound
out of them"; (G&B 2/02). "Jimmy Page recoded the whole of the first Led-zeppelin album using a
Telecaster; the guitar sound on this album is exactly that of a Les Paul"; (G&B Fender-special-issue).
Mark Knopfler: "If I look for a thicker sound, I use my Les Paul; it simply is more dynamic. That
doesn’t mean that I couldn’t do the same thing with a Stratocaster"; (G&B Fender-special-issue).
Gary Moore: "Some people think that a Fender Stratocaster is heard on 'Ain't nobody'; actually, that
is my own Gibson Signature Les Paul"; G&B 7/06 p.91.
Big mass of wood (3,9 kg): Due to the big mass of wood, the response seems a bit ponderous and the
notes don’t get off the starting blocks that fast; (G&B 7/06).
Even heavier (4,15 kg): the guitar vibrates intensely, responds directly and dynamically, each chord or
note unfolds crisply and lively; (G&B 8/06).
Despite the enormous mass of wood (3,85 kg), almost every note responds crisply and dynamically,
and unfolds very swiftly; (G&B 7/06).
The lower mass can be more easily made to vibrate; (Thomas Kortmann, gitarrist.net).
A slender guitar body also creates a slender sound; (G&B 7/02).
Thinner body = less bass; (G&B 4/04).
Thick neck = sonic advantages; (G&B 8/02). Thin neck = round, fat tone; (G&B 10/05). Thin neck:
the lower the mass that needs to be moved, the more direct and quickly response and tone-unfolding
get off the starting blocks; (G&B 3/05). Crisp and direct in the response, every note gets quickly and
lively away from the starting blocks, despite the immense mass of the neck (that needs to be first set
into motion, after all); (G&B 9/05). A thin neck has no acceptable vibration behavior whatsoever;
(G&B 3/97). Sonically advantageous is that the neck weighs in with a lot of mass; (G&B Fender-
special-issue). The Ibanez JEM 777 sports an extremely thin neck design: the basic tonal character is
powerful and earthy; (Fachblatt, 6/88). Of course, the shape of the neck contributes to the tonal
character of the guitar, as well; (G&B, 12/06). What is absolutely not true is that thick necks will
sound better than thin ones. I have already built the same guitar with the thick and a thin neck and
could not find any difference; (Luthier Thomas Kortmann, gitarrist.net)

8.7.3 Tactile vibration perception
There is scarcely any guitar test-report that does not praise the exorbitant vibration-happiness
of the electric guitar under scrutiny: "The design shows considerable resonance properties,
after each picking of a string it vibrates intensively and clearly noticeable.” G&B 9/06. Or: “
From a vibration-engineering point of view, the MTM1 ranks at the highest level because the
whole structure resonates intensively into the last wood fiber after each picking of a string;
this results in a slowly and continuously decaying sustain.” G&B 8/06. Or: “Combined with
the given open freedom of vibration (sic), we achieve a beaming sound color.” G&B 8/06. Or:
“Less mass can more easily be made to vibrate.” Luthier Thomas Kortmann, Gitarrist.net. Or:
“At Fender they even proceeded to build bodies from several wood-parts … Of course, the
ability of the wood to resonate is restricted by such a number of differently sized pieces.” And
loc. cit.: “That Ash moreover has almost optimum resonance properties was thankfully
acknowledged at the time. It does not bear contemplating that Leo Fender might have opted
for mahogany back in the day.” Day et al. Or: “Clearly noticeable right into the outermost
wood fibers, both Strat and Tele show very good resonance properties.” G&B 4/06.
Mind you: we are discussing electric solid-body guitars here, and not acoustic guitars. The
clearly noticeable vibrating of the guitar is taken as a criterion for quality. Why don’t we let
one of the fathers of the solid guitar, Lester William Polfuss, speak: "I figured out that when
you've got the top vibrating and a string vibrating, you've got a conflict. One of them has got
to stop and it can't be the string, because that's making the sound." Mr. Polfuss sought to let
only the string vibrate, and not the guitar top. O.k., one could object that the man was a
musician, not an engineer. Still, he was a musician that replied to the question of who had
designed the Gibson Les Paul with "I designed it all by myself". The string is intended to
vibrate, and the body should just shut up and be quiet. Only the very nit-picking ones will
throw in at this point that only the relative movement counts, i.e. if the strings remain at rest,
and instead the body would … no, enough with the theories of relativity, it does work better
the other way ‘round. However: what does that mean – better? What characterizes a better
sounding guitar? In his dissertation [16], Ulrich May cites D. Brosnac with the insight that a
guitar made of rubber would absorb the vibration energy of the string within a short time and
therefore would not sound right. That is understandable but does not prove that ash (or maple,
etc.) is better suited. Evidently, there are unsuitable body-materials that will withdraw an
unbecomingly big amount of vibration energy from the strings. Rubber is among these
materials – but who would want to build a guitar out of rubber? Presumably, damp towels♣
also rank among the unsuitable materials. Or, fresh from the sleep-lab: since a bed of a length
of 1,45 m (about 5 feet) is uncomfortable for most grown-ups, a 2,12-m-long bed has to be
more comfortable than a bed of 2,05 m length. Or, more guitar-specifically: what the luthiers
have learned for the acoustic guitar cannot be wrong for the electric guitar. A guitar has to
resonate. Right into the outermost wood-fibers. Intensively and clearly noticeable.
So, what can we feel – as human being in general, and as a guitar-tester in particular? That
depends, of course, on the stimulus and on the receptor. However, in terms of vibrations, the
subcutaneous Pacini-corpuscles react most sensitively to stimulus frequencies of 200 – 300
Hz; they sense vibration amplitudes as low as 0,1 µm. That also implies that the sense of
vibration becomes increasingly less sensitive above frequencies of about 250 Hz. Sound-
shaping harmonics remain largely hidden from the tactile sense.
♣
because of the high „damp“-ing ...

Fig. 8.47 shows the frequency dependence of the vibration threshold, i.e. the vibration
amplitude that needs to be reached in order to generate any vibration perception in the first
place. Besides the dependency on frequency and amplitude, the exact shape of the curve
depends also on the area of the vibrating surface, and on the location that is stimulated. The
given graph can be seen as typical for the thenar. Thus, if a guitarist feels a vibration in the
neck or the body of the guitar upon striking the strings, it will be a case of low-frequency
vibrations. To check via a calculation: if we take 10 N as force at the bridge, a mass of 4 kg,
and 250 Hz as stimulation frequency, we get a displacement of 1 µm. It is therefore no
wonder that vibrations result that are felt, even without any resonance-amplification.
Fig. 8.47: Vibration threshold. Only values

above the threshold lead to a perception of
vibration. According to this graph, a vibration of
an amplitude of 0,4 µm can be felt at 300 Hz; at
800 Hz it would not be felt anymore.
“Schwingungsamplitude” = vibration amplitude;
“Vibrationsschwelle” = vibration threshold
Therefore it is less a question of whether noticeable vibrations can emerge, but more how
these should be assessed. If we take up again Les Paul’s idea, any body-vibration to speak of
would be counterproductive. With a lot of mass (ten-pounder Paula), this ideal can be
approached at the cost of wearing comfort, and disregarding natural modes (eigenmodes) that
amplify the vibration. The guitar neck in particular must not be too heavy; it will resonate to a
noticeable degree in any guitar. What would in fact happen if guitar body and guitar neck
could be manufactured to be vibration-free? On every guitar of this kind, comparable strings
would vibrate in an identical manner given comparable picking! Individuality is
imperfection, and it would fall by the wayside. In the acoustic guitar, the luthier seeks to
form the transfer function frequency-dependently, and thus let some frequency ranges be
radiated better, but conversely let other frequency ranges be radiated worse. An individual
sound does result that way. The same principle could be applied for electric guitars, as well,
and neck and body could be made to vibrate more at certain frequencies, i.e. the vibration
energy would be more strongly dissipated. Whether this is indeed desired can only be judged
in an overall consideration of all sound-shaping elements. Still, it would be a remarkable
coincidence if exactly those frequency ranges for which the tactile sense is particularly
sensitive would require the strongest damping. For one thing is certain beyond all doubt: the
vibration energy that is felt, it is sourced from the string. The more intensive “the whole
structure resonates”, the less the string vibrates. One may agree or disagree with Les Paul’s
ideas – the law of conservation of energy should rather not be objected to.
Whether we would like to contradict Day et al., however, is again left up to us: “the vibrato-
system itself was given a knife-edge-type shape at the six holes foreseen for the screws
retaining it. The whole system was therefore mounted optimally in a very low-friction manner
but still could transfer the vibrations of the strings optimally to the body.” Indeed, this path is
known: “because the tawdry goes down to the corpus unsung” … Schiller, Nänie. Or
something like that.

The almost-empty page
Hi there old-timers,
Hey by ‘n’ large I can unnerstand and confir yer notions.
Old gittars just sound diff’rent than new ones, and it’s in the ear and the fingers of the beholder to decide if a
gittar has some upwards potential or not. Course, the bottem supstanse has to be right butthat is then a matter of
espirience so you can tax that. An axe who wont resone at all when its played won’t be impressd by that after
1000ds of plain hours, eithe.
So I pay atentsion to vibro-and reso-behavor in the newones.
Fact is too that plaing lots impoves yer own s kill and that change the sound agin (hope to the better).
Have read the article by U.P. He cam to the conlusion that such attack-apparatuss really change the gittar (cause
stuff is done to it fysicaly like things happen in that cryo-tuning – you can see on youtube what Joe does at G-
Cener):
What the two metods make clear, it is not really possible to esimate how the sound’ll differ. To show that they
would have to record the gittars before and after the treatement with a fixed recording setup to make adifference
somewhat possible to hear objectively.
If in the studio someone has tried to record one and same song on two days, he knows that really there can be
differences in the sound if you jus drop yer axe after a cool session and leave everthing like that. The next
morningthat super-sound sounds suddenly not as super even though you have changed nothin. Hereit’s again the
subjetive hearin. As such even such a differece can not be felly objectively spread. We will have to try out such
stuff ourself sometime if we really want to do that.
I kinda more think it better to play ones instrument so that one gets better instead all the time to run around in
search for perfect soiund. Course too a new instrument, ol or new, needs to be won over by much plaing!
Play It, hear It and Sound like yourself !
Jack J.
“Play-in” the guitar and find the desired sound (loosely translated from the G&B-Forum of 11.01.2014)

9. Guitar circuitry
In its original form, the electric guitar was equipped with one, two or three pickups. The
voltages of these could be selected or combined with switches. Guitars fitted with four
pickups did surface occasionally but proved to be of little interest – apparently switching
between the pickups gave too little the sonic difference. With the standard circuitry normally
in use, the switch on the guitar allows for the selection of a pickup or the parallel-connected
combination of two pickups. Later variations on this arrangement additionally offered series
connections and phase reversal. Controlling volume and tone was usually achieved via the
installation of simple RC-networks. For manipulating tone, one occasionally one finds more
complex filter networks (e.g. in the Gibson ES-345) or battery-powered amplifier and filter
circuits. The following descriptions relate to simple passive circuits – more extensive
information can e.g. be found in the book "Electric Guitar – Sound Secrets and Technology"
by Helmuth Lemme.
9.1. Potentiometers
In the guitar, potentiometers (i.e. adjustable resistors) are connected to the pickups to control
volume and tone. The respective values are in most cases ca. 250 or 500 kΩ, less frequently
used are 100 kΩ or 1 MΩ. The tone potentiometer allows for shunting a capacitor (typically in
the order of 20 - 50 nF) in parallel to the pickup. As one turns the tone knob counter-
clockwise to the end position, the potentiometer reaches 0 Ω, and the now directly connected
capacitor further reduces the resonant frequency of the pickup and cable in combination to
values below 1 kHz. Turning the knob to the other extreme position leaves the full resistance
of the potentiometer connected to the circuit. This results in a minor dampening of the
resonance with the capacitor acting like a short circuit (i.e. having no audible effect in itself).
Some guitars sport a special potentiometer which completely switches the resistance out of
the circuit in the clockwise end position – in this case the resonance is fully retained without
any dampening. Normally, however, it is safe to assume that the volume and tone controls do
have a load-effect on the pickup. To be certain, one would have to make a measurement or
have access to the schematics. The latter are also advantageous if the guitar holds a battery
and an amplifier the input impedance of which would be the effective load to the pickup.
Fig. 9.1: Schematics for an electric guitar

T = pickup, P = volume control, R = tone control, C = capacitor.
The figure on the right shows the effective electrical situation for clockwise position of the controls ("full up").
The clockwise end-position of the potentiometer taps in the figure on the left is at the upper end of the resistor
for the volume control and on the lower end for the tone control

9-2 9. Guitar Circuitry
Fig. 9.1 shows a typical guitar circuit. By twisting the knob, the potentiometer tap (the arrow
in the figure) can be moved continuously between the end points. The rotational angle usually
extends over about 270°. For linear potentiometers, the tapped resistance is proportional to
the rotational angle while for logarithmic potentiometers, the resistance change rises
progressively (see Fig. 9.2). Theoretically, the potentiometer characteristic can be shown as
exponential function. The logarithm of the exponentially growing resistance is proportional to
the rotational angle: thus the designation logarithmic potentiometer. In practice, substantial
deviations from the exponential function are likely because for cost reasons this desirable
characteristic is only approximated.
Theoretical dependency on angle of rotation:
R/Rmax = kx-1 x = 0 ... 1 k = 50 ....300
Fig. 9.2: Resistance characteristic for a linear

potentiometer (straight line) and logarithmic
potentiometers (hatched area).
The dashed line shows a typical characteristic of
potentiometers used in vintage Fender guitars
Potentiometers of recent production typically have tolerances of about +/-20%, i.e. the actual
value of a 250-kΩ-Potentiometer lies between 200 and 300 kΩ. Even 150 to 350 kΩ values
can occur as outliers – especially with older guitars which appear not to have been subject to
any excessive quality control. If a tone pot has a value of 350 kΩ rather than 250 kΩ the
guitar sounds more brilliant. If this is not desired, turning down the pot slightly (for the purist:
twist the knob counterclockwise) will compensate. Connecting a 0.9 MΩ resistor in parallel to
the pot will do the same job. On the other hand, a pot having merely 150 kΩ will make the
guitar sound duller. In this case the only remedy will be exchanging the pot. Still really
dramatic differences are not to be expected (see Fig. 9.3). The most important parameter for a
potentiometer are resistance and angle-over-resistance characteristic. The power rating
(usually 0,1 - 0,5 Watt) is unimportant since the pickups will generate merely a few micro-
Watts. All other parasitic electrical effects (capacity, inductivity) can be neglected in the
audio range. Good contacts (i.e. no drop-outs across the turning range) go without saying
when using brand potentiometers. The latter will cost in the order of $ 2.- to 4.-. Prices of
more than $ 100.- for "vintage parts" are not justifiable from an engineering point of view.
Fig. 9.3: Influence of different potentiometer

values for a Fender Stratocaster. Tone and
volume pots were (both!) assumed to be having a
value of 300, 250 and 200 kΩ.

9.1 Potentiometers 9-3
In Fig. 9.4, the effect of tone and volume control is shown for a Stratocaster. Merely turning
down the volume slightly will already make the resonance peak disappear. The sound
becomes duller. The reason for this is that a part of the resistance of the volume control is
now connected between the pickup's coil inductance and the capacitance of the cable to the
amplifier. This series resistance dampens the resonance. When turning down the volume
further, a further resonance at a higher frequency appears but this is not really usable since the
signal level is very small. Turning down the tone control first also reduces the resonance peak
but – at fully CCW-position – then leads to a resonance at a lower frequency (typically around
350 Hz).
Fig. 9.4: Stratocaster: Volume control (left), tone control (right); 600-pF-cable; 1-MΩ-amplifer-input
Even more extreme is the situation with the Fender Jazzmaster (Fig. 9.5). Here, the high-
impedance volume pot (1 MΩ) kills the treble radically already when turning down the
volume just a bit. Of course, the resistance changes only have the shown effect if a high-
input-impedance amp is connected to the guitar. A typical input impedance for tube amplifiers
is 1 MΩ (this is indeed considered "high"). Smaller input impedances of the amp will reduce
the Q-factor of the resonance circuit and therefore the resonance peak.
Fig. 9.5: Jazzmaster: Volume control (left), tone control (right); 600-pF-cable; 1-MΩ-amplifer-input
As a potentiometer is turned fully CW or CCW, a resistance remains between the tap of a

potentiometer and the connection close to it. This might also deserve some consideration.
High quality pots have a very small remaining resistance (< 50 Ω). Audible effects can be
expected if the remaining resistance is more than ca. 500 Ω – however potentiometers
showing this are very low grade and should be discarded.

The treble loss perceived with turning down the volume pot can be reduced by soldering a
bridging capacitor between the tap and the CW end connection of the pot (Fig. 9.6). For low
volumes (i.e. a turned-down volume pot) a stronger treble boost can be achieved. When in
1967 the Telecaster was fitted with a 1-MΩ-volume-pot, Fender evidently discovered the
strong tone change this pot can result in: the guitar received a bridging capacitor (1 nF).
Fig. 9.6: Stratocaster: volume pot with bridging cap; 150pF (upper left), 1nF (u. r.), 1nF//100kΩ (l.l.);
Jazzmaster: 1nF//150kΩ (lower right); all diagrams with 600pF-cable and 1MΩ amp input impedance
Selective tone changes are possible with LC-filter-networks installed in the guitar. One
example is shown in Fig. 9.7: for some Gibson guitars a 8-H-coil is fitted. A rotary switch
connects various capacitors in series with this coil creating a resonant shunt connected in
parallel to the pickup. The result is an attenuation of a narrow band of frequencies. It appears
that the tonal control achieved this way did not enthuse many guitar players since the demand
remained rather low.
Fig. 9.7: Vari-Tone-Filter of the Gibson Lucille: 6 frequency responses selected via a rotary switch (left, cable
capacity 700 pF ). Right: the cable capacity is varied (330 pF, 680 pF, 1000 pF) for Vari-Tone position no. 3

9.2 Tone Capacitor & 9.3 Connecting Wires 9-5
9.2. Tone capacitor
The tone capacitor is in a series connection with the tone pot and allows for an attenuation of
the treble frequencies. Values of 20 - 50 nF are often used, more rarely one finds 100 nF (in
old Fender guitars). Capacitors can be characterized by their capacity value (measured in units
of Farad) as a first order approximation. Additional parameters may be important – this
depends on how exact the description needs to be.
A capacitor stores separated (positive and negative) charges. At the same time it converts a
small amount of the electrical energy into heat and thus has the effect of a loss resistance. In
the overall balance energy cannot be "lost", however the generated tiny thermal energy is not
available anymore as electrical energy – thus the use of the term "loss". There are several
reasons for capacitor losses: insulation resistance in the dielectric, connection-wire- and
electrode-resistances, polarization losses (the oscillation of the dipoles in the dielectric re.
their rest position causes a warming, see 10.9.3)
Simple models for a capacitor extend the capacitor schematic by a resistor (Fig. 9.8). A
characterizing value is the dissipation factor d. The arctangent of d results in the dissipation
angle ∂ which describes the phase shift due to the loss.
d = G/ωC; d = R·ωC
d = tan∂ = dissipation factor
Fig. 9.8: Simple capacitor equivalent circuits: NEB (left), HEB (right)
In literature, the GC-parallel circuit is designated as the low-frequency equivalent circuit (in
German Niederfrequenz-Ersatzschaltbild: NEB) while the RC-series circuit is designated as
the high frequency equivalent circuit (in German Hochfrequenz-Ersatzschaltbild: HEB). For
the NEB, d has a reciprocal dependency on frequency, while for the HEB this is proportional.
Measurements show that the NEB is not suitable at all for the audio range because the
dissipation angle increases with frequency and does not diminish (Fig. 9.9). On the other
hand, the HEB reproduces the frequency dependency only very roughly – the quantitative
correspondence is unsatisfactory.
Fig. 9.9: Dissipation factor

d(f). Measurements of various
22-nF-capacitors. The high-
frequency equivalent circuit
results in the dashed straight
line (10Ω in series with 22 nF).
MKC = Polycarbonate
MKT = Polyester
KP = Polypropylene
KS = Polystyrene = Styroflex

Fig. 9.9 shows that capacitors can have rather different electrical characteristics - even if their
capacitance values are the same. However, it would be wrong to reason based on this fact that
the sound of an electrically amplified guitar would vary correspondingly. Components other
than the capacitor determine the overall losses in the electrical circuit. With the tone control
"fully up" (i.e. the tone pot has its maximum value) there are typically 250 or 500 kΩ in series
with the tone cap. Compared to this it is insignificant whether the capacitor losses are 500 Ω
or merely 10 Ω. Even if one would radically replace the tone cap by a short, the transmission
factor changes les than 0,01 dB in the relevant frequency range. This does not mean, however,
that the tone cap has no effect at all if the tone pot is "fully up". It does: relative to the tone
pot it works – as a very good approximation -– as a short. It makes no difference whether its
value is 20 or 60 nF, and it makes no difference whether the dissipation angle is 0,1% or 5%.
With the tone control "fully closed" (i.e. the tone pot has its minimum value), losses are
dominated by the pickup and the volume potentiometer connected in parallel. For the
Stratocaster the pickup inductance works with the tone cap towards a slight resonance peak
around 350 Hz (Fig. 9.10). Only here the capacitor losses have any effect. In Fig. 9.10 the
corresponding transmission factors for both an ideal, lossless capacitor, and an extremely
lossy capacitor are shown. Lossless implies a d = 0 while for this example d = 60% is taken
for the lossy component. Such a "bad" capacitor is not normally soldered into any guitar. If
one would opt for one of the "bad" capacitors from Fig. 9.9 (e.g. choose a d = 0,1%), the level
differences in comparison to the lossless capacitor would amount to ∆L < 0,1 dB i.e. they
would be inaudible. Therefore, for the tone control fully closed it is still true that the
dissipation factors of customary available capacitors have no audible consequences on the
sound whatsoever. This does not only hold for the Stratocaster but for other guitars. Indeed,
even the tone caps in a Les Paul are subject to the same laws of physics – irrespective of the
price they command on the vintage market. To take a quick look at the remaining resistance
of the tone pot: a fully closed potentiometer will of course not result in an ideal short-circuit,
however even the remaining resistances (< 100 Ω) of low-cost pots will easily suffice and do
not lead to audible differences.
Fig. 9.10: transmission factor of the

Stratocaster, tone control full down
two different capacitors (solid line
= lossless)

If capacitor losses have no audible effects on the sound – how come there are som many
reports from guitar players who state that their instrument sounds "totally different" after
changing the capacitors? Discarding those cases where the guitarist (or sound guru) also
changed the strings as well (since everything was taken apart anyway), enough cases remain
which merit consideration. Could there be – other than the dissipation factor – other (possibly
undiscovered) parameters to describe the electric effect of a capacitor? Is this question already
to restrictive again? Could a capacitor generate non-electric effects? In principle yes: from a
mechanical point-of-view it is a mass suspended from springs (the connecting wires). Thus it
could co-vibrate. This observation encourages to go further: Does the John-Lennon-Casino
sound authentically only when the knob has been lost? Is the original E.C. sound only
generated if a cigarette is clamped between strings and headstock? Does then the sound
change because the mass of the co-vibrating cigarette goes up in smoke over time? There
would also be microphonics and tribo-electricity (Ch. 9.4) ... this will not be considered here.
But back to the electrical parameters: the modeling of a capacitor via an RC-network is only
admissible if insinuate linear behavior. However, the moment a voltage is applied to the
capacitor's electrodes, attraction forces appear which reduce the electrode distance - and an
increase of the capacity follows suit. The systemic quantity "capacitance" becomes dependent
on the signal fed through the capacitor, and this situation points to a nonlinear system
behavior. Distortion factor measurements show, however, that such non-linear processes
are insignificant: at 2 Vpp the measured distortion amounted to less than 0,01% for film
capacitors and 0,1% for ceramic caps. Consequently, this aspect can be excluded as a reason
for audible differences between capacitors in electric guitars.
So, what remains? The capacity itself, of course! With all the considerations regarding
capacitor characteristics we must not forget that the capacity is subject to production
tolerances. A new capacitor of nominally 50 nF may well have a real capacitance of only 40
nF. In the mid-20th century, tolerances of +/- 20% were not uncommon, and even today
tolerances of 1% are commercially available but certainly not the standard. Fig. 9.11 depicts
the effects of a capacity tolerance of 20% for a Stratocaster – and such level differences are
without a doubt audible. Therefore it is conceivable that a guitarist who changes the el-cheapo
capacitor fitted into his guitar for a $50 "replica cap" indeed notices a change in sound. This
change would have been achievable with a regular MKP capacitor costing a full 18 Cents as
well .... but of course an "original bumblebee" exudes are radically different aura (i.e.
"mojo"), and everybody should reach happiness after their own fashion. The after-market
industry as well lives off those who furnish their $100 guitar with four Centralab pots ($100
each) and two replica caps ($50 each) – which helps to distinguish oneself from the many
unenlightened.
Fig. 9.11: transmission characteristic for a

Stratocaster, tone control turned down, two
different capacitors: 60 nF and 40 nF. When the
Strat was released originally, Fender fitted 100
nF capacitors, from 1970 50 nF were used,
followed by 22 nF from 1983. The smaller the
capacitance, the higher the resonance is in
frequency and the stronger it is pronounced.

The decision for or against a certain tone cap will always depend on subjective preferences.
Rumor has it that there are those Jazz guitarists who continue to ask themselves in puzzlement
what might have caused Gibson to include a bridge pickup on an ES-335 might be. Why then
shouldn't there be a Stratocaster owner who exchanges the 22nF tone cap for one with 100
nF? The guitar could be closer to what Leo Fender devised originally, or it could sound less
shrill .... the larger the capacitor, the darker the sound with the tone control turned fully CCW
(Fig. 9.12). On the other hand, rotary switches are available which allow for selective
connection of smaller capacitors (e.g. 1 - 10 nF. To each his/her own – motivated by no-
physical thinking (same as Jeff Beck has), paraphysics ( ), pragmatism (was already
installed, is ok), or devotion (was recommended in the March issue of "Guitar Picks &
Licks"). Those who require the exact nominal value but have no capacitance meter at their
disposal could buy a 1%-tolerance-MKP-capacitor for 60 Cents. Those who are happy taking
a risk buy a handful of 5%-tolerance-MKP-caps (20 cents each) and check whether they can
already hear differences between the capacitors. From the dielectrics listed in the following
table, polypropylene and polycarbonate are particularly suitable, but MP, KT, MKT or NDK
may be used without audible deterioration. Of course, the capacitor needs to be undamaged. A
styroflex cap which got too close the soldering iron may well be much worse than an
unscathed MKT-capacitor.
Designation Abbreviation d in % comment
Glimmer Mica >0,1 difficult to obtain, large, unpractical for guitar

Polystyrene = Styroflex KS, MKS 0,1 very high-grade
Polypropylene KP, MKP 0,3 highly suitable
Polycarbonate KC, MKC 1 highly suitable, very good temperature coefficient
Paper MP 4-8 well suited
Polyester KT, MKT 5 - 10 well suited
Ceramic class 1 NDK < 1,5 well suited
Ceramic class 2 HDK < 30 unpractical for guitar
Ceramic class 1 - < 60 unpractical for guitar
Table: Dissipation factors of commonly used dielectrics
Fig. 9.12 shows the effects of the tone cap with the tone control all the way "down". Cable
capacity is 500 pF; input impedance of the connected amplifier is 1 MΩ. With the tone
control all the way up one gets the dashed line.
Fig. 9.12: effect of different tone caps: 100 nF, 50 nF, 22 nF, 10 nF, 3,3 nF; dashed line: tone control fully CW

9.3 Pickup Connecting Wires
One would think that the wire connecting the pickups with the switches and controls do not
have any significant influence on the electric parameters of a guitar. In most cases this
assumption would indeed be correct – however there are exceptions.
In Fender guitars, the internal wiring is often done via single stranded wires which are paired
and soldered to the pickups as a so-called two-wire line. Such a connection can - for the audio
range - described with very good approximation as a pure capacity having about 50 pF/m. A
length of 20 cm (as it would typically occur inside a guitar) would thus yield a capacitance of
10 pF which is a value that is clearly negligible relative to the capacity of the guitar cable.
Losses, as well, do not play any role: even if one would assume d = 0,01, the loss resistance in
the equivalent circuit would be more than 100 MΩ.
As an alternative to the two-wire line, coaxial wiring may be used. An insulated internal
conductor is surrounded by a concentric shielding braid or stranded wire. Depending on
geometry and the dielectric, capacities of 50 - 200 pF/m will occur – which is already more
than what the two-wire line exhibits but still immaterial for the typical small lengths in the
guitar interior. But then, there's Gibson. Many of the pickups of this manufacturer sport a
coaxial cable with astounding characteristics. When we measured the 50-cm-long cable of a
P90 pickup for the first time, our spontaneous reaction was: our PM6303 instrument is clearly
broken. The display showed 700 pF in parallel with 500 kΩ at 1 kHz – which is a whole order
of magnitude away from the expected value. However: Philips again proved to be dependable:
the instrument worked flawlessly. The cable capacity was indeed that high (Fig. 9.13).
Typical insulators have a dielectric constant of between 2 and 4 – this could not explain such
a large capacitance. There is however a substance with a high dielectric constant of about 80
that could help to explain what was going on: water! If indeed the fibrous insulating material
is hygroscopic und absorbs water, such a large capacitance could actually result. We tried and
heated the cable to 75° C for 5 hours – and, alas, the (cooled down) capacitance dropped to
160 pF.
Such a "special" cable hits back in several ways: the high capacitance exceeds possibly even
that of the guitar cable and this audibly reduces the resonance of the pickups, plus the high
losses dampen the resonance. These effects are dependent on humidity! in the humid
basement the guitar sounds duller than in dry, heated rooms – and this is due to a cable, not
due to the wood! We would have liked to print here a comment by the manufacturer .... but
those concerned preferred not to reply to an inquiry.
Fig. 9.13: capacitance of a 50 cm long Gibson

pickup cable (pickup disconnected). The two
curves were measured at different points in time.

9.4 Guitar Cables
Guitar pickups are usually connected to the guitar amplifier via a cable several meters long. In
more rare cases an amplifier is already built into the guitar or even into the pickup - wireless
connections are also used. The customary guitar cable determines the sound. Its capacitance
generates - in cooperation with the capacitance and the inductance of the pickup coil - the
pickup resonance which lends a characteristic color to the transmission.
The guitar cable holds an internal conductor (in some cases two). This is constructed as a thin,
flexible stranded wire (sometimes two) outwardly insulated cylindrically e.g. by foamed
polyethylene. Around the insulator there is a concentric braided shielding which sometimes
holds conducting synthetics in addition. For high-quality cables double shielding is
customary. Every differential little piece of cable can be described by four elements: a series
resistance, a series inductance, a parallel capacitance and a parallel resistance.
The series resistance amounts to just a few Ωs - it may with very good approximation be fully
neglected re. the source impedance (kΩ). The series inductance (ca. 1 µH) is so small, as well,
that is will not play any role here. As a rule, the parallel resistance is large (> 100 MΩ) to the
extent that it, too, will have no audible effect. On the other hand, charge displacements and
corresponding very small mechanical deformations will occur in the dielectric (the insulating
synthetic). This will lead to mounting energy losses with increasing frequency. Such effects
cannot be captured with a normal insulation measurement which is normally done with
direct current. For this reason, more elaborate equivalent circuits feature not just one simple
(real) parallel resistor but a complicated RC-array modeling the complex parallel
conductance. On other words: the cable capacitance is frequency dependent to a small degree
and decreases a little with increasing frequency, while the cable losses are strongly frequency
dependent and mount with increasing frequency.
Lossy capacitances are described in a simplified manner by an RC equivalent circuit. In the

lower frequency ranges an RC parallel circuit is employed while in the higher frequency
ranges an RC parallel circuit is used. The energy stored in the capacitor can be recalled,
however the resistor irreversibly converts electrical energy into thermal energy – thus the term
loss. In the complex admittance plane, the admittance real component represents the
conductance due to the loss while the admittance imaginary component is the susceptance due
to the capacitance. Instead of the Cartesian coordinate system with conductance and
susceptance the polar coordinate system with magnitude and phase may also be used. The
magnitude is the admittance while the tangent of the complementary phase angle ∂ is the
dissipation factor d.
d = tan ∂ = 1/R · 1/(ωC) = 1/(ωRC) ∂ = dissipation angle (Fig. 9.14)
For high quality capacitors the parallel resistance R is very large, and consequently the
parallel conductance 1/R very small. The result is a very small value for ∂. Data sheets show
e.g. values of d = tan ∂ ≈ 10-4. Inserting into the above formula a frequency-independent
resistance R and a frequency-independent capacitance C should lead to reciprocal dependency
of tan ∂ on frequency. However, in reality tan ∂ is more or less constant at lower frequencies,
and for many insulators even an increase with frequency is found (see also Chapter 9.2, tone
capacitor).

9.4 Guitar Cables 9-11
The measurements are in clear contradiction with the formula shown above. If frequency
dependent components (the function of which is difficult to understand) are to be avoided, the
only solution is to extend the equivalent circuit to multiple components. Depending on the
desired accuracy a rather large RC array may be required. Fig. 9.14 shows one simple and two
extended equivalent circuits.
Fig. 9.14: Two-pole equivalent circuits of different complexity for a guitar cable. (see also 5.9.2).
Cable capacity and pickup inductivity cooperate to generate a resonance in the frequency
range between 2 and 5 kHz. The cable capacity is an indispensable partner in this resonance
circuit and determines the sound. Cable losses dampen the resonance – however this is
negligible for good cables. As has been discussed already, it is not proper to use the (very
high) insulation resistance for considering the cable loss; rather a loss simulation dependent
on frequency is required. Since the pickup/cable resonant circuit has its highest impedance at
resonance frequency, a dampening resistance in parallel has the biggest effect. High quality
cables will yield loss resistances > 50 MΩ in the frequency range of the resonance. Compared
to other losses and in particular compared to typical potentiometers used in guitars (250 kΩ),
such cable losses are consequently negligible. This does not mean that cable losses are
negligible in general. For radio-frequency transmission other criteria are valid. Guitar cables
however are operated in the audio range – and here only the cable capacitance is of
importance. High quality cables cost a couple of $/m – add some high quality plugs and the
cost can be some $ 20.-. That should be it. "Monstrous" prices are not justifiable from a
physics point of view.
The cable capacity usually is around 100 pF/m (+/- 30%). Normally used cable lengths thus
yield capacitances of 300 - 600 pF. For very long cables this could rise to up to about 1,5 nF.
Special low-capacitance cables go as low as 70 pF/m. For comparison measurements we had
access to a 40 year-old guitar cable (i.e. truly "vintage" ☺). It was 4 m long and sported a
rather remarkable 1050 pF, plus a similarly noteworthy loss resistance of only 500 kΩ.
Compared to the low-capacitance cable mentioned above with 4 x 70 pF = 280 pF there is a
large and clearly audible difference. The effects of the low loss resistance can be (just)
audible for high-impedance guitars, as well. The "vintage" cable is however not typical for
modern cable production.
Next to the above elementary electrical parameters there are some other properties of
importance: shielding effect, mechanical resilience, flexibility, safety against fracture, flexural
strength and low noise performance. It may be surprising that a cable can generate noise.
Bending and straining changes the mechanical tensions in the insulator which can lead to
charge displacements. The latter can manifest themselves as crackling noise (tribo-electric
effect) - for high quality cables this is not audible, though.

The sound of an electric guitar can audibly change when the cable is switched. Unless very
low-quality cables are used, the reasons are solely found in the different cable capacitances.
Relaxation phenomena (orientation polarization, inertia of dipole rotation in the frequency
range f > 1GHz), dispersive signal-propagation or non-linear effects are insignificant in the
audio range. Physics are neither applicable nor competent in the area of esoterism – nor are
psychoacoustics.
Old coil cords could often be stretched to 5 m. The actual cable length was even longer – 8 m
were probably not untypical. Capacitances of about 2,1 nF and loss resistances of 250 kΩ
could be the result. If someone would like to reproduce specifically these old "vintage"
characteristics but is shying away from laying 21 m of modern cable 21 x 100 pF = 2,1 nF)
could solder an additional capacitor to the cable. The effect of the loss resistance can be
reproduced by turning down the tone control to some degree. The final evaluation should be
done via a listening test. To exclude prejudice and bias, a blind test with direct A/B-
comparison is recommended.
A very flexible solution can be obtained by connecting different capacitors via a rotary
switch to a short low-capacitance cable – the resonance frequency is now adjustable. The
connection of a capacitor is indispensable in particular if a usual magnetic pickup is to be
connected to an amplifier (e.g. an on-board pre-amp) without a long cable. The resonance
frequency of the pickup without the loading by the cable capacitance is too high such that the
sound becomes "glassy" or "too sharp". If this sound is not actually desired, a capacitor of 300
- 1000 pF needs to be connected in parallel to the pickup in lieu of the cable.
Fig. 9.15: Low-pass transmission for varying cable capacitances: 1200 pF, 600 pF, 300 pF, 0 pF. The solid lines
show loss-free cables, the dotted line refers to a 500 kΩ loss resistance. For modern high quality cables the loss
resistance of R>50 MΩ in the range of the resonance frequency is certainly negligible.
In Fig. 9.15 we see the influence of the cable capacity on the HUv-transmission function (low-
pass model). Elongating the cable has the effect of a capacitance increase proportional to the
length increase. This reduces the resonance frequency. The resonance peak at the same time
becomes stronger. The figure is meant to exemplify the effect in principle. Additional data are
found where the specific pickups are discussed.

9.4 Guitar Cables 9-13
Fig. 9.16 depicts the loss factors for a number of guitar cables. The five lines in the upper
region of the figure are the results of guitar cables from old production, and of modern
microphone (!) cables. These cables can result in an audible dampening of the resonance. The
cables of the makes Horizon, Straight, Klotz LaGrange and Gibson will clearly not decrease
the resonance peak, nor will the RG58-CU used in measuring and instrumentation (it would
however not be flexible enough as a guitar cable). The rather thin George-L's-cable can be
seen as a borderline case: the dissipation factor should not exceed 2% in the range of the
resonance frequency (2 - 5 kHz).
Microphone cables are generally not suitable as guitar cables. They are usually a two-wire
line and optimized for connection to a differential amplifier input; the issue of low
capacitance is not much considered. The survey measurements revealed microphone cables
sporting a rather sizeable 250 pF/m and dissipation factors of 10%. When operating a
dynamic 200-Ω-microphone one can still get very good results with a 10-m-long-line, but for
a high-impedance guitar such a cable should not be used. This does not imply that
microphone cables are generally unsuitable for electric guitars – there are indeed also very
good microphone cables. The suitability should therefore be checked in the specific situation.
It should be general knowledge that loudspeaker cables are unsuitable for guitars. Most often
speaker cables are constructed as thick stranded cables, and they are not shielded. The pickup
might be susceptible to noise, but at least the cable should be silent.
Fig. 9.16: Dissipation factors of guitar cables. The sample shows just a few examples, there are many more
manufacturers.

It is no surprise that inconceivable nonsense is sometimes found in advertisements – rather

this appears almost to be part of the specific charm of the genre. In contrast, editorial test
reports published regularly in musician's magazines carry a high weight since one assumes an
independent expert author to be behind it. To illustrate here an excerpt from a well-respected
guitar and bass magazine: in an issue in spring 2000, the well-known author notes, in the
framework of a test of instrument cables, with some prudence: "According to the
manufacturer, the Monster-Cable Performer 500 Rock supposedly is distinguished by an
aggressive sound, while the Bass Instrument Cable (Performer 500 Bass) supposedly has a
particular strength in the low frequency range and achieves an extended-dynamic
performance." Supposedly! This wording leaves room for interpretation of the kind we know
to be smart when dealing with the tourist industry: "given that the hotel is located in the
immediate vicinity of the airport, it supposedly is relatively quiet." If you still go there, it's
your own fault. In the next issue of the same magazine the same author writes: "While the
cables for bass and Rock distinguish themselves through emphasizing the cutoff frequencies
and aggressive presence, respectively, the Performer 500 Jazz establishes itself audibly more
succinct in the lower mid frequencies and presents the character-defining timbre-range with
extensive emphasis but remains pleasant and round .... compared to other Monster Cables,
the Studio Pro 1000 appears a tad softer - this can be traced to a particularly balanced
transmission without any emphasis or peculiarities."
So much for diplomatic restraint: here is the opinion of the man carrying out the test. Of
course, he is entitled to it and may publish it. However, he will then have to put up with the
question whether he indeed has any clue at all of the electric function of a cable. "Without any
emphasis"? Does that mean the cable has no capacitance? That would probably not work, and
it is not desirable, either. What actually is the capacitance of these wonder-cables? One could
easily and cost-effectively measure them and publish the result - the reader would take away
much more than he will profit from speculations about cutoff frequencies. In any case the
price of these wondrous cables does not remain in the dark (remember, this is in 2000, and
below we are using a conversion rate of slightly more than 1 $ to the EUR):
- Performer 500 Monster Bass Guitar Cable 6,4 m: ca. $ 70.-

- Performer 500 Monster Rock Guitar Cable 6,4 m: ca. $ 70.-
- Performer 500 Monster Jazz Guitar Cable 6,4 m: ca. $ 85.-
- Studio Pro 1000 6,4 m: ca. $ 180.- No typing error: onehundredandeighty bucks!
It is of course understood that a Jazz-cable will be more expensive than a Rock-cable. If that
weren't the case, the marketing manager should be laid off without notice. The step size is
o.k., as well: one quarter more expensive. You do see it the same way, dear Jazz guitarists,
don't you? But what about the actual level of these prices?? The very high-grade Klotz
LaGrange cost at the time about $30.-. Same length of 6,4 m, and a capacitance of 67 pF/m,
with two Neutrik plugs. And that cable, as well, will not have been sold without profit ....
It may be that the special capacitance of the Monstercables generated a special sound during
the test which led to the mentioned description. Of course nobody will will imply that the
Author may have simply copied the advertisement texts provided by the manufacturer and
then signed with his name. However the special capacitive load could have been achieved at
less expense: for $180.- one could buy 1000 capacitors and as many resistors to go with them.
That would have been sufficient for a whole lot of set-ups to emulate ANY cable, even the
Monster-ones. And a loc-cap cable with two Neutriks would have been thrown in ...
To cite an author/tester from another magazine: "The idea of an instrinsic sound of cables as
propagated by the industry is, in my opinion, a load of total BS." Stated by a well-respected
studio owner and regular author with this other magazine.

9.5 Mounting Plates 9-15
9.5 Mounting plates
Since the magnetic alternating field is not limited to the interior of a pickup, it is possible that
metal parts mounted in the vicinity of the pickup influence the mechano-electric transmission
parameters. Examples for such parts are the rectangular bridge plate of the Telecaster lead
pickup, or pickguards made from metal. The eddy currents induced in these part dampen the
pickup and reduce the inductivity L and the resonance peak. Some Stratocaster pickguards
are entirely made of plastic – no eddy currents can happen here. However, often more or less
thin metal foil or even metal sheets are glued underneath the pickguard for shielding
purposes. The thicker these foils or sheets, the more they dampen the resonance. Particularly
"efficient" in this way are pickguards entirely fabricated from metal (e.g. aluminium). The
dampening effect can be audible in a direct A/B comparison – the range of brilliance that is so
important to the "Fender Sound" is attenuated by about 2 dB. (Fig. 9.17)
Fig. 9.17: Transfer characteristic of a Stratocaster pickup without and with aluminium pickguard
Similarly, the transfer characteristics of the Telecaster bridge pickup will change if a well
conducting bridge plate is mounted (Fig. 9.18). The differences resulting from the comparison
between two bridge plates are however so small that they will normally not be registered. If
that happens nevertheless: a thin slit in the bridge plate effectively prevents the eddy currents
from flowing.
Fig. 9.18: Transfer characteristic of a Telecaster pickup without and with brass bridge plate (Gotoh).

10 Guitar amplifiers
Electric guitars per se radiate only very little sound – to be decently heard, they require
special amplifiers and loudspeakers. Indeed, one is well advised to better regard amplifier and
loudspeaker as an integral part of the musical instrument: in a manner of speaking, the electric
guitar does extend up to the loudspeaker. Guitar amps create distortion and usually feature a
frequency dependent transmission-factor; the attached speakers create distortion, as well, and
do show an uneven frequency-response – and only if, on top of everything, the loudspeaker
enclosure has pronounced resonances will the guitar player “be satisfied”. However, there are
also ugly, buzzy distortions, and not every resonance or kink in the frequency-response will
sound good. We have seen innumerable attempts to improve the primitive circuits of the first
guitar amplifiers – alas, in many cases the circuits may have improved but the sound got
worse. Textbooks on circuit-design teach about avoiding the non-ideal characteristics of
circuits; for example: how negative feedback will reduce the nonlinear distortion of the
power-amplifier. Some famous guitar amps, however, achieve their great sound especially
because indeed they dispense with all negative feedback in the power section – the VOX AC-
30 being a most prominent example. On the other hand, to conclude that to this day science
has failed to understand the functionalities of a tube amplifier – that would be far from the
truth. Indeed, systems-theory, circuit design and instrumentation technology are powerful and
successful areas in electronics … the issue here is the definition of the task at hand. It must
precisely not be the aim to “linearize” the frequency-response (i.e. to render the transmission-
factor frequency-independent), but we need to e.g. follow up the question how dents in the
frequency-response influences the sound. It is this subjectively perceived sound that is
important, not so much the physically measured sound. The question whether 2nd-order
distortion makes an electric guitar sound better than 3rd-order distortion is not included in the
tube manual. Whether such distortion should happen in the pre-amplifier or rather in the
power stage is not discussed, either. It is in particular this interaction of the individual system
parts that renders the circuit-analysis and -design so complex and difficult. The simplifying
description as LTI-system can be only a first step and needs to be followed up by further
steps. Thus, circuit analysis of guitar amps requires significant effort; the following
elaborations therefore confine themselves to a few generic circuits1.
10.1 Preamplifier
In circuit design, the circuit section grouped around an amplifier tube is i.a. designated an
amplifier stage. Typical are: preamplifier, tone-filter, phase-inverter, power amplifier, and
power supply. Each of these partial circuits contributes to the sound of an amplifier, or, rather
to its transfer characteristic. In the preamplifier (or input amplifier), the preamp tube amplifies
the signal by a factor of 20 – 50. During the first two decades of amp-history, large octal tubes
(tubes with an 8-pin socket) were used, but in the mid-1950’s the smaller 9-pin noval tubes
were introduced to the booming amp market. Especially the high-gain 12AX7 (ECC83, 7025)
has established itself as a standard still recognized over half a century later.
1
Translator’s note: In this chapter, measurements taken on a number of typical amplifiers are shown. These
amplifier specimen were for the most part newly built exactly according to the schematics and layout of the
historic originals, i.e. they were in a state as if they had just come off the assembly line (comparable to
„N.O.S.“). In one case (VOX AC-30) an original form the 1960’s was available and used; this amp had been
restored to its original working condition using new components of accurate values where needed. N.B.: the
Fender Super Reverb used for the measurements had an output transformer with both 2-Ω- and 8-Ω-
outputs – this made comparisons with the other amplifiers easier.

10-2 10. Guitar amplifiers
10.1.1 Preamplifier tube
Fig. 10.1.1 juxtaposes the circuit of a triode as it is typically used in input stages of guitar
amplifiers, and an N-channel-JFET circuit. These circuits are not equivalent but the
comparison will assist the solid-state-expert to easily access to the world of tubes. The three
electrodes of the triode are designated cathode, grid, and anode or, more commonly used,
plate. They correspond to source, gate, and drain in the JFET. Unlike with the FET, the tube
requires a heater current (about 0,3 A at 6,3 V) that normally is not shown in the schematic.
Tubes operate using very high supply-voltages (UB = 200 – 400V) i.e. 10 times the value for
the FET. On the other hand, the currents flowing in the plate- and in the drain-circuit are
comparable: for input stages, they amount to about 1 – 2 mA. The voltage between grid and
cathode (gate and source) constitutes the control quantity; for small drive levels, the input
(grid, gate) is of very high impedance – i.e. the grid-(gate)-current through the grid-resistor
(gate-resistor) Rg, is negligible. The cathode-(source)-current therefore is equal to the plate-
(drain)-current. For a cathode-(source)-current of 0,8 mA we find a voltage of 1,2 V across
the cathode-resistor Rk (source-resistor Rs), and consequently the control voltage Ugk (Ugs)
amounts to –1,2V as long as the input voltage Ue remains at zero. For guitar amps, the input
impedance Re is often 1 MΩ, and the series resistor Rg often amounts to 34 kΩ (two 68-kΩ-
resistors in parallel), while the plate-resistor Ra will be between 100 kΩ and 200 kΩ.
Fig. 10.1.1: Input-circuit of a tube amplifier (center) compared to a FET-amplifier (left). The right-hand picture
shows typical characteristics (data-sheet of the double-triode-tube ECC83). The term “control voltage” is used in
various ways – here, the grid/cathode-voltage and the gate/source-voltage is meant.
At the operating point (i.e. without drive signal, Ue = Ug = 0), the control voltage Ugk (Ugs) is
(at e.g. -1,2 V) negative for both circuits. Positive (i.e. less negative, e.g. -1V) control
voltages make both amplifying elements conduct better: plate-(drain)- and cathode-(source)-
currents rise. Without any cathode-(source)-capacitor, the cathode-(source)-voltage would
consequently increase and thus counteract the drive signal, effectively causing negative
feedback and decreasing the gain. Since early guitar amps had to make do with few tubes,
high gain was required and such a negative feedback was uncalled for. Therefore, the
cathode-resistor was bridged via an electrolytic capacitor (typically 25 µF) eliminating any
AC-voltage at the cathode (within the relevant frequency-range): Ck acts as an AC-short. With
increasing input voltage Ue, the plate-current rises, and this increases the voltage across the
load resistor Ra such that – for a constant supply-voltage UB – the output voltage Ua
decreases. An AC-voltage at the input will cause an amplified, opposite-phase AC-output-
voltage shifted by a constant DC-voltage (e.g. 250V – 100kΩ⋅0,8mA = 170V). The
achievable voltage gain depends significantly on the type of tube: the often-used ECC83
allows for an AC-voltage-gain of about -50.

10.1 Preamplifier 10-3
The AC-voltage at the plate is out-of-phase with the grid-AC-voltage – which is why
occasionally we find a “minus”-sign (v = –50) in the gain specification. It would also be
possible to define the voltage gain as the quotient of two RMS-values: now we would always
get a positive gain-factor, e.g. v = +50 (RMS-values are always positive). Still, even for
positive gain specification, the plate- and the grid-ac-voltages will remain out-of-phase – at
least for the common-cathode-configuration (cathode ac-connected to ground) that is
ubiquitous in pre-amplification stages in guitar amplifiers.
The voltage gain actually obtainable with a tube circuit depends on the circuitry, on the
power-supply, and on the individual tube. The open-loop gain (designated µ or u) given in
data books characterizes a very specific operational state (no load at all at the output) that
does not occur for a typical preamp stage. Both the plate resistor (also called load resistor)
connected between plate and supply-voltage, and the input impedance of the subsequent
amplifier stage, reduce the theoretical voltage gain to the level of the real closed-loop gain
(often simply called gain). For the ECC83 (a typical preamplifier tube), an open-loop gain of
µ = 100 (or –100) is given; the closed-loop gain that actually obtainable is smaller: typically,
20 … 50 may be expected. Since tube-data change their characteristics as they age, the gain
does not remain constant over the years. Tube production is sometimes subject to con-
siderable tolerances (electrode material, wiring, cathode coating, etc.), the gain of two off-the-
shelf ECC83 may easily differ by 10% to 20%, and even larger tolerances are not unheard of.
To analytically describe the function of an amplifier tube, simplifications are required.

Typical models for tubes are based on the following idealized modeling laws:
Driving a tube does not require any power (the input impedance is practically infinite); the
tube is a linear and time-invariant system; the upper cutoff frequency is so high that the (low-
pass-limited) signal from the guitar does not receive any additional filtering; “the” tube data
are found in the tables of the data books.
None of these assumptions are, however, applicable to typical guitar amplifiers – at best,
they are merely useful in the framework of a rough orientation. The following chapters are a
short description of those tube-characteristics that are of particular importance for guitar
amplifiers. Included are typical concepts for circuits, as well. Standard text-books [e.g.
Barkhausen, Schröder, RCA-handbook] give supplemental basic knowledge. It is however
vital to consider that, while the classic standard works discuss in detail and to some extent
very theoretically the operational behavior of the tube, they do not mention with a single word
the “abuse” (i.e. the umpteen-fold overdrive) that is regular practice in guitar amps. Modern
text-books concentrate on semi-conductors and special tubes (technical tubes), and are not
helpful in the context of guitar amps. Those books that in fact do discuss the idiosyncrasies of
a tube-powered guitar amp are often kept rather general; they hardly offer any measurement
results and rarely any theoretical calculations. In the worst case, mere assumptions are
circulated as they are now found almost deluge-like on the Internet. “The cathodyne circuit
sounds much tighter than the SEPP or the long-tail because already Leo Fender introduced it
in the 5E6a” N.B.: here, it appears that this circuit sounds tighter (whatever that means) not
because of any technical characteristic but because it is spiritually connected to Leo Fender …

10.1.2 Tube input-impedance
Together with the cable capacitance, the input impedance of a guitar amplifier is connected in
parallel with the source impedance of the guitar pickup. Looked at in a simplified manner, the
amplifier input may be represented by a high-value resistor: for guitar amps about 1 MΩ is
customary. The resulting damping effect on the pickup-signal is small. If, however, the input
impedance of the amp is significantly lower, an audible damping effect does happen that
makes itself felt (or rather heard) as a loss in brilliance. Entirely different scenarios may occur
with effects boxes (e.g. treble booster, distortion device, or wah-wah) connected between
guitar and amp. Their input impedance often is rather low but this needs to be seen as part of
the effect.
Aside from the regular standard input (designated “1” or “Hi”), many classic tube amps offer
a second input of lower sensitivity (“2” or “Low”). Due to the smaller input impedance
(typically 136 kΩ), this second input makes the guitar sound less brilliant. Also, a 50%-
signal-attenuation involving a voltage divider with two 68-kΩ-resistors is included, reducing
down preamplifier distortion. When the standard input (“1”) is used, the two 68-kΩ-
resistors are connected in parallel with each other, and in series with the tube input. They have
the effect of a low-pass filter that however only cuts out high-frequency radio transmissions –
in the audible range, the low-pass effect is insignificant.
The input capacitance of customary guitar amplifiers is small but not always negligible
compared to the cable capacitance. For a tube amplifier, the input capacitance of the preamp-
triode will be around 80 – 150 pF due to the Miller-effect. Depending on the wiring within
the amp, further line capacitances of about 50 pF may need to be added. With the standard
input-circuitry for tube amps, the guitar is galvanically coupled to the grid of the first tube –
there is no coupling capacitor. Only few amps (in particular very old ones) generate the grid
bias via the leakage current of the grid, and therefore separate guitar and tube via a coupling-
capacitor of 10 – 20 nF. The effect of this capacitor is negligible in the framework of the
linear model – the operating point of the tube in this configuration is, however, not very stable
at all.
The tube grid is connected neither to the plate nor to the cathode, and since the glass container
insulates very well, we could indeed surmise a tube input of very high impedance. However,
while plate and cathode are not connected, there is still an electric current flowing between
them. This is due to the glowing cathode emitting electrons that fly – through the vacuum in
the glass container – to the positively charged plate. A flow of electron is an electrical current:
negative charges flowing from the cathode to the plate make (applying the technical current
direction) for a positive current from plate to cathode. The electrons travelling from the
cathode land on the plate and not on the grid because the plate is charged positively relative to
the cathode, and the grid negatively – for the customary operating point, anyway. A
cathode-current of e.g. 0,8 mA (Fig. 10.1.1) flows in the absence of an input signal, and with
this the grid-potential is 1,2 V more negative than the cathode-voltage. However, in the case
that the grid becomes positive relative to the cathode, the electrons find two attractive landing
sites: the highly positive plate and the weakly positive grid. Since the plate-surface is much
larger than the grid-surface, and since the plate-voltage is much higher than the grid-voltage,
most of the electrons will fly to the plate. However, a small part of them does land on the grid
and causes a grid-current. This grid-current exits the grid as a negative electron flow, i.e. it
enters the grid as technical current.

There are several reasons for the flow of this grid-current: finite insulation resistances
grid/plate and grid/cathode, ionization of the remaining gas in the glass container (deficient
vacuum), thermal grid-emission due to high grid-temperature, and the already mentioned
pickup of a part of the electron cloud emitted by the cathode. The individual effects
superimpose (in part with inverse signs) and result in a non-linear input characteristic; the
grid-current depends on the grid/cathode-voltage Ugk in a non-linear fashion. For input
voltages♣ Ue of above about +0,7 V (Ugk > –0,5 V) there will be an observable grid-current
leading to a voltage across the grid-resistor Rg. Consequently, the grid-voltage Ug decreases.
This effect makes itself felt especially for strongly positive input voltages: for example, we
may find only about +1,2 V instead of +4 V at the grid (Fig. 10.1.2).
Fig. 10.1.2: Non-linear correspondence between input voltage Ue, grid-voltage Ug, and grid-current Ig.
Measurements of real tube voltages and tube currents show a hysteresis caused by capacitive
coupling between plate and grid. Within the tube, the grid/plate-capacitance (about 1,6 pF)
has an effect, and external stray-capacitances depending on the build of the circuitry weigh in.
In conjunction with the grid-resistor, a low-pass in the feedback branch is created, i.e. the
plate-voltage is (approximately) differentiated and the result superimposed onto the generator
voltage. Since the plate-voltage is strongly limited for the drive signal shown in the figure
(Chapter 10.1.3), this feedback becomes effective predominantly close to zero. Idealized
characteristics are shown in Fig. 10.1.3: Ugk is the voltage between grid and cathode i.e. the
actual control voltage of the tube. For the example it amounts to about -1,2 V in the operating
point (i.e. without drive signal).
Fig. 10.1.3: Grid-current depending on the input voltage (or on the grid/cathode-voltage).
The right-hand picture shows measurements (----) in addition to the idealized curve.
.♣
Ue between input and ground, Ug between grid and ground, Ugk between grid and cathode.

In order to assess the non-linear behavior, it is of course the magnitude to the actual input
voltage that needs to be considered. If this were no larger than 100 mV, we could ignore the
non-linearity. However, normal magnetic pickups can easily generate voltages in excess of
0,5 V, and even 4 V is not unheard of – therefore the non-linearity merits a discussion. The
distribution function of the occurrence of pickup voltages is shown in Fig. 10.1.4. It may be
nicely described by a Laplace-distribution (just like speech): the larger the amount of the
pickup voltage, the less frequent it occurs♣. “Loud” pickups (Chapter 5.4), heavy strings and a
strong picking attack may generate considerable voltages. The distribution function of this
special example shows that 95% of all voltage values are smaller than 1 V, and 98% are
smaller than 2 V. The relatively low likelihood of crossing these borders must not lead to the
conclusion that the non-linearity may be neglected. Strong amplitudes especially happen with
the plucking of a string (Fig. 10.1.5), and the immediate subsequent attack-process in the
signal is analyzed by the hearing system with particular precision. The two signals shown in
Fig. 10.1.5 do sound differently. However, the amplitude-limited signal – surprisingly – does
not sound more distorted but less trebly than the original signal. The plate-voltage looks
entirely different, again (Chapter 10.1.3), and what always holds is: the isolated portrayal of
an individual non-linearity says little about the output signal of an amplifier.
Fig. 10.1.4: left: Distribution function (cumulative) of the pickup voltage (Strat, SDS-1 in bridge position).
Right: Non-linear correspondence between input voltage Ue and grid-voltage Ug for a sine signal.
Fig. 10.1.5: Voltage-over-time at the terminals of an SDS-1 pickup (left); with limiting similar to a tube (right).
♣
Strictly speaking, the probability density is zero for discrete values of the continuously distributed voltage; to
arrive at a probability (other than zero), integrating over a range is required.

10.1.3 Characteristic curves of the triode
The two circuits shown in Fig. 10.1.1 include a number of similarities; however, this must not
lead to the conclusion that their behavior is equivalent. Already simple standard models
indicate differences: for the tube, the correspondence between plate-current and control
voltage is described via a power function with an exponent of 1,5 while for the FET, the
corresponding exponent is 2. In reality, the characteristic curves of both amplifiers do deviate
from this idealization – but not in the sense that they would become more equal.
Often, simple model-calculations for the triode start from the Child-Langmuir-law♣:
Triode characteristic
In this equation, Ugk designates the voltage between grid and cathode, and Ua designates the
voltage between plate and cathode. K and µ are constants relating to the specific tube while Ia
is the plate-current. As simple as this law is: it is as inappropriate for guitar amplifiers.
Differences relative the real triode already show up in the range of the characteristic curve
that could be seen as reasonably linear; for the overdrive range, the Child-Langmuir-law
utterly fails (it was not put together for this scenario, anyway). Fig. 10.1.6 compares idealized
and real triode characteristics – the differences are significant. In literature (e.g. JAES), we
find several improvements of the above equation that brings it closer to reality (i.e. closer to
the characteristics given in data books), but the resulting complex equations do not only
require two but six or even more modeling parameters. If the latter are optimized to model the
linear and the weakly non-linear drive range, we may still not assume that the extreme
overdrive conditions♥ in guitar amplifiers are also suitably modeled. The following depictions
therefore do not orient themselves according to tube models but are based on actual precision-
instrumentation-measurements taken from amplifier-typical circuits. This included all the
associated uncertainties … whether this exact circuit or this tube-specimen was typical
enough, whether the capacitors had been run-in long enough, whether the moon had already
risen (or set, or was in the correct house) …
Abb. 10.1.6: Tube characteristics. Left: idealized according to Child-Langmuir. Right: data sheet info.
♣
D. Child: Phys. Rev., Vol. 32 (1911), p.498. I. Langmuir: Phys. Rev., Vol. 2 (1913), p.450.
♥
The drive-limit for the linear range may easily be exceeded by a factor of 30.

The plate-current Ia depends on both the control voltage♣ Ugk, and on the plate-voltage Ua
(Fig. 10.1.6). In a real amplifier circuit (Fig. 10.1.1), both of these values change, and
consequently the transmission behavior may not be taken as such from Fig. 10.1.6. Fig. 10.1.7
therefore directly indicates the mapping of the input voltage Ue onto the plate-voltage Ua. In
the left part, the input voltage (multiplied by a factor of -58) is included, as well, in order to
clearly show the effect of the non-linearity: both half-waves experience limiting. The latter is
explained only from the overall transmission behavior, and not merely from the input
characteristic. The right hand part of the figure shows the plate-voltage for input voltages of 1
Veff and 4 Veff respectively.
Fig. 10.1.7: ECC83: non-linear distortion of the plate-voltage; on the left with vertical offset. Ra = 100kΩ.
The figure shows how the negative half-wave is flattened first as the drive level increases; for
strong overdrive, heavy clipping is introduced for the positive half-wave. The plate loading
(5 MΩ-probe) is the reason why the plate-voltage does not fully reach the supply-voltage (250
V). That the minimum voltage is not closer to zero is due to the grid-resistor – it attenuates
positive input voltages on their way to the grid (Fig. 10.1.4) and prohibits full drive of the
tube. Fig. 10.1.8 shows the influence of the grid-resistor: without Rg, larger plate-currents and
smaller plate-voltages are possible – this kind of operation is, however, not typical for input
stages of customary guitar amplifiers, and it will not be investigated further. What does
require consideration is the plate-load that has, in the measurements so far, been very small
(at 5 MΩ). In the classic tube amps (Fender, VOX, Marshall), the input tube often feeds the
tone-control stage that exerts considerable loading onto the plate.
Fig. 10.1.8: transmission characteristic Ue → Ua. For a 33-kΩ-grid-resistor (left); for shorted grid-resistor (right).
Ra = 100 kΩ (plus 5 MΩ load).
♣
May take different meanings; in this case: grid/cathode-voltage.

The input impedance of the tone-control stage is complex, and therefore the analytical
description now begins to become complicated (non-linear and frequency-dependent
behavior). As a first approximation, however, we may replace the input impedance of a
typical tone-control network by the series connection of a 100-kΩ-resistor and a 0,1-µF-
capacitor – this enables us to describe the important effects already pretty well. More precise,
amplifier-specific models would go beyond the scope of these basic considerations. The
cutoff-frequency of the load-two-pole is low enough that the plate is loaded with 100 kΩ in
the steady-state condition. Compared to the situation without load (as it has been looked at so
far), the AC-plate-voltage is reduced by about a third (Fig. 10.1.9). The measured small-
signal-gain, i.e. the gain for small drive levels (e.g. 0,1 V), amounts to -42. In theory, the
small-signal-gain results from the multiplication of the transconductance S (data sheet: S = 1,6
mA/V) with the operational resistance. The latter consists of the parallel-connection of the
internal impedance of the tube (data sheet: 63 kΩ), the plate-resistance (in the present
example 100 kΩ), and the load resistance (again 100 kΩ). We calculate a small-signal-gain of
-45, from this i.e. a reasonable correspondence. What needs to be borne in mind, though: the
data-sheet information may be taken only as a guide number: swapping a tube for another can
easily change the small-signal gain by 3 dB! The drive limits are specific to the respective
tube specimen, as well.
Fig. 10.1.9: Plate-voltage for input voltages of 1Veff and 4Veff (with Rg, and plate loading, left);
transmission characteristic (with grid-resistor Rg and plate loading, right).
The comparison between Figs. 10.1.7 and 10.1.8 has already shown how important the
internal impedance of the signal generator is. Whether the tube grid is driven from a low-
impedance source (Rg = 0), or via a grid-resistor (Rg = 33 kΩ) makes a big difference. Of
course, the serially connected impedance of the signal generator needs to be considered in
addition. Active pickups (e.g. EMG) feature internal impedances similarly low as those of the
generators used for the measurements; however, most guitars have high-impedance passive
pickups. For an exact analysis, the operation with a 50-Ω-generator is therefore not indicative
of the behavior when driven by an electric guitar. The latter may easily show an internal
impedance of 100 kΩ in the range of the pickup resonance (2 – 5 kHz). Since the internal
impedance of the electric guitar is frequency dependent (e.g. 6 kΩ at low frequencies and 100
– 200 kΩ at resonance), and since the input impedance of the tube is non-linear, complicated
interactions between the different systems occur already in the input stage of a tube amp.
Such an amp will make the guitar see an entirely different load compared to a “modeling
amp”. In the latter, the guitar-signal will be normally fed - via a high-impedance OP-amp-
stage - to the AD-converter, and all signal processing will be taken care of in the digital
realm. However, which tube characteristics will in the end lead to audible differences can
only be investigated via listening-experiments.

For two different source impedances, the grid-voltage limiting is shown in Fig. 10.1.10. It is
evident that even relatively small positive voltages are visibly reduced. The internal
impedance of the generator is, however, purely ohmic in this measurement – which does not
correspond to the situation for a connected electric guitar fitted with passive pickups. To
better simulate this operational condition, a small transmitter coil was laid on top of the
pickup of a Stratocaster (with original wiring). Driving this coil with a power amplifier
generated a magnetic AC-field inducing a sinusoidal voltage into the pickup. The internal
impedance of this arrangement therefore realistically corresponded to the actual operation.
Fig. 10.1.10: Limiting of the generator voltage (Ue) in

the grid circuit. Internal generator impedance = 0 (–––),
and 100 kΩ (----), plus grid-resistor Rg = 33 kΩ. Two
periods with different voltage-amplitude shown (1V and
0,4V).
In its left-hand section, Fig. 10.1.11 shows the corresponding measurement results. The
dashed line relates to the source voltage of the guitar corresponding to the open-loop voltage
generated by the unloaded guitar. With the load of the tube amplifier, the guitar voltage is
bent out of shape; however, this does not happen such that the positive half-wave would
simply be compressed (as it would be the case for an ohmic source impedance). Rather, the
complex guitar impedance leads to phase shifts between the spectral distortion components
(especially in the 1st and 2nd harmonic), and thus the voltage curve is also changed for the
negative half-wave. The grid-voltage changes correspondingly (right hand section of the
figure), and in the plate-voltage the duty factor is shifted (compare with Fig. 10.1.9). These
measurements show that already the first interface between guitar and amplifier-tube has an
effect on the signal. Precise observation indicates that the tube input is not of ideally high
impedance but acts as a non-linear load-resistance already at moderate voltages. Whether the
corresponding changes in the signal are audible compared to other non-linearities, is another
question and can, however, be determined only for the individual case.
Fig. 10.1.11: Mapping of the guitar-source-voltage ULL onto the terminal voltage of the guitar (left). Load
resistance for the guitar is the input-circuit of the tube, Rg = 33 kΩ, f = 2kHz. On the right, the corresponding
grid- and plate-voltages are shown; the plate is loaded as given in Fig. 10.1.9.

It is almost impossible to describe the transmission behavior of a guitar amplifier in its

entirety by formulae and diagrams. This is not because the relations and connections would be
unknown, but rather because too many dependencies would have to be defined. While the
small-signal behavior can easily by specified via the frequency-response, there is – strictly
speaking – not even a transfer-function for the large-signal operation because this function is
only defined via the LTI-(linear time-invariant)-model. Mixtures of small-signal frequency-
response and harmonic-distortion characteristic are either incomplete or too extensive. Non-
linear distortion is dependent on frequency and on level, and thus is a bi-variant quantity.
There are, in fact, many bi-variant quantities: 2nd-order (k2) and 3rd-order (k3) harmonic
distortion, as well as 2nd-order and 3rd-order difference-tone-distortion, just to name the most
important ones. For the – frequently occurring – strong-overdrive condition it is not adequate
in the least to assess distortion up to merely the 3rd-order; rather, it would be necessary to
determine a multitude of individual harmonic-distortion- and difference-tone-factors, and
represent this as a function of two variables. And even if we would make such an effort: the
result would be all but impossible to interpret. For example, how would we evaluate a circuit
change that results in a reduction of the 3rd-order harmonic distortion at 0,5 V and 1 kHz,
while the 2nd-order harmonic distortion increases at 0,8 V and 2 kHz? While at the same time
the 4th-order harmonic distortion at 0,8 V (2 kHz) drops strongly but the 2nd-order difference-
tone-distortion generally grows stronger? Is this desirable or counter-productive? General
judgments such as: for tubes, 3rd-order distortion (= good) dominates, for transistors 2nd-
order (=bad) does are far too unsophisticated, but unfortunately they keep getting copied
again and again from textbook to textbook. Listening experiments remain indispensable. Still,
a few fundamental relations can be taken from the theoretical models, after all – even if it the
result is not much more than the insight that the circuit layout (than cannot be derived from
the schematic) can be highly important, or that tube data have a considerable scatter range. In
the following analyses, we will give some data on harmonic distortion for a tube driven via an
ohmic source impedance, all the while remaining fully aware that only part of the topic can be
covered this way, and that additional research would be highly desirable.
10.1.4 Non-linearity, harmonic-distortion factor
Here is a simple example regarding the topic of non-linearity: an amplifier generates – at an

input voltage of 1 V – pure 2nd-order harmonic distortion with k2 = 5%. Let’s set its gain
factor to v = 1. Now, a second amplifier (also with v = 1) also generating k2 = 5% at 1 V is
connected in series with the first one in a non-reactive fashion. How big would the harmonic
distortion of the overall system be?
Would that be: k2 = 10%, or 7%, or an unchanged 5%?
It is not even possible to answer this question without supplementary data: we do not know
the phase of the distortion. In case the 2nd-order distortion is generated in-phase in both
amplifiers, k2 is doubled, but if it happens to be in the opposite phase, the 2nd-order distortions
all but cancel themselves out. In both cases an additional 3rd-order distortion appears at k3 =
0,5%. If there is a random phase-shift between the two amps, k2 can assume any value
between 0 and 10%. Already this simple example shows that it is very difficult to derive any
statements about the distortion of the overall system from the non-linear behavior of the
single amplifier stages.
So, are you having fun yet, dear audio-engineers? O.K. then – let’s go for a second example:
now both amplifier stages feature pure 3rd-order distortion at k3 = 5%. Right … use the above:
the series connection results in k3 = 10% for the in-phase condition, and for the out-of-phase
condition in k3 = 0%; plus additionally k4. Hm … are you sure? Then do turn the page!
For the pure 3rd-order distortion, the overall system does not distort with k3 = 10%, but with k3
= 12,3%, and k5 = 1% is generated in addition, rather than k4. Given the anti-phase-condition,
we do not see a cancellation but k3 = 7,5%! Even examples as simple as these show that the
results of connections of non-linear systems are rarely understood based on intuition.
Moreover, tubes do not exhibit pure 2nd-order distortion or pure 3rd-order distortion; there will
also be distortion of higher order, and in addition the signal will be subject to filter stages –
with the result being a highly complex signal processing despite the relatively simple
circuitry.
Often, modeling a non-linear circuit starts with the simplification that the system is memory-
free. With the investigated system not including any signal memory, the output signal
exclusively depends on the input signal at the same instant – with the dependency between
both signals described by the transmission characteristic (not the transmission function!).
This transmission characteristic y(x) is curved (compare to Fig. 10.1.8), but it is time-invariant
and excludes any hysteresis. The characteristic may be expanded into a power-series
(Taylor/MacLaurin) around the operating point – the smaller the drive levels, the more
precisely this works. Put in another way: the more the amp is driven, the less the power series
is appropriate. This is easily understood: a limiting characteristic has two horizontal
asymptotes, which is incompatible with a power-series converging towards infinity. In this
situation, wouldn’t the arctan-function seem to be a much better starting point? Yes, indeed –
but it would be one that is far from intuitively accessible: how is e.g. mapped
again onto y? With . O.k., I see. So how does the harmonic distortion
depend on the drive levels? Well, we would have to develop a series-expansion of the arctan
… Phewww – that means we might as well expand the transmission characteristic into a
series:
Series expansion of the transmission characteristic
In this expansion, a0 is the DC-offset that is separated in most circuits by the coupling capa-
citors; we ignore this offset. a1 is the gain – for an input tube this might be e.g. -54. Now we
get to the non-linearity: using, for example, the pure 2nd-order distortion (i.e. for i > 2),
we obtain
Due to the non-linearity, the DC-component has changed but we can again ignore it. There is
now a new spectral line at twice the frequency. The ratio of the RMS-values is
nd nd
designated the 2 -order harmonic distortion k2. is the RMS-value of the 2 -order
harmonic (at 2ω) and is the RMS-value of y. Let us set, as an example, a1 = 1 and a2 = 0.1
– this yields . Connecting two such systems in series, a series-
transformation z(y(x)) is the result:
Assuming again x = sin(ωt), the amplitudes (or rather the RMS-values) of the individual
harmonics can be calculated. What is striking in view of the second bracket of the equation is
that the offset (a0) now not only influences the DC-component but the 1st and 2nd order
harmonic, as well! Moreover, we notice the generation of a 4th-order harmonic due to x4 –
although its amplitude is so small that it may be disregarded. From x2, a DC-component and
the 2nd-order harmonic result, and from x3 we derive the 1st-order and the 3rd-order harmonics.
x4 generates a DC-component plus the 2nd-order and 4th-order harmonics. So, everything
depends on everything else, more or less.

In summary: for a0 = 0, the levels of the 1st, 2nd, and 3rd harmonic depend on a1 and a2, and
only the 4th harmonic depends solely on a2. A simplification yields:
; for a0 = 0 and a1 = 1.
Neglecting higher-order effects, we can state that the series connection indeed leads to double
the 2nd-order harmonic distortion. Also, if one system is set to be and the
other is set to be , k2 can be approximately fully compensated. The
simplifications we introduce here are purposeful – they do not generate any large errors.
For the purely 3rd-order distortion, the mapping is: The offset a0 is set to zero
in this case, and the linear term is set to 1 (a1 = 1) as a simplification. The series connection
yields:
Again, we could disregard all higher-order terms and assume that the 3rd-order harmonic
distortion will be doubled. However, the 1st and the 2nd harmonic are dependent on all
summands, and the resulting effect is not at all that minor:
,
The summation of all terms of the expansion has the effect that the RMS-value of the 3rd
harmonic is not only doubled but rises by a factor of 3,7. At the same time, the RMS-value of
the overall signal increases by half, yielding k3 = 12,3%. If the sign of a3 in one of the two
systems is inverted, x3 can be reduced to zero but the remaining members of the series deliver
a significant contribution to the 3rd harmonic. The amplitude of the latter therefore does not go
down to zero but decreases merely to 7,5%.
If the offset (a0) is not set to zero, the situation becomes even more complicated. The same
happens if we do not keep the limitation on purely 2nd- and 3rd-order distortion, respectively.
So: with two nonlinear systems connected in series, both generate 2nd- as well as 3rd-order
distortion.
O.k. – computing this is not impossible; we could multiply that out. The added value would
not be that big, however. It is already clear now that the RMS-value of each harmonic will be
dependent on many coefficients. Also, there will be cancellations of components if there are
opposing algebraic signs. These cancellations will be drive-level-dependent, though – or at
least there will be drive-level-dependent maxima and minima. The individual harmonic
distortion components will not simply experience a monotonous increase with rising drive-
levels but can pass through complicated curves. The individual system may generate
exclusively even-order distortion (k2, k4, ...), but the series connection of two such systems
may still show a predominant odd-order distortion. If we now consider that even simple guitar
amplifiers do not contain one but four tube-stages, and that each tube introduces distortion
both at its input and at its output, and that moreover tone-filters will change amplitudes and
phases … this is where we start to catch a glimpse of how complex a guitar amp in fact is.

The following figures present the dependency of individual harmonic distortion components
on the input signal level. N.B.: Another way of quantifying distortion is expressing how much
lower the level of the distortion is compared to the original (undistorted) signal. This
approach yields the so-call harmonic distortion attenuation aki calculated from the harmonic
distortion factor ki (as we have been using it so far) as:
for e.g. .
All measurements were done with a regularly heated ECC83 with the intrinsic distortion
factor of the analyzer being negligible (k < 0,001%, CORTEX CF-100). The cathode of the
tube was connected to ground via 1,5 kΩ // 25 µF (Fig. 10.1.1), and the plate to UB via 100
kΩ. To model the load, the plate was additionally connected to ground via a 0,33-µF-100-kΩ-
series-circuit. The signal was fed to the grid from a low-impedance generator (CORTEX CF-
90) via the grid-resistor Rg. For one row of measurements, Rg was 33 kΩ (corresponding to a
classic tube amp scenario that is fed from a low impedance source), and for the other row it
was Rg = 133 kΩ (corresponding the additional source impedance of 100 kΩ as it can be
present if a guitar with a passive pickup is operated around its resonance frequency, compare
to Fig. 10.1.10). The supply-voltage UB amounted to 200 V and 250 V, respectively, i.e.
typical settings for input stages (Fig. 10.1.12).
Abb. 10.1.12: Distortion attenuation as function of the generator level, Rg and UB vary. 0 dBV 1Veff.
These graphs are reserved for the printed version of this book.
While the ak3–curve maintains its shape and predominantly experiences “merely” a shift, the
minima and maxima of the 2nd-order distortion change rather drastically. So will you tell me
how that sounds, already? would be an obvious question … however: nobody actually listens
to the plate-voltage of the preamp-tube, and therefore the sound of that signal is irrelevant.
Highly relevant would be how the differences mentioned above affect the loudspeaker
voltage, but this would require the consideration of a myriad of additional parameters and go
beyond the constraints give here. Unfortunately.

Another question relates to the tube: RCA, Tungsram, Telefunken, Chinese, Russian, NOS,
little/much used, and whatever other difference there might be? Simple answer: the tube came
out of the box that served here as container for tubes since 1965, and was re-stocked many
times since. An ECC83 cost DM 7,50 (about € 3,25) in Germany in 1965; today is offered for
€ 6.-. It could also set you back € 25, or even more, though. Without a doubt, tubes of the
same type can differ a lot – the label “ECC83” does not indicate any special sound. Selection
processes performed by the supplier may be helpful but do not have to be. Pricy tubes are not
necessarily in principle better than cheap ones; in particular, “NOS” (i.e. the tube that has
spent 50 years on the shelf without being touched) does not guarantee a “super-sound”.
In Fig. 10.1.13 we see differences that can occur when we change tubes (all measurements
taken with ECC83s). A tube was simply unplugged and another was plugged in, instead. It is
intentional that manufacturers are not identified here, since we do not have a representative
sample. We did not investigate whether an old 80-$-NOS-tube delivers similar or entirely
different curves – confronted with its measurements, it might have experiences a kind of final
deadly shock. Plus, strictly off the record: for the analyst, this is somewhat like the situation
experienced by Galileo’s colleagues who did not even want to look through the telescope to
see Jupiter’s moons – some of us in fact don’t really want to know.
Fig. 10.1.13: Differences in harmonic distortion attenuation caused by swapping tubes. UB = 250V, Rg = 33kΩ.
Now back for the record: already at an input voltage of 300 mVeff, the harmonic distortion in
the input stage of a guitar amp can reach 3%. For small input voltages, 2nd-order distortion is
predominant while from 0,25...1 V, 3rd-order distortion dominates. The location of the border
between the two distortion types depends on the grid-resistor, on the supply-voltage, and on
the ECC83-specimen. The distortion is not inherently unwelcome but rather typical for a
guitar amp of this construction.

So far, we have varied, as parameters, the grid-resistor, the supply-voltage and the tube itself.
As the supply-voltage changes, the plate-current, the plate-voltage, and the grid/cathode-
voltage change, as well. Of course, more parameters vary – but right now we look only at
these three. For example: increasing the supply-voltage from 200 V to 250 V increases the
plate-voltage from 131 V to 165 V, and Ugk decreases from -0,97 V to -1,23 V. Another
method to vary the anode current is the so-called “cathode clamp”: here, the cathode-voltage
is imprinted (i.e. kept constant) using a separate power supply. One could think that the
cathode-voltage could not change anyway due to the capacitor connected in parallel – but in
fact, it can: a 2nd-order-distorted sine tone will generate a DC-component (f = 0) that shifts the
operating point. The following figures show the effects of a relatively small change in the
grid/cathode-voltage on the distortion /(Fig. 10.1.14).
Abb. 10.1.14: Harmonic distortion-attenuation dependent on the input level with varying grid/cathode-voltage.
These figures are reserved for the printed version of this book.
It is clearly visible that even apparently minor changes in the operating point have
considerable effects on the non-linear distortion. For reasons of clarity, no higher-order
distortion products are included in the figures; it can be stated, however, that they are highly
similar. The operating point of the tube is far from fixed but drifts while the amp is being
played. One cause for this is found in the non-linearities already mentioned, and another lies
with the time-variant supply-voltage. The latter depends on the plate-currents of the power
stage and the internal impedance of the power supply and will change depending on the
output power of the amp at the given moment (see Chapter 10.1.6). For the Fender Deluxe we
investigated, this variation was as much as between 210 and 247 V, after all . . .

Now, what is so special in a tube amplifier compared to other amps? Looking at the
preamplifier, there are differences in particular in the non-linear behavior. There are, in
addition, compressor effects and linear filtering – this will be elaborated upon a bit later. The
operational amplifier (OP) appears to be a modern alternative to the tube. It has an
operational range to above 1 MHz and its harmonic distortion may be reduced to 0,001%.
These are, however, all properties that a guitar amplifier should not actually have! An OP
may only be considered as an alternative if additional circuitry simulates the non-linear
behavior of the tube. That this is not entirely trivial was shown in the preceding paragraphs.
Fig. 10.1.15 depicts the drive-level-dependent increase of the distortion for the ideal OP in
comparison with to the tube. The hard amplitude limiting (“clipping”) leads to a steep
distortion increase that is atypical for a tube.
Fig. 10.1.15: Distortion for tube-typical limiting (left) and hard OP-clipping (right). By the way: the designation
“ideal OP” does not imply that the OP would be ideal for playing guitar through it. NB: The OP-offset was
adjusted for an asymmetry similar to a tube.

The rise of the distortion will be considerably flatter if the signal limiting is realized not by
the OP itself but by two silicon diodes (1N4148) in an anti-parallel connection (Fig. 10.1.16).
If the two measured diodes were perfectly identical, only odd-numbered distortion products
would appear; due to small production-related differences, we also obtain even-numbered
distortion products in this example. The 3rd-order distortion of this diode circuit already shows
strong similarities to the triode circuits measured in Fig. 10.1.15 but the 2nd-order distortion is
not reproduced yet. It is not very demanding to design – using a combination of germanium
and silicon diodes – a non-linear two-pole the distortion behavior of which sounds similar to
that of a tube. The exact reproduction of tube-distortion is not even required for this; an
approximate modeling suffices.
Fig. 10.1.16: Signal limiting using two anti-parallel silicon diodes (1N4148) fed from a 20-kΩ-resistor. The level
reference on the abscissa of the right-hand picture is chosen to match the representation in Fig. 10.1.15.
It is not only the harmonic distortion that is different in tube and OP, but the compression is,
as well (Fig. 10.1.17). This difference is not big, but may be compared to the so-called
“sagging” – a modulation caused by the power supply (Chapter 10.1.6). In the attack phase of
a tone, a tube amp may lend that extra little bit of power that can be decisive when competing
with other instruments. That tube amplifiers can be louder than transistor amps rated at the
same power is due in particular to the higher output impedance, but may also have to do with
the weaker compression (= increased dynamics). Of course, this is not generated by the
preamp-tube alone but by the overall circuit.
Fig. 10.1.17: Signal-limiting in a tube and an ideal OP. Equal small-signal gain.

We will dedicate a later chapter to the power-delivery, but at this point a fundamental aspect
of the connection of non-linear systems may already be briefly introduced: the transformation
caused by the individual systems are generally not commutative, i.e. the individual systems
cannot be simply interchanged in their sequence. For this reason, it is not possible to replace
an amplifier consisting of a plurality of non-linear and linear systems by a single non-linear
stage and a single filter-stage. Special consideration needs to be given to the fact that already
the coupling capacitor that taps the signal from the plate is such a filter-system, even if the
associated cutoff frequency of this high-pass is very low.
⇑ ⇓ via coupling-capacitor
Fig. 10.1.18: Half-wave limiting in a sequence of systems; top: w/out coupling cap, bottom: with coupling cap.
Fig. 10.1.18 shows an example: a two-tone signal first passes a stage limiting the positive
half-wave and then a second stage limiting the negative half wave. If these two stages directly
follow each other, the result is a signal limiting on both sides as depicted in the upper row of
pictures. However, if a coupling capacitor is connected in between the two limiting stages,
we obtain an entirely different output signal (lower right picture). With the coupling capacitor
connected ahead of the first limiter stage, it would have no effect because the two-tone signal
is already without any DC-component. The same result would be obtained with the capacitor
positioned after the two limiting stages. However, connected between the two stages, the
capacitor will change the signal even if the cutoff frequency is far below the two frequencies
contained in the two-tone signal. Now, let’s put this in the context of a guitar amplifier: since
the plate-voltage is 150 – 200 V even without any drive signal, a coupling capacitor is
required to split off the AC signal. Together with the input impedance of the subsequent
stage, this capacitor forms a high-pass. In many circuits, its cutoff frequency is so low that it
does not seem to have an effect. For example, in the Fender Bassman (held in highest regard
also by guitarists), we find fg = 3Hz (50nF/1MΩ) which is way below any normal frequency
found in the guitar. However, Fig. 10.1.18 shows that this coupling cap has an effect despite
its low cutoff frequency: the non-linearity will generate extremely low frequencies (0 Hz if
you wait long enough …) that are split off by the high-pass. Taken by themselves, these low
frequencies would be inaudible. However, they do determine the position of the operating
point and therefore influence the distortion of the subsequent stage. The specific value of the
cutoff frequency also has a significance because it determines how fast the transient processes
run (Chapter 10.1.6). This example very clearly shows that design rules valid for linear
operation can lose their relevance in an overdrive scenario.

In our second example (Fig. 10.1.19), a signal of two sine signals close in frequency and
beating against each other is first distorted with a tube characteristic, and then subject to hard
clipping. The scaling of the ordinate is chosen for all pictures such that equal amplitudes
result for small-signal operation. Given such an extreme clipping one could surmise that the
“soft” tube distortion occurring first would not have any actual effect since subsequently we
have clipping, anyway. As long as there is no coupling capacitor between the two distortion
stages, this assumption is indeed correct. However, as a coupling capacitor is introduced, the
signal changes – in particular in the low-frequency region and in the area of the summation
frequency of the two sinuses (in this example around the 31st harmonic).
Fig. 10.1.19: Beat-signal (1st column).

Distorted with tube characteristic (2nd column),
then subject to hard clipping (3rd column).
The spectrum given in the 3rd line shows
the effect of the coupling capacitor.
To round off this section, let us bring the “round” tube distortions face to face with the typical
OP-clipping. If we trust literature, then the latter is the reason for the “harsh” transistor-sound
– as opposed to the soft tube sound. Sure, there are differences in the spectrum (Fig. 10.1.20),
but in fact we also find similarities. In any case, the visual impression (“a round signal shape
will sound more round, as well”) should not be overrated; tube- and transistor-amps differ in
much more than just the rounding of the signal-shapes. Only the connection of several
systems makes for the amp. Or, rather, for the sound …
Fig. 10.1.20: Tube distortion (ECC83) compared to hard OP-clipping, driven by a sinusoidal signal; the distortion
levels below 60 dB correspond to a harmonic distortion of < 1% in this example.

Special consideration needs to be given to slew-rate-limiting since such non-linearity does not
occur in tube amplifiers. The slew-rate SR is the speed of change in the signal i.e. the
derivative dU/dt, usually given in V/µs. For a sinusoidal signal, the maximum slew-rate is
present at the zero-crossing: . A voltage amplitude of 13V (a typical OP
amplitude) results in a slew-rate of just about 1 V/µs at f = 12 kHz. If the maximum slew-rate
that the amplifier can provide is smaller than the signal slew-rate, then non-linear distortion
results. In contrast to a low-pass (the linear transformation of which can alternatively be
specified by a cutoff-frequency and a time-constant), the slew-rate-limiting is a non-linear
transformation that changes the signal shape in particular close to the zero-crossing (Fig.
10.1.21).
Fig. 10.1.21: Sine-functions of different amplitude (–––), non-linear transformation w/slew-rate-limiting (–––).
Although in principle the slew-rate may be limited for rising signals to a different value
compared to falling signals, both values are almost equal for most operational amplifiers: for
example for the (outdated) LM-741: SRmax = 0,5 V/µs, or for the TL-071: SRmax = 13 V/µs.
With a SRmax = 0,5 V/µs, the maximum frequency for distortion-free, full-drive-level
operation is only 6 kHz. One could assume that this would suffice for a guitar amp since most
magnetic pickups limit their spectrum at the most at this frequency. However, this assumption
overlooks the possibility of overdrive: if this 6-kHz-tone overdrives the OP by a factor of 10,
then the signal-slew-rate is also 10 times as quick at 5 V/µs. Fig. 10.1.22 shows that slew-rate
limiting and clipping are two different kinds of non-linearity: clipping limits the too-large
values of the signal while slew-rate limiting confines the value of the slope of the signal. If
both types of distortion happen in one and the same stage, the sequence needs to be
considered: the two transformations are not commutative!
Fig. 10.1.22: Sinusoidal signal (–––), slew-rate limiting (––– left). Sinusoidal signal with clipping (middle).
Sinusoidal signal with slew-rate limiting and subsequent clipping (right).
The principles of circuit design are the reason that we get slew-rate limiting with an OP but
not with a tube (in a comparable manner, anyway). In the OP a number of subsequent stages
generate a very high amplification (e.g. 100.000) that is then reigned in by negative feedback
to e.g. 50. The same gain is accomplished in tube amps in a single stage without or with very
little negative feedback. The high gain of the OP forces another difference: in order for the
overall feedback-loop to remain stable, a low-pass characteristic (e.g. with a cutoff at about
100 Hz) is required in the forward branch (i.e. in the pure OP without the feedback network).

Working in the framework of a model, we may replace the typical OP by the building blocks
shown in Fig. 10.1.23: a comparator (subtractor) is followed by a first amplifier, a first
limiter, a 1st-order low-pass (with e.g. the aforementioned 100 Hz cutoff), a second amplifier,
a second limiter. A (negative) feedback branch connects the output to the other input of the
subtractor). The DC-gain is e.g. 100.000; the gain drops off with 1/f from 100 Hz, reaching
the value of 1 at 10 MHz (the so-called transit frequency).
Fig. 10.1.23: Block-diagram of a typical

operational amplifier with negative feedback.
Limiting can occur at two places in this amplifier with different effects on the external world:
limiting in the output stage will introduce clipping, while limiting occurring ahead of the low-
pass will introduce slew-rate distortion. We may see the low-pass approximately as an
integrator – this will give us an easily understandable model for the limiting of signal rise-
times. If the amplifier is driven with a low-frequency signal, the output stage will limit first
and we get clipping. With a high-frequency signal, the stage ahead of the low-pass will limit
first and slew-rate-type limiting happens.
It really gets interesting for a mixture of tones, e.g. with a two-tone signal consisting of a 1st
and a 2nd harmonic (Fig. 10.1.24). This signal has the same peak value both on the negative
and the positive side and would be symmetrically limited given a point-symmetric limiter-
characteristic. However, since the zero-crossings have slopes of different steepness, the slew-
rate distortion has a different effect on the two half-waves, resulting in a shifting of the signal:
it moves away from the zero-line and becomes asymmetric. In the OP, the negative feedback
would immediately become active (the loop gain is indeed very high at low frequencies), and
a counteractive offset voltage would result, with the slew-rate limited signal losing its quality
of being DC-free, and experiencing a shift towards the negative (middle picture). Now the
clipping is added that predominantly limits the negative half-wave – despite the fact that the
original two-tone signal is in fact symmetric with regard to the horizontal axis. The
processing of the second signal – a superposition of 1st and 3rd harmonic (right-hand picture) –
is just as interesting: the slew-rate distortion does not only reduce the signal but distorts it in a
non-linear fashion. Still, dents remain at the location of the extreme signal values – in contrast
to the effect of pure clipping. These examples show that the slew-rate distortion occurring in
OP-circuits has a very different effect compared to pure clipping process that is often seen as
the sole reason for distortion. In the typical operational amplifier, slew-rate distortion does,
however, not appear by itself but always in combination with clipping.
Fig. 10.1.24: Slew-rate limiting: left and middle: 1st and 2nd harmonic. Right: 1st and 3rd harmonic.
Two-tone signal (–––), slew-rate-limited signal (–––).

The maximum slew-rate that an OP can handle may vary dramatically – depending on the
OP-type: in this respect the old µA709 (one of the first universally usable operational
amplifiers) was particularly inept. Its maximum slew-rate was (at 0,25 V/µs) so small that full
drive was only possible up to 3 kHz at most. Since at the time of the introduction of this OP
(1966), harmonic-distortion measurements were usually carried out only at 1 kHz, the slew-
rate distortion often remained undiscovered. The µA741 introduced two years later was able
to deal with double the slew-rate but that was still not enough: 10-fold overdrive with a 3-
kHz-signal requires 2,5 V/µs. Only later OP-amps – such as the TL071 at 13 V/µs – reach
faster regions. By the way: what is in fact the typical slew-rate of the voltage generated by a
magnetic pickup? Of course, this depends on many parameters; in Fig. 10.1.5, for example,
0,06 V/µs is reached. Feeding this signal to a Music-Man♣ guitar amplifier, it will be
amplified 20-fold in the first OP. To avoid any slew-rate distortion, the OP would have to be
able to take on 1,2 V/µs. However, the LM1458 used in some Music-Man amps cannot go
beyond 0,5 V/µs without distortion (just like the LM307H used as an alternative). Not all
Music-Man amplifiers used these slow LM1458 or LM307H in their input-circuits: some
work with the fast TL071 (13 V/µs) … but then feed the signal to a LM1458 in the third
amplification stage. Worse: for the input-OP, the gain in the treble range can be increased
from 20 to 120 via the “Bright”-switch, increasing the necessary slew-rate value by another
factor of six. The distortion generated by this is therefore tube-untypical. That the Music-Man
amp has a tube power amp ahead of the loudspeaker will therefore not guarantee the same
sound compared to an amplifier working exclusively with tubes in its signal path.
Of course, tubes are not infinitely fast, either; however in most cases in tube circuits the rise-
time is already limited in the grid-circuit via a low-pass. While this low-pass is non-linear due
to its (Miller-) capacitance depending on the voltage-gain of the tube, this non-linearity has an
entirely different effect compared to slew-rate limiting.
The following table lists the slew-rate values for some operational amplifiers. Depending on
the manufacturer, the numbers differ somewhat: for the LF356, for example, we find both 10
V/µs and 13 V/µs. The first letters in the designation may indicate the manufacturer (e.g. LM
741, or SG 741, or µA 741), while the last letters specify housing types, or temperature
ranges, or amplifiers with selected data (e.g. LM 307 and LM 307H). These supplementary
letters are, however, not standardized but specific to the respective manufacturer. For some
types, the open-loop gain (and thus the slew-rate) can be changed via an externally connected
capacitor (so-called compensation, e.g. in the LM 301A).
35 V/µs: HA 5147, OPA 404,

13 V/µs: TL 071, LF 351, LF 353, LF 356,
10 V/µs: LM 302, LM 301A (uncompensated),
6 V/µs: NE 5534, LF 355,
0,5 V/µs: LM 107, LM 207, LM 307, LM 741, µA 748, RC 1458, RM 1558,
0,2 V/µs: OP 07,
0,1 V/µs: LM 108, LM 208, LM 308 (each compensated),
Table: Slew-rates of some selected operational amplifiers (guide values).
♣
Amplifiers and instruments, founded in 1972 by Leo Fender (and Tom Walker), sold to Ernie Ball in 1980.

10.1.5 Frequency limits
The frequency limits of the spectrum of an electric guitar are located at about 82 Hz and about
5 kHz; a guitar amplifier does not have to reproduce any lower or higher frequencies. That is
a common and not entirely wrong opinion. The open low E-string vibrates with a fundamental
frequency of 82.4 Hz, and the spectrum is limited towards the higher frequencies by the
pickup-resonance often located between 2 and 5 kHz. However: bandwidth of the electric
guitar and bandwidth of the amplifier are two different things. It is not necessary in a first
modeling step to look into the issue that a guitar generates time-limited sounds and therefore
the associated spectrum cannot even become zero below 82 Hz. What does require in-depth
consideration is the fact that an amplifier with a non-linear distortion-characteristic will
generate difference-tones of a frequency far below 82 Hz. Using operational amplifiers (OPs)
it would be possible to DC-couple the output of each amplifier stage with the input of the
following stage and thus transmit any desired low frequency (down to 0 Hz if we wait long
enough …). Such an arrangement is sometimes seen as ideal for recording studio technology
because there will be neither phase- nor amplitude errors in the low-frequency region.
However, as already previously mentioned: the guitar amplifier is a part of the instrument, it
is supposed to generate lots of errors. “Errors” from the point of view of classic circuit design,
that is, which in the present context are better termed with “signal alterations”. The latter
should be of the right kind, i.e. those that sound good – and only those. What sounds good or
bad is of course a matter of subjective judgment. If a guitarist wants to hear low-frequency
difference-tones, amp and speaker need to reproduce these. This feature is, however, not the
norm, because the resulting sound will be assessed by many players as “undifferentiated” and
“mushy”. In your typical guitar rig, we therefore see even whole bunch of high-pass filters
taking care of effectively attenuating the very low frequencies: several RC-high-passes, the
output transformer, and the loudspeaker. An extreme case was already mentioned in Chapter
10.1.4: in some Fender amps, you will find an RC-cutoff as low as 3 Hz. But then there is the
other extreme: the 600-Hz-high-pass in the VOX AC-30.
Low frequencies may be attenuated not only in the plate-circuit where the RC-coupling works
as a high-pass, but also in the cathode circuit. To obtain the highest possible gain, the
cathode-resistor is often bridged by a capacitor. This cathode-capacitor will, however, only
have an effect as long as its impedance is not significantly higher than the value of the resistor
it bridges. Since it is not possible to make this capacitor indefinitely large, two cutoff-
frequencies appear: below the lower cutoff-frequency, the capacitor is almost without any
effect and the gain here is vT, while above the upper cutoff-frequency, the gain is vH, with a
monotonous increase in between (Fig. 10.1.25).
For the small-signal model, the tube is replaced by an AC-voltage-source of the voltage U0 =
µ⋅Ugk. Here, µ is the open-loop gain of the tube – a theoretical parameter amounting to about
100 for the ECC83. The internal impedance Ri of the tube is connected in series to this source
(internally within the tube); for the ECC83, its value is about 50 – 100 kΩ. If we postulate
that there is no current though the grid, the plate-current equals the cathode current and is
calculated as Ik = U0 / (Rk + Ri + Ra). Ik generates a negative feedback voltage across the
cathode-resistor. The input voltage Ue decreases by the amount of this feedback voltage
. This enables us to calculate the plate-voltage :
Voltage gain (without load)
µ = 100, Ri = 72 kΩ, Ra = 100 kΩ, Rk = 1,5 kΩ, yields .

Including the cathode-capacitor (which is taken as a short in the high-frequency region) sets
the voltage gain to ; this again is for the unloaded tube. A load resistor is
simply connected in parallel to the plate-resistance: with a load of e.g. 100 kΩ, Ra is reduced
to 50 kΩ and the voltage gain drops to, and ,
respectively. For the tube without load, the cathode-capacitor will generate a treble-boost of
5,5 dB, and for the tube loaded with 100 kΩ, the boost will be 7,1 dB (Fig. 10.1.25). The
capacitance of the cathode-capacitor – in conjunction with the remainder of the circuitry –
determines in which frequency-range the transition from vT to vH happens. We could surmise
that, besides Ck, it is only Rk that sets the treble-boost because this is the resistor that Ck
bridges. However, in fact the cathode needs to be considered as load of this two-pole, as well.
The relative treble-boost is:
Relative treble-boost
The center-frequency fZ (marked with a small circle in the figure) computes to:
Center-frequency
If the cathode-resistor is bridged with a “large electrolytic cap” of e.g. 25 µF or more, the
center-frequency is located so low (e.g. 5 Hz) that the gain receives a broadband increase –
this being the normal approach for Fender amplifiers. Typical examples for small capacitor
values (e.g. 0,68 µF) are found in some Marshall amps (fZ = 150Hz, ΔG = 8dB).
Ra = 100 kΩ
Ri = 72 kΩ
Rk = 1,5 kΩ
Ck = 0,68 µF
CL = ∞
RL = 100 kΩ
Fig. 10.1.25: Input-circuit of a tube amplifier (left), effect of the cathode-capacitor (right).
In the circuit according to Fig. 10.1.25, the coupling capacitor CL is taken to be of infinite
capacitance in order to be able to show the effect of the cathode-capacitor by itself. In guitar
amps, CL often has a value of 22 nF, but larger values (0,1 µF) are used, as well, as are
smaller capacitances (500 pF). When calculating the resulting high-pass cutoff frequencies, it
should be considered that the internal impedance of the tube circuit is not zero but is given by
the Ra and Ri connected in parallel.
The classic guitar amplifier contains 4 tube-stages and thus has 3 coupling capacitors – the
output of power stage is not picked up via a capacitor but via the output transformer. While it
is easy to calculate the effect of the coupling caps on the low-frequency-response, the output
transformer constitutes a complex system the data of which cannot be seen in the circuit
diagram. The upper cutoff-frequency is not apparent, either.

We could assume that the upper cutoff-frequency is always sufficiently high to reproduce the
guitar signal (which is low-pass-limited by the pickups), and that therefore it would be not
necessary to specially consider it. This assumption is, however, only admissible as long as the
amplifier is considered as a linear system. If an overdrive situation occurs, we get signal
components in the ultrasonic range. These would be inaudible by themselves – however, as
ultrasonic signals hit a non-linear amplifier stage, difference tones may be formed that may be
audible, after all. A tone-pair constituted of two ultrasound signals (e.g. 24 kHz and 25 kHz)
is inaudible at normal levels. Feeding the tone-pair to a 2nd-order distortion-characteristic will
generate (among other components) a 1 kHz 2nd-order difference tone that may be audible.
This effect should not be seen as all that dramatic, but it should not be entirely disregarded,
either. Whether the 1-kHz-tone is in fact audible depends on its level and the levels of further
neighboring tones which may have a masking effect. After all, the two ultrasound-tones are
not generated in isolation, but are part of a spectrum generated by preceding amplifier stages,
and they will not have very large levels. However, since guitar amps may include a very
strong emphasis in the high frequency register, a bit of out-of-the-box thinking is advised. We
have distortion, treble-boost and subsequently more distortion: there is potential for audible
sound differences the reason for which may lie in the ultrasonic region.
Why do we not find any upper cutoff-frequency in the data-sheets of tubes? Some manuals
will give 300 MHz for triodes, or – depending on the type – 1 GHz; however, specifically for
the ECC83 this field is usually left empty. The reason is actually rather trivial: the upper
cutoff-frequency is determined by the circuitry around the tube. Let’s speculate a bit how all
this started: the first guitar amps had to be economical regarding the use of power – that made
(after the octal-socket-era had passed) the 12AX7 with only 1 mA plate-current highly
welcome. As a result, the circuitry had to be of rather high impedance, with 1-MΩ-
potentiometers (Fender, Marshall) necessary so as not to load the plate circuits too much.
With the center-tap of such a potentiometer set in the middle of the range, its internal
impedance is about 250 kΩ♣. Connecting this center-tap to the next high-gain triode with an
input capacitance (enlarged by the Miller-effect) of about 150 pF, we get a low-pass with a
cutoff frequency at about 4,2 kHz. That is kilohertz, not megahertz! You would not want to
include such a low upper cutoff frequency in a data-book – it would look quite bad. The
relatively high input capacitance is generated by the capacitance between grid and plate
(12AX7: Cga = 1,6 pF) that is enlarged by a factor given by the voltage gain. With vU = 50,
this already amounts to 80 pF, and since the wiring leading up to the tube also has a
capacitance, 150 pF are easily reached – or even more. The low-pass mentioned above is not
always there, though: if the center-tap of the 1-MΩ-pot feeds a cathode-follower (common-
plate circuit) the cutoff-frequency will be much higher. In Fenders “Twin-Reverb” (just to
name one example), however, the center-tap of the potentiometer directly connects to a
common-cathode circuit the input capacitance of which is relatively high. In many Marshall
amplifiers there is even a 470-kΩ-resistor in series to the center-tap (summation-stage, total
series resistance = 320 kΩ). At this location in the circuit there was also an opportunity to
include a low-cost supplement increasing the treble response: a fixed capacitor (Marshall) or a
switchable capacitor (“Bright”-switch, Fender). The overall actual cutoff frequency resulting
from this hodge-podge of frequency-boosts and frequency-cuts can be calculated via
complicated models but depends on many parameters – not least on the layout. The distance
between lines leading to grid and plate does influence, via the Miller effect, the input capacity
and the upper cutoff-frequency.
♣
The internal impedance of the tube will also make a small contribution.

Fig. 10.1.26 shows a section of the layout of a Fender amplifier (Super-Amp). Resistors and
capacitors are soldered to eyelets on a carrier board, and wires lead from the long side of the
board to other sub-assemblies (connectors, potentiometers, tubes, transformers). Some wires
are laid out below the board at only a small distance to the components above. For example,
the wire connecting the grid of a tube is located directly below the coupling capacitor
connected to the plate of the same tube – this certainly is not the best possible decoupling
approach. Even more extreme is the situation with three wires coming out of an access-hole
(in the top section of the picture): two of these are connected to the input jacks, the third
carries the plate-AC-voltage of the corresponding input tube. The resulting capacitive
coupling is not particularly strong but we need to consider that the grid-plate-circuit is
especially sensitive, and that such coupling has the effect of a low-pass. It cannot be excluded
that such a low-pass is in fact intended, but comparisons with many other Fender layouts do
not really support this assumption. The various wires seem to too arbitrarily keep or change
their positions over the years.
Fig. 10.1.26: Fender component board (excerpt,

above).
Third-octave spectrum of the power tube-g1-voltage
(Fender Deluxe). Stratocaster, Stratocaster (left).
As distortion occurs, frequencies above 5 kHz result. The above third-octave-diagram shows
this – it is taken from the grid of a power-tube; similar situations can be present at other tubes,
as well. The effect of the input capacitance of a tube is shown in Fig. 10.1.27 using the
example of volume-pot: as it is turned down we obtain a low-pass-effect. The cutoff-
frequency is lowest at an attenuation of about 6 dB. A “Bright”-capacitor bridging the pot
(from the anode to the grid of the subsequent tube) compensates this treble loss but as the
control is turned down further, an over-emphasis of the treble occurs. The individual
characteristics are strongly dependent on stray-capacitances and on the gain of the individual
tubes.
Fig. 10.1.27: Transfer-function of a volume-pot loaded by a capacitance (1 MΩ); tube-input-capacitance 150 pF.

10.1.6 Time variance
Many theorems of systems-theory are valid only for linear and time-invariant systems.
Guitars amplifiers are neither the former nor the latter. There is non-linear distortion, and the
characteristics are time-variant: they change due to aging of the components (this being a
well-known aspect), and they are subject to short-term shifts of the operating points (this
aspect being not recognized much). Trivial time-variances relate to components that change
their characteristics dependent on temperature, or that age relatively quickly (such as tubes).
These time-variances will not be addressed here – for the present consideration, all
components are assumed to be time-invariant. Short-term shifts in the operating point,
however, can nevertheless occur, because non-linear processes lead to a re-charging in
capacitors. If there were no capacitors, this chapter would be omitted – or the other way
round: every capacitor is a potential source for variances.
Tubes in guitar amplifiers are often overdriven – they are non-linear systems. Even for
seemingly undistorted (“clean”) sounds, the attack may easily be slightly distorted♣. All even-
order distortions (k2, k4, k6 ...) generate an additional DC-component that shifts the operating
point for a short time – the transfer behavior thus becomes time-variant. For example, the
cathode-resistor is often bridged by a capacitor in order to reduce the negative feedback. The
DC-component generated by even-numbered distortion of the cathode current changes – in a
time-variant manner – the cathode-voltage and correspondingly the operating point. A further
variable is the supply-voltage fluctuating (“sagging”) dependent on the power fed to the
loudspeaker. While these shifts are low-pass-filtered, they are not regulated out; they have a
backwards effect on the plate-voltages of preamp and intermediate amp. In Fig. 10.1.28, we
see measurements of the supply-voltage of a Fender amplifier (Deluxe). The amp is fully
driven from t = 0,3 to 1,7 s and the supply-voltage drops from 247 V to 210 V. As a
consequence, maximum signal level and harmonic distortion change as already shown in Fig.
10.1.12. Many guitar players demand this effect since they feel that it renders the guitar sound
livelier. However, measures are also taken to reduce this sagging – via changes in the filter
capacitors and associated resistors. In early amps, the filter capacitor (CB) was rather small (8
µF) and was later increased to up to 50 µF. For full removal of the effect, a stabilizer-tube is
required. The sagging is not primarily caused by the larger current consumption of the
preamp-tube but by the current used up by the power stage that reduces the voltage of the
power supply. The exact shape of UB over time therefore depends on many parameters.
Fig. 10.1.28: Tube input stage (left); drive-dependent sagging of the voltage supply (right).
♣
It is the high art of amplifier-design to make such distortion sound good.

Another variance shows for the coupling capacitors. Only for a negative grid-offset does
driving a tube not require any power (i.e. no current at the input is required). Typically, we
find an offset of Ugk = –1,2V in preamplifier tubes, and thus strong pickups can drive the tube
into the range of non-negligible grid-current. The subsequent tubes are all the more likely to
experience grid-current. Most amplifiers do not use a coupling capacitor at the input; the
pickup is usually connected via a resistance of 34 kΩ to the grid of the first tube. If the pickup
were connected via a capacitor, only the grid-current could charge it (as a unipolar current
creating a negative voltage at the grid) and shift the operating point towards a smaller plate-
current. Of course, this is only temporary because the capacitor could discharge via the grid-
resistor and the internal guitar-impedance – but exactly these transient processes are NOT
present in the input stage of the classic guitar amps (Fender, VOX, Marshall), if we disregard
some very early variants using the splash current (i.e. using grid-current bias).
We find a much different situation in the coupling circuits between the individual tubes: here
there is almost always a coupling capacitor (the only exception actually being the galvanic
coupling in cathode-follower circuits). Since the AC-plate-voltages can be much larger than
the voltages allowable for operation without grid-current, a temporary re-charging of the
coupling capacitors is almost inevitable. The grid-currents themselves will not in principle
lead to audible effects, but the shift in the operating point can lead to audible changes in the
harmonic content. We can roughly estimate the speed with which the re-charging processes
run: for the “charging” (grid-current flowing) there is a non-linear process because the input-
impedance of the tube becomes non-linear. As an approximation, the load-resistance of the
preceding tube may serve – in conjunction with the capacitance of the coupling capacitor.
Depending on the specific amplifier, the re-charging will happen over the course of a few
milliseconds. The “discharging” cannot happen via the grid but only via the leak-resistance
(in the order of at least 1 MΩ). This leads to an effect occurring over a time of 20 ms i.e. a
time comparable to what is used in studio-compressors (in a “fast” setting). Thus: even if the
value of the coupling capacitor is large enough that the high-pass it constitutes is effective
only at frequencies far below those of usual guitar signals – the recharging times are defined
by these capacitances (and the resistors in the circuit).
Given the sheer variety of tube amplifiers available on the market, it is difficult to specify the
typical cause for the “tube sound”. Even when only asking about the typical characteristics, a
range of different answers is offered; this will happen even more if we look for the
corresponding causes. An often heard verdict would be: the tube amp is alive, it plays more
dynamically, sounds more lively, reacts better to changes in the expression. The opinion
regarding a transistor amplifier often is the opposite, it is said to sound sterile, impersonal,
analytical, dead. The perceived “liveliness” connected to the tube may well have its base in
the shifts of the operating point as described above. Even if a unidirectional current as small
as 10 µA flows through a 22-nF-capacitor for a mere 1 ms, the resulting voltage change will
be 0,45 V. Such a shift in the operating point would drastically alter the transmission behavior
of an ECC83. It is not that such a behavior would not be possible to obtain with a
semiconductor amplifier, as well – however “modern” circuit design sees big advantages in
direct coupling between amplification stages (i.e. without capacitors). That is indeed a
conducive approach to minimize artifacts, but in guitar amplifiers that is exactly NOT the
issue (or at least not a main one).
__________
P.S.: The term time-variant chosen here is valid for short-term considerations; in the long term the shifts in the
operating point as discussed above are indeed time-invariant i.e. they run an identical course given identical
excitation. This distinction is, however, only important in a strictly systems-theory-oriented approach.

10.1.7 Noise, hum, microphonics
The noise of the input-amplifier tube can be modeled with good approximation by a noise-
source connected in series with the tube grid and generating white noise with a noise-voltage
density of about 5 (compare to Chapter 5.12). Note that this model does include
only the stochastic component of the overall interference, and not the hum component
generated by the AC-heater of the tube. Moreover, in a typical input-circuit it is not only the
tube itself creating the noise: in addition, there is the grid-resistor (34 kΩ) resulting from the
parallel connection of the two 68-kΩ-resistors in the input-circuit. In fact, this resistor is the
actual culprit and acts as the main noise-source with a model voltage-density of no less than
24 ! Consequently, it is pointless to consider tubes with lower noise as long as we
cling to the classical input-circuit. By the way: the overall noise-voltage may not be
calculated by simple summation because the signals from noise-sources are not correlated.
Rather, a square-root summation needs to be performed:
Clearly, the noise from the tube contributes almost nothing to the overall noise. However,
before taking out the grid-resistor and connecting the pickup directly to the grid of the input
tube, you should consider that this resistor does have some other jobs to do, too: it limits the
grid-current and influences the non-linear distortion of the preamp-tube. Moreover, together
with the input capacitance, it does form a low-pass that suppresses unwanted RF (This is
Radio Free Europe ...). In many cases the noise generated by the grid-resistor will be less than
the noise generated by the guitar circuit; the latter may certainly reach voltage-densities of
40 (or even more) in the frequency-range important for the hearing system.
Fig. 10.1.29: 1/3rd-octave noise-spectrum (ECC83).

The two dashed lines mark the spectrum belonging
to white noise; the dotted line shows the typical
noise-spectrum generated by a Stratocaster. All
spectra are referenced to the tube input.
In Fig. 10.1.29 we see the measured third-octave spectrum of an ECC83 in comparison to the
theoretical characteristics. For the measurement, the grid was shorted to ground and the tube
received DC-heating. Hum of around 0,1 µV is typical for simple shielding; this is much less
than the interference caught by magnetic pickups. Without the grid-resistor, the tube creates –
across the whole frequency-range – less noise than the pickup measured for comparison
(Chapter 5.12). Including the grid-resistor, the pickup noise dominates only in the range of the
pickup resonance. The third-octave levels measured at the plate are, compared to the levels
given in the figure, larger by the gain factor (33,4 dB in our example). The broadband input-
noise voltage below 20 kHz amounts to about 1µVeff (with shorted grid); this is equivalent to
a noise-voltage of about 47 µVeff at the plate.

Every guitarist can find out for him/herself which noise-source dominates in a given guitar-
amp setting: just compare the noise with shorted input to the noise that occurs with the guitar
plugged in and fully turned up. In case both signals are approximately equal in strength, one
needs to indeed question the quality of the input tube (or that of the amplifier concept); if
there is more noise with the guitar turned up, the interference is caused there. How can we
achieve a short circuit at the input? The best way is to use a plug with both connecting pins
soldered together. Alternatively, a metal potentiometer shaft (6,3 mm) or a similar short-
circuit-pin could be plugged into the input jack. Or, very simple: plug in the guitar and turn
the volume control (on the guitar) to “0”. Note, however, that this works only if the guitar
cable actually reaches the center-tap (middle) connector of the pot (as is the case for Strats or
Les Pauls with the customary circuitry). Instruments that have the pots in the so-called
“reverse” connection (such as the Fender Jazz Bass) are not suitable for this approach
The second unwanted signal generated in an amplifier is hum. It is caused by the power
system (230V/50 Hz, or 110V/60, or other voltages/frequencies depending on the country)
that contaminates the more sensitive circuit sections via capacitive or inductive coupling♣.
Faulty design of the layout of the ground-connection can be a reason, as well – especially in
the power-rectifier circuit. In the typical tube amplifier we have relatively strong heating
currents (preamps tubes: 0,3 A, power tubes 1 – 2 A) the magnetic fields of which can feed
into the sensitive plate circuits. DC-heating would be an option for (the customary) indirectly
heated tubes but is implemented rarely. It is not really necessary, either: using twisted wiring
for the heating and a correct layout of the (electric) ground, every tube amplifier can be
constructed in a sufficiently hum-free way such that – for normal use – the hum caught by the
magnetic pickups of the guitar dominates.
Microphonics is a term characterizing the tendency of a tube to react to sound (i.e.

mechanical vibrations), whether transmitted via air, or structure-borne. Combo-type amps –
with loudspeaker and amplifier housed in the same cabinet – are particularly prone to
associated problems. The amp may sound as if there is always a bell operating in the
background, and at high volumes a howling, uncontrollable feedback may occur. The cause of
microphonics is a deformation in the tube-interior, in particular in the (control) grid. The
ultra-thin grid wires start to vibrate as sound impacts on the tube, and this in turn modulates
the plate-current and generates interfering noises. Every tube is microphonic – but not always
to the extent that problems result. Preamp tubes with their very small signal voltages should
have especially low microphonics, and tubes specially selected towards this goal are
available.
In an orientating measurement, a double-triode (12AX7) that generated a clearly ringing tone

at 630 Hz when tapped was subjected to sound coming from a loudspeaker. At 130 dB SPL (a
sound pressure level easily reached in a combo), an interference voltage of about 1 mV (when
referenced back to the input) occurred. A 12AU7 was even considerably worse at 30 mV!
Even without fully turning up an amp, such a tube will start to bring some undesirable
accompaniment, and feedback whenever the amplification is high. Vibration can get to the
tubes not only via air but also via the tube-socket. Consequently it is advisable to consider – at
least for the preamp – mounting the respective tubes in sockets using rubber or a similar
mechanical absorbent material. The latter should be able to withstand heat while not being
prone to embrittlement.
♣
This happens not only at 50/60 Hz but also at the multiple frequencies, i.e. at 100/120 Hz, 150/180 Hz, etc.

10.1.8 Noise processes
Noise belongs to the stochastic signals – it is not possible to predict its exact course. The
simplest quantitative description involves RMS-value, bandwidth, and spectral-envelope
characteristic (dP/df = const. or 1/f). Supplementary specification regarding the time function
is given by distribution (= probability-density-function) and cumulation (= probability-
density-distribution), further information regarding the spectral distribution results from DFT-
and 1/3rd-octave-sprectra. The literature listed at the end of this chapter may serve as guide to
the theoretical principles for the description of random signals – the following listing in short
introduces the most important noise processes.
a) Thermal noise (white noise i.e. dP/df = const.)

The temperature-dependent random-movements of free charge-carriers in a conductor (or in a
resistor) lead to a thermal open-loop voltage at the connecting terminals (without any load);
the RMS-value of this voltage is computed to:
Open-loop noise-voltage density en and RMS-value of open-loop noise-voltage for Δf = 10 kHz at resistor R:
R = 58.8 100 200 1k 10k 100k 1M Ω

en = 1.00 1.30 1.8 4.1 13.0 41.2 130
= 0.1 0.13 0.18 0.41 1.3 4.12 13 µV
b) Shot noise (white, i.e. dP/df = const.)

Shot noise occurs in semiconductors and amplifier tubes. It is caused by statistic fluctuations
of the current-flow through an interface layer between potentials. As an example, the electron-
emission at an amplifier cathode may be modeled by a Poisson-distribution, with the current
not continuously flowing but having statistic fluctuations. The real tube-noise is (given the
space-charge conditions) slightly less than the theoretical maximum value calculated below
for saturation [Meinke/Gundlach]:
2e = 3.204⋅10-19 As
Noise-current density iS, (RMS) noise-voltage across a 10-kΩ-resistor for 10 kHz bandwidth, generated by
DC I0: [f = Femto = 10-15, p = Pico = 10-12]
I0 = 10 n 100 n 1µ 10 µ 100 µ 1m 10 m A
iS = 56,6 f 179 f 566 f 1,79 p 5,66 p 17,9 p 56,6 p
= 56,6 n 179 n 566 n 1,79 µ 5,66 µ 17,9 µ 56,6 µ V
The relation between shot-noise voltage and thermal noise-voltage depends on the DC
voltage across the resistor and on the temperature voltage:
U0 is the DC-voltage across resistor R; 2UT = 2⋅26 mV = 52 mV.

c) Flicker noise (approximately pink, i.e. dP/df ~ 1/f)

This is low-frequency 1/f-noise caused by inhomogeneities in the material, deficiencies from
manufacture, contaminations, and charge-fluctuations at surfaces. The designation stems from
the burn spots jumping around (flickering) on the cathode of an amplifier tube. Simplified, the
power-density decreases towards high frequencies with 1/f (pink noise). However, also
observed were noise processes the spectral density of which does not correspond exactly to
the 1/f-hyperbola. Flicker noise is only relevant in the low-frequency range.
The 1/f-noise caused in resistors carrying DC is characterized by the Noise-Index NI. Metal-
film resistors (homogeneous crystal lattice structure) feature a small NI, while carbon
composition resistors have large NI-values. In general, resistors with a high power-handling
capacity (and requiring a larger volume) generate less noise than their low-power cousins of
the same basic build.
represents the DC-voltage across the resistor, is the resulting 1/f-noise-voltage (RMS
value) per frequency-decade; NI = 0 dB ⇒ 1 µV/V.
Abb. 10.1.30: left: noise-index NI for two different resistor types (Kohleschicht = carbon layer, Metallfilm =
metal film). The grey areas show the scatter range between typical average values and typical maximum values.
Right: measured 1/3rd-octave noise-; dashed: intrinsic noise of analyzer. Pink noise results in frequency-
independent 1/3rd-octave-level voltage levels; for white noise the 1/3rd-octave-levels rise with 10 dB/decade.
Kohlepresswiderstand = carbon-composite resistor; Metallfilmwiderstand = metal film resistor; Analysator-
Eigenrauschen = intrinsic noise of analyzer.
In Fig. 10.1.30, NI is listed for different resistor types. The areas marked in grey can only give
very approximate orientation-values since the individual build has significant influence on the
NI. In the right-hand section of the figure, we see measurements taken with two serially
connected 68-kΩ-resistors carrying a DC of 1 mA. The two incoherent noise currents of the
two resistors need to be added via a Pythagorean summation, and the mutual loading plus the
loading via the analyzer (100 kΩ) has to be considered, as well. The metal-film resistors show
a thermal white noise in the high-frequency region, and a current-dependent pink noise at low
frequencies. In the carbon-composite resistors, current-dependent pink noise dominates
throughout practically the whole frequency-range. These measurements give a noise index of
the carbon-composite resistors of -11 dB, and an NI for metal film resistors of -32 dB. At low
frequencies, the noise power densities of these two resistor-types therefore differ by a factor
of 126. This factor is current-dependent; 1 mA is typical for plate-currents in preamplifiers.

Despite this considerable current-noise, carbon-composite resistors are listed as “absolute

high-end” in the catalog of a retailer; one is very temped to interpret this as “absolute upper
range of the resistor noise”. The high-end fan must furthermore not be irritated by the fact the
carbon-composite resistors have also considerably larger tolerances (compared to metal film
resistors): maximum ±10% (carbon) vs. maximum ±1% (metal). Measurements confirm this:
+ 7% (carbon) vs. -0,3% (metal). What about the price-difference? As expected, carbon-
composite resistors are about 10 times as expensive as metal film resistors. Say no more: it’s
about more noise – more tolerance to resistance – more money …
What remains is the question whether differences in the current-noise play any role at all
compared to the shot-noise generated in the tube. For an ECC83 (12AX7), the equivalent
input-noise voltage-density may be set to about 5 as a good approximation. With a
voltage gain of 34 dB, this is equal to 250 at the plate, corresponding to a third-
octave-level of 11.6 dBµ at 1 kHz (bandwidth 232 Hz). In comparison, the thermal noise from
the grid-resistors (68 kΩ // 68 kΩ = 34 kΩ) typically found in the input-stages of guitar amps
is five times as much (Chapter 10.1.7), reaching some ample 26 dBµ in the third-octave band
around 1 kHz. And how are we doing regarding the resistor-noise created by the plate-
current? Given a 100-V-voltage-drop across the plate-resistor, and including a noise-index of
NI = -11 dB, we would be confronted with a 1-kHz-third-octave-level (open loop) of no more
than 19 dBµ. With the loading by the internal impedance of the tube, this would decrease to
about 11 dBµ. Consequently, the current-noise of a carbon plate-resistor (NI = -11 dB) at 1
kHz is lower than the noise of the preamplifier by 15 dB. For higher frequencies, this
difference will grow even bigger, and only below 31 Hz, the current-noise would become
dominant for the present model.
So: The current-noise of customary carbon resistors is inaudible in the investigated circuits
But: Supposedly there are carbon-composite resistors with NI not at -11 dB, but at 0 dB,
Or even higher – that could then just become audible.
Question: Is that worth 10 times the price? Answer: sure, the retailers are happy.
Two advantages are often highlighted to scientifically support the apparent superiority of the
carbon-composite resistors: high power capacity with impulses, and small inductance. There
may be scenarios in which the relative long thermal time-constant of the carbon-composites
helps to avoid overheating, but pre-amp stages in guitar amps are not even remotely in the
playing filed here. O.K. then: the reported low inductance of composite resistors will be
crucial, won’t it? No, sorry, that aspect is utterly insignificant in the relevant frequency-range!
The impedance of a 100-kΩ-resistor will increase by 0,000000002% at 100 kHz (with a
inductance of 1 µH as a baseline). This increase should be seen relative to the manufacturing
tolerances in carbon-composites: 10% according to data sheets. Plus: do not forget that 1 µH
is already a high value; in data sheets we often find the entry “a few nano-henry”. BTW, our
metal-o-phobic friends prefer not to mention capacitive reactive values, although these exist
in carbon resistors, as well. Do you need to consider those? Course you do … if you want to
look beyond 1 MHz, where the reactive currents start to achieve some significance.
The never ending Internet saga of Carbon Comps: “Smooth, creamy sound…Are unstable, should not be
used…Very clean and natural sound…Should be avoided…Taut and 3-dimensional sound…Make the working
point drift away…Are the only choice for guitar amps…Never heard any difference in sound…Light-years
ahead.” More examples are available …
Literature: Motchenbacher/Connelly: Low-Noise Electronic System Design, Wiley 1993. Connor: Rauschen,
Vieweg 1987. Hänsler: Statistische Signale, Springer 1991. Bendat/Piersol: Random Data, Wiley 1986.

10.1.9 Pentode-preamp
Of all amplifier tubes, the pentode is the most widely used. Compared to the triode, the
technical advantage of the pentode used in input stages is based on the high internal
impedance and the very small capacitance between grid and plate. Large voltage gain is
possible without running the danger of self-excitation [Meinke/Gundlach]. However, as we so
often see it: what is true for classical circuit design does not necessarily hold as guideline for
guitar amplifiers – the latter most often employ a triode in the first stages. There are
exceptions, though: the VOX AC-15 or the Fender Champ may serve as examples. In these
rather early amps, we find a pentode in the preamplifier. We will look into the technical
details of this five-electrode-tube a little later; as a simplification, it functions similar to the
triode: the plate-current is controlled by the voltage at the control-grid, the extra screen-grid
(g2) is connected to a constant (high) voltage, and the suppressor-grid is joined with the
cathode. The transconductance of the 6 SJ 7 pentode used in the Champ is rather
comparable to that of an ECC83 (1,6 mA/V) but the internal impedances are very different:
1000 kΩ in the 6 SJ 7 but merely 63 kΩ in the ECC83. Purely by way of calculation, this
yields – e.g. for Ra = 200 kΩ – an operational gain of 267 (6 SJ 7) and 48 (ECC83). The
operational gain-factors therefore differ by 15 dB! The EF 86 as it is deployed in the AC-15
features even larger values for transconductance (2 mA/V) and internal impedance (2500 kΩ),
and we get an additional 3 dB gain.
It was the susceptibility to oscillations that made VOX replace the EF 86 by a triode, after all:
The EF 86, although excellent electronically, was susceptible to mechanical damage through
vibration and would soon begin adding it's own ringing, rattling accompaniment
[Petersen/Denney]. Another reason could lie in the seeming advantage of the pentode: its
high voltage gain is helpful when dealing with small input signal. However, when confronted
with pickups able to deliver in excess of 1 V, this advantage can easily backfire: the preamp
will generate considerable distortion that is not generally desired.
Fig. 10.1.31: Pentode-input-stages in guitar amplifiers: VOX AC-15 (left), Fender Dual Professional (right).
Fig. 10.1.31 shows the input-circuits of two early guitar amps. The AC-15 employs the more
modern pentode with the noval-socket while the Dual Professional (developed more than 10
years earlier) still relies on the octal-tube. Only shortly thereafter Fender changes to the dual-
triode 6SC7, and in the following generation to the 12AY7.

10.2 Intermediate amplifier
In the signal path, the intermediate amplifier operates between input stage and power stage –
but it is not the tube immediately ahead of the power tubes that is meant here (that would be
the tube commonly known as phase inverter in push-pull arrangements, see Chapter 10.4). In
the classical guitar amp, the typical intermediate amplifier is the second amplification stage.
Between the first and the second stage we find the tone-filter … or the volume control … or
both. In fact even in the classic amp-forefathers we already find different concepts.
Which are the (dis-) advantages of these various topologies, and what are the sonic
differences? That is quite difficult to answer. It is easier to address the question what the
reasons could have been to implement the respective topology. Fig.10.2.1 shows the most
important ones – there are more but we will not investigate them here. In almost all guitar
amps, the signal from the pickup is directly fed to the first tube. This is because any circuitry
connected between pickup and the first tube would have to be high-impedance and thus would
unduly increase noise. If the volume control is located directly after the first tube and the
tone-filter subsequently (as shown in the first variant), then the source circuit (the volume
control) feeding the tone-filter would have an internal impedance that depends on the position
of the potentiometer’s center-tap. Moreover, the potentiometer load (= filter-input) would be
frequency-dependent. The effect of the filter would therefore not only depend on the settings
of the tone control, but also on the setting of the volume control. Presumable it was this
interdependency that precluded the corresponding topology from become really widespread.
Fig. 10.2.1: Some obvious circuitry-topologies. BMT = tone-filter, arrow = volume-pot.
We will look more closely at the second and third topology-variants; these two are found
most often in tube amplifiers. The fourth variant would work without any problem, as well,
but was apparently not seen as directly superior and was thus rarely used. Variant two and
three differ in the position of the tone-filter: ahead of the volume control or after it. The
sequence of the subsystems in a guitar amp would be rather unimportant if the amp were a
linear system. However, as Chapter 10.1 has shown, non-negligible harmonic distortion
happens as early as the very first amplifier stage; the system is non-linear in quite a
complicated way. Moreover, a further non-linear effect needs to be considered: the noise that
every component generates. Non-linear system need to be source-free i.e. they must not
include any noise-sources, either. If the volume-pot is positioned late in the signal-flow (close
to the power amp), almost no noise will be audible when the volume control is turned down.
However, there is now considerable danger that one of the preceding amplifier stages will be
overdriven in case the connected guitar has a high-output – and this danger cannot be reduced
by turning down the volume control. If, conversely, the volume control is located directly
after the first stage, any potential overdrive of subsequent stages is fully controllable – but
there may be a considerable noise level even with the volume set to zero. Of course, no
guitarist plays his or her amp with the volume fully turned down so this would probably not
be a problem. Rather, the sales department that makes demands here: in the music store, it’s
no good if the amp creates such a racket even though no-one is even playing through it. Still,
the amp needs to be “clean” at low volume. Only later amplifier generations include “Fat”-
and “Boost”-switches, and master-volumes to get more sound-options; the early amplifier-
variants had to do without that. Obviously, “sound” won out over “noise”: in the circuits, the
volume control was close to the input (mostly before the second tube stage).

10.2 Intermediate amplifier 10-37
10.2.1 Intermediate amplifier in common cathode-circuit
The standard version of the intermediate amplifier contains one tube (almost always a triode)
in common-cathode configuration. The circuit is similar or even identical to the first preamp-
stage. And why not – the signal has been attenuated by tone-filter and/or volume control and
needs to be re-amplified, with the common-cathode configuration being highly suitable.
Sometimes, the developers see a need for an impedance conversion in the second amplifier
stage – this aspect we will cover in the next section (10.2.2).
In the common-cathode circuit, the cathode is connected to “common” i.e. to ground. The
required grid-offset is usually generated “automatically” by a cathode-resistor (Chapter 10.1).
A capacitor is connected across this resistor in order for the latter to be active only for DC,
and to avoid any AC-voltage across it (which would introduce negative feedback). As long as
there is not grid-current, this circuit features very high input impedance – although a non-
negligible input capacitance (100 pF minimum) does require consideration. The output
impedance (internal impedance) results from the parallel connection of the internal impedance
of the tube (about 60 kΩ) and the plate-resistor (100 kΩ); the gain factor is about 35 dB (or a
bit less if there is significant loading).
Fig. 10.2.2 shows two famous amplifier concepts in comparison: in the Fender circuit, the
volume potentiometer directly follows the tone-filter and feeds the intermediate stage, while
in the VOX, the intermediate stage is placed between volume pot and tone-filter. Fender
follows the simple line of thinking: take care of all control efforts at one and the same
location. The interaction between the directly connected volume control and tone-filter
remains within reasonable limits because the pot is of relatively high impedance (1 MΩ).
With the VOX, we find an entirely different approach: a special intermediate amplifier with
high-impedance input (common cathode configuration) and low-impedance output (common-
plate configuration, see 10.2.2) follows the volume pot.
Fig. 10.2.2: Comparison between a typical Fender-circuits (left) and a VOX-circuit (right).
*) There are VOX amps that do not include the cathode-capacitor for the 2nd tube.
Pushing the discussion of the tone-filter into Chapter 10.3, we will first analyze the 2nd tube-
stage of the Fender circuit. Both 1st and 2nd tube-stages are fundamentally similar but there
are differences regarding the cathode circuit: in the Super Reverb (under scrutiny here), the
cathode-RC-circuit also feeds the corresponding cathode of a tube in the other input-channel.
Other Fender amplifiers include the same component-saving detail. In the figure, the second
tube is not included but an arrow indicates the connection to it. For the grid-offset of the
tube(s) to remain at the desired value, the value of the cathode-resistor common to both tubes
is approximately halved at 820 Ω (instead of 1,5 kΩ). Since both triodes are feeding relatively
high impedance circuits, they have similar voltage gains. Given a regular ECC83, each triode
will yield about 32 – 34 dB. The harmonic distortion, however, will be different because the
source impedances (ahead of the grid) differ.

10.2.2 Intermediate amplifier with cathode-follower
The VOX-circuit (Fig. 10.2.2) differs from the Fender-circuit not just in the sequence of the
partial systems but also in the build of the second amplifier stage. It deploys two triodes: the
first generates the required voltage gain while the second acts as a current amplifier
(impedance converter = cathode-follower = common-plate circuit) and achieves a low output-
impedance (= internal impedance). Strictly calculating the internal impedance according to
text-books we get 1/S (S = transconductance); for the present circuit this would be 600 Ω. An
output-impedance of such a low order would not be mandatory, though: the load imposed by
the VOX-tone-filter is always larger than 100 kΩ. Before we go into further detail regarding
the rather special dimensioning of the VOX-circuit, let us quickly review the history of the
cathode-follower: Leo Fender outfits his tweed amps with this circuit from the mid-1950’s
(albeit not using the 12AX7 but the 12AY7, Fig. 10.2.3).
Fig. 10.2.3: Intermediate amplifier with cathode-follower; family of output-characteristics of the 12AY7,
UB = 170V … 275V.
For the 5D8-Twin, the layout specifies [Funk] a supply-voltage of UB = 170 V, for the later
5E6-Bassman this has risen to 235 V, and in the 5E6-A we find even 275 V. With the increase
of the supply-voltage, the quiescent current of the triodes also mounts; this is indicated in Fig.
10.2.3 as a dot on the load-line. For UB = 170 V, the travel of the plate-voltage of the first
triode is limited to about 35 V towards small values by the Ugk=0V-characteristic. For even
smaller Ua (i.e. larger Ia), the grid would have to become positive relative to the cathode but
this is only possible to a small extent: the grid-current is kept low by the high-impedance
feed. If the first tube were in blocking mode, its plate-voltage would be the same as the
supply-voltage (with no load present). However, since in the second triode there is a grid
current (200 µA), the plate-voltage of the first tube rises only to about 150 V. Corresponding
characteristics result for a supply-voltage of 275 V (Fig. 10.2.4).
Fig. 10.2.4: Grid-current (left, measured for three different tubes); transmission characteristic (right). The first
tube is driven via a 100-kΩ.grid-resistor.

With the change from the E- to the F-series, Fender replaced the 12AY7 by the 12AX7 (=
7025 = ECC83) – presumably because the latter has higher gain, or simply in order to
standardize. Bassman 5F6, Super 5F4, and Twin 5F8 still included the common-
cathode/plate-circuit for their intermediate amplifiers but received the 12AX7 instead of the
12AY7. In the Super 5F4 the associated components remained identical; for the other amps
Rk1 was decreased from 1.5 kΩ to merely 820 Ω. The differences between the two double
triodes are shown in Fig. 10.2.5: the 12AX7 sports the larger open-loop-gain (µ = 100 vs. 44)
but also has the larger internal impedance: 63 kΩ vs. 25 kΩ. Since the tubes are not operating
under open-loop conditions, the gain in reality differs not that much but still considerably: 50
vs. 30, i.e. 34.0 dB vs. 29.5 dB.
Fig. 10.2.5: Output characteristics (according to data sheets) of the 12AY7 (left) and the 12AX7 (right).
The transmission characteristic of the 5F4-circuit is shown in Fig. 10.2.6. Besides the steeper
slope (= higher voltage gain) it is especially the much stronger curvature that stands out – it is
the reason for strong non-linear distortion. The change to the smaller cathode-resistor (5F6)
balances the operating point somewhat but cannot change anything about the curvature. It
may be due to this non-linear behavior that Fender’s Super-Amp 5F4 received additional
negative feedback – but the Bassman 5F6 (and its successor 5F6-A) had to do without the
negative feedback. It needs to be noted that in particular this Bassman had a lasting influence
on the British amplifier industry: it was the amp that Jim Marshall modeled his JTM amps
after from 1962 (with cathode-follower, with 820-Ω-resistor, without additional negative
feedback).
Fig. 10.2.6: Left: comparison 12AX7 vs. 12AY7 (1.5 kΩ // 25µF). Right: comparison 820Ω vs.1.5 kΩ (// 25µF).
As in Fig. 10.2.4, the first tube was driven via a 100-kΩ-grid-resistor.

The first cathode-resistor (Fig. 10.2.7) determines the operating point of the first tube but the
individual tube data also have significant influence. In Fig. 10.2.7, we see the results of
measurements taken from several 12AX7 (Siemens, Valvo, Brimar, Mazda, Ultron, TAD).
There are clear differences in the transmission characteristics as well as in the time-functions
– this of course does dramatic effects on the level-dependencies of the harmonic distortion.
Still, the attributes good or bad may be assigned with great caution only. Whether single-
sided signal-limiting is preferred or objected to is a matter of taste, and the same holds for
whether new or old tubes are utilized. A stringent correlation between tube data and tube age
must not be expected – a clear correlation between tube price and tube age may be, though.
Fig. 10.2.7: Characteristics of 6 different 12AX7 tubes. Right: time functions (two different 12AX7).
First cathode-resistor = 820 Ω bridged with 25 µF; drive signal fed via a 100-kΩ-grid-resistor.
While we are on this subject: the opinion that tubes produced back in the day (NOS) are
better, and consequently of course more expensive, holds only for the latter part. There might
be something to the idea that the descendants of the old geniuses have plainly misplaced the
recipes and do not know anymore how to build a high-quality tube. New tubes might have
issues with microphonics, noise, a short lifespan, leaky seals, unsuitable getter♣, just to name
a few criteria. But variations in the transconductance? The formula higher transconductance
= better does certainly not work out, and a corresponding link to the price remains unclear, as
well. The overdrive-behavior that is so important for guitar amplifiers is not specified in any
data sheet for triodes, and, generally, neither is the grid current. A 12AX7 bought in 2008
may cost 6 € (advertised with tight bass, punchy mids and silky top end), or more than 13 €
(tight bass, punchy mids and silky top end with overall definition and brightness). Or it could
be priced at 25 € (great for warm clean tones and creamy overdrive). That is too expensive?
Here is a 20-€-tube with "great warm clean tones and fat overdrive with smooth top end".
Still not in your price-range? Hm … then maybe the 7-€-tube with "better gain and warm
tone", or the 8-€-tube with "good gain, lots of treble and tight bass response"? Blimey – I’ve
shelled out a 20-€-surcharge♥ for the tube-supplier scraping off the original labeling and
replacing it by his company logo – shouldn’t I be entitled a source to read up on the criteria
that the tube (now knighted as “selected”) will actually meet? Not a chance - "good gain", or
"slightly better gain than Nr. 5" … that’ll have to do. Or simply: "comes in the original RCA-
boxing". That will set you back at least 30 €, though. But the real winner is: "12AX7;
enlarged grid giving a better articulation in the bass-range. The helix-shaped heating
filament takes care of excellent noise-behavior and lowest microphonics” – at no less that 42
€ per piece! Hopefully that extended bass-range is worth this kind of money-drain – given
that the regular 12AX7 already extends down to 0 Hz. Of course, this sort of premium-stuff
might be exactly what you were searching for forever. But then, the 5-€-no-name tube might
have done the exact same trick. Faites vos jeux, ladies and gentlemen.
♣
materials that bind gas residues and improve the vacuum that way.
♥
Dear lawyers (including partners and colleagues in your firm scattered throughout the ROW): this is all just
irreal satire. Ain’t no spondu-licks coming through these tubes …

But now back to our actual topic: Fig. 10.2.8 shows the harmonic distortion of the signals in
Fig. 10.2.7. The differences between the left and right sections in Fig. 10.2.8 are due to just
swapping tubes: take out the 12AX7 – plug in another 12AX7. Left, the 2nd-order distortion
dominates up to -2.5 dBV; above that we see mainly 3rd-order distortion. Right, things are
very different: 2nd-order distortion up to -11 dBV; from there on, about the same share for 2nd-
and 3rd-order distortion. The closer the operating point gets to the end of the characteristic, the
more dominant the 2nd-order distortion becomes for small drive levels. An ideal one-way
rectifier (as an extreme example) would show only even-order distortion (k3 ≡ 0).
Fig. 10.2.8: Harmonic distortion of the signals of Fig. 10.2.7. 1st tube driven via a 100-kΩ -grid-resistor.
Harmonic distortion attenuation ak = 20 ⋅ lg(1/k), k = harmonic distortion factor. Larger dB-values indicate
smaller non-linear distortion. These figures are reserved for the printed version of this book.
Given such variances, wouldn’t it be worth the while to use selected tubes, after all? That
question is reason enough to check some offerings. A sample of 6 tubes sourced from a tube
supplier was measured using the circuit seen in Fig. 10.2.7; the results are shown in Fig.
10.2.9. The small signal gain varies from vU = 34.8 to 35.6 dB, and the operating points differ
by as much as 20 V. The differences in the maximum and minimum achievable voltage are of
similar magnitude, and thus in the symmetry of the curves, as well. “Asymmetry” would be
the better term: in this circuit, this type of tube will be the source of pronounced single-sided
distortion. Not that that’s entirely undesirable in a Marshall amp … however, the precise
reproduction of special distortion characteristics clearly is NOT warranted by the “selection”
of tubes – as can easily be seen from Fig. 10.2.12. Except for the attribute “selected tube”, no
actual selection criteria are made public, and we can only speculate what the basis of the
surcharge asked for these tubes could be. Maybe there is a selection for reduced microphonics
– not entirely useless, but not a first priority in an intermediate amplifier stage, either.
Fig. 10.2.9: Characteristic of 6 selected 12AX7 (supplier A); 4 of them in normalized presentation (right).

Fig. 10.2.10 shows measurements taken from 6 tubes provided by another supplier. The
curves are indeed closer to each other although there are still differences in the details. Small
signal gain is between 35.7 and 36 dB – a better match compared to our first example. The
voltage limits, however, include a similar scatter so that we do not have a uniform distortion
characteristic across several tubes, either (Fig. 10.2.12).
Fig. 10.2.10: Characteristic of 6 selected 12AX7 (supplier B); 4 of them in normalized presentation (right).
Last, let us take a look at 4 unselected tubes (all 4 from the same manufacturer), bought at a
low price from a component discounter (Fig. 10.2.11). The small-signal gain vU varies
between 33.3 and 33.4 dB i.e. the gain factor this is 2 dB less than in the other samples. This
can by no means be seen as a general deficit: whether the user prefers or dislikes the
corresponding (small) reduction of distortion is a purely subjective rating.
Fig. 10.2.11: Characteristic of 4 unselected 12AX7; 3 of them in normalized presentation (right).
In Fig. 10.2.12, again normalized transfer characteristics and harmonic distortion are brought
face to face. The first sample of “selected” tubes shows measurable variance in the gain and –
in particular – strong differences in the harmonic distortion; a common characteristic,
however, cannot be established. The second and the third samples show a group-specific
characteristic, but the variations within each group are still considerable – whether with or
without “selection”.
Of course, these measurements do not allow for the conclusion that all selected tubes offered
on the market do not merit the term; the samples used here are too small for that. Still,
inquiring about what the selection process in fact entails would appear to be highly advisable.

Fig. 10.2.12: Normalized transmission characteristics (left); harmonic distortion (right). Comp. Figs. 10.2.9-11.
Starting from the first Fender-circuits, the cathode-follower was subjected to two important
changes until it arrived in Jim Marshalls JTM: 12AY7 → 12AX7, and 1500 Ω → 820 Ω. For
the VOX AC-30TB, a third modification was added: the cathode-resistor at the cathode-
follower tube was reduced from 100 kΩ to 56 kΩ, with the result that even without any drive
signal, no less than 3 mA already flow through this tube. That is no laughing matter for a tube
specified to carry 1,2 mA in its operating point. It won’t be destroyed, but such a high current
cannot be generated without the presence of a grid-current. This cathode-follower tube does
not have a high-impedance input anymore but represents a non-linear load for the plate-circuit
of the preceding tube. The latter is required to deliver a grid current of almost 1 mA which,
considering that the plate resistor has a value of 100 kΩ, is no mean feat, and which will be
the source of a special non-linearity. As is always the case with this special amplifier type,
that might, however, not be generally undesirable.

Fig. 10.2.13 depicts the measurement results taken from the VOX-circuit: even without drive
signal, the cathode-follower requires a grid-current of 185 µA. Measuring the differential
input impedance (AC-resistance) of the cathode-follower resulted in the surprisingly small
value of a mere 90 kΩ! This impedance-converter apparently does not feature the “extremely
high” input-impedance typically found in such circuits, but is – due to its relatively high
quiescent plate-current – even of quite low impedance. At high plate-voltages (Ua1), it loads
down the preceding stage just like a 90-kΩ-resistor, and reduces the voltage gain of that stage
by a quite sizeable 28%. With decreasing plate-voltage (Ua1), the input-impedance of the
cathode-follower increases, after all; it therefore represents a non-linear load impedance. The
transmission characteristic is strongly curved and the output-voltage swing is relatively small.
This means that for large output voltages, the cathode-follower cannot provide enough
current, and for small input voltages, the first tube is not sufficiently low-impedance. Not
when using the 12AX7, anyway.
Fig. 10.2.13: Lefts: VOX AC-30TB. Middle: transmission characteristic of the overall circuit. For the
measurement, the first tube is driven via Rg1 = 100 kΩ. Right: grid-current of the cathode-follower tube.
Fig. 10.2.14 compares the summation- and the distortion-levels. The left-hand section shows
the situation at the un-loaded 1st tube while the right-hand section describes the non-linear
loading. The reduction of the summation level LΣ by 2,8 dB and the growth of the distortion is
clearly visible. Already at an input level of -15 dBV (178 mV), the 2nd harmonic (distortion)
is a mere 30 dB below the level of the primary signal (i.e. k2 = 3,2%). It will come as no
surprise that the internal impedance (output impedance) of this cathode-follower is not at a
by-the-book-value of 600 Ω but brings no less than 7 kΩ to the market: the operating point is
not positioned by-the-book, either! Nevertheless: 7 kΩ are o.k. for the VOX-circuitry.
Fig. 10.2.14: Output-summation-level LΣ, L2 and L3 of the VOX-circuit. Left: first half of the intermediate
amplifier only (i.e. without cathode-follower). Right: complete circuit with cathode-follower.

The unusual selection of the operating point of the cathode-follower is the reason for strong
2nd-order distortion (k2) showing up in the intermediate amplifier of the VOX. It is, however,
difficult to surmise that there is any intentional design in this – the details too clearly fail to be
reproducible. The non-linearity depends strongly on the supply-voltage, and on the individual
tube in use, and it therefore appears – from one individual amp to the next - with varying
distinction. (Fig. 10.2.15).
Fig. 10.2.15: Level (left) and harmonic distortion (right) of the VOX intermediate amp; 8 different 12AX7.
Grid-resistor in the first tube: Rg1 = 100 kΩ. Supply voltage: UB = 290V (compare to Fig. 10.2.13).
All distortion measurements of the VOX intermediate amplifier were done with Rk1 being
bridged with a capacitor. During the history of the AC-30TB-circuit, there has, however, been
a variant that fails to include this capacitor. With Ck = 25 µF, practically the whole relevant
frequency-range receives an increase in gain of about 7.5 dB, while the 0.68-µF-capacitor
found in some Marshall amps boost only the mids and highs (compare to Chapter 10.1). The
treble-loss occurring upwards of 10 kHz happens in the first tube (Rg1 plus Miller-effect). Fig.
10.2.16 compares the frequency-responses measured with and without the cathode-capacitor.
Fig. 10.2.16: Effect of the cathode-capacitor. In the VOX-circuit (left), the cathode-resistor is either bridged with
25 µF or left without a capacitor in parallel.
In the framework of discussing nonlinear distortion, the actual drive level is, obviously, of
significance – there is no consistent benchmark for this, though. Guitar, playing style, setting
of the tone- and volume-controls … all this determines the voltage arriving at the cathode-
follower. Subtle playing may bring down the voltage level to below -20 dBV: in this case the
non-linearity of the cathode-follower does not play any role. However, just turning up the
volume control halfway generates – with a Stratocaster played in a normal way – easily
voltage amplitudes of in excess of 1 V at the grid of the first tube in the two-tube-cathode-
follower circuit. In particular the picking-attack will generate strong non-linear distortion in
this scenario.

10.2.3 Mixing stage
Most guitar amplifiers feature more than a single “channel” i.e. there are several input jacks
that are associated with different amplifier branches. These branches may vary in sound, in
the distortion and/or in switchable effects. All branches are, however, fed to one and the same
power amplifier, and this requires that the respective signals be added. Rather then the term
“adding”, the term “mixing” is often used – note that this does not refer to the process of the
same name used in RF-engineering and designating circuits for frequency conversion. For the
present context, we mean: mixing = adding.
Fig. 10.2.17: Circuit concepts for signal addition: reverse-mode, standard-mode, active-mode (left to right).
Three often-implemented circuit concepts are shown in Fig. 10.2.17. The so-called reverse-
mode was often found in early amplifiers; it was soon replaced by the standard-mode. Passive
circuitry has the general disadvantage that the potentiometers influence each other: if the
volume control in one channel is fully up (α = 100%), and if the second volume control is
now also turned up (β = 100%), the gain factor of the first channel can be reduced by up to 6
dB because of the mutual loading between the two channels. Fig. 10.2.18 shows this influence
dependent on the center-tap position of the respective other potentiometer (β).
Fig. 10.2.18: Mutual influence of the two potentiometers; α = CH1, β = CH2. Figures assigned as in Fig. 10.2.17.
Potentiometer = 1 MΩ, mixing resistors = 220 kΩ and 270 kΩ, respectively. Passive modes: gain up to the
summation point. Active mode: gain incl. tube stage (v = -50).
The internal impedance of the sources (amounting to about 40 kΩ for triode-amplifier stages
in common-cathode configuration: tube // plate-resistor) has an effect on the “counter-side” as
the potentiometers are turned up, and attenuates the “other” signal. Additional summation-
resistors (in series with the potentiometer center-tap) reduce this effect for the standard-
mode. In the Fender Deluxe 6G3, for example, we see 220-kΩ-resistors at this point in the
circuit, but there are also amps that use 470 kΩ (e.g. the Bassman 6G6). Larger summation-
resistors give a higher-degree independence of the controls but do have the disadvantage that
noise is likely to increase, and that the treble-response will probably get worse. In the third
variant, the active-mode, a negative-feedback-resistor reduces the gain as well as the input
impedance (current-voltage-feedback). Given high open-loop gain and strong feedback, the
contra-lateral influence can be practically eliminated. A small dependency remains in the
typical tube amp with v = – (30 ... 50) but this is practice is of no bother. As another effect of
the negative feedback, maximum gain and harmonic distortion decrease.

Active mixing-stages are not often seen in guitar amplifiers: they surfaced in Fender amps in
the mid-1950’s (5E4, 5E5-A, 5D6-A) but disappeared again shortly afterwards. The standard-
mode is by far the most often used, with mixing resistors of 220 – 470 kΩ. Moderately
reducing the mixing resistors does not bring much advantage regarding the gain but increases
the upper cut-off frequency (while deteriorating the mutual interaction). With the
potentiometer center-tap positioned mid-way, the source-impedance that the following tube-
grid “sees” is about (P/4 + R)/2, with P = potentiometer-resistance and R = mixing resistor.
Typical values of this source impedance are found to be in the region of 250 kΩ. In
conjunction with the tube-input-capacitance (up to 150 pF due to the Miller-effect), a 1st-order
low-pass with a cutoff-frequency of 4 – 8 kHz results. In some amplifiers, the corresponding
slight treble-loss is counteracted via a bridging-capacitor that bridges potentiometer and/or
mixing resistor. This may be implemented only in one of the two channels because otherwise
the effect would suffer. Manufacturers like to designate the channel modified that way with
terms such as “Bright” or “Treble or “Instrument”, while the other channel is dubbed
“Standard” or “Normal”.
In Marshall's JTM-45, a guitar amplifier from the early 1960’s, the signal addition is done
via two 270-kΩ-resistors in the beginning – just like in the Fender the JTM was modeled
after. Soon, however, there is a change to 470-kΩ-resistors; these remain for several model
generations. To compensate the associated treble-loss, bridging capacitors with model-
specific value-variations are installed. The early Marshall amps were available in versions for
guitar (lead), for organ, for bass and for use as PA, with the technical distinction between
them mainly being the differing values of the bridging capacitors and the mixing resistors.
Fig. 10.2.19: Marshall-amplifier, adding stages with different-value components.
Fig. 10.2.19 shows three versions of the mixing stage; for the first (on the left), Fig. 10.2.20
indicates the frequency-responses for different positions of the respective volume-control.
The grey areas depict the ranges of mutual influence of the two controls. Depending on one’s
position in the hierarchy of Marshall-ites, these results may be interpreted as testimony to
genius manifoldness, or as ghastly circuitry-botch-up.
Fig. 10.2.20: Marshall JTM-45, mixing stage. Left: frequency-response of the “High Treble” channel,
right: “Normal”-channel. The grey areas show the mutual influence between the two volume-pots.

Fig. 10.2.21: Marshall Type 1987, mixing stage. Left: frequency-response of the “High Treble” channel,
right: “Normal”-channel. The grey areas show the mutual influence between the two volume-pots
In Abb. 10.2.21 we see the frequency-responses of the circuit shown on the right in Fig.
10.2.19. The change to the unusually large 5-nF-capacitor results in a special low-cut. Also, in
the upper range of the volume control (i.e. where the user usually “lives”), it operates almost
solely as an adjustable bass-cut. That is quite successful, as one can hear. The reduction of the
coupling capacitor to 2.2 nF makes for an additional low-cut. Since apparently the sound was
still not aggressive enough, the cathode-resistor was bridged not (as Fender would have it)
with a large electrolytic capacitor, but with a 680-nF-capacitor (Fig. 10.2.22) that makes this
stage run at maximum gain only for higher frequencies. At low frequencies, there is a slight
negative feedback. Some Marshall amps had a further capacitor to bridge the cathode-resistor
in the pre-amplifier, other completely dispensed with these caps. There is, after all, neither
“the” Marshall-circuit nor “the” Marshall-sound.
Fig. 10.2.22: Left: cathode-resistor bridged by a capacitor in the Marshall amp types 1987 and 1959. The right-
hand picture shows the treble boost resulting from the cathode-capacitor.

10.3 Tone controls 10-49
10.3 Tone-Controls
Just to state it right upfront: the secret of a great-sounding guitar amp does not lie in its tone
controls (tone-filter). Of course, these modules are necessary to adjust bass, middle and treble
to the subjective desires, but modifications to the tone controls normally will not convert a
bad amp into a great one.
The first guitar amps often had merely a simple treble-control. Fender’s Champ even had only
one solitary knob: Volume. If sound variations were indeed indispensable, you had to do them
on the guitar. The Deluxe at least already had a treble-control, and over the years, further
controls were added. In the 1950’s, your standard helping of tone-control included a Bass and
a Treble-knob, and later some chosen few received a middle-control in addition. Marshall
copies Fender’s tone-control circuit (with minor modifications), and in Jennings’ VOX-amps,
a comparable filter-stage is found. And there you have it: the glorious Big Three – most
subjectively chosen, of course. Trying to put together even only an approximately
representative selection of all tone-controls developed over the years would go WAY beyond
the scope planned here, and so we will limit ourselves to a only few circuits.
Set to their middle (“neutral”) position, the tone controls in a HiFi-amplifier need to give a
frequency-independent reproduction. The tone controls in a guitar amplifier do not have to
perform that way, because the amp is – together with the loudspeaker – still a part of the
sound generator and contributes to the sound. Although the tone controls may include
frequency-selective filtering of more than 20 dB, it is not the only filter-stage in a guitar amp.
The input capacitances of the tubes have (in conjunction with the usually high-impedance
circuitry) the effect of a treble-cut. Bridging capacitors (over-) compensate this via a treble-
boost. Intentionally small coupling capacitors attenuate the lows, as do small cathode-
capacitors. Frequency-selective negative feedback in the power stages yields brilliance, output
transformers may contribute resonance-accentuations and/or bass-cuts, and at the end of the
transmission chain we have the loudspeaker with its only weakly dampened resonances. No,
this transmission is everything but frequency independent – and that is what makes it so
desirable.
10.3.1 Bass-Middle-Treble
As an example for a passive tone control we chose a circuit that is included in many Fender-
amps, but (in more or less modified versions) also has found its way into amps by Ampeg,
Kitty Hawk, Marshall, Mesa Boogie, Music Man, Randall, Rickenbacker, Roland, Selmer,
Solton, VOX, and many more. The term “passive tone control” indicates that the frequency-
dependent filtering is done exclusively via passive components, i.e. by resistors and
capacitors. The tube-stages grouped around the tone control contribute frequency-independent
gain. As an approximation, we may ignore for now that this is not fully correct. In an active
tone control, the RC-network is integrated into the feedback loop of a tube, and corresponding
circuits have a significantly different structure. Fundamentally, inductances also count as
passive components – but they are not liked, due to their relatively large build. At the most
they are included as exotic birds, if at all.

In Fig. 10.3.1, we see a good example for a simple passive tone-filter. This circuit was
deployed in early Fender-amplifiers (e.g. the 5E4) but may be found in variations also in
radios and similar devices. Turning down the bass control (i.e. moving the tap in the figure
fully to the right) results in a readily comprehensible situation. What remains now is merely a
complex-valued voltage divider that can be further simplified if we take the load resistance as
infinite. The current becomes independent of the position of the center-tap, and depends on
the frequency only as a 1st-order function (cutoff frequency = 653 Hz), despite the presence of
two storage-elements. The output voltage, as multiplication of this current with the transverse
impedance, also merely has a 1st-order dependency on p = jω, and an appropriate adjustment
of the treble control even results in a 0-order-system with frequency independent transmission
(32.2 dB attenuation). The right-hand diagram in Fig. 10.3.1 shows the transmission functions
of the divider without load; the position of the center-tap is the parameter.
Fig. 10.3.1: Simple treble-filter. The circuit on the left was used in the 5E4 Super-Amp; the circuits to the right
are simplifications for the bass-control turned down. See also Fig. 10.3.3.
Introducing a load-impedance yields a 2nd-order transmission-function that, as an

approximation, can be seen as load-less divider with an additional high-pass (fg = 70 Hz). The
middle picture shows this scenario as a Bode-diagram with approximation lines. In the left-
hand picture, we see the complete magnitude-frequency-response. In a real circuit, it will be
necessary to consider the input capacitance of the subsequent tube; this capacitance can easily
amount up to 100 pF due to the Miller-effect. The resulting minor treble-attenuation is only
felt above about 10 kHz. Fig. 10.3.2 shows a peculiarity of the Fender-circuit that sets it apart
from the tone-filters usually found in audio-engineering: while the latter keep the cutoff
frequency constant and fan out the curves symmetrically, the cutoff-frequency for the Fender-
filter changes as the treble control is adjusted.
Fig. 10.3.2: Comparison of magnitude-frequency-responses: guitar amp (left), audio amplifier (middle and right).

Fig. 10.3.3 depicts the effect of the Fender tone-filter in 6 diagrams. Again, pronounced
differences compared to classic audio-filters are apparent: treble- and bass-attenuation
influence each other, and for bass and treble fully turned up, a rather selective mid-cut results.
The latter is a specialty that will remain in almost all later Fender amplifiers.
Fig. 10.3.3: Frequency-responses of the Filter circuit acc. to Fig. 10.3.1 (Fender Super-Amp 5E4, ca. 1955)
The structure of this tone-filter has some similarities to the mixing-stage discussed in Chapter
10.2.3: treble and bass are divided up into two parallel channels, then high- and low-pass
filtered, respectively, and finally added up again at the output. The 5-nF-capacitor shorts high
frequencies to ground; as such it has a function similar to that of the 10-nF-capacitor.
Combined with a desire to cut cost, it was presumably this similarity that led to a merging of
the two capacitor-branches. To keep the effect of the Treble filter when the Bass-control was
turned down, a resistor was required between the 10-nF-capacitor and ground … and you got
a tone-filter that makes do with only two capacitors (Fig. 10.3.4).
Fig. 10.3.4: Frequency-responses: tone-filter of the Super-Amp 6G4. Circuit given in Fig. 10.3.8.

Supposedly, the control options of this simple filter were seen as too limited, after all, because
very soon there was the revision 6G4-A (Fig. 10.3.5): an updated filter-circuit with no less
than 4 capacitors, and with a special treble-potentiometer sporting an additional tap.
Apparently, this development was worth the effort since the Tremolux (6G9) received it as
well, and since it was also used in the Bandmaster (6G7-A) and the Vibrolux (6G11), albeit
with small component modifications in the latter two.
Fig. 10.3.5: Frequency-responses: tone-filter of the 6G4-A. Circuit as in Fig. 10.3.8.
Nevertheless, the pot with the special tap disappeared again already in the following amplifier
generation, and around 1963 a circuit was developed that would go down in history as the
mother of all tone-filters – to be found in this or very similar configurations in VOX, Marshall
and many other guitar amplifiers (Fig. 10.3.6). In fact, the range of settings is not that big, but
it apparently fits the combination Fender-guitar + Fender-amplifier perfectly. The individual
component values are subject to variations at Fender as well as for the many copycats (in
particular the “mid-scoop” is shifted back and forth in its frequency position), but the basic
topology is now set.
Fig. 10.3.6: Frequency-responses: tone-filter at he beginning of the 1960’s. Circuit as in Fig. 10.3.8.

Also, the new filter circuit (AA763, Fig. 10.3.8) allows for the addition of a Middle-Control
in addition to Bass- and Treble-Controls. The required effort is rather small: the fixed 6.8-kΩ-
resistor of the first version is simply replaced by a 10-kΩ-potentiometer.
Fig. 10.3.7: Frequency-responses of the tone-filter with middle-control (RM = 500Ω). Compare to Fig. 10.3.6.
Fig. 10.3.8 documents the development of the Fender tone-circuit. The number of capacitors
changes from two to four until a simple 3-capacitor-circuit is found the topology of which to
this day is seen as a standard.
Fig. 10.3.8: Fender tone-filter circuits: 5E4, 6G4, 6G4-A, AA763 (left to right).
As mentioned, there were changes now and again in the values of the components of the
AA763-tone-filter: apart from the variations on the 6800-Ω-resistor (middle-pot), the 47-nF-
capcitor was subject to several modifications and varied from 22 nF to 33 nF and on to 47 nF.
The effect of this change in capacitance is shown in Fig. 10.3.9: if the Bass control is not
entirely turned down, the spectral components below 500 Hz are boosted by the reduction of
the capacitance. With the bass-pot at “0” nothing changes because in the relevant frequency-
range the parallel connection with the 100-nF-capacitor acts approximately as a short
compared to the 100-kΩ-resistor. This holds for 22 nF as well as for 47 nF. It is difficult to
find a clear criterion for the choice of this capacitor-value in Fender amps. Some amplifiers
such as the Showman or the Twin start with 47 nF in 1963 and keep that value. The
Bandmaster receives the 47-nF-capacitor in 1963 but 5 years later this is changed to 22 nF.
The Pro-Amp sports a 33-nF-capacitor to begin with (AA763), but that is changed to 47 nF
in the same year (AB763) – and 6 years later we find a 22-nF-capacitor. In the Super-Amp,
the capacitor-history is different: it starts out with 33 nF (AA763), then in the same year sees
the change to 22 nF. Yet another approach in the Deluxe: 33 nF in the AA763 and the change
to 47 nF in the same year. Must be magic …

Fig. 10.3.9: Differences between 22 nF (fine line) and 47 nF (heavy line) in Fender tone-filters; RM = 6800Ω.
These figures are reserved for the printed version of the book.
The frequency-responses shown so far do not consider the peripheral circuitry. The source
impedance of the preceding stage and the input impedance of the subsequent stage change the
curves – but not fundamentally; in fact the differences are rather marginal. In Fender
amplifiers, the source impedance typically amounts to 30 – 40 kΩ, which is low enough that
we approximately have a stiff voltage source. The load of the tone-circuit is either a high-
impedance tube-input or the volume-potentiometer. The latter is at 1 MΩ (rarely also 500 kΩ)
of sufficiently high impedance; the output of the tone circuit therefore can be seen as
operating without load. For the uppermost frequency-range, however, we do need to consider
the input capacitance of the subsequent tube. Due to the Miller-effect this has to be assumed
to be 100 – 150 pF, leading – in conjunction with 250 kΩ (see below) – to a cutoff frequency
of 6.4 or 4.2 kHz, respectively. The corresponding loss in brilliance makes itself felt most
when the center-tap of the volume potentiometer is set mid-way, because here the internal
impedance of the pot is largest (R/4 = e.g. 250 kΩ). In order to counteract the treble-loss,
already the first guitar amps had a bright-capacitor installed that bridged the upper part of
the volume pot. In its left-hand section, Fig. 10.3.10 shows the treble-loss due to the
capacitance, and in the right-hand section the effect of the bright-capacitor. The figure focuses
on the transmission characteristic given by source impedance (38 kΩ), volume pot (1 MΩ),
bright-capacitor (120 pF), and input capacitance (150 pF); the additional attenuation of the
tone-filter is not shown to maintain a straightforward display.
Fig. 10.3.10: Treble-loss due to capacitive loading of the volume pot (left), treble boost via bright-C (right)
Source impedance RQ = 38kΩ, 1-MΩ-potentiometer, input capacitance of the subsequent stage: 150 pF.

The Fender tone-filter designated AA763 is again shown in Fig. 10.3.11, this time in
comparison to two competitors that originated at approximately the same time: the VOX AC-
30TB and the Marshall JTM-45. The basic structure is identical but there are characteristic
variations in the details. For example, the Fender filter shuts the signal off completely with all
controls turned down fully – the other filters avoid this awkward property. The individual
component values differ substantially so that three distinct circuits emerged, after all – despite
all similarities.
Fig. 10.3.11: Comparison of tone-filter

circuits: Fender, VOX, Marshall. The
filters are loaded differently: high
impedance for Fender and MARSHALL,
360 kΩ for VOX (Miller capacitance to
be added to each).
In Fig. 10.3.12 we see the transmission characteristics of the VOX-Filter (AC-30TB). The
low-cut is particularly conspicuous; it is due to an RC high-pass not shown in the figure. The
Marshall-filter (Fig. 10.3.13) is different, again: the aim here apparently was a small
attenuation of the filter stage. (Translator’s note: incidentally, this Marshall-tone-circuit is a
direct copy of the circuit found in the last tweed Fender Bassman 5F6-A that had – in the
tone-control-department – similar advantages and disadvantages.) This attenuation is further
reduced in the subsequent versions of the amplifier (JTM-50, Fig. 10.3.14) by replacing the
56-kΩ-resistor by 33 kΩ and the 250-pF-capacitor by 500 pF.
Fig. 10.3.12: Frequency-responses: the tone-filter of the VOX AC-30TB (incl. 580-Hz-high-pass).
The differences in tone of the three amplifiers under scrutiny here are, however, not
principally based on the different filter circuits. Only several stages cooperating make for the
individual sound. For example, the high-impedance power-amp output of the AC-30TB
results in a strong bass-boost (Chapter 10.5.7) that is found in Fenders only to a much smaller
degree. Marshall amps, on the other hand, offer the presence filter integrated into the power
amplifier stage; it brings a special treble-boost that the VOX lacks. We find further
differences in the overdrive-behavior and in the loudspeakers used: the latter typically work in
an open combo-cabinet in the Fender and VOX amps – for the Marshall, however, the bass-
heavy 4x12-enclosure is employed. While the tone-filter is a substantial part of the overall
system, its respective special realization should not be credited with any exaggerated
importance.

Fig. 10.3.13: Marshall JTM-45. The Treble-boost from preceding stages is not considered.
Fig. 10.3.14: Marshall JTM-50. The Treble-boost from preceding stages is not considered.

The following examples show that, in tone-filters, “more” is not necessarily “better”: in the
Fender circuit we find two to four capacitors but the Sound-City-filter has six of them! Or
even 10, as shown in Fig. 10.3.15. Not bad, but short lived. If this filter structure were
superior, it would have asserted itself in products by the competition, as well – but that didn’t
happen, and the circuits disappeared again from the market.
Fig. 10.3.15: tone-filters in

Sound-City amps:
Left: CS100B.
Right: L/B 120 Mark IV.
Simple tone-filters do not stand in the way of creating a convincing amplifier, as the
Marshall 18-Watt-amp (examined in the following) proves. This amplifier was produced
from 1965 – 1967 and has a lot of fans despite its rather spartan filter-network. In the
“Normal”-channel we find a single tone control: cut either treble, or bass – that’s it. Similarly,
there is only one simple Tone-knob in the “Tremolo”-channel: more treble or less treble,
interactively coupled to the volume-pot.
Fig.10.3.16: Frequency-response of the Marshall 18-W-amp; left: “Normal” channel; right: “Tremolo” channel.
Very similar circuit concepts are found already 10 years earlier in the Fender “Deluxe-Amp”
amplifier; the volume-pot is merely connected “in reverse” to facilitate the connection of a
second channel. Even today, these very simple old amps are not at all “out” but enjoy cult-
status in the use for club-gigs or in the studio. Very obviously, a complicated tone-filter is not
necessary to amplify an electric guitar. Question to Lenny Kravitz♣: "How do you get this
tone?“ Answer: "Well, you just plug an Epiphone into a Tweed Deluxe, crank it to 10 … and
that’s it.”
On the other end of the spectrum of complexity we find amplifiers that offer almost infinite
variability using multi-band graphical and/or parametric equalizers (Chapter 10.3.2). They are
predestinated for the creation of very “different” sounds, but the majority of guitar players
seem to be able to do without them.
♣
Gitarre&Bass 06/04

10.3.2 Equalizer (EQ)
A filter that allows for narrow-band changes in the spectrum (or in the transmission function)
is called an equalizer. Besides a basic gain that we assume to be 1 ( ) in the following,
there are 3 parameters that define the transmission behavior of an equalizer: center-frequency,
boost and Q-factor (Fig. 10.3.17) The center-frequency fx is the frequency at which the gain
assumes it maximum (or minimum) value, the boost β specifies the gain at fx, and Q-factor Q
determines the bandwidth. For a so-called parametric equalizer (EQ), all three parameters are
adjustable while for a so-called graphic EQ, only β is variable, with fx and Q fixed at
predetermined values.
Fig. 10.3.17: Equalizer characteristic. B = 20⋅lg(β) = [-12 -9 -6 -3 0 3 6 9 12]dB, fx = 1 kHz.
In Fig. 10.3.17 we see two different groups of curves. fx and B are self-explanatory, but the Q-
factor requires some supplementary comments. Often, the Q-factor is determined from the
relative bandwidth measured as the distance of the -3-dB-points on the graph. This definition
is, however, useless for an EQ e.g. because for a 2 dB-boost no -3-dB-points can be defined at
all. The correct definition results from the transmission function H:
As can be seen, this filter has a pole-Q-factor QN and a zero-Q-factor QZ. For f = fx, we get
b = QN / QZ. In order to define one single Q-factor for an equalizer, an infinite number of
possibilities present themselves; customary are two (different!) definitions. Either we keep the
denominator-Q-factor constant and vary the boost-factor via the numerator-Q-factor; this
filter-type is called constant-Q-equalizer, and the denominator-Q-factor is specified as the
Q-factor of the equalizer. Or we link numerator- and denominator-Q-factors via
and ; in this case we specify as Q-factor of the equalizer: .
Connecting two equalizer of the second variety in series with fx and Q correspondingly
identical in both EQs, and the boost-factors set reciprocally (β1 = 1/β2), the effects of these
two equalizers compensate each other completely. They are inverse to each other, and
therefore this EQ-type is also called inverse EQ (the filter shown in Fig. 10.3.17 is of this
type). For the constant-Q-equalizer, however, a corresponding series-connection does not lead
to a complete compensation: the attenuation is of a smaller bandwidth than the amplification
(Fig.10.3.18). These differences (if they are of any importance at all) play a role only for
graphic EQs, because all parameters can be freely adjusted in the parametric EQ, anyway.

Fig. 10.3.18: Characteristic of a Constant-Q-Equalizer. The specified Q is the denominator-Q.
The constant-Q-equalizer is held in high esteem because the Q-factor does not increase as the
boost-factor grows but remains constant independent of the boost. It should be added that it is
the denominator-Q-factor that remains constant because the numerator-Q-factor of course
does change. It is not entirely far-fetched to give priority to the denominator-Q over the
numerator-Q because the decay-coefficient determining the time-envelope of a step- or an
impulse-response indeed does depend only on the denominator-Q. However, whether it is in
fact desirable that abutting EQ-bands show a boost-dependent, more or less pronounced
overlap as depicted in Fig. 10.3.18, needs to be determined on a case-by-case basis according
to individual preferences.
Fig. 10.3.19: Series-connection of two constant-Q-equalizers. Single filter (----) and series connection (––––).
For the gain to add up to 0, both Q-factors need to be reciprocal (right-hand picture).
Fig. 10.3.20 shows a circuit often utilized for designing graphic EQs. The frequency-
dependent impedance Z of the resonant circuit may be realized in a passive (RLC) or an active
manner; the latter via adding an additional amplifier. The boost-factor can be controlled with
the potentiometer P, the center-frequency and the Q-factor are pre-set by the circuit design.
Fig. 10.3.20: Active EQ-circuit. The series-resonance-circuit (Z) may be realized via either active circuit.
The active resonant circuits are approximations of an ideal series-resonance circuit

The circuit presented in Fig. 10.3.20 offers the possibility to vary Q (within certain limits)
depending on the boost-factor (Fig. 10.3.21). As can be seen, we obtain inverse behavior with
a bandwidth varying in detail. Relatively high impedance in the potentiometers results in the
characteristic as show on the right, and low-impedance pots give the curves on the left. For
linear potentiometers, the boost-value changes predominantly towards the end to the control
path – therefore special pots with an S-shaped characteristic are required.
Fig. 10.3.21: Transmission characteristics of the EQ-circuit according to Fig. 10.3.20.
A multi-band graphic EQ may be designed with little effort by adding into the circuit
according to Fig. 10.3.20 further potentiometers with corresponding different resonant
circuits. Fig. 10.3.22 has the corresponding diagrams for various settings.
Fig. 10.3.22: Octave-equalizer: single filter (upper left). Six-band EQ, boost only in the 1-kHz-channel (u. right).
Boost only in 3 bands (lower left). Boost increasing with frequency (l. right).

10.3.3 Presence-Control
In studio-electronics, the term “presence” often characterizes the frequency-range between

about 1 kHz and 4 kHz, and a “presence filter” designates an equalizer operating in this
range. In guitar amplifiers, however, the presence-control represents an alternative to the
treble-control. An early variant of the presence-control is found in Leo Fender’s Bassman:
already the early versions (e.g. 5B6) include negative feedback in the power amplifier, and
this becomes frequency-dependent in the model 5D6. Presumably an additional treble boost
was desirable. There already was a treble-control so a different designation had to be found:
presence-control.
Having picked the Bassman as a model for his JTM-45, Jim Marshall (or rather Jim’s tech
Ken Bran) adopts this presence-filter, as well. Only VOX takes the opposite approach: since
the AC-30 already boosts the treble almost too much, the power amp here receives a treble-
attenuator designated with “Cut”. In the Fender- and Marshall-amps, the presence-filter
operates on the basis of a simple principle: a low-pass integrated into the negative-feedback-
loop diminishes the loop-gain for high frequencies, and boosts the treble that way. However,
despite their simple function, the circuit includes two special aspects. First, the loudspeaker
needs to be considered as part of the negative-feedback-loop: its impedance contributes to the
effect of the presence-filter. Second, the power-amplifier of a guitar amp is often subject to
overdrive. The presence filter becomes part of a non-linear system the tonal effects of which
are different from those of the treble-control.
Fig. 10.3.23: Effect of the presence filter in the Marshall JTM-45. In the measurement on the left, the 16-Ω-
output was loaded with a 16-Ω-resistor whereas on the right the load was a 4x12 speaker box (1960 AX).
In Fig. 10.3.23 we see measurements on the JTM-45. The generator-signal was fed to the
input of the differential amplifier; measurements were taken at the output of the power-stage.
In one case the load was a 16-Ω-resistor; in the other case a loudspeaker-box was used. The
latter is specified at 16 Ω, as well, but does not have constant impedance; rather, its
impedance is frequency dependent.

10.4 Phase-Splitter
A single power tube (class-A operation) allows only for small output power. High power
needs push-pull operation (Chapter 10.5). A push-pull output stage requires two drive signals
shifted by 180° relative to each other. These two anti-phase signals are generated in the so-
called phase-splitter circuit using one or two tubes. In essence, there are three circuit-
concepts: the tube operating with v = –1 in common-cathode configuration (paraphase-
circuit), the cathodyne circuit, and the differential amplifier in common-grid configuration.
10.4.1 Common-cathode circuit (paraphase)
This is a simple concept: one triode provides amplification with its plate-voltage serving both
as drive-signal for one of the two output tubes, and – attenuated via resistors – as drive signal
for the other triode. The latter feeds its (opposite-phase) plate-voltage to the other power tube
(Fig. 10.4.1).).
Fig. 10.4.1: Phase-inverter in common-cathode configuration. Right: modified version with negative feedback.
This basic paraphase circuit is predominantly found in early guitar amplifiers (e.g. the 1947
Fender Deluxe). It was soon first modified and then replaced by the cathodyne circuit. The
advantage of the paraphase circuit lies in its high voltage gain and the relatively large output
voltage swing of the two tubes. Disadvantageous is that the magnitudes of the output voltages
are not exactly equal but depend significantly on the individual tube data. Matching the
divider resistors leads to an individual symmetry, but this would have to be checked and re-
checked as the tube ages. Of course, it is an entirely different question whether a guitar
amplifier actually sounds best with complete symmetry of the output stage – however even if
a lack of symmetry would be desired, this would have to be specific and not subject to
random tube-variance.
The typical paraphase circuit – as it is found e.g. in the old Fender Deluxe (5B3) – attenuates
the output AC-voltage of the first tube with a 250-kΩ/7.0-kΩ-divider by a factor of 1/44. For
a precise calculation, the internal impedance of the first triode must be added in – this is
approximately 50 kΩ. The second triode amplifies this attenuated voltage by a factor of -44,
making available two AC-voltages of equal amplitude and opposite phase that drive the
output tubes. That would be the ideal case, anyway – in reality, however, the gain of the
second tube has significant scatter.

10.4 Phase splitter 10-63
If the voltage gain of the second tube is not at its nominal value but e.g. too small by 20%, the
two half-waves generated by the power amp also differ by 20%. The consequence is that this
effect alone is cause for harmonic distortion of 4%. One may feel good or bad about such
asymmetry – at Fender, it was not liked. The voltage divider at the grid of the second triode
was replaced by a current/voltage negative feedback: the plate-voltage is tapped (via 270 kΩ)
and generates an additional current in the grid-circuit. Fig. 10.4.2 depicts the circuit of the
Fender Deluxe 5D3; it is also found on other Fender amps of the same era (Super Amp 5D4,
Pro Amp 5D5, Twin 5D8).
Fig. 10.4.2: Paraphase-circuit with current/voltage negative-feedback (Fender Deluxe 5D3, 1954).
The principle of the current/voltage negative-feedback is also used in the inverting OP (right-
hand section of the figure): for an OP-gain approaching infinity, the voltage across R2
becomes close to zero; U2/U1 is merely defined by the relationship of the resistances and not
by the gain anymore [e.g. Tietze/Schenk]. For a tube circuit, this simplification holds only
approximately – but the basic operation is the same: if the open-loop gain of the second triode
changes by 10%, the ratio of the two (opposite-phase) output voltages changes by merely 1%
due to the negative feedback. The latter stabilizes the ratio U2/U1 of the two output voltages –
the circuit is termed “self-balancing paraphase circuit”.
The negative feedback has a further effect: it reduces the internal impedance of the right-
hand triode. With a load, the plate-AC-voltage of the triode on the right becomes smaller and
consequently the voltage fed back via the 270-kΩ-resistor decreases also, resulting in a
overall larger voltage gain. To some extent at least, the load-dependent decrease in the plate-
voltage is compensated. The internal impedance of the triode-circuit on the left (Fig. 10.4.2) is
simply the parallel connection of the internal impedance of the tube (e.g. 63 kΩ) and the plate
resistor (e.g. 100 kΩ) – i.e. about 39 kΩ in our example. Considering the load (about 220
kΩ), as well, brings us to Ri1 ≈ 33 kΩ for the overall circuit. For the right-hand tube, the
calculation yields Ri2 ≈ 12 kΩ (including load). The negative feedback has therefore reduced
the internal impedance of the second triode-system to about 1/3rd. As long as the loading of
the two paraphase outputs is negligible, the differing internal impedances do not play any
role. However, the input capacitances of the power tubes and the occurrence of grid-currents
can lead to load situations that cause considerable asymmetries.
Furthermore, it is necessary to consider that the input signal to one output tubes passes one
RC-high-pass, while the input signal to the other output tube passes though two such filters,
causing phase shifts in the low-frequency range. Similar effects happen at high frequencies:
the detour via the second triode-system acts as an additional low-pass that causes phase shifts
in the high-frequency range.

Fig. 10.4.3 shows the output voltages of a paraphase circuit having no negative feedback. For
small drive levels, we indeed get two phase-opposed voltages of approximately equal
amplitude. With increasing drive levels, triode-clipping starts to become visible – this shifts
the operating point across the coupling capacitor. In the lower line of the figure, we see
power-tube grid-currents (occurring from about +20 V) that limit the voltage-curves in the
direction of positive values. Because the signal of the second triode is derived from the
clipped plate-voltage, the second output signal is limited towards negative values, as well.
The overdrive of the output tubes consequently is asymmetrical.
Fig. 10.4.3: Measurements on a paraphase-stage without negative feedback: 1st tube (–––), 2nd tube (---).
Top: no grid-current limiting. Bottom; grid-current happening from 20 V. Supply-voltage for the triodes: 260 V.
Fig. 10.4.4 represents the corresponding measurements of a paraphase stage with negative
feedback. We again see the different drive situations of the two power-tubes in non-linear
operation. Also, the change in the duty-factor already recognizable in Fig. 10.4.3 reappears.
Fig. 10.4.4: Measurements on a paraphase stage with negative feedback: 1st tube (–––), 2nd Tube (---).
Top: no grid-current limiting. Bottom; grid-current happening from 20 V. Supply-voltage for the triodes: 235 V.

10.4.2 Cathodyne-circuit (split-load)
The cathodyne circuit takes advantage of the opposite-phase-situation of the AC-voltages at

cathode- and anode. Assuming a drive situation with a grid-current of zero, the cathode-
current is equal to the plate-current, and therefore voltages across equal cathode- and plate-
resistors will also be of the exact same amount – irrespective of any tube variances.
Textbooks on circuit design tend to explain the cathodyne configuration by separating the
plate-resistance into two “exactly” equal halves that then result in the new plate-resistance and
cathode-resistance, respectively. It is possible that this approach led to the designers using
high-precision resistors in the cathodyne-stage. For example, the schematic for the Ampeg B-
42-X specifies: all resistors 10% – however, the caption of the 47-kΩ-cathodyne-resistors and
the subsequent 100-kΩ-load resistors reads 5%. There were even amplifiers requiring a
resistor-tolerance as low as 2% for this circuit.
; ;
Fig. 10.4.5: Cathodyne-circuit. Signals taken directly from the cathode as is typical for Fender.
In Abb. 10.4.5 we see a guitar-amplifier-typical cathodyne-circuit. In Fender amps, both load

resistors (R) normally have a value of 56 kΩ with Rk = 1.5 kΩ and a grid-resistor of 1 MΩ.
Several Fender amps received this circuit in 1955 (Deluxe, Super, Pro, Bassman, Twin) but it
was only about two years until the arrival of the differential amplifier (more in chapter
10.4.3). The grid-resistor Rg of the circuit in Fig. 10.4.5 is connected to the split cathode-
resistor rather than to ground. This negative-feedback arrangement substantially increases the
input impedance RE (in the example to about 18 MΩ). It is questionable whether the designer
at Fender was aware: the coupling capacitor feeding the grid is, after all, 20 nF, just as
customary with 1-MΩ-inputs. The 1-MΩ-resistor is, however, not connected to ground but to
an almost equally big coherent AC voltage, and thus the effective input impedance increases
(bootstrap). The 10 nF and 18 MΩ component values results in a high-pass cutoff-frequency
of 0,4 Hz – quite generous for a guitar amplifier. Gibson used, in their GA-19-RVT, a
capacitor of merely 500 pF for the cathodyne input capacitor – maybe they knew more?
The voltage gain from grid to cathode is about 1 – 3/µ, with µ = open-loop gain of the tube.
For the ECC83 follows, with good approximation: vK = 0.97. As is typical for Fender, the
amount of the plate-AC-voltage is slightly less, about vA = –0.945. The internal impedances
of both outputs are, however, highly different: at the plate we have (with good approximation)
56 kΩ (negative current-feedback at the cathode), while no more than about 1.2 kΩ are
present at the cathode (cathode-follower). Amplifier tubes are often said to present no load to
the preceding circuits, and if that were always correct, the differences between the internal
impedances would be irrelevant. However, grid-currents may flow in the power tubes, and if
that is the case, plate- and cathode-voltages in the cathodyne stage start to be different.

An AC-relevant plate- or cathode-load has different effects on the respective other electrode:
a cathode-loading would increase the plate-current and thus grid-to-plate gain, while a plate-
loading would decrease this gain. Both types of loading would however have only little
impact on the grid-to-cathode gain (negative feedback). The cathodyne-stage does experience
loading by the output tubes. The latter are showing a high input-impedance only as long as the
power-tube grid is sufficiently negative relative to the power-tube cathode. At full drive
levels, and in particular in a state of overdrive, grid-currents do flow, and the cathodyne stage
operates with a non-linear load.
Fig. 10.4.6 shows the time-functions of the plate- and the cathode-voltages for different drive-
levels – first without the loading effect the output tubes have. Compared to the paraphase-
circuit, the maximum voltages are smaller but the symmetry is better. As we include the
loading by the power tubes (6V6, Fig. 10.4.7), the shape of the plate-voltage changes due to
the grid-current-drain via the cathode – this increases the plate-current and consequently the
voltage drop across the plate-resistor. In the cathode-voltage, there is practically no
corresponding protrusion because the voltage gain of the cathode-follower is only marginally
influenced by the plate-resistance. A typical effect found in tube amplifiers is shown in the
last line of the figure: the supply-voltage decreases with increasing overdrive (“sagging”).
Therefore, the minimum voltage is not constant but depends on the filter-circuit in the power-
supply.
Fig. 10.4.6: Cathodyne-stage without load; AC-component. Plate-voltage (----), cathode-voltage (–––).
Fig. 10.4.7: Cathodyne-stage with load; AC-component. Plate-voltage (----), cathode-voltage (–––).
The bottom right-hand picture shows the situation after longer-term overdrive.

10.4.3 Differential amplifier (long-tail)
This type of circuit unites two different basic tube-amplifier-concepts: the first tube works in
a common-cathode configuration with current-based negative feedback; the second tube
operates in common-grid configuration and is driven by the first tube via the cathode. In
Fender-history, the differential amplifier represents the final step in series of developments:
paraphase (1946 – 1951), paraphase with negative feedback (1951 – 1954), cathodyne (1955
– 1957), and differential amplifier (from 1956). Other manufacturers, such as e.g. VOX
(1958) or Marshall (1962) that start amplifier production more than a decade later than
Fender, use the differential amplifier right from the start.
Fig. 10.4.8: Differential amplifier with negative feedback via the cathode (Geko = negative feedback).
The left section of Fig. 10.4.8 shows the basic arrangement of the differential amplifier.
Driving the left tube with an AC-voltage changes its plate- and cathode-currents and thus
creates a voltage-drop at the plate- and cathode-resistors. The cathode-voltage of the left tube
changes the drive-voltage of the right-hand tube, as well, and also here causes changes in the
plate- and cathode-currents (common-grid-circuit). An example: if the grid-voltage (defined
against ground) of the left tube rises by 2 mV, the cathode-voltage increases by 1 mV. Its
grid-to-cathode-voltage therefore has increased by 1 mV while the grid-to-cathode-voltage of
the right tube has decreased by 1 mV. For identical transconductances of the tubes, the result
would be plate-voltages of the same amplitude but opposite phase. Text-books like to use this
example – but it does have a flaw: the sum of the changes of the plate-currents would be zero,
and the cathode-potential would remain constant, i.e the right tube would not receive a drive
signal. We can introduce a small correction to make the example work: the left grid-potential
rises by 3 mV, the cathode-potential by 1 mV, the plate-voltages are of opposite phase … but
not of the same amplitude anymore! Given typical component values, the AC-voltage-gain of
the right tube would be only about half of that of the left tube, plus it would be rather strongly
dependent on individual tube data. For this reason, the cathode-resistor is increased. This
reduces the gain of the two tubes, but also the dependency on the individual tube (current-
based negative feedback). The middle section in Fig. 10.4.8 shows such a circuit (VOX AC-
30), the right-hand section also presents an input for a negative-feedback (NFB) loop that
would be closed via a line from the output transformer (Marshall, Fender from 1956).
For the typical tube for the differential amplifier, Fender uses the 12AX7 (7025, ECC83) first
but then changes (in the Blackface era) to the lower-impedance 12AT7 (ECC81). VOX uses
the ECC83 (12AX7); Marshall does, as well.
DATA-SHEET SPECIFICATIONS: Internal impedance = 30 kΩ (ECC81) and 63 kΩ (ECC83).

An exact analysis of the differential-amplifier circuit shows that the voltage gains of the two
tubes are different, despite the negative feedback. In a typical Fender configuration (Pro Amp
AA763: Ra = 100 kΩ, Rg = 1 MΩ, Rk = 470 Ω, RB = 27 kΩ), this difference is about 7%. It is
likely that for this reason one of the plate-resistors (Ra1) was changed to 82 kΩ in a later
model (Pro Reverb AA 165). For the following variant (AB 668), the plate-resistors are again
equal in value but have merely 47 kΩ – and this arrangement remains for some time. VOX
uses two resistors of equal value (and completely dispenses with any overall negative
feedback!), while Marshall mostly employs the 82k/100k-pairing, and a frequency-dependent
overall negative feedback.
The grid-resistor Rg of the first tube usually has 1 MΩ; this value was probably also seen as
the input impedance. With a 10-nF-coupling-capacitor (e.g. Fender Twin 5F8A), a high-pass
cutoff-frequency of 8 Hz would result – that is very low for a guitar amp but certainly
compatible with the HiFi-preachings of the day. The negative feedback (RB), however, does
not only decrease the voltage gain, but it also increases the input impedance (bootstrap) from
1 MΩ to 2 MΩ, pushing the cutoff-frequency to a subsonic 4 Hz. That would more than
suffice even for a bass amplifier, and indeed the 5F6-Bassman includes the 20-nF-coupling-
capacitor, as well. But: a few years later the 6G6-B-Bassman receives a coupling-capacitor of
a mere 500 pF! The calculation would yield a high 160 Hz as the lower cutoff-frequency, but
we must not overlook that a second negative feedback loop is operating besides the feedback
via the cathode. This complicates the calculation because further phase-shifting RC-circuits
are in the game, and in particular the output transformer requires consideration. We had only
the schematic of the 6G6-B-Bassman and no original amplifier at our disposal so no
quantitative elaborations shall be included here. Just this general statement: Fender used very
different capacitances (250 pF – 20 nF) for the input capacitor (C1) of the differential
amplifier; the actual high-pass cutoff-frequencies of these different circuits should be
measured and not just calculated from the schematics. By the way: C1 is 47 nF in the AC-30
and 22 nF in the Marshall.
In Fig. 10.4.9, the grid-voltages of a Fender Super-Reverb are shown for three different drive
levels. For a small drive level, the two signals show minor differences in their amplitudes but
at high drive levels there is a significant asymmetry. We could ignore the differences in the
limiting towards negative voltages because the respective output tube will be in cut-off state
anyway; however, due to the differences in the DC-component in the two drive-signals the
two coupling-capacitors are polarized differently, leading to different duty-cycles in the plate-
currents of the power amplifier. In Chapter 10.4.4, we will take an in-depth look at this
asymmetry caused by the grid-current.
Fig. 10.4.9: Measurements at the differential amp of a Fender Super-Reverb (AB-763, negative feedback
deactivated). Power-tube bias = –50V. Grid-voltage of the 1st power tube (V7 = –––), and of the 2nd power tube
(V8 = ----). On the left, undistorted cosine-oscillations are shown for comparison.

10.4.4 Half-wave anti-symmetry
Each of the two power tubes generates both even-order and odd-order distortions; however, as
the two separately generated half-waves are superimposed, the even-order distortions cancel
each other out (half-wave anti-symmetry, Fourier-transform). This would be the ideal
scenario that would require:
• the output voltages of the phase-inverter to be as similar as possible,
• the power-tubes to be as similar as possible (i.e. paired),
• the primary windings of the output transformer to be as equal as possible.
Classical amplifier technology offers solutions for signal amplification with as little distortion
as possible, and regards the minimization of the even-order distortion as an advantage of the
push-pull power stage. We will not investigate here whether even-order distortion (i.e. k2, k4,
etc.) sounds good or bad in a guitar-amplifier – that would be a subject for psychoacoustics
(Chapter 10.8). The following analyses will focus on the question how far the distortion-
minimization is in fact successful.
Within the push-pull Class-B power stage (Chapter 10.5.3), the signal is spit into two parallel,
opposite-phase signal paths – each power tube amplifies only one half-wave. The
superposition towards the overall signal happens in the output transformer (Fig. 10.1.10).
Ideally, no error at all would occur in this process with all spectral lines except the 1st
harmonic cancelling each other out in the superposition. Of course, the splitting and re-
composition will not work flawlessly in reality, and non-linear distortion will appear.
Fig. 10.4.10: Time functions (left) and spectra of the half-wave signals. The signs of the Fourier-components are
the same only for the 1st harmonic, and consequently only this component remains after the addition.
An obvious error results from the unequal amplification of the two half-waves (Fig. 10.4.11).
The compensation of the even-order harmonics is incomplete and even-order distortion
remains (k2 ≈ 8% in the picture).
Fig. 10.4.11: Time function and spectrum of a signal with different amplification of the two half-waves.

For the time function shown in Fig. 10.1.11, the two half-waves have different amplitudes –
they are, however, not half-wave anti-symmetric. Half-wave anti-symmetry stands for a
time-periodic signal repeating itself, with inverted sign, after half a signal-period: u(t) = –u(t +
T/2). From the rules of the Fourier-transform, it directly follows that such a signal can only
contain odd harmonics. Consequently, only distortion products of odd order (k3, k5, k7 etc.)
can be generated as long as the transmission characteristics of the two half-wave transmission
branches are equal. “Asymmetry♣”, however, already starts in the phase-splitter stage for the
drive signals. The two gains in the paraphase-branches (Chapter 10.4.1) are as different as the
two tube-systems in the double-triode – that’s why quite early on the doctor (or rather Leo F.)
ordered a negative-feedback loop. Cathodyne-circuit and differential amplifier show much
less dependency on the individual tube data, and in fact they could deliver two signals equal
in amplitude and opposed in phase with sufficient precision – but only as long as there are
negligible grid-currents. Why do we find asymmetries already in the schematics, why do the
gain factors differ for the two half-waves, even for ideal tubes? Answers have been and
remain speculative:
1. the designers of early circuits were not yet that well versed in electronics, and later the
archetypes continued to be simply (and indiscriminately) copied.
2. these intentional “asymmetries” were supposed to give a special sound.
3. these asymmetries were supposed to correct other asymmetries in the circuit.
4. guitar amplifiers are no instrumentation devices; high accuracy was not that important.
Ad 1: This assumption cannot entirely be brushed off. Leo Fender’s explanations regarding
magnetism are … well, to be fair … they’re what you would expect given that he was
originally trained as a bookkeeper (one with aspects of a genius, without a doubt). But early
on improvements creep into the circuits (whoever developed them): the paraphase circuit
with negative feedback appears around 1954 in the Fender Deluxe i.e. it was desirable that the
asymmetries created by the tube-variances didn’t take over too much. Balancing a power
amplifier can be done without any grand network-analysis: with an oscilloscope and a
resistor-decade you come already pretty far, and such equipment was probably available even
in the labs (or workshops, rather) of the early protagonists.
Ad 2: That is an alluring thought but it asks for a bit of dispute. On the one hand: your regular
musician (or customer) will not be able (or willing) to un- and re-solder resistors after each
tube-change. If the asymmetry mentioned above were decisive for the sound, it would be
purely accidental because no circuit will totally equalize out the tube variances (in particular
those of the power tubes). We would have a contradiction to the objective of achieving a
special, sought after sound. On the other hand: this is exactly why musicians will choose that
one best-sounding amp from a group of 5 Deluxes (or Super-Reverbs, or Twins …).
Understandably, you are not allowed to ask whether this amp can be switched on ever again at
all (so that the tubes may not age, and to preserve the incomparable sound). “Just buy some
more NOS-tubes” – that’s what advertising will recommend.
Ad 3: there may be some truth to that, was well – possibly connected to 1. A designer
discovers that the phase-splitter stage needs to work in an un-balanced mode to obtain a fully
symmetric signal at the speaker output. Maybe the output transformer has a special
asymmetry? Not because the winding-machine has failed to count correctly, but because there
are slightly different (magnetic) coupling factors. Indeed, that may be compensated via the
phase-splitter stage – but of course only as long as the transformer data always remain the
same.
♣
we could call this “un-anti-symmetry“ just as well

Ad 4: Of course, every designer gets to the point where additional effort is not sensibly
warranted anymore in view of the costs additionally incurred. Although: a 100-kΩ-resistor
costs just as much as an 82-kΩ-resistor. Following-up the development of resistor-values in
the phase-splitter over the years, we easily recognize the fight for the “optimum solution”
(Chapter 10.4.3). Overall-negative-feedback approaches that include even asymmetries in the
magnetic fields bear testimony to the desire for reducing non-linearity as much as at all
possible. There are counterexamples, though, such as the AC-30 with a power amp that must
make do without any negative feedback – and this surely not just because of the cost-factor.
So, there we are. As already mentioned; the answers were always and remain speculative.
Maybe the following mixture was a typical situation: the expressed objective was a symmetry
as good as possible, ergo little k2, and so the prototype in the workshop was modified until the
result was something the designer could be proud of – and hopefully sounded good, as well.
And off to production … the next project awaits. Creating statistics about parameter variances
was likely to be as popular in the 1950’s as it is today – and it was apparently not necessary,
either.
Unless we are checking out a completely out-of-control paraphase circuit, the tolerances (“un-
anti-symmetries”) occurring in a typical phase-splitter stage for small-signal operation are
rather insignificant, especially compared to the idiosyncrasies in the large-signal behavior.
In order to get from the high plate- (or cathode-) potential to the low grid-potential of the two
power tubes, every usual phase-splitter stage uses two coupling capacitors (coupling-C’s)
carrying the two signals driving the power-tubes. The coupling-C “separates the DC-
component” and carries a constant DC-voltage across it – tells us theory, anyway. It ain’t so!
As distortion (not actually forbidden in guitar amps!) occurs in the output tubes, the latter
experience a non-negligible grid-current which changes the DC-voltage across the coupling-
C’s and thus also the operating point of the output tubes.
Fig. 10.4.12: Simple model-circuit to simulate grid-currents.
Fig. 10.4.12 presents a simple circuit enabling us to discuss the basic behavior in case of
occurrence of a grid-current. US is the signal-source (i.e. the tube of the phase-splitter) with its
internal impedance RS, C is the coupling capacitor. Rg stands for the gird-resistor of the output
tube (e.g. 220 kΩ); the non-linear input impedance of the output tube is modeled by the diode
and the DC-voltage source (e.g. U0 = 20 V). As a first step, it is conducive to assume the AC-
voltage source not to have an additional DC-offset.
As long as the amplitude of the AC-voltage US is smaller than U0, the diode (thought to be
ideal) is in blocking mode. Only a minimum AC-voltage and no DC-voltage is found across
the coupling-C (assuming operation significantly above the high-pass cutoff-frequency).
However, as the AC-amplitude ÛS rises above the DC-voltage U0, the diode starts to conduct
and limits the signal across Rg. The diode now carries an impulse-shaped current flowing only
in one direction and thus having a mean value different from zero. We could also say: a DC-
free AC-current with superimposed DC-current flows through the diode. The DC-current-part
can, however, not pass through the capacitor and has to flow in total through Rg, generating a
(negative) voltage across the resistor. The source (US) remains free of any DC-voltage (stiff
voltage source), but across Rg we get a DC-voltage, and consequently the DC-current
polarizes the coupling capacitor.

This polarization of the coupling capacitor is a non-linear process that could be described via
a non-linear differential equation. As a simplification, we can also look at the final process-
state and assume the polarizing voltage across the coupling-C to be constant (but dependent
on the drive level). Fig. 10.4.13 shows several corresponding time-functions: the amplitude of
the source voltage is 35 V in both sections of the figure; in the left-hand section the signal is
only limited, and in the right-hand section it is additionally shifted towards negative values.
This voltage-shift is the polarization-voltage across the capacitor.
Fig. 10.4.13: Potential-shift due to grid-current in the output tubes. Left: AC-voltage limited to merely 20 V;
right: AC-voltage limited and shifted (capacitor-polarization).
Only for strong drive levels, or for overdrive, any relevant grid-current starts to flow in the
output amplifier, and only these currents lead to a re-charging of the coupling capacitors, and
thus to a shift in the operating points of the output tubes. In Fig. 10.4.14, we see this
polarization voltage given for two different series-resistors as a function of the signal
amplitude.
Fig. 10.4.14: Average grid-voltage-bias UDC in dependence on the drive-voltage-amplitude (model).
In contrast to this model, we find – in the real-world push-pull power amplifier – a voltage
across the capacitors even without any drive signal. This is the difference between the plate-
voltage (e.g. 250 V) and the grid-bias voltage of the output tube (e.g. -50 V). In Fig. 10.4.15
the mean value of the grid voltage of the output tubes is shown as a function of the drive
level. As mentioned above, the grid becomes more negative as the grid-current increases. For
the 2nd output tube (V 8), there are potential shifts already at small drive levels. This is not
due to any grid current, but caused by shifts in the operating point of the differential amplifier.
Fig. 10.4.15: Fender Super-Reverb, grid-bias-voltage of output tubes (mean); 3 different operating points. Drive
voltage (abscissa) is the grid-voltage of the left-hand differential-amplifier tube.

The mean values of the plate-voltages of the phase-splitter do not remain constant as a drive-
signal is applied; they shift even for moderate levels (Fig.10.4.16). Consequently, the
polarization-voltage levels of all four capacitors change – with very different time-constants
taking effect. For example, C2 = 0.1 µF is recharged via Rg = 1 MΩ, resulting in τ = 0.1 s. The
capacitors branching off the plates need to be re-charged, as well, and thus re-charging
currents flow through the grid-resistors (not shown in the figure) of the output tubes.
Consequently, the operating points of the output tubes are shifted due to two mechanisms: the
potential shifts in the differential amplifier, and the grid-currents flowing in the output tubes.
Fig. 10.4.16: Shift of the operating point in the differential amplifier of a Super-Reverb (negative feedback
deactivated). The mean-value of the plate-voltage for the right-hand triode shifts towards lower voltages.
We can see from Fig. 10.4.17, that these drive-dependent re-charging processes in the
differential amplifier do not happen in a symmetrical fashion: for small drive-levels, both
mean values of the plate-voltages decrease, while for strong drive-levels the mean plate-
voltage of tube 1 increases while the plate-voltage for tube 2 decreases. Switching off the
drive signal makes the grid-voltage at the 1st output-tube (V7) jump to more negative values
while this jump is to more positive values for the other output tube (V8). Consequently, there
will be a superposition of interferences of very low frequencies on top of the useful signal.
We could ignore this because neither the output transformer nor the loudspeaker nor the
hearing system is susceptible to such low-frequency excitation – still, we must not generally
neglect these side-effects because corresponding operating-point shifts can lead to envelope
modulation and time-variant non-linear distortion.
Fig. 10.4.17: Mean values of the voltages at the plates of the differential amplifier (left) and at the output tube
grids. During 0 < t < 2 s, the signal level rises by 20 dB, at t = 2 s the signal is shut off. Super-Reverb.

Fig. 10.4.18 shows corresponding loudspeaker voltages of a Super Reverb that had its overall
negative-feedback loop (via the output transducer) deactivated. In the left-hand part of the
figure, a 1-kHz-tone that overdrives the power-amplifier is switched on at t = 0. At t = 100
ms, the level of the tone is reduced♣ by 20 dB which makes the loudspeaker voltage collapse
for a short time. We should not dramatize such effects (compare to the post-masking effects in
the hearing system) but we should not generally ignore them, either, because there may be
individual cases with longer time constants, and because music does not really consist of
exclusively 20-dB-jumps. In the right-hand section of the picture, the loudspeaker voltage
is depicted for almost full drive and for overdrive. Caused by the potential shifts connected to
the grid-current, saddle-point-shaped distortions appear for overdrive-operation at the zero-
crossings. These distortions cannot be traced to insufficient biasing or output-transformer
saturation, as it is sometimes surmised in literature.
Fig. 10.4.18: Super-Reverb, loudspeaker-voltage (overall feedback-loop deactivated).
The saddle-points (also termed crossover-distortion) appearing at the zero-crossings occur if

the half-waves, separately processed by the output tubes, cannot be joined precisely enough.
The superposition does not work sufficiently with the tube-characteristics moving apart due to
the shifts of the mean voltage-values (Fig. 10.4.19). For supplements, see Chapter 10.5.8.
Fig. 10.4.19: Dynamic (drive-level dependent) crossover distortion (compare to Chapter 10.5.8).
♣
The power-amplifier still remains overdriven

10.5 Power Stage 10-75
10.5 Power-Stage
The power-stage is the last amplification stage in the signal chain; it delivers the power
required to drive the loudspeaker. In most cases, it operates with rather pathetic efficiency
because normally less than half of the power produced by the power stage is actually fed to
the loudspeaker – the remainder is converted into heat within the power-tube(s). In order to be
able to sufficiently dissipate this power loss, the tube(s) deployed in the power-stage is (are)
larger than typical preamplifier tubes. The thermal load capacity of power tubes typically
amounts to 12 – 45 W with their physical volume up to 10 times that of a preamp-tube. Since
power-tubes can deal with high voltages but not with high currents, they are almost never
directly connected to the loudspeaker. Rather, the plate-currents of the power-tubes are fed to
the output transformer that takes care of an impedance matching towards the speaker.
A good overview is provided by the family of output characteristics of the power-tube (Fig.
10.5.1) showing the relation between plate-voltage and plate-current. Multiplying these two
quantities yields the power-dissipation at the plate Pa, i.e. the power heating up the plate of
the tube (in addition to the heating done by the tube filament). If the specified maximum
dissipation at the plate is exceeded for long periods of time, the tube begins to glow and may
be destroyed. The so-called power-hyperbola is given in Fig. 10.5.1 as the dashed line,
indicating the largest permissible plate-current for the respective plate-voltage. To the right,
the characteristic finds it limitation in the largest allowable plate-voltage; larger values will
cause sparking and damage. Towards the top, the maximum specifications of plate-current
and/or grid-drive provide a ceiling; the lower limits are given by the blocking behavior of the
tube. Normally, tubes are rather good-natured regarding overload situations (much more so
than transistors) because the associated thermal time constants are much longer. However,
this behavior must not be interpreted as general “indolence”: continuous overload will reduce
the lifetime (Chapter 10.5.9).
Fig. 10.5.1: Family of output characteristics of two typical power-pentodes. Screen-grid-voltage Ug2 = 250V.
It should be noted that the characteristics given in Fig. 10.5.1 are sourced from datasheets (as
is the case for all tube characteristics); to a degree, the individual tube-specimen will look
different. In addition, it needs to be considered that power-tubes are almost always tetrodes or
pentodes, and consequently their behavior is defined by both control-grid and screen grid. For
data-sheet specifications, the screen-grid-voltage is assumed to be constant – however, reality
shows that it depends on the drive-levels, after all. On the one hand, this is due to the fact that
the supply-voltage drops somewhat as the drive-levels increase (“sagging”), and on the other
hand, it is because there is a voltage-drop across the grid-resistor.

10-76 10. Guitar Amplifiers
10.5.1 Single-ended (class A)-operation, tetrode, pentode
In the single-ended, class-A power-stage, one (single) power-tube operates in common-

cathode configuration with the output transformer being part of the plate circuit (transformer-
coupling). Without AC-drive (“quiescent state”), a stable balance appears – it is called the
operating point (OPP). The characteristics shown in Fig. 10.5.2 yield an OPP at 250 V and
48 mA, if a voltage of -7.5 V between (control) grid (g1) and cathode is chosen. This can be
done e.g. by using a cathode-resistor of 142 Ω. The cathode-current (the sum of the 48-mA-
plate-current and the 5-mA-screen-grid-current) will then generate a positive cathode-voltage
of + 7.5 V (relative to ground). With the control-grid at ground-potential (Ug1 = 0) a control-
grid-to-cathode-voltage of -7.5 V results (i.e. the control grid is negative vs. the cathode).
Fig. 10.5.2: Output characteristics of the EL84, power-stage circuit (single-ended class-A operation). AP = OPP
As a drive signal appears (Ug1 ≠ 0), plate-voltage and –current change. As a first approach, it
will be sufficient to consider the transformer in the plate-circuit as a large inductance
connected in parallel with an ohmic resistor (Chapter 10.6). In this model we have only pure
DC flowing through the inductance, and only pure AC flowing through the resistor. With a
drive-signal present, the Ua/Ia-point will move along the load-line given in Fig. 10.5.2: as the
grid-voltage is enlarged, the plate-current increases and the plate-voltage drops until a limit is
reached at 17 V / 92 mA with Ugk = 0. Fig. 10.5.3 shows that the relation between input- and
output-magnitudes is non-linear: merely with small drive-signals around the operating point
we can obtain an approximate image of the input signal with small harmonic distortion. In
addition, we need to bear in mind that in reality, the power-tube is rarely driven via a low-
impedance source. Often, the driver-tube ahead of the power-tube is operating in common-
cathode configuration i.e. with a relatively high internal impedance (e.g. 50 kΩ) – in this case
the grid-current of the power-tube already distorts the control (drive) signal.
Fig. 10.5.3: Transmission characteristic; plate-voltage and plate-current for sinusoidal Ugk (from a stiff voltage source).

The output transformer takes the AC-component from the plate-circuit and generates the
loudspeaker-current that is enlarged by the turns-ratio factor TR (secondary current, Chapter
10.6). The AC-load of the power-tube results from the inclination of the load line; in Fig.
10.5.2 this is 5208 Ω (19.2 mA / 100 V). From this load-impedance at the plate, and from the
loudspeaker impedance (e.g. 8 Ω), we get a first approximation of the transformation ratio
(ratio of turns in the two transformer windings) TR of the transformer: .
In view of the transformer losses, this value should be decreased by about 10 % – now we
arrive at approximately TR ≈ 23 [for more exact calculations see e.g. Schröder, Vol. II].
In quiescent state (i.e. without any drive signal), the plate is at 250 V and the plate-current is
48 mA. Multiplying these two values gives us the dissipation loss at the plate of Pa = 12 W.
Since the (idealized) load-impedance was assumed to be an R/L parallel-circuit (= short-
circuit for DC), the supply voltage is calculated as UB = plate-voltage + cathode-voltage =
257.5 V. This is a bit too much “lab jargon” and we need to get more precise. What the data
books term “plate-voltage” is in fact the voltage drop Uak between plate and cathode; it is also
called plate/cathode-voltage. In a series connection to it we have the voltage drop occurring
across the cathode resistor, also termed cathode-voltage: UB = Uk + Uak. Without drive signal,
the cathode resistor (142 Ω) absorbs 0,4 W while the plate absorbs 12 W, and the screen grid
250 V ⋅ 5 mA = 1.25 W. Consequently, the power supply needs to deliver, in quiescent state,
13.65 W. With a drive signal, the plate-current becomes time-variant und oscillates between
two limit-values, e.g. 5 und 92 mA (Fig. 10.5.3). If we ignore the non-linear re-shaping, the
average of the current remains constant, which implies: the power that the power supply needs
to make available is approximately constant i.e. independent of the drive signal level!
Multiplying the AC-components of the plate-voltage and the plate-current (Fig. 10.5.3) results
in the effective power pushed into the load-impedance: PN = 6 W. Given an ideal transformer,
this power fully arrives at the load-impedance (the loudspeaker); in reality a loss of 20% is
likely. Only about 4.8 W arrive at the loudspeaker and the remaining 1.2 W are converted into
heat in the transformer.
In summary: the power supply needs to deliver about 14 W independently of the drive
signal, which leaves just under 5 W output power at full drive level – with the output signal
being already subjected to substantial non-linear distortion (strong THD). The efficiency of
this circuit is 35%, at best – or even as low as 26% if we include in our considerations the
tube heating. The latter is necessary to operate the EL84, and gobbles up another 4.8 W.
As inefficient this circuit may be – it was indeed used in some early guitar amplifiers. One of
the first VOX-amplifiers, the AC-4, generated 4 W from a single EL84 in a single-ended
class-A configuration. The first smaller Fender amps uses the single-ended Class-A circuit, as
well – we find it e.g. in the Champ, Bronco, Princeton and Harvard amps, although these used
the 6V6-GT, a 12W beam-power tetrode rather than the EL84. Over the years, the Fender
amps in particular were subject to various modifications. Among these the increase in supply
voltage is especially striking: early versions had 305 V; an increase to 305 V followed, and
finally there was as much as 420 V. Can we boost the output power that way? Which is the
optimum operation point to achieve the maximum power output? Which load impedance is
optimal for the tube? Using simplifications in the tube- and transformer-data, the calculation
for optimum working conditions is unproblematic. However, in the real world one needs to
consider deviations from these ideal conditions. In particular the maximum current-load of the
power tubes is subject to manufacturing tolerances, and transformer losses (build-size!)
determine the eventually achievable output power, too.

With the idealized assumption that, in the plate-circuit of the power tube, the power-hyperbola
is the only limitation, the left section of Fig. 10.5.4 shows two load-lines that each are
tangents to the hyperbola. The division UAP / IAP yields the optimum operating point (OPP),
corresponding at the same time to the negative slope of the hyperbola at the OPP. The
maximum possible voltage deflection at OPP1 is 400 Vss, resulting in 6 W, with a load
resistance of 3333 Ω. The same power can be achieved in OPP2: the voltage deflection is
indeed larger at 600 Vss, but the current is correspondingly smaller. If we define the power
hyperbola as limit, the achievable maximum power is always exactly half of the maximum
dissipation-power at the plate – independent of the OPP. For a real circuit we need to factor
in that the plate-current cannot become indefinitely large. In the right-hand section of the
figure, the output characteristic of a 6V6-GT is indicated as limit for the case that the
grid/cathode-voltage is zero. This curve must not be seen as the absolute limit – even larger
plate-currents would be possible if the grid/cathode-voltage were positive. However, the
typically used drive-circuits could not deliver the necessary current, and consequently it is
purposeful to define, in addition to the power hyperbola, Ugk = 0 as the limiting factor. Now,
the maximum voltage deflection reachable at OPP1 is not 400 Vss anymore but decreases to
334 Vss, and the OPP is not located in the middle of the load line any longer. A conducive
shift of OPP1 from 200 V to 233 V does enable us to establish symmetry with regard to the
maximum drive level. However, the reduction of the maximum voltage deflection by 16.5%
decreases the maximum power-offering by 30% (in our example from 6 W to 4.2 W). For
OPP2, the reduction of the voltage-deflection makes itself less strongly felt (5.6 W instead of
6 W), and we can expect the operation with a higher voltage to bring somewhat more power.
Fig. 10.5.4: Output characteristic with two different operating points; the power hyperbola is the limit.
The above calculations regarding the achievable power-output deliberately are of a rather
“principle” character in order to illustrate basic functions within the power stage. If we do not
consider the power hyperbola as limiting factor, the circuit will deliver 50% of the maximum
power dissipated at the plate to the output transformer – irrespective of the tube used. This
upper power-threshold can only become smaller (and never bigger) as individual tube-limit-
data are incorporated. Besides the maximum power-dissipation at the plate, in particular the
maximum tolerable plate-voltage and the maximum allowable power-dissipation of the screen
grid need to be considered. For a supply voltage of 300 V, up to 600 V may occur at the plate,
and even as much as 840 V for 420 V supply voltage. Also, even higher voltages may appear,
since the load impedance (loudspeaker) is not a purely ohmic 8-Ω-resistor but will become
inductive (and thus larger) at high frequencies. Even if the insulation within the transformer is
exemplarily well done: at too large voltages, arc-over is possible in or at the tube, and it can
lead to destruction.

So much for an introductory, basic description of the behavior of a single-ended power-stage

– now on to the details. For the triodes deployed in preamplifiers, a simple power law was
formulated as an approximation (Child/Langmuir, Chapter 10.1.3):
Triode characteristics
The plate-current Ia depends on the grid/cathode-voltage Ugk, on the plate-voltage Ua, and on
the open-loop gain µ, the latter also known as durchgriff: D = 1/µ. To get even more into
detail: the free conducting electrons in the metal cathode are highly mobile but cannot leave
the metal in its cold state. A special coating combined with red-heating enables a significant
portion of the electrons to leave the metal and form, in the immediate vicinity around the
(heated) cathode, a kind of “electron-mist” – also called “space-charge cloud”. The more
electrons accumulate in front of the cathode, the more negative this space-charge area
becomes, and the more effectively further electrons are inhibited to move against this negative
potential – an equilibrium results. A positively charged plate will superimpose an electron-
accelerating plate-field over the electron-inhibiting space-charge field, and the former field
will suck electrons away from the cathode and draw them to the plate. The space-charge
decreases, enabling more electrons to leave the cathode. The electrons leaving the cathode
form the cathode-current, and the electrons arriving at the plate form the plate-current. A
(control-) grid (three-electrode-tube = triode) introduced between cathode and plate will
create, via its electrical potential, an additional field. Consequently, on top of the space-charge
field, two fields that are controllable via the electrodes act on the electrons and therefore
influence the current: one generated by the (control-) grid, and the other generated by the
plate. Since the grid is positioned closer to the cathode, it exerts the bigger influence: the plate
needs to first “reach through the control grid to the space-charge” – hence the term
“durchgriff” (the term taken from German, meaning “reaching through”). For the ECC83, the
datasheet indicates a rather small value at D = 0.01. However, with the plate-voltage being
about 100 times the value of the grid/cathode voltage, both Ua and Ugk influence the plate-
current. Textbooks on practical circuit-design see the grid as control-electrode and designate
Ugk as control-voltage. More theoretically oriented oeuvres combine the summands
, using the same term control-voltage for the combination i.e. this term may
have two different meanings! In the formula above, USt is the theoretical control-voltage
considering both the influence of grid and plate, with K2 being a tube-specific constant.
One may consider it a problem that the plate-current of the triode does not only depend on the
grid/cathode-voltage but also on the plate-voltage. A solution can be found by inserting an
additional screen grid (g2) between control grid (g1) and plate, and connecting it to a high
positive voltage – this way the electrons are predominantly accelerated by the control grid and
the screen grid, with the plate-potential retaining merely a minor significance. For the
resulting 4-electrode tube (= tetrode), the potentials of all electrodes can be described via a
theoretical control-voltage:
Control-voltage of the tetrode
The tube parameters D1 and D2 – both considerably smaller than 1 – can again be interpreted
as durchgriff. D1 ⋅ D2 shows the (intended) small influence of the plate-voltage.

As an example: if the control-grid-voltage has to change by 1 V in order to change the plate-

current by 10 mA, then for the same plate-current change the screen-grid-voltage would have
to be changed by 20 V, or the plate-voltage by 400 V. To map the control-voltage onto the
plate-current, we could use the power law for the tetrode, as well, but we would need to
introduce considerable corrections to obtain a good match to the actual behavior. A main
reason for this discrepancy between simple theory and practice is the release of secondary
electrons from the sheet-metal of the plate. As soon as the electrons arriving from the cathode
are accelerated with more than 10 V difference in potential, they have enough energy to
knock, as they hit the metal, further electrons from the plate – these are the secondary
electrons. With the screen-grid-potential lower than the plate-potential, this process is not
disruptive because the secondary electrons return to the plate. However, for higher screen-
grid-potential the secondary electrons fly on to the screen grid – correspondingly decreasing
the plate-current and increasing the screen-grid-current. This is the reason why an enormous
bump appears in the Ia/Ua-characteristic of the tetrode for small plate-currents. This bump is
undesirable (Fig. 10.5.5).
Tetrode Pentode
Fig. 10.5.5: Output characteristics (Ia vs. Ua) of a tetrode (left) and a pentode (right).
Corrective action is provided by yet another electrode, the suppressor grid (or retarding grid)
located between screen grid and plate. Its job is to push back the secondary electrons en route
from the plate so that they will not land on the screen grid. This only works if the suppressor-
grid potential is much lower than the screen-grid potential, and therefore the suppressor grid is
normally connected to ground. The fast electrodes emitted by the cathode are pretty much
unaffected by the this suppressor grid while the slow secondary electrons knocked out of the
plate are not able to overcome the potential difference to the suppressor grid and return to the
plate. Staff at Philips developed the first commercial version of this five-electrode tube (=
pentode), with a corresponding patent filed in 1926. For a short time, pentodes are also found
in pre-amplification stages of guitar amplifiers but these were soon replaced by triodes
(Chapter 10.1). In contrast, we find almost exclusively power-pentodes in the power-stages,
for example the EL84 (e.g. VOX), or the more powerful EL34 (e.g. Marshall).
The London-based tube manufacturer MO-V (MO-Valve or Marconi-Osram Valve Co Ltd.)

was not allowed to manufacture pentodes due to the patent owned by Philips, and developed
(around 1933) a serious alternative to the pentode: the Beam-Power-Tetrode. Its baffles
concentrate the electron-stream such that strong space-charges strongly deemphasize the
characteristic tetrode-bump. It appears, however, that there was not that much confidence in
the concept at MO-V, and the corresponding rights were sold to RCA in the United States.
RCA used them to very successfully develop the beam-power-tetrode 6L6, and this again
forced MO-V to act all the more. They introduced the KT-66, the “kink-less tetrode”. Both
the 6L6 and the KT-66 were manufactured in many variants that can differ substantially.

The power tubes employed in guitar amplifiers may be divided into three main groups:
pentodes, British beam-tetrodes and US beam-tetrodes. Among the pentodes, there is the
EL84 for low-power applications, and the EL34 for high power. The KT-66 and the more
powerful KT-88 are the British beam-tetrodes. Their American counterparts are the
smaller 6V6 and the larger 6L6. All these tubes have undergone multiple redesigns since their
introduction to the market; that is why we cannot speak of “the” 6L6. First came the
development step from steel- to glass-container, then there were changes in the shape of the
container, but also in the electrodes and thus in the electrical parameters. The RCA 6L6-GB is
rated with a maximum plate-dissipation of 19 W, the Tungsol 6L6-GB is rated at 22 W. Can
the Tungsol-tube therefore carry a higher load? That is difficult to say, because we read in the
RCA datasheet: Design-Center Values, but in the information by Tungsol: Design Maximum
System (more about these rating systems in Chapter 10.5.9). The Sylvania 6L6-WGA is
specified at 19 W (Design Center), but also at 21 W (Absolute Maximum). As a first
approximation, these are all tubes that are the result of a development from the 6L6, via the
6L6-G and the 6L6-GA, to the 6L6-GB, that related predominantly to the shape. Only for the
6L6-GC we see a pronounced power upgrade to a plate dissipation of 30 W (Design
Maximum Values); this is probably based on changes in the metal sheet of the plate. None of
these tubes were developed specifically for guitar amplifiers – that market was much too
small at the time. Rather, we read: For Radio Receivers. There were also particularly robust
military tubes designated with a supplementary W, e.g. the 6L6-WGB. The corresponding
electrode-build was optimized to withstand the stringent MIL-testing.
The KT-66 is the British counterpart to the 6L6. It is specified with a maximum plate
dissipation of 25 W in the Osram data-sheet; we find the same data in the Marconi data-sheet,
and checking the info from MO-V yields 25 W (Design Max) or 30 W (absolute Max),
respectively. MO-V is the moniker for the Marconi-Osram-Valve-Company, that offered the
KT-66 globally under the GEC label. This is GEC = General Electric Corporation of
England, not to be confused with General Electric USA. Both the 6L6 and the KT-66 are
beam-tetrodes, i.e. tubes without a suppressor grid. Because the beam-forming sheets can be
seen as a fifth electrode, after all, these tubes are often labeled as pentodes, too (despite the
lack of an actual suppressor grid). The EL34, however, is a true pentode, specified at 25 W –
or at 27.5 W (“at maximum drive level”). All these tubes show similar data regarding the
maximum load, but we may not conclude that they can be arbitrarily interchanged – their
control characteristics show considerable differences, after all.
Before we delve more deeply into the area of tube characteristics, let us take a short look at
other power-tubes. Around 1950, Tung-Sol develops the 5881 and advertises it as an
advancement of the 6L6 (or the 6L6-GA). In 1962, the maximum plate dissipation of the 5881
is still specified at 23 W (Design Center System) – but by that time, the 6L6 has enjoyed the
further development into the 6L6-GC (30 W), and the 6L6-WGB (26 W) has been available at
least since 1955. It is not surprising that not everybody regards the 5881 as the “better 6L6”.
And then: what does “better” mean in this context? Is this from the point-of-view of the MIG-
pilot demanding full function even after a rough touchdown? Or from the point-of-view of the
aficionado of classical music expecting the least possible distortion? Or from the point-of-
view of the Jazz guitarist having just discovered that the tone control does not have to be
stuck at “0” all the time? Or from the point-of-view of the Eddie-epigone overdriving his
equipment (his “rig”) exactly “VH-like”? To state “the 5881 is the better 6L6” is just as
misguided as “6L6 = KT66 = 5881”. “The” 6L6 does not exist, just as “the” KT66 or “the”
5881 do not exist. It is not just that the datasheets indicate differences – today many a KT66
internally is but a 6L6-variant.

When evaluating tubes in general, and power tubes in particular, two criteria offer themselves:
the sound, and the operational lifespan. Sure, price and availability also figure – but we will
tackle that later. The lifespan may be five hours or five years; it has its own chapter dedicated
to it (Chapter 10.5.9). The sound is advertised with “powerful bass” or “clear treble”, and
consequently many guitarists presume that tubes would feature a frequency-dependent
transmission characteristic – like that found in a loudspeaker. However, this assumption is not
correct as such: tubes can process frequencies as low as desired♣, and frequencies as high as
they come; whether the upper cutoff-frequency is found to be 100 MHz or 200 MHz is
immaterial in the present context. On the other hand, to deduce that all tubes would sound the
same is incorrect as well. It’s not that the tube itself would have a “sound”, but it does
influence the transmission behavior of the power stage as a whole. It does make a difference
to the loudspeaker whether it is driven by a source of high or low impedance, and the
character of the distortion is tube-specific, as well. The generally publicized view seems to be:
tubes will sound somehow, expensive tubes will sound better, and old tubes will sound best.
Cheapest are so-called industrial tubes i.e. tubes manufactured for industry. Well – of course
it’s not only industry that gets them, because how else would they be offered in minimal
quantities to musicians. “Industrial tube” probably is supposed to indicate that the musician
will receive these tubes in the same condition that industry would receive them: without
additional value added by the retailer. Without added value does not mean without an add-on
to the price tag, though – that a business makes money from this commodity, too, is the
legitimate result of mercantile aspiration. Besides industrial tubes, there are selected/matched
power tubes. They carry mysterious numbers on their sockets and/or on their carton, and they
were “paired”. At least they are being boxed with a label indicating that. That such an ado will
cost extra is again the result of mercantile operation. A set of 4 EL84, for example, will cost
30 Euro if you ask for industry tubes but set you back 70 Euro if you are being handed a
“matched quartet”. How this “matching” is done will normally not be disclosed. How well it
works out: that shall be the subject of the following pages. For those of us who regard 70 Euro
as an insult to their virtuosity, NOS-ware is available. These would be tubes that have not
only miraculously be hidden away in basements and warehouses but actually were even able
to reproduce, and are offered – since many years – with the supplemental encouraging
remark: one of the very last originals! Their sound is portrayed as unrivaled, this assessment
being supported by the intuitively fair enough reasoning that the old tube experts were
scrapped together with the old manufacturing plants. In individual cases that may have been
accurate (while not entirely trivial, after all), but it is – frankly – nonsense to conclude that a
tube would be better just because it has spent 50 years in the basement without use. It will
possibly deliver exactly the desired sound; just as possibly it will, however, sound worse than
a low-cost industrial tube. You will only know after you’ve bought it.
It is difficult for the buyer to verify whether a particular tube indeed hails from ancient stock
or is merely a modern el-cheapo imitation. Internet-forums about “faked tubes” are of some
help here. Whether a tube does meet the given requirements will be revealed (subjectively) by
listening tests and (objectively) by measuring its data. At this point we shall not yet
investigate to which extent a conclusion from one to the other is legitimate – let’s look at the
technical data first. According to conventional wisdom, most important are plate dissipation
and transconductance (plus of course the socket needs to fit). Plate dissipation = maximum
load (e.g. 30 W), transconductance = gain (e.g. 5 mA/V). That, however, will not be good
enough to select a power tube – in a guitar amp there are further criteria to base this choice
on.
♣
Only the lifespan of the tube stands in the way of that this range not extending to exactly 0 Hz.

To assign the power-needs is relatively easy as long as we look at the bare essentials: low
power = EL-84, 6V6-GT; medium power = 6L6-GC, 5881, KT-66, EL-34; high power = KT-
88, 6550. There are of course more tubes, and some tubes were/are offered in several power
categories (e.g. 6L6-GB vs. 6L6-GC), but we will not go into that here. Similarly, a
discussion about the proof voltage will be omitted – the corresponding statements in the
datasheets are too obscure and contradictory.
Power tubes are rated with about 10 – 50 W regarding the maximum power dissipation at
the plate. This value must not be mistaken for the power output of the amplifier! There are
100-W-amps that draw their output power from 2xEL34 (Pa,max = 25W), and there are 40-W-
amps using 2x6L6-GC (Pa,max = 30W). Fig. 10.5.6 shows the output- and transmission-
characteristics of the most important power tubes. All curves are for Ug1 = Ugk = 0V, i.e. for
full drive level. Applying positive control-grid voltages, it would in principle be possible to
achieve even higher plate-currents but the usual driver stages are of too high an output-
impedance for this. Besides the control-grid voltage, it is also the screen-grid voltage that
determines the shape of the output characteristic. In order to be able to compare, we choose
Ug2 = 350 V, although of course not all amplifiers operate with this voltage value. The GE-
datasheet even specifies as little as 285 V for the 6V6-GT – but that didn’t hold back Fender
to subject the 6V6-GT in the Princeton to a proud 415 V.
Fig. 10.5.6: Output characteristics (left) und transmission characteristics (right) of some power tubes.
We can see from Fig. 10.5.6 that – for comparable operating conditions – the maximum plate-
currents differ quite substantially, after all. The transmission characteristics, as well, show
pronounced individuality, and therefore a KT-66, for example, must only be switched for an
EL34 after suitable modifications in the circuitry. In any case, it is important to bear in mind
that such characteristics remain general, simplifying illustrations.
Fig. 10.5.7 proves this via measurements with
3x6L6-GC and 2xKT-66. Given just the data-
sheet info, similar characteristics in all 5 tubes
would be expected – reality is quite different,
however. In some circles, the looks (i.e. the
shape of the glass container) of a tube will be
given more attention than the actual electrical
function. Comparison tests that are not
considering significant electrical differences
as those shown above are thus not only not
helpful, but just plain useless. More on that in
Fig. 10.5.7: Measured output characteristics Chapters 10.5.11 and 10.11.

The following table compiles some tube data. The respective year was taken from literature
i.e. it does not necessarily indicate the true time when the respective tube was first issued to
the market. The transconductance (mA/V) depends much on the specific operating point, and
therefore the given value is for rough orientation only: detailed information is offered by the
characteristic curves (Chapter 10.11).
The maximum permissible plate dissipation is also to be seen for orientation only: the
specification in the datasheets of different manufacturers deviate to some extent, and
moreover, back in the day the specification was done using two different standards: Design
Center System, and Design Maximum System (in brackets, compare to Chapter 10.5.9).
Type Pa,max / W mA/V Manufacturer Year

6V6 12 (14) 4 RCA 1937
6V6-G 12 (14) 4 RCA 1941
6V6-GT 12 (14) 4 RCA 1944
6V6-GTA 12 (14) 4 RCA 1962
6L6 19 (---) 5.3 MOV ⇒ RCA > 1933

6L6-G 19 (---) 5.3 > 1936
6L6-GA 19 (---) 5.3 > 1943
6L6-GB 19 (22) 5.3
6L6-WGB 20 (23) 5.3 Tung-Sol 1950
6L6-GC --- (30) 5.3 1954
5881 23 (---) 5.3 Tung-Sol 1950

7027 25 (---) 6 RCA 1958
7027-A --- (35) 6 RCA 1959
6550 35 (---) 11 RCA 1962
6550-A --- (42) 11 GE 1972
KT-66 --- (25) 6.3 Marconi 1956 (> 1937)

KT-66 --- (25) 7 MOV 1977
KT-77 --- (25) 11 MOV 1977
KT-88 --- (35) 11 MOV, GEC 1957
KT-90 --- (45) 11 Ei
EL84 12 (---) 11 Philips Ca. 1955

EL34 25 (---) 12 Philips Ca. 1952
EL51 45 11 Philips 1953
EL151 60 13 Telefunken 1943
QB3.5/750 250 4 Philips

Table: Power-tube data-sheet information: maximum plate dissipation and transconductance.

10.5.2 Push-pull class-A operation
The single-ended power stage introduced in Chapter 10.5.1 turned out to be relatively weak in
terms of power delivery: With a 12-W-tube we could get at most 6 W output power from it.
For a greater output power more powerful tubes would be available, but then there is another
disadvantage of the single-ended circuit: even without any drive signal, a relatively strong
DC-current runs through the output transformer, and the latter needs to operate under
unfavorable conditions due to the resulting DC-pre-magnetization. We could insert an air
gap into the iron core of the transformer and reduce the DC-field dependency of the reversible
permeability – but then we would in total reduce this permeability to a value smaller than the
one for the core without air gap. Moreover we need to consider that the load on the power
supply is independent of the drive levels for the single-ended class-A power stage. In other
words, even at rest, the power supply experiences maximum load, and therefore the supply
voltage is not constant but oscillates around a mean value with a frequency of 100 Hz (given a
two-way rectifier). This AC-component generates an AC-current through the output
transformer, the output tube being of high impedance but still no ideal current source. The
result is an undesirable interference tone at 100 Hz or 120 Hz (depending on the local power).
Using a push-pull class-A circuit (Fig. 10.5.8), some of the disadvantages of the single-ended
class-A circuit can be avoided. The term “push-pull” is derived from the opposite-phase grid-
drive of the two output tubes. The rising grid-voltage at one tube increases its plate-current
while at the same time the decreasing grid-voltage at the other tube reduces the plate-current
there. Ideally, the former plate-current increases by the same ΔI that the latter plate-current
decreases by; the sum of the currents sourced from the power supply I= remains constant (DC
current), independently of the drive levels. At rest, this DC-current splits up into the two
plate-currents of equal strength that each generates a magnetic DC-field in the transformer
core. Since the two DC-fields have opposite directions, they compensate each other within the
core, and the latter remains field-free (without pre-magnetization). No air gap is necessary. A
corresponding compensation also happens for the residual ripple in the supply voltage: the
100-Hz AC-current generated by it causes opposite-phase AC-fields that cancel each other
out, and cannot result in hum in the loudspeaker.
An entirely different situation exists for the AC-currents at the plate that are created by the
opposite-phase grid-drive: they are, in terms of the reference-arrows defined in Fig. 10.5.8,
of opposite phase, but therefore correspond in-phase to the primary AC-current (defined in
one and the same direction): . The equality of these two AC-currents also
results from the power supply (ideally) delivering pure DC: if no AC-current is leaving the
winding at the tap of the primary winding, both primary AC-currents need to be equal.
Assuming an ideal transformer with identical primary windings, the total primary voltage will
be double the AC voltages at the plate; the two tubes thus operate in series.
Fig. 10.5.8: Push-pull class-A power stage; on the right, the phases of currents and voltages are illustrated.

In Fig. 10.5.9 we see the idealized voltages and currents relating to Fig. 10.5.8. Each
individual tube operates in single-ended class-A mode with the operating point located in the
middle of the useable load line (Chapter 10.5.1). The two tubes are driven by opposite-phase
signals. With the reference-arrows defined according to Fig. 10.5.8, both the AC-voltages at
the plate and the AC-currents at the plate are of opposite phase, as well. The (overall) primary
voltage Uaa is the difference between these opposite-phase AC-voltages ,
and the AC-current flowing through both primary windings is .
Fig. 10.5.9: Voltages and currents of the individual tubes (left); primary AC-voltage and -current (right)
If we neglect the residual plate-voltage (for Ug1 = 0) and assume an ideal output transformer,
the obtainable effective power corresponds to the maximum power dissipation in one tube;
two EL84 will thus yield 12 W at the most (and in the ideal case). In practice (i.e. considering
residual voltage and transformer losses), about 10 W may be expected. Ideally, the power
taken from the power supply is independent of the drive level and corresponds to the
maximum dissipation of both power tubes – for 2 x EL84 that would be 24 W due to the
plate-currents, plus about 2.8 W screen-grid dissipation, plus about 0.8 W dissipation in the
joint cathode resistor. This resistor should be chosen such that in idle the cathode-current
(0.11 A in this example) generates just the required grid-bias (7.3 V).
The optimum load-impedance at the plate, and with it the transformation ratio of the
transformer, is to be derived from the gradient of the load line, just as it would have to be for
a single-ended class-A amplifier. For both the latter and the push-pull class-A amp, every tube
needs to “see” the same load impedance Ropt. When designing the push-pull class-A amplifier,
consideration needs to be given to the fact that both (half) primary windings “see” two load
impedances: the secondary winding as passive load, and the other (half) primary winding as
active load! For this reason, the load of the individual tube in the push-pull class-A circuit
may not be simply calculated from the transformation ratio (impedance paradox, Chapter
10.5.5)! If the transformation ratio for a single-ended class-A amplifier is e.g. Np : Ns = 24 : 1,
it will be Np1 : Np2 : Ns = 17 : 17 : 1 for the push-pull class-A amp (given otherwise equal
conditions). The datasheet specifies an optimum load-impedance at the plate of Ropt = 5.2 kΩ
for single-ended class-A operation of an EL84-amplifier with UB = 250 V, and thus the
(overall) primary impedance for the push-pull class-A amp amounts to Raa = 10.4 kΩ.
Examples for specific amplifiers are presented at the end of the chapter.

10.5.3 Push-pull class-B operation
In a push-pull class-B amplifier, the operating point is not positioned in the middle of the load
line but at its lower end. Without a drive signal, only a small idle-current flows though the
power tubes. There is lesser load on the power supply and the tubes do not get as hot –
however the obtainable output power is still higher than that of the push-pull class-A
amplifier.
To keep the plate current small while there is no drive signal, the grid bias voltage needs to
be on the rather negative side. Due to the small current, this cannot be achieved as a voltage
drop across a cathode resistor anymore, and therefore both cathodes are set to ground
potential while a separate DC voltage-source generates the required negative bias-voltage
(measuring, after all, in the order of -15 … -65 V) at the grid. This DC voltage-source is
designated Ug1 in Fig. 10.5.10, and it is fed to the circuit via two high-impedance resistors
(e.g. 220 kΩ) connected across, and two grid resistors (e.g. 1 … 5 kΩ; there are circuits
without these grid resistors, as well).
Fig. 10.5.10: Output characteristics of the EL84; circuit of the push-pull class-B power stage.
We obtain the optimum gradient of the load line (and thus the optimum load impedance) if
the intersection of the Ugk = 0V-characteristic and the load line has the maximum distance
from the operating point. Fig. 10.5.10 shows two different load lines: the flatter line (b)
relates to a 2.0-kΩ load-impedance while for (a), the load-impedance is 1.6 kΩ. It is of no
issue that the hyperbola designating the power limit is intersected: the tube is only subject to
strain during one half-wave, and the plate dissipation remains within the tolerable limit on
average. Static drive signals, or drive signals of extremely low frequency are not to be
expected with guitar amplifiers since the power stages are fed via a high-pass.
The maximum power yield does not differ much between the two variants: it is about 19 W
for (a), and 18 W for (b). As a comparison: a corresponding push-pull class-A power stage
could only deliver about 11 W. Besides the maximum power yield it is, however, also the
power required from the power supply that merits consideration, especially in the case when
there are small drive signals. For the push-pull class-A power stage, the load on the power
supply is independent of the drive level, e.g. 24 W for 2xEL84. In contrast, the power supply
needs to deliver as little as 3 W in the push-pull class-B power stage (depending on the bias
setting). Fig. 10.5.11 shows the power balance – albeit without considering the power
dissipated in the screen grid that would amount to about an additional 3 W at full drive levels.
In the class-B mode, the output power is larger and the power losses in the tubes are smaller:
the maximum plate dissipation in class-B mode is only about half of that found in class-A
mode.

Fig. 10.5.11: Link between power-supply load (without g2-dissipation) and output power PAudio. “Gegentakt” = push-pull.
The difference between the two curves corresponds to the plate dissipation 2⋅Pa of both tubes. (UB = 300V).
The relatively high efficiency of the push-pull class-B circuit results from the fact that each
tube carries a large plate current only when power is actually delivered to the load. For this,
the operating point needs to be set at the lower end of the load line. However, the bias-voltage
at the grid must not become too negative because this would result in crossover distortion
(Fig. 10.5.12). Given a sufficiently large bias-current (left-hand section of the figure), the two
tube characteristics superimpose to a reasonably smooth curve, while for too small a bias-
current a saddle point appears (middle and right-hand sections of the figure). This saddle point
will increase the odd-order distortion on one hand, and on the other hand leads to an
undesirable (progressive) drive-dependency of the slope of the characteristic (Chapter 10.5.8).
Special consideration needs to be given to the fact that the supply voltage decreases with
increasing drive levels – the screen-grid voltage therefore decreases as well, and this further
emphasizes distortion (Chapter 10.5.8).
Fig. 10.5.12: Characteristics at different bias-current settings. Crossover distortion. On the right, the distortion relating to the middle picture
is depicted for two different amplitudes.
Literature does not give an exact definition for the load line in push-pull class-B operation.
Rather, it mentions “small plate-current”, and occasionally even a plate-current for which the
operating point is set “almost to zero”. This did not keep Siemens and other tube
manufacturers from specifying a bias current for the EL34 of no less than 35 mA. They do
have a point because the theoretical case that the plate-current approaches “almost zero” has
next to no bearing on low-frequency applications. 35 mA: that is indisputably “somewhat
more than almost zero”, and whether this mode of operation may in fact be still called “push-
pull class-B” is subject of controversial discussions. Alternatively, the term “push-pull class-
AB operation” is used, or the term “push-pull class-D operation” – it is important to know that
these designations are ambiguous! (Details are found in Chapter 10.5.4).

In class-B operation, the two power tubes conduct simultaneously only for small drive levels;
at higher drive settings each power tube conducts predominantly only during one half-wave.
This needs to be taken into account when choosing the cross-section of the wire in the
transformer. If we assume a sinusoidal drive signal, and a peak plate-current of 141 mA
(Fig. 10.5.10), the RMS-value is not 100 mA but only 50 mA.
The plate-voltage without drive signal is, for class-B operation, just slightly less that the
supply voltage UB (e.g. 300 V). During the half-wave at which the tube is conducting, the
plate-voltage drops as far as the residual voltage (e.g. 30 V). During the other half-wave
(blocked mode), the plate-voltage does not remain at the level of the supply voltage but
increases to almost double of that value (e.g. 570 V!) This is because the primary winding of
the output transformer sees practically no load when the respective tube is in blocking mode,
while the magnetic flux generated by the other (active) tube induces a high voltage in this
winding without load. In power stages that operate with higher supply voltages, voltages that
are even much more dangerous can result: e.g. 850 V in Fender amplifiers, or 1100 V in
Marshall 200-W-amps. These high voltages are not contradictory to the information given by
datasheets where the maximum plate-voltage is specified e.g. at 800 V; the values expressed
there are meant as idle voltage (without drive signal). For example, the datasheet of the EL34
determines the maximum plate-voltage at 800 V but allows for maximum peak voltages of
2000 V in the blocked mode. Such high voltages can in fact occur easily if the amplifier is not
connected to its nominal load but operated with a higher impedance, or no load at all at the
output. In this case, spark-over or arcing between the connector-pins 3 and 2 (plate and heater
filament) can easily happen – which is likely to damage the tube socket and/or the tube holder
irreversibly. Even more dangerous is an insulation-destroying spark-over within the output
transformer because an adequate replacement for this component may not be at hand.
A few comments regarding seemingly “useless” circuits-components: that they are included
often needs to be credited to practical insights. The grid-resistors (2 – 5 kΩ) connected in
series with the (apparently high-impedance) control grid will reduce the tendency of a power
stage to self-oscillate. High-frequency self-oscillations may occur – but they do not have to.
The power stage may well operate perfectly without these resistors, as well; however, it is
advisable not to simply omit them. With each tube- or loudspeaker-change, different stability-
criteria creep in, and the small additional investment for these resistors can very quickly pay
off. The same holds for small capacitors (10 – 100 pF): if they are not directly connected to
the tone-filter stages, they presumably are supposed to suppress RF-oscillations. It is indeed
possibly that they were (had to be?) chosen with a value that audibly cuts into the brilliance of
the guitar sound. If that is the case, we find a wide field of possibilities to improve the sound –
but we are also confronted with a good chance that we operate a powerful radio-transmitter as
we change or remove such capacitors. Since power-stage oscillations can easily occur in the
FM-range (100 MHz), it is recommended to check the stability with a broadband oscilloscope.
Evidently, we must not discard such oscillations as “inaudible” and therefore irrelevant: on
one hand, operating such a transmitter may be unlawful, and on the other hand the power
tubes may be overloaded massively. Moreover, there may well be secondary symptoms that
are audible.

10.5.4 Push-pull class-AB operation, push-pull class-D
A push-pull class-A power stage operates in push-pull class-A mode for small drive levels,
and for high drive levels in push-pull class-B mode – that far, literature agrees. In detail,
however, differences appear and we find three definitions that we will designate old,
alternative, and new. According to the old definition, the class-AB operation is a class-B
operation with a somewhat enlarged bias-current; there is a distinction between AB1 (without
grid-current) and AB2 (with grid-current). A specific guideline where to set the operating
point does not exist; it may be located (in the output characteristic) “somewhere between” the
A-operating point and the B-operating point. This has often led to defining the location of the
AB-operating point exactly in the middle between the two (A- and B-) operating points.
Example: if the bias-current is 50 mA for class-A operation and 10 mA for class-B operation,
then it must be 30 mA for class-AB (according to the datasheet).
The literature from “back in the day” does not specify whether the bias-voltage at the grid of
the class-AB circuit is generated “automatically” via a resistor at the cathode, or via a separate
voltage source. The alternative definition seeks to be more precise. In the class-AB
amplifier, the operating point can shift dependent on the input signal: with increasing drive
level, the cathode-current will become more and more asymmetric (due to the non-linearity of
the characteristic). Consequently, the voltage-drop across the cathode resistor (bridged by a
capacitor) increases and shifts the average grid/cathode-voltage more into the negative. That
way, class-A operation changes into class-B operation as drive levels increase (Fig. 10.5.13).
The alternative definition moreover designates all those power stages with the term push-pull
class-D amplifier that generate their bias voltage at the grid (exclusively) via a separate
voltage source, and that feature an increased bias current relative to the class-B operation [e.g.
H. Schröder, W. Knobloch]. This definition does not generally consider the polarizations of
the coupling capacitor also leading to a drive-level dependent shift of the operating point.
Fig. 10.5.13: Output characteristics and operating point: push-pull class-A and –B amplifiers (left), push-pull class-D amplifier according to
the alternative definition (center), push-pull class-AB amplifier according to the alternative definition (right).
Under the moniker class-D operation, the new definition considers something entirely
different: it designates a switching amplifier (using pulse-width modulation PWM) with the
term D. According to the new definition, class-AB operation is a class-B operation with
increased bias-current and a fixed operating point. Presumably, this “new” terminology came
in when bipolar transistors started to supersede power tubes. Setting the operating point for
transistor circuits is done according to different criteria compared to tubes; there is no drive-
dependent operating point anymore, and the meaning of the terms changed.
In contrast to HiFi power amplifiers, the minimization of distortion does not have priority in
typical guitar amps. For this reason, we see a domination of old-school class-AB power amps
with the bias-current set according to special criteria (Chapter 10.5.8). The AC-30 also
belongs to this group, and not to the group of push-pull class-A circuits (Chapter 10.5.12).

10.5.5 The impedance-paradox
The output transformer matches the low-impedance loudspeaker (e.g. 8 Ω) to the higher-
impedance tube circuitry. For push-pull circuits, this transformer has two serially connected
primary windings. Could you say what the input impedance of these two windings is? Let’s
take as an example turns-ratios of 10:10:1, and 8Ω as secondary load: for the ideal
transformer the input impedance of the whole primary winding is Raa = 202 ⋅ 8 Ω = Raa = 3200
Ω. Is now the input impedance of one of the two windings half of this value i.e. 1600 Ω? For
the push-pull class-A operation, we assume as much because here the same AC-current flows
through both primary windings. However: calculating the impedance transformation for half
the primary winding, we get: Ra = 102 ⋅ 8 Ω = 800 Ω. What is the correct value?
The push-pull transformer is a three-port network i.e. a system with three pairs of
connections. The two primary ports are connected in series so that overall only 5 connecting
points show. To calculate the input impedance of one port, the two other ports need to be
considered as loads. If we connect only one 8-Ω-load resistor to the secondary winding, and
leave the one primary winding open, we will measure 800 Ω at the remaining other primary
winding. This is because the transformer does now operate merely as a two-port network (=
quadripole). However, if we connect both primary windings – as it is done for the push-pull
class-A operation, then each primary winding “sees” two load impedances: the secondary
load, and the other primary winding.
Fig. 10.5.14: The rigidly-coupled output transformer as three-port. Right: simplified circuitry; ü = TR.
Fig. 10.5.14 shows a simplified equivalent circuit of a transformer. R1 and R2 are the
resistances of the windings, RL is the secondary load resistance, L stands of the main
inductance. In the middle frequency range the effect of the main inductance may be neglected,
and the secondary resistances can be transformed via (TR)2 into the equivalent impedance Z.
The equivalent circuit given in the right-hand section of the figure can now easily be
calculated:
It is clear that for I2 = –I1 the input impedance becomes independent of Z (or rather of RL),
because here the voltage across Z approaches zero. In a push-pull class-A power stage,
however, the two currents are in opposite phase (and ideally also equal in their magnitude) so
that the input impedance is increased. Neglecting R1, the input impedance doubles as we bring
the second primary winding into the circuitry.
Let us include some numbers into the above example:

In the push-pull class-A power stage, the input impedance of (half of) the primary winding
is half of the input impedance Raa of the total primary winding, i.e. 1600 Ω. In contrast, only
one winding is active at any given time in the push-pull class-B power stage: when one of the
tubes is conducting, the other blocks. Therefore, in this case the input impedance of (half of)
the primary winding is only a quarter of Raa, i.e. 800 Ω. The impedance Raa does not appear
physically in the push-pull class-B power stage; it remains a pure calculation value.

10.5.6 Negative feedback
We talk about feedback if a part of the output signal of an amplifier is channeled back to the
input and superimposed there onto the input signal. Same-phase feedback is termed positive
feedback while the designation of opposite-phase feedback is negative feedback (sometimes
abbreviated with NFB in the following). Since there are two output signals (current and
voltage), and correspondingly two input signals, four different ways of negative feedback may
be defined. In a typical guitar-amp power-stage we predominantly find negative feedback of
the voltage-voltage kind: a percentage of the output voltage is fed back and superimposed on
the input voltage. This superposition results in a control circuit: as the output voltage
decreases (e.g. due to loading), less voltage is fed back – resulting in more gain so that the
voltage loss is partially compensated. This special negative feedback (termed g21-negative-
feedback in circuit-design) stabilizes the voltage-gain factor and reduces the linear internal
impedance♣, and also broadens the small-signal-bandwidth, and reduces harmonic distortion.
Fig. 10.5.15: Basic schematic of a feedback loop (left). For positive gain v, negative feedback results. The most important values are the
internal impedance Ri and the voltage-gain-factor vU.
In Fig. 10.5.15, k determines the degree of the negative feedback i.e. its effectiveness. For k =
0, the negative feedback is without effect; with rising k, the effect of the negative feedback
increases. R represents the internal impedance of the power stage without feedback; in tube
amplifiers, this is considerably larger than the load impedance. The Fender Super Reverb,
for example, reveals Ri = R = 180 Ω, and v = 160; with a load of 8 Ω1 (RL) the voltage gain
will be vU = 6.8. The factory-set negative feedback is k = 0.056; with it R drops to Ri = 18 Ω,
and the gain to vU = 4.9 (measurement: Fig. 10.5.16). The low-frequency range reveals an
interesting twist: due to phase-shifts, a positive feedback comes into play here! Enlarging the
input capacitor of the differential amplifier (from 1 nF to e.g. 100 nF) will, however, keep the
output impedance in the low-frequency region (with NFB) almost constant (Chapter 10.4.3).
Fig. 10.5.16: Left = magnitude of the output impedance. Right = frequency-response, from the input of the differential amplifier (ahead of the
1-nF-capacitor) to the loudspeaker (4xP10-R).
“Endstufen-Innenwiderstand” = power stage impedance; “Betrags-Frequenzgang” = magnitude frequ. fesponse;
“ohne Gegenkopplung” = without NFB; “mit Gegenkopplung” = with NFB.
♣
For non-linear operation (overdrive), any negative feedback will loose its effect since the control value (here:
output voltage) can practically not change anymore.
1
Note that the Super-Reverb-specimen investigated here had an output transformer with not just the customary
2-Ω-output, but also an additional output for 8-Ω speaker-matching.

Not all power stages include negative feedback: the VOX AC-15 and AC-30 (and some
others, too) completely dispense with any feedback. In Fender amps, the situation is mixed:
very early power stages have no feedback, the Bassman 5B6 receives NFB around 1952, the
Deluxe 5E3 only as late as 1955 [D. Funk]. The Bassman acquires special significance due to
its Presence-control included in the negative-feedback circuit: it enables the frequency
dependency of the feedback to be set by the user (Chapter 10.3.3). Legend has it that Jim
Marshall and Ken Bran were particularly inspired by the late 1950’s Fender Bassman when
they developed their amps, and therefore we find a power stage with integrated presence
potentiometer in Marshall amps, as well.
As early as 1943, Frederick Terman describes, in his remarkable "Radio Engineers'

Handbook", the effects of power-stage feedback on gain, internal impedance, and harmonic
distortion. The reduction in gain that accompanies negative feedback certainly was not an
express objective of the circuit designers, but they put up with it in order to reduce the non-
linear distortion of tubes and output transformers. In the 1950’s, there was no Heavy-Metal
music scene, and playing was mostly “civilized” i.e. undistorted. Presumably, the pioneering
developers observed the output signal of their amplifiers with an oscilloscope, and tried to
reproduce sine curves as perfectly as possible: “by the book”, as D. Funk writes. The more
negative feedback is introduced, the less an amplifier distorts – that’s what the book said. It
was also known that strong phase shifts may turn NFB into positive feedback – although not
every designer would or could do much of the required calculating. In any case, the designer
would soon discover that, with too strong a negative feedback, the amp would start to self-
oscillate, and so the NFB was adjusted empirically to a degree that would avoid instability
within the framework of the given manufacturing tolerances.
In Fig. 10.5.17, we see the frequency responses of a bandwidth-limited system: in the left-
hand section for slightly different filter flanks, and with and without negative feedback. The
more narrow-band system (dashed lines) does not only receive the expected gain-reduction
but also considerable resonance peaks at the frequency limits. “Negative feedback” means
superposition of an opposite-phase signal. However, the phase-shifts that live in every circuit
with bandwidth-limiting will have the effect that around the band-limits, the opposite-phase
correction-signal can turn into an (almost) same-phase positive-feedback-signal that increases
gain. Increasing the gain in the forward branch (right-hand section of the figure) will
disproportionally increase the overall gain (blue) in the fringe ranges. In fact, this may occur
as an effect of just a tube change. The new power-stage tubes will generate a stronger bass
response due to their slightly higher transconductance, and right away the test report in the
music-mag will read: “the KT-X delivers more bass than the 6L-Y”. This characteristic,
however, needs to be always seen in connection with the specific individual circuit. Power
tubes will transmit from 0 Hz to about 100 MHz – but only the teamwork also including
transformer, speaker-load and NFB-network will result in the individual frequency response!
Fig. 10.5.17: Effects of negative feedback when a bandwidth-limitation is present. “Gegenkopplung” = NFB; “mit” = with; “ohne” = without.

10.5.7 Internal impedance of the power stage
Tube circuits are of high impedance while loudspeakers have a low impedance. The output
transformer – with the term “matching” appearing in its description – serves as mediator
between these different impedance levels: indeed, the output transformer matches the
different impedance levels to each other. Usually, the term “matching” indicates that source-
and load-impedance are of equal magnitude. For a tube amplifier, this would mean that its
internal impedance is decreased to the level of the loudspeaker by the output transformer, i.e.
is brought down to e.g. 8 Ω. This strict definition of the term “matching” should, however,
not be used in the context of tube amplifiers; the impedance levels are brought closer to each
other, but they are not actually (power-) matched. In a tube amplifier specified for a nominal
loudspeaker-impedance of 8 Ω, the internal impedance of the amplifier (the source
impedance) will normally not be 8 Ω but much more, e.g. 100 Ω. Tube amplifiers operate
“almost as current sources”, and not with power-matching.
For HiFi-loudspeakers, such a current imprint is usually unwanted because an emphasizing of

the loudspeaker resonance will result. Customary is the operation from a source with very low
impedance that reduces the undesired high Q-factor [3]. This ideal can be achieved almost
perfectly with transistor power-stages, while in tube power-stages, the impedance can be
reduced via negative feedback – but not to the same degree as in transistor amplifiers (phase-
shifts, tendency to self-oscillate). The question whether negative feedback is at all desirable
in guitar amplifiers has received quite different answers in the past: no negative feedback in
almost all VOX amplifiers and very early Fender amps; inclusion of negative feedback in
almost all Fender amps from the early 1950’s. The amplifier with NFB reacts “more
civilized” with lower non-linear distortion compared to its feedback-free counterpart – at least
as long as it is not overdriven. Whether the lower distortion is felt to be an advantage is a
matter of taste and shall not be the subject of an evaluation here. However, since the negative
feedback does not only influence harmonic distortion and dampening of the loudspeaker, but
also has an effect on the source impedance of the guitar amplifier, another question becomes
obvious: can the output power be increased via negative feedback? The NFB as we see it in
power stages does decrease the (“too high”) internal impedance of the amp – it should be
possible to interpret this aspect as an improvement of the power-matching situation.
Things are not that simple, though – the tube is a non-linear component that is only
inadequately described by the theory of linear two-ports. Experience shows that it is
conducive to distinguish between (approximately) linear and (strongly) non-linear operation.
For small drive-levels, the power stages works approximately in a linear fashion♣. In this case
the internal impedance of the tubes can be estimated from the slope of the output
characteristic. Depending on the type of tube and on the operating point, we can expect an
internal tube-impedance of 10 – 100 kΩ. If we would now chose – in order to achieve power
matching – the load resistor at the plate exactly as big as the internal impedance of the tube
(e.g. 100 kΩ), then the AC plate-voltage would have to be 3.1 kVss in order to reach P = 12 W
… no normal tube could withstand that. Equal impedance definitely is not the desired goal;
rather, the output power is to be maximized while considering the given limit values. In the
chapters on the specific push-pull power stages, we will give guidelines for calculating the
optimal load resistor at the plate – typically, values around 1 – 2 kΩ are the result.
♣
Given that a sufficient bias-current has been set in the case of push-pull operation.

The optimal load-resistance for the plate is considerably smaller than the internal impedance
of the tube, and therefore the internal impedance of the transformer is considerably larger than
the nominal impedance of the loudspeaker. The loudspeaker voltage consequently depends
strongly on the loudspeaker impedance. This is not a disadvantage, though, because – in
contrast to HiFi-loudspeakers – emphasizing speaker- and enclosure-resonances is not
generally undesired in guitar amplifiers. In fact, it is seen as a special sound characteristic and
even asked for in many cases. Still, it needs to be considered that the stiff-current-source
feature dies a rather sudden death as clipping occurs. The power stage is of high output
impedance only while it remains in linear mode; for overdrive, the plate-voltages of the push-
pull amplifiers (and therefore the output voltage of the output transformer, as well) hit a
relatively rigid border: the residual voltage of the tube (e.g. 50 V). Fig. 10.5.18 shows the
output characteristic of a 6L6-GC in combination with a few internal impedances. Just like
any other tube, the 6L6 does not have one single internal impedance. Rather, the latter is
strongly dependent on the drive-signal level, and it changes by up to two orders of magnitude
as the operating point shifts. The tube will be of high impedance (about 70 kΩ) in the range of
usual bias currents (e.g. 350 V, 30 mA), but become of lower impedance at the overdrive-
limit (e.g. 50 V, 200 mA). Of course, this is not really that surprising because the tube is a
strongly non-linear component. We need to always remain aware of this, especially since the
theory of LTI-systems with its relatively simple calculation methods is all too alluring – just
like it is also deceptive. Connecting a 14:1-transformer to the 6L6-GC shown in the figure,
the transformation will be from 72 kΩ to 367 Ω, which is a high impedance in comparison to
an 8-Ω-speaker. The transformer will, however, transform the 2 kΩ to 10 Ω.
Fig. 10.5.18: Output characteristics of the 6L6-GC, including a few internal impedances. Ug2 = 300V, Rg2 = 0.
The internal impedance specified in the datasheet of a tube is an orientation-value that may
be used for small-signal considerations as a rough approximation. More extensive calculations
using it are not advised; first, because the power stage rarely operates under small-signal
conditions, and second, because the internal impedance is specific to a given operating point,
and on top of that it also depends on the voltage at the screed-grid. For the 6L6-GC, the
datasheet specifies an internal impedance of 33 kΩ (class-A). This is a good match to the
measurement data given above but remains usable only for very few guitar amplifiers because
they normally operate in class-AB-mode. For the latter, datasheets usually do not give any
internal impedance – rather, the optimum load impedance is given. This optimum load
impedance – and not the internal impedance – may serve to calculate the transformation ratio
of the output transformer. To calculate the source impedance RQ (as it is “seen” by the
speaker) for class-AB operation, several peculiarities need to be considered. For small drive
levels, the two power tubes cooperate and RQ is halved: with a 10:10:1 transformer, we obtain
Ri = 60 kΩ → RQ = 300 Ω. For high drive levels, only one tube is active at a time (for each
half wave). Moreover, we need to consider that the transformer is not at all ideal, either: RQ is
reduced by the (non-linear!) main inductance and the capacitance of the winding.

Fig. 10.5.19 shows measurements of a power stage (JTM-45, KT-66, GZ-34). The uppermost
curve results from switching-off the supply voltage (Standby): the tubes (with the heater in
operation) do not insulate perfectly but are of rather high impedance – transformer-
capacitance and -inductance determine the impedance. Switching on the supply voltage but
keeping the negative feedback deactivated (no NFB) reduces the internal impedance RQ
because the tubes are now operated in the operating point. As we switch on the negative
feedback, RQ experiences another, very pronounced drop. Since the frequency response of the
feedback loop can be modified by the Presence potentiometer, various characteristics may be
realized. The strong resonance peak at 7 kHz is due to phase-shifts caused by the presence-
filter (low-pass within the negative-feedback loop, see also Chapter 10.3.3). The increase in
the no-NFB-condition does not happen proportionally to the frequency: this is due to the non-
linear main inductance that depends on the drive levels, and on the operating point within the
hysteresis (Chapter 10.6).
Fig. 10.5.19: Internal impedance of the JTM-45-power-stage. Left for the 8-Ω-output, right for 16 Ω. The right-hand picture also shows the
effect of the bias on the internal impedance.
The impedance also depends on the bias-current of the power tubes, and on the power-tube
type. If the power tubes are not equal, pre-magnetization effects of the transformer core weigh
in, as well. Because of these dependencies, it is advisable to take from Fig. 10.5.19 not more
than the fact that output impedances around 100 Ω occur without negative feedback. Any
exact data or frequency responses would be too much connected to the individual amplifier.
On the other hand, taking into account that normally the feedback-loop in the KTM-45 is
closed, the differences in regular operation may not be that big, after all. With closed NFB-
loop, we see an astonishingly small internal impedance (= magnitude of the output
impedance) of the power stage of merely 2 Ω (for the 8-Ω-output). This power amp does have
efficient NFB! Well, that’s the case at least if we don’t turn up the Presence-control too much
… Who would have though that Marshall (not actually known for any HiFi-designs) would
decrease the output impedance via negative feedback (that will decrease distortion) to values
that are significantly below the load impedance!
At Marshall, this take on things would not always remain, as the 18-W-amp developed later
proves. Its two EL84’s operate in a power stage that entirely does without any negative
feedback. VOX immediately comes to mind, but allegedly the “Watkins Dominator” was the
inspiration for Ken Bran [Doyle]. Still, there are no big differences to the AC-30 with regard
to the internal impedance, as Fig. 10.5.20 shows. Also generally valid: such measurement logs
are snapshots – every tube-swap will change the amplifier-parameters and as such also the
internal impedance. Chapter 10.5.11 will elaborate on how far selecting the power tubes helps
to avoid inter-individual differences. Moreover, the effects of the bias-setting are discussed in
Chapter 10.5.8.

Of the output-impedances documented in Fig. 10.5.20, three are taken from power stages that
do not include negative feedback: VOX AC-30, Marshall 18W and Tweed Deluxe; the Super-
Reverb does have NFB but was additionally measured with open negative-feedback loop.
Fig. 10.5.20: Internal impedances of several guitar amplifiers (Super-Reverb with and w/out negative feedback).
In view of these significant internal impedances, we could ask where the energy necessary for
their operation is actually sourced. If an 8-Ω-resistor absorbs 30 W and if it has serially
connected an impedance of 75Ω (the internal impedance of the amplifier), will the latter then
absorb 281 W? Is that why the VOX gets so hot? No – with this thinking, we would in fact
abuse a model. With respect to a specific given problem, an equivalent circuit shows the same
behavior as the real structure [20]. In our case, however, the given problem (the real source
and its replacement by an ideal source including an internal impedance) is not the energy
balance. Rather, the difference between source voltage and terminal voltage is to be
illustrated. We see this right away as we replace the voltage source (with a serially-connected
internal impedance) by a current source (with a parallel-connected internal impedance): with
open terminals, the internal losses for the current source are at a maximum, and zero for the
voltage source. The model of a source with internal impedance is well equipped to explain the
dependency of the output voltage on the load impedance (loudspeaker impedance): for a low-
impedance source, the terminal voltage is practically independent of the load, while for a
high-impedance source, the terminal voltage is practically proportional to the load impedance.
Model and reality are a good match for visualizing the given problem. However, the model is
not suitable to determine dissipation power in tubes: the actual (real) voltages and current in
the tubes need to be considered for this.

10.5.8 The bias-current in the power stage (bias-setting)
The bias-current in the power stage is the current that runs through a power tube when it is “in
idle”-state. “In idle” means that all supply voltages are present (in contrast to the “standby”-
state) but no input signal is fed to the circuit (volume = 0). In some circuits, the bias-current
sets itself automatically (cathode-resistor, Chapters 10.5.2 and 10.5.2), in others it can be
adjusted within certain limits by a potentiometer (bias pot). For such an adjustment,
instructions are required: what is the optimum setting, and how (and what) do we measure?
What is measured? The bias, of course! “If the mains voltage drops from 230 V to 220 V
(translator’s remark: or from 110V to 105 V, if your are in the corresponding part of the world), the bias
changes by a few milliamps, and the sound does change”. Of course: the bias! Adjusting that
just right, the sound will be right! The bias; that’s apparently the idle-current in the power
stage. But which one – the cathode- or the plate-current? This is where it already gets tricky
(and slippery) for most experts. Recommendation 1: you measure the bias by disconnecting
the plate from the output transformer and putting an ampere-meter in between. This expert
clearly targets your plate-current. And if, while you’re measuring away, suddenly Jimi H.
appears and invites for a jam: then the insulation was inadequate. Because plate-voltages can
be – SERIOUSLY! – absolutely DEADLY! Measurements of this kind are not something
the layperson can do; real expert knowledge is required. Recommendation 2: You measure the
bias by connecting an ampere-meter in parallel to the primary winding of the output
transformer. Our second expert also targets the plate-current and sees a measurement error
(due to the copper resistance) of 5 – 10% as unproblematic. These primary windings are not
of that high an impedance, possibly as low as 30 Ω. For an orientation measurement, this is
good enough, though. Recommendation 3: You insert (solder) a 1-Ω-resistor into the cathode-
connection of every power tube and you measure the voltage drop across it. Ooops – now
we’ve jumped to the cathode-current, i.e. the sum of the plate- and the screen-grid-current.
For a plate-current of 35 mA, the screen-grid-current may well be 5 mA with the result that
the cathode-current is 40 mA. If we think of a 5%-change in the mains voltage as substantial,
we should not include a 14%-error in our current measurements. Most serious datasheets
specify the plate-current in the operating point; measurements at the cathode resistor would
give us the cathode-current. That in fact is no problem if the screen grid (g2) is operated with
a grid-resistor in series (e.g. 470 Ω): in this case the screen-grid-current can easily be
calculated from the voltage drop across this resistor. Still: CAUTION! This measurement, too,
can have a deadly conclusion … the same danger that always exists when doing measurement
on the opened-up amplifier. Do observe all regulations!
Instead of recommendations relating to the plate- or cathode-current in idle state, we also find
hints towards an optimal setting of the bias-voltage at the grid: adjust to -42 V at the grid (g1)
of the power tube. Indeed, this also is a workable approach: measure – using a volt-meter with
high input impedance – the voltage between grid and cathode with no drive signal present: the
more negative this value, the smaller the plate current, i.e. at -50 V there is less current
through the tube than at -40 V. The actual current value is, however, not revealed this way.
So, what is the correct voltage or the correct current? Answers fill many thousand pages on
the Internet; it’s a science in itself. Correction: it’s a playground for self-proclaimed experts,
not a science as such. Searching for Ohm’s law, you will consistently find U = RI. Looking
for rules to set the bias, results are contradictory. One advice might be to use an oscilloscope
and “turn up the bias until the kinks in the curves disappear”. Plausible, that one: the spelling
is correct – must be a studied person. But the next entry calls exactly this method: “couldn't be
further from the truth”. Is it even more plausible because the guy has 1532 postings?

Well then, let us add version 1001 to the 1000 existing ones. First, however, and as always,
we need to suffer through some basics. In push-pull power stages (and only those are
discussed here, anyway), the audio signal is first dissected into two parts that are amplified
separately and then re-joined. The separation- and re-joining processes are error-prone, and it
is here that the bias-current adjustment helps out. Changing the bias-current may improve the
sound – or make it worse if you don’t do it right. If the bias-current is set too low (cold
biasing), distortion of the not-so-nice kind appears. At the same time the power-stage has an
expander-effect: a lightly plucked string will be reproduced too softly, and with a stronger
attack the amp suddenly roars. For the bias-current set high (hot biasing), the sound is good (if
there are no other issues). So, should we set the bias-current as high as possible? No, don’t –
that will reduce the power-tube lifespan (which anyway is relatively short) even further, and
could possibly destroy the power supply if it is under-sized. This would be the main effects.
In the details, 2nd-order-effects show up, as well. For small bias-currents, the filter-capacitors
in the power supply get charged to a higher voltage, which might give the preamplifier- and
intermediate-amplifier-stages a different operational behavior. We should not expect big
effects from this but it should be mentioned for completeness sake. A small effect could also
manifest itself in terms of the impulse-power i.e. the power measured at the onset of a tone. If
the filter caps are charged to a higher voltage, the impulse power, too, will be a little higher. It
is, however, not purposeful to reduce the bias-current just because of such effects – the
distortion connected to the readjustment is normally not acceptable. If we do not start with the
details but stick with the main effects, we have a simple rule: low bias-current = distortion,
high bias-current = premature death of the tubes.
But then, we also find: high bias = distortion, low bias = tube-death. How can that be? Simple,
actually: the experts, in particularly the self-proclaimed ones, writing (rather: allowed to
write?) their columns in the guitar-magazines do have very different educations♣. The term
bias is not always meant to refer to the actual bias-current but may be used as for the bias-
voltage fed to the grid of the power tubes in idle. This is where a mix-up may well happen,
and even a double mix-up at that, because for the negative bias-voltage, it is easy to confuse
the magnitude of the given number and the actual value (with a “-“-sign). A lower (more
negative: e.g. -50 V instead of -40 V) bias-voltage leads to smaller bias-current (and vice-
versa), but this means that the larger absolute number (50 vs. 40) corresponds to the smaller
bias current (and vice versa). All this is now connected to the one term “bias” in many not-so-
professional publications. What does an author seek to express when he/she writes “turn up
the bias”? Should it be more bias current (idle-current) i.e. plate-current (or even cathode-
current!) when no input signal is present? Or more voltage fed to the grid via the bias pot? If
the latter: more voltage in absolute numbers (i.e. go to from 40 V to 50 V, both voltages being
negative), or higher voltage in terms of physics (i.e. go from -50 V to -40 V)? It’s all rather
complicated, and one person implies this while the other understands that – but only because
(and if) unclear terminology is used. Therefore, let’s talk about idle-current, or bias-current, or
grid-bias-voltage or even bias-voltage (with a clear “-“-sign, and watching the polarity of our
meter); but let’s avoid “bias” without further specifics. That term is simply not precise
enough.
♣
The corresponding scale (no lower boundary) includes the rating „has not a single clue whatsoever, at all“.

In the following, the term “bias-current“ is used to designate the plate-current flowing in one
power tube when no input signal is present. Alternative terms for “bias-current” would be
"quiescent plate-current" or "idle current". We will use the term “bias-voltage” to indicate the
DC-voltage fed to the control-grid of the power tubes when no input signal is present. It is
always a negative voltage. An alternative term for “bias-voltage” would be “grid-bias-
voltage”. This terminology will lead to e.g. the precise statement: “At a bias-voltage of -50 V,
a bias-current of 38 mA flows.” Only one possible pitfall remains: it needs to be clear that the
bias-voltage is the voltage measured from grid to cathode. In case the cathode is connected to
ground via a resistor (and possibly a capacitor in parallel), the grid/cathode voltage is not the
same as the grid/ground-voltage. When taking a measurement, this is important. In the
following, “bias-voltage” always indicates the grid/cathode-voltage.
Cold Biasing indicates that the bias-current is set to a relatively low value – corresponding to
the “very negative” bias-voltage. In line with what has been said above, it is a bit problematic
to use the terminology “small bias-voltage” because not everyone may understand that -50 V
is smaller than -40 V. For a thermometer, the situation would be clear: -20 C° (or F°) is colder
than -10 C° (or F°), and so -20 C° (or F°) is the colder/lower/smaller temperature. Despite a
clear separation into topological and metrical scales, the Internet community has found its
own interval scaling, and we may read the terminology “turning up the grid voltage” as a
readjustment from -40 V to -50 V. Whether the voltage or the magnitude of the voltage is
increased – it does make a difference. Hot Biasing is the other extreme: high bias-current, and
a “less negative” bias-voltage. In other words, to safely avoid all doubts: the smaller the
magnitude (i.e. the numeric value) of the (negative) bias-voltage, the hotter the tube is run. 10
mA/-60V is cold, 80 mA/-40 V is hot – just as an example! Because we will see in the
following that these numbers are circuit-specific; one circuit’s “cold biasing” may well be the
other circuit’s “relatively hot”.
In Fig. 10.5.21 we see both the characteristics of the individual tubes (dashed) and the overall
characteristic generated by superposition. The center picture shows a “cold operating point”
i.e. “cold biasing”. With little change in the drive level (the voltage indicated on the
abscissa), neither of the tubes feels animated towards much activity – both still are in blocking
mode and the overall current remains small. Only for larger input voltages, the tubes start to
move into the respective (alternating) conducting state and the current increases. The result is
a saddle-shaped crossover-distortion. In the left-hand picture, the situation is different: the
bias-currents in the operating point are higher, the overall (summed) characteristic retains its
incline over a broad range and only curves clearly at the overdrive-limit.
Fig. 10.5.21: Characteristics for two different settings of the bias-current. On the right, the distortion relating to the characteristic in the center
picture is shown (“crossover-distortion”).

The cold biasing shown in the center picture has two effects: undesirable non-linear
distortion, and an expansion just as unwanted. In contrast to a compressor, an expander
increases its gain with increasing signal level – not a stylistic device many guitarists welcome.
Corresponding measurement data are shown in Fig. 10.5.22: in a Super-Reverb, the bias-
current (Ia) of the power stage was varied between 10 mA and 53 mA. With a bias-current set
to 10 mA, the output level increases by 27 dB as the input level rises by 10 dB – this is
already a noticeable expansion. The right-hand picture shows the corresponding 3rd-order
distortion level – again there are clear differences.
Fig. 10.5.22: Fender Super-Reverb (2x 6L6-GC), bias-voltage varied between –65V and –40V.
Left: output signal level vs. input signal level; right: distortion level vs. input signal level. NFB disabled.
In order to document the effects of the bias-current on the forward branch, the measurements
for Fig. 10.5.22 were done with the negative feedback loop left open. With active NFB (Fig.
10.5.23) we see similar curves with a minimally weaker expansion and slightly lower
distortion. For the “hot” operating modes the difference in the distortion is clearly visible; for
the operation with low bias-current we need to consider that for equal input levels, the output
levels differ considerably, after all.
Fig. 10.5.23: Fender Super-Reverb (2x 6L6-GC), bias-voltage varied between –65V and –40V.
Left: output signal level vs. input signal level; right: distortion level vs. input signal level. NFB enabled.
Usually, the Super-Reverb will not be operated with a bias-current as small as shown via the
blue curves. A larger bias-current (about 35 – 45 mA) would be normal. However, the bias-
current should not be much larger, either, because the plate-loading would then possibly enter
the critical range (see Fig. 10.5.26).

Variations of the bias-current in the power stage will change distortion, gain and dynamics,
and also alter the internal impedance. We have already seen in Chapter 10.5.7 that the internal
impedance of a tube is not constant but depends on the operating point. The internal
impedance transformed by the output transformer therefore depends on the OP, too. With a
high internal impedance of the power stage, the loudspeaker experiences less dampening, and
resonances influence the transmission behavior more strongly. Moreover, since the speaker
impedance rises towards high frequencies (Chapter 11), the high internal impedance results in
a treble boost. Fig. 10.5.24 shows measurement results for a Super-Reverb.
Fig. 10.5.24: Super-Reverb. Left: magnitude of the output impedance; right: transfer-function of the power stage.
NFB active; amplifier loaded with loudspeakers (4x Jensen 4xPR-10). Different bias-currents.
Operating the amplifier with a small bias-current (cold biasing, Ia = 15 mA), the internal
impedance of the amplifier (with active negative feedback) amounts to 30 Ω – this is
relatively high compared to the load impedance. The resonance peak and the treble boost are
more pronounced than for the “hot biasing” shown in red (Ia = 50 mA).
A different picture emerges for the Marshall power amplifier which features stronger negative
feedback. The output is of significantly lower impedance, and the loudspeaker impedance
maps onto the output voltage to a much lower extent (Fig. 10.5.25). For usual settings of the
bias-current, the 16-Ω output of the JTM-45 is lower in impedance compared to the Super-
Reverb by a factor of five! However, it would be wrong to conclude that the Marshall
could/should be operated with a loudspeaker of smaller impedance (or the Fender with a
loudspeaker of higher impedance): the optimum load-resistance is not directly derived from
the internal impedance but from the limit data of the tubes.
Fig. 10.5.25: JTM-45. Left: magnitude of the output impedance; right: transfer-function of the power stage.
NFB active; amplifier loaded with loudspeaker (Marshall 1960-AX). Different bias-currents.

The above pictures show that the transmission characteristics (and therefore also the sound) of
a power stage depend on the setting of the bias-current. The latter also influences the power
dissipation, and the following is dedicated to this issue. The less negative the bias-voltage is,
the larger is the plate-current and the larger the power dissipation at the plate. We frequently
read that the power dissipation at the plate (without input signal) should be 70% of the
maximum permissible power dissipation. As an example: for the 6L6-GC, specified at 30 W,
this would be 21 W (e.g. 47 mA at 450 V). It is not purposeful to search for the origin of the
70%-rule – that would be much too speculative. More conducive is to build an example
explaining the strain that the tubes experience. Fig. 10.5.26 shows three load lines of the 6L6-
GC drawn into the output characteristic. For the upper line, we assume a bias–current of 47
mA, and for the lower line one of 33 mA. We calculate (for a plate-voltage of 450 V) a power
dissipation in idle of 21 W (70%), and 15 W (50%), respectively. Although the dissipations in
idle differ by as much as 40% (21 = 1.4 ⋅ 15), the maximum power for Ra = 1.3 kΩ varies by
only 7%. Only changing the load-impedance from 1.3 kΩ to 1 kΩ brings larger differences in
the strain on the plate. What about the mean values of these curves? They depend on the
individual drive levels. The worst-case would be a square-shaped plate-current of an
amplitude of 200 mA (1.3 kΩ); the determined instantaneous power would have to be halved
because each power tube conducts only for one half-wave. For Ra = 1.3 kΩ, the maximum
allowable power dissipation at the plate is not reached – it is, however, already slightly
surpassed for Ra = 1.0 kΩ.
Fig. 10.5.26: Output characteristic of the 6L6-GC. AP = operating point without input signal.
Right: power dissipation at the plate dependent on the plate-current. Load impedance = 1.3 kΩ / 1.0 kΩ.
In conclusion: regarding the strain on the plate, the load-impedance is much more important
than adjusting the bias-current to the second decimal. If the load impedance becomes too
small, the plate will be overloaded. Of course, the type of drive- (or overdrive-) signal plays a
role, as well – as does the voltage at the screen-grid … and as does the plate-voltage. Most
everything that can change does change. Therefore there is no harm in calculating a load line
once in a while – but only if we do not seek to adhere slavishly to the results. In Fig. 10.5.26,
the operating point was assumed for 450 V. However, with the presence of a drive signal, the
voltage delivered by the power supply does not remain constant but may easily change by as
much as 50 V depending on the load… the strain on the plate will change correspondingly.
Also, the load line will deviate even much more from the normally assumed straight line. No
loudspeaker has a constant and purely ohmic impedance (Chapter 11). Rather, the
loudspeaker impedance is complex (voltage and current are phase-shifted re. each other), and
its magnitude can easily vary by a factor of 10 depending on frequency. Calculations using
straight load lines are highly idealized models – nothing more but also nothing less. Reality is
very different, in any case.

In reality, a guitar amplifier is neither driven by sine-tones nor is it loaded with a purely
ohmic resistance. Neither its supply voltage nor the voltage at the screen-grid are constant.
Fig. 10.5.27 shows a first step in the direction of reality: these output characteristics were not
determined on the test-bench but from a real guitar amplifier.
Fig. 10.5.27: Output characteristic; Super-Reverb with purely ohmic load (left), and with complex load (right).
For the measurements shown on the left, the amplifier was loaded with a purely ohmic 8–Ω–
resistor. Small drive levels yield a small, straight line through the operating point. This line
grows as the drive level increases, bends and shifts to the left. Consequently, even an ohmic
load does not generally warrant assuming a straight line passing through the operating point.
This is because on one hand the supply voltage drops, and the other hand the coupling
capacitors are polarized due to the current flowing through the grids (Chapter 10.4). The
curves on the right are for a complex loudspeaker-load (f = 3 kHz). For small input levels we
see ellipses encompassing the operating point; large drive levels result in sharply bent curves
that extend into the range of 30 W – which has been specified as limit. Since that value needs
to be seen as short-term power average, this transgression does not generally indicate a
thermal overload of the tubes (Chapter 10.5.9).
Fig. 10.5.28: Output characteristic, Super-Reverb with loudspeaker load using guitar tones (Stratocaster).
Even closer to reality are the curves depicted in Fig. 10.5.28: with a loudspeaker as load, the
Super-Reverb was played with a guitar – and no sign of any straight load line at all remains!
Rather, there is a myriad of highly different loops that only with great difficulty allow for any
conclusion regarding the setting of the bias-current (white circle). Therefore, the load line is
unsuitable to establish any connection between bias-current and power dissipation in the tubes
– to do this, true measurements of the power dissipation are necessary.

In order to measure the power dissipation at the plate, the plate-voltage and the plate-current
need to be recorded. Just multiplying the RMS-value of the plate-voltage with the RMS-value
of the plate-current is not sufficient because that way we would merely determine the
apparent power [20]! So: anybody connecting a volt-meter to the plate, and an ampere-meter
serially into plate-connection, will indeed measure Ua and Ia, but the product of these values
will only give information about the strain on the tube in a DC-situation. With an input signal
present, however, AC results – and here we need to distinguish between effective power,
reactive power and apparent power. It is the effective power (the product of plate-voltage and
plate-current averaged over time) that heats up the plate. It is important to understand that it
makes a difference whether the multiplication comes first (ahead of the averaging – correct
for the present considerations) or averaging comes first (ahead of the multiplication –
incorrect in the present case). Fig. 10.5.28 has impressively shown that for this 6L6-GC, the
short-term dissipation at the plate exceeds 100 W – more than three times the value specified
as a maximum. A tube needs to be able to take such a short-term overload if it is to be
successfully deployed in a guitar amplifier. As we switch on the plate-current, the temperature
of the plate begins to rise – thermal energy is supplied. At the same time, thermal energy is
dissipated via radiation, and after some time a steady equilibrium, i.e. a constant plate
temperature, is reached. If this temperature is too high (with the plate glowing brightly), the
tube dies. If we do not wait until the equilibrium is in place, the temperature remains below
the steady final value. Compare this to a car: stepping on the accelerator for only 2 seconds
will not give you maximum speed. For a tube, though, 2 seconds would already be relatively
long – in any case the typical short-term overload situations in a guitar amplifier will be of
lesser durations.
Fig. 10.5.29: Instantaneous power dissipation at the plate (black), sliding 10-ms-average (magenta).
The lower section shows excerpts from the progress of the plate-voltage, -current, and -power-dissipation.
Fig. 10.5.29 shows two examples for a loudspeaker-loaded Super-Reverb. The instantaneous
values of the plate dissipation (sampled at 48 kHz) reach about 140 W. Averaging over 10 ms
still gives values significantly in excess of the specified 30-W-limit. The lower line in the
figure indicates that this strain happens in a frequency range unusually low for guitar tones:
interpreting 10 ms as half a period, we get 50 Hz. This is not connected to the mains
frequency but results from effects of overdrive-related recharging processes in phase-inverter
(Chapter 10.4) and power stage. Since such short-term overloads can happen repeatedly, it is
purposeful not to set the bias-current too high so that little effective power is fed to the tube in
between such high-power phases. The question now remaining is: how high is a bias-current
set “not too high?

So, at last: what is the correct setting of the bias-current (the “bias setting♣ ")? Unfortunately,
there is no formula that will generally hold – the circuits, tubes, loudspeakers, and ways of
playing that can occur are too diverse. Nevertheless here are some basic recommendations:
For group 1 including laypersons i.e. persons without any education in electronics: whoever
is not clearly aware of the reasons why in a 40W-amp voltages of over 800 V may occur, and
who does not know how to protect him-/herself against the corresponding deadly dangers,
must not open up an amplifier. Studying the manual of a multi-meter must not be understood
as an education in electronics, and the same holds for confident handling of a screwdriver.
Not everybody who removes an amp chassis from a cabinet instantly keels over dead – but
this fact must not lead to the conclusion that this will never happen. If we are not allowed to
open an amp, we can merely resort to measurements using a socket-adapter. The latter should
be certified and re-checked regularly according to local regulations because it is subject of the
same high voltages. Equipped this way, we now (more or less incorrectly) consider ourselves
part of group 2.
Group 2 includes appropriately trained persons (e.g. certified electricians) who have simple
measurement devices at their disposal. They should be in the position to adjust the bias-
current without being in danger, should be able to recognize whether a power stage operates
in true class-A mode (BIG exception), and then be able to find – using an oscilloscope – the
middle of the load line. For an amplifier working in class-AB mode, the only helpful approach
is a mixture of listening-tests and simple measurements of the power consumption in idle: if
the amp already sounds good at 50% of the allowable plate-dissipation (e.g. 450 V, 33 mA for
the 6L6-GC), you should just let it be. If your hearing (or the musician looking over your
shoulder) demands more, you can run the thing a bit hotter – but at 70% most practitioners
will raise an admonitory finger although there is no theoretical foundation for this limit. In
any case, the power tubes need to be looked at in the dark to check whether, during any phase
of extensive and multifaceted testing, grids and/or plate are visibly glowing. That this testing
is not to be done with just a sine-generator and simple load-resistor should be – in view of the
above – crystal clear by now.
Group 3 includes persons belonging to group 2 who have special instrumentation equipment
in their arsenal, for example a current clamp that can measure with a resolution of 10 mA or
better, and in the frequency range of 0 – 10 kHz. Seriously: 0 Hz – because the DC-
components need to be measured, as well, and thus a frequency limit of 1 Hz is useless.
Suitable would be e.g. the Tektronix AM305B/A6302, with the offset continuously
monitored. Given such a current-measuring device and a high-voltage test-probe for the
voltage measurement, you can then capture the factors determining the plate dissipation,
digitize them in the calibrated front-end, store them in the computer and derive the actual, true
loading of the tube. Once you’ve gotten that far, inevitably the question will arise whether
today’s tube manufacturers will actually still adhere to the tube data from the 1950s, and will
warrant e.g. 800 hour MTBF for their products. The other immediate question is whether
indeed every tube-wholesaler who allegedly cooperates in the development of “his” special
tubes will expend such an effort.
In the case that “no” is the answer to these questions, we quickly move to become members of
group 0. Here we join all those who have noticed that old Marshalls or Fenders did not even
have any means to adjust the bias-current – but still did their job admirably. And so we
change the tubes, if need be, and that’s it.
♣
too much of a bias never is a good thing – that seems to hold for all aspects in life.

10.5.9 Tube-strain and -aging
During regular operation, the power tubes of a guitar amplifier become very hot: 250 °C can
be easily found at the glass container, and a much higher temperature within it. It is this heat
that makes the utilized materials age, that will destroy coatings and consequently deteriorate
the operational data. Of course, tubes may also break – but that is not normally the reason for
their failure. Common is:
• The cathode coating evaporates from the cathode and deposits on other electrodes;
overload accelerates this process.
• An insulating intermediate layer may form on the cathode between the carrier (e.g.
nickel) and the emitting layer (mostly barium-oxide).
• Gases released from the electrode metals impair the vacuum.
• Mechanical vibrations bend or destroy electrodes.
In the MOV-datasheet for the GEC-KT-66 we find the lifetime specified at “minimum 8000
hours”. Consequently, an amateur playing 10 h per week at high volume would not need to
be worried for 15 years, and even the pro (with 8 h per day of loud playing) would be able to
enjoy that set of power tubes for 3 years. Author Helmuth Lemme [1995] quotes entirely
different time-spans: he recommends to tentatively change the power tubes after 100 hours.
Edward van Halen apparently acts even more rigorously: allegedly, he has (or had) all power
tubes replaced after every gig. Checking the MOV-definition for “end of tube life” more
closely, the 8000-hour-euphoria is brought down a peg or two: we find that the output power
has gone down to 50% i.e. as long as a 100-W-amp will yield 50 W, the tubes are deemed o.k.
A guitar player will hardly be content to work with half the specified power, though, and
therefore the MOV-definition is lacking in practical relevance. MOV is silent about the
characteristic of the drop in power, we only find that reducing the strain on the tube by 40%
will extend the lifetime by 25%. That does not help us to draw any conclusion about the
lifetime under overload conditions. The latter often appear in guitar amplifiers; operating
outside of the specified ranges is often the case. The recommendation to replace the output
tubes after 100 h to check is therefore not entirely without merit. We will happily avoid
discussing the occasionally found (wacky) idea that tubes should be “run in” for about 100
hours before they sound right; rather we will opt to clarify the question which operational
state puts the highest strain on the tubes.
Just switching on an amplifier can be detrimental: subjecting the still cold tube to the full
plate-voltage may cause parts of the cathode-coating to detach. Here, the good old rectifier
tube did have an advantage: only once it had heated up, the full supply voltage was available,
and by that time, the other tubes generally were at operating temperature, too. On the other
hand, just keeping the filaments powered up without any current through the cathode should
be avoided, as well, because it supports the build-up of the impeding intermediate layer
(exceptions are the so-called long-life tubes). Whether an amp should, during breaks in
playing, continue to run with the plate-voltage switched on, or switched off (i.e. on standby),
or should be powered down completely is discussed controversially. Complete shut-down
does reduce the “hours in action”, but it brings numerous strong temperature fluctuations that
also reduce the tube-lifespan – better leave the power on, then. Regarding the use of the stand-
by mode, there are only assumptions: advantages and disadvantages are more or less in
balance. In amplifiers with a high-bias current, the stand-by mode can be purposeful because
in these amps the tubes are under the highest strain without an input signal. An example
would be the VOX AC-30: in idle-mode, the maximum strain on the plate of the EL84 is
usually already exceeded – but it is exactly this amp that does not have a stand-by switch.

What does put the strain onto the power tubes? The heating (or even over-heating) of the
electrodes! Without a technical-education background, one could assume that the amp being
operated at full drive would bring the power tubes to their strain-limit; overdrive would then
cause overload. That is, however, not correct per se: relevant for the power dissipation at the
plate is the product of plate-current and plate voltage. For example: in the idle-state, there are
450 V between plate and cathode♣, and the bias-current is 40 mA. The plate dissipation then
measures 18 W. If the tube can take 30 W, this strain in idle is not critical. As a drive signal is
fed to the amp, both Ua and Ia change. The change, however, in opposite directions: as Ia rises,
Ua falls. At the drive-limit, the plate dissipation would even converge to zero (at least for
idealized conditions): either the tube conducts; in that case there is a plate-current but the
voltage drop across the tube is zero. Or the tube is in blocking state; now there is a high
voltage across the tube but the current is zero. No power tube is that ideal, though: at
maximum plate-current, we find a voltage of about 50 V between plate and cathode, or even
more. Still, even with a plate-current of Ia = 0.3 A, this would imply merely 15 W i.e. totally
uncritical. The danger lurks in the intermediate range at around half of the drive-level range:
225V ⋅ 0.15A = 34 W. With a 30-W-tube, we would be already outside of the specified strain
limit. The latter needs to be seen as an average value, though – the tube is not operated
statically in this state for the duration since the input signal changes all the time. Here, we find
strain calculations that determine the plate dissipation for sinusoidal drive signals. This is not
entirely unreasonable, but not typical, either: the signal delivered by an electric guitar is not
sinusoidal. Alternative calculation methods assume a square drive signal and result in a 23%
higher strain on the plate – however, the guitar signal is not always of a square shape, either.
In any case: even assuming worst-case as continuous operational state, the plate will take on
the thermal strain quite nicely – at least as long as the load impedance fits.
But then there’s the screen grid! Contrary to the situation at the plate, the voltage at the
screen grid decreases insignificantly in the presence of a drive signal, therefore it remains an
ideal landing spot for the electrons when the plate-voltage is small. Consequently, current and
dissipation in the screen grid increase as the plate-voltage drops. Since the maximum
allowable dissipation in the screen grid is smaller than the maximum allowable plate-
dissipation, the screen grid can easily be made to glow. Moreover, it is much more difficult to
observe this compared to a glowing plate because the screen grid is surrounded by the plate.
Here is a numeric example: the 6L6-GC is specified with Pa,max at 30 W and Pg2,max at 5 W. At
full drive level (UB = 400 V, Ug2 = 350 V), the plate experiences a strain of 19 W, and the
screen grid of no less than 15 W! Such an extreme overload must only be present for short
periods unless we want to run the risk of the screen-grid wires melting and the tube dying.
This is why, during the circuit design, the screen-grid dissipation is measured, as well, and
measures to limit it are taken if necessary. The tried-and-tested method here is to introduce a
screen-grid resistor in series with the grid. In fact, this resistor has two functions: to suppress
RF-oscillations, and (given the appropriate resistance) to decrease the voltage at the screen
grid for high currents through the screen grid. Some old Marshall- and Fender-amplifiers
lack any screen-grid resistor, and the lifetime of the tubes can become extremely short – in
particular if the EL34 is deployed. This tube is a true pentode and the current through the
screen grid can easily reach values 2 – 3 times as high compared to the 6L6-GC. If a screen-
grid resistor is present, it often is given a value of 1 kΩ. This resistance is recommended as
adequate to avoid RF-oscillations, but the screen-grid dissipation may not be reduced enough.
You could raise the resistance to 5 kΩ, but that may have effects on the output power and the
sound (the latter of course being a matter of taste). Sound or safety – you choose.
♣
That would be for class-AB operation; for class-A operation, the strain on the tube is highest in idle.

The family of output characteristics gives us a clear overview regarding the relation between
current and voltages (Fig. 10.5.30). Changing the drive levels shifts the operating point along
the load line and changes the power dissipation at the plate. The fact that the load line
traverses the power-hyperbola, and that a peak value of 44 W results, is not critical because
the operating point advances into the range of such high dissipation only for one half-wave. In
the worst case (a square wave-shape), the thermal strain on the tube is 22 W, which is still
clearly below the 30-W-limit.
Fig. 10.5.30: family of output characteristics of the 6L6-GC for a load impedance of 1.1 kΩ; time functions for a sinusoidal drive-signal. In a
push-pull output-stage including an output transformer, the plate-voltage would not be limited to 440 V (as indicated here) but rise to 700 V
(Chapter. 10.5.3).
Fig. 10.5.30 is for a purely ohmic load-impedance of 1.1 kΩ. If this value changes, the
gradient of the load line changes, as well, and with it the strain on the tubes. Increasing the
load-impedance (smaller load-line gradient) reduces both plate-current and dissipation at the
plate. Decreasing the load-impedance increases the strain on the plate: a power stage specified
for 8 Ω should therefore not be operated at 4 Ω for extended periods of time.
We find entirely different functions for the current through the screen grid (Fig. 10.5.31).
Given a constant voltage at the screen grid, the maximum power dissipation at the screen grid
rises to 125 W – if at all, this is only allowable for impulse-operation: according to the
datasheet, 8 W should not be exceeded. Even with 1.5 kΩ connected between voltage source
(again 350 V) and screen grid, the allowable screen-grid-dissipation is, at 20 W, considerably
exceeded. The same holds for the 6L6-GC (Fig. 10.5.32). On the other hand, the question
remains whether this state can actually happen during real operation?
Fig. 10.5.31: Plate- and screen-grid-current dependent on the plate-voltage.

Fig. 10.5.32 depicts the average power dissipation at plate and screen grid, dependent on the
drive levels (square-shaped drive signal). The highest plate dissipation Pa appears at Ug1,max =
-14 V; for Ra = 1100 Ω the average dissipation is Pa = 22 W, which is safely below the strain
limit. A higher load-impedance (1500 Ω) reduces the strain on the plate, and we generally
find that a higher-impedance loads relieve the plate. The maximum power dissipation at the
screen grid Pg2 happens at Ug1,max = 0 i.e. at a fully overdriven power tube. For a load
impedance of 1100 Ω, Pg2 is 8 W, and for a 1500-Ω-load we already see 14.5 W – three times
the allowable power dissipation at the screen grid. For overdrive conditions we therefore find:
higher-impedance loads put more strain on the screen grid.
Fig. 10.5.32: Symmetric square-shaped drive around the operating point of 45 V / 30 mA, power dissipation at plate and screen grid
(averaged over time). No screen-grid resistor
Consequently, a tube amplifier can in no way be seen as totally immune against a load-
mismatch. The 8-Ω-output should indeed be connected to an 8-Ω-loudspeaker! Too small a
load-impedance (e.g. two 8-Ω-speaker-boxes in parallel = 4 Ω) would increase the strain on
the plate – although this would not lead to immediate failure. On one hand, there is some
overhead here, and on the other hand, the higher stain will reduce the supply voltage
(depending on the power supply circuit) such that the strain is not quite as high. Too high a
load-impedance (e.g. a 16–Ω-speaker connected to an 8–Ω-output) will increase the strain on
the screen-grid – in particular if the power stage is operated often under overdrive-conditions.
In this case, there is little reserve, as everybody measuring the power stage (connected to a
loudspeaker) with a sweep may notice right away: every speaker will turn high-impedance at
high frequencies, irrespective of its nominal impedance (Chapter 11). With such a high-
impedance load, even a single measurement can lead to immediate failure of the power tubes.
Looking at the datasheet and considering a load-impedance of double the optimum value, we
find a static power dissipation at the screen grid of 40 W: Pg2 = 0.1 A ⋅ 400 V (at Ug1 = 0 V).
The tube is in active state during only one half-wave such that on average 20 W remain, but
that is still much too high compared to the allowable max. 5 Watt. Also, the speaker
impedance may not merely double – a 10-fold increase is possible, as well.
In view of all this, the impression manifests itself that the classic power-stage circuits were
not developed for Hardrock but for radio programs. And indeed, how could it be otherwise:
70 years ago, the place of action for a 6L6 was a radio receiver in most cases (or maybe an
amplifier in a cinema, at most). Guitar amplifiers were few and far between. Even if such a
tube found its way into such an exotic job site, it still lived a relatively tranquil life – the
rocker “turning everything to 10” was only just about to be born. He (or she) arrived only
later, but then used amplifiers that were – with unbelievable tenacity (or ignorance) – built as
if reproducing moderate dance music without distortion were the only way of life. Marshall
model 1987 amps of certain vintage include two EL34 but no screen-grid resistors. Lucky are
those who can after each gig afford a new set of tubes (at 50 to 100 Euro).

We can insinuate that the developers of the early “original circuits” strove to avoid
overloading the tubes too much at least in idle mode. No guitar player will buy an amp for its
idle-mode qualities, though, and will “turn it up” at some point. Several parameters determine
how the strain on the tubes will then change: the drive levels, the loudspeaker impedance, the
specific tubes used, and the circuit variant. To start with the latter aspect: most amplifier
circuits went through a number of stages of advancement, often driven by the demand for
more power. When the Fender Twin entered the market in 1952, it was specified at 18 W.
Only 3 years later, this power output had grown to 30 W, then 60 W and 80 W, and finally to
100 W (and beyond). The 5C8-Twin feeds 370 V to the plates – in the AC568 this is 470 V. A
VOX AC-30 expects its EL84 to each put up with 14 W in idle, if it is operated with the
rectifier tube customary back in the early days. Exchanging the rectifier tube for silicon
diodes (a design development in later AC-30’s) pushes the power dissipation at the plates to
17 W each. Marshall amps may or may not sport screen-grid resistors (25 Ω, or 470 Ω, or 1
kΩ), and the tube complement could include KT-66, 6L6-GC, EL-34, KT-88 or 6550.
Therefore even if amps look similar, the strain on the tubes may differ significantly.
Fig. 10.5.33: Characteristics of two tubes both sold as KT-66; plate-current (left), screen-grid current (right).
Occasionally, even the tube designations may be just as unspecific: the measurements shown
in Fig. 10.5.33 were taken from two fresh KT-66. They show significant differences both in
the achievable maximum output power of the amplifier, and in the strain on the tubes, despite
the fact that allegedly this is the same type of tube. To be sure, most tubes sold today will be
roughly in the ballpark of the datasheet specifications; however, the amount of deliberately
sold “selected” defectives unfortunately is not petty – to put it mildly.
It requires no emphasizing that the strain on the tubes depends on the speaker impedance, too:
loudspeaker impedances are complex and strongly dependent on frequency (Chapter 11.2),
and therefore the load line deforms into an ellipse in real operation, rendering immaterial all
calculations of power dissipation for nominal load. In addition, the input signals almost never
correspond to the textbook-sinusoidal shape: power stages in guitar amps are often
overdriven. Not always, admittedly – but even with a seemingly “clean” sound, the string
attack can drive the power stage into short-term limiting. More than a few guitarists
appreciate their tube amps especially because of the power-stage limiting which is not easily
imitated via small effects boxes (in contrast to the pre-amplifier distortion). Under overdrive
conditions, a current flows through the control grid polarizing the coupling capacitors such
that the operating point drifts back and forth depending on the drive signal – not an effect that
is typically discussed in circuit-design textbooks. The latter will give you guidance to push the
THD below 1%, and will exemplarily calculate the whole HiFi power stage. Continuous
overdrive is not a topic here, nor was it in 1940 for the amp-forefathers. Only today, in
everyday stage-life, it very much is.

In order not to succumb to the temptation to join in and explain the operating characteristics
of a guitar amp only in the linear range, let’s turn to 10 now – proud ‘n’ loud. The candidate
for our measurements is a Fender Super-Reverb, its two 6L6-GC generating an output power
of approximately 40 W – or more if we go into overdrive. Which we do: Fig. 10.5.34.
Fig. 10.5.34: Average power dissipation at plate and screen grid of a 6L6-GC, Super-Reverb. IBias = 50 mA.
Sinusoidal 1-kHz-tone with a level increasing by 20 dB from 0 – 1 s; switch-off at t = 1 s. Measurements taken with 8 Ω load (purely ohmic)
for plate dissipation, and 16 Ω load (purely ohmic) for screen-grid dissipation.
The drive signal is a sine-tone with a level increasing by 20 dB from the time “0” to the time
“1 s”; at t = 1 s we switch it off. During the last quarter of the measurement, the power stage
is overdriven, which does not harm the plate at all: its strain decreases to about 10 W with
increasing drive level. After (!) switching-off, however, there is a short-term plate dissipation
of 50 W due to the settling of the polarization of the coupling capacitors. The tube does not
die right away because this overload happens only for a short time. If such short-term
overload conditions repeat themselves quickly one after the other, they could, however, pose a
problem, after all. Also, 50 W is not really the end of the line: corresponding measurements
with a real loudspeaker as a load resulted in more than 100 W!
We see a rather different behavior for the power dissipation in the screen grid: it grows with
increasing drive level, and approximately at the point where overdrive occurs it crosses over
beyond the maximum value of 5 W. Therefore: as soon as the power stage is overdriven, the
screen grid enters the danger-zone. If we could address one of the design-forefathers with
this problem, the answer would probably be: “you don’t overdrive the power stage!” Yeah
you do, these days. The argument that the Super-Reverb is an amp for rhythm-guitar that
should be played “clean” could easily be countered in that the power-stage design for this amp
corresponds to the Fender-standard of the 1960’s – the power stage of the Bassman (as just
one example) is in no way more less prone to be overdriven. This was all by-the-book design.
At the time.
Overdrive is the joint cause for putting excessive strain on plate and screen grid; the exact
effective mechanisms are specific to the respective electrode. Normally, the plate-voltage in a
power tube decreases with increasing plate-current; at full drive level (Ug1 = 0), the plate-
voltage will be minimal. At this point, however, the plate becomes rather unattractive as a
landing-site for the electrons (emitted by the cathode). The electrons are much more attracted
to the screen grid that remains at a high potential (high voltage), and they land (at full drive
level) on the thin screen-grid wires. The latter promptly heat up under this bombardment and
start to glow. Even datasheets do not shy away from specifying a power dissipation of 100 W
or more for the screen grid (at Ua = 0) – and at the same time they will give a maximum strain
of 5 W. This is not a contradiction, because for short-term strain (impulses), the dissipation
limit is higher. How high is not specified, unfortunately. At the plate, entirely different
processes are significant: as long as the load-impedance of the power stage is not too low, the
plate does not run into danger even under dramatic overdrive conditions. However, the
coupling capacitors will vary their average DC-voltage during periods of overdrive, and
during the following balancing processes, danger looms, after all.

The source for the balancing processes just mentioned is the phase-inverter (Chapter 10.4).
The two signals generated by the phase-inverter are equal in magnitude and opposite in phase
only for moderate drive levels. Strong drive levels shift the operating points in the phase-
inverter, and the coupling capacitors change their average DC-voltage. As the input signal is
switched off, the coupling-capacitors potentials return to their quiescent state within 2 s♣ – it
is here that the peaks in the power-tube-specific strain occur.
Another problem may present itself at the fringes of the transmission range: at very high and
very low frequencies, the two signals from the phase-inverter do not maintain exact opposite
phases. The output transformer can generate its impedance-transforming magnetic field only
in an efficient manner if the plate-currents support each other. If both power tubes conduct at
the same time, the transformer has the effect of a bifilarly wound coil – with the effect that the
inductance goes to zero, and merely the copper resistance of the primary winding remains as
load impedance for the plate (e.g. 50 Ω): the plate will possibly be overloaded.
In Fig. 10.5.35 we see once more the family of output characteristic of the 6L6-GC, with an
Ia/Ua-characteristic measured for a JJ-6L6-GC. For a load impedance of 1200 Ω, the “knee”
of the curve is almost exactly met; the strain on the screen grid at this point is 350 V ⋅ 46 mA
= 16 W. As we increase the load impedance to 6000 Ω, the strain on the screen grid grows to
54 W. Assuming that this maximum strain occurs only during one half-wave, we may half
that value – but the remaining 27 W still overshoot the allowable maximum value
considerably. The approximate load-impedance for the specified match is 1200 Ω, i.e. 8 Ω for
the Super-Reverb2. However, loudspeaker measurements show that the magnitude of the
speaker-impedance will be larger that than the specified impedance both at high frequencies
and at the speaker resonance. A primary load of 6000 Ω corresponds to a secondary load
impedance of 40 Ω – this can easily be achieved with a loudspeaker. A guitar will not
normally generate continuous tones at 15 kHz, but sustaining notes in the range of the speaker
resonance are possible – and may be dangerous.
Fig. 10.5.35: output characteristic of a 6L6-GC (JJ) for two different loads on the plate (left),
loudspeaker impedances (Jensen, right), plate-current (–––) and screen-grid-current (---) for Ug1 = 0.
In order to remain datasheet-compliant, the screen-grid resistor in Fig. 10.5.35 is assumed to

be 0; in return, the screen-grid-voltage is only set to 350 V. A Fender-typical resistor-value
would be 470 Ω, connected to 400 – 450 V. As a first-order approximation, we find similar
strains on the screen grid; in the detail, there are differences that however cannot be calculated
to the last watt.
♣
In theory, this asymptotic recharging takes an infinite time; 2 s should be seen as a specific guidance value.
2
As already mentioned, this specific Super-Reverb carried a transformer with both 2-Ω and 8-Ω-outputs.

Often, the screen-grid resistors are connected to the first or second filter capacitors in the
power supply. Without any signal at the amplifier input, these electrolytic capacitors charge to
400 – 450 V; with a strong input signal this voltage drops a bit (sags). How much the voltage
drops depends on the load impedance and the internal impedance of the power supply. The
sagging effect will be stronger for power supplies with a rectifier tube, and weaker for ones
with silicon diodes. A power supply with rectifier tube and small caps (10 µF) will go easy on
the screen grid while silicon diodes and 100 µF represent a challenge. Unfortunately, the
datasheets for usual power tubes do not reveal any thermal time-constants of the screen grids,
and therefore any calculation of the impulse-strain will remain speculative. Only with the
triumph (?) of power transistors, impulse-diagrams enter the datasheets. The 2N3055 (aka
BD-130), for example, is specified with 100 W continuous power dissipation at 20°C of the
casing, with 320 W for 1 ms, and even with no less than 900 W for 30 µs. For the 6L6-GC,
we find 5 W as a limit-value for the screen grid power dissipation, without any further details.
In the semiconductor area, there is at least a rough guideline (in case you want to avoid much
calculation) that the lifespan doubles if the operating temperature is decreased by 10°C. Given
tubes, we need to rely on completely flaky speculations. How long did that Mullard survive
100% overload at the screen grid, and how does its modern Chinese remake fare? The remake
that raises the suspicion that it’s mostly the cosmetics that is important (it’s got the brown
base!). Caution, though! With such prejudice, you may well do very wrong by those sino-
factures. Not everything that originates in China is bad – just as is the case for any other
country. As he developed the 5881, was the Tungsol-R&D-guy in the US really interested in
how strongly the screen grid would be overloaded in guitar amplifiers, and was that tube
therefore marketed as the “better 6L6”? As late as 1962, the Tungsol datasheet specified:
“Maximum Grid #2 Dissipation: 3 Watts”. That’s not really a lot, either, isn’t it?
These days, acquiring a hand-wired boutique amp will easily set you back 4000 or 5000 Euro.
That’s without speaker, of course. Maybe the manufacturer boasts using only output
transformers with original insulation-paper (with worse breakdown rating) and slightly rusted
transformer sheets – to get the ‘brown’ sound? Maybe he will put a 1-A-fuse in the mains line
(just as in the original) without realizing that converting from 110V to 220V the value of the
fuse should be halved? It all has to be original- that’s the main thing. Or, the focus is on using
the same circuit that made the Bassman (or the Deluxe, the Twin, the JTM – you name it)
famous. Including all the grid-destroying characteristics of these old amps. Amazing how the
oldest cows are the most holy ones. Maybe it was the CBS-engineers who, by introducing
protective circuits, discredited just these circuits. A guitar amp sounds best if it gobbles up a
set of power tubes each evening – it’s a cast-iron credo.
Rating Systems [Langford-Smith & RCA-Receiving-Tube-Manual]:

The absolute maximum system originated in the early days of valve development and was based on the voltage characteristics of battery
supplies. Battery voltages could fall below their nominal values but seldom appreciably exceeded them, so that valve maximum ratings set on
the basis of specified battery voltages were absolute maximum ratings that should not be exceeded under any condition of operation.
The design center system was adopted in the U.S.A. by the Radio Manufacturers Association in 1939 for the rating of receiving valves and
since then has become the standard system for rating most receiver types of American design. Under the design center system, ratings are
based on the normal voltage variations which are representative of those experienced with […] power lines. Design center ratings should not
be exceeded under normal operation. These ratings allow for normal variations in both tube characteristics and operating conditions.
The design maximum system was adopted for receiving tubes in 1957. Design maximum ratings should not be exceeded under any condition
of operation. These ratings allow for normal variations in tube characteristics, but do not provide for variations in operating conditions.

10.5.10 How does the 6L6 sound?
"6L6 = silky, clear treble combined with a well-defined, deep bass contingent ", states the
advertisement. Or: "EL-34 = delicate treble and well-defined bass and midrange." Or: "The
6L6 have more of a midrange tone". Or: "The KT-90 will give a little more bass and treble, as
compared to a EL-34." On the other hand, we read in circuit-design textbooks that amplifier-
tubes transmit their signal from 0 Hz up into ranges not measured in kHz but MHz. Osram,
for example, recommends the KT-66 for radio-frequency amplifiers "for frequencies up to 30
Mc/s". Much depends on the circuit the tube cooperates with, after all. In a first step, the
parameters of the tube need to be taken as frequency-independent. That does not mean,
however, that the whole power stage has no dependency on frequency. Also, changing tubes
may well result in a different frequency response. That, however, does not allow for the
conclusion that a special tube generally delivers more (or less) treble.
It is possible to investigate the behavior of a tube power separately for linear and non-linear
operation, even though that is not entirely unproblematic for a guitar amplifier deliberately
pushed into overdrive. But then, after all, there are guitar players who seek a sound as
undistorted as possible. Also, if we focus on a system that distorts both linearly and non-
linearly, we run into problems to describe it clearly (since no transfer function can be defined
for such a system, amongst other reasons). In the approximately linear range, several
components determine the (magnitude-) frequency response of the power stage: coupling
capacitors (in combination with their load resistors), output transformer, and loudspeaker. For
a guitar amplifier, we may not establish one frequency response of the power stage and one
frequency response for the loudspeaker, and hope that these two diagrams would describe the
overall system. For a loudspeaker fed from a stiff voltage source, we could define one
frequency response (on axis), and for the speaker fed from a stiff current source, too – but
these would be different frequency responses. For an internal impedance of 10 Ω of the
amplifier, we would obtain yet another frequency response of the loudspeaker. For the power
stage, the situation is similar: for an 8-Ω-load, the frequency response is different compared to
a 16-Ω-load, and considering the loudspeaker loading results in different curves yet again.
One criterion in which power tubes may differ is their internal impedance. In pentodes (that
are operated as such!) it is typically rather high: think 30 kΩ or so – but do keep in mind that
this depends on the operating point i.e. on the bias current (Chapter 10.5.7 and 10.5.8). As the
internal impedance of the tubes changes, so does the internal impedance of the amplifier.
However, negative feedback (NFB) enters the stage at this point. For power stages with
strong NFB (Chapter 10.5.6), a change in the internal impedance of the tubes has little effect
on the internal impedance of the amplifier, but for power stages without NFB, these effects
are considerable. Consequently, we may not conclude that characteristics found in an amp
without NFB are also found in an amp that includes NFB. The higher the impedance of the
amplifier output, the more resonances and treble range are emphasized. This, however, does
not justify the purchase of expensive tubes (even though that may be suggested in ads):
inexpensive components allow for varying the frequency response of the power stage within a
broad range, as long as the linear characteristics are the issue. It may well be that a
replacement tube has less gain that its predecessor – that can easily be compensated for by
turning up the volume. If the power stage includes NFB, changes in the loop gain could
influence the frequency response (Chapter 10.5.6), but it is easy to get a handle on this, as
well – and with simple means. In linear operation, any frequency response can be achieved
with any power tube; that is standard engineering-knowledge.

Things get more difficult in the non-linear range. Both phase-inverter and power stage may be
overdriven. If high-gain power tubes are used, we can assume that these will distort first but if
they have only a small gain, the distortion of the phase-inverter comes to the fore. Talking
about power-amp distortion, we must therefore also keep the phase-inverter in mind. Your
typical cathodyne circuit (Chapter 10.4.2) can generate up to about 40 V of voltage-
amplitude, which is not enough if the grid-cathode-voltage of the power tubes is set to - 50 V.
The paraphase- and long-tail-circuits, however, can easily drive the power tubes to their limit
(and beyond) even for -60 V. It is necessary to always consider the power stage as a whole.
One possibility to describe the non-linear behavior of tubes is offered via the family of output
characteristics, an alternative to this would be the transfer characteristic Ia(Ug). All power
tubes have frequency-independent amplification parameters throughout the audio range –
their maximum currents are, however, dependent on the load. Since the load impedance (i.e.
the loudspeaker) is frequency-dependent, there is also a tube-specific dependency of the
maximum obtainable power output. Fig. 10.5.36 shows the output characteristics of two
power tubes. For a load impedance of 2200 Ω, tube A will give more maximum power, while
tube B will have a higher output power at 550 Ω. At 1100 Ω, both offer the same maximum
power. In this example, the maximum power is impedance-dependent, and since the
impedance is frequency-dependent, so is the maximum power. The right-hand picture
indicates that for higher load impedances a current-saturation happens already at Ug1 = -8 V –
the transfer characteristic turns into the horizontal.
Fig. 10.5.36: Family of output characteristics: plate-current vs. plate-voltage.
Already these initial considerations show that exchanging the power tubes can change both
the linear and the non-linear behavior of the power stage. Still, we may not deduce that a
specific type of tube (e.g. the EL-34) has a special frequency response. Again: all tube
parameters are frequency-independent throughout the audio range. However, cooperating with
a special circuit-environment, every tube can and will result in a system that in its entirety is
frequency-dependent. The number of circuit variants for power stages in guitar amplifier is
not infinite, and therefore findings from the investigation of one amplifier may be applied to
some other amplifiers. For example, if a special 6L6-GC sounds trebly in listening
experiments with a 4-Ω-Tremolux, it is likely to sound that way in a 4-Ω-Bandmaster because
the same output transformer is used in both amps. Even here, though, an imponderability
remains in that the impedances of the two loudspeakers may be different, and another one in
the fact that even 6L6-GC’s sourced from the same manufacturer may be different. Just
crowned the test-winner in a Fender-amp, one and the same 6L6-GC may disappoint
completely when plugged into a Marshall. Or it may be fine – that depends on the personal
taste, the musical style, the specific circuitry, the individual loudspeaker and the individual
tube. Blanket-judgments such as “the KT-66 is a HiFi-tube” are non-sense, if they seek to
refer to resistance against distortion. Because: all tubes were originally developed for HiFi,
weren’t they?

Fig. 10.5.37 shows the transfer characteristics of typical power tubes. These measurements
can (and are supposed to) merely indicate the behavior of such curves in principle - no
recommendation regarding which tube to purchase may be derived. For one, because e.g. KT-
66’s from two manufacturers can differ drastically, and second because even KT-66’s from
one and the same manufacturer can vary in their parameters. In order to obtain a reasonably
reliable statistic, numerous tubes would have to be acquired – at more than 50 Euro per piece,
this is not a really desirable undertaking.
Fig. 10.5.37: Difference in the transfer behavior (measurements), ohmic load 8 Ω (–––), 16 Ω (----). M50A.
The measurements were done using a TAD M50A output transformer and an ohmic load-
impedance. The differences in the gain (large with the EL34, smaller with the 6L6-GC and the
KT-66) are obvious, which is not surprising: the datasheets show the same under the entry
“transconductance”. We find large maximum currents for the EL34 and the 6550, allowing for
inferences regarding loudness, and we see similar curves with the 6L6-GC for both 8Ω and
16Ω, but larger differences between the two load conditions in the 6550. Taking the frequency
response of a specific speaker as a basis, we can deduce basic sound-variations from the
difference in the respective two curves for each tube. Both curves similar = load-independent,
stiff current-source; pronounced differences = less power at a higher-impedance load i.e. less
treble. For such statements we certainly need to look into the linear behavior, too; it is only
with such an overall consideration that we arrive at a reliable conclusion.
Can these measurements support the notion that an EL34 will distort rather early while the
6550 remains "solid and clean", as Pittman writes in his Tube Amp Book? Before diving into
the slightly different 16-Ω-curves, lets first linger and check Pittman’s bias-voltages: -50 V
for the EL34, and -68 V for the 6550. That does, however, not match Fig. 10.5.37, at all!
What could A.P.’s approach have been here? Probably, his reference amplifier works with a
higher supply voltage than the one used for the above measurements. An increased voltage at
the screen grid could explain a more negative control-grid-voltage, however: how generally
valid are Pittman’s statements then? His subjective evaluation of the sound shall not be put
into question, but his GT-Electronics-Dual-75-Amp is not really that ubiquitous, and at -50 V
(EL34) it is not set typically, either. As Pittman notes a few pages later: in a 50-W-Marshall
there’d be -43 … -40 V.

Also, another striking aspect in the Tube Amp Book: two tubes (6L6-OS vs. 6L6-B) differ
slightly in their distortion: "The 6L6-OS clips a little sooner". Every single tube is, however,
also selected according to its distortion characteristic and designated with numbers: 1 – 3 =
early distortion, 4 – 7 = normal distortion, 8 – 10 = late distortion. This is because of
unavoidable scatter in manufacturing even tubes of one and the same type will have different
data. Now, does a 6L6-OS with a rating of 10 still distort earlier than a 6L6-B with a rating of
1?? As commendable as Aspen Pittmans’s approach towards quantification is, one has to get
lost in this vast desert: the circuits are different, and so are the output transformers, the
loudspeakers, the tubes – and in any case on top of that the subjective expectations, as well. In
a Marshall, Pittman opines, the 6550 sounds "loud and crunchy". If you don’t favor that, you
change the circuit to install EL34’s, because that tube sounds "smoother, with warmer
distortion". Turning over a page or two, we also read that the EL-34 may sound "gritty and a
little squashy", and the 6550 may sound "fat and clean". Or it may sound "extremely harsh",
as we read in publications from German authors. We however also find the latter attesting the
EL34 a “warm sound” which doesn’t really fit “gritty”. On the other hand, German
advertisement has the EL34 giving a “dynamic attack”, while the US-colleagues arrive at the
verdict: "The EL-34-setup seemed to lack dynamics". For the 6L6-GC, evaluations stretch
from “a fat, more mid-rangy singing distortion sound” to “more unstable and mushy”.
In fact, it is not only purposeful but imperative to judge the tone of a sound source according
to auditory criteria. Measurements are (hopefully) precise and objective, but they are not
necessarily directly linked to our subjectively perception. That’s why we carry out auditory
experiments. If both too much and too little dynamics are attested to the EL34, several reasons
are conceivable. Different amps may have been used, or different music played, or just
different EL34’s may have found their way into the experiment. Indeed, it seems that
everything that sports a glass cylinder of 8 – 9 cm length and 33 mm diameter may call itself
EL34. Tube retailers do not angrily send back to the manufacturer all rejects that fail the
specifications in the Philips-/Siemens-/Mullard-datasheets, but sell this junk – with a markup
– as "specially selected" merchandise. Which isn’t totally inaccurate, either, somehow. Why
didn’t others think of that? “For a premium of an additional 500 Euro you can get a selected
TV-set the right-hand screen-half of which remains dark.” Wouldn’t that be a cool idea? It is,
for amplifier tubes. It is even legal, because today your EL34 is not just designated as such
but it’s now called EL-34-SVT, EL-34-Cz, EL-34B-STR, EL-34C, or EL-34R, etc. – and any
notice of defects can be averted. That does not mean that none of these tubes meet the
specifications; some even exceed it – but some will remain 20% below the given current
specification. Others may reach the specified current but fail regarding the transconductance.
Apparently, the scatter is big enough for Groove-Tubes to designate one of ten (!) subgroups
to each tube. These 10 subgroups will have to be significantly different, too – otherwise e.g.
three groups would have been sufficient. Now lets consider, on top of this, that the power
tubes are fed by driver-tubes the data of which are also subject to a noticeable spread.
Furthermore, the power is supplied from circuits including rectifier tubes that may by called
(despite individual “selection”) rejects (see Chapter 10.7.4). In view of all this the question
“what does the 6L6 sound like” can only be answered with a sobering “beats me – no clue –
not a hunch”. Sorry, folks, thou ask'st the wrong man.
"...I have to point out that my experiments trying to map the sonic differences between various tube-types to
sound-files did not meet satisfactory results. The recording/reproduction-process minimizes the differences to a
minimum such that almost nothing remains of the described differences. We can hardly conclude anything
comparable to what is experienced as a difference when playing.” (Gitarre&Bass 6/09).

10.5.11 Match Point
Amplifier tubes run through a multi-stage manufacturing process that does not tolerate any
major errors. If a process step does not go according to plan, the tube parameters deviate from
the values given in the datasheets. The manufacturers (and some retailers, too) therefore test
all manufactured tubes, and weeds out the sub-par specimen. The remaining “good ones” are
ennobled with the attribute “selected” and sold to the consumer. However, rumor has it that
there were some singular cases in which deficient tubes have found their way to the musician.
For this reason, the consumers of the “hot goods” from time to time stage a match in which
the “matched tubes” have to compete against each other: comparison tests. These typically are
a merry ado, extensively covered in the trade magazines. Translator’s note: From the point of
view of the scientist and engineer, these reports often have a rather special “quality”
bordering on the dysfunctional. That, however, does not seem to bother the testers nor many
of the readers even if the process is repeated in the exact same way – in fact the contrary
appears to be the case. Is this testimony to the “magic of the tube”?
10.5.11.1 Selecting and Matching
Translator’s note: I choose not to translate the 1st paragraph here because it is a send-off
targeting the often excessive use of English terms in German music trade-magazines.
Corresponding German terms would be available, so often English is brought in just to sound
cool, or to hide a lack of proper understanding of the subject matter behind impressive
English terminology. A translation of this paragraph would therefore almost by definition not
work in English. Having said that, isn’t a term like “transconductance” just marvelously sexy
and seductive if we write about guitar amps? Anyone talking about guitar technology should
put “transconductance” to good use in any conversation. Seriously though, “transconduc-
tance” is a parameter that tubes are “matched” by – so let’s get back to our book …:
Both chemistry and mechanics are involved in the manufacture of tubes, and in both areas,
technical tolerances exist. The cathode coating, the metals of the electrodes, the wound grid-
wires, the getter, the insulators, the vacuum – varying parameters wherever we look, and
therefore all tubes differ in their operational behavior. The really bad ones get to be thrown
away, but the parameters of the useable tubes are still subject to scatter. Consequently, they
are individually measured (‘selected’), and for use in power stages they are paired up
(‘matched’). It is customary to operate the power tubes in the typical operating point ('at idle')
and to specify the plate-current (PC) flowing at a manufacturer-specific supply voltage (e.g.
PC = 41 mA). Usually, the supply voltage is not indicated but this is not that necessary if it is
typical for the amp (and consistent). It is, however, not sufficient that two tube characteristics
coincide in a single point since the tube is subject to a drive signal, and both voltage and
current will change accordingly. It is therefore purposeful to check also the dynamic behavior
– on top of the static behavior. Enter the transconductance. Barkhausen [Lit.] put his tube-
formula together using it: durchgriff (see Chapter 10.5.1) x internal impedance x
transconductance = 1. The transconductance indicates how strongly the plate-current
changes with variations of (only) the grid-voltage. Since the Ia(Ug)-correspondence is non-
linear, the transconductance can only be determined (as differential quotient) for small drive
levels: S = dIa / dUg, for constant Ua. The information of e.g. S = 5 mA/V consequently
expresses that plate-current changes by 5 mA if the grid-voltage is changed by 1 V. In this
scenario, the plate-voltage must not change i.e. the load impedance must be zero – therefore
the more extensive, alternate term would be short-circuit-transconductance.

Two straight lines are identical if they share one point and have the same slope. If
characteristics of tubes were straight lines it would be sufficient to measure one point (PC =
plate current) and the slope (S = transconductance). However, tube characteristics are not
straight lines and therefore two tubes that have been ‘matched’ via one PC- and one S-value
may very well differ. Fig. 10.5.38 shows corresponding measurement results. The 35-mA-
operating-points of the curves shown in the left-hand section correspond, but the plate-
currents of these two tubes differ significantly for grid-voltages converging towards zero. The
two tubes documented in the right-hand section show – at the 35-mA-operating point –
approximately the same transconductance but their bias-voltage differs. The EL84s on the left
are specified with a transconductance of 9 mA/V, the EL84’s on the right with 10 mA/V – not
a big difference. Given such similar ‘matching specs’, we would not expect curves differing
as strongly.
Fig. 10.5.38: EL84-tubes with ‘matched’ transfer characteristics. UB = 350 V, Ug2 = 300 V, Ra = 2 kΩ, Rg2 = 0.
Or maybe we should. Maybe this is why the H&K sales department states: “for our H&K-
amps, we had purchased selected, i.e. matched tubes, but we still experienced a high rejection
rate because many tubes did not meet our requirements (Gitarre & Bass 4/09).” Please note:
if you match tubes in only one point of the characteristic, they are not necessarily a match at
other points. In practical operation a tube does receive a drive signal, and here not only
operating point and transconductance play a role, but among other things also the behavior in
the extreme ranges: how well does the tube insulate in reverse operation, how much current
will it draw when fully driven, how big is residual voltage caught in the tube. All this should
also be tested, shouldn’t it? No, not as a rule it isn’t – because often there is not even any
insight that such measurements would be required. Frequently, the equipment is lacking, too –
available is merely your no-name tube-tester indicating “bias” and “transconductance”, and
that’s it: done! To compensate, the plate-current is determined to the tenth of a milliampere,
and consequently the PC-values of the ‘matched’ tubes correspond to the tenth, too. You want
to avoid the risk that a musician complains because 36,6/36,7 mA is offered as perfectly
matched. Assuming an allowable scatter of the plate-current of ±5 mA, a bin width of 0,1 mA
results in 100 different bins. If the transconductance is to be matched with three digits, as
well, there might be 100 “transconductance bins”. And so the matching person (is he/she a
matchmaker, then?) is confronted with 100 x 100 boxes, and bags ‘em: every pair matched to
a percent. In some cases, this process, tube pairs of astonishing synchronization will go on
sale, as shown in the left-hand section of Fig. 10.5.39. But then there will also be badly
matched ones, like the example given in the right-hand section of the figure. If anyone
absolutely is of the opinion that power tubes need to be matched: here you are being served –
either way. Incidentally, the two EL34’s are, at 43 Euro (per pair), not low-cost but of
“excellent quality”.

Of course, the transconductance-data printed on the boxes of the measured EL34’s correspond
to a tenth of a percent: 11.28 mA/V, for both tubes. That may even be correct in some point of
the curves – it is, however, unlikely that these tubes were tested at all under full load. And so
this “matching” is of little use.
Fig. 10.5.39: Output characteristics of ‘matched’ tubes. Left: 2x EH-6550, right: 2x EL-34B-STR.
Shelling out e.g. 35 Euro rather than 13 Euro for a pair of EL84’s because they are ‘selected’
and ‘matched’, it seems only fair to expect not just well matching characteristics but also a
correspondence to the curves published in the test certificate. In Fig. 10.5.40, we see a
comparison between a reference (Philips) and two newly-developed EL-84-STR. According
to promo, the latter are supposed to introduce a new standard, and guarantee minimum
production scatter. While the idea of a standard may be interpreted this or that way, the fact
that the scatter in not minimal in the new tubes is clear enough to be recognized by even the
most cloth-eared head-banger. It seems hardly possible for any retailer to more efficiently
shoot down his own highly-praised “premium dynamic matching”.
Fig. 10.5.40: Output characteristics (Ug1 = 0) of two EL-84, 'selected' and 'matched'. Right: 2x JJ-6L6-GC.
(“Datenblatt” = datasheet)
The right hand part of the figure shows that it is also possible that datasheet-specs are
exceeded: both measured JJ-6L6-GC perform better that they need to. That is gratifying, but
the screen grids are under more strain, as well, and the gratification about the high
performance will quickly vanish if the tubes fail after a short time. That’s the gratification of
the musician that’s gone, cause the dealer will continue to be gratified as the next
duet/quartet/sextet/octet/duodectet needs to be acquired. A hint on the side: why not buy,
instead of 12 ‘matched’ tubes, 35 industrial tubes for the same money and check whether they
won’t do the job just as well, either?

10.5.11.2 Comparison tests
Since the market for tubes is entirely non-transparent, every purchase amounts to a gamble.
Good to have the “tube gurus” who give recommendations, or even organize comparison
tests. Which one is the best tube? In many corresponding test reports, the guitar used is
presented with much enthusiasm (1958 faded vomit green), as is the involved amplifier (1962
brown Deluxe with Marcus Hotsteam’s Mod No. 17), and the speaker (a pair of Pinkywinky-
Tubbys, with more than 150 h of running-in by playing Blackmore-licks) so that everybody
realizes that true experts are on the job here. Popular is also to point to the jurors (even Blind
Fat Broonzy had a listen), and the location of the affair (we all met in Hamburg) – possibly
because of the specific given air pressure. Then there’s some dignified unpacking: 2 pieces
6L6-GA, 2 pieces 6L6-WGB-STR, 2 pieces GE-6L6-WGC-NOS (loaned ‘em from Crack
Snootshack’s pal), and many another elitist precious’. Plug ‘em in, warm ‘em up, listen to
‘em. "Most of us arrived at the opinion that the WGB is a touch louder but doesn’t give the
oomph of the WGC; some liked the GA better, though. Everybody agreed, however, that
somehow the NOS very clearly sounded the tightest." Man, those tube tests – one could get
addicted to them. Really informative, somehow.
Not to be misunderstood: this is (so far) a free country; hey, any minister of finance can tell
tall tales about his state bank – so why shouldn’t aging guitar-slingers just as well publish a
tube test or two? Is it sufficient to use a mere two specimen per tube type? Well, at $ 200 per
pair, we get that. It is an irrefutable axiom, that listening tests are imperative – just as the fact
that never ever will any measurement data be published. As a rule, the tester will have
procured that "Faded Vomit Green" easily worth 6 numbers among friends – but there is no
adequate instrumentation. And even if that were available, the tester could not be bothered to
get an understanding of how inter-modulation distortion and difference-tone distortion is not
the same thing. Rather, an impulsive “forget all that theoretical baloney” will be included in
the test-report – and that will not be entirely off the mark, either. Amplifier tubes are
designed to be listened to, not to be measured. However, it is the measurement that allows for
elegantly objectifying any differences. As a supplement to the listing test, of course. “Of
course NOT”? Well, it’s a free country (see above).
Such listening tests convey the impression that every type of tube has its own sound-shaping
characteristic. Indeed, the sound of an amp can noticeably change as the power tubes are
swapped – and so each tube must have its special frequency response, musn’t it? It will boost
of cut the treble, won’t it, or it will amplify the bass with particular force. An analogous
conclusion would be: as we feed more air to the Bunsen burner, the flame will become hotter
– therefore air is combustible. Well, it ain’t – and in just the same way, all tube parameters are
frequency-independent throughout the audio range. We should not give highest priority to
thermal infrasound effects, nor to MHz-effects. As every electronics-undergraduate learns in
the circuit-design course: changing a frequency-independent gain-factor in a system with
negative feedback may well change the overall transfer function in a frequency-dependent
manner. The same can happen if the internal impedance changes by a frequency-independent
factor. The frequency response does depend on the tube, but interactively, specific to the
amplifier and speaker. Comparison test for tubes are always flawed in that one never knows
how far the results are at all applicable to another amplifier. Moreover, one needs to be afraid
of a complete and utter disregard of the basic rules of psychometric test-methodology: the test
persons are plain prejudiced because no blind testing is done. Or, the test signals are changed
in addition to the tube-changes: someone/anyone plays something/anything on the guitar.
Reproducibility? No such luck … dream on!

Everyone in the process of purchasing tubes and banking on general verdicts such as “for
distortion sounds, the Sovtek 5881 WXT was the absolute winner in the test” should know that
the parameters of tubes are subject to scatter (due to the manufacturing process). The left-
hand section of Fig. 10.5.41 shows the output characteristics of 6 6L6GC-tubes made by
Ultron. That’s only 6 tubes and therefore not enough to indicate the maximum scatter that can
be expected. This sample is, however, sufficient to recognize that these 6 arbitrarily chosen
tubes all manufactured by Ultron vary about as much as the TAD 6L6-WGC differs from the
Tungsol 5881 (i.e. two different suppliers!) in the right hand section of the figure.
Fig. 10.5.41: Output characteristics: 6x Ultron 6L6-GC (left), Tungsol 5881 and TAD 6L6-WGC (right).
Unfortunately, the “matching” of tubes does not get rid of the problem. A comparison test as
mentioned above elucidates: “moreover, the tube pairs need to be optimally matched. In other
words, the idle-current needs to be the same in both tubes as exactly as possible. The pair
supplied by Tube Amp Doctor was perfectly in tune. We measured a deviation of only 2 mA. A
mismatch of more than 5 mA would cause crossover distortion, and weak and inharmonic
sound.” Figs. 10.5.38 & 39 have already shown that equality in idle-current (bias-current)
does in no way guarantee equality in the characteristics. Also, the term “transconductance”
does not even show up once in the test report, just as power measurements, frequency
responses or characteristic curves are completely foregone. Rather, the insights won are
limited to blanket judgments such as “the KT-66’s are, in principle, HiFi-tubes and were
deployed in 200-W-Marshall tops.” Now, that indeed is a surprise. Not so much because ‘in
principle’ all tubes should be HiFi-tubes, but more so because 4 KT-66 can hardly generate
200 W. Michael Doyle writes in his Book on Marshall that KT-88’s were used in the 200-W-
power-stages – that makes much more sense. And another citation for all students of
psychology who need a quick additional example for their exam: “Steve Ray Vaughan had a
quartet of KT-66 in his famous Dumble Steel String Singer amp. Anyone who knows Stevie’s
album ‘The sky is crying’ already knows fundamentally how these tubes sound.” In principle
like HiFi, of course, don’t they? Once you are aware of that, tube tests become rather
dispensable – at least in principle.
There are some indications of this insight filtering through a bit; in a more recent test
(Gitarre&Bass 3/2009), we read: “Another problem in my testing was the possibility of a
complete reversal of the results, depending on the amplifier”, and “occasionally, only little
remains of the clear differences that are experienced directly in front of the amp.” What does
remain is least one question: is it possible that a Chinese KT-66 can be “through and through
authentically” sounding like the old MOV-originals, although its data (at Ua = 50 V) differ by
a factor of three (!) from those in the old MOV datasheet? No, this is not proof that datasheets
have no connection to the sound: every sound is based on voltages and currents, the
correspondence of which is depicted in characteristic curves. And if that weren’t necessarily
so, we wouldn’t have to so carefully match the plate-current, either, now would we?

10.5.12 Special tube power-stages
In the following, a few selected tube power-stages are presented and discussed with respect to
some parameters. In doing so, we will always consider that the behavior of every tube amp
will depend on its individual components. All measurement curves shown were taken from a
specific amp – even if an amplifier of the same type is built according to the same schematic,
it can still behave differently.
VOX AC-30
The VOX AC-30 and its predecessor AC-15 are the guitar amplifiers often seen as “the”
prototypes for the class-A push-pull power stage. We will not investigate the fact that there
was a series of similar amps (e.g. from Gibson) – but we will look into the issue whether the
AC-30 actually is powered by a class-A push-pull output stage. Technical literature
consistently defines this type of operation via two aspects: the power tubes must not be driven
into reverse operation, and the operating point must be located in the middle of the load line.
What is the situation lined up in the VOX?
Four EL84 are employed in the AC-30, two each in a parallel configuration to double up the
current. Fig. 10.5.42 shows the output characteristic of this power pentode, with the operating
point set to about 310 V / 47 mA – at least for the early variants. After the silicon rectifier had
superseded the rectifier tube, voltages of more than 360 V found their way into the amp, but
lets pick the “original VOX”, the way it was built at the beginning of the 1960’s, as object of
our investigation. Without any drive signal, we find, at the operating point as given above, a
power dissipation of 14 W per tube – mind you, that’s 2 W in excess of what the datasheet
allows. Still, that is just about tolerable (if we agree to a reduced life expectancy of the tube).
However, a symmetric drive-situation (i.e. text-book class-A-operation) is not possible for
this operating point: at a control-grid-voltage of about –10 ±6 V, the power tubes start
limiting to one side of the signal, and therefore the provisional conclusion needs to be: the
VOX AC-30 does not feature a class-A push-pull power stage.
Fig. 10.5.42: Output characteristic of the EL84, ideal load line (4 kΩ) at a supply voltage of 310 V (left).
On the right measurement results for a VOX AC-30 are given (ohmic 8-Ω-load at the 8–Ω–output).
A more exact analysis of the load line confirms this diagnosis (Fig. 10.5.42, right hand
section). For small drive levels the expected load line occurs, with a slope resulting from the
4-kΩ-load-impedance. For increasing drive levels, however, the OP wanders off into the
lower ranges i.e. to smaller current values, and the slope changes from 4 kΩ to 2 kΩ. This
indeed needs to happen, because the setting of the grid-current in the power tubes will
polarize the coupling cap (Chapter 10.4.4), and also because each of the tubes now practically
works in push-pull class-B mode (Chapter 10.5.3 & 10.5.5). If the AC-30 power stage indeed

were a class-A push-pull circuit, current

would have to flow during the full signal
period. That is not the case, however, as
the measurements shown in Fig 5.4.43.
clearly prove. High drive-levels let current
flow in the power tubes only during half
the period, and therefore the AC-30 power
Fig. 10.5.43: Primary currents in the output transformer.
stage is not a class-A power amp.
It is astonishing how tenaciously the fairytale about the allegedly unique push-pull class-A
power stage keeps being repeated. In the book about VOX [Petersen/Denney 1995], this starts
already in the introduction written by Brian May: "The VOX AC-30 ... uses a Class A
configuration." Co-author Denney should know better: he has developed this amplifier, after
all. The same with tube-vendor TAD: “The sound of the class-A operation has been made
legendary by the VOX AC-30! Class-A operation yields several advantages: a thick, ‘three-
dimensional’ sound with pleasant, slight compression, singing sustain and harmonic,
controllable distortion are typical”. Aspen Pittman opines in his collection of schematics:
"Contributing to the amp's smooth tone in both the clean and distorted modes is its very
unusual Class A circuit designed by Dick Denney". Well, this power-stage circuit was not that
unusual: two power pentodes, a common cathode resistor (i.e. automatic generation of bias-
voltage) – we can easily find that years before in Fender amps (e.g. the Deluxe 5B3), and in
Gibson amps (e.g. the GA-40); this was textbook-standard. It was only the value of the
cathode resistor that varied – it set the operating point and made for a “hotter” or “cooler”
operation of the amp (Chapter 10.5.8). And indeed, here the VOX does show a peculiarity: it
operates at the hottest possible tail-end, with a power-dissipation of 14 W (average value) at
the plate (the datasheet gives a maximum value of 12 W). However, a hot operation (or
cathode resistor, respectively) does not automatically imply push-pull class-A.
Why push-pull class-A in the first place? To get the least non-linear distortion! Due to the
superposition of differently-curved tube-characteristics, the non-linear components
compensate each other, the THD decreases. Literature explicitly points out, however, that this
only holds for triodes: "For pentodes the push-pull A-circuit does not yield a significant
improvement relative to the THD of the single tube [Schröder]". As a reminder: the EL84 is a
pentode! The literature has more to in store: "In a correctly balanced push-pull A-amplifier a
capacitor is not required to bridge the cathode resistor. In a AB-amplifier it is, though
[Langford-Smith]". The AC-30 does possess such a capacitor. Lastly: we find the voltage at
the cathode resistor specified in the VOC-schematic; it is 10 V without input signal, and 12.5
V at full drive level. If this were a push-pull-A-circuit, this voltage would remain constant. An
old AC-15-schematic from back in 1955 reveals a common cathode resistor amounting to 130
Ω, and Raa = 8 kΩ for the load impedance at the plate. The Siemens-datasheet (from 1955)
recommends, for a plate-voltage of 300 V, a common cathode resistor of 130 Ω, as well as Raa
= 8 kΩ. Coincidence? Of course not – the circuit designers were wise enough to follow the
recommendations of the tube manufacturers. Siemens, Telefunken, Philips – they all specified
Rk = 130 Ω and Raa = 8 kΩ for the EL84-push-pull power stage. No, not for push-pull class-A
configuration! These recommendations from Siemens, Telefunken and Philips are given for
push-pull class-AB configuration. The AC-30 included four EL84 instead of two, i.e. double
the current, and thus half the value of the cathode resistor. Old plans show an Rk of 80 Ω to
begin with, but it soon was reduced to 47 Ω. Half of 130 Ω would have been 65 Ω – and so
they opted for slightly higher output power (and slightly less tube endurance).

Also, the decision had been taken to do without any negative feedback (NFB). The typical
Fender amplifier from the late 1950’s fed back a part of the output signal to the input of the
phase-inverter and reduced the non-linear distortion of the power stage that way. The AC-30
(from 1958) dispenses with that kind of negative feedback; for this reason, some presume that
the distortion in the AC-30 would be “extremely high”. Well, it’s not – as Fig. 10.5.44 shows.
Granted, 5% THD is not exactly studio-standard, but the AC-30 was never intended to power
studio monitors. At small drive levels, the harmonic distortion is a low as k3 = 0.3%, and with
increasing drive levels, the distortion gradually rises. This is in contrast to power-stages that
feature strong NFB and, correspondingly, a sudden increase of the distortion at the drive-
limit. Apparently, VOX-guitarists prefer the gradually rising distortion.
Fig. 10.5.44: AC-30, 8-Ω-output: distortion below signal (left), power (right); Abscissa referenced to P = 30 W.
It has already been mentioned that tube power stages cannot be described with one single
characteristic curve because the operating points shift due to re-charging effects. Fig. 10.5.45
shows corresponding measurements taken with varying drive levels. With increasing drive,
the transmission characteristic flattens out, with a saddle-point appearing in the origin. For a
load of 16 Ω (right hand part of the figure), the curves generally run steeper (high internal
impedance ≈ current source). The flattening of the curves can be interpreted as a kind of
compressor that reduces the gain of the power stage as the signal level increases. The load-
dependency of the output voltage results in emphasizing the loudspeaker resonance and the
high-frequency signal-components (compare to Chapter 11). In contrast, a power stage with
strong NFB would have a drive-level-independent, sharply bent characteristic similar to the
one discussed in Chapter 10.1.4. The maximum power-yield merits some attention, as well:
with a stiff voltage-source (low internal impedance), the power-limit for a 16-Ω-load would
be half of that for an 8-Ω-load (P = U2 / R); the AC-30, however, reaches more than 80%.
Fig. 10.5.45: Characteristic curves of an AC-30 power stage. Left 8 Ω, right 16 Ω load (at the 8-Ω-output).
These figures are reserved for the print-version of this book.

This special power-stage characteristic is also documented by sweep measurements. For the
latter, the AC-30 was connected to a Marshall 1960-AX – not your typical AC-30 speaker but
able to take significantly more punishment than the fragile and overly expensive blue
Celestions in the combo. Fig. 10.5.46 shows the voltage level measured at the 16–Ω–output,
on the bottom for small drive level and on top for overload.
Fig. 10.5.46: Frequency response of an AC-30 power stage, 16-Ω-output loaded with 1960-AX,
Cut CCW (---) , Cut CW (–––). Right: loudspeaker impedance (in a reflecting room). Compare to Chapter. 11.8.
It has been repeatedly noted that the power tubes deployed in the VOX do not only suffer
when the amp is overdriven, but are under strain already with no input signal present at all. In
Fig. 10.5.47, we see the power dissipation at the plate and at the control grid for drive levels
rising by 30 dB. Without input signal, the power dissipation at the plate is about 14 W in each
EL84. The strain on the plate decreases as the drive level rises, and after switching off the
input signal there is a short peak in the strain. At idle, the strain at the screen grid is just
below the allowable limit; with an input signal present, the limit value is very easily exceeded,
especially with a high-impedance load (for typical loudspeaker impedances see Chapter 11).
Fig. 10.5.47: AC-30: plate dissipation (left), screen-grid dissipation (right). The level of the sine tone at the input linearly rises by 30 dB from
0 … 3 s, at t = 1.9 s nominal power is reached for an 8–Ω–load. At t = 3 s the input signal is switched off, subsequently there are balancing
processes in the capacitors of the power-stage.
The measurements for Fig. 10.5.47 were taken with an AC-30 that had a tube rectifier (GZ-
34) in its power supply. Replacing the GZ-34 by silicon diodes will lead to an increase of the
strain at the plate in idle to about 17 W; the peak after the switching-off reaches 30 W. The
maximum stain on the screen grid exceeds 6 W for an 8-Ω-load, and 10 W for a 16-Ω-load!
Since real loudspeaker impedances (including the so-called 16-Ω-speaker) can become higher
than 16 Ω (Chapter 11.2), even stronger overload needs to be expected.
The power supply merits attention for another reason: the operating voltage for the power
pentodes is directly taken from the cathode of the rectifier tube – the voltage has a
corresponding ripple. This is not a problem at small drive levels, but it is for strong drive
levels since clearly noticeable amplitude modulations result.

Abb. 10.5.48 depicts the time-function of the output voltage, with the 8-Ω-output loaded with
a purely ohmic 8-Ω-load. As long as the output voltage is not limited, any fluctuations in the
supply voltage represent a (superimposed) common-mode interference that is largely
suppressed by the output transformer – the output voltage of the transformer remains un-
modulated (left hand section of the figure). In overdrive-operation, however, an asymmetric
limiting appears: the maximum plate-current depends on the supply voltage while the
minimum plate-current does not (it is practically zero). As a result, an envelope depending on
double the mains-frequency is generated – it may be seen, as a first approximation, as a 100-
Hz-amplitude-modulation (AM). It is not that a 100-Hz-tone is superimposed onto the input
signal; rather, the latter is changed (modulated) in its amplitude. The envelope over time
(which does not actually exist but is an imaginary auxiliary line) is shown dashed in the right-
hand section of the figure.
Fig. 10.5.48: Voltage at the 8-Ω-output for an 8-Ω-load. Drive level: half (left), overdriven (right).
Spectrally seen, this 100-Hz-modulation does not make itself felt as a line at 100 Hz. Rather,
modulation-lines next to the signal-lines result. As a model, the AM can be illustrated as the
multiplication ‘signal x envelope’, corresponding to a convolution in the frequency domain.
Since the envelope is not of an exact sine-shape (Chapter 10.7), we not only get a single pair
of additional lines (±100 Hz), but several pairs. The level-spectra related to Fig. 10.5.48 are
shown in Fig.10.5.49: in the left-hand picture, the modulation lines (lateral lines) have a level
of 45 dB below the carrier – the (3rd-order) distortion (at 1.5 kHz) is 1%. For the overdrive
operation chosen in the right-hand section of the picture, the 3rd-order distortion amounts to
20% with the level-distance between the modulation lines having decreased to 27 dB.
Abb. 10.5.49: Level-spectra related to Fig. 10.5.48.
Not every amplitude modulation that can be measured is necessarily audible – the AM shown
in the right hand section of the figure can, however, be assumed to be noticeable as an
additional roughness. In the area of psychoacoustics [12], the term “roughness” designates
auditory perceptions created by fast signal fluctuations that could be labeled as a kind of
buzzing sound. Measurements of harmonic distortion normally do not encompass modulation
distortions; therefore, dedicated measurements are necessary for this type of distortion.

Marshall JTM-45
James Marshall opened his drum shop in London in 1960, and soon began to sell amplifiers
alongside the drum kits. First, the rather expensive Fender amps and others, but from 1962
also the first Marshall amps that his technician Ken Bran assembled as close copies of the
Fender (tweed) Bassman from 1959. Young guitarist James Marshall Hendrix was a
customer, and both had laid the foundation of their respective careers: one went by the name
Jim Marshall, the other called himself Jimi Hendrix. It was Eric Clapton, however, who first
attracted worldwide attention to the Marshall amp. He recorded, together with John Mayall,
an album the cover of which included a picture showing in the background a Marshall combo:
the legendary JTM-45, with model number 1962.
What’s so special about the JTM-45 power stage – what is it that creates the legendary sound?
A sound – as Gitarre&Bass 7/06 puts it (for once not onomatopoetically but tribo-poetically⊕)
– of a "fat and creamy crunch-tone", but "never a Marshall-typical distortion sound". Excuse
me?!?! A non-Marshall-sounding Marshall? Although we are told that the JTM-45 includes
"all the ingredients responsible for the plexi-sound♣ that achieved legendary status later"?
Presumably, you can see these ingredients – but you can’t actually hear them. 18 months
before (G&B 2/05), the JTM-45 was described as "even hotter and more aggressive", and 6
months before (G&B 2/06) with "clear and fat, with a soft spectrum in the mids".
Clearly, fat may be hot – why not. How this hot-fat sound originates is subject of innumerably
speculations. It starts with Clapton’s Les Paul for which pertinent literature holds in stock a
vintage of '58, '59. or '60. Shouldn’t that be all the same to us? No way – that makes for one
heck of a difference: after all, the frets changed over these production-years (they got wider),
the neck angle also (it increased), and the cross-section of the neck, as well (it got more
narrow). All this should be, of course, "tone-affecting", shouldn’t it? And so we would expect
E.C. to answer the question which model he bought back in the day (in June 1965 according
to G&B 9/08) with an immediate: the ’58, of course, because of the big neck that – as we all
know – improves richness of tone and sustain [G&B Gibson-Special]. However, he does not
answer anything of the like but merely notes: "No idea". No idea? Geez, Eric (as musician,
one is on a first name-basis right away), you should know that: the increased neck-angle of
the ’60-Paula (as this type of guitar is designated in circles of experts) alone would have
ruined sustain, and the thin neck of the ’60 "has no acceptable vibration characteristic
whatsoever [G&B 3/97]". Very strange that Eric does not remember. Thank Eric we do have
recordings surviving from those “Clapton is God”-times – Beano and such – so we should
easily be able to pick out what the deal was. Here’s the latest level of knowledge: "Today the
general opinion is that the guitar concerned was a ’60-model since both Clapton and Peter
Green describe the ‘slinky’ neck [G&B 9/08]." Indeed these are tough times for guitar
experts: on the one hand they continuously are required to explain that the smallest details of
a Les Paul (varnish, frets, neck, pots or tone-caps) have an immense influence on the sound,
but on the other hand there is not anyone in the world who could recognize, on the basis of
these sound-specifics, and from listening to the Bluesbreaker-LP, the version of the guitar.
The stopgap solution then is to reason the guitar-type from on memories regarding the neck-
profile. What an odd, make-believe world of Gods and experts … and stopgaps.
Well then: we don’t know any specifics about the guitar, but the amp is known: a JTM-45,
2x12-Combo Type II, in all likelihood fitted with alnico-speakers (G&B 9/2008).
⊕
Tribology = teachings of friction and lubricants
♣
Hopefully, the plexi will sound like a Marshall – or still not, either??

In all likelihood? Again, nobody knows exactly. Alnicos will yield "particularly sweet and
harmonically rich treble", that is known the latest since G&B 8/05, and so we should be able
to hear from the Beano-LP whether alnicos or ceramics are at work. But again, the LP denies
any analysis, and although ceramic-powered speakers sound "entirely different" compared to
alnicos, nobody can pick out from the record which speakers were recorded. Unfortunately, it
was just in those days that Marshall started to switch over to the ceramic-Celestions so that
both types would be eligible. Again, let’s ask Mr. Clapton – he should … no? Again, no
memory? Well-well, dear Eric: did various substances abound that much already back then?
Indeed?! Then we shall not insist. And on to a look into the expert literature: "Because
Clapton ran the amplifier at full volume, the Alnicos may have been damaged. He may have
replaced them with the higher wattage, ceramic magnet Celestion Greenbacks." This is the
voice of Premier Guitar (February 2008). Clapton replacing his alnicos by ceramics? His
alnicos, those that will produce – according to Premier Guitar – "sweet warm tones and a
smooth midrange"? And that generate, according to G&B, "particularly sweet and
harmonically rich treble" and do “sound harmonic and with a bite”. Entirely differently then,
compared to the subsequent ceramic-Celestions that yield "plenty of midrange crunch" but
"...sounded very different from the Alnico type speakers used in other Marshalls [David
Szabados]." Of course nobody knows whether Eric did actually change the speakers: "He may
have", and he himself can’t remember. That should be not a real problem, though, because
there is that LP, and from it we should be able to pick out the speakers due to: "sounded very
different." It remains difficult, though, because on the one hand the ceramics sound somehow
very different – but not really, on the other hand, because otherwise we would be able to pick
them out. In conclusion: we don’t know anything in detail about the speakers, either.
We do know one thing, though: the output transformer was sourced either from Radio-
Spares, or from Drake. That is certain: either / or. It is also known with certainty that the two
transformers were not equivalent: the Drakes were "rougher and more distortion-happy, more
mid-rangy, darker than the R.S.". Unfortunately we cannot pick out from the recording which
transformer was on duty for Mr. Clapton, and therefore the retrofit-supplier offers replicas of
both transformer, just to be safe. They cost about 250.-- USD (that’s for one, not for both),
thank you very much, plus customs and shipping, and there you are, another step nearer my
God to Thee. You gotta understand why these transformer are so expensive: hand-made!
Encouragingly, the core-sheets are not sawed out with a jigsaw – that would have made them
seem a bit overpriced. Around 300 USD, that’s o.k. – it’s a detailed copy of the Clapton-gear,
after all. In all likelihood – because we still do not know whether Drake or RS, and moreover
the resident expert at G&B offers yet another variant: Mr. Clapton may have operated a pair
of speakers having (in conjunction) an impedance of 8 Ω from the 16-Ω-ouput of the
transformer. That’s a factor of two, so a 100-%-mismatch – or is it 50%? Sorry, it is not easy
to theoretically get a handle on these things, so we better draw some conclusions: Clapton’s
JTM-45-sound is legendary, we all agree on that. If you want to copy that sound, you acquire
either a '58- or a '59- or a '60-Les Paul (allegedly differing audibly in sound), fit your JTM-45
with either a Drake- or an RS-transformer (allegedly differing audibly in sound), and install
two alnico- or two ceramic-Celestions (allegedly differing audibly in sound) – and now you
should firmly, certifiably reside in the midst of Beano-tone. Wow!
Clapton’s Bluesbreaker-sound is great – how it originated is uncertain. Readily overlooked: a

guitar player was involved of for the time extraordinary skill and talent, and of course studio
technology will have had an influence. Clapton’s Marshall-combo has disappeared – its specs
are unknown. What remains is to use schematics and replicas, knowing that a schematic does
not document all details. In the following we will analyze what the hand-drawn sketch in
Doyle’s Marshall-book reveals.

Marshall’s (or rather Bran’s) first amplifier was the JTM-45, with two KT-66’s operating in a
push-pull class-AB configuration in its power stage (with a few exceptions). For this mode of
operations, the GEC-datasheet specifies an output power of 30 W. The number “45” after the
JTM therefore is not an indication of the RMS-power but just promises a seemingly 50-%-
advantage over the AC-30. The JTM-45 power stage includes a negative feedback which is
relatively strong for a tube amp, with several consequences: non-linear distortion is reduced,
loudspeaker resonances have less of an effect, and the amp may oscillate in the RF-range, in
particular with the presence control turned down. Feedback functions as negative feedback
(NFB) if the signal led back to the input is added to the control signal in opposite phase. In
the high-frequency ranges, however, phase-shifts may occur (e.g. in the output transformer),
and the negative feedback can turn into positive feedback: the amp will oscillate. These
oscillations may only happen in a certain range of the drive-signal range where the specific
gain and phase-shifts (both being drive-level-dependent) make for a loop gain of larger than 1.
It is necessary to avoid such oscillations even if they are located in an inaudible frequency
range to begin with: first, because they result in the operation of an illegal RF-transmitter, and
second, because they put unnecessary strain on the power stage.
Fig. 10.5.50: 1-kHz-tone with superimposed RF-oscillation.
Fig. 10.5.50 depicts, in principle, the shape of an “RF-infested” audio signal. The RF (often
around 150 kHz) is not always recognizable as a clean oscillation – it may result merely in a
widening or a smearing of the curve of the audio signal. Small capacitors may be found in the
circuit as a “brute-force bug-fix”, soldered-in at “appropriate locations” to squash the malady.
Much better would be a textbook RC-compensation reducing the loop-gain at high
frequencies without adding significant phase-shifts (a LP with limited, defined attenuation at
high frequencies). Sure, this is not a trivial topic because with every tube-replacement, the
condition for the oscillation is newly negotiated – on top of also being dependent on the
loudspeaker. Those who want to address this issue in a somewhat less sophisticated manner
find a cooperative partner in the form of the Presence control. Just turn it to the right (up,
CW) and the annoying RF is gone. It may be surprising that increasing the gain at high
frequencies will choke the oscillation – however this reduces the loop gain that determines the
tendency to self-oscillation.
Besides the signal-feedback via the NFB-network designed into the amp, another factor may
support RF-oscillations: capacitive coupling of non-shielded components. In fact none of the
components found internally in a Marshall are shielded, which is why even the position of
individual wires can co-determine the tendency to oscillate. Incidentally, this is another detail
that cannot be found in the schematic.

With the JTM-45 power stage operating in class-AB mode, there is, besides the choice of
tubes, another degrees of freedom: the bias-current (or the offset-voltage at the grid). It would
take us too far afield to show all significant characteristics for all appropriate tubes at several
bias-current-settings, and therefore just a few examples shall do. Fig. 10.5.51 shows
measurements of the harmonic distortion (ak3), without NFB in the power stage (left) and with
NFB as found in the original circuit. In his comments regarding the Marshall circuit, Ken
Bran does not make any secret of the fact that the Fender 5F6A-Bassman was used as a
model. Therefore, it is not surprising that in both amplifiers, a 27-kΩ-resistor feeds back the
signal to a 5-kΩ-presence-control. However, in the Bassman the 2-Ω-winding of the output
transformer is the source, while in the JTM-45, the 16-Ω-winding is tapped for this. The
negative feedback in the Marshall therefore is three times as efficient (impedances are
transformed with the square of ratio of the windings). Whether this was by chance, or due to
ignorance, or intentional … who would know 50 years later? In any case, for the successor-
models of the JTM-45, the degree of NFB was reduced again – for whatever reason.
Fig. 10.5.51: JTM-45, harmonic distortion without (left) and with (right) negative feedback in the power stage. An 8-Ω-resistor was
connected to the 8-Ω-output for the measurements, Raa = 8 kΩ, f = 500 Hz.
In Abb. 10.5.51, the abscissa is set such that at 0 dB and for the specified loading, the signal
is just starting undergo limiting. A THD < 1% (i.e. with the generated harmonics 40 dB below
the signal) are surely irrelevant for the auditory perception – presumably, 30 dB difference
(i.e. 3% THD) would still be inaudible in a guitar amp. There is no binding limit value,
though, because too many parameters decide about the audibility of nonlinear distortions. At
first glance, the JTM-45 power stage distorts similarly to a transistor power stage – due to the
strong NFB. It remains practically distortion-free⊗ for the non-limited signal, and shows
textbook increase of harmonic distortion above the drive-limit. This sentence should in fact be
carved in stone: “Marshall’s power amp distorts like a transistor power stage” – considering
that all those amplifier gurus keep praising the specially-bred Marshall distortion! But then,
where should we find something special when we have a copy of an American amp the circuit
of which was taken from tube-manuals? The JTM-45 power stage includes a textbook
differential amplifier as phase-inverter, two beam-tetrodes with a textbook drive-scenario, and
an output transformer as it was offered to a clientele that we would call “hobbyists”.♣ It must
not surprise us that secret forces are entrusted to these “Radiospares-Deluxe-Transformer” by
its fan base – it is, after all, in the sacred company of Ken Bran’s special solder the atoms of
which always automatically redirect themselves towards Hanwell. Caution, though, dear
buddies: after lugging the amp around you gotta wait for 4 minutes – as we learn in the
chemistry course, tin and lead have 4 valence-electrons … they are the so-called inert
(passive) heavy metals, and the redirecting of the atoms will take a little while.
⊗
60 dB level difference between the generated distortion and the signal corresponds to a THD 0,1%
♣
In fact, many guitar amp designers had/have a background in ham-radio.

Are you startled yet? Relax, there are, after all, differences to a transistor amp – we must not
conclude a general equivalence from the similarity of two distortion-curves. The combination
of power tubes and output transformer results in a special power- and distortion-characteristic
that in this manner cannot be found in transistor amps. In the JTM-45, the two KT-66 operate
towards an output transformer with a rather high primary impedance. Presumably, it first was
the RS-Deluxe transformer – nobody can remember which transformer resided in Clapton’s
“Bluesbreaker”-amp. A: it’s been a long time, and B: RS was not a manufacturer but a
retailer, and therefore several manufacturers are possible. An American retrofitter (who is not
adverse to selling his replacement-transformers) surmises that, in the original JTM-45, an RS-
Transformer with Raa = 6.6 kΩ was at work. The magazine Gitarre&Bass supposes an RS-
Transformer with 8.0 kΩ included in the amp (7/2006), but also considers an 8-kΩ-Drake to
be a possible candidate (9/2008). Why would there be such high impedances? The RS-
transformer used to begin with was an all-round device intended for applications as universal
as possible. Consequently it offered four different primary connections: for KT-66 and EL34
with additional ultra-linear connections Raa = 6.6 kΩ; for 6L6, 6V6 and EL-84 Raa = 8.0 kΩ,
or Raa = 9.0 kΩ. The Marshall JTM-45 did not have the ultra-linear configuration, but the KT-
66 with Raa = 6.6 kΩ or 8.0 kΩ is today seen as historically correct. By the way, what does the
KT-66 datasheet specify? We find Raa = 7 kΩ (ultra-linear), or 8 kΩ for the regular class-AB
power stage; for both versions with a cathode-resistor, though. The JTM-45 did not include
such a resistor! For this bias-variant, the KT-66 datasheet specifies 5 kΩ but the supply
voltages do not entirely match. Conclusion: neither the output-transformer manufacturer nor
the tube manufacturer supplied any exactly matching guidelines to the Marshall developers.
Anything else is speculation.
Measurements of the family of output characteristics show that Raa = 8 kΩ is not really
conducive for an instrument amplifier (Fig. 10.5.52). With a load of 8 Ω connected to the 8-
Ω-output, the load line meets the output characteristic of the KT-66 at rather too low a point.
In our example, the KT-66 has a scarily high residual voltage but that is another matter. With
half the load impedance (right-hand section of the figure), we would close in much better on
the ideal condition – and so the conclusion is: for the KT-66 in the JTM-45, Raa = 4 kΩ
would be optimal. That is, at least if a high-power yield is requested. For minimal harmonic
distortion, higher primary impedances could be considered, too … but in a Marshall? A 4-Ω-
load at the 8-Ω-output would be approximately equivalent to an 8-Ω-load at the 16-Ω-output,
a variant also thought possible in Clapton’s amp by G&B (09/2008).
Fig. 10.5.52: Load characteristics, Raa = 8.0 kΩ, 8-Ω-output at an ohmic load of 8 Ω (left), and at 4 Ω (right).
In its measured data, the TungSol-KT-66 corresponds approximated to the GEC-datasheet,

while the TAD-KT-66 fails to deliver the required power due to its excessive residual voltage.
On the other hand, the latter distorts somewhat less as already shown in Fig. 10.5.52. It is,
however, not possible to say how long these evaluations hold: such data change too often.

Fig. 10.5.53 shows the output power of the JTM-45, dependent on the level of the input
signal. Fitted with the TAD, it struggles to climb over the 30-W-mark even when overdriven
(and loaded with the nominal impedance), while with the TungSol-tubes, it is o.k. to call it a
“30-W-amp”. Only when overdriven, and with a mismatched load, it gets close to 45 W. Its
daddy, the Fender Bassman, could to offer more (Fig. 10.5.62), until Marshall later
outperforms it again using the EL34.
Fig. 10.5.53: JTM-45, output power at the 8Ω-output, loaded with 16 Ω, 8 Ω and 4 Ω, purely ohmic.
In Fig. 10.5.54 we see the frequency response of the power-stage loaded with a real
loudspeaker. With the presence control turned down, the characteristic is almost frequency-
independent, despite the frequency-dependent load. As the power stage is overdriven (30-dB-
curves) the presence control looses its effect. While the JTM-45 has strong negative feedback
(NFB), this apparently was not seen as “the” secret of the Marshall-sound – otherwise it
would have been retained in later models. But just that does not happen: rather, the NFB-tap
drifts from the 16-Ω-winding to the 8-Ω-winding, and later even on to the 4-Ω-winding; at the
same time Marshall increases the feedback resistor from 27 kΩ to 47 kΩ and later even to 100
kΩ. Both these changes reduce the NFB – that, however, affects the successors fitted with
EL34’s.
Fig. 10.5.54: Frequency response of a JTM-45 power stage, 16-Ω-output loaded with a 1960-AX speaker (left). Magnitude of the loudspeaker
impedance: 1960-AX measured in a room with reflecting surfaces (right). )
We had seen in Fig. 10.5.52 that an 8-kΩ-transfromer does not really challenge the power
tubes much. For the power-curves shown in the following, we therefore used a 4-kΩ-
transformer. Connecting the latter, at its 8-Ω-output, to a 16-Ω-load, we arrive
approximately at the original conditions (Raa = 8 kΩ). With an 8-Ω-load, the load line just
about meets the “knee” of the output characteristic of the tubes (Raa = 4 kΩ). Chapter 10.5.9
already illustrated the effects of such load changes: the smaller the load impedance, the higher
the strain on the plate; the larger the load-impedance, the larger the strain on the screen grid.
Valid for the JTM-45: Raa = 8 kΩ is the presumed original value; Raa = 4 kΩ would be
optimized in terms of the power yield.

The strain on the power tubes is shown in Fig. 10.5.55 for an ohmic 8-Ω-load at the 8-Ω-
output. As the drive level mounts, the strain on both plates first decreases. Then, however, the
strain on one of the two power tubes rises again. At the moment the input signal is switched
off, we see a high peak in the strain resulting from charge-balancing processes in the coupling
capacitors. Since this peak only has a short duration, it is not particularly dangerous to the
tubes. Conversely, the screen grid is in more danger: as soon as the power stage is overdriven,
the power dissipation in the screen grids mounts: ongoing overdrive does overload the tube
for the duration, and its lifetime is shortened.
Fig. 10.5.55: JTM-45: power dissipation at the plate (left), and at the screen grid (right). From 0 to 3 s, the level of the input sine-tone rises
linearly by 30 dB; at t = 1.5 s a power of P = 30W at an 8-Ω-load is reached.
At t = 3 s the drive signal is switched off; balancing processes in the capacitors of the power stage follow.
The power dissipation of one of the power tubes (TAD KT-66) is shown in black, the one of the other in blue.
The transmission characteristic from the input of the differential amplifier to the power output
is shown in Fig. 10.5.56. As soon as the power amp is overdriven, the curve looses its point-
symmetric shape, and the duty cycles change. The reasons for this are potential shifts in the
differential amplifier (phase inverter) and the grid-current flowing in the power tubes. Until
just short of the drive-limit, the output signal is proportional to the input signal as can be seen
in the left hand picture. As overdrive occurs, the output voltage experiences limiting but also
becomes increasingly asymmetric, and consequently the characteristic curve shifts (the
average value needs to remain zero). Since the limited signal does now include several rather
than a single frequency, phase-shifts occurring in the output transformer (acting as a high-
pass) start to take an effect. The transmission characteristic is not memory-free anymore but
decomposes into a rising and a falling branch. To retain sufficient clarity, Fig. 10.5.56 does
not show the corresponding hysteresis-loops but average values. If a loudspeaker were to be
connected rather than the ohmic load resistor, the complex impedance would result in even
more complicated curves.
Fig. 10.5.56: Idealized transmission characteristic (left); time-function of output for ohmic nominal load (right).

The output transformer (OT) does influence both the frequency response and the non-linear
distortion of the power stage – but there is no mystery about that. In order to conduct
measurements as well as listening tests we put together a test-circuit that allows selecting
between 10 different output transformers via a switch. Among others, we had in the running: a
Marshall-JTM-45-transformer, a TAD MJTM45A, a Hammond 1750Q, an IGOT-JTM45 plus
a few 4-kΩ-transformers. In the frequency range relevant for the guitar, all transformers
showed practically the same frequency response. Regarding maximum power and harmonic
distortion, there is no more than a just-about-noticeable difference between a 4-kΩ-OT und an
8-kΩ-OT. Consequently, these results cannot be the basis for advising a swapping of
transformers. When comparing two OT’s, the most important parameter is the transformation
ratio. If two top-quality transformers produce audible differences in sound, this is most likely
due to a different transformation ratio. However, turns-ratios can be set and checked very
easily and precisely, and therefore any exorbitant pricing of a transformer is not justifiable
merely on the basis of a special transformation ratio.
This is a good point to take a short side-trip into advertising psychology: as an exceptionally
gifted transformer winder, what can you do to increase your turnover? You could write: “we
are the best!” … but they all write that. Rather, you could motivate an independent trade-
journalist to write an editorial contribution about, say, “Restoring Marshalls”. You then find a
well-known musician not happy with the sound of his/her Marshall – and off you go. Taking
stock: a boring sound, odd harmonics (!), the worst Marshall since dinosaurs (of any kind)
roamed the planet. After this diagnosis, on to the therapy: swap components! You will want to
grab: genuine carbon-resistors, yellow or orange capacitors (depending on which supplier
forks over a more generous subsidy), and of course: a new mains transformer (it supplies all
that power, after all), and a new output transformer (all that music needs to pass through it),
and, since we’re at it, throw in a new choke. Ah – now we’re in brown-sound-city: the best
Marshall ever heard! Last, make the well-known musician rave about the unbelievable
improvement in sound, and make him/her recommend that everybody installs these wicked
transformers. Now, it only remains to hope that nobody checks www.tone-
lizard.com/marshall-myths, where a discussion can be found mentioning – with relish – that
for repairs frequently a damaged Marshall-transformer was exchanged for a low-cost no-name
transformer … and not a single complaint was ever received. Delightful stuff.
It is normal and necessary that manufacturer advertise their products; that they hire musicians
to praise the unrivalled sound may be criticized but there’s not much that can be done about
that. From a technical point of view, nothing stands against swapping a correctly working
Marshall transformer for an expensive clone. It is easily conceivable that a guitarist feels
better after the swap than before – but that has different reasons then.
The JTM-45 and its output transformer have achieved cult-status. Supply (meager) and
demand (high) now regulate the price (enormous). In Doyle’s Marshall-book we read,
however, that the differences to the 5F6-A-Bassman are in essence due to the different
loudspeakers (Celestion 12” vs. Jensen 10”), the different input tube (12AX7 vs. 12AY7), and
the higher negative feedback in the power-stage of the JTM-45. The RS-output-transformer is
not the reason for a special sound, as Doyle cites the design-director of Marshall. Hopefully,
nobody still believes that a steel-chassis will make the amp sound different compared to an
aluminum-chassis. You over there still do? Be informed that this is another myth. Aluminum
has paramagnetic characteristics while steel is ferromagnetic?! So? The effects on the sound
are about as dramatic as the color of the control-knobs is.

There is, however, a sound-determining parameter that has so far been investigated too little:
the 2nd-order harmonic distortion. In a transistor amplifier we usually pay close attention to
symmetry, and consequently even in overdrive mode the 2nd-order distortion is reduced to
insignificant levels. The tube power-stage, on the other hand, shows a rather different
behavior (e.g. Fig. 10.5.56): as the overdrive increases, the duty-cycle changes and k2 may not
be neglected anymore. Fig. 10.5.57 shows distortion measurements: the differences between
the individual curves are rather substantial. What is the reason? These are different tubes (all
KT-66). TungSol, TAD, and several original GEC-KT66 from the good old days. They are
accredited with qualities that allegedly are not achievable anymore today, and so a pair of
GEC-KT-66 may be offered (at the time of writing) for 280 Euro. That could be $699, as
well, if we jump to the other side of the Atlantic (Ebay, December 2013). Stiff prices, indeed.
Fig. 10.5.57: 2nd order (left) and 3rd order harmonic distortion, KT-66, Raa = 8 kΩ, 8-Ω-load at the 8-Ω-output.

The first line in Fig. 10.5.57 shows the scatter-width across 8 different KT-66-pairs. For pairs
c and d, only one plate-resistor was changed: 82kΩ /108kΩ were included rather than the
usual 82kΩ /100kΩ (differential amplifier). Immediately, the power-stage anti-symmetry
changes, as does the 2nd-order harmonic distortion. For e and f, the offset voltage at the grid
of the power tubes was changed from -53V/-50V (13mA/13mA) to -48V/-48V
(25mA/19mA). This KT-66-pair was ‘matched’ only to a rather lukewarm degree, despite a
4-digit-coincidence of the numbers on the sticker. Increasing the plate-current increases the
3rd-order distortion; the asymmetry in the current to some extent compensates for the disparity
in the tubes (ak2). For g and h another KT-66-pair was used, and the bias-current was changed
from 13mA/13 mA to 17mA/17 mA. It is surprising that the KT-66 requires such a small
bias-current for low distortion (it was operated with Rg2 = 1.5 kΩ for these measurements).
The bias-current could be easily adjusted in this JTM-45; no potentiometer is, however,
foreseen to set the symmetry. In fact, there are two values of relevance here: differences in the
drive circuit (plate-resistors in the differential amplifier), and the offset-voltages at the grids
of the power tubes. Of course, the respective individual KT-66 adds in, as well. Frequently,
carbon film resistors are recommended in order to achieve the original sound. However, such
resistors normally have a tolerance of ±10%! The variation of a plate resistor from 100 kΩ to
108 kΩ is comfortably covered by this tolerance span, but it will change k2 (at c) by more than
a factor of 10! If such variations are indeed considered to be relevant, it is not necessary to
shell out more than 15.000.- Euro for an old JTM-45 – one or two potentiometers added into
the circuit of a reissue amp will do fine (at a price of 3 Euro per piece).
The notion that KT-66’s produced today will not match the original data holds, in this
experiment, only for the TAD-tubes (which in the meantime may well be supplied by another
manufacturer – we did not investigate this aspect). The TungSol-KT-66’s are not generally
worse than the original GEC-tubes – quite the contrary. We had 8 GEC-tubes at our disposal
for the measurements. True vintage! Correspondingly, they were handled with great care. One
of these tubes was practically useless, two others had a gain so low that they could not be
used. The remaining KT-66’s worked well but their data corresponded only very moderately,
despite the "7500/7500"-marking on one pair and the "6500/6500"-marking on another pair.
The seemingly 4-digit correspondence (“matching”) did not keep the tubes from featuring
different gain – which will influence the harmonic distortion.
The above observations warrant the warning not to acquire vintage tubes (so-called NOS)
from unknown sources. This especially holds if the prices are significantly higher than those
of new tubes. At present, a pair of KT-66 is about 65 Euro, and it would be unwise to pay
much more. While it is indeed possible that vintage tubes on the market are well paired and
have been used little, and also feature small grid-currents and a good vacuum, they may just
as well be of abysmal quality. What can you do if (prepaid with an enormous sum) a parcel
arrives from far-away lands with tubes for which merely the label is correct? It may be rather
profitable to re-sell a cosmetically perfect replica (bought for a few Yuan) for $699 per pair –
but let’s mention that only in passing. It does of course not imply, that all NOS-tubes
necessarily are fakes (for more information search the web for “faked tubes”).
To avoid that the results shown in Fig. 10.5.57 are interpreted as stellar peculiarity of the
JTM-45: please take note of another warning. Similar curves are to be found with Fender
power stages, as well – this is not exclusive to Marshall! The plate resistors in the differential
amplifier, the degree of pairing of the power tubes, the bias current, the negative feedback –
all this determines the behavior of the power stage. The equation "vintage = great" does not
compute!

Fender – VOX – Marshall, the holy trinity: sure, it will not command the respect of every
guitarist, but the constantly recurring chorus in the “vintage”-columns of magazines has
generated the widely held opinion that the primordial VOX (or the proto-JTM-45, or the
ancient Bassman) is unequalled sonically, and easily justifies the $10.000 or 20.000 asked
today for the old originals. And of course, it is alluring to elect the top dog of this troika:
“compared to the 1959 Fender Bassman it is modeled after, the JTM outperforms its alter ego
with ease” (Gitarre&Bass, 7/2006). Onto the podium – long live the myth.
How should we imagine the scenery at the beginning of the 1960’s, when this legendary amp
came to life? Maybe like this (as some would have it): 39-year-old Jim sits behind his drum
kit, squeezing some sophisticated triplets between the screaming guitar-arpeggios, and thinks
‘that doesn’t sound like Hardrock at all – I’m gonna build ‘em a new amp with the right brit-
brown sound’. And then he tells Ken: ‘get on with it’ – and the result is the JTM-45 with its
unparalleled distortion sound? Maybe it was like that, with Jim the Rocker? This image does
not really fit the picture found in the books about Marshall: a friendly gentleman sporting suit
and bow-tie who probably makes his sticks dance across the skins in a more gentle manner.
Wikipedia sees the start of the Hardrock-era in 1969 but not in 1962. We know that Ritchie
Blackmore, Jimi Hendrix, Pete Townshend and many others came to fame using Marshall
amps, and it is easily imaginable that they voiced requests for more power – could that have
been in 1962, though? Townshend played (according to Wikipedia) in a Dixieland-band in
1959, then graduated to Skiffle, and the Who gets off the ground only as late as 1964. Deep
Purple forms in 1968, Hendrix starts his Experience in 1966, and Clapton plays with the
Yardbirds in 1963, miles away from any Beano-like tone. It is also sufficiently well
documented that Brian Poole (with his Tremeloes) was not an early exponent of Hardrock.
No contest: Jim Marshall has deservedly earned his medal as amp-pioneer – summa cum
laude, without any doubt. That does not imply, however, that the JTM-45 was developed and
optimized as distortion-heavy amp, even if this rumor is circulated within fan circles. Folks,
read closely what Ken Bran states in the Marshall book: "It was a bass amp we originally
wanted ... but the guitar sound was too good to pass up." The differences existing between the
bass, guitar, organ and PA-variants of the early Marshalls are limited to two small bridging-
capacitors to boost the treble. Had the JTM-45-circuit been developed to generate special
distortion, it would have also distorted vocals amplified by the PA-version – except for the
different treble gain, all these amps were identical. Many guitarists found (and continue to do
so) that the JTM-45 sounds really good when overdriven, but already in the description of the
distorted sound we find differences: according to Wikipedia, the Bluesbreaker combo (Model
No. 1962) was the amp that ”first led to the breakthrough of the typical Marshall sound”.
However, in Gitarre&Bass (07/2006) we find the statement that this same amp produces
”never a Marshall-typical distortion sound”. The author, writing a monthly column about
vintage amplifiers, is somewhat of a Nostradamus-of-the-tube-amp (i.e. not looking into the
future but backwards-oriented – we are talking vintage here!), and in terms of interpretation
simply congenial. A sample: “and the result (JTM-45) differed, in the end, strongly from a
Fender Bassman” (G&B 07/2006) versus “ the first of the so-called JTM-models were
therefore rather authentic copies (of the Fender Bassman)” (G&B 2/2005). Just like with
Nostradamus: it all depends on the year. Not a problem for anybody bred and raised in
Munich, Bavaria, and familiar with local poet Karl Valentin who wrote: “it has expertly been
calculated that the Lake of Starnberg (a well-known beautiful lake south of Munich) is, at the
same time, deep, shallow, long, short, narrow, and wide.

Fender Super-Reverb
A typical medium-power Fender amplifier holds two 6L6-GC’s, and the Super-Reverb is a
good example. The cathodes of the power tubes are connected directly to ground, and a
separate diode generates the negative offset-voltage: no doubt at all – this is textbook-class-
AB-operation. Ahead of the power tubes we find a differential amplifier, following them the
output transformer with a connection to the negative feedback loop. All in all it is a model for
the way Fender power stages looked like in the 1960’s. Still, there are individual
idiosycracies: the driver-tube may change (12AT7 instead of 7025), the coupling capacitors,
too; small blocking capacitors are discarded, then they return again – and even a “Presence”-
control is found in the ‘Super’ for a short time. The Super-Reverb investigated in the
following has the AB-763-circuit originating in the ‘Blackface-era’ i.e. in the golden 1960’s.
Fig. 10.5.59 depicts the output characteristics for ohmic loading of the 8-Ω-output (this
specimen of the amp had a transformer with such a connection installed). For the specified
load, the “knee” of the 0-V-curve is almost exactly met, indicating an optimum transformer
dimensioning. As the drive level rises, the curve is shifted towards smaller voltages.
Fig. 10.5.59: Characteristics for ohmic load of 8Ω (left) and 16Ω (right).
Note: the output transformer used here also had an 8-Ω-output on top of the regular 2-Ω-output.
The negative feedback in this power stage is not as strong as it is in the JTM-45, and therefore
the loudspeaker impedance is more clearly represented in the transmission frequency response
(Fig. 10.5.60). For all these diagrams, it is important to recognize that the exact shape of the
curve depends on the specific loudspeaker: the loudspeaker resonance, which is about 75 Hz
in the given example, may rise to over 100 Hz with other speakers. This of course has an
effect on the sound (compare to Chapter 11). If the power stage is overdriven, the influence of
the speaker diminishes and the characteristic becomes closer to that of a voltage source. This
is shown in upper curve of the left-hand picture.
Fig. 10.5.60: Frequency response of a Super-Reverb power amp, 8-Ω-output loaded with 4xP10R (left).
Right: magnitude of the loudspeaker impedance (4xP10R, cabinet set up in reflecting surroundings).

The transmission characteristic of the power stage for 8-Ω-loading is shown in Fig. 10.5.61.
As the drive level rises, the curve decomposes into two branches that slide apart. As has
already been noted, the reason is the polarization of the coupling capacitors.
Fig. 10.5.61: Idealized characteristic (left), output time-function with ohmic nominal load (right).
We can see from Fig. 10.5.62 that the output power of 40 W (as specified e.g. in the 1968
catalog) is actually achieved. This is in sharp contrast to the JTM-45, the replica of which is
advertised by TAD (in 2008) with “about 45 Watt” but reaches merely 30 W. The minimum
of the harmonic distortion is due to the progressively curved characteristic that changes the
direction at the onset of distortion. The strain on the power tubes is similar to the JTM-45: the
screen grid is overloaded for overdrive operation with a high-impedance load (Fig. 10.5.63).
One significant difference is found in the input capacitor: if it is only 1 nF (AB763), the plate
is overloaded less (compare to Fig. 10.5.55). There are, however, also Fender amplifiers with
a larger input capacitor (e.g. 10 nF).
Abb. 10.5.62: Super-Reverb: harmonic distortion, output power at the 8-Ω- output with 8-Ω- and. 16-Ω-load.
Fig. 10.5.63: Power dissipation at the plate for both output tubes (left); power dissipation at the screen grid for two different load impedances
(right). The level of the input signal (500 Hz) rises linearly from 0 – 3 s, switch-off occurs at t = 3 s. From t = 1.3 s, the power stage is
overdriven.

Comparison of power stages
The 1960’s holy trinity: VOX, Marshall, Fender. Of course there are also Gibson, Ampeg,
Hiwatt and many more, but the ‘big three’ stand out. So, what makes for the difference
between these amplifiers or, rather, between the respective power stages (to do this chapter
adequate justice)? This question cannot be answered generally because there is not the one
Marshall- or Fender-amp. Even at VOX, the AC-30 ran through several production variants.
For Fender, Dave Funk lists 250 pages of schematics – and still has not captured all Fender
amps. Practically every amplifier model (e.g. the Bassman) was, over the years, built in many
variants, and there are many models to begin with. It is therefore impossible to speak of one
Fender-typical circuit, or of one Fender-typical sound. The situation is similar for Marshall –
only the AC-30 remains reasonably true to itself, although even here there are modifications,
e.g. the models developed for the US that only seemingly were similar to the UK-standard.
Even when concentrating on only three special power stages, a comparison turns out to be
difficult due to many small differences in detail. Most important are category of output-
power, negative feedback (and correspondingly the internal impedance), and balancing
processes during overdrive conditions. Even the loudspeaker needs to be considered although
it is not part of the power stage: its impedance determines the load on the power stage and
thus the frequency response of the latter. The circuitry preceding the power stage plays a
considerable role, as well: is it of high or low impedance, and what voltage can it offer
without distorting? If the power stage were a linear and time-invariant system, we could
record its frequency response and have a good starting point for comparisons. However, guitar
amps are subject to overdrive (i.e. they are operating as non-linear systems), and therefore a
small-signal analysis allows for only very limited conclusions on their behavior.
To illustrate the problems appearing when comparing amplifiers, let us look at the VOX AC-
30 and the Fender Super-Reverb. The VOX offers 30 W, the Fender 45 W. In the VOX-
cabinet we find two 12”-loudspeakers while four 10”-speakers are deployed in the Super-
Reverb. If we allow for each power stage to work with its original speakers, we not only
compare the power stages but also the loudspeakers. Should we consider connecting the
VOX-speakers to the Fender, we risk blowing them because Celestion specifies only a 15-W-
load for each speaker. Moreover, the nominal impedance the Super-Reverb is specified for is
2 Ω, while it is 16 Ω for the VOX. One could re-solder the VOX-speakers to a 4-Ω-
configuration, but that would result in yet another different scenario. How about the other way
round: operating the VOX with the Fender speakers? That would work in terms of power
capacity, but the issue with the different output power remains: it could result in differing
loudspeaker distortion (with the sub-harmonics being level-dependent).
Therefore, the chosen approach would have to be to use only one and the same loudspeaker
for all amps to be compared. What would remain now as power-stage specific differences?
First, the internal impedance: it is high in the VOX, medium in the Fender and low in the
Marshall. The speaker impedance will therefore more or less shape the frequency response. At
resonance, the loudspeaker impedance can rise to 40 Ω or even 150 Ω, implying a voltage-
level difference of almost 12 dB for a high-impedance source and an almost unchanged level
or a low-impedance source. This is an enormous difference that is neither due to the power
stage by itself nor caused by the loudspeaker by itself (Fig. 10.5.64). Even though the power
stages are neither pure voltage sources nor pure current sources, the corresponding difference
between an AC-30 (Ri ≈ 80 Ω) and a JTM-45 (Ri ≈ 2 Ω) is considerable.

Fig. 10.5.64: Frequency responses from phase-inverter input to

loudspeaker output.
AC-30 with 2x12"-Celestion in combo-enclosure,
Super-Reverb with 4x10"-Jensen in combo-enclosure,
JTM-45 with 4x12"-Celestion in separate 1960AX enclosure.
The reason for the different internal impedances is the negative feedback (NFB) in the
power stages: it is strong in the JTM-45, somewhat less strong in the Super-Reverb, and non-
existent in the AC-30. Besides influencing the internal impedance, the NFB also has an effect
on the non-linear distortion of the power stage: this distortion is stronger in the AC-30 and
smaller in the Super-Reverb and, in particular, the JTM-45. The type of distortion varies, as
well: with increasing overdrive, the duty cycles in the JTM-45 and the Super-Reverb change,
and correspondingly 2nd-order distortion mounts. Conversely, the output signal remains
largely half-wave anti-symmetric in the AC-30, with k3 remaining dominant.
The output transformer influences the output signal, too – though less than first expected
(see Chapter 10.6.5). In the low-frequency range, harmonic distortion caused by the
transformer can become audible – but only for really low-quality transformers. All
transformers investigated here gave no cause for complaint. Because not all transformers
have the same turns-ratio, the frequency responses differ a little; this, however, is no secret
science – in essence this is a matter of the number of turns in the windings.
The ratio of impulse power to continuous power, and the hum-interference-modulation is

not alone a characteristic of the power stage but the power supply is involved, as well. The
Super-Amp 5F4 had a capacitor of 16 µF connected after the rectifier tube, and another 16 µF
after the choke. That was indeed rather modest, so the successor receives 40 µF / 20 µF. With
Marshall, the JTM-45 first included 32 µF / 32 µF, but the model 1987 filtered with an ample
100 µF / 50 µF. Started out with 16 µF / 16 µF, the AC-30 was upgraded to 32 µF / 32 µF
later. It is a well-known fact that all these electrolytic capacitors often had considerable
tolerances.
The plate resistors in the phase inverter are a science in themselves: we have 82k/100k with
the 7025 in the 6G4, 100k/100k with the 12AT7 in the AA763, 47k/47k with the 12AT7 in
the AB568. In the Marshall, an ECC83 with 82k/100k is at work, and in the VOX an ECC83
with 100k/100k. The anti-symmetry of the phase-inverter outputs influences the even-order
distortions in the power stage. The scatter of component values can have an extremely strong
effect, and, of course, the equality of the power tubes plays a role, as well (“matching”).
The power tubes: EL84, 6V6GT, 6L6GT, KT66, EL34, KT88, and relations. This is a
difficult topic because there is not “the” 6L6GT – the scatter can be very wide. The
acquisition of a large number of 6L6GT (say 12 pieces) does not help here, either: if all
twelve tubes are from the same production batch, they might have similar parameters, but if
we later buy another pair, the parameters might well be entirely different. It has already been
elaborated that “selecting” and “matching” are no cure-alls, either (Chapter 10.5.11). The
measurement results listed in the following are therefore to be taken merely as a snapshot to
provide orientation values of limited general validity.

10.5.13 Comparison of power tubes
What happens if the two 6L6GC in a Vibroverb are swapped for a pair of KT66 – or for a pair
of EL34? Citations of how these tubes allegedly sound have already been given on the
preceding pages. Let us first disregard the sound – how do the electrical data change? First,
there is the heating: two 6L6GC require a filament current of 1.8 A, two KT-66 demand 2.6
A, and two EL34 already push this to 3 A. The mains transformer is therefore put under
different strain, but let us insinuate by all means that it can take the additional load for the
short term. The bias voltage at the grids (i.e. the bias current) needs to be adapted, of course –
and now what? Does the frequency response change substantially due to the tube-swap? What
about the harmonic distortion? More generally: which part do the tubes play in the operating
behavior of the power stage?
The simple solution: it is the output power that depends on the power tubes – and that’s it.
You may or should add a few bits here and there, but in essence, this is the sobering answer.
We do find differences already with regular instrumentation, but the relevance to the sound
remains very modest. The measurements discussed in the following were taken from a
Marshall power stage that was, however, operated via a stabilized 400-V-power-supply. One
of the two plate resistors of the differential amplifier was adjustable in order to balance out
different gain of the power tubes. The primary impedance of the output transformer was Raa =
3.5 kΩ; the resistors at the screen grid had 1.5 kΩ each. To emphasize any differences, the
negative feedback in the power stage was deactivated. Nominal load implies that an 8-Ω-
resistor (purely ohmic) was connected to the 8-Ω-output. The signal generator was directly
connected to the input of the differential amplifier (→ vU).
Fig. 10.5.65: Output power vs. voltage gain.

6L6GC from GEC, TAD, JJ, Ultron, TungSol, Sovtek;
5881 from Sovtec, TungSol;
KT66 from TAD, TungSol, Marconi;
EL34 from TubeTown, TAD, JJ, EH, Valvo;
KT88 (and 6550) from Sovtek, GEC, EH, SED;
Power measurements were taken at 500 Hz at the onset of clipping and with
nominal load. Any influences due to the power supply were eliminated using a
stabilized plate-voltage (400 V)
Fig. 10.5.65 gives an orientation regarding the output power and the gain of the power stage
with different tubes. The sample was very small (20 6L6GC, 10 EL34, 10 KT66, 8 KT88),
and therefore it is to be expected that the market will offer specimen with data lying outside of
the grey areas. Already the tubes measured here show considerable scatter in the maximum
power: a fresh pair of 6L6GC may yield the datasheet-conform 40 W, or a meager 27 W. For
the EL34, the span extends form 38 W to 55 W, and consequently the changeover 6L6GC →
EL34 could bluntly double the power … or reduce it some. In any case, the probability that
the voltage gain goes up by about 2 – 5 dB is high. The higher maximum power could lead to
stronger distortion in the connected loudspeaker – but this should not tempt us to generally
attest more distortion to the EL34. If at all, these would be indirect tube characteristics. How
much the power stage itself distorts, that will be subject to the following analyses.

All these analyses were done at 500 Hz and with nominal load. Before we investigate the
harmonic distortion, let us take a look at the transmission characteristic i.e. the mapping of
the generator voltage onto the output voltage generated across the 8-Ω-load-resistor. With a
small bias-current (20 mA), we see a saddle-point for small drive levels, in other words a
progressive curvature of the characteristic. This changes into a degressive curvature with high
bias-current (60 mA). All three curvatures are depicted in Fig.10.5.66 in an idealized manner.
The progressive characteristic rises from the origin with
increasing slope, the proportional dependency shows a
constant slope, and for the degressive curve the slope
decreases with increasing drive. At 25 V the so-called
clipping (limiting of the ordinate values) sets in. These curves
are, however, idealized; for the real tube there is no perfect
proportionality: all curves are “somehow bent”. The exact
shape depends on the geometry of the tube-electrodes and on
how equal or unequal the two power tubes are; it will
therefore be different for each push-pull power stage. See
Fig. 10.5.66: Transfer characteristics Chapter 10.5.3 for the basics of push-pull operation.
From the idealization on to real tubes: Fig. 10.5.67 shows three transmission characteristics of
a Sovtek 5881♣. It corresponds best to the above idealization (which does not necessarily hold
for all tubes of this type, and much less for Sovtek in general). Seeking the least distortion, we
would have to choose the middle curve (40 mA). The Groove-Tubes 6L6GC shown next also
allows for a good proportionality, although it requires 60 mA cathode-current, which makes –
at 400 plate-voltage – already for a pretty hot operation. The Tung-Sol 5881 again is more
similar to the Sovtek 5881 – so much so that the conjecture finds support that the two tubes
may differ only in the labeling. The next tube, an old Marconi, can keep up well with the
others although this is not a 6L6GC but a KT-66.
Fig. 10.5.67: Measured transmission characteristics, nominal load, IK = 20, 40, 60 mA.
♣
The 5881 is the professional variant of the 6L6GC.

For the next tube, the family of characteristic curves looks similar, too, as it does for the last
specimen in this overview. These are entirely different tubes, however: here we have a pair of
EL34’s and a pair of KT-88’s. We do see some differences at the drive limit but the basic
curves are very similar indeed. For these 6 pairs, that is! The tubes from Fig. 10.5.68 show
that more strongly bent curves exist, as well – to a varied degree. For the JJ-6L6GC we see a
pronounced ripple while the TAD has a more degressive characteristic – this shows that the
difference between a 6L6GC and a KT-66 is not necessarily bigger than the difference
between two KT-66’s.
Fig. 10.5.68: Measured transmission characteristics; nominal load, IK = 20, 40, 60 mA.
The x-vs.-y-depiction is not very suitable to clarify how well the transmission characteristics
can follow the ideal proportionality – the analysis of the harmonic distortion delivers better
results here. The level of a 500-Hz-tone was increased by 30 dB over the course of 4 seconds,
and at the same time the levels of the first 20 harmonics were extracted from the output signal
of the power stage (software Cortex VIPER). The results are shown in Fig. 10.5.69.
Fig. 10.5.69: 3rd-order harmonic distortion vs. output power, 500 Hz, nominal load, NFB deactivated.

In Fig. 10.5.69, a maximum of the distortion attenuation (= distortion minimum) appears

around 40 W. This is an effect of the change in curvature in the range of the clipping-onset;
the 3rd harmonic changes its algebraic sign here. As different the curves may look – the non-
linear distortion can be easily reduced to inaudible levels by choosing the appropriate bias-
current. However, “appropriate” means a whopping 60 mA for the GT-6L6GC, but no more
than 30 mA for the JJ-6L6GC. In the case that distortion is heard when comparing power
tubes: that may simply be due to an inappropriate bias-current!
We have all heard or read opinions related to the ‘tube-sound’: "simply unplug the two 6L6GC
and plug in two KT-66 – you gotta hear that difference!" Just like that, without considering
the bias-current? It is almost certain that, with such a makeshift setup, differing modes of
operation are evaluated rather than the difference in the tubes per se. Setting the bias-current
via the bias-voltage at the grids will not remove the problem: mind you, for the same bias-
voltage at the grids, three "premium matched" 6L6GC-pairs show sizeable variance in the
bias-current (30 mA vs. 40 mA). The voltage gains for the two half-waves did not match,
either♣, with 10% difference. This problem persists in particular with the famed NOS-tubes
because normally there will not be 20 of them available to choose the ones matching best.
Fig. 10.5.69 compared four beam-tetrodes. Does the EL34, a real pentode, show other curves?
Yes and no, as we see from Fig. 10.5.70. In the details, there are strong differences and in
particular there is more power, but we cannot speak of a generally different behavior. If we
take k = 3% as the limit for audible non-linear distortion in a guitar amp, all tubes generate
audible distortion only just before going into clipping, if the bias-current is set correctly. They
also show the same type of increase of the THD. Moreover, we must not forget that the THD
will decrease significantly as we (re-) activate the negative feedback.
Abb. 10.5.70: As Fig. 10.5.69, but for EL34 and KT-88; without NFB, 20 – 60 mA.
♣
Amongst tube retailers, the readiness to match and pair (up) does not seem to be particularly distinct ...

Time for an interim statement: yes, the tubes under investigation show variances, but in
essence this is limited to power output and gain. While we find differences between 6L6GC
and EL34 regarding the individual distortion characteristics, similar differences are observed
between two pairs of 6L6GC. Is it, however, sufficient to analyze only these parameters?
What about the internal impedance (also termed source impedance)? Fig 10.5.71 shows the
corresponding measurements. Pentodes are of high impedance, the frequency dependency
mainly stems for the output transformer (Marshall). The resonance of the winding receives
differing dampening from the power tubes; this makes for a different height in the maxima.
The left-hand picture depicts the impedance curves for six pairs of 6L6GC (lines), two pairs
of 5881 (dashed), and one 6L6WGC-pair (dotted). The right-hand picture shows the results
for two KT-66-pairs and for three pairs of EL34. What’s interesting: the lines close to each
other are the dashed line (KT-66) and solid line (EL34) i.e. they are not the ones that would
“belong” together. So again, we do not see a general difference.
Fig. 10.5.71: Frequency response of the internal impedance measured at the 8-Ω-output; IK = 40 mA.
The internal impedance (which is also dependent on the bias-current) influences the
dampening of the loudspeaker and thus the transmission frequency response. This however
holds mainly for power stages without negative feedback. Chapter 10.5.14 elucidates how
much this effect looses its significance as soon as the negative feedback is in action.
Was that it? No! It is a widespread error to limit testing of power amplifiers to merely the
nominal impedance as a load. Loudspeaker impedances are frequency-dependent♣, and
therefore supplemental measurements with a load of 32 Ω follow here. The main difference
occurs between beam-tetrode and pentode (Fig. 10.5.72): with a true pentode (e.g. the EL34),
the current through the screen grid increases within the distribution area towards much higher
values. This conversely implies a reduction of the plate-current i.e. a pronounced sharp bend.
Fig. 10.5.72: Measured transfer characteristics, 32-Ω-load at the 8-Ω-output, IK = 10, 30, 50 mA.
♣
It was already documented in Fig. 10.5.28 that the speaker does not have of a real but a complex characteristic.

The sharp bend appearing in the EL34-characteristic (Fig. 10.5.72) generates a different curve
of the harmonic distortion – see Fig. 10.5.73. Still, we again need to heed here that the
negative feedback is deactivated. As typical NFB is brought in, this effect looses its
significance, as can be seen from the dashed lines.
Fig. 10.5.73: 3rd order distortion attenuation vs. output power, 500 Hz, 32 Ω at the 8-Ω-output.
Negative feedback activated (dashed) and deactivated (solid line).
As different as the individual distortion attenuations might be: if we take 30 dB as audible

limit, only maximum power and gain are left as main criteria (with the NFB activated).
Because the EL34 has (in the present series of measurements) 2 – 5 dB more gain than the
6L6GC, the frequency response of the amp with active NFB changes somewhat. The higher
the gain (or the transconductance) is, the stronger the effect of the NFB, and the smaller the
influence of the loudspeaker impedance on the frequency response (Fig.10.5.74). In case you
regard these differences as essential: simply change the negative-feedback circuit .
Fig. 10.5.74: Transmission frequency response.

Power stage with negative feedback, 16-Ω-output,
Voltage level at the Marshall-Box (1960-AX).
Sovtek 5881 (red), TAD El34 (black).
For the upper two curves, the power stage is overdriven.
Conclusion: a representative comparison is not possible because even the data of selected
tubes include a scatter, and because the vendors do not guarantee any limit values. The sample
analyzed above shows measurable differences between various 6L6GC, 5881 (6L6WGC) und
KT66, but these will most probably lie within the assumed production scatter. They are also
of secondary importance in everyday studio- and stage-operation. Within the sample, the
EL34’s distinguish themselves regarding maximum power and transconductance, and from
this a marginal difference in the frequency response results. No investigation could be carried
out regarding the lifetime. For example, in order to check the 10.000 h propagated by MOV,
14 months would be needed! If we would set up 15 power stages per tube-type in order to
meet the minimum statistical requirements, the cost for mains power alone would amount to
20.000 Euro – that is not reasonable. Who can guarantee that after the conclusion of the test,
the tube vendor will not have “his” special 6L6-WGC-STR-XXL-premium-selected built by
another manufacturer? Due to the even better offered quality, as the vendor writes … or rather
because he did not buy a sufficient quantity off the first manufacturer…

For the “small” power tubes 6V6GT and EL84, another degree of freedom is added: besides
amplifiers with a fixed offset-voltage at the grid (e.g. the Deluxe reverb), there are also amps
with a cathode resistor (e.g. the AC-15). This resistor has several effects: the DC current
flowing through it generates the offset-voltage, the power dissipated in it is lacking in the
loudspeaker, and in overdrive mode it changes the operating point.
In the AC15, the cathode-resistor has a value of 130 Ω generating about -10 V offset voltage,
and somewhat more than 40 mA per tube. The tweed Deluxe has 270 Ω, -23 V and more than
40 mA, respectively (depending on rectifier tube and power tubes). With increasing drive
level, the average cathode-voltage rises (due to the non-linearity in the tubes), and the
operating point shifts towards the ”cooler” range. The transfer characteristic becomes flatter.
The following measurements were again taken using a stabilized power supply (300 V), with
Raa = 6.2 kΩ, no negative feedback, Rg2 = 470 Ω, ohmic nominal load.
Fig. 10.5.75: Transmission from the phase-

inverter to the load-impedance. Power stage
with cathode-resistor.
EL84: 120 Ω, 6V6GT: 270 Ω,
bridged with 250 µF.
Three different drive-levels.
We can see from Fig. 10.5.75 that the EL83 and the 6V6GT differ in gain by 7.5 dB, and that
the gain drops by 4 dB with increasing drive-level. With a fixed bias-voltage, such a gain-
reduction cannot be observed (compare to Fig. 10.5.45). As the overdrive increases, a saddle
point in the origin appears for both tubes – here the true pentode differs from the beam-
tetrode, though: in the EL84, the characteristic has an almost horizontal slope in the origin
(Fig. 10.5.76) while for the 6V6Gt, this gain-decrease is much weaker.
Fig. 10.5.76: Transmission from phase-

inverter to load impedance.
With RK (black) and without (red).
Overdriven power stage.
This dip in the transmission curve has several reasons in a guitar amplifier: as the drive-level
increases, the supply voltage decreases, and Ug2 with it; the coupling capacitors towards the
phase-inverter change their polarization; if a cathode resistor is present, the voltage drop
across it increases. All three effects build up in the same direction and shift the operating
point towards the “cooler” range, and consequently the crossover distortion close to the origin
increases. Last, the screen-grid resistor needs to be considered, as well: in the EL84, the
currents through the screen grid are larger than in the 6V6Gt, and therefore the voltage drop
across the screen-grid resistors is bound to be different.

Fig. 10.5.77: Attenuation of distortion ak3, ak5 and ak7 with nominal load. With (black) and without (red) RK.

In Fig. 10.5.77 we see the first three odd-order distortion attenuation characteristics. For k5
and k7, the situation is clear: without cathode resistor, the power stage distorts more than with
this resistor, and the EL84 distorts less than the 6V6GT. For the 3rd-order distortion, such
tendencies are less pronounced. Strongly overdriven, the amp again has the larger output
power without RK. At around 10 W, however, k3 is subject to strong fluctuations that are
different from tube to tube.
The frequency response of the internal impedance is inconspicuous; the measurements did
not lay open any significant differences between 6V6GT and EL84. With regard to the
operation with and without cathode resistor, no significant differences in the internal
impedance could be found, either – as long as the idle-currents were set comparably.
Fig. 10.5.78: Output power vs. gain (left); transmission characteristic (center, right). “ohne”=w/out, “mit”=with.
In the left-hand section of Fig. 10.5.78, output power (at 300 V) and gain are shown; again
this is only for a small sample. The centre section depicts the effect of the internal impedance
of the power supply (RNT). “ohne RNT“ (without RNT) indicates the stabilized 400-V-power-
supply, being used; “mit RNT“ (with RNT) indicates operation from a stabilized 460-V-power-
supply, but via a 240–Ω-resistor, and buffered with 47 µF. With the internal impedance of the
power supply present, the supply voltage to the power stage drops to 400 V at full power; the
bias-current, however, is set for a supply voltage of about 440 V. As the drive-level increases,
the operating range wanders off towards “cooler” regions – comparable to the operation
without RNT, but with a bias-current of only 20 mA (right). One highly essential difference
remains: the red curve is not static! Its slope (= gain) drops with decreasing drive-levels.
Fig 10.5.79 again documents (for ak2) how strong the scatter within new, selected tube pairs
can be. Anybody who believes that a THD of 1% is relevant needs to buy better-
selected/matched tubes.
Fig. 10.5.79: 2nd order distortion attenuation; in each picture two newly bought, “matched” tube pairs.

10.5.14 Pentode/Triode/Ultralinear
Pentodes are at work in the power stage of a typical guitar amplifier: EL-34 in the Marshall,
EL-84 in the VOX, 6L6-GC in the Fender; or comparable tubes (5881, KT-66, KT-88) – but
always pentodes, and not triodes. That some of these tubes actually are beam-tetrodes shall
not bother us here, because their beamforming plates are in fact some kind of fifth electrode,
as well – even though differences to a true suppressor-grid remain if we apply strict theory.
Since these differences are of no significance in the following, we will treat pentodes
synonymously with beam-tetrodes.
In a triode, the drive-level-dependent plate-voltage accelerates the electrons, and therefore the
gain of the tube is small for low plate-voltage. Conversely, the plate-voltage has only little
influence on the emission-current in the pentode because the screen grid is on a high potential
independently of the drive-level. The output characteristics of the pentode are therefore
almost horizontal (except for the initial distribution area), and the internal impedance is larger
compared to triodes. From an overall investigation of efficiency, internal impedance and
harmonic distortion, HiFi-developers noticed that both pentode and triode were operating in a
sub-optimal border range, and they looked for a compromise. The latter could be found in the
ultra-linear circuit: here, the screen grid of the power tubes is connected neither to a
constant potential (pentode operation) nor to plate potential (approximately corresponding to
triode operation), but in between. Since all voltages between supply voltage and plate-voltage
are available at the output transformer, it is merely necessary to include a suitable “tap” from
the primary winding. This is why an ultra-linear output transformer does not have three but
five connections on its primary side. The back-channeling of the signal to the screen grid (g2)
has the effect of a negative feedback that was seen as advantage in HiFi-amplifiers. It appears
the same did not happen for guitar amps since only few experiments made the jump to
production, such as the 1979 Twin Reverb, some Sunn-variants or – will wonders never
cease! – the 200-Watt-Marshall: yep, there’s an ultra-linear power stage, designed for the
least amount of distortion. Later, though, the JCM-800-series amps went after their business
again without ultra-linearization.
Due to the reduction of the screen-grid-voltage, the obtainable maximum output power drops.
This can be used to convert a 100-W-amp into a 50-W-amp. Two switches are installed at the
screen grids such that either the full supply voltage or the plate-voltage is connected to the
screen grids. Sure, it would also be possible to simply reduce the gain if it gets too loud, but
power-stage distortion happens only when overdrive occurs. If the screen grids are connected
to the (corresponding) plates, the power pentodes operate in a kind of triode-mode: with
smaller maximum power, but also with smaller internal impedance. The switch from pentode-
to triode-mode therefore does not only change the maximum power (loudness) but also the
sound. The operation with high internal impedance emphasizes treble and speaker-resonances,
and in the triode-mode the sound looses brilliance and volume. To which extent this is in fact
audible depends on (besides the screen-grid-voltage) the negative feedback of the power
stage. Changing the screen-grid-voltage implies changing the gain (i.e. the loop-gain) and
therefore changing the negative-feedback-factor. It may consequently be that, besides the
screen-grid-voltage, the NFB-loop needs to be switched as well. The pentode/triode-switch
has no bearing on the operating point because at idle, the plate-voltage is almost the same as
the supply voltage (the primary winding is of low-impedance of DC current).

Fig. 10.5.80 shows an example for the difference between pentode and triode. An EL34-
power-stage with Marshall-transformer (JTM-50) is operated with a stabilized voltage of 400
V, and a Marshall-Box 1960-AX as loudspeaker. First, the investigation targets the influence
of the screen-grid resistor at 1.5 kΩ, 470 Ω, or 0 Ω. A small effect shows up with the gain: for
Rg2 = 0 Ω, the gain is 1.3 dB larger than for Rg2 = 1.5 kΩ. The strain on the screen grid is
heavily affected: with overdrive, the screen grid glows barely visibly♣ for Rg2 = 1.5 kΩ, but
lights up to bright red with Rg2 = 470 Ω, and to bright yellow at Rg2 = 0 Ω! In the interest of
a long tube life, a sufficiently large Rg2 should always be used. The flipside is that the power
stage generates, with Rg2 = 0 Ω, 1/3 more power compared to Rg2 = 1.5 kΩ. Let’s now look at
the figure: the triode configuration reduces the internal impedance, which makes the gain drop
in a particularly strong manner for a high-impedance load (Chapter 11.2). The gain is smaller
by about 5 dB at 400 Hz in the triode-mode; in the figure, this was balanced out for small
drive level (normalization to 400 Hz).
Fig. 10.5.80: Voltage transmission for pentode- and triode-operation of the power tubes (EL34, 1960AX).
The measurement curves at small drive level were normalized to 400 Hz; no normalization was done for high drive level. On the right the
frequency responses of source-impedances of the power stage are shown (internal impedance at the output transformer). For all these
measurements, the negative feedback was deactivated.
At small drive level (i.e. for linear operation), the power tubes are of high impedance (about
30 kΩ) in pentode-mode, and the source impedance measured at the output is co-determined
by the output transformer. In triode-mode, the internal impedance of the tubes drops to about
1.2 kΩ: now, the source-impedance is predominantly determined by the internal impedance of
the power tubes. As the power stage is driven to the extent that limiting occurs (measurement
curves for high levels), we see the differences in maximum power output. On the other hand,
the characteristics of the curves clearly become more similar. Conclusion: In triode-mode,
brilliance and emphasis of the speaker-resonance drop for undistorted operation. The
maximum power-output drops to about 1/3, and when overdrive occurs, the frequency
responses become similar.
For the power stage with active negative feedback, the differences in the frequency
responses is much smaller for linear operation while the differences in the maximum power
output remain similar. It is well known that negative feedback has no effect on the maximum
power yield: the non-linear distortion is reduced somewhat but the limit values of the tubes
cannot by modified via NFB. However, even merely moderate NFB decreases the source
impedance so strongly that differences in frequency response between triode- and pentode-
mode become meaningless. Simply put: negative feedback transforms the power stage from a
current source to a voltage source.
♣
As seen with the JJ-EL34; different strain-situations can occur with other tubes.

Fig. 10.5.81 shows how gain and source impedance are reduced by the negative feedback.
The NFB-factor (= 1 + loop-gain) depends, for a tube amplifier, on the load (speaker
impedance). Measurement and calculation (circles at 400 Hz) match very well. For an NFB-
factor of 1, the power stage has no negative feedback, while an NFB-factor of 6.73 already
represents a strong feedback for power stage based on tubes. The frequency characteristics of
pentode- and triode-mode become very similar already for moderately strong NFB, and
therefore in practice all that remains is a small difference in gain.
Fig. 10.5.81: Gain from the input of the phase-inverter to the output of the transformer (16 Ω, 1960AX).
Left: pentode-mode, right triode-mode. Lower line of pictures: source impedance; “GKF” = NFB-factor.
Besides the drop in output power and the change in frequency response, switching from
pentode- to triode-mode brings a further consequence: the non-linear behavior changes. The
triode-characteristic has multiple bends while the pentode characteristic is degressive. This
has an effect on the distortion-attenuation: a reversal of the sign of the curvature (2nd
derivative) leads to a zero in the harmonic distortion in triode-mode (Fig. 10.5.82).
Fig. 10.5.82: Transmission characteristic (left), 3rd-order distortion attenuation (right). 2xEL34,
nominal load (ohmic), no NFB. The exact curve depends on the individual tubes and on the bias-current.

10.5.15 … and the current flows on while you are long dead
Suggestions regarding modifying a tube power-stage will entice to do just that. Swap your
power tubes, install different filter caps, modify that negative feedback. To cite myself: The
fact that not everybody who removes an amp chassis from a cabinet instantly keels over dead
must not lead to the conclusion that this will never happen [Chapter. 10.5.8]. A tube amp
operates on the basis of life-endangering voltages, any musician screwing (sic!) with it, as
well. Therefore, let me repeat: working on a tube amp requires a specialist education. And
even if the courageous/experienced/lucky customizer is left unharmed: it is bad enough if the
power transformer gets fried. Or if the loudspeaker expires right in the midst of the most
important solo of ones life … because it could withstand the original 40 W, but decided to
succumb to those after-mod-80 W.
Books, magazines, and fora on the world-wide-web, are filled with recommendations how to
customize your amp. More crunch, more bass, more treble, more oomph, more of everything.
Swapping the output transformer can lead to additional strain on the mains transformer and it
can overload the rectifier tube (if one is in the game). The expert can size up all this but the
layperson can’t. 6L6-GC and KT-66 may be swapped for each other, as long as the bias-
current is correctly adjusted afterwards. A change from the 6L6-GC to the EL34 represents
already a potential power increase, and needs to be carefully considered and implemented.
The socked-connections need to be checked when doing this because there are differences
here. Power stages with the 6V6 are particularly dangerous candidates: whoever – hoping for
the triple output power – plugs in two 6L6-GC (or even EL34’s) instead of the 6V6-GT acts
negligently. If the object of experimentation were a tweed Deluxe, we would first need to
have a look at the rectifier tube: the 5Y3-GT is a good partner for the 6V6-GT, but not for the
EL34. So that needs to be changed, too: instead of the 5Y3-GT, the 5U4G gets to be plugged
in – or should the GZ34 be used, yielding limitless current? But then there’s the cathode
resistor: 270 Ω. The EL34 easily exceeds a cathode-current of 300 mA, so that’s 24 W
dissipated in the cathode resistor. Which actually is a power increase of some kind – but
probably not the desired one. For the Deluxe Reverb, this problem disappears: there is no
cathode resistor. Still, the mains transformer needs to be watched: it not only needs to supply
an additional 2 A of filament-heating current, but also the desired additional output power.
Likely to be forgotten is the increased power dissipation in the tubes: it’s about 15 W for the
6V6-GT but double that for the EL34. The EL34 is a true pentode but the 6V6-GT is not.
Those who are “in the know” can do such conversions. But then you read in a forum: my new
transformer has a wire more than the old one … what should I do? Or: the big resistor is shot.
Where can I get a new one? Or can I just leave it out♣? Simple answer: HANDS OFF!! You
don’t get a kidney transplant done in your auto-shop, either, now do you?
♣
In fact, what we read is: That biggie resister is in ashees where do i get anew 1. Or cann I jus leave it of?

10.6 Output Transformer 10-157
10.6 Output transformer
Typically, your customary power tube will have an optimum load-impedance in the kilo-ohm-
range i.e. about 1000 times the impedance of a loudspeaker. If, for example, an 8-!-load-
impedance were to be connected to a source having an internal impedance of 8000 !, then
99,9% of the generated power would be dissipated via the internal impedance, and only 0,1%
would arrive at the load-impedance. That is of course not acceptable. Tubes operate at high
voltages (400 V) but can digest only small currents (0.2 A). With loudspeakers, the situation
is exactly the other way round: a 4-!-loudspeaker requires 16 V to take on 64 W, with a
current of 4 A flowing through it. The output transformer (OT) has the task to match the
high-impedance tube circuit to the low-impedance loudspeaker. As a matter of principle, the
OT at the same time works as a filter that rejects high and low frequencies, and it generates
special non-linear distortion. While the matching function of the OT is relatively easily
calculated, the non-linear distortion eludes an exact description. The corresponding models
are therefore either inadequate, or not at all readily understood, or both. The following
elaborations try to give a clear picture on the basis of specific measurements. For the latter,
genre-typical output transformers were used – they do, however, not represent any selected
sample-median.
10.6.1 The linear model
Impedances (complex resistances) are only defined within the linear model [20], and therefore
the impedance transformation can be calculated only for a linear output transformer. The AC-
source is the tube circuit that is assumed to be a voltage-source with a (series-connected)
source-impedance RQ. The load is given by the loudspeaker-impedance RL (Fig. 10.6.1), and
both source- and load-impedance taken to be purely ohmic for our first investigations.
Fig. 10.6.1: AC voltage-source with load-impedance; with & without an ideal matching transformer.
The transformer shown here is of ideal characteristics, and completely described by the two
equations given above; is the turns-ratio, also termed transformer-ratio. The
windings shown in the schematic therefore must not be interpreted as inductances but have a
purely symbolic character. The idealization mentioned above may be in sharp contrast to
reality: the ideal transformer can transmit DC – something impossible for a real transformer.
For our first forays into transformer-land, this discrepancy is not a problem – we can (and will
have to) expand the model as needed. According to the idealization, the transformer is also
loss-less: U1 ! I1 = U2 ! I2. In the interior, energy is not stored, nor dissipated into heat. This is
another difference to the real transformer: its windings do generate heat – which is not (yet)
considered in this simple model. The latter is not able to simulate the non-linearity (magnetic
hysteresis) caused by the iron core, and the same holds for winding capacitances and leakage
flux. All these specific characteristics will need to be incorporated in a realistic model, and we
can already now anticipate how complex this is likely to become.

The power matching, on the other hand, may very well be shown using the ideal transformer:
the source (voltage-source with source impedance) “sees” as load the input-impedance RE of
the output transformer (OT):
Impedance-transformation
The secondary load-impedance (RL) is mapped (transformed) via the OT into the primary
input-impedance of the OT. If this input-impedance RE is very small relative to RQ, the major
part of the power is fed to RQ, and not to RL. Conversely, if RQ is large, almost all power is fed
to RL, but due to P ~ 1/ RE, this power becomes smaller as RQ becomes larger. Therefore,
equal internal- and load-impedance is often sought as an optimum for matching: RQ = RE.
With internal impedance and load-impedance known, the transformer-ratio can easily be
calculated from this simple condition: . Given RQ = 7200 " and RL = 8 ", we
get, for example, a transformer-ratio (turns-ratio) of TR= 30 (tube amplifiers Chapter. 10.6.2).
So, how exactly does the output transformer accomplish this transformation, how does it
generate the secondary quantities from the primary ones? This is done via the magnetic
coupling of two windings the turns-ratio of which corresponds to the transformer-ratio TR.
The primary current I1 flowing through the primary coil generates a magnetic field that, in an
ideal transformer, entirely permeates the secondary winding and induces the secondary
voltage U2. If the transformer has a load coupled to its secondary winding (as it is normally
the case), there is also a current in the secondary circuit that itself generates a magnetic field
entirely permeating the primary winding (in the ideal transformer) and inducing a voltage
there. Both coupled processes (current # field # voltage) can and need to be superimposed;
this is the basis for the calculation of the general case [4, 7, 17, 18, 20]. However, a wire
configured as a winding needs to be represented in the equivalent circuit diagram (ECD) at
least via a resistor (copper-resistance) and an inductance (magnetic field) – which leads to a
first extension of the ideal transformer-schematic. Since the magnetic coupling of the two
windings is an indispensable basis, it needs to find its way into the transformer-ECD, too.
How this ECD is derived from the physical interrelations shall not be elaborated here
explicitly – extensive literature already exists for this (see above). Basically, the real
transformer can be represented by a special ideal transformer and several supplementary two-
poles. The special ideal transformer is fully described by its transformation ratio TR, and what
has been stated in Fig. 10.6.1 does hold for it. The supplemental two-poles approximately
model the characteristics in which the real transformer differs from the ideal one. Still: these
are approximations the applicability of which needs to be checked in each individual case.
The most important characteristics modeled by the supplemental two-poles are: resistive
losses, inductances, and flux-leakage. Losses are due to the copper wire and the magnetic
core, inductances result from (coupled) windings, and flux-leakage happens because, in the
real transformer, not the whole magnetic flux generated by one winding permeates the second
winding, but a part misses it. The leakage-factor ! defines the extent of the flux-leakages;
alternatively, the coupling-factor can be given. A leakage-factor of ! = 0%
corresponds to complete coupling (= ideal tight coupling), while a leakage-factor of 100%
indicates non-coupled windings. There are different equivalent circuit diagrams; the
individual factor TR may deviate from the physical turns-ratio.

Two of the most important ECD’s are shown in Fig. 10.6.2. R1 and R2 represent the ohmic
components of the winding-impedances and model the copper-resistances. L1 and L2 are the
inductances of the primary and the secondary windings, respectively. For a secondary open-
circuit, the measurement of the primary input-impedance yields R1 + j"L1. For a primary
open-circuit, the measurement of the secondary output impedance yields R2 + j"L2. The
inductance designated M in the right-hand ECD is the mutual inductance. The following
relationships hold: , , .
Fig. 10.6.2: ECD’s for transformers. The transformer in the ECD on the left is ideal (and thus free of
inductances). The inductances in the ECD on the right may become negative; this does not restrict the validity.
Besides the three ohmic resistances that can be easily determined from a DC-measurement,
the ECD holds three degrees of freedom: L1, L2, and k. L1 and L2 may be ascertained e.g. via
an impedance-measurement with contra-lateral open circuit. The coupling-factor can be
determined with contra-lateral short-circuit. Measuring the primary DC-resistance R1 of the
OT is most unproblematic, while regarding the secondary resistance we need to bear in mind
that it may by of very small magnitude (possibly R2 < 0.1"). When measuring the inductance,
the fact that the ECD mentioned above has only limited applicability in practice requires
consideration: stray- and winding-capacitances influence the impedance, as well (Fig. 10.6.4).
In both ECD’s given in Fig. 10.6.2, the inductance in the parallel branch will short any DC
voltages – the result is a high-pass. Accordingly, the parallel inductance needs to be as large
as possible in order to allow for low-frequency operation. The inductance rises approximately
with the square of the turns-number of the winding, and therefore a winding with a high turns-
number would be desirable – however, this brings along mounting copper-resistance, and
correspondingly increasing losses. To keep the copper-resistance low, the cross-section of the
deployed wire needs to be large – requiring the dimensions of the transformer to be large, as
well. Simple conclusion: transformers that handle high power and low frequencies need
to be large. For the selection of the cross-section of the wire, the current-density supplies a
first step of orientation: given an RMS primary current of 0.11 A, a 0.2-mm-wire would be
suitable for 3.5 A/mm2. The latter value is just for orientation: for large transformers,
somewhat smaller current-densities will have to be assumed, especially if the surrounding air
is heated up by the tubes. The current I2 flowing in the secondary winding is larger than the
primary current I1 by the factor of TR; however, the secondary turns-number is 1/TR-fold
smaller than the primary turns-number; the product of current-strength and turns-number
therefore is the same for primary and secondary winding. This holds at least for the ideal
transformer – in real transformers there are small deviations that may, however, be
disregarded for a first consideration. Given equal current-densities for primary and secondary
winding, it follows from the equation I1N1 = I2N2 that the cross-sectional areas of the
windings should be equal for both windings. The total cross-sectional area of the winding
(amounting to e.g. 2.2 cm2 for the M55-transformer) therefore is made available with 50%
each to both primary and secondary winding. Depending on the application, transformers
need to meet certain requirements, for example with a proof-voltage of more than 1000 V
(and corresponding supplementary insulation layers), or a special low-capacitance winding
(with different build), or additional taps (requiring more contact wires and thus space). This
shows that transformers may have manufacturer-specific differences that are not obvious at
first glance.

The M55-transformer cited as an example has a winding-surface of 2.2 cm2 i.e. 1.1 cm2 per
winding. This value must, however, not be simply divided by the cross-sectional area of the
wire because wire-insulation and -spacing also require space. Nevertheless, it should just
about be possible to accommodate 2000 turns of 0.2-mm-wire. Applying the current (e.g. 0.11
A) as calculated from the current-density yields a magnetomotive force of 220 A, and a
magnetic field-strength of 1.7 kA/m (as a first-order approximation). From a thermal point-of-
view, this may be o.k. – from a communication engineering point-of-view, it is not: the
materials normally used for cores in transformers are all but “saturated” at such high field-
strengths, and the magnetic flux cannot increase anymore if the field-strength is further
increased. Strong non-linear distortion would be the result. Schröder recommends in Vol. 1 of
his book Elektrische Nachrichtentechnik a maximum magnetic field-strength of 0.1 kA/m.
Consequently the overdrive found in our above example would be massive. Alternatively, the
maximum magnetic flux-density could also be calculated:
Peak value of the magnetic flux-density.

N1 = primary turns-number,
AFe = cross-sectional area of iron.
It is clear from the reciprocal dependency on frequency that, for a primary voltage U1 sourced
form a stiff voltage-source, the flux-density decreases with increasing frequency – therefore
problems may result in particular for low frequencies. We will get back to the non-linear
behavior in Chapter 10.6.4; first, the behavior for small drive-levels is under scrutiny. The
(linear) ECD’s introduced in Fig. 10.6.2 enable us to approximately describe impedances and
transmission behavior of an output transformer. In the higher-frequency region, however,
noticeably deficits remain because capacitive coupling among the windings and iron losses
are not considered yet. Strictly speaking, every differential section of the winding is
capacitively coupled to every other section, but a single substitute capacity is sufficient to
model this infinite number of coupling capacitances. The iron losses (hysteresis- and eddy-
current-losses) may be modeled via a resistor with good approximation, as well, and an
extended equivalent circuit diagram shown in Fig. 10.6.3 represents a good compromise
between complexity and accuracy. Calculations with the approximation TRi $ TR are always
acceptable: the transformers considered here rarely have a leakage-factor of in excess of 1%.
C1 = capacitance of the winding,

L1 = primary inductance,
R1, R2 = copper-resistances,
RFe = iron losses,
LS = leakage inductance.
Fig. 10.6.3: Equivalent circuit diagram of transformer% (linear model). Non-linear behavior: see Chapter. 10.6.4.
Fig. 10.6.4 shows comparisons between measurements and calculations carried out on the
basis of the above model. Since all these transformers are used in push-pull output stages, the
respective primary winding is divided in two halves. Calculation and measurement was
respectively done for one half of the primary winding. For secondary open-loop operation, the
primary impedances of the two winding-halves are practically identical; there are differences
for secondary short-circuit, though – these are due to different coupling of the windings. For
low-impedance loading (i.e. for loudspeaker-loading, as well) the push-pull drive-signal
therefore is not symmetrical anymore in the higher-frequency region.
%
The capacitance may also by connected on parallel to L1; the differences are small.

Fig. 10.6.4: Comparison of impedance measurements (------) and model calculations (–––––), each for one half of
the primary winding (Ra). The two open-loop impedances are practically identical; the short-circuit impedances
differ due to different coupling-factors.
Measurements and calculations in Fig. 10.6.4 are practically identical over a wide range but
there are some sections in which differences become apparent. In principle it would not be
difficult to extend the model by a few further components such that a good correspondence
would be achieved across the whole frequency range. However, in the interest of general
applicability, the ECD as developed above shall remain unchanged. The divergences are
rather limited, anyway.
We can also see from Fig. 10.6.4 that – at least for the transformers investigated here – the
ECD is well suited to model the primary load-impedance (i.e. the strain on the power tubes)
for linear operation. However, output transformers work linearly only for very small output
power, typically P < 1 mW. For your regular output power, the parallel inductance (L1), in
particular, depends very strongly on the drive-level. As simple as the linear equivalent circuit
diagrams are, their applicability still remains strongly limited. For this reason, Chapter 10.6.4
will elaborate more extensively on the non-linear behavior.

10.6.2 Impedance-matching and transmission
Frequently, the term “impedance matching” is interpreted such that, for a maximum of power-
yield, the source- and the load-impedances need to be equal (or conjugate). The datasheet of
the power-tetrode 6L6-GC lists an internal impedance of 35 k! so that we could conclude
that the primary impedance of the output transformer should also amount to 35 k!. At the
same time, however, the datasheet specifies a so-called “optimum load impedance” at no
more than 1.4 k!. What follows is this: the 6L6-GC is (like all tetrodes%) a high impedance
source and operates approximately as a current source. The power delivered by a current
source is proportional to the load-impedance: the higher the latter the higher the power-yield.
However, this simple relation is limited by three non-linear conditions: the maximum
allowable plate-dissipation, the maximum allowable plate-voltage, and the residual voltage at
the plate. The optimum load-impedance (= external impedance) results from these non-
linear conditions, and not from the equality of internal- and load-impedance. It is sufficient, as
a rule, to assume the internal impedance of the tube to be large relative to the load-impedance;
the optimum load-impedance (per plate) for push-pull stages usually is about 1 – 2 k!.
The output transformer enlarges the secondary load-impedance (typically, this is the
loudspeaker impedance) by the square of the turns-ratio, for example:
An 8-"-load-impedance is transformed – for TR = 12 – into 144 x 8 " = 1152 ".
Usually, there is no need to distinguish between the turns-ratio of the windings TR = N1/N2,
and the transmission ratio TRi in the equivalent circuit diagram, because in most cases the
respective values differ by less than 1% (Fig. 10.6.3). The internal impedance Ri of the tube is
transformed with TR2, as well: the internal impedance of the replacement source driving the
loudspeaker amounts to Ri / TR2 (in the example 35 k" / 144 = 243 "). As long as the power
stage is not overdriven, it will operate the loudspeaker approximately as a stiff current-
source – if the power stage does not involve negative feedback (NFB). The voltage/voltage-
NFB implemented in many amplifiers reduces the internal impedance of the power amplifier.
Still, perfect behavior as a stiff voltage-source is not accomplished by tube power-amps
(however, most transistor power-amplifiers will achieve this – but they are not a object of the
present investigations).
Fig. 10.6.5 shows the family of output characteristics for a power-pentode known from
Chapter 10.5, plus some load-dependent transmission characteristics. Given the secondary
impedance (e.g. 8 !), the slope of the operating characteristic may be changed as needed.
Abb. 10.6.5: Transmission characteristics (left), frequency-response at the 8-"-output for a load of 4/8/16 ".
%
As far as they are not operated in triode-mode (triode-mode: g2 and plate are directly connected).

It may be matched to the family of characteristics discretionarily by varying the transmission

ratio (TR): a larger TR results in a flatter curve for the load-line i.e. a smaller plate-current and
a larger voltage swing.
The internal impedance of the tube transformed via TR2 is, however, not the source
impedance relevant for the loudspeaker across the whole frequency range. The equivalent
circuit diagram presented in Fig. 10.6.2 shows that the parallel inductance L1 determines the
impedance at low frequencies: it shorts the source for low frequencies and has the effect of a
high-pass. Moreover, we need to consider that this inductance is non-linear, and therefore
we do not have a conventional high-pass here (Chapter 10.6.4). The transmission curves given
in Fig. 10.6.4 involve a demagnetized transformer core; however, this can be achieved only at
untypically small drive-levels of about 1 µW. Nobody will play a 45-W-amp at such a small
power level – the tube amp will not be able to shape the sound in the way for which it is
designed. Still, the curves shown in Fig. 10.6.5 had to be measured approximately at this
power level, otherwise the main inductance L1 would have become dependent on drive-level
in a rather unbecoming way. The small-signal ECD so popular in communication engineering
it in a bit of trouble due to this, but it can be rescued by a special modeling at low frequencies
(Chapter 10.6.4). Basically, the parallel inductance looses its impact with rising frequency,
and the transmission becomes frequency-independent (for an ohmic load). At very high
frequencies (that can however barely, if at all, be reproduced by a typical guitar-loudspeaker),
the incomplete field-coupling and the winding-capacitances may start to have an effect – but
in all likelihood this will not be dramatic or noticeable at all.
Power amplifiers are always specified for a real (ohmic) nominal load-impedance although
the impedance of a loudspeaker is always dependent on frequency. For this reason, Fig.
10.6.6 depicts transmission frequency responses for loading with a loudspeaker; the mapping
of the frequency-dependent loudspeaker impedance onto the frequency response is clearly
visible. The power stage of a Super-Reverb normally has negative feedback but for these
measurements it was deactivated – otherwise the characteristics of the output transformer
would have been suppressed too much (operation with negative feedback: Chapter 10.5). The
operation with a loudspeaker results in a treble boost (voice-coil inductance), and between 10
and 100 Hz we observe a narrow-band boost due to the loudspeaker resonance. For both
operational states, attenuation shows up in the bass range for very small drive-levels (P <
1mW): this is due to the main inductance (see also Chapter 10.6.4).
Fig. 10.6.6: Transmission frequency response; transformer with a secondary load of 8 " (left), and loaded with a
real loudspeaker (right). NFB deactivated. 8-"-load yields a voltage level of –20 dBV & P = 1.25 mW.

10.6.3 Winding-capacitances & -asymmetries
In order for the push-pull power-stage to assemble the two half-waves of the signal correctly
with respect to magnitude and phase, the primary windings of the transformer need to be
completely similar. Which of course they are not, because they cannot be located at one and
the same position on the winding-former. If first one primary winding is wound, and then the
second on top of the first, the difference in wire-length is immediately apparent. Furthermore,
measurements in the high-frequency range will reveal differences in the coupling- and
leakage-factors, and in the winding-capacitance. To moderate these problems, the windings
are subdivided (Fig. 10.6.7), and the subsections are alternately wound on top of each other
(or next to each other in multi-chambered transformers).
Fig. 10.6.7: Construction of the

winding. In the interleaved
winding (right), the sub-sections
of different windings alternate. In
transformers with a sophisticated
build, we find multiple “nestings”
of primary and secondary
winding.
In the RL-equivalent-circuit-diagram of the transformer (Fig. 10.6.2), the relative bandwidth

(fH / fT) is inverse to the leakage-factor; with a favorable build of the winding three frequency-
decades can be covered which is sufficient even for HiFi-quality. However, the winding
capacitance must not be completely ignored – in order to describe the high-frequency
transmission characteristic, at least one capacitance is required (e.g. Fig. 10.6.3). It is this
capacitance that determines (together with other parameters) the upper cutoff frequency, and
it is just as important as the stray-inductance. As an example, two transformers were
examined that are both offered for the Fender Tweed Deluxe: the 1750E from Hammond and
the TAD-1839. Fig. 10.6.8 shows the transmission frequency responses measured for loads of
8 ! and 80 ! at the secondary output (with a stiff current source driving one primary
winding). Both transformers show a resonance-emphasis at high frequency: the effect of
stray-inductance and winding-capacitance. Since loudspeakers do not merely represent simple
ohmic resistances (Chapter 11), supplementary measurements were taken with an 80-!-load.
This suddenly revealed serious differences, and consequently specifications at nominal load
are a necessary but insufficient criterion.
Fig. 10.6.8: Frequency response with a stiff current source (0.16 mA) driving one primary winding.

A short diagnosis of Fig. 10.6.8 could read: the Hammond lacks in treble, and the TAD lacks
in bass. That is too simplified, though, and we need to dive a bit more into the details. The
measurements in fact happen at a rather small primary current and, according to Fig. 10.6.6,
the main inductance (see Fig. 10.6.6) is relatively small here. Also, a loudspeaker impedance
of 80 ! is, in reality, not actually reached at high frequencies. Therefore, supplementary
measurements are required with loading by a real loudspeaker. These are shown in Fig.
10.6.9, with a Jensen P12N (mounted in a Deluxe-cabinet) loading the output transformer.
Using a stiff current-source again reveals a slight deficiency of the TAD-transformer in the
bass-region although this becomes less significant as the drive-level increases. The treble-
deficiency of the Hammond-transformer remains relegated to ranges which – for a 12”-
speaker transmitting frequencies up to about 5 kHz – have no practical bearing. Our revised
conclusion therefore is: in the transmission range important for electric guitars, the Hammond
1750E offers a marginal advantage versus the TAD-1839 – this would possibly justify a small
mark-up for the Hammond. Surprise, though: at the time of this writing (AD 2012), TAD
charges a stout 86,20 Euro for the 1839 while the Hammond 1750E sets you back a mere
34,70 Euro at Tube-Town. Both TAD and Tube-Town offer a whole range of further output
transformers; Chapter 10.6.5 includes corresponding measurement results.
Fig. 10.6.9: Frequency responses with loudspeaker-loading: stiff current-source (left), power-stage (right).
20 dBV at 8 " yield => P = 12.5W, P = 10 W corresponds to a voltage level of 19 dBV. At voltage levels
around 20 dB, this 6V6-GT-power-stage already shows significant non-linear distortion.
Figs. 10.6.8-9 show the transmission from one primary winding to the secondary winding –
there are, however, two primary windings that feature different magnetic and capacitive
coupling to the secondary side. Fig. 10.6.10 considers this and shows both transmission
functions. Again, it becomes apparent that an ECD of pure RL-build is not adequate, although
the figure also clarifies that the differences are limited to ranges that are not relevant for guitar
amplifiers.
Fig. 10.6.10: Frequency responses of transmission. Primary stiff current-source; asymmetric primary windings.

10.6.4 The non-linear model
Ampère’s circuital law describes the connection between the magnetic field-strength H and
the electric current I, while the law of induction characterizes the relation between electric
voltage U and magnetic flux-density B. Both laws are time-invariant mappings. The tie-in
between B and H, however, is given by a non-linear, time-variant mapping: . In the
ferromagnetic sheet metals used in transformer cores, the permeability µ is a non-linear
quantity the magnitude of which depends both on the field-strength and on past values
(compare to Chapter 4).
A first indication of this non-linearity of the core emerges when measuring the transformer
impedance. Changing the sinusoidal AC-current flowing through the primary winding of an
output transformer, and concurrently measuring the voltage across this winding, we get a
quotient depending on the current (Fig. 10.6.11). The time-curve of the voltage (or of the
magnetic flux-density) indicates strong non-linearity already at moderate amplitudes, i.e.
there are deviations from the sinusoidal shape resulting from the warping in the hysteresis-
curve (Chapter 4).
Fig. 10.6.11: Measurement at the primary winding (EI-96). The “inductance” given in the section on the left is a
special non-linear quantity. Right: secondary voltage (LL) and flux-density for input from a stiff current source.
The relation between B and H is, however, not just non-linear but in a sense time-variant, as
well: on the one hand there is an infinite number of hysteresis-loops, on the other hand these
can be cycled through only in one direction – for one and the same field-strength there are
two corresponding (different!) flux-densities. Of course, the material in the core reacts in the
same manner each time if we start from the totally demagnetized state: as such the system is
time-invariant. After switching off an external source, however, the core material remains in a
partially or fully magnetized state for any length of time, and as we re-start driving the
material, an individual characteristic results that is dependent on the previous drive-state – as
such there is time-variance. Fig. 10.6.11 includes two curves: the upper was measured with a
fully de-magnetized core while the lower resulted from the core having first been strongly
magnetized by a DC-field that was switched off for the L-measurement – i.e. a degree of
magnetization remained (remanence). Last, we need to consider that small drive-states run
around an offset-point do not follow the large hysteresis curve (see Chapter 4.10.3, reversible
permeability). All these non-linear and time-variant effects give measurements with output
transformers a certain challenge. Moreover, the data of the transformers under scrutiny are, as
a rule, not known and can be (non-destructively) determined only approximately – the curves
shown in the following will therefore include tolerances.

Ferromagnetism is a characteristic of the crystal lattice: the elementary magnets are grouped
as Weiss domains, and in demagnetized ferromagnetic materials the orientations in space of
these domains are randomized i.e. their combined effects on the outside world cancel each
other out. An exterior magnetic field (e.g. caused by an electric current) shifts the borders of
the Weiss domains (Bloch walls), and a polarization results. These wall-shifts (in part
reversible and in part irreversible) depend in strongly non-linear fashion on the magnetic
field-strength – this is the basis for the non-linear electrical behavior. The relation between
field-strength H and flux-density B is shown, for small drive-levels, in Fig. 10.6.12: it is
evident how the hysteresis-loop tilts upright with increasing drive-level, and how
consequently the permeability increases. The right-hand picture indicates the field-strengths
measured with imprinted flux-density: already at small drive-level a deviation in shape
occurs, as does an increasing phase-shift relative to the flux-density curve (dashed line,
sketched in without scaling).
Fig. 10.6.12: Hysteresis loops. Right: time-functions of field-strength measured with imprinted sinusoidal flux-
density; dashed: the time-curve of a flux-density (no scaling).
The imprinted flux-density shown in Fig. 10.6.12 is easily achieved: driving a winding from a
stiff voltage-source results in an imprinted flux% (due to the law of induction). In this mode
of operation, the voltages transferred to the other windings are also sinusoidal with good
approximation – however, this is not the typical case for tube power stages. The latter (as
current sources) imprint a priori the current, and this leads to non-linear distortion in the
voltages across the windings. This mode of operation is depicted in Fig. 10.6.13: already for
relatively small field-strengths, non-linear distortion in the flux occurs, leading (as the
derivative) to distortion in the voltage. This is not crossover-distortion from the tubes, but
pure hysteresis-distortion (imprinting the field-strength works almost distortion-free here).
Fig. 10.6.13: Sinusoidal field-strength H (imprinted via the primary current) and corresp. flux-density B (left);
non-linear distortion in the voltage Uw across the winding resulting from this H and B (right).
%
The voltage-drop across the copper-resistance may be compensated, if necessary.

The curves shown in Fig 10.6.13 were measured at an EI-96-core for a secondary open-loop
circuit. With a load connected to the secondary winding, this kind of non-linearity
increasingly takes a backseat as the frequency rises. If we exclude the transmission of high
frequencies for the time being, the equivalent circuit-diagram (Fig. 10.6.3) may be drastically
simplified: the secondary copper-resistance R2 ($ 0,5 ") is added to the nominal loudspeaker
resistance, and the leakage-inductance may be omitted, just as the winding-capacitance C1.
The model thus has a purely ohmic secondary loading. Transforming this secondary load via
the transformer with TR2, we get – on the primary side – an equivalent load-impedance R' =
ü2!(R2 + RL) connected in parallel to L1. We may take as guide value for this primary load-
impedance about R' = 1 k", as long as we involve one primary winding%. Relative to this
value, the iron-losses (RFe) may be neglected, and only three elements remain in the ECD: the
primary copper-resistance R', the non-linear parallel inductance L1, and the transformed load-
impedance R' (Fig. 10.6.14). The primary current therefore splits up into two parts: the non-
linearly distorted magnetizing-current (through L1), and the current through the load.
Compared to the current through the load, the magnetizing current becomes increasingly
smaller with rising frequency and looses its significance: the non-linear distortion decreases.
Fig. 10.6.14: Equivalent circuit for the transformer (left); two-pole simplification for low frequencies (right).
It has already been mentioned that this parallel inductance is non-linear; therefore, strictly
speaking, no transmission function can be established. The quotient of RMS-source-current
and RMS-output-voltage may still be determined, and it is shown in Fig. 10.6.15 (left-hand
section). In the right-hand section, two peculiarities stand out: the slope is not 20dB/decade,
and the cut-off frequency is drive-level dependent: with increasing drive-level, the low-
frequency response improves. As can be seen, it is not purposeful to determine the main
inductance based on the initial permeability (as it would be called for according to the
classical dimensioning-rule). This approach would land us in the "W-range, which is rather
academic in the world of guitar amps. Rather, one could (and should) orient oneself according
to the saturation-behavior of the core-material, and determine, for high drive-levels, the flux-
density. The saturation of the latter gives hints towards the dominating magnetic distortion.
Fig. 10.6.15: Left: drive-dependent main inductance (––– core demagnetized, ----- with remanence).
Right: drive-dependent non-linear high-pass (fed from a stiff-current-source, core of transformer demagnetized).
The specified power is fed to the ohmic nominal impedance (4 !) at 1 kHz.
%
For both primary windings the quadruple value (not the double) is to be used (Chapter 10.5.5).

Before we occupy ourselves in more detail with the magnet distortions, first a comment
regarding the pre-magnetization and de-magnetization of the core: we must not expect that
the core is always operated free of remanence. At some point, there will be a strong
magnetization (even if it happens only as the switching-on impulse occurs), and from this the
operating point will return to a point on the hysteresis that does not necessarily correspond to
the flux-free origin of the coordinates. Another issue merits attention: only for exactly
corresponding plate-currents will the output transformer in push-pull power-stages not
experience any pre-magnetization. In most case, the plate-currents will be different, and the
resulting difference-current will magnetize the core. Consequently, the main inductance will
become smaller, and the even-order distortions will increase.
For the demagnetized core (!), the hysteresis loops are point-symmetric, and therefore the
distortion spectrum contains only odd-order harmonics. Usually, the 3rd order distortion-
suppression ak3 is stated; given certain circumstances also the 5th harmonic may be evaluated.
The levels of the higher-frequency harmonics are often negligible in comparison. Fig. 10.6.16
shows the 3rd-order distortion-suppression versus the RMS-power (fed to a purely ohmic
nominal impedance). In the power-range important for stage-use (over 0.1 W and over 100
Hz), the distortion-suppression remains above 40 dB i.e. the THD remains below 1%.
Compared to the distortion generated by a tube power-stage, this is not a dominating effect.
Only for lower frequencies and high power output, the transformer distortion rises again
steeply – this, however, will usually be outweighed by tube distortion. Of course, the guitarist
is at liberty to demand a powerful and distortion-free reproduction of the fundamentals of
his/her 7-string guitar. For this scenario, however, a look at loudspeaker-distortion and
loudspeaker frequency-responses (Chapter 11) immediately opens the path towards bass-
amplifiers and –loudspeakers.
Abb. 10.6.16: Distortion-suppression ak3 of a 50W-output-transformer for high-impedance drive-signals and

nominal load. The non-linear distortion is generated exclusively by the transformer and not by the driving
amplifier. The hysteresis loop shows the relation between magnetic field-strength and flux-density (20 Hz).
A summary in short: the output transformer shows several characteristics that distinguish it
from linear, time-invariant components: 1) its main inductance depends on the drive-level; the
deep bass is reproduced weaker as the signal level drops. 2) The harmonic distortion is
frequency- and drive-level-dependent: the lower the frequency and the higher the signal level,
the larger the harmonic distortion; the side-maximum at around 1 mW has little bearing on
guitar amplifiers. 3) Harmonic distortion and bass-reproduction depend on the remanence i.e.
the previous history of the core-magnetization. 4) How equal (or unequal) the bias-current in
the power tubes is, determines the amount of even-numbered distortion components – the
matching of the power-tubes is a critical factor here.

The reason for the strange behavior of the output transformer is its warped transmission
characteristic. Each of the two transformer windings% may be assigned a current and a voltage
that are mapped onto each other via transformer and load-impedance. This is classical
systems-theory: systems map signals onto each other [7]. If a system always reacts the same
way, it is time-invariant; if principles of superposition and proportionality hold, and if the
system is source-free, it is linear. The transformer is neither – nor. The following
considerations concentrate on two (of the four) signal quantities; in a transformer this could
be input-current and output-voltage. The nomenclature of mathematical analysis likes to
denote the input quantity x and the output quantity y. A so-called “linear function” is defined
via y = 5# x + 3. From the point of view of systems-theory, the corresponding system is,
however, not linear because “source-free”-condition is not adhered to, among others aspects:
in a linear system y = 0 has to follow for x = 0. A further term needs to be introduced for the
consideration of functional dependencies: in a memory-free system, the output quantity (y)
may, at each and every instant, only depend of the input quantity (x) at that instant. Each pair
of values (xi, yi) may then be seen as a point on the xy-plane. The entirety of all points forms
the graph of the function – this graph is called transmission characteristic in systems theory
(and it is something completely different from the transmission function). The ideal amplifier
features, as transmission characteristic, a straight line traversing the origin. The slope of the
straight line is a measure for the amplification factor. The transmission characteristic of the
tube (Chapter 10.1.3) is, conversely, bent; the tube therefore amplifies in a non-linear fashion.
It is somewhat popular to deduce from this the theorem: “curved transmission characteristics
lead to non-linear distortion” – however things are not that simple.
Let us look at the transmission behavior of a simple RC high-pass. Its elements (R and C) are
linear components, and therefore the transmission behavior needs to be linear. However, as
we plot, for a sinusoidal input-signal, the output quantity versus the input quantity, an ellipse
(Fig. 10.6.17) is generated, i.e. a curved line. On top of that, this curve will change shape if
the input signal is not sinusoidal anymore. From these simple examples alone, we observe:
transmission characteristics are purposeful if the system is memory-free – in dynamic
(memory-containing) systems, there is no static transmission characteristic but, if anything, a
signal dependent function-graph.
Fig. 10.6.17: Transmission characteristic of a linear system (left) and of a non-linear system (center).
For dynamic (memory-containing) systems (right) two drive-levels are depicted.
So, how does that fit with our transformer? Globally viewed, we have a degressive functional
relation between magnetic field-strength (abscissa) and magnetic flux-density (ordinate),
similar to the curve shown in the middle section of Fig. 10.6.17. In addition, the curve splits
into two loop-shaped branches. A family of degressively clinched ellipses is the result (Fig.
10.6.16). Without a doubt this is non-linear, and it is dynamic (memory-including). Still, it is
very different compared to the simple RC high-pass.
%
For the present considerations the primary winding is not subdivided.

The dynamic behavior of the RC high-pass results from recharging processes in the capacitor:
after e.g. a step in the input voltage it takes a while until the capacitor has recharged to the
new voltage%. This “while” (i.e. this delay) leads to phase shifts, and these are the reason why
the straight line becomes an ellipse. In the ferromagnetic iron core of the transformer, the
magnetic flux instantly follows the field-strength, any inertia effects (that in fact exist) do not
play a role at the very low frequencies considered here. The contoured, s-shaped hysteresis-
curve holds for quasi-stationary processes, as well, i.e. for arbitrarily low frequencies.
Fig. 10.6.18: Relationship between magnetic field-strength H and magnetic flux-density B.
The left-hand section of Fig. 10.6.18 shows the B/H-relationship for an initially totally
demagnetized core – both H and B are zero. With increasing field-strength, the flux-density
first follows on a progressively bent curve, and on a degressively bent curve. If – starting
from any one point – the field-strength is now reduced, the corresponding B-value does not
wander back along the curve it followed on the upwards path, but it takes a significantly
flatter backwards-curve (middle section of the figure). If the field-strength oscillates between
two values equal in magnitude, the BH-curve encloses the origin, as shown in the right-hand
section of the figure for four cases. The quotient of B and H (the slope of the curve) is
proportional to the inductance L.
For a very small drive-level, the hysteresis curve has a shallow shape (but is not horizontal),
and the inductance is relatively small. In this range, the B/H-relationship may be described via
two parabolic branches that themselves can be approximated by a flat ellipse (Fig. 10.6.19).
The parabolas result in a non-linear mapping while the ellipse is linear. As the drive-level
increases, the parabolas (or the ellipses) raise themselves up more steeply, and the inductance
increases until, at high drive-level, the core material is increasingly saturated, and the slope of
the curve becomes flatter again. While this non-linear behavior does not seem to be very
complicated, we need to also consider that the loudspeaker-voltage does not depend on the
flux-density B but on the time-derivative of it (U ~ dB / dt). If the drive-signal is not generated
by an ideal voltage- or current-source, both voltage and current will be non-linearly distorted
and shifted in phase, and on top of this the non-linear inductance is dependent on the drive-
level.
Fig. 10.6.19: Approximations using parabola (right) and ellipse (center). Limits of the ellipse-approximation as
saturation sets in (right).
%
Strictly speaking, it takes infinitely long but we do not need to exactly look into this issue here.

The equivalent circuit diagram developed in Fig. 10.6.14 is helpful to understand these linear
and non-linear mappings. For small levels and low frequencies, the main inductance L1
remains relatively small. For constant output power (e.g. 1 "W), the primary current is (due
to U ~ "LI ) inverse to the frequency; in the measurement of the distortion-suppression shown
in Fig. 10.6.16, the current-level therefore needs to drop by 3.5 dB while the frequency is
increased by a factor of 1.5. Since, as a first approximation, the 3rd-order distortion depends
on the drive-signal amplitude according to a square law, the distortion-suppression will
correspondingly increase by 7 dB – this can be measured with good accuracy for small power
levels (e.g. 1 "W). As the power increases (while the frequency is kept constant), the
distortion rises, but at the same time the inductance will, above a certain value of the current,
start to increase (Fig. 10.6.15). As soon as the impedance of this growing inductance has
reached the size of the transformed load-impedance, the distorted magnetizing current looses
significance and the distortion decreases. In Fig. 10.6.16, this is the case at about 1 mW for
the 90-Hz-curve. As the power (or, more precisely, the flux-density) continues to increase, the
range of non-linear flux-limiting is reached at about 1 T – the distortion suddenly increases.
The rather capricious distortion-behavior seen in Fig. 10.6.16 is explained that way, at least as
far as the pure transformer-distortion is concerned. It has already been elaborated elsewhere
that power tubes and loudspeakers will also operate in a non-linear fashion, and that in
particular the loudspeaker impedance may have a strongly non-linear characteristic.
The cause for all non-linear transformer-distortion is found in the non-linear permeability of
the core metal sheets: it is conducive to examine their magnetic parameters more closely. To
guide a magnetic field with low resistance, a material with very high permeability is required:
ferromagnetic material with its main ingredient being iron (ferrum). Unfortunately, iron also
conducts electrical current relatively well, and for this reason eddy currents can develop their
dampening effect at high frequencies without much hindrance (see also Chapter 5.9.2.4). In
order to hamper this, a few percent silicon are mixed into the iron. Already merely including
1% Si, the electrical conductivity can be halved; it even drops to 1/5th with 5% Si. This is
desirable, but the instruction leaflet points to side effects: the saturation limit decreases with
increasing Si-content, and the metal becomes more brittle. According to Heck [21], at more
than 3.5% Si the metal will break when bent cold, and hot-processed sheets contain 4.5% Si at
most. Fig. 10.6.20 shows commutation curves of typical sheet metals for transformer cores.
These curves result as the reversal points of the inner hysteresis curves are connected; they
correspond practically to curves for previously demagnetized material (dashed in Fig.
10.6.18). Including silicon has a further advantage: the permeability at small drive-levels
increases, and the re-magnetization losses decrease (Chapter 4.10.4). The main reason that
the ideal values presented in the datasheets are not reached in practice is found in the
unavoidable butt joints: due to the very big difference in permeability between air and core-
sheet, even very short air gaps (0.1. mm) deteriorate the magnetic resistance.
Fig. 10.6.20: Magnetic commutation curves of various core metal sheets; impact of the butt joints.

If the core laminations are reciprocally layered – as it is indicated in Fig. 10.6.20 – there will
be 4 overlapped butt joints per magnetic circuit in an EI-core. At each butt joint, the flux
density in the neighboring sheet is doubled, and the saturation limit consequently decreases.
For the example in the picture, an effective gap-width of 0.2 mm was assumed; the geometric
gap-width is even smaller. What’s clear here: a sloppy manufacturing process can quickly
cancel out any advantage that low-loss core sheets may bring.
How big then are these core-losses, anyway? For , the datasheets specify a power
dissipation of 1 – 2 W/kg, i.e. 0.5 – 1 W for your regular 18-W-transformer (500 gFe). This is
for 50 Hz. The often-voiced fear that these re-magnetization losses would rise proportionally
with frequency (because the hysteresis loop is traversed more often as the frequency
increases) fortunately is incorrect: the voltage is approximately constant vs. the frequency%,
and therefore the drive-level decreases with increasing frequency. Besides, if a transformer
was to ‘loose’ 1 W at 50 Hz, it would have to ‘loose’ 200 W at 10 kHz. No – while these
losses do exist (in one transformer somewhat more, in the other somewhat less pronounced),
they are not creating any existential danger. It is therefore not necessary, either, to use NiFe-
sheets with the 20-fold price tag. Already 50 years ago, H. Schröder wrote: time and again it
shows that, for transformers that need to transmit high power, it does not lead anywhere to
use materials with high permeability such as permalloy or permenorm. These materials are
much too easily overdriven [Lit.]. That’s not entirely wrong but requires a supplement:
permalloy is a NiFe-alloy with 70 – 81 % nickel-content. It allows for very high permeability
values but has a rather meager saturated flux density of 0.8 T. In permenorm (as mentioned
by Schröder), the nickel content is lower (36%) and the saturated flux density higher (1.4 T).
These days, 50%-NiFi-alloys reach as much as 1.6 T – almost as good as FeSi-sheets (2 T).
The saturated flux density is often connected to the maximum power that can be transmitted
– unjustly so in most cases, as the following example will show: the primary winding is
connected to a voltage-source, the secondary winding is without load (open circuit), and the
primary current mostly depends on the main inductance. We now connect a secondary load-
impedance (purely ohmic), and the primary current increases. The smaller the secondary load,
the higher the primary current: the more the hysteresis curve is pushed? Given Ampère’s law,
isn’t that correct? In fact, it isn’t: the now flowing higher secondary current generates a
magnetic field, as well, and this one is oriented in the opposite direction of the primary field
(Chapter 10.7.6). The core-drive depends on: voltage, frequency, and inductance $ % U / "L.
In the power stage, the maximum amplitude of the voltage is determined by power supply,
and by the tubes – it is, as a first approximation, constant. Given this, and a specific frequency
(e.g. 100 Hz), the drive-level in the core is halved as the permeability is doubled. Relative to
FeSi sheet metals, datasheets specify a 10 – 20-fold higher permeability for NiFe-sheets – a
slightly smaller maximum flux density would not be of any bother here, would it? Indeed it
wouldn’t – if the core actually had such a high permeability. However, the larger the
permeability of the material, the more the unavoidable air gaps make themselves felt. NiFe
sheet metals are therefore purposeful predominantly for tape-cores. According to Boll, EI-
cores are almost exclusively fabricated from FeSi-sheets, and M-cores in small number from
NiFe-sheets. In the end, an optimization is required that considers, apart from permeability
and saturation flux density, also iron-losses, build-size and – especially – cost. Whether a core
costs 7 Euro or 100 Euro is crucial. If there is too much distortion, a slightly larger FeSi-core
should also be considered (instead of the NiFe-core). It would be far less pricey. At the time
of this writing (2012), sheet metals with high nickel content cost about 60 Euro per kg – given
a minimum purchase of 50 kg.
%
It’s not perfectly independent of frequency, but U ~ f certainly does not hold, either.

Another alternative approach that may be taken is found in grain-oriented transformer

sheets. Applying special milling and annealing, these sheet metals receive a preferred
orientation (texture); they are anisotropic. In a specified direction, their permeability is
higher than that in isotropic SiFe-sheets, and the re-magnetization losses are correspondingly
smaller. In tape-wound cores and split-tape cores, this advantage takes full effect. In EI- and
M-cores, the additional price needs to be carefully weighed against the quality-increase
because here the magnetic flux will in places run transversely to the preferred orientation. Fig.
10.6.21 contrasts hysteresis curves as published by the manufacturer of base-materials with
measured curves. The shapes do not match exactly for a number of reasons: 1) stamping will
deteriorate the material properties at the stamping-edges; 2) The butt joints (unavoidable in
EI-cores) decrease the maximum magnetic flux; 3) with grain-oriented sheets (M165-35S),
the flux is oriented in unfavorable directions also, e.g. transverse to the preferred orientation.
It is rather striking here that the data of the base-materials are not achieved.
Fig. 10.6.21: Material characteristics (Waasner, left), measurements (EI96a, right). The material characteristics
are valid for the base-materials; stamping will change the values; for the influence of butt joints: see Fig. 10.6.20.
Fig. 10.6.22 shows how big the orientation dependency in grain-oriented transformer sheets
is: at an angle of 60° and 90° we obtain curves as they would result for regular, non-grain-
oriented sheet metal. It is consequently not surprising that the good values featured by the
base material are not achievable with EI-cores – even with meticulous assembly. All too
easily the impression could be created that the air-gap between the E and I of an EI-core (Fig.
10.7.14) could be avoided if both these sheets were only pressed together tightly enough.
However, these are non-planar, non-parallel surfaces that meet. The boundary surfaces result
from stamping, and they are slightly arched such that even with peak compression, gaps
remain. The datasheets have info about which tolerances are desirable: 5 "m are seen as good
quality; this is a value that cannot be achieved with stuck-together EI-sheets. Even for split-
tape cores, this could only be obtained with optimum bracing – the long-term sustaining of
which is not at all trivial.
Fig. 10.6.22: Magnetization curves: grain-oriented sheets (left), isotropic sheets (right); base material.

We manufactured three transformers using the three core sheets mentioned above with 900
turns on the primary winding and 79 turns each on the respective two secondary windings%.
The inductance ((U/"I, measured via the primary RMS-current) is shown in Fig. 10.6.23:
although the grain-oriented sheet metal does not reach the nominal data of the base-material,
it still clearly outperforms the isotropic sheets. It is, however, also significantly more
expensive. As an effect of the enlarged inductance, we obtain a smaller harmonic distortion,
as depicted in the right–hand part of the figure. In the budget-priced M530-50A, and at 80 Hz
and 50 W, the THD is four times that found in the M165-35S. Before we elect a favorite,
though, it is wise to take a look at Chapter 11.6: the non-linearity of regular guitar
loudspeakers is much higher that that of the transformers examined here.
Fig. 10.6.23: Inductance of one primary winding (N = 900, RMS current); distortion-suppression at P = 50 W.
M330-50A and M530-50A are isotropic FeSi-sheets, M165-35S is a grain-oriented FeSi-sheet. EI-96a.
Besides the harmonic distortion, the frequency response is of course also of interest – the
windings% were not nested, after all – so according to popular HiFi-lore no usable outcome
could be expected. Fig. 10.6.24 shows, however, how viable the result turned out to be. The
transformer was connected to a secondary load of 8 !, and for each measurement one of the
two primary windings was driven via an internal impedance of 8 k!. Nesting the windings
will drive up cost, and make the filling factor of the copper drop. The Cu-resistances of the
transformer investigated here are Raa = 53 ", and 0.17 " for the 8-"-winding. This is not bad
at all, compared to the industrial products examined in Chapter 10.6.5, the Cu-resistances of
which are two to three times as high, with correspondingly higher thermal copper-losses. The
iron losses cause few problems: for the investigated EI96-transformers, we found as little as
1.2 W (M350) and 0.55 W (M165) at 1 kHz and 50 W. As expected, the grain-oriented sheets
win out – but the advantage is, absolutely taken, insignificant. Simple conclusion: in a guitar
amplifier, expensive core sheets have a hard time pushing their advantages. The M330-sheet
represents a good compromise.
Fig. 10.6.24: Frequency response for an 8-!-load. Primary drive via 8 k!, P = 1/4 W.
Both secondary windings (1 mm ') are connected in parallel, EI-96a core, core sheet M165-35S.
%
Since no 1,5-mm-wire was at hand, 2 secondary windings were set up using 1-mm-wire.

10.6.5 Comparison measurements
From 2012 to 2016, the university at Regensburg (Germany) offered a practical course on
tube amplifiers for which a modular guitar amplifier was developed. It included a 15-W-
power-stage with the possibility to directly switch between up to 10 different output
transformers, and a 50-W-power-amp offering a choice between 13 OT’s. The candidates are:
Transformer Zaa / k" Raa= / " R 8" = / " Core Amplifier !

1 Conrad ELA 10W 7,9 280 0,70 EI-48/16 Ela 6,90
2 Conrad ELA 20 W 7,0 180 0,33 EI-48/24 Ela 9,50
3 Hammond-1750E 8,8 300 0,45 EI-57/19 Deluxe Tweed 34,70
4 TAD-1839 9,1 560 0,70 EI-66/22 Deluxe Tweed 86,20
5 TAD-125A1A 6,9 330 0,44 EI-66/22 Deluxe Reverb 69,00
6 Hammond-1760H 5,9 400 0,83 EI-66/22 Deluxe 'upgrade' 54,39
7 Hammond-1750J 8,2 180 0,35 EI-75/24 Tremolux 38,65
8 TAD-MJTM18WA 9,1 670 0,60 EI-75/24 Marshall 18Watt 79,00
9 Hammond-1750Y 6,8 300 0,50 EI-75/38 VOX AC15 77,30
10 NSC 401318-T 7,1 196 0,50 EI-66/22 e.g. Fender 17,80
11 TT-SLO50 4,5 100 0,43 EI-96/40 Soldano 50W 88,90
12 Hammond-1760L 4,1 100 0,41 EI-96/31 Bassman 'upgrade' 82,30
13 Marshall JTM-50 3,5 86 0,54 EI-96/40 Marshall 50W 86,56
14 Hammond-1750N 3,5 80 0,51 EI-96/40 JCM800 77,50
15 OTH M330-50A 3,5 53 0,17 EI-96/36 university lab --
16 Hammond-1750V 4,2 140 0,70 EI-96/40 VOX AC30 86,50
17 Hammond-1750Q 7,9 140 0,61 EI-96/40 JTM-45 92,25
18 Marshall JTM-45 7,8 155 0,42 EI-96/40 JTM-45 100,30
19 IG-Wickeltechnik 8,2 218 0,49 EI-96/40 JTM-45 106,20
20 Toroid mains transf. 3,5 60 0,21 '81x35 Mains transformer 15,--
21 TAD-MJTM45A 8,1 360 0,49 EI-96/40 JTM-45 129,50
22 TAD-018343 4,7 100 0,20 EI-96/34 Super Reverb 110,00
23 TAD-M50A 3,7 150 0,48 EI-96/40 Marshall 50W 89,90
The ‘small’ transformers (upper group) are operated at either 2xEL84, or 2x6V6-GC while
the ‘big’ ones work with either 2xEL34, or 2x6L6-GC, or 2xKT-66. Using the easily
accessible datasheets as a basis, the optimum load-impedance (plate-to-plate, Zaa) across the
entire primary winding should amount to 8 k" for both the 2xEL84- and the 2x6V6-GT-
complement. Checking a bit more thoroughly, we find as a boundary condition e.g. for the
6V6-GC: a plate- and screen-grid-voltage of 285 V. However, the Deluxe in fact was operated
already in its initial versions at 350 V, and later with as much as 420 V. This slight !
overload has not killed it (the datasheet allow for a maximum of Ua = 315 V) … but what
about the optimum load-impedance at these voltages? The datasheets are silent about it –
presumably because of the limit value mentioned above. These days, transformers produced
for these amps mostly have about 8 k! for the early Deluxe-variants and 6.7 k! for the later
ones. The measurements in the table indicate that these target specifications are ‘generously’
interpreted. For the ‘big’ amps there is agreement that the correct load-impedance for a JTM-
45 should be exactly 8000 ! … that does not prevent TAD to include a 3,7-k!-transformer
with the JTM-45-kit. Well, you are free to reorder the 8-k!-varaint for an extra 130 Euro.
Over to the 2x6L6-GC or 2xEL34: here, impedance-values of around 4 k! are customary, and
you are in good hands with this for the AC30 (4xEL84), as well. It is recommended to take
the impedance specifications with a pinch of salt – they are frequency-dependent, and the tube
data that are supposed to be a match to these impedance values scatter rather strongly, too.

Fig. 10.6.25 shows the measured frequency responses of the impedance. The transformers
were loaded at their 8-!-output with 8 !, and primary impedance of the entire winding (Zaa)
was measured. The Tremolux-OT (7) is actually specified for 4 k" / 4 ", and it was tested
with 8 ! at its 4-!-output, which approximately doubles the primary impedance. The two
ELA-transformers were not actually specified for operation with a push-pull power stage but
their windings allow for comparable transmission ratios. Still, it needs to be emphasized that
these transformers were designed for an operation with 100 V and not for 250 V as it
regularly occurs with power stages (Ua, under regular operation). Corresponding experiments
therefore require adequate safeguarding. All measurements were taken with very small power
such that, for the low-frequency impedance, the initial permeability is significant. The latter
is particularly small for the ELA-transformers; but this was to be expected in the face of the
very small build-size. Also, it must not be forgotten that the other transformers are about 10
times the price! The impedance increase at high frequencies is due to winding-resonances and
–capacitances, and the scatter in the middle frequency-range is due to differences in the
transformation ratio (turns-ratio).
Fig. 10.6.25: Frequency response of the impedance (Zaa) for drive from a stiff current-source (10 "A) and a
secondary load of 8 !. The numbers in the figure relate to the above table.
It is not imperative to assume that the different transformation ratios result from bad
manufacturing quality. The number of the turns of the wire can easily and precisely be
checked; divergences are, with high probability, intentional. The suppliers indicate e.g. Zaa =
8.1 k", but apparently a result of 9 k" will not be the end of the world. What seems to be
more important: manufactured according to the original specs using authentic materials.
That’s the reason for the high price. For the 5E3-Tweed-Deluxe, you will find a vast variety
of output transformers; these all wait to be lovingly assembled by hand (and with authentic
materials) first, and that costs. One single variant for all 18-W-amps would probably also do
– but only for the very un-emotional customer.

Because a purely ohmic 8-!-load is required but not sufficient, the corresponding figures with
loading by a loudspeaker are also included (Fig. 10.6.26). As already elaborated in Chapter
10.5.8, the (straight) load line is a first approach – reality is more complex (in the true sense
of the term). The power tube does not “see” a constant resistance but a complex load the
magnitude of which varies between e.g. 7 and 30 k!. This could as well be a range from 9 to
50 k! – or whatever else the transformer offers as a load. Depending on the transformer and
the loudspeaker, the optimum operational range of the amplifier therefore resides within
different frequency ranges, and consequently, the output transformer influences the sound.
Again: this ain’t no secret science: with the turns numbers, and the size of the core, you have
the main ingredients already on the table.
Fig. 10.6.26: Frequency responses of the impedance (Zaa) with drive from a stiff current source (10 µA),
load = Jensen C12N in an enclosure. Left: first-group transformers (1 – 10); dashed = Conrad-ELA-transformer.
Right: 50-W-OT’s of the second group; dashed = JTM-45-transformer (8 k").
The impedance graphs give an impression of the strain on the tubes; more important,
however, is the power transmission (Fig. 10.6.27). At small output power (0.2 W / 1 kHz), the
variation within the transformers is not that big anymore; even the ELA-transformers provide
sufficient bass-reproduction. In the range of the power-limit (right-hand section of the figure),
however, differences show up, after all. No. 6 (Hammond 1760-H) has the smallest primary
impedance (5.9 k!) and therefore delivers the highest output power in the frequency ranges
where the loudspeaker is of high impedance. The opposite is represented by No. 4 (the TAD
Tweed-Deluxe-transformer): its forte is in the area of low speaker impedance i.e. in the
middle frequency-range. At small and medium output power, the sound can be shaped via
filters almost at will. However, if the power stage is operated in the range of its power limit,
we find: for a brilliant sound, the output transformer should show low primary
impedance, and for a more mid-range-y sound, it should feature higher impedance.
Fig. 10.6.27: Transmission from the phase-inverter input (NFB disabled) to the loudspeaker (P12N).
Normalized to 1 kHz, small drive-level (left), high drive-level with power-stage overdrive (right).

In Fig. 10.6.28 (not normalized), the effect of the transformer establishes itself clearly; No. 6,
with small primary impedance conversely generates the largest secondary source-impedance.
This is why the loudspeaker impedance maps itself relatively strongly onto the transmission
frequency-response. If the internal impedance of the power stage were zero (ideal current
source), the figure would show a horizontal straight line. Relative to this theoretical “ideal”
situation (that for a guitar amp would generally be held as not ideal), the source impedances
in the figure increase with the sequence 4-3-9-6. The compression of the curves follows
almost the same sequence; it is only No. 4 that gets out of line: the TAD-transformer offered
for the Tweed-Deluxe (4) has the highest DC-resistance and therefore somewhat higher
copper losses. To compensate, it is the most expensive one of them all. And who knows:
maybe it is the most authentic one, as well.
Regarding the strain on the power tubes, the following holds: the higher the primary
impedance, the more the screen grid is likely to be overloaded (Chapter 10.5.9). Thus, if you
run your 4-!-amp into a 16-!-speaker, better keep a watchful eye on your power tubes.
Fig. 10.6.28: Transmission from phase-inverter input (NFB disabled) to loudspeaker (P12N).
This figure is reserved for the printed version of this book.
Let us take another look at the differences between linear and non-linear operation. Between
power tube and loudspeaker, there is no tone-stack – if the power tube is clipping, only the
output transformer is left to have any impact on the transmission behavior. Therefore, the
transformer-ratio is important to the sound. The tube power stage has a relatively high output
impedance. If it were as small as it is in a transistor power stage, the output power would
increase as the load-impedance decreases. Conversely, the output power increases, in a tube
amp, as the load-impedance increases. This will not be the case without limit, though – at
some point, the tube hits its limit and then the situation reverses. We see this in Fig. 10.6.29
for the Tremolux-transformer (7), while Fig. 10.6.30 gives on overview over the remaining
measurement results.
Fig. 10.6.29: Output power dependent on the load-impedance for various drive-levels. At the dashed line, the
power-stage overdrive starts. Power stage without negative feedback. 1 kHz.

Fig. 10.6.30a: Maximum power vs. (ohmic) load-impedance; power stage overdriven by 14 dB, 1000 Hz.
Two different 6V6-GC pairs: Ultron (––––), TAD (-----); RK = 270 " // 250 µF.

Fig. 10.6.30b: Maximum power vs. (ohmic) load-impedance; power stage overdriven by 14 dB, 1000 Hz.
Three different EL-84 pairs: JJ (upper curve), Ultron (------), TAD (lower curve); RK = 120 " // 250 µF.

The preceding diagrams indicated that output transformers may result in different operational
behavior – even if offered for the same amplifier model. One parameter in this context is the
frequency response under full load (Fig. 10.6.28), another is the maximum power (Fig.
10.6.30), and a third is the harmonic distortion. The Hammond transformer investigated in
these measurements works (in conjunction with Ultron 6V6-G) most efficiently at a load of
12 !. This result is documented in Fig. 10.6.31, as well. However, using other tubes, different
values were obtained which again shows that the cooperation of several components
determines the transmission behavior of the power stage.
Fig. 10.6.31: Power stage: distortion-suppression for different load-impedances: Hammond Deluxe upgrade.
A distortion-suppression of 20 dB corresponds to a harmonic distortion of k = 10%.
Fig. 10.6.32 shows the measurement results for two different tubes (6V6-GT, EL-84) and two
different cathode circuits. EL-84 with RK is typical for the VOX AC-15 and the 18-W-
Marshall (Model 1958). EL-84 without RK reflects the Mesa/Boogie Studio-22, and 6V6-GT
with RK corresponds to e.g. the Tweed Deluxe. 6V6-GT without RK is exemplified in the
Deluxe Reverb. Because of their differing turns-ratios, the transformers exert a different load
onto the power tubes and generate different distortion-suppression that way. This is for 1 kHz,
though! Here, another parameter enters the scene: the frequency. This now is the point where
the depictions start to become confusing. It shall be mentioned only in passing that on top of
everything, the plate-voltage, the screen-grid resistor, and the phase-inverter may also vary.
Fig. 10.6.32: Distortion-suppression: power stage with different output transformers. On the respective upper
right the cathode circuit is indicated (common RK bridged by 250-"F-capacitor, and fixed bias, respectively). 8 !.

In all these transformer-measurements, only the turns-ratio has shown itself as relevant so far.
However, distortion measurements in the low-frequency range redirect the attention to the
main inductance, or the core-material and –size. If the turns-number is too small, the bass-
reproduction becomes weak and distorted. An increase in the turns-number, however, can
only be achieved (due to the limited space for the winding) by reducing the diameter of the
wire, in turn increasing the copper-resistance. If the secondary copper-resistance amounts to
0.83 ! (as it is the case in the Hammond 1760H), 10% of the generated power remains in the
secondary winding. Approximately the same percentage will again be dissipated in the
primary winding. If both high efficiency and good bass-response are the objective, only
changing to a better core will help, resulting in higher weight and/or price. No magic here: the
small Conrad-ELA-transformer (1) features merely a cross-section of the iron of 2.4 cm2 in its
small core, and no attention was given to achieving a minimum air gap, either. The result can
be seen in Fig. 10.6.33: very strong distortion in the bass. With its proud 8.7 cm2, the AC-15-
transformer of course has a much easier life here. It is no contradiction that the 20-W-
transformer (2) is even worse than transformer (1): (2) is of particularly low impedance and
therefore has an even smaller L. Again: these are ELA-transformers!
Fig. 10.6.33: Left: distortion-suppression in the power stage for different OT’s. 8-!-load at the 8-!-output.
Right: distortion-suppression of the OT (without power stage) as a function of frequency.
The distortion shown in the left-hand section of Fig. 10.6.33 is generated in part by the output
transformer and in part by the power tubes, while the distortion shown in the right-hand
section stems from the output transformer only. The ELA-transformers experience strong
overdrive at low frequencies and are good for distortion sounds, if anything at all. Their
primary impedance in the mid-frequency range is a good match for the tubes but their
inductance is too small. However, all other transformers are suitable for guitar amplifiers,
whether they cost 18 or 86 Euro. Unless blatant errors are made, the following theorem holds:
in the frequency range important for the electric guitar, the turns-ratio (i.e. the primary
impedance Zaa) is the decisive parameter; everything else is of minor importance. Indeed,
the manufacturers do use different core sheets, and, yes, they do invest much time in
“authentic” replicas. They procure old (i.e. outdated) insulation paper, search for wire
insulated in an antiquated fashion, copy scary nesting for the winding, and of course they
need to be royally remunerated for the whole hoopla – it is, after all, almost one-off
production. Mindless reproduction of outdated technology on the basis of misunderstood
context? Yes, for the odd transformer this impression does force itself. However, let’s not
take such a narrow view. Maybe we should consider the approach of the placebo-
pharmacologist: where there’s a will, there’s a market. So: at the latest as we have outgrown
our 18-W-shoes and have dragged our 30-W-whopper to the stage despite the slipped disc, we
sigh contently: two really fat transformers, thus really fat sound.

We have given the ‘small’ 18-W-transformers a lot of space – almost too much since more
diagrams do not necessarily mean more clarity. Therefore, a short description shall suffice for
the ‘big’ 50-W-transformers. The frequency responses have already been shown – maximum
power and distortion have similar characteristics as with the 18-W-OT’s, just with a higher
power level. Overall, the quality is somewhat higher, because for the larger transformers the
inductance-determining relationship iron-surface-to-iron-length is more favorable. All 50-W-
transformers investigated here perform well, whether they have 3.5 k! or 4.7 k! (Raa each).
Supplementing Fig. 10.6.26, Fig. 10.6.34 depicts the transmission frequency responses of the
complete power stage employing EL34’s. It is clear that, in the frequency range important for
the electric guitar, all transformers work almost equally well%.
Fig. 10.6.34: Transmission from the phase-inverter input to the loudspeaker (Vintage-30 in enclosure).
Right: transformers No. 11 – 16; left: transformers No. 17 – 19.
We see larger differences for the non-linear distortion (Fig. 10.6.35). As will be elaborated
later, the 8-k!-transformer is unsuitable for a power stage deploying EL34’s. All other
transformers show a similar behavior at and above 90 Hz; it takes a backseat compared to the
effect of the tubes. The self-wound M330-50A (compare to Fig. 10.6.23) was a first foray into
building an OT – it is suitable, as well. With the addition of 10% more turns, this transformer
could have been brought into the range of the other transformers (there would be enough
space even with the same wire diameter) – however, this step was not deemed necessary. The
red curve refers to a very special “output transformer”: a mains transformer. Indeed, this
works, as well! Not with just any mains transformer – we needed to look around a bit, but this
one fits the bill. It’s a toroidal transformer costing all of 15 Euro – is smaller, more efficient,
lighter by 1.5 kg, and much less expensive (due to large-scale manufacture). Why do we then
still need an EI96? Maybe because it has been done like that for more than 60 years? And
because micro-entrepreneurs do like to make the odd 100 – 300 Euro …
Fig. 10.6.35: Distortion-suppression. Left: whole power stage with 10 different OT’s; right: OT’s only; blue:
M30-50A; red: mains transformer as OT.
%
The difference 3.5 k" vs. 8 k" will be discussed later.

Mains transformers are optimized to achieve optimum efficiency – exotic issues such as
“harmonic distortion” are of zero interest in this area. You just wind thick wire and decently
drive the core in order to find a good compromise between power-losses and cost. The
copper-resistance of the 9-V-winding – which we abused as 8-!-winding – reads only 0.2 !,
compared to 0.4 to 0.7 ! for the real OT’s. Given a few more turns, the main inductance
could be increased without significant deterioration (and especially with next to no additional
cost), and the harmonic distortion could correspondingly be lowered. This is not intended as a
general call: “guys ‘n’ gals, just load your power stage with a low-cost mains transformer”,
but it means to say that, given the correct calculation, and fabricated in industry-correct
quantities, a toroidal transformer can be a small, light-weight and inexpensive alternative.
And what about the frequency response? It’s fully in the green, as shown by Fig. 10.6.34. For
the sake of completeness, the measurements with an ohmic load are shown in Fig. 10.6.36.
All OT’s are perfect – including the mains transformer..
Fig. 10.6.36: Frequency response at an ohmic load (8 !); right: 8-k!-transformers. On the left, the frequency
response of the mains transformer (abused as output transformer) is of course also included.
In order to preempt misunderstandings, here a short afterthought: guitar amplifiers are no

HiFi-systems. The latter require a significantly wider frequency range and a significantly
lower distortion. The message here is not that generally a mains transformer will do as output
transformer, but rather that the high prices of output transformers result from their small
production numbers, and from the more or less authentic replication of out-of-date historic
examples. If authenticity is not the main objective, a mains transformer in the output stage of
a guitar amplifier may be a low-cost alternative to the dedicated special output transformer.
Each of us has to find out (!) on his/her own what is deemed suitable – the expectations vary
too much.
A peculiarity: the special JTM-45-transformers wound – very authentically – to an Raa = 8 k"

specification. This certainly is inappropriate for an EL34-power-stage, but in the original
JTM-45 we do not find EL34’s but two KT-66’s. Do these then require an 8-k"-transformer?
Yes. Or no – it depends on the source. According to Doyle’s Marshall-book, Radiospares
was the first purveyor to the court with their "De Luxe Output Transformer". Radiospares,
however, was not a manufacturer but a distributor (they became RS-Components later). Who
actually manufactured these early transformers is the object of escalating discussions
(allegedly up to five manufacturers may be in the running). The RS-transformer was a typical
universal transformer featuring a choice of several primary impedances: 6.6 k" (with ultra-
linear tap) for EL34 and KT66, and 8 k" or 9 k" for 6L6, 6V6 and EL84. The power stage of
the JTM-45 does not operate in ultra-linear mode; the experts consider 2xKT-66 / Raa = 8 k"
to be the nominal complement. The Drake-transformer used after the RS-transformers
operates with this primary impedance, too. And the GEC-datasheet of the KT-66 (1956), as
well, specifies 8 k", but does this for "cathode-bias" which is not used for the JTM-45.
Fig. 10.6.37: Output power at 8 ", for a distortion-suppression of 30 dB; transformer-numbers acc. to the table.
Fig. 10.6.37 shows the measurement results for a supply-voltage of 400 V. Two EL34 will
yield well over 50 W, given a primary impedance of about 3.5 k". At 8 k", the power output
drops to a meager 15 – 16 W – for sure, this is not optimal. Using two KT-66’s, about 25 W
are achieved with 8 k" impedance, which is about in agreement with the datasheet. We
obtained more power operating our JTM-45 with two KT-66’s and a 3.5-k!-transformer: just
under 50 W with a Russian TungSol-KT-66, significantly less with a TAD-KT-66
(measurement results in Fig. 10.11.3).
Besides the maximum output power, the source impedance shows differences, as well.
Pentodes are of high impedance, and therefore the source impedance of the power stage (the
internal impedance) is relatively high, too. It will be around 100 – 200 ! with two KT-66
cooperating with a 3.5-k!-transformer, but only 40 – 80 ! with an 8-k!-transformer (each at
the 8-!-output with the negative feedback disabled). The effects have already been discussed
several times; they show up e.g. in Fig. 10.6.34.
Well then, it’s getting to be after hours – time to go home for dinner. It’s been quite a while.
You want a recommendation? Because, according to an OECD-study, many readers have
difficulty to hang in there when confronted with longer texts? Ok, here we go:
Loud = 2xEL34 with 3.5-k" output transformer;

Authentic = 2xKT66 with 8-k" output transformer;
Prepared to take a risk = 2xEL34 with (special) mains transformer as output transformer;
Moronic = expensive replacement transformer from faraway lands.
Is that short enough, and intelligible despite three multiplication signs?

You are welcome – happy to comply.
It is not the things that delight us,

but the opinion we have about the things%
%
loosely based on Epiktet

Power stages for the practical course on tube amplifiers
At work in the tube lab with a somewhat disengaged participant

10.7 Power Supply
The power supply delivers the operating-voltages (and -currents) required by the amplifier to
be able to work: plate-voltage, filament-voltage and, if applicable, bias-voltage for the grids.
The most important components of the power supply (mains transformer, rectifier, and filter
capacitor) will be investigated in the following. Power supplies in guitar amplifiers fitted with
tubes generate 500 – 1000 V, and consequently observing pertinent safety regulations is
imperative: touching of live components or wires may be fatal! For this reason, only
trained professionals are allowed to work on such amplifiers. Particular consideration needs to
be given to the fact that even devices that are switched off and disconnected from the mains
power may be storing deadly voltages for hours. Again: such equipment may be opened by
qualified personnel only!
10.7.1 Tube filament
The cathode of a tube will emit the required stream of electrodes only as it glows. A dedicated
secondary winding of the mains transformer delivers the necessary power (2 – 16 W
depending on the tube) for the associated heating. Most tubes are heated with 6.3 V~ , rectifier
tubes with 5.0 V~, as well. DC-heating is possible but uncommon. In order to minimize the
effects of capacitive coupling between filament-circuit and signal-circuits, the heating voltage
often is of symmetric configuration, either via a middle tap in the filament-winding of the
mains transformer, or via two resistors or a potentiometer.
The connections to the tube-filaments are mostly designated with f (from the Latin filum) in
the socket-diagrams; in the actual circuit diagrams, they are not included to keep the drawings
neat. Filaments are positive-temperature-coefficient (PTC) resistors – their resistance
increases by a factor of 7 – 8 when heated. It can therefore be beneficial to a long tube-life to
limit the switch-on current – but this is not mandatory. On the other hand, the filament
voltage should be neither too high nor too low: ±5% is stated as acceptable tolerance and
±10% would already be too much. The reason is that at too high a voltage, part of the cathode
material evaporates, and at too low a voltage, undesirable intermediate layers form.
The filament circuits carry large AC-currents, possibly upwards of 5 A. At a distance of 2 cm

from a wire subject to such a current, we find a magnetic flux-density of 50 !T, i.e. there will
be 100 !T between two wires positioned at a distance of 4 cm. This magnetic field will
induce, into a conductor loop of 3 cm2, a hum-interference of 10 !V at 50 Hz (or 60 Hz,
depending on your geographic location). Such an interference voltage will not be a big
problem in a power stage, bit it might in the preamplifier. The filament supply-wire pairs
therefore usually are installed twisted around each other; the magnetic fields generated by the
individual wires largely compensate each other that way, as do the induced interference
voltages.
Fig. 10.7.1: Filament connections of some selected tube sockets (seen from below). “Pentoden” = pentodes

10.7 Power Supply 10-189
10.7.2 The charging circuit
In tube amplifiers, the typical supply-voltage lies in the range of 200 – 500 V DC. The mains
transformer can only deliver AC and therefore a rectifier is required. Older guitar amplifiers
mostly include a tube rectifier while newer ones often (but not always) sport a silicon
rectifier. The main difference is that a tube rectifier necessitates a filament heating while Si-
rectifiers do not. Moreover, the Si rectifier will generate a voltage drop of about 1 V in the
flow-direction while this will amount to 40 V or more in a tube rectifier.
In most guitar amplifiers, both halves of the sine wave are rectified, this approach being
termed a two-way or full-wave rectifier. The two secondary voltages generated by the
transformer are in opposite phase so that each of the two rectifier-diodes conducts only during
one half-wave (Fig. 10.7.2). However, this does not happen during the complete half-wave
but only close to the maximum voltage, because the supply-voltage generated at the cathodes
of the diodes is smoothened by an electrolytic filter capacitor. From an idealized point of
view, the diode will only conduct if the anode/plate-potential (at the transformer) is higher
than the cathode-potential (at the capacitor). With none of the diodes conducting, the
capacitor-voltage will decrease exponentially: ; here, ! is the time-constant
given by the capacitance and the load-resistor, e.g. ! = 32 µF ! 2000 " = 64 ms. If this time-
constant is large relative to half the cycle-duration, there will be only a small voltage
decrease, and the current through the diode will flow only during a short time (small angle of
current flow). It follows according to the law of charge-conservation, that the peak current
will be the higher the smaller the angle of current flow is.
Fig. 10.7.2: Full-wave rectifier. Voltages and currents for two different angles of current flow.
In strongly simplified terms: if, given a load current of 200 mA, a current is flowing through
the diodes only during 1/5th of the time, then this current needs to be five times as strong as
the load current i.e. 1 A. In reality, load- and diode-currents can only be described with
relatively complicated formulas, but the factor five mentioned here is a good benchmark. If a
power supply is designed for 400 V / 100 mA, a peak current of 1 A may flow through the
diodes. This peak current is specified in extended datasheets – in the abbreviated versions,
however, only an allowable average current value is given. In the above example, this average
is 100 mA per diode. The following table indicates both the peak current î, and the average
IDC for a number of diodes. For the tube diodes, IDC is the load-current (all listed rectifier
tubes are double-diodes), while for the Si-diodes, IDC is the average current per diode. Also
included is the internal impedance of the transformer RTr (per secondary winding).
Together with the capacitance CL and the load resistance RL, this impedance determines the
actual peak current î. If RTr is made too small, of if CL is too large, the rectifier tube may be
overloaded under certain conditions! Normally, RTr cannot easily be changed – transformers
are mostly picked on the basis of their power. If the value of RTr turns out to be too small, the
simple solution is to connect a resistor in series! The given maximum capacitance values are
taken from the datasheet of the manufacturers, and a bit of modesty is called for here: if we
install – in order to further reduce the remaining ripple – a 100-µF-cap instead of the
permitted 32 !F, then the tube will be operated outside of its specifications. Depending on the
quality, it will hold up for some time – or not.

Type Filament UTr / Veff û/V IDC / mA î / mA CL / µF RTr / "

EZ80 (= 6V4) 6,3 / 0,6 2 x 350 1000 90 270 50 2 x 300
5Y3-GT 5,0 / 2,0 2 x 350 1400 125 440 10 2 x 50
EZ81 (= 6CA4) 6,3 / 1,0 2 x 350 1000 150 450 50 2 x 240
5V4-G 5,0 / 2,0 2 x 375 1400 175 525 10 2 x 100
5U4-G 5,0 / 3,0 2 x 450 1550 225 800 32 2 x 75
5AR4 5,0 / 1,9 2 x 450 1700 225 825 40 2 x 140
GZ34 5,0 / 1,9 2 x 350 1500 250 750 60 2 x 75
2 x 450 250 750 60 2 x 125
2 x 550 160 750 60 2 x 175
5U4-GB 5,0 / 3,0 2 x 450 1550 275 1000 40 2 x 67
83 (Hg-vapor) 5,0 / 3,0 2 x 450 1400 250 1000 40 2 x 50
BYX 90 – 2 x 2kV 7500 0,55 A 5A # #
1N 4007 – 2 x 300 1000 1A 10 A # #
BY 133 – 2 x 390 1300 1A 10 A # #
1N 5399 – 2 x 300 1000 1,5 A 10 A # #
1N 5062 – 2 x 240 800 2A 10 A # #
BY 255 – 2 x 390 1300 3A 20 A # #
1N 5408 – 2 x 300 1000 3A 20 A # #
Table: Operational data of mains-rectifiers (from datasheets; please consider manufacturer-specific details!)
For silicon-rectifiers (#), the internal impedance of mains transformers are normally always big enough, and the
maximum load-capacitance does not represent any constraint, either, at (typically) > 200 !F.
The elaborations above have shown that in the charge circuit – between transformer, rectifier
and filter capacitor – a peak current of 1 A can easily flow. Multiplying this value with a
resistance of 1 m" yields a voltage drop of 1 mV. For a 0.5-mm-wire, 1 m" is reached
already with a length of about 1 cm – this merits some consideration: if we contact the
ground-conductor of the charge circuit at two different points that are 1 cm away from each
other, a potential-difference of 1 mV is generated. For an input of high sensitivity, the full-
drive input-voltage is in the same order of magnitude! Of course, the capacitor connections
also have a non-negligible resistance, but here it is only the ripple that is marginally
increased. If, according to the motto “ground is ground”, the input-socket ground is
connected to one point of the filter-cap feed, and the input of the pre-amplifier not to the same
point but to another one off by 1 cm, severe hum-interference is bound to occur.
It is recommended for the wiring of an amplifier to draw up a plan in which all ground-wires
are shown as resistors – this gives a good idea about unwanted voltage drops. By the way,
similar problems may pop up in the secondary circuit of the output transformer, because here,
too, the current may reach several Amperes. Therefore note: connect the output transformer
directly to the output-socket; avoid channeling the loudspeaker current through the amp
chassis.

In the following, measurements taken from amplifier power-supplies will be introduced – a

calculation would be possible, as well, but requires a lot of effort because both the mains
transformer and the rectifier are non-linear components. As a first example, we will look into
the power supply of the TAD-Deluxe kit. It uses the 5Y3-GT as a rectifier tube; according to
the datasheet, it has to make do with a 10-!F-filter-cap. For the measurements, this filter-cap
was loaded with 10, 5 and 3 k" – the corresponding charge-current amounts to 45, 85, and
130 mA, respectively, and the peak current through the diodes amounts to 180, 280, and 380
mA, respectively (Fig. 10.7.3).
Fig. 10.7.3: TAD-Deluxe: voltages and currents for the full-wave tube-rectifier; CL = 10 µF.
As is easily seen, the relationship of peak-current to average current is no bigger than a factor
of 2 – mainly due to the mains transformer at work here. Its secondary windings have a DC-
resistance of 225 ", and thus the voltage breaks down strongly under load, and the angle of
current-flow is relatively large. The thin line in the figure belongs to the open-loop
transformer-voltage; depicted below it is the voltage under load. The voltage across the filter
cap oscillates up and down in the shape of a saw-tooth wave; the forward-voltage of the tube
is highlighted in grey. For the measurements, a rectifier tube of recent production was used; it
shows a voltage drop in flow-direction of 30 – 40 V. In the U/I-diagram, two old datasheet-
curves are entered with a dashed line, and in addition a measurement curve taken with an
RCA NOS tube. NOS = new old stock: this designates tubes that have been remained unused
on the shelf for decades, and which now are deployed for the first time. A tube with a voltage-
drop of more than 100 V in the flow direction will indeed help the amplifier to a different
operational behavior: the supply-voltage collapses even further than shown in Fig. 10.7.3. If
this “sagging” is, in fact, desired, but no NOS-tube is available: a 200-"-resistor will do the
same job.

The curves shown in Fig. 10.7.3 indicate that the data of a tube-type may not be simply taken
from one datasheet. First, there are manufacturer-specific idiosyncrasies, and second, the
production methods were always subject to an ongoing change. On the other hand, the almost
inflationary multitude of designation letters (5Y3-G, -GT) often merely refers to differences
in the glass container and not to electrical data. From this point of view it may be justified to
ask horrendous prices for certain old tubes – tubes from today’s production indeed do have
different data. However: there are reports that it may be very easy to have any desired tube
replicated even in relatively small numbers by the gentleman with the name consisting of
merely two letters$. At least as far as the cosmetics are concerned: perfect! Even the old GEC-
sticker is superbly imitated. Okay … the electrical data … well, you can’t have everything,
can you? Back to the roots … or to revenue. Revenue, mostly, though, ‘cause the NOS-tube
built for 5 $ can easily be sold for 50 $ via the internet (foun in grandads atic no garratee).
The odd tube will bring in excess of 500 $ – here the financial suffering alone will
automatically take care of an “unparalleled sound experience”. At least for rectifier tubes,
such escapades are not required from the point-of-view of physics: any characteristic may be
approximated with a few diodes and a few resistors.
Apart from manufacturer-specific and vintage-specific differences, manufacturing scatter

within one lot also occurs, and so the luxury-tubes are individually measured i.e. selected. If
you order selected tubes and they are delivered without a “selected” label, you can complain.
You cannot complain if you receive tubes that are not selected. That is because “selected”
merely indicates that the tube is labeled “selected”. Whether, and how, a selection process
happens – that mostly remains in the dark realm of trade secrets. Two “selected” GZ-34
acquired from a German tube distributor both were defective. How is something like that
possible? A broken glass-container would be understandable – that can happen post-selection.
But too low a power-capacity? That had to stand out during selection – because the label
reads, after all: GZ34-STR Selected. How can anyone select without testing each tube? Only
the third specimen of this supplier could deliver the current customary for a GZ34. This is in
sharp contrast to the unselected Ultron-tubes: each of the three acquired tubes was perfect.
Fig. 10.7.4 shows measurement diagrams taken from “selected” GZ34’s. RC-loading was 32
!F, the load current was 200 mA. The high-quality tube (A) has both systems operating with
almost identical characteristic while the other two tubes (B/C, D/E) are expensive rejects.
Fig. 10.7.4: GZ34 (full-wave rectifier). U/I-characteristic of "selected" tubes of varying quality.
A = tube o.k. B1 and B2 designate the two systems of a bad tube; C1 and C2 belong to a very bad tube the
characteristic of which changes from bad to very bad within a few seconds.
[badly coated cathodes % Schade, O.: Analysis of rectifier operation, Proc. IRE, July 1943, 341-361].
$
The RCA-tube measured in Fig. 10.7.3 was not sourced there - it could be located in the basement of the
author’s home.

10.7.3 The internal impedance of the power supply
Strictly speaking, the constant DC supply-voltage generated by the power supply does not
remain constant at all: it varies dependent on the load, and in addition, a hum-interference is
superimposed onto it. Talking about a constant DC-voltage, we in fact refer to the arithmetic
mean-value of the supply-voltage, measured across a short period of time, e.g. 20 ms. Only
when not connected to a load does the power supply generate a supply-voltage that has no
superimposed ripple. This maximum voltage corresponds almost to the peak value of the
secondary mains transformer voltage in idle, e.g. 500 V. As a load is connected and draws
current (e.g. 200 mA), this voltage decreases to e.g. 460 V – a behavior that may be equated
to an ideal voltage-source with an internal impedance: in the above example Ri = (500 –
460)V / 0.2A = 200 ". The internal impedance depends on the transformer, the rectifier, the
filter capacitor and the load-impedance, but unless the load changes dramatically, the load-
dependence may be ignored, and the internal impedance may be seen as a constant
characteristic of the power supply.
Fig. 10.7.5 indicates the dependency of the supply-voltage on the load-current for different
configurations. The power supply (seen as ideal) contains an ideal voltage-source and an ideal
rectifier; the load-dependent voltage-drop is exclusively due to the capacitor discharge. The
other two curves were measured at a power supply with a real transformer having an internal
impedance of 2x40 ". The reason for the fact that such small resistances can already have
such a considerable effect is found in the high peak currents (Chapter 10.7.2).
Fig. 10.7.5: Dependency of the supply- (left) and of the hum-interference-voltage (right) on the load current.
a = ideal power supply, b = Si-rectifier (1N4007), c = tube rectifier (GZ34). Pure RC-loading.
Of course, it makes a difference whether the supply-voltage sags by 30 V or by 100 V,

because the operating points of the power tubes depend on this value. Depending on the
filtering, such voltage fluctuations can have an effect even up to the preamplifier tubes
(Chapter 10.1). The largest sag in Fig. 10.7.5 is found in the voltage at the tube rectifier: for I
= 150 mA it amounts to U = 75 V. With U / I follows the absolute internal impedance (500
"), while dU/dI yields the differential version of it (300 "). Using a larger filter capacitance,
both internal impedances may be reduced but this increases the peak current flowing through
the diodes (compare to Fig. 10.7.2).
The hum-interference superimposed onto the supply-voltage is, according to Fig. 10.7.5,
smallest for the tube rectifier because the relatively large internal impedance causes a large
angle of current-flow. In push-pull power stages, the hum-currents compensate each other
within the output transformer in the ideal case (Chapter 10.5.2) whereas in single-ended
power stages, the hum-voltage causes audible interference.

10.7.4 Rectifier tubes
Datasheets very rarely indicate the voltage drop across a rectifier tube in the forward- (flow-)
direction, although the internal impedance of the power supply very significantly depends on
this. Superficial consideration of such tables easily gives rise to the impression that the main
criteria are limited to the maximum allowable voltage and current. Fig. 10.7.6 shows how big
the differences can be. Surprisingly, already the datasheet-information differs: a Philips-GZ34
of 1952-vintage is something entirely different than today’s Philips-GZ34, and a modern
5Y3-GT has little in common with a 5Y3-GT built 50 years ago.
Fig. 10.7.6: Rectifier tubes operated in forward direction: U/I-diagrams.
Most tubes offered today have no binding datasheets; the reason is probably the following (as
an imaginary example): expanding guitar wholesaler Pick-o-Might, Ltd.$ decides to also carry
tubes in the future. The sales assistant has projected colorful PPT diagrams including some
very tasty pie charts, both the sales manager and the senior sales manager, plus in particular
VP “Sales”, had euphorically agreed, and the chief exec had nodded. So, tubes it is. The order
is commissioned with Mr. Li$ (or Mr. Wu$ or Mr. Ly-ing$); the logo (cost some serious
dough to be put together by the designer) is included. 4 weeks later the first “our own – made
by Pick-o-Might!”-tubes are delivered. In the meantime, the ad has been devised: "A genuine
GZ34 in its entire powerful glory" and "First-class, old-school-bulby GZ34: audibly sweeter
sound – a reproduction of the classic Philips tube". Datasheet? Better not, cause if Messrs.
Li$, Wu$, or Ly-ing$ should some day not be able to deliver anymore, the order will quickly
have to be redirected to Mr. Wassili$ (or Mr. Wischnorschow$ or Mr Slochisow$) – of course
with the same logo. Datasheets would only tend to get in the way. Nastrowje!
$
Any similarity to past, present of future names is purely circumstantial.

10.7.5 The filter-chain
As Fig. 10.7.5 shows, a sawtooth-shaped AC-voltage (hum interference) with a peak-to-peak

voltage of easily 20 – 40 V is superimposed onto the supply-voltage. Such a high AC-
component is problematic for the preamplifier because it will contaminate the signal-flow via
the plate-resistor. This is the reason why the supply-voltage needs to be cleaned up using
several filter-stages. Such a “filter-chain” typically consists of several consecutive low-pass
filters containing, in the series branch, high-power resistors (e.g. 10 k", 2 W), and in the
respective parallel branch high-voltage electrolytic capacitors, e.g. 50 !F. In high-grade
power supplies, the first series element is in fact not a resistor but an inductor. A filter choke
(e.g. L = 3 H) is used here because its AC-resistance Z is much higher than its DC-resistance
R. At 100 Hz, a 3-mH-choke has about Z = 1885 " which is about the nineteen-fold of the
copper-resistance (typically about 100 "). In combination with a 32-!F-filter-cap, the choke
results in a 2nd-order low-pass with a cutoff frequency of 16 Hz and an attenuation rising with
a slope of 12 db/oct above that frequency.
That would be the case in an ideal world. In reality, we need to consider that filter-caps may
loose part – or all – of their capacitance at higher frequencies (the may even become
inductive). Therefore, it is recommended to connect high-voltage foil-capacitors (10 – 47 nF)
in parallel with the filter-caps. Hold on: higher frequencies in a power supply operated at 50
Hz (or 60 Hz)? Sure: the rectifiers operate as a kind of switch, and every switching action
represents a broadband event. In particular, the Si-rectifiers will interrupt the current-flow
abruptly as the voltage at the filter cap drops below the voltage provided by the transformer.
Integration$ results in a sawtooth-shaped voltage that contains significant spectral lines up
into the kHz-range. The reverse recovery time of the rectifier diodes may possibly cause
additional interference: it takes a few !s until the charge carriers are “cleaned” out of the
depletion layer, and during this minor time, needle-shaped peaks occur in the current-flow.
With a correct circuit-layout, the interference-effect will, however, be rather small. If
problems still ensue, it is possible to either use fast-recovery diodes, or to bridge the diodes
with appropriate capacitors.
Fig. 10.7.7: Power supply with filter-choke. The two diagrams show the voltage curve at the filter capacitors: left
at the first capacitor, right at the second capacitor. Top: with a faultless rectifier tube; bottom: with defective tube.
In some countries, the mains voltage will of course be different from 230 V (e.g. 110 V).
Fig. 10.7.7 indicates the filter-cap-voltages of a power supply operating with a GZ34.
Cooperating with a 32-!F-cap, the choke (3 mH) reduces the ripple to about 0.5 Veff, although
only with a faultless rectifier tube. In the defective “selected” tube measured in comparison,
the two diodes had very different characteristics, and a strong 50-Hz-component dominated
the voltage. Higher-frequency signals are not apparent in this example. The faultless tube
generates (on top of the DC-voltage) an almost perfect 100-Hz-tone the amplitude of which
can be further reduced via the subsequent RC-filters.
$
I = dQ / dt = C ! dU / dt.

10.7.6 The mains transformer
The mains transformer generates, from the mains voltage (e.g. 230 Veff, depending on the
country), the AC-voltages required in the amp; it also galvanically separates these voltages
from the mains line. A transformer typical for a guitar amplifier has, on its secondary side,
two filament-windings (5 V and 6.3 V) and a plate-winding with a middle tap (e.g. 2x350 V).
The following elaborations refer to the relationships between one primary and one secondary
winding; further details may be found in textbooks on the subject [e.g. 7, 17, 18, 20].
One main criterion to design or to choose a transformer is the power P that needs to be
handled. Here, it is not the audio-power (e.g. 50 W) that is meant but the total power
necessary to operate the amp (e.g. 140 W). A considerable part of this power taken from the
mains line is converted into heat in the amplifier; the power delivered to the loudspeaker
usually is the comparably smaller part. The power required by the filaments can be easily
taken from datasheets; an example would be: 2xEL34 = 19 W, 4xECC83 = 7.5 W, 1xGZ34 =
9.5 W, in total 36 W. Next we get to the power absorbed by the tubes and resistors – that can
be estimated only approximately: the power fed to the triodes of course depends strongly on
the operating point; for an ECC83 we may use 1 W as a first order approximation. The two
output pentodes absorb about as much power as they make audio-power available: 50 W in
our example. This leads to the power-balance: P = 36 W + 54 W + 50 W, in total about 140 W.
This simple calculation does not include the efficiency of the transformers – for it, about 90%
would again be purposeful (although it may be less in individual cases). If the amplifier is to
have 50 W audio-power, and the output transformer will dissipate 5 W as thermal energy, the
power consumption will be not 50 W but 55 W. Broadly speaking, the power required from
the mains transformer will rise to 147 W, and if the mains transformer also has a 90%-
efficiency, it will draw 164 W from the mains power. When the power stage is overdriven,
this value can increase further; it is therefore recommended to estimate, as a benchmark for
the power consumption, the four times the value of the audio output-power. A 50-W amp
therefore will require a 200-W-transformer. If saving is an objective, a 150-W-version might
also do: an amplifier is not continuously overdriven, is it? Oh, it is?! In that case it is worth
the while to go for a few reserves and include a larger transformer right away.
Determining the transformer-size is a complicated optimization process: mains transformers

are heavy, big, and costly so that any carefree over-dimensioning needs to be under scrutiny.
On the other hand, transformer failure will require a complex repair process that might ruin
the company-image. A main criterion for the transformer-dependability is (besides adequate
proof-voltage – that is taken as a given here) the temperature of the winding. This depends on
the build-type and the load, but also on the temperature that develops within the amplifier. If
the transformer is operated close to hot tubes (70° C air-temperature are not out of the
ordinary), the maximum electrical strain will be lower compared to a fan-cooled amp.
Consequently, it is not untypical to find a 250-W-specified transformer in a 50-W-amp. An
entirely different question, however, is whether the prices asked for such transformers are
justified. The corresponding sums are not always based on special safety-reserves, but on the
fact that old (but famous) predecessors from the 1950’s and -60’s are replicated. Do not let
yourself be restrained if you desire to pay 250 # for such an old-school 250-W-mains-
transforner; it should be noted, however, that 300-W-toroidal-transformers can be had already
from 50 # – and these even meet the present CE- and other international regulations.

A mains transformer constitutes a voltage-source that is defined by its open-loop voltage and
its internal impedance. It has been shown in Chapter 10.7.3 that a load-dependent sagging of
the supply-voltage depends on the internal impedance of the transformer, among other
factors; in that sense it may well be desirable to copy old models. Alternatively, it is however
just as possible to increase the internal impedance of the transformer (up to a nominal value)
simply by connecting the secondary center tap not directly to ground, but via a resistor.
Fig. 10.7.8: The transformer as ideal quadripole. Voltages and currents for secondary open-loop circuit. (I2 = 0).
Fig. 10.7.8 clarifies the principle of operation of a transformer. The primary winding (here
represented via merely a single turn) carries the primary current I1; it generates the magnetic
flux ". First, the secondary circuit shall be considered without load (I2 = 0). The magnetic
flux is proportional to the primary current (Ampère’s circuital law); both shall be purely
sinusoidal. Any non-linearity shall be disregarded at this point. The change of magnetic flux
over time has the effect that a voltage U2 is induced in the secondary winding – U2 is
proportional to the flux-amplitude (law of induction: U2 = d" / dt). The secondary voltage and
the magnetic flux are shifted in phase by 90°, and so are primary current and primary voltage,
because the primary winding represents – for an open-loop secondary circuit – a pure
inductance (for the time disregarding the copper-resistance of the windings).
Things change as a load is introduced on the secondary side, because now a secondary current
is flowing that in itself generates an (additional) magnetic flux. Assuming a stiff voltage-
source as input (the mains supply is of low impedance), the secondary load must not change
the magnetic flux – the primary current and the magnetic flux are interdependent via the law
of induction, after all. The temporal course of the magnetic flux can, however, be maintained
only if the magnetic flux "2 generated by the secondary current is compensated by a further
magnetic flux "1 that is in opposite phase to "2. "1 needs to be generated by an additional
input current. In summary: " does not change as a secondary load is connected; an ohmic
secondary load will, however, have the effect that a primary active current joins the primary
reactive current. This would be the ideal point-of-view. Models that are closer to reality also
consider the ohmic resistances, the magnetic leakage, winding-capacitances, and the non-
linear behavior of the core material.
The winding resistances depend on winding length, turns number, wire-cross-section, and
specific (copper-) resistance; these are one cause of transformer losses$, i.e. of the fact that a
transformer will convert a part of the received power into heat. For the primary winding of a
mains transformer, 6 " is not an unusual value. This resistance will, however, not be the only
component of the primary impedance, because: 230 V / 6 " = 38 A – that input current would
be too high an order of magnitude. For a secondary open-loop circuit, the main component of
the primary impedance is an inductance$, connected (in the simple equivalent circuit) in series
to the winding resistance. With a secondary active load, the magnetic field transmits active
power that is taken from the primary circuit via an additional active resistance.
$
Losses in the iron will be looked into later.

In the interest of a high efficiency, and as a carrier of reactive and – in particular – active
power, the magnetic field generated by the primary winding should completely penetrate the
secondary winding as far as at all possible. In order to guide the field, ferromagnetic material
(see Chapter 4) that has a much smaller magnetic resistance (relative to air) is used for the
transformer core. However, this magnetic resistance is highly non-linear: for strong magnetic
flux densities, saturation effects occur. As the limit of saturation is exceeded (not a sudden but
a gradual process), the ferromagnetic material increasingly looses its good magnetic
conductance: for the magnetic flux beyond the limit, the ferromagnetic acts merely with a
conductance as bad as the normally much worse conducting air. Moreover, the core material
shows the non-linearity not only at high flux densities but very strongly at small drive-levels,
as well – this is in sharp contrast to many classic non-linearities. The mains transformer is
operated with an AC voltage that will not change significantly (nominal local mains voltage),
and it is therefore not purposeful to devise a small-signal equivalent circuit diagram – what
it would show would be unsuitable for the typical mode of operations.
Fig. 10.7.9 shows the dependency of the primary current on the primary voltage for a mains
transformer (EI 105c) without load, and also the dependency of “a kind of impedance” on the
voltage. Mind you: the impedance in the classical sense is only defined for sinusoidal signals!
The curves shown here are supposed to give an impression of the strong dependency on the
voltage; H$ and " are the units for non-linear elements.
Fig. 10.7.9: Left: primary RMS-current vs. primary RMS-voltage for transformer w/out load; the other two
diagrams show the quotient of two RMS-values: U / 2#fI and U / I. (H = Henry$).
In the small-signal equivalent circuit diagram, the primary impedance of the un-loaded
transformer would (with U & 0) result from a series circuit of copper-resistance (6 ") and
inductance L = 0.4 H. However, already at U = 1 V, L rises to about 1 H, and further increases
up to 11 H with increasing voltage. Already at U = 1 V there is a clearly visible non-linearity
between current and voltage that increases its influence further with mounting voltage. This
system is strongly non-linear and in a sense even time-variant: the operating point on the
magnetic hysteresis loop depends on the previous drive-levels, and it can run through the
curve in one and the same direction (counter-clockwise) only. If switching-off happens at an
instant of high field-strength, a different operating point ensues compared to a slow decrease
of the AC-field to zero. Currents and impedances measured for small voltages are normally
not reproducible if in the meantime a strong field has been present. However, since the mains
transformers in guitar amplifiers are not operated at 1 V but at (230 V (or 110 V), the
behavior with strongly varying voltage will not be elaborated on. Specialist literature does
offer supplementary information on this.
$
The unit Henry (H) must not be confused with the formula symbol of the field strength (H).

Fig. 10.7.10 depicts measurements of a mains transformer that is offered for replica Marshall-
amplifiers (TAD-JTM45). First, there is no load; measured are primary current and voltage.
While the primary voltages – generated from the mains line via a variac – are approximately
sinusoidal, the primary current show voltage-dependent deviations from a sinusoidal curve.
With increasing RMS-value of the primary voltage, a bulge emerges in the time-curve of the
current, with a maximum close (in time) to the zero-crossing of the voltage. Explanation: at
these times, the magnetic flux becomes largest (Fig. 10.7.8), the core material is saturated,
decreases, I increases.
Fig. 10.7.10: Time-curve of primary voltage (left) and current (right) for a mains transformer without load.
Here, the inductance is not a constant anymore but dependent on the drive-level.
The time-course of the current with its non-linear distortion can be dissected into sinusoidal
components, and with a few partials, the basic behavior can already be explained: the active
power is made up from the sine-voltage cooperating with the sine-component of the current,
while the cosine-component of the current forms, together with the sine voltage, the reactive
power:
This simple summation is indicated via a dashed line in the second half of the right hand
section of the figure; the basic usefulness can be recognized – if necessary, further partials of
higher order must simply be added in. Taken by themselves, the sine-components (uS, iS) in
combination yield the active part (convolution theorem of the Fourier-transform), while the
mean-value of the sine- and cosine-oscillations (orthogonal to each other) always results in
zero – these products therefore represent the reactive power. Understandably so: the
magnetic leakage-flux exiting the core includes purely reactive power as long as no eddy-
currents are generated. The re-magnetization of the core sheets, however, requires active
power, in the example this would be about 13 W. This power is irreversibly fed to the core
and generates heat. Compared to these re-magnetization losses, the copper-losses in the
primary winding are, at 0.07 W, insignificant. No heat is generated in the (un-loaded)
secondary winding, either. The following holds with good approximation: in idle, only iron-
losses appear. It should be stressed that a reactive current, too, does flow in reality: it
generates a co-phase voltage-drop across any ohmic resistance it traverses; this voltage drop
implies an active power in conjunction with the reactive current. In the above example,
however, the primary copper-resistance is (at 6 ") so small that the active power generated at
it has no bearing (for secondary open-loop operation). This will change as a load is connected
to secondary side. The primary current may now be as high as more than 1 A.

For the following measurements, the transformer received to a secondary load in the form of a
10-k"-resistor fed from the series-connected high-voltage windings (2 x 350 V). Fig. 10.7.11
shows the resulting primary voltages and currents: in contrast to Fig. 10.7.10, the current
follows a much more sinusoidal curve because an additional active current joins the
magnetizing current – the active power taken from the secondary winding (49 W) needs to be
delivered by the primary side as active power, as well. The total active power fed to the
primary side is now 63 W, and again about 14 W of this remains in the transformer and is
dissipated (into heat). The magnetic flux is approximately independent of the load, and
therefore the re-magnetization losses$ (here: 13 W) are also load-independent. The copper-
losses need to be added – they increase proportionally to the yielded power.
Fig. 10.7.11: Primary voltage (right) and current (left) over time for the mains transformer connected to a load.
Although the superposition-principle is not applicable in non-linear systems, it is still possible

in good approximation to separate the primary current into a load-independent magnetization-
current and a load-current. Fig. 10.7.12 gives two examples: the mains transformer mentioned
above is given a load of 10 k", and 5 k", respectively, and an active current proportional to
the primary voltage is subtracted from the primary overall current. What remains is in all
three cases (incl. the condition w/out load) the same amount of magnetization current.
Fig. 10.7.12: Primary magnetization current for three different-sized loads (measurements).
The relationship of primary voltage vs. current may also be depicted as a Lissajous diagram
– although here we do not directly have the voltage but rather the integral of the voltage vs.
the current. While the magnetic field-strength H is directly proportional to the current, the law
of induction necessitates proportionality between magnetic flux density B and the integral of
the voltage. This allows for a mental picture of the drive situation in the magnetic circuit,
although no exact quantitative scaling is included: the induced voltage corresponds to the
product of flux-change and turns-number – the latter is not known for the investigated
transformer. That is why the ordinate of the following pictures does not show the flux ".
$
This simplified discussion does not distinguish between hysteresis- and eddy-current-losses.

Fig. 10.7.13 shows the integrated primary voltage vs. the primary current. With an un-loaded
transformer, we would see – for a purely inductive primary winding – a straight line passing
through the origin (due to the integration no circle shows), but now the non-linear core
material causes the hysteresis loop. With a load present, the active-current-component
increases and the curves widen to become more circle-like. Subtracting the voltage-
proportional load-current from the total primary current again yields the primary
magnetization current with very good agreement. The exact turns-number for the measured
transformer is unknown; manufacturer datasheets indicate a value of just short of 400 turns.
Fig. 10.7.13: Integrated primary voltage vs. primary current. Left: sec. open circuit; center & right: 10 k" load.
Right: the load current (proportional to the primary voltage) was subtracted from the primary current.
The ordinate shows the interlinking flux using the unit Vs, the actual magnetic flux" is smaller by a factor of N1
(N1 = turns number, about 400).
Summarizing the results so far gives an ambivalent picture: on the one hand, the mains
transformer is a complicated non-linear system with inhomogeneous field-distribution, on the
other had, it may be rather well approximated by a simple voltage-source with internal
impedance. For a secondary open-loop circuit, we find at the secondary connections an
approximately sinusoidal voltage that drops (sags) as a load is connected. This drop is not
dramatic but noticeably: for a secondary current of I2 = 300 mAeff, the RMS-value of the
secondary voltage decreases from e.g. 350 Veff to 338 Veff corresponding to the source
impedance of Zi = 40 ". This impedance is approximately ohmic; apart from the secondary
copper-resistance and the primary copper-resistance transformed via TN2, there is also a small
(leakage-induced) inductance. The below table lists some fundamental parameters of typical
transformer-builds; more details can be found in the chapter about output transformers.
Complete transformer-design is not the objective of the present discussion: regarding that
topic, enough special literature already exists.
Besides size, the transformer power probably is the most important parameter in the table.
In literature, mention of this power usually refers to the secondary power but occasionally
also to the primary power that is about 10% higher. The transformer power i.e. the allowable
maximum power is directly coupled to the heating-up of a transformer. If the latter becomes
too hot, the winding wires may burn through, and/or the insulation may be hurt. Over the last
decades, better core materials have become available (less hysteresis and eddy-current-losses,
and consequently less heating-up), and the resilience against high temperatures was improved,
as well. These reasons are responsible for a power span of as wide as 30 – 57 W in a M-65-
mains transformer. Admittedly, this is a pretty considerable range, but again: no recipe
without exact information about the ingredients – that holds for transformers, too. The
following table includes two values for the power; they may be interpreted as typical limits:
the smaller value represents the way the old heroes were constructed back in the day, while
the larger value is found in the datasheets of A.D. 2008 (and may be even exceeded by
another 10 – 20% using special core materials). For the core data (Fe), the smaller value
holds for a fill-factor of 90%, the larger value relates to 100%.

P /VA b / mm h / mm Cu / cm2 Cu / cm Fe / cm2 Fe / cm Fe / grams

M20 ca. 0.5 20 5.0 0.33 3.6 0.22 / 0.25 4.6 11 / 10
M30a ca. 1.5 30 7.0 0.83 5.4 0.43 / 0.49 7.0 33 / 30
M30b 2-3 30 10.5 0.83 6.1 0.65 / 0.74 7.0 45 / 40
M42 5-7 42 14.7 1.75 8.7 1.55 / 1.76 9.8 119 / 108
M55 15-21 55 20.3 2.68 11.3 3.04 / 3.45 12.9 310 / 279
M65 30-50 65 27.0 3.90 13.8 4.75 / 5.40 15.2 560 / 500
M74 60-82 74 32.2 5.17 16.0 6.52 / 7.41 17.3 890 / 790
M85a 79-105 85 32.2 5.29 17.0 8.22 / 9.34 19.8 1240 / 1120
M85b 106-140 85 44.8 5.29 19.6 11.4 / 13.0 19.8 1730 / 1560
M102a 135-180 102 35.0 7.87 19.7 10.5 / 11.9 23.8 1910 / 1770
M102b 193-240 102 52.5 7.87 23.2 15.7 / 17.9 23.8 2880 / 2640
core / mm P / VA b / mm h / mm Cu / cm2 Cu / cm Fe / cm2 Fe / cm Fe / grams
EI-30 ca. 1 25 10.0 0.50 6.0 0.9 / 1.00 6.0 45 / 41
EI-42 2.5-4.5 35 13.7 0.95 8.2 1.73 / 1.9 8.4 120 / 108
EI-48 5-9 40 15.7 1.30 9.3 2.25 / 2.5 9.6 180 / 162
EI-54 9-15 45 17.7 1.65 10.5 2.88 / 3.2 10.8 260 / 234
EI-60 14-22 50 19.9 2.06 11.7 3.6 / 4.0 12.0 360 / 324
EI-66a 21-31 55 21.9 2.49 12.8 4.32 / 4.8 13.2 480 / 432
EI-66b 31-46 55 33.5 2.49 15.1 6.63 / 7.4 13.2 733 / 660
EI-75 34-48 62.5 25.2 2.90 14.2 5.67 / 6.3 15.0 710 / 639
EI-78 40-60 65 26.4 3.16 15.1 6.17 / 6.9 15.6 805 / 725
EI-84a 58-80 70 27.9 3.85 16.2 7.03 / 7.8 16.8 985 / 887
EI-84b 75-106 70 41.9 3.85 19.0 10.6 / 11.7 16.8 1480 / 1332
EI-92a 70-91 74 22.9 9.4 16.8 4.7 / 5.3 19.4 770 / 693
EI-92b 95-123 74 31.9 9.4 18.6 6.6 / 7.3 19.4 1070 / 963
EI-96a 100-128 80 34.0 4.9 18.8 9.8 / 10.9 19.2 1567 / 1410
EI-96b 125-170 80 44.0 4.9 20.8 12.7 / 14.1 19.2 2033 / 1830
EI-96c 160-215 80 58.0 4.9 23.6 16.7 / 18.6 19.2 2678 / 2410
EI-105a 120-160 87.5 35.0 5.8 20.3 11.0 / 12.3 21.0 1930 / 1737
EI-105b 150-210 87.5 44.8 5.8 22.3 14.1 / 15.7 21.0 2470 / 2223
EI-105c 190-260 87.5 56.0 5.8 24.5 17.6 / 19.6 21.0 3088 / 2779
EI-106a 135-180 85 31.9 10.6 20.5 8.3 / 9.3 21.8 1520 / 1370
EI-106b 184-239 85 44.8 10.6 23.1 11.7 / 13.0 21.8 2130 / 1920
EI-108 140-180 90 36.1 6.2 21.0 11.7 / 13.0 21.6 2110 / 1900
EI-120a 200-250 100 40.0 7.6 22.9 14.4 / 16.0 24 2889 / 2600
EI-120b 250-320 100 52.2 7.6 25.3 18.8 / 20.9 24 3756 / 3380
EI-120c 320-400 100 72.1 7.6 29.3 25.9 / 28.8 24 5200 / 4680
EI-130a 250-320 105 36.1 16.7 24.2 11.3 / 12.6 27 2570 / 2310
EI-130b 290-400 105 46.1 16.7 26.2 14.5 / 16.1 27 3280 / 2950
EI-150a 340-480 120 40.1 20.9 28.1 14.4 / 16 31 3720 / 3350
EI-150b 430-580 120 50.1 20.9 30.1 18.0 / 20 31 4650 / 4180
EI-150c 500-670 120 60.1 20.9 32.1 21.6 / 24 31 5550 / 5000
EI-150Na 400-510 145 47.9 13.4 28.5 21.5 / 24.0 30 5400 / 4860
EI-150Nb 500-600 145 64.9 13.4 31.9 29.2 / 32.5 30 7300 / 6570
EI-150Nc 630-700 145 90.9 13.4 37.1 40.9 / 45.5 30 10200 / 9180
Table: Transformer-data; b = width, h = height of the metal-sheet-package.
Cu-data: cross section of winding, length of winding.
Fe-data: cross-section of iron; path-length in iron, core-mass (fill-factor 90%.) T’formers in italics: low wastage.

It is customary not to specify the gross-dimensions of a mains transformer, but to give the
core-dimensions. The most-often-used core sheets either have the M-format, or the EI-
format. The core sheets of an M85-transformer feature an edge length of 85 mm (all M-sheets
are of a square shape), and usually are of a thickness of 0.35 mm (0.5 mm is also possible).
For the thickness (stacking-height) of the whole core, there are two nominal values
(designated a or b): M85a = 32.2 mm, M85b = 44.8 mm. In order to minimize the effects of
the air gap, the sheets are stacked alternately from opposite directions. With the EI-core, there
is – besides the common wastage-free core shape – also the low-wastage shape. Compared to
the M-cores, EI-cores have three air gaps, and therefore tend to have higher fringe-losses –
but they are easier to assemble. In the wastage-free cut, the punching-out of two E-pieces
exactly yields two I-pieces (without clippings). However, the cross-section of the winding is
smaller compared to the low-wastage cut. Fig. 10.7.14 shows all three shapes of core sheets;
the table just shown summarizes the dimensions. The geometric and magnetic properties of
the core sheets are standardized according to various standards; still, it may not be taken as a
given that all manufacturers on the globe produce their transformers according to DIN or EN.
Fig. 10.7.14: Core sheets normalized to equal width: M-core (left), low-wastage EI-core (middle),
wastage-free EI-core (right). There are different standards for the mounting holes.
Kühn R.: Der Kleintransformator. C.F. Winter 1964, Prien.

Hanncke W.: Kleintransformatoren und Eisenkerndrosseln, Vogel 1970, Würzburg.
Klein P.: Netztransformatoren und Drosseln, Franzis 1979, Munich.
Feldtkeller R.: Theorie der Spulen und Übertrager, Hirzel 1971, Stuttgart.

10.8 Effects
10.8.1 Reverb
In every (interior) space♣, floor, ceiling and walls reflect sound. Individual reflections with a
temporal distance of more than about 50 ms are perceived as individual echoes. Reflections
arriving with smaller intervals create a perception of reverb. Since sound propagation in a
room (i.e. in air) is – with very good approximation – a linear process (LTI-system), the
system “room” may be described by its impulse response (Fig. 10.8.1), or by its transfer
function.
In order to obtain the impulse-response of a room, the room is acoustically excited via a
sound-impulse: this could be an electrically generated spark (spark-plug), or a bursting air-
balloon, or a hand-clap, or something similar. In reality, such an excitation signal is not the
Dirac impulse known from systems theory but a real sound impulse of a duration larger than
zero and an amplitude smaller than infinity. A microphone picks up the sound pressure at the
measuring location and its magnitude is depicted as a graph over time (Fig. 10.8.1). Since
there are any number of excitation- and measuring-locations in a room, there is also a
corresponding multitude of impulse responses. With passing time, the reflections become
weaker and their density increases. The first reflections (early reflections) serve the auditory
system to obtain information about the size of the room. The speed at which the reflections
decrease is a measure for the absorption in the room. Slow decay results in a reverberant
impression, the opposite impression is called dry.
Fig. 10.8.1: Impulse response of a room (reflectogram). Left: model, right: examples for a real room.
Reverb-springs, reverb-plates, magnetic-tape devices or electronic delay-systems are used to

simulate real room reverb. For guitar amps, the reverb spring has established itself as a
standard from the early 1960s. After being available for some time in a stand-alone device
only, it was first integrated into Fender’s Vibroverb-Amp from 1963. Two steel-wires wound
into helical springs serve as delay lines with mechanical waves running back and forth within
them. The basics of this delay-principle had been investigated at Bell Labs, and employees at
the Hammond Company had developed it into a product ready for series production. The
“reverb can” (or “reverb pan” or “reverb tank”) as it is manufactured today by Accutronics
holds steel wires of a diameter of 0,4 mm that are wound into a helix of 4.2 mm outer
diameter. On the side of the actuator, an electromagnet creates forces within a small
permanent magnet that deforms the wire; the sensor side operates similarly: the movement of
a permanent magnet induces a voltage in a sensor-coil.
♣
In the following, the term „room“ is used; it is always associated with a space having reflecting boundaries
(room, hall).
Translated into English by Tilmann Zwicker © M. Zollner 2007

10.8 Effekte 10-205
The originally used reverb system included 2 subdivided springs; newer systems are also
available with 3 springs (Fig. 10.8.2). The connecting point (which is not located exactly in
the middle) between the subdivisions creates additional reflections. As a current flows
through the actuator coil, the permanent magnet creates torsion in the steel wire that results in
a flexural wave. The latter runs along the wire and reaches the other end after about 30 – 40
ms. With the two springs having slightly different dimensions, the delays differ as well (43
ms and 41 ms, according to the manufacturer).
Fig. 10.8.2: Reverb system with 2 subdivided

springs (Accutronics). Alternatively, systems with
3 springs are also in use. The flexural waves travel
along the springs and are reflected at both ends. At
the connecting points, reflections are created, as
well – these are, however, less pronounced.
Compared to real room reverb, spring-reverb shows a significantly different behavior: sound
propagation in space is three-dimensional and non-dispersive, while in the spring, it is one-
dimensional and dispersive. If we had only a single spring without subdivision, we would get
merely a sequence of echoes (e.g. after 30, 90, 150, 210 ... ms). These echoes would be
equidistant, separated by the time it takes the sound to travel back and forth in the spring.
Conversely, the average reflection density in a real room increases with time squared t2. Each
reverb spring consist of two parts connected via a ring. If we take the spring to be a
mechanical line (compare to Chapter 2), the ring acts as mass loading which reflects in
particular the higher-frequency waves. Thus, two echo-systems are connected in series, and
the reflected wave obtains a t-proportional component at high frequencies. The second
(subdivided) spring connected in parallel, on the other hand, merely doubles the density of the
reflections without changing the exponent of the time-dependency.
Fig. 10.8.3 shows spectrograms of the impulse response: in the left picture for the two-spring
system, and in the right picture for the same system but with on spring clamped down (such
that no vibration could be formed). Clearly, there is not really any one delay-time per spring.
Rather, a frequency dependent group-delay is created due to the dispersive propagation: high-
frequency components require about 50% more time to arrive than low-frequency
components.
Fig. 10.8.3: DFT-spectrograms of the impulse response: two reverb-springs in parallel (left), one spring (right).
These figures are reserved for the print-version of this book.
© M. Zollner 2007 Translated into English by Tilmann Zwicker

The usable upper frequency limit of a spring-reverb is about 5 kHz – fully adequate for a
guitar amp. Fig. 10.8.4 depicts a third-octave analysis taken from an excitation with pink
noise [3]. The different curve shapes in the high-frequency range result from different loading
of the sensor coil: the inductive source impedance acts, in conjunction with the load
capacitance, as a second-order low-pass. This low-pass can generate a slight resonance peak
at 4,5 kHz (----) given the appropriate dimensioning. A measurement with a sine sweep
enables us to take a closer look at the fine structure but requires consideration of the
extremely long attack and decay times. Even using a sweep-duration of 2 minutes, the system
cannot actually “settle”: the exact position of the maxima and – in particular – the minima
depends on the measurement parameters (resolution, sweep duration).
Fig. 10.8.4: Accutronics reverb 4AB3C1B, current input. 1/3-octave-analysis (upper left), sweep measurements.
Fig. 10.8.5: 4AB3C1B, sweep-analysis; 2 springs in parallel (left), one of the 2 dampened (right).

10.8 Effekte 10-207
In Fig. 10.8.5 we see the lower-order natural frequencies with enlarged scaling. The lowest of
these natural frequencies are located at 12,5 and 15,6 Hz, respectively, for the two parallel
springs, and the higher natural frequencies are found at multiple integers thereof (circles and
stars in the figure). At 63,5 Hz, there is an interaction of the 5th and 4th natural frequencies, the
result being a beat-effect of a very long periodicity. The right hand section of the figure
clarifies that the minima are the effect of destructive interference: with only one spring active,
the comb-structure is much more even. This regularity also supports the hypothesis (on the
basis of transmission-line-theory) that the ring positioned in the middle of the string (and
connecting the string subdivisions) works as a scattering body predominantly at high
frequencies.
The decay of the reverb is usually expressed as the reverberation time T60: this is the time it
takes for the (1/3-octave) level to decrease by 60 dB after switching off the excitation signal.
In the high-frequency range, we find a by-the-book behavior (Fig. 10.8.6): the level decreases
linearly with time. The reverberation time is 2,5 s. At low frequencies two superimposed
decay-processes reveal themselves: an early fast decay and a subsequent slower decay. In
such cases the perception-relevant early-decay-time is specified as six times the duration it
takes the signal to decay from -5 dB to -15 dB. The reference level is the averaged level for
the steady state excitation. We can see in the figure that this EDT (T10) is, again, 2,5 s for the
chosen 1/3-octave band – the subsequent slow decay can be attributed a reverberation time of
about 12 s. In guitar amps, the frequency range below about 300 Hz does not have a particular
importance (for the reverb signal): usually a high-pass will effectively dampen the lows in
order to suppress any annoying booming.
Fig. 10.8.6: 1/3-octave decay analysis, Accutronics spring-reverb 4AB3C1B. 250 Hz (left), 2000 Hz (right).
The lower pictures show the frequency-dependency of the reverberation times.

The kink in the reverberation curve is probably an effect of the scattering mass in the middle
of the spring. The waves reflected at it reach the (partially absorbing) end of the spring twice
as often as the waves passing through, and we get two types of standing waves. This
hypothesis, however, was not extensively tested via experiments – a full dampening of one of
the two springs did in any case reveal clearly that the kink is not a result of different
absorption in the two springs: the individual (subdivided) spring shows the same kink-
behavior as shown in Fig. 10.8.6.
In order to achieve a reasonably frequency-independent transmission with the investigated

Accutronics reverb system, driving it with a stiff current source is conducive (Fig. 10.8.4).
With a voltage source, a strong treble-loss would occur due to the substantially inductive
input impedance of the actuator coil. A high impedance source is automatically made
available by tubes in a common-cathode circuit. However, the Accutronics system has such a
low input impedance (8 Ω at 1 kHz) that an extreme mismatch would result. The optimum
current drive happens with a source impedance of about 100 Ω – this is obtainable from a tube
plate only using a transformer. Impedances are transformed with the transformation-ratio
squared: a transformer with a 25:1 ratio will yield – for a tube output impedance of 62,5 kΩ –
the appropriate secondary source impedance of 100 Ω. Fender’s stand-alone reverb unit 6G-
15 employs a 6K6-GT to drive the reverb-transformer; this low-power pentode features an
internal impedance♣ of 90 kΩ (the 6V6-GT has 50 kΩ). If the reverb is integrated into a guitar
amplifier, it is almost always a 12AT7 that is deployed; it has an internal impedance of
merely about 40 kΩ per triode. Since both triodes in the tube are connected in parallel (!), the
source impedance drops to 20 kΩ. On top of this, the reverb transformer 125A20B has a
transformation ratio of 50. As a consequence, the reverb system is effectively driven from a
voltage source above 1 kHz, and a corresponding treble-loss.
If the reverb system were a linear device (in the sense used in systems theory), we could insert
a corrective filter at any point in the amplification chain and boost the missing treble.
However, both the reverb spring and the tubes are non-linear devices, distorting at high
signal levels, and generating noise and rumble in the small-signal range. Filter-design
therefore always includes a component of dynamics-optimization, as well. In a typical Fender
amp, predominantly the low frequencies are attenuated ahead and after the reverb spring – the
treble-boost-enabling current-drive is only rudimentarily taken advantage of.
In Fig. 10.8.7 we see the transmission frequency response from the ECC83 ahead of the
reverb branch up to the plate of the ECC81. The filtering is done via two sections: the RC-
high-pass ahead of the ECC81 (fg = 320 Hz), and the inductive input impedance of the reverb-
transformer. Since the double-ECC81 has quite a low output impedance, we see the reverb
driven by a current source only up to about 1 kHz – in the upper frequency range, the tube
acts as a voltage source. The voltage transmission-factor of the reverb system is shown in Fig.
10.8.9 – in contrast to the situation with a current source (Fig. 10.8.4) we find a pronounced
treble-loss. The connection to the reverb-potentiometer is done via a 3-nF-capacitor resulting
in another high-pass filtering (350 Hz). The overall reverb-branch has a bandpass charac-
teristic centered around 600 Hz while the direct signal only receives a mild treble-boost. The
circuit depicted in Fig. 10.8.8 is typical for many Fender amplifiers – some do have a 2-nF-
capcitor connected in parallel to the reverb-pan output in order to create a small resonance
peak (Fig. 10.8.4).
♣
Data-sheet specifications. ♠
The 12AT7 (= ECC81) is working at an atypical operating point!

10.8 Effekte 10-209
Fig. 10.8.7: Transfer characteristic from the ECC83 up to the ECC81 (left); transformer input (right).
Fig. 10.8.8: Circuit of a typical

Fender reverb. The reverb-
spring input is of low
impedance, the reverb
transformer has a 50:1-ratio.
Fig. 10.8.9: Smoothed voltage transmission factor of the spring-reverb system (left). Transmission factor of the
overall reverb-branch (right ––––), and of the direct signal (right -----). Both curves in the right-hand picture
show the transmission from the plate of the ECC83 ahead of the reverb branch up to the last ECC83 in the reverb
branch. (Fig. 10.8.8). These figures are reserved for the print-version of this book.
The Accutronics reverb-pans are coded with 7 characters, e.g. 4AB3C1B. The individual characters indicate:
1st position: type. At the time of writing the Types 1, 4, 8 and 9 are available.
Accutronics reverb-pan Type 1 and Type 4 Accutronics reverb-pan Type 8 and Type 9
2nd position = Zin 3rd position = Zout 2nd position = Zin 3rd position = Zout
A=8Ω D = 250 Ω A = 500 Ω A = 10 Ω D = 310 Ω A = 600 Ω
B = 150 Ω E = 600 Ω B = 2250 Ω B = 190 Ω E = 800 Ω B = 2575 Ω
C = 200 Ω F = 1475 Ω C = 10 kΩ C = 240 Ω F = 1925 Ω C = 12 kΩ
4th position = reverberation time: 1 = short, 2 = medium, 3 = long.
5th position = chassis connected to: A = Input + Output; B = Input; C = Output; D = chassis insulated.
6th position = pan lock: 1 = no lock.
7th position = preferred mounting orientation: A = ; B = ; C = ; D = ; E = ; F = ;

1 2
3 4
5 6
These figures are reserved for the print-version

of this book.
1: Accutronics (4AB3C1B), 2 springs, subdivided.

2: Accutronics (9AB2C1B), 3 springs, subdivided.
3: low-cost, 2 springs (17 cm), not subdivided.
4: low-cost, 2 springs (14 cm), not subdivided.
5: digital reverb, "Spring-Reverb".
6: digital reverb, "Spring-Reverb".
7: digital reverb, "Room-Reverb".
7
Fig. 10.8.10: Various reverb spectra.

10.8 Effekte 10-211
In Fig. 10.8.10 we find a comparison between the reverb spectra of various reverb systems.
The two Accutronics tanks represent the spring-reverb standard: they measure about 36 cm
and are subdivided in the middle, with clearly visible individual reflections and dispersion.
The spring-reverbs 3 and 4 are only of half the size compared to the Accutronics; they have
no subdivision. The dispersion here is much stronger compared to systems 1 and 2, and the
bandwidth is smaller, as is the reflection density – and the price is, of course, lower also. The
pictures 5 to 7 show spectra of digital reverb systems. System 5 is marketed as “Spring
Reverb” but has little similarity to an actual spring-reverb. In system 6, we may at least
surmise that the developers sought to model the spring-typical dispersion although the result
is not very authentic. System 7 is offered as “Room Reverb” and it does differ from the
previously shown systems in that the strong periodicity is gone. The limited bandwidth of
only 2,5 kHz is probably due to the computation power: the larger the bandwidth, the more
load on the signal processor. The reverb spectra of a real room (Fig. 10.8.11) show, in
comparison to the models, a much higher reflection-density and no discernible periodicity.
The spectrogram has only limited meaningfulness here: since the DFT on which it is based
cannot provide a high selectivity at the same time in both the frequency-domain and the time-
domain. Nevertheless, the spectra shown enable us to get a basic insight into the individual
reverb structures.
Fig. 10.8.11: Broad-band spectrum of a real room (left). Digital reverb of a studio-grade effects-processor,
“large room” program with pre-delay and slight treble attenuation (left)
We should stress that every one of the reverb-systems discussed here can serve to generate a
quite useful reverb for guitar. The responses following a short impulse may sound somewhat
strange, but with a guitar such an excitation does not normally occur. Of course, compared to
a real room, the wavering wash created by a spring-reverb has a somewhat outlandish sound
at the first moment. However, the “room reverb” sounds just as peculiar in comparison if we
have just listened to a Fender spring-reverb. There is a good reason that professional reverb
devices offer a multitude of special reverb parameters to adjust such that the sound can be
taylored to individual tastes and needs. In most cases, it is real rooms that are to be modeled
(living room, hall, church, stadium), while an authentic digital simulation of a spring-reverb is
not found that often. Maybe this is the reason why there are still “real” spring-reverb system
and devices on the market.

10.8.2 Vibrato / Tremolo
From a systems-theory point-of-view, a vibrato- or tremolo-system is a modulator with time-

variant transfer-characteristics – it changes the signal-amplitude or -frequency. Leo Fender’s
usage of these terms for his guitars and amplifiers has created a big mix-up that the music
world has not really recovered from even today; a clear assignment between the terms and the
respective function has not reestablished itself: does the vibrato effect in fact change the pitch
– or is it the volume that varies? Fender’s Stratocaster, protected since 1954 by US-Patent No.
2741146, according to the patent description holds a “tremolo device” to change the pitch. 50
years later, Fender brochures still use the term in the same sense. However, in a Fender
service manual from 1968, the corresponding unit on a Mustang guitar is suddenly called
vibrato, although on the same page the term tremolo is used for the Stratocaster and the
Bronco. Similar confusion happens with the amps: vibrato is originally used for amplitude-
modulation (change in loudness), and the Vibrolux amplifier indeed includes this function.
How about the Tremolux? Same – it’s the identical effect. Does that feel complicated? Yep,
without a doubt: there is a Tremolux with a tremolo-pedal♣, and also a later one with a
vibrato-pedal. And, sure enough, there is a Vibrolux-version with a tremolo-pedal – and also
one with a vibrato-pedal. The circuit that generates the effect, is always based on the same
principle: originally it was a time-variant grid-bias that varied the amplification factor; later a
light dependent resistor (LDR) illuminated by a blinking light – in any case the typical
amplitude-modulator that was most often termed “vibrato” in the Fender brochures. Not
always, though: what does the 1968 Fender brochure designate the built-in amplitude
modulator for the Princeton Reverb (sporting a “Vibrato” pedal)? Right you are: it is called a
tremolo.
Fender did offer not just this one modulation effect: in 1959, the Vibrasonic amp received a
special circuit generating a mixture of frequency modulation (FM) and amplitude modulation
(AM). In the mid-frequency range there was mainly FM, and in the treble and bass ranges an
AM working in opposite directions: as the treble got louder, the bass got softer, and vice
versa. This same circuit could be found at the beginning of the 1960s also in the Concert,
Bandmaster, Pro, and Super amps, and in a slightly modified version in the Showman and the
Twin. Its reign was short, however: it soon was replaced by the LDR-amplitude-modulator.
With one exception all these effects were designated “vibrato” at Fender; just for the
Princeton reverb the same effect was called “tremolo” – as mentioned above.
In summary: at Fender, “tremolo” is often (but not always) used for FM, and “vibrato” often
(but not always) stands for AM. The classical (and scientific) definition is the other way
‘round: tremolo = AM, and vibrato = FM.
How are these two effects perceived differently in our hearing? Surprisingly: not to a big
degree – as long as the modulation is not too strong. The reason is that pure FM does not
occur in normal situations: due to selective resonances in speakers and, especially, in the
rooms we listen in, FM always generates an additional AM [e.g. 3]. The latter may even be
detected (for small modulation indices) more easily by the hearing system. It is difficult to
generate FM with a strong modulation index while it is much easier for AM. Here we may
find the reason why Fender says good-bye to FM in the early 60s, and fits the low-cost and
highly efficient LDR-modulator into all his amps. The following circuit descriptions focus
predominantly on the well-documented and trend-setting Fender amps – well aware that other
manufacturers have also developed and successfully marketed vibrato/tremolo-circuits.
♣
Label of the footswitch-jack

10.8 Effekte 10-213
In your typical tube amp, a triode generates the low-frequency signal (LFO = low frequency
oscillator) while the modulation itself happens in another tube (or more tubes), or in the LDR.
The LFO is of a relatively simple build: a tube in common-cathode configuration with
frequency-dependent feedback. The tube inverts the signal from plate to cathode, and
consequently the feedback circuit needs also to invert; both inversions result in a phase shift
of 2π, which is the requirement for self-excitation. In addition, the loop gain needs to be
larger than one – easily achievable with a tube. Fig. 10.8.12 shows a circuit as it is frequently
utilized (Fender, VOX, and many others). The feedback branch consists of a 3rd-order high-
pass with a variable resistor that adjusts the oscillation frequency between about 3 Hz and 11
Hz. Since there is no amplitude control, the generated signal is not of perfect sine-shape – the
system is non-linear and therefore there is, strictly speaking, not really a transfer function as
such. This should not be seen as a problem, however, since the approximation achievable with
the linear model is perfectly practice-oriented and therefore adequate for the present context.
Fig. 10.8.12: LFO-circuit in a tube amplifier; magnitude- and phase- characteristics of the feedback network.
From about 1963, Fender amplifiers were fitted with an amplitude-modulator that used an
opto-coupler: an LDR intermittently illuminated by a glow-lamp. The required control signal
was tapped (with high impedance) from the circuit described above, and fed to the glow-lamp
via the second half of the double triode (ECC83). Due to the operating point chosen for this
second triode, a significant current is flowing only during a relatively short part of the LFO-
period, and the glow-lamp lights up only for a short time. The resistance of the LDR
decreases when lit and causes – integrated into the parallel branch of a voltage divider – a
signal-attenuation (Fig. 10.8.13). Significant slurring of the envelope occurs due to the
relatively long recovery time of the LDR – this is, however, rather beneficial to the auditory
perception.
Fig. 10.8.13: LDR-modulator. LFO-signal at the plate of the oscillator-tube (left), 600-Hz-sine-tine modulated by
the LDR (right). Dashed: imaginary effect of a modulator with zero recovery time.

Preceding the LDR-era, Fender deployed tube modulators. There are three types of circuits:
screen-grid modulation in the power-stage, modulation in the phase-inverter, and, for single-
ended power amps, modulation in the intermediate amplifier stage. The amplitude-modulation
is achieved simply by shifting the operating point: a superposition (addition) of an AC voltage
of very low frequency periodically moves the operating point into the end ranges of the
characteristic curve. Here, the slope of the latter (and thus also the gain) is smaller than in the
middle range: the gain becomes time-variant. The non-linear signal distortion also created
could be accepted as an additional effect; the low-frequency parasitic signal also occurring
(even without input signal from the guitar), however, requires including additional high-pass
filters. For push-pull output stages, there is an elegant workaround: since the output
transformer constitutes the difference of the two anti-phase signals, all common-mode signals
cancel each other out (as it happens in every differential amplifier). The guitar signal is fed
out-of-phase into the two halves of the push-pull stage while the LFO-signal is fed in-phase to
the two sections. The result is that the guitar signal is doubled while the spurious LFO-signal
is cancelled.
The control-grid voltages♣ of the power tubes offer themselves as the “last possibility” to
achieve the mentioned shifting of the operating point; it is implemented e.g. in the Tremolux
5G9). Synchronously pushing both grid voltages into the negative makes both tubes block: the
audio signal is attenuated. Apparently, this power-tube control was seen as superior. It is
found in several Fender amplifiers, and it superseded the driver-stage control (e.g.
Tremolux 5E9-A) introduced a few years before and feeding the LFO-signal to the cathode of
the phase-inverter. In both circuits an in-phase excitation of a differential amplifier is
accomplished which (ideally) will avoid any LFO-signal coming out of the loudspeaker.
In the Fender Vibro-Champ (AA764), this LFO-compensation does not work because it has a
single-ended power amp. Here, the LFO-signal is fed to the cathode of the driver-tube, and
it is amplified together with the guitar signal, resulting in a low-frequency interference. The
high-pass inserted directly ahead of the power-tube provides merely limited relief.
In contrast to the amplitude-modulator described above, the AM/FM-circuit first included in

1959 into the Vibrasonic is not understood prima facie. Here, the guitar signal is fed to a
frequency crossover and separated into a high-pass branch and a low-pass branch♥. The
effect is mainly a change in the loudness of the partials, but to a small degree there is also a
change in phase, and therefore in pitch. The momentary angular frequency is, in fact, the
derivative of the phase angle ϕ [3]. Since a 1st-order high-pass changes the phase by up to
90°, and a 1st-order low-pass does this by up to -90°, phase-shifts occur – as we change from
the high-pass filtering to low-pass filtering – of up to about 120° (in the Fender-typical
circuit). A pitch modulation with a frequency-swing of about ±10 Hz is possible with this
approach, allowing for definitely audible changes in pitch. The threshold for just noticeable
frequency changes is about ±2 Hz for FM-tones [12]. In Fig. 10.8.14, we see the magnitude
and phase characteristics of the Vibrasonic-circuit (5G13); the schematic is given in Fig.
10.8.15. In later amplifiers (e.g. 6G13-A), the resistive voltage divider in the high-pass branch
was dropped, with a gain of 7 dB in this branch.
♣
In principle, the screen-grid voltages of the power-tubes could be modulated, as well, but this would require a
higher control-power.
♥
Strictly speaking: high-pass and bandpass, but the bandpass center frequency is, at 60 Hz, very low.

10.8 Effekte 10-215
Fig. 10.8.14: Magnitude- and phase-characteristics of the frequency crossover (top), and of the overall system
(bottom). 5G13.
This small frequency modulation realized in the Vibrasonic et al. was perfected in the VOX
AC-30 (Fig. 10.8.15) using an all-pass circuit from the Wurlitzer organ [Petersen/Denney]
that generates mainly FM but almost no AM. The required filter network is considerable: it
uses 6 capacitors, 6 resistors and 3 amplifiers. It may nevertheless be divided into simple
partial systems for a calculation purposes. The schematic shows two active 2nd-order bandpass
filters of the same structure, differing merely in the values of the components. The signal
mapping from U0 to Ua is easily understood by omitting R3 and C3, to start with. What
remains is a capacitively bridged voltage-divider determined by 4 components (4 degrees of
freedom). One of the latter is the impedance level which tube-typically is chosen to be in the
100-kΩ-range. The second degree of freedom is the attenuation factor (about 3). Pole/zero-
compensation yields the third degree of freedom (R1C1 = R2C2), and the cutoff frequency
(about 1 kHz) yields the fourth. The result is a passive system of zero (!) order that generates
a frequency-independent attenuation (of about 10 dB) across the whole frequency range.
Fig. 10.8.15: Frequency crossover of the Vibrasonic 5G13 (Fender, left), and of the AC-30 (VOX, right).

Now we add R3 and C3, driven by an inverter (v = –1). Given the correct dimensioning, the
output signal (Ua) is of the same phase at low and high frequencies. The active branch (R3, C3)
cannot have an effect at low frequencies since C3 is of high impedance relative to the divider-
resistors. At high frequencies, no effect is there, either, because here R3 is of high impedance
relative to the divider-capacitors. It is only in the range of the cutoff frequency that R3 and C3
determine the transmission and have the effect of a phase shift from 0 to π. The components
of the active branch (R3 and C3) can be determined such that the magnitude of the transfer
characteristic becomes frequency-independent: this corresponds to a true all-pass.
An all-pass is a filter that changes only the signal phase but not the signal amplitude. This is
achieved if the numerator of the transfer function is the complex-conjugate of the
denominator of the transfer function; the magnitudes of numerator and denominator are equal
for this condition, the magnitude of the transfer function becomes a constant (i.e. it is not
dependent on ω). In the case of the filter circuit described above (Fig. 10.8.15), we get a 2nd-
order all-pass the numerator- and denominator-polynomial of which contains p at the most
with the power of two.
Second-order all-pass, p = jω
A 2nd-order transfer-function has 5 degrees of freedom. Two of these are required by the all-
pass characteristic: the same behavior for f → 0 and f → ∞ results in a = c, and the complex
conjugation of numerator and denominator yields b = –d. The remaining 3 degrees of freedom
are defined by: basic gain (H0), cutoff frequency (a) and Q-factor (b). The components of the
AC-30-filter in the original circuit were chosen such that not a perfect all-pass resulted but a
slight magnitude change did also occur (about 3 dB). The reason for this is unknown; possibly
the additionally generated AM was desirable.
An all-pass in itself does, however, still not generate a frequency modulation (FM) – it only
creates a stationary (time-invariant) phase shift. For this reason there is a second all-pass (Fig.
10.8.15) with a cutoff frequency of a factor of 4,5 lower than the cutoff frequency of the first
all-pass (1040 Hz vs. 4700 Hz). There will be a significant phase difference between the
output signals of these two all-passes that can be turned into a time-variant phase-shift by a
LFO-controlled cross-fading between the two outputs. If we take the phase modulation to be
approximately sine-shaped, the maximum of the frequency modulation generated this way
corresponds to the product of modulation-frequency (LFO-frequency) and phase-change
amplitude: . With fmod = 10 Hz and a maximum phase-change amplitude of 55°
(= 0.3π), we obtain a frequency-change amplitude of Δf = ±9.4 Hz.
We know from psycho-acoustical experiments that the threshold for just noticeable frequency
differences is about ±2 Hz at low frequencies; the auditory system becomes increasingly less
sensitive to absolute frequency changes only at frequencies above 500 Hz [12]. The frequency
modulation generated by the AC-30-Modulator is therefore clearly audible; in addition we
need to consider that the modulator circuit, and loudspeaker- as well as room-resonances,
additionally generate amplitude modulation. In conclusion, it should be noted that in the AC-
30, one of the two all-passes can switched-off such that the amplitude modulation becomes
the dominating effect.

10.8 Effekte 10-217
Fig. 10.8.16: AC-30-modulator. Top: magnitude and phase characteristic of both all-pass filters.
Middle: Magnitude- and normalized phase characteristic of the overall modulator.
Bottom: Magnitude characteristic incl. 4th-order high-pass; frequency change 2⋅Δf achievable with fLFO = 10Hz.
“Hörschwelle fmod=10 Hz”: threshold for just noticeable FM at a modulation frequency of 10 Hz
In Fig. 10.8.16, calculations regarding the transfer behavior are depicted. The LFO-controlled
crossover between the all-passes generates level changes of up to 5 dB, and phase changes of
up to 110°, resulting in frequency changes of up to 19 Hz at 10 Hz modulation frequency. In
the lower right picture, the FM-perception-threshold is shown (dashed) for comparison [12];
the achieved modulation is clearly above threshold. The 4th-order high-pass added in for the
picture on the lower left follows the modulator in the AC-30 to detach the remaining LFO-
signal from the guitar signal. A compensation of the LFO-signal is implemented in the
summation stage but this can never be perfect due to unavoidable tube tolerances.

10.8.3 Phaser / Flanger / Chorus
Since the first electric guitars came into service, there was also the wish for sound-
modifications. First, there was only a “tone”-control (a potentiometer with a capacitor), then
more sophisticated sound filters were added, followed by electronic vibrato, tremolo, echo,
and reverb. The typical guitar-echo results from periodic signal repetitions (about 50 – 500
ms delay-time), simple reverb combines several echo sequences of differing periodicity,
high-quality reverb is generated by springs (10.8.1) or digital signal processors (in the studio,
reverb-plates or reverb-chambers are used, as well). Phaser, flanger and chorus are
electronic effects based on a short delay. The delay is a linear system that delays signals. A
short time-delay sets the signal back by a few milliseconds, and therefore is different from an
echo-system.
For phaser-, flanger- and chorus-devices, the delayed signal is added to the original signal
such that a comb-filter results. The name is derived from the fact that the magnitude-
frequency-response has a remote similarity to the teeth of a comb (Fig. 10.8.17). Plotted
against a linear division of the frequency axis, the maxima and minima alternate in equal
frequency distances; the figure, however, shows the logarithmic frequency scaling as it is
preferred in electro-acoustics. Apart from the basic gain (not that important), two parameters
determine the filter behavior.: the delay-time τ and the delay gain k. Varying τ will change
the frequencies at which the maxima and minima occur (i.e. the distance between the notches
in the frequency spectrum), while k governs by how many dB the gain factor changes (i.e.
how deep the notches are). For a negative k, the first minimum is at f = 0.
Fig. 10.8.17: Comb-filter: frequency responses

and signal-flow diagram (block-diagram).
The comb-filter is in fact a typical interference filter: for a sine signal, a delay of half a
period leads to a cancellation (or attenuation). A delay of a whole period causes amplification.
Maxima and minima repeat with the period of the frequency: a minimum occurs with a delay
of 1.5, 2.5, 3.5, … (or generally n + 0.5 with n = 0, 1, 2, ...) periods of the sine-signal. For the
maxima, the situation is similar. In systems theory, such a filter is also termed FIR-filter, due
to its impulse response which is finite in the time domain: Finite Impulse Response filter.

10.8 Effekte 10-219
The special aspect about phaser / flanger / chorus is, however, not really to be seen in the
periodical frequency response but in the variation of the latter over time. A low-frequency-
oscillator (LFO) changes the delay-time τ periodically. For example, τ swings back and forth
once per second between 1 ms and 2 ms, making each minimum and each maximum sweep
across a certain frequency-range as a function of time. Strictly speaking, we encounter here a
time-variant system the description of which is not entirely trivial – but the quasi-stationary
approximation of the shifting comb-teeth (or notches) is good enough in practice. Adding
original and delayed signal (positive k) positions the lowest-frequency minimum at the
inverse of twice the delay-time (1ms ⇒ 500Hz). For too short a delay-time, there is barely any
audible effect because small changes occur in the high frequency region only. For a flanger, a
typical delay-time range is 1 … 5 ms, in extreme cases this may extend from 0,3 to 15 ms.
As the delay-time is increased to above about 20 ms, a new auditory perception is generated:
the chorus-effect. As a first-order approximation, both flanger and chorus can be described
with the block-schematic as given above. Due to the very short delay-time, the flanger
generates relatively broad minima in the signal spectrum and thus predominantly changes the
color of the sound. Conversely, the delay-time of the chorus approaches already the value
where single echoes might be discernible. This occurs at about 50 ms delay-time; our auditory
system can not yet distinguish echoes as such at τ = 25 ms, but it recognizes already a “fellow
player”. This effect is the aim of the chorus: the slightly delayed repetition is intended to
create the fuller sound of not just one but two instruments playing. In addition, the delay-time
is modulated by the LFO (as it is in the flanger), creating an impression of a whole
instrument-ensemble. The term chorus is derived from “choir”; in the latter the individual
voices start at slightly different times and sing slightly different pitches. The pitch change
(more exactly the frequency change) is the result of the time-variant delay-time τ (t). As τ
increases, f decreases, as τ decreases, f increases. The relative de-tuning is calculated as the
change of the delay-time over time: . As an example: if τ rises linearly by
10 ms within 0,5 s, the frequency of the delayed signal is decreased by 2%. A delay
modulation in the shape of a triangle generates a back-and-forth sweep in the pitch. With a
subtle mixing-in of the chorus (slow modulation, small frequency shift) the desired wavering
choir effect is generated. For extreme settings a whining frequency modulation becomes
audible.
The phaser is similar to the flanger but uses all-pass circuits to generate the delay; these all-
passes were originally created using active circuitry (Fig. 10.8.18). The RC-combination
determines the delay – with the R being the controllable element (as LDR or FET). Since a
1st-order all-pass can only shift the phase by 180°, several all-pass circuits need to be
connected in series: n = 6…10 would be a typical number. In contrast to the flanger, the
minima are not equidistant, and fewer interference notches of greater width are created..
Fig. 10.8.18: All-pass phaser

10.8.4 Wah-Wah-Pedal
The Wah-Wah-pedal is an effects device performing speech-like formant-filtering. Formants

are maxima in the speech spectrum that classify the speech-sounds [3]. The frequency F1 of
the lowest formant is about 400 Hz for the spoken vowel /o/, while for an /a/ it is around 800
Hz. If, while playing a guitar, a band-filtering is introduced with a time-variant center-
frequency, and if the latter sweeps between 400 Hz and 800 Hz, we obtain a sound change
that can be described by the vowel-sequence /oaoaoa/, or with “wahwahwah”.
In some devices the filtering was achieved via an LC-filter, the coil inductance of which (or
rather the air gap) was variable via moving a pedal. Most Wah-pedals, however, put to use an
active filter circuit in which the filter capacitance is varied by changing the gain (Miller
effect). This arrangement allowed for a sweep between about 400 Hz and 2 kHz –
measurements with an old VOX-wah yielded 0.44 – 2.3 kHz. The boost of the frequency
range around 2 kHz is typical for the formant of an /i/, so that using the full range of the pedal
results in a vowel-sequence akin to /oaiaoaiao/. More sophisticated devices (marketed with
the designation 'Yoy-Yoy' or 'Doing-Doing') offered two synchronously tunable filters –
presumable to more precisely imitate the human voice. Tempi passati – long bygone times.
Fi. 10.8.19: Wah-Wah-pedal (VOX).
Fig. 10.8.19 shows the circuit as well as some transmission-frequency-responses of a VOX-

Wah-Wah. The inductance (about 0.5 H) and the capacitance C (10 nF) determine the centre-
frequency of the filter – the capacitance is however enlarged in its effect by the factor of the
gain (0…27). The capacitance effective for the filter is thus 10…280 nF resulting in a pole-
frequency of 0,44…2,3 kHz. Apart from some copper- and ferrite-losses, the resistor (33 kΩ)
that is connected in parallel to the coil determines the Q-factor of the filter; the latter also
depends on the centre frequency. From the systems-theory point-of-view, a pole-Q-factor and
a (different) zero-Q-factor could be specified, but in practice the “Q-factor” usually is
determined using the 3-dB-down-bandwidth. For the above circuit, this definition yields Q =
3.3…15.
"Auto-Wah" is the designation for a Wah-Wah-filter that automatically controls its center
frequency. The control parameter is the signal strength i.e. approximately the loudness of the
guitar signal. Without any signal, the system tunes to the lowest possible center frequency. As
the strings are plucked lightly, the centre frequency rises slightly, for strong picking the band-
filter quickly sweeps from low to high frequencies and returns more slowly to the starting
state. This picking-strength-dependent filter-control enables the guitar player to use the wah-
wah-effect without having to operate a pedal. There will be less versatility but also less stress
for the foot.

10.8 Effekte 10-221
10.8.5 Distortion devices
In communication engineering, two types of distortion are specified: linear and non-linear
distortions. Linear distortion is created in systems with a frequency-dependent transfer-
characteristic, i.e. for example in sound filters♣ (also called tone controls, or EQ). Non-linear
distortion occurs in a non-linear system. For a system to operate linearly, it needs to have
proportionality, lack of sources, and the possibility for superpositioning. If only one of these
conditions is missing, the respective system is non-linear.
• “Proportionality” means that an n-fold increase of the input signal will result in an n-
fold increase of the output signal (doubling the input results in doubling of the output).
• “lack of sources” means that without an input signal, the output signal will be zero.
• “possibility for superpositioning” means that a transformation of a sum of signals
effected by the system will correspond to the sum of the transformed individual
signals :T(x+y) = T(x)+T(y).
The description will be simpler if linear and non-linear systems occur strictly separately, i.e. if
every sub-system will perform only linear or only non-linear mapping. A non-linear system
that does not cause any linear distortion includes no memory – this is because only devices
including memory (inductances, capacitances) generate frequency-dependent resistances and
thus create a frequency-dependent transmission. In a memory-free system, the output signal
will therefore not depend in any way on the past input signal but exclusively on the input
signal occurring at the very same moment. The transmission characteristics can be described
via a characteristic y(x). For non-linear behavior to be present, we need to have a curved
characteristic (strictly speaking, an offset also introduces non-linearity).
The amplifying elements (tube, transistor) used in guitar amplifiers all feature a curved
characteristic, and therefore every guitar amplifier operates as a non-linear device. According
to the rules of classical amplifier technology, these non-linearities are supposed to be as small
as possible, and therefore negative-feedback circuits reduce the gain and at the same time
perform a linearization. Many guitar players were satisfied with the resulting so-called
“clean” sound, but some forced non-linear distortion by overdriving their amplifiers, creating
“crunch”, “distortion”, or “fuzz”. In many amps this required using their full power and thus
very high loudness – but sometimes even at maximum gain, the resulting non-linear distortion
was not pronounced enough. This is the reason why additional devices for the generation of
non-linear distortion were created under various monikers: fuzz-box, distortion-pedal,
overdrive … Before long, the effect was not limited to additional devices: with an increasing
number of tubes, guitar amplifiers themselves soon offered possibilities to control the desired
degree of distortion.
The distortion devices described in the following are systems that add non-linear distortion
to the guitar sound. Whether this happens in the amplifier itself or in a separate device is not
distinguished to begin with. With the distortion, the guitar sound becomes fuller, sustaining,
more shrill, more aggressive, buzzing, more alive – this is always depending on the chosen
settings. There is not “the” distortion sound. The distortion also changes the dynamics of the
signal in the sense that sustain is extended. Since practically all distortion devices have a
degressive, limiting characteristic curve, any level-differences in the guitar signal are reduced
and differences between loud and soft are evened out. The originally percussive guitar sound
becomes steadier, and takes on some sound-characteristics of horns (saxophone, trumpet) or
strings (cello).
♣
M. Zollner: Signalverarbeitung. Hochschule Regensburg, 2009.

There are two types of drive-situations for customary tube- and transistor-circuits: low drive-
levels with weak distortion, and overdrive with strong distortion. In the left section of Fig.
10.8.20, we see the characteristic curve of an ECC83, and in the right-hand section the time-
function of the sine-shaped input voltage and the distorted output voltage. For small drive-
levels (up to about 1 VS), there are only small differences between input- and output-voltage.
For high drive-levels, we find a strong – and in this case asymmetrical – limiting of the signal
(“clipping”).
Fig. 10.8.20: Non-linear distortion for an ECC-83 (compare to Chapter 10.1.4).
Now, the human auditory system is not a sensor for directly analyzing the time-function.
Rather, it detects in a first step the time-variant short-term spectrum (Chapters 8.2.4 & 8.6) to
determine the sound color. If a sine-tone (e.g. 1 kHz) undergoes non-linear distortion, new
spectral lines are created at the integer multiples of the fundamental frequency, i.e. at 2 kHz, 3
kHz, 4 kHz, etc. Distortion of a signal composed of several partials will generate sum- and
difference-tones as multiples of the largest common divisor of all partials♣. If the sound of a
single guitar string were of strictly harmonic content, distorting it would still result in a
harmonic spectrum. The level and the phases of the partials would change, and so would the
sound color, but the frequency of the partials would remain unchanged. However, the
spectrum of every real string-vibration is spread in-harmonically, and it is here where we
find the key to understanding the impact of a distortion device.
If, for example, a complex tone consisting of a 100-Hz- and a 202-Hz-partial undergoes 2nd-
order distortion, additional partials at 0 Hz, 102 Hz, 200 Hz, 302 Hz, 404 Hz are created. The
0-Hz-component may be ignored because it is DC that the circuit will not transmit further.
The partials at 302 and 404 Hz will brighten up the sound if they are strong enough, but the
main effect will be close to the primaries: the partials at 100 and 102 Hz will beat against each
other, and so will the 200- and 202-Hz-partials. The non-linear distortion will, on one hand,
enlarge the spectrum towards high frequencies (i.e. emphasize the treble more), and on the
other hand the amplitudes of the primaries will start to fluctuate due to the beat effects: the
sound becomes more lively. If not only 2nd-order distortion occurs but higher-order distortion
as well, a large number of additional partials is created and correspondingly many and
possibly strong fluctuations. These fluctuations bring a kind of noise-character to the sound;
we get an effect as if additional noise would be superimposed. Strictly speaking, noise in its
usual definition belongs to the group of stochastic (random) signals, while the fluctuation of
partials generated by non-linearity is not stochastic but determined. Since this special noise
has a periodicity, it is called pseudo-noise.
♣
M. Zollner: Frequenzanalyse. Hochschule Regensburg, 2009.

10.8 Effekte 10-223
Fig. 10.8.21 depicts the zero-symmetric characteristic curves that will be used for distortion in
the following. They are odd-order functions, i.e. functions that can be expanded into a series
containing only members of odd-order power (x, x3, x5, ...).
Fig. 10.8.21: Characteristic curves of the systems used for distortion.
Simple periodic functions are distorted in Fig. 10.8.22 using characteristic curve 2 (an arctan-
function). Distortion of the sine-signal (uppermost row in the figure) results in a spectrum
with only odd-numbered harmonics. However, merely adding a second partial to the primary
signal (2nd row in the figure) may generate even-numbered harmonics – although this is not
generally the case, as demonstrated by the third row in the figure. Only a signal of half-wave
anti-symmetry will, in its spectrum, contain exclusively odd-numbered harmonics. Such a
half-wave anti-symmetric signal, if distorted via an odd-order characteristic curve, remain
half-wave anti-symmetric, and will not gain any even-numbered harmonics. In the lowermost
line of the figure, a signal of three partials is distorted. Due to the 2nd harmonic, this signal
cannot be half-wave anti-symmetric, and therefore the spectrum of the distorted signal
contains even-order harmonics as well.
Fig. 10.8.22: Time-function of undistorted and distorted signal (left), spectrum of undistorted signal, spectrum of
distorted signal (right). Time-functions and spectra individually scaled. Characteristic curve 2.

For the analyses discussed in the following, synthetic guitar sounds were used; their rules of
construction were nonetheless derived from a real guitar. Synthetic sounds were used because
all parameters of these sounds are known – which is not the case for real sounds. For the
harmonic signal, 45 partials were added. The frequencies of these partials all had an integer
relationship to the fundamental of 82 Hz, and their levels and decay time constants were taken
from a real guitar tone (open E2-string). The inharmonic signal was synthesized with the
same levels and time constants, but the frequencies of the partials were slightly spread out
according to the formula outlined in Chapter 1 (dispersion due to the bending stiffness of the
strings). The spectrograms of the undistorted and distorted synthetic guitar sound are shown
in Fig. 10.8.23. The spectrum of the harmonic sound ends at 3,7 kHz in view of the usage of
45 partials; the spectrum of the inharmonic sound goes up to 4.1 kHz due to the frequency
spreading (again 45 partial were used).
Distorting the harmonic sound results in additional partials that however are positioned
exactly within the harmonic grid. On one hand, the new partials fill up the frequency range
above 3.7 kHz, and on the other hand they change the level of the primary partials. The
degressive curvature of the distortion characteristic has the effect that the partials decay more
slowly, for some there is even an initial growth. The changes of the partial-levels are slow,
with change speeds similar to those of the primary levels. For the inharmonic sound, the
distortion generates many new partials positioned closely to the primaries such that the DFT-
analysis (and the auditory system) cannot recognize them individually anymore. The spectral
pooling of these undistinguishable lines results in fast signal modulations bearing some
resemblance to a stochastic noise process but being (strictly speaking) determined (pseudo-
noise).
The distortion has three effects on the sound: the treble-content grows (a more brilliant
sound), the dynamics are compressed (longer sustain), and the partials are pseudo-
stochastically modulated (creating a “buzzing” and “raspy” character). The pseudo-stochastic
modulation happens only for inharmonic sounds and is dependent on the string-parameters.
The thicker the string, the more noise is created. Maybe we should explicitly mention that
this holds for single tones, because for chords, the spectrum is not harmonic in a simple
fashion anymore, anyway.
The level evolutions shown in the picture below indicate that the pseudo-stochastic
modulations increase if the characteristic is more strongly curved. Analyses for characteristic
3 are not included; a similar picture as for characteristic 2 would emerge, as long there is
strong overdrive. Larger differences become apparent for less overdrive: the change from
“undistorted” to “distorted” happens abruptly for characteristic 3, and more gradually for the
other characteristics. Moreover, the spectrum becomes more treble-heavy if abrupt signal-
limitation (clipping) occurs.
It should be expressly mentioned here again that despite a zero-symmetric characteristic

(“odd-order function”), distortion products of even-order do occur. The assumption, that an
odd-order-characteristic would generate only odd-order distortion products, only holds for
half-wave anti-symmetric signals (e.g. for a sine tone). For real guitar sounds, even-order
distortion products can very well result from odd-order characteristics. In the same manner,
the assumption that tubes would generate predominantly even-order distortion products is
wrong as a general statement. Tubes do not generate “better” distortion than transistors –
otherwise nobody would ever have used a Range-Master ahead of the amp. In the Range-
Master, pure transistor distortion is generated (Chapter 10.8.5.3).

10.8 Effekte 10-225
Fig. 10.8.23: Spectrograms, harmonic (left) und inharmonic signal, 0 - 5.5 kHz, ΔL = 40dB.
1st row: no distortion, 2nd row: characteristic 1, 3rd row: characteristic 2. Bottom: level-evolution of 15th partial.

Digital emulations of analog distortion devices merit some special consideration. It is

generally known that an anti-aliasing filter needs to be included ahead of an A/D-converter in
any time-discrete system. This is in order to avoid spectral back-convolutions (sampling
theorem). If the signal is already appropriately bandwidth-limited, this filter may of course be
dispensed of. For example, there is no need to include a filter if a 5-kHz-tone is sampled at
44.1 kHz. If the A/D-converted 5-kHz-tone is, however, distorted in the digital realm with a
digital distortion characteristic, new frequency lines are generated – above half of the
sampling frequency, as well. Since time-discretization has the effect of spectral periodization,
these new lines appear around all multiples of the sampling frequency. If the distortion
characteristic is point-symmetrical, new partials are generated at 15 kHz, 25 kHz, 35 kHz and
further odd-numbered multiples. The distortion products are mirrored with respect to the
sampling frequency (e.g. 35 kHz re. 44.1 kHz), a new distortion line appears at 9.1 kHz that
would not be generated by an analog distortion device. At 0.9 kHz, and at many other
frequencies, further partials appear, sounding rather unpleasant (as a rule, if their level is high
enough). It is therefore insufficient to digitally emulate an analog distortion characteristic in
order to create a digital equivalent. The higher the frequency of the signal to be distorted, and
the more angularly shaped the distortion, the more disappointing the emulation will be.
To avoid such back-convolutions, the sampling frequency needs to be increased. Whether a

ten-fold increase is adequate or whether even much higher sampling rates are necessary,
depends on the signal, the distortion characteristic, and the quality requirements. Here is a
simple estimate: if a sine signal undergoes hard clipping with a symmetric rectangular
characteristic, new partials are generated following an si-envelope. The level of the 11th
partial is 21 dB below the level of the primary, the 99th partial is 40 dB below the level of the
primary. If the sampling frequency is 100 times of the frequency of the tone to be distorted,
the back-convolution creates an interfering tone which is 40 dB down relative to the primary-
level (Fig. 10.8.24). Of course, not only this one interfering tone is back-convoluted, and back
convolution does not happen only at the sampling frequency. The figure shows merely one
back-convolution such the lines can still be associated properly.
Fig. 10.8.24: Spectrum of a strongly distorted sine tone; time-continuous (top); time-discrete with one back-
convolution (bottom). The frequency of the back-convoluted lines is strongly dependent on the relation f / fa .

10.8 Effekte 10-227
10.8.5.1 Diodes
In order to achieve non-linear distortion of a signal, at least one non-linear component is
required. This may be a tube, a transistor, or – in the simplest case – a diode. In the following,
the term diode is meant to refer to a semiconductor diode and not to a tube diode. The latter
could also be used for distortion but this does not happen in practice. Simplified, diodes
conduct current only in the forward-direction; with the reverse-polarity, the diode blocks.
More accurate models consider that a few 100 mV are created across the diode in the forward-
direction, and moreover include the reverse-current flowing for reverse polarity. Dynamic
models add capacitances (possibly of non-linear nature) – a model for a diode can in fact get
quite complicated. As a first approximation, the so-called Shockley-characteristic suffices:
; Shockley-equation
The diode current I grows exponentially with the voltage across the diode in the forward-
direction U which is referenced to the temperature-voltage UT; IS represents a theoretical
reverse-current. Real diodes may strongly deviate from this idealization, and corrections and
supplements are necessary. In particular, it is necessary to modify the temperature-voltage: for
a real diode, measurements yield values for UT of up to more than 60 mV; moreover a track-
resistance in forward-direction needs to be considered. In the left section of Fig. 10.8.25, the
forward characteristic of an 1N4148 diode is shown with linear scaling, the middle section
shows the same characteristic but with log-scaling along the horizontal axis. The exponential
function leads to a strong curvature for the linear scaling – this led to the term “threshold
voltage”; for silicon diodes this is often specified at 0.7 V. A scaling for large currents indeed
shows a sharp bend of the rounded characteristic at 0.7 V; a scaling for smaller current shifts
the “kink” to 0.4 V or even smaller values. Do note: an exponential function does not have a
kink – the value of the threshold is arbitrary!
Fig. 10.8.25: Pass-characteristic of a 1N4148 (data from Fairchild). The upper x-axis-scaling in the left picture
holds for the dashed line, the lower scaling for the solid line. Right: AA113 (Siemens).
The right-hand picture shows the forward characteristic of a Germanium point-contact diode
(AA113). In contrast to the silicon diode 1N4148, smaller voltage occur across the diode for
small currents in the forward-direction. For currents above 3 mA, however, the voltage
across the Ge-diode is higher than that across the Si-diode because spreading resistances
make themselves felt more.
For all these characteristics, we need to bear in mind that there will be scattering due to
temperature- and manufacturing-fluctuations. Increasing the temperature by 20°C may indeed
double the forward-current, and exchanging a diode for another of the same type (!) may
change the current flowing at a given voltage within a range of -70/+200%. It is therefore not
purposeful to count on highly specific data from the data sheets.

Fig. 10.8.26 shows further forward-characteristics (from data sheets). All diodes marked AA
are Ge-diodes, all others are SI-diodes.
Fig. 10.8.26: Forward characteristics of various semiconductor diodes (data sheet specifications).

10.8 Effekte 10-229
The diode currents flowing in typical circuits for distortion devices are small (max. about 1
mA). The characteristic curves (reaching over 100 mA in the data sheets) are therefore only
relevant in the lower range – which is relatively well described via the Shockley-equation. To
achieve a limitation of both sides of the signal, two diodes are interconnected in an anti-
parallel fashion; this results in a zero-symmetrical (odd-order) characteristic. In theory, that is:
manufacturing tolerances are felt considerably in real diodes. Driving the diode-pair by a stiff
current source results in a signal rounded off at both sides, as depicted in Fig. 10.8.27). This
ideal-current-source-drive is, however, not possible (and not necessary) in reality, since the
impedance of the driving current source cannot become infinite. It is helpful to use, for the
model, an ideal current source and extend the diode pair by a parallel resistor (left-hand
picture) – this changes the behavior in particular at small drive-levels. In all 3 pictures we see
the characteristics of two diode pairs: for the small-signal diode 1N4148, and for the power
diode 1N4003.
Fig. 10.8.27: Left: forward characteristics (half-log representation), diode with resistor in parallel. Middle: two
anti-parallel diodes. Right: voltage limiting for sinusoidal current input.
Describing the diode characteristics with only the Shockley-model requires merely two
parameters: the reverse current and the factor of the temperature-voltage (1…2). Both
parameters can be seen as scaling factors for the current and the voltage, and consequently the
following holds: within the framework of the Shockley-model all diodes show the same
behavior as long as variable gain is available at both the distortion device input and the
distortion device output. Working with this model does not require choosing a special diode
because every diode allows for the same distortion characteristics. However, this does not tell
us anything about the dynamic behavior, which is not described via a static characteristic
curve. A diode will go into the blocking state only once “all” charges have left the barrier
layer output, and this takes a moment: a relatively long moment for power diodes and a
relatively very short moment for RF-diodes. Partnered up with the distortion-introducing
diodes operating within the feedback branch of an operational amplifier, we often find an
additional capacitor in parallel to the diodes; this leads to the conclusion that it is in fact not
even desirable that the diodes act very quickly. In the Tube-Screamer, for example, there is a
50-pF-cap in parallel that will have an effect only in the highest frequency range – and only
with the gain turned up. In the Boss DS-1, however, we find some quite respectable 10 nF in
parallel with the diodes! This rather huge capacitance pushes the switching behavior of the
diode somewhat into the background. Moreover, we must not forget that even a capacitor of
this size impresses a diode only as long as very small currents are flowing. At e.g. 2 mA (a
current value that certainly may occur), the differential resistance of the diode is a mere 20 Ω,
and compared to that even 10 nF are of relatively high impedance.
Different diodes are connected in an anti-parallel manner if a non-zero-symmetrical

characteristic is desired – e.g. a Ge- and a Si-diode, or special parallel/serial-networks. The
individual characteristic becomes more important in this scenario, because it is not possible to
do an individual current/voltage-scaling anymore for each diode.

10.8.5.2 Transistors
With transistors, we need to distinguish between two crystal types (Ge, Si) and two types of
doping (PNP, NPN). Strongly simplified, the main differences♣ are: NPN-transistors require
(in common-emitter configuration) a positive operating voltage, and PNP-transistors require a
negative one. The typical base-emitter-voltage is 0.1 V for a Ge-transistor and 0.6 V for a Si-
transistor. There is a vast multitude of the most different transistors – and among them a
surprising number of compatible equivalents. PNP vs. NPN is, however, incompatible, as is
Ge vs. Si, even though there may be instances where the latter swap will work. An OC44 may
be exchanged for an AC151 without any problem, but an AC187 is incompatible with an
AC188. For not too old specimen of Europe-built transistors, the first letter in the designation
specifies the crystal-type: A for Germanium, B for Silicon. The second letter stands for the
recommended usage: C for audio frequency preamplifiers, D for audio frequency power
amps, F for RF-amplifiers, S for switching stages. The American (2N) and Japanese (2S)
designations do not allow for such a distinction.
Fig. 10.8.28: Transistor-schematics, connections.
Fig. 10.8.28 shows the circuit diagram for transistors, and the connector-pin assignment (seen
from below). Not all transistors have this assignment – in case of doubt the data sheets of the
manufacturers help. For an NPN-transistor, the current flows into the base, out of the emitter
and into the collector (the technical direction of the current-flow), and for the PNP-transistor
out of the base, into the emitter and out of the collector. The usual collector-current values in
distortion devices are smaller than 1 mA; the base current is about 1% of the collector current.
The quotient of collector current and base current, i.e. the current gain B, is strongly
dependent on the manufacturing process and the temperature. For commonly used transistors,
B is about 40 … 300. It is therefore possible that the behavior of a circuit changes if one
transistor is swapped for another (of the same type!).
In the idle state (i.e. without input signal) the collector current is about 0.1…1 mA. This of
course depends on the specific circuit – as example we assume it to be 0,2 mA. For B = 100
the base current will amount to 2 µA. The input, i.e. the “gate” between base and emitter, is
best described as a diode operated in the forward-direction – a Si-diode for the Si-Transistor,
and a Ge-diode for the Ge-transistor. A forward-current of 2 µA yields a forward-voltage of
about 0.1 V for the Ge-diode and of about 0.5 V for the Si-diode. Again, this is a first point of
reference – depending on the manufacturing process these values may vary. If the base-
voltage for an NPN-Transistor is more than 1 V larger than the emitter-voltage, that transistor
is shot. If UBE is negative for an NPN-transistor, the transistor will be in blocking mode, and
the collector-current will be approximately zero. The same correspondingly holds for a PNP-
transistor and negative base-emitter-voltages. The collector-current will, however, not be
exactly zero since a reverse current will still flow – in Ge-transistors this can reach sizeable
values. For example, the Siemens data-book specifies a reverse current of max. 200 µA for
the AC188 (for the emitter-diode in blocked state) – corresponding to the current in the
operating point for the above example! In addition, the reverse current has the unpleasant
characteristic of exponentially growing with increasing temperature. All this has created in
particular for Ge-transistors the image that they are solitary, hard-to-handle lone wolves.
♣
For practice-oriented details see e.g. Tietze/Schenk: Electronic Circuits – Handbook for Design and
Application; Springer.

10.8 Effekte 10-231
In Fig. 10.8.29 a PNP-transistor is operated in common-emitter configuration, i.e. with the

emitter connected to ground. If the transistor is in blocking mode, there will be (almost) no
collector current and the collector voltage will be -9 V (left picture). If the transistor is fully
on, there will be only a small voltage left at the collector of e.g. -0.2 V. As a first approach,
an operating point in the middle of the characteristic curve would be selected; the collector
voltage would be set to -4.6 V. From this, we obtain a voltage across the collector resistor of
4.4 V resulting in a collector current of -0.44 mA for a resistance of 10 kΩ♣. The base voltage
would be -0.1 mV in this example.
Fig. 10.8.29: Transistor in common-emitter circuit.
Both base and collector are not at 0 V without signal input, and a coupling capacitor each is
necessary for connection to the dc-free outside world. The base voltage required for the
operating point is set via the voltage divider at the base (middle picture). This circuit would
not support a stable operation, however, even if the operating point would be set individually
for each transistor specimen. With just a few degrees of temperature drift, the operating point
would shift, and the sound would change. The means of choice countering thermal drift is
negative (i.e. inverse-phase) feedback. This is implemented either via an emitter-resistor
(increasing the input impedance), or via a resistor from the output back to the base (lowering
the input impedance), or via other measures too extensive to be covered in the present context
[see e.g. Fliege]. The following pages will show examples of transistor-circuits employing
negative feedback – see e.g. chapter 10.8.5.3. Only with purposeful negative feedback,
multistage amplifiers such as the one in the above right-hand picture can be put together. In
the version shown, the first transistor would have to operate with too small a collector
voltage: since the base voltage of the second transistor can not grow above about 0.2 V, the
collector voltage of the first transistor is subject to the same limitation. This is why a resistor
(of e.g. 1 kΩ) is introduced into the emitter branch of the second transistor; this resistor
increases the input impedance and the input voltage.
Negative feedback decreases the gain but also stabilizes it, i.e. it becomes less sensitive to
fluctuations in temperature or due to manufacture. Circuits that need not operate down to a
frequency of 0 Hz allow for a separation of AC- and DC-negative-feedback. A strong
negative feedback for DC will stabilize the operating point, while at the same time a weaker
negative feedback for AC will ensure that the gain does not drop too far. One thing that needs
to be considered for all amplifiers is the phase-shift that occurs at high frequencies: it can turn
negative feedback into a positive one: the circuit may start to oscillate and inadvertently
become an RF-generator.
♣
The in fact quite important area of reference arrows and algebraic signs will not be elaborated upon in this
context – reference is made to literature, e.g. [20].

10.8.5.3 Range-Master (Dallas Arbiter)

One of the famous distortion devices from the early days of hard rock is the “Range-Master”
built by the Dallas-Arbiter-company in Britain. It was also called a treble-booster, although
it did not just increase the gain at high frequencies (as it happens when a “Treble”-knob is
turned up) but performs this job in a non-linear fashion, with a rich seasoning of distortion.
The circuit of the Range-Master (Fig. 10.8.30) is a simple as they come: the input signal is fed
via a relatively small 6.8-nF-capacitor to a transistor providing an amplification of a factor of
approximately 60. Since normal pickups generate several 100s of mV (even up to 4 V are not
impossible), this transistor is almost always overdriven. However, two peculiarities need to be
considered already at the input (U1): The input impedance is – at about 10 kΩ – rather low,
and for this reason the coupling capacitor has the effect of a strong attenuation of the low
frequencies. To calculate the cutoff frequency, it is not only the input impedance of the range-
master that needs to be taken into account, but also the pickup impedance that is part of the
mesh. 6.8 nF and 10 kΩ would result in a cutoff frequency of 2.3 kHz; depending on the
pickup this value drops to 1 – 2 kHz.
Abb. 10.8.30: Range-Master: circuit (left), old Germanium transistors (right).
The OC44 used in the Range-Master is a Germanium-RF-transistor from the dawn of solid-
state technology, and was probably available at very low cost – deemed outdated already back
in the day given the high-speed progress in technology. Even in fully conducting mode, the
collector current remains below 1 mA (for normal loads); the quiescent current flowing
through the output potentiometer is merely 0.2 mA. The corresponding base current is 1 – 2
µA, the base-emitter-voltage is smaller than 100 mV. The quotient of collector- and base-
currents (the current-gain B = IC / IB) has a large scatter-range due to manufacturing
tolerances; values of 50 … 200 are possible. Thus, the operating point is also subject to
scatter: typically, we find -6.8 … -7 V at the collector. For a new battery, that is – the power
source will also influence the transmission behavior.
Connecting the Range-Master to an amplifier input of 1 MΩ impedance, the output capacitor

(10 nF) of the Range-Master creates – in conjunction with this load – a high-pass with a
cutoff frequency of 16 Hz. Choosing the low-sensitivity input of the amp (typically 136 kΩ
input impedance) pushes the cutoff frequency up to 114 Hz. In total, two high-passes have an
effect: the first at the Range-Master-input, the second at its output. The emitter resistor is so
effectively bridged for AC that the resulting high-pass may be ignored: it creates no
attenuation of the bass frequencies. It may not be ignored, however, regarding its effect on
shifting the operating point. This happens when an input signal is present because now charge
reversals of the emitter capacitor take place. Due to the non-sine-shaped emitter-current, the
emitter-voltage shifts by about 0.2 V towards the negative at high drive-levels. Similarly, the
polarization voltage of the input capacitor changes: the asymmetric base current flowing at
overdrive-levels decreases the average voltage at the input cap. These shifts in the potentials
are the “secret” of the Range-Master – not the purportedly unique behavior of the OC44.

10.8 Effekte 10-233
An original OC-44 may be microphonic and/or be very noisy, its reverse current may be
beyond good and evil – still it is being traded at prices 50 times of what modern transistors
cost. What about the unique sound? That can be achieved with other transistors, as well. Of
course, the latter need to be PNP-Germanium transistors, and an individual check is warranted
for modern merchandise, too. An original OC-44, however, does not need to be put into the
Range-Master – the operational behavior of different transistors can, in fact, be surprisingly
similar. We may find large differences in the maximum ratings; for example with the
collector voltage: 20 V, or 100 V. The same for the maximum collector current: e.g. 10 mA,
or 2000 mA. And of course for the β-cutoff frequency: 150 kHz, or 10 kHz. However, all
these values are of secondary importance for a distortion device operated at 9 V. Important is
the current gain (static and dynamic) and the reverse current – and both these are not in any
way special for the OC-44. Why else could B. C. Meiser♣ recommend as replacement the
AC122, or the AC128, or – particularly suitable – the AC151. It should be noted that this
recommendation is not the writing of a blind amateur (as it often is the case in magazines),
but the well-founded opinion of a seasoned, experienced circuit designer. N.B.: the AC128 is
recommended in the data sheets “for slow switches” or for “small audio power stages”, the
AC122 for audio preamplifiers, and the AC151 for audio-preamps and driver stages. The
OC-44 was designed as RF-transistor for AM-radio usage … and still it can be swapped for
these other transistors. Of course, there are piles of other possible transistors – the special
sound is not due to a special transistor, but due to the asymmetric and drive-dependent
transmission characteristic (which in itself is not even that unusual).
Fig. 10.8.31: Range-Master: transmission characteristics; U0 ⇒ U1 (left), U1 ⇒ U2 (right).

This figure is reserved for the printed version of this book.
Fig. 10.8.31 shows, in its right-hand part, the transmission behavior from input (U1) to output
(U2); the operating point is indicated via the coordinate axes. For small drive-levels, the
correspondence between input- and output-voltage is approximately linear (short, straight
section of the characteristic); with increasing amplitude the characteristics grow longer and
curved, and shift to the left towards more negative input voltages. As already mentioned, the
reason for the shift is the charge reversal in the capacitors. Without the shift, limiting on both
sides would occur already at U1 = -0.2 V; with the shift it happens only at U1 = -0.4 V. Put
another way: the shift of the operating point renders the characteristic less symmetric and
emphasizes even-numbered distortion (for a sine-input). Another effect is the dependency of
the operating-point-drift on the input signal amplitude: it changes transmission parameters
more strongly than a fixed characteristic: the guitar-sound increases in liveliness. To
emphasize it again: all this is not a special OC-44-characteristic – every suitable transistor
will take care of these effects. But know this: many are suitable, but few are chosen …
♣
Gitarre&Bass, 01/2002.

In the left section of Fig. 10.8.31, the correspondence between U0 and U1 is show. A voltage
source was connected via 6.8 kΩ to the input of the Range-Master; the non-linear input
impedance of the latter is the reason for the curve in the characteristic. For a (passive)
magnetic pickup, the Range-Master represents a very special load: non-linear and of low
impedance such that the pickup resonance cannot manifest itself. Fig. 10.8.32 indicates the
effect assuming linear filtering: the low-impedance load attenuates the treble upwards of
about 1 kHz, the series capacitor attenuates the bass, and, from the pickup-source-voltage (in
this case a Strat) to the Range-Master input, we obtain a bandpass characteristic with a
center frequency of 1,25 kHz.
Fig. 10.8.32: Low-pass model of the loaded pickup (see Chapter 5.5.4). In the left picture, the Strat-pickup is
subjected to a real load of 10 kΩ, in the right-hand figure the load is the series circuit of 6.8 nF and 10 kΩ.
Fig. 10.8.33 depicts the non-linear distortion for a sine-shaped input signal. The lower half-
wave of the collector voltage is cut off first; the characteristic is not point-symmetrical and
the duty cycle therefore is not 50%. These measurements were taken at 500 Hz, and the input
capacitor was enlarged to 680 nF in order to be able to clearly separate linear and non-linear
distortion. For the regular operation (6.8 nF), linear and non-linear distortion interact.
Fig. 10.8.33: Range-Master: collector voltages for sine-shaped input signal; f = 500 Hz.
In the following table transistors are listed that may serve as replacements of the OC44; the
limit values are taken from data sheets (Va, Te, Si, and others).
OC44 AC122 AC125 AC126 AC128 AC151 2N508 2N527 OC71 OC75 OC77
kHz 150 15 17 17 15 15 45 35 10 8 3.5
mA 5 100 200 200 2000 200 100 500 10 10 125
V 15 18 32 32 32 32 16 45 30 30 60
Table: transistors comparable to the OC44 [B.C. Meiser, Gitarre&Bass 1/02]. It follows from the data variance
that practically every Ge-small-signal transistor is suitable; optimum-β = 80-110.

10.8 Effekte 10-235
10.8.5.4 Tube-Screamer (Ibanez)

Whether you want to call the Tube-Screamer a distortion device, an overdrive or a treble-
booster is a matter of taste – there is no fixed rule. The unit goes between guitar and amplifier,
and it has 3 sections: an impedance converter consisting of a transistor, a distortion section
with low-cut, and a tone filter. Fig. 10.8.34 shows – grouped around an operational amplifier
– the distortion-section components. A high-pass at the input takes care of a defined operating
point – its cutoff-frequency is so low that it has no impact on the frequency response. The
negative feedback circuit of the OP-amp includes the distortion-significant diodes. They have
merely a negligible effect at very low signal levels; here, the circuit operates as an amplifier
with a high-pass cutoff-frequency of 720 Hz. However, as soon as the voltage across the
diodes becomes sufficiently large such that (relative to the potentiometer) a significant
forward-current occurs, non-linear distortion starts to manifest itself, and the voltage across
the potentiometer is subject to limiting. In fact, the output voltage is composed of two parts
(potentiometer-voltage, and voltage across the RC two-pole), the consequence being that a
part of the undistorted signal is superimposed on the distorted signal. This is a peculiarity of
the Tube Screamer (and many similarly constructed devices on the market): it does not only
distort but mixes in a bit of the original signal. An easier-to-interpret equivalent circuit is
obtained by referencing the output voltage not to ground but to the input connection, and
compensating this via adding the input voltage to the output voltage (lower part of the figure).
Fig. 10.8.34: Tube-Screamer: small-signal frequency-response and schematic of the distortion section.
Now, the two-part output signal becomes evident: there is the inverted input voltage, plus the
(also inverted) high-pass-filtered, amplified and distorted input voltage. Of course, we arrive
at the same conclusion using Kirchhoff’s loop-rule, and assuming the differential input-
voltage of the OP-amp as zero.
The potentiometer controls the basic amplification of the distortion branch, but not the
amount of the distorted output signal, and not the amount of the undistorted signal, either. The
balance between distorted and undistorted output signal is pre-set and cannot be changed
without changing the circuit. If, for example, the amount of the distorted signal is to be
enlarged, 4 diodes instead of two could be included: two each in series and the two series-
circuits in an anti-parallel connection. Using one diode in one direction and two in the other
direction creates an asymmetric clipping with a sound that could be considered somewhat
fuller and assertive that that of a point-symmetric characteristic. Any preference will be a
matter of taste and can – in case of doubt – be changed for very little money. And while we
are in the process of changing diodes: a mixture of Ge- and Si-diodes can sound very
attractive, and even LED’s are deployed these days by Marshall (and everybody else) to
achieve distortion.

In Fig. 10.8.35 we see the time function of the OP-amp output voltages for three different
drive-levels. The clearly recognizable phase-shift is not the result of the high-pass at the input
but that of the 4.7-kΩ-47-nF-two-pole. The right-hand picture shows a type of distortion that
should be avoided at all cost: a piercing-through to the opposite voltage limit. The exact
reason (latch-up?) will not be discussed here, but if this happens, another OP-amp-type needs
to be brought in. The curves shown here were measured using a TL-072 that – being a FET-
OP-amp) apparently is susceptible to such effects. In defense of this actually very good
analog IC it should be said that this effect happens only at rather high drive-levels. But if it
indeed happens, the sound is so horrible that it probably is usable only as a special effect.
Fig. 10.8.35: Time function for different drive-levels (f = 500 Hz). UBatt = 9V.
In his Gitarre&Bass-article (11/01 – recommended reading!), B.C. Meiser lists several OP-
amp types suitable for the operation in the Tube Screamer (e.g. the NE 5532). He also points
to the fact that the NJM 4558 is not suitable. A fundamental problem of all pedal-type devices
is the requirement that they have to run off small battery-voltages. For the TL-072, the
recommended supply voltage is 30 V; in the Tube-Screamer it has to make do with a meager
9 V – and even this only for a fresh battery. The manufacturers do allow for smaller operating
voltages, but they do not specify which parameters will then deteriorate. If the specified
operating-voltage for the LM1458 is 10…36 V, 9 V is simply too little. The NJM4558 is
supposed work from 8 V – but how well will it do the job? In some data sheets we find: use
from 12 V. For the Texas RC4558 we read: from 10 V. With regard to the slew-rate, the data
given are: for the NJM 4558 = 1V/µs, for the RC 4558 = 0.5 V/µs, and for the MC 4558 = 1.5
V/µs. All these values are specified for 30 V supply voltage and not for 9 V. Trial and error is
the only way to find out how well (or how poorly) an OP-amp performs at 9 V; the data
sheets give too little information on this. Also, we need to consider that distortion – as it is
practiced in the Tube-Screamer – originally was seen as off-limits by the manufacturers.
Word has gotten out only rather late that an OP-amp needs to sound good also when
overdriven. So: try out some OP-amps – these ICs don’t cost a lot.
Fig. 10.8.36: Tube-Screamer-frequency responses: distortion unit, sound filter. Overall circuit. The transfer
function of the sound filter is easily calculated by Y-delta-transforming the OP-amp input circuit.

10.8 Effekte 10-237
10.8.5.5 Fuzz-Face (Dallas Arbiter)

As a typical representative of the group of brute-force distortion-devices, we will in the
following analyze the Fuzz Face, a small battery-powered effects device offered by Dallas
Arbiter♣ from 1966. In the original version, 2 Germanium transistors (AC 128) took care of
gain and distortion (Fig. 10.8.37); the output voltage could be controlled by the “Volume”-
potentiometer while the “Fuzz”-potentiometer adjusted the degree of the distortion.
Fig. 10.8.37: Fuzz-Face: circuit (left), pickup-frequency-response (right, compare to Chapter 5.5.4) Details re. the
circuit-board construction (and replication) are described by Martin Thewes in Gitarre&Bass, 09/2009.
The circuit is peculiar – starting with the input: due to the current-feedback (100 kΩ), the
input shows very low impedance. For the usual pickup, it practically presents itself as a short,
especially when the 1-kΩ-potentiometer is set such that the tap is connected to ground. For all
measurements described in the following, the generator providing the input signal was
connected via a 6.8 kΩ resistor. In this configuration, there is almost zero input voltage – this
does not mean, however, that the circuit is not receiving any drive signal. In fact, the input
operates under current control as a so-called “zero-ohm-node” known from recording-studio-
technology. Due to the frequency-dependent source impedance of a magnetic pickup, the
result is a veritable low-pass radically attenuating all treble above 500 Hz. The treble is
revived, however, in the form of strong non-linear distortion-products generated via the high
gain-factor of 100…2000 (Fig. 10.8.38). Because of the current-control, the gain must not be
referenced to the input voltage, but to the quotient of collector voltage (T2) and generator
voltage ahead of the 6.8-kΩ-resistor. This resistor is required due to the small input
impedance; it models the pickup-resistance. Whether the resistor has a value of 5.2 kΩ or 7.3
kΩ is of no importance. As was the case for the Range-Master, charge-reversals in the
capacitors cause shifts in the characteristic curve – this to a somewhat lesser extent but with
the same tendency.
Fig. 10.8.38: Fuzz-Face: collector-voltage of the 2nd transistor for different drive-levels (500 Hz).
♣
First under the "Arbiter Electronics"-moniker, then under "Dallas Arbiter".

Translator’s note: the following page contains a satirical send up of some of the absurdities
occurring when transistors and tubes are discussed among guitar “experts”. Much of the
satire relies on the German language and culture, and it is all but impossible to do a simple
translation. Even the title accordingly plays with words: in German, the word (substantive)
for “tube” is “Röhre”. Incidentally, there is a similarly spelled verb in German: “röhren”
that is not related at all in its meaning which is “to roar”. The title of this page – literally
translated – would thus be: “As the transistor roars”. The translated intended pun “As the
transistor tubes” does not work, of course, as it does in German where the terms for “tube”
and “to roar” both make a kind of sense in the context.
Therefore this page is left un-translated. Reading it is recommended to all who have a little
command of the German language. We might consider translating it at a later point, after all
– and even find a native speaker who can come up with correspondences and matches that
really trans-late the intended meaning into American or British context and culture.
10.8.5.6 Wenn der Transistor röhrt

Heute besuchen wir, die Guitar-Licks-und-Tricks-Redaktion, einen Exponenten der deutschen Verstärker-Szene:
Markus Dampfmeister, den Kölner Amp-Wizard. Mit seinen Marshall-Mods war er schon Anfang der 80er
aufgefallen, ist gar bis Straubing gekommen, seither hat sich sein Ruf sogar über die Landesgrenzen hinaus
verbreitet. "Markus, ich darf doch Du sagen?" "Natürlich, wir sind ja so eine Art Kollegen. Was liegt an?" "Wir
wollten ein altes Thema aufgreifen: Röhren für die Gitarre, oder Transistoren?" "Röhren, keine Frage. Transis-
toren geben dem Klang eine kalte Sterilität, deren Dimension man schnell erahnt, wenn man nur mal den Finger
auf eine Röhre legt. Dieses Urfeuer, das da im Innern brennt, dieses Elektronenbombardement der Anode, das ist
es, was den heißen Ton ausmacht. Mit Transistoren ist das nie erreichbar, die werden, wie der Fachmann sagt,
schon bei viel niedrigeren Temperaturen eigenleitend." "Und dann leiten sie?" "Nein, das heißt, schon, aber dann
sind sie kaputt. Röhren halten da viel mehr aus, die kann man so quälen, dass das Anodenblech glüht. Den damit
verbundenen Höllensound bringt der Transistor einfach nicht, einen glühenden 2N3055 hat noch niemand
gesehen. Und schon der Fachbegriff: Halbleiter! Gibst Du dich mit halben Sachen zufrieden?" "Nee, drum sind
wir ja hier, um endlich einmal die volle Packung zu bekommen. Also Röhre?" "Nur! Die alten sind die besten!
>Hast Du Tungsol in dem Fender, gibt’s nur eins: Return to Sender. Dieses Mumpfen, dieses Dröhnen –
schauerlich, zum Abgewöhnen. Ist es aber eine Mullard, kriegst du einen Sound, der pullert<. Ich habe vor
Jahren alte UK-Bestände aufgekauft, die ich jetzt gegen echtes Geld in meine Customs einbaue." "Und Röhren
haben keine Nachteile?" "Nee. Sie rauschen sogar weniger als die Transistoren. Weil: Die ganzen
zwangsläufigen Verschmutzungen, die im Silizium (Sand!) unvermeidlich sind, die hat die Röhre nicht. Sogar
die Luft wird rausgesaugt, damit ja nichts stört. Und nochmals: Die alten sind die besten – damals war einfach
auch die Umwelt noch nicht so versaut, da konnten sie noch hochwertige Materialien verbauen. Ergo: Sound
pur."
Der Mann wusste, wovon er redete, das war noch einer der alten Schule. Thorben, unser Fotograf, hatte bisher
Pics von der Location geschossen, nun meldete er sich plötzlich mit einer Frage: "In den Verzerrern sitzen aber
Transen, oder?" "Ja, drum rauschen die auch so stark. Da muss man aber höllisch aufpassen, da gibt’s Si und
Ge. Ge-Transistoren klingen gut, Si kratzt und sägt. Bei Si ist nämlich der sog. Bandabstand größer, da tun sich
die Elektronen schwerer um rüberzuhüpfen. Mit einem alten OC44, der schon mal für 10 Euro gehandelt wird,
lebt jeder Verzerrer auf, da entstehen einzigartige Klangwelten, das lebt, das röhrt, wenn man so sagen will."
"Aber vor einem Monat waren wir beim BCM, und der meinte, anstelle eines OC44 kann man locker einen
OC71 reinstecken. Oder einen OC75." "Ja, das sind alles alte Germanen, da ist praktisch dasselbe drin, nur mit
anderer Aufschrift." "Oder einen AC122, oder einen AC128..." "Ja gut, was der BCM wo reinsteckt will ich jetzt
nicht kommentieren, wenn das der BCM meint ... das sind europäische Vergleichstypen." "Oder einen OC76,
oder einen OC77. Ganz gut soll der amerikanische 2N508 sein, sagt der BCM, oder der 2N527 ... ich versteh die
Sache mit der Einzigartigkeit noch nicht so ganz." Thorben, der Schrecken der Redaktion, immer gut für einen
Eklat. Markus verlor zusehends die Lust, über die Problematik ubiquitärer Einzigartigkeit nachzudenken, er
wollte schließlich seinen neuen Custom in diesem redaktionellen Beitrag unterbringen, in Zeiten wie diesen ist
man um jede kostenlose Werbung froh. Doch an Thorben führte kein Weg vorbei: "Ich hab noch nicht
verstanden, warum einerseits die glühende Röhre für zerrende Höllensounds am besten sein soll, andererseits
der Clapton (und nicht nur der, d. Red.) vor seinen Röhren-Amp einen Range-Master gesteckt hat, in dem doch,
wenn ich mich nicht täusche, ein Transistor sitzt?" Das fasste Markus nun allerdings als unfreundlichen Akt auf,
der Talk endete unplanmäßig. Schade eigentlich, ich hätte noch Fragen gehabt. Macht aber nix, nächsten Monat
geht’s zum Piepenbrink, von diesen Exoten hat's ja noch mehr.

10.9 General operating characteristic 10-239
10.9 General operating characteristic
The topic of the previous chapters was the performance of individual tube stages or individual
circuit sections; in the following the focus shall be shifted to higher-level considerations. The
operating characteristic of a guitar amplifier can be looked at from two different angles: from
the point of view of the circuit design (i.e. how does the circuit function?), or from the point
of view of auditory acoustics (i.e. how does the amp sound?). Of particular interest is to
causally interconnect the two approaches – that, however, is also the most difficult task.
10.9.1 Tube sound vs. transistor sound
In fact, transistors seem to have only advantages over tubes: they are smaller, cheaper, have
no fragile glass containers, do not need heating. Apparently, they have the single disadvantage
that guitar amps designed using them do not sound good. Of course, this is a highly subjective
judgment, and of course there are other opinions – however, in particular the early transistor
amps had few advocates, and aside from all mysticism there are without a doubt differences
from the point of view of systems theory. Still, there is no “tube sound” as such, just as there
is no “transistor sound” as such. A guitar amplifier does not sound better merely because it is
fitted with tubes, and a transistor amp does not need to inherently sound bad. It might, though.
Fender’s Solid-State-Series – the one advertised in 1968 with 'superb sound' letting you skim
the waters of musical greatness – was rather unsuccessful. 'Curious refrigerators' was the
term used by the German Gitarre&Bass-magazine in their Fender special edition. To vindicate
the name “Fender”, one could argue that “this wasn’t Fender, this was CBS”; however, after
the sale of his company to CBS, the same Leo Fender designed and produced (with his new
company Music-Man) hybrid amplifiers featuring a transistor pre-amp and a tube power-amp.
These amps – at least today – by far fail to achieve the fame and glory of Fender’s black-
faced heroes. The same with VOX: the company did not become famous with the
transistorized Defiant, but with the all-tube AC30. Guitarists and transistor amps: not a love at
first sight.
Transistor amps sound sterile, impersonal, lifeless, they buzz, crackle, sound scratchy, and on
top of everything, at the same wattage they are not as loud as tube amps. These subjective
judgments elude any circuit analysis. Who wants to stipulate how a guitarist should perceive
his guitar sound? Plus: even if this is pure imagination, it is easily conceivable that this kind
of imagination has repercussions on the virtuosity. Electrical engineering with its many
disciplines is actually only challenged when causal links are brought in: 'hot tubes for a
warmer sound', or: 'tubes do a rounder limiting and thus sounds less sharp', or: 'tube amps
sound better because 2nd-order-distortion dominates in them'. Still, it is not that simple. If the
audible differences in sound could be traced to a single reason, we would probably see
exclusively transistor amps today. As is the case for public address systems: who would make
the effort to stage several hundred tube power amps? But nine tube amps lined-up behind a
guitar player: even today, that is not very strange. “Very loud” in the case of nine AC30, and
“VERY LOUD” if six Super-Twins are stacked to a pyramid. Why do they do that – what is
the secret of the tube? With such a presumptuous question, the answer can only end up in
hybris … anyway: the secret i.e. the undiscovered country is in the diversity, in the interaction
of a multitude of non-trivial components and characteristics, respectively.

Harmonic distortion, slew-rate, frequency response, input- and output-impedance, shifts of

operating points … and all that in combination. What is the effect of the level-dependency of
the 4th-order harmonic-distortion on the sound? Does the 4th-order distortion have to be even
considered at all, and if yes, up to which order are distortions relevant? To measure the
distortion is simple, but to determine its effect on the sound is difficult. For one,
comprehensive auditory tests are required, and then each judgment is dependent on many
boundary conditions: on the setting of the tone controls, on the loudspeaker, on the listening
location, on the guitar, and of course on the generated tones. Because this variety of
parameters is vast, the developers of transistor amplifiers do not only have to develop circuits
but also “survival strategies”.
One of these strategies is: as long as the frequency response is identical, the sound has to
correspond, as well. That thought is too simple. Here’s another: since it is unknown how the
characteristics of each individual tube stage affects the sound, every detail of the tube circuit
needs to be modeled. Sounds as if it would be on the safe side – and it would be if indeed
every detail were known. And another variant: we tune and retune until everybody is satisfied,
even if the new circuit has no relations to anymore to tube circuits. Possible, but easily
subject to diffuse criticism: something is missing! No one knows exactly what it is that’s
missing, but everybody is convinced: that is not the ideal tube-sound. Or the opposite
approach: the designers are happy (sounds quite good and doesn’t go up in smoke anymore),
the management as well (remained even 2% below the pre-calculation), and the sales
department agrees (finally they finished it). It’s juts that not only the calculation but also the
turnover is below plan. It is a difficult market: the original Bassman is a legend but the Music-
Man is not. Despite the fact that behind both the mastermind was Leo Fender.
This book with its focus more on guitars will not answer the question which type of distortion
will render the sound sparkling-creamy-wooden-throaty – that topic belongs to a book
exclusively dedicated to amplifiers. Still, there is room for a few basic thoughts. The previous
chapters dealt with the non-linear behavior of the amplifier; from the point of view of the
author, this is a main theme. In tube amplifiers, several linear and non-linear systems interact:
high-passes in the coupling-capacitors, in the output transformer and in the loudspeaker, low-
passes in every tube, in the output transformer, in the loudspeaker. In every tube, in the output
transformer, and in the loudspeaker we also find non-linearity. It all makes for an almost
unfathomable system – even without negative feedback. It’s not that we couldn’t describe the
individual sections of the system – it is the overall judgment that is so difficult. Something
that is routine in the LTI-system develops into a vast problem for coupled non-linear systems.
For example, it is only in the linear system that it makes no difference whether filter-poles are
realized in the electrical of the mechanical domain – here we can compensate e.g. a treble-loss
of a loudspeaker by an electrical filter. If indeed linear behavior is desired, equal-sounding
amplifiers can easily be built with both tubes and transistors. With non-linearity entering the
picture it gets very complicated, though.
After several decades of searching for the right sound, transistorized guitar amps today have
matured to the point where the acceptance can to said to be good. Nevertheless there are still
innumerable tube amps on the market, and many guitarists are likely not to buy anything else
for decades to come, ready to invest those 200 € now and again in a quartet of tubes. The
manufacturers have learned from the mistakes of the early years, and offer well-sounding
circuits and amps. However, the guitarists have also educated themselves and now can hear
details that 50 years ago would have been classified as unsubstantial.

What adds to the problems that non-linear amplifier circuits can pose are psychometric
issues: how do we measure auditory perceptions? Here, the range starts with “plug in, turn up,
listen” and ends with round-robin-tests carried out on a global scale. For guitar amps, we
mostly find the former experimental method: listening test in the store, in the rehearsal room,
in the editorial office. The results of such test are often ignored or contested on the side of the
acousticians (who typically have a scientific background) – not so much because of the
involved jurors (normally not that wee known) but because of the unscientific approach
lacking objectivity and reproducibility. Is the hand-wired boutique-amp praised primarily
because it hails from California and sets you back 5000.- €, as every person testing the amp is
made to know first of all? Would the amp be as convincing if it would remain unknown and
hidden behind a curtain? We find a nice example for such a scenario described by Uli
Emskötter in SOUND-CHECK magazine [issue May, 2000]: in a test of guitar cables, all
involved perceive “pronounced differences in sound”. A week later, in a repetition – this time
as a blind-test – shows: “bewildering result – the judgment came out as entirely different”. It
is nothing new to psychologists that the type of presentation procedure will influence the
result of perceptional tests. The insights these experts have regarding experimental
methodology are beneficial for listening experiments, as well (see Chapter 10.9.4).
Guitarists not only perceive the sound of their guitar but they also evaluate it. For the
perception process, a relatively small inter-individual variance may be assumed, however the
evaluation process always depends on various boundary conditions. We all know this: first
the new CD by the latest superstar is lauded to be among the 10 best releases of the year, next
thing we know it appears in a TV-show listed among the most embarrassing oeuvres.
Although the auditory event remains exactly the same, the evaluation of it changes. Another
example that every studio musician is familiar with: you do a mix-down, find a suitable
mixer-setting, everybody is delighted and calls it a day. The next day you listen again –
without any changes to any setting – and everybody is disappointed: the vocals are too loud,
the drums are trebly, the bass too fat … or the other way round. The reasons for this change
are rarely found the technical issues (loudspeakers cooled down, humidity different) but with
all likelihood the difference is found in the changed judgment-standard. The underlying
processes may develop over minutes or even hours, but time-invariant, systematic differences
(bias, offset) are also known: there is a tendency to set a value controlled by the subject to
high [12, loudness scaling].
Here is an episode showing how much our value judgments are affected by cognitive
processes. It may be a singular case, but is likely to happen quite often in a similar way. After
a gig, a young musician addressed me speaking in highest terms of the "super-sound" my
guitar-rig had: "You can’t beat the good old AC-30 – that’s pure tube-sound." I am sure the
people at VOX would have loved to hear that – although I wasn’t using an AC-30. At the
beginning of the 21st century, this legendary amp is not as ubiquitous as it used to be, and the
younger generation apparently is not as familiar with it. Indeed it was a VOX I was playing
through, as the gilded badge on the front of the amp confirms, however it rather was a AD-60-
VT. That amp features a transistor-preamp and a transistor-supported 1-W-tube-output-amp.
It doesn’t sound that horrible, either, in fact it sounds pretty darn good, and its AC-30TB
model cooperates most harmoniously with the Historic Les Paul. In any case, associating the
terms VOX = AC-30 = tube amp = super sound appears to be hard-wired into many (though
not all) a musician’s brain. Had the amp-label not read VOX but Solid-State-MOSFET, the
judgment could easily have been “doesn’t sound too bad – for a transistor amp”. Indeed the
psyche plays an important partner-role in the wide and colorful world of psychophysics. The
psyche’s counterpart in this area, i.e. physics, and more specifically circuit technology, will
now get some attention as well.

From the many amplifier circuits we chose a Fender- and a Music-Man-circuit (Fig. 10.9.1)
because Leo Fender was involved both. He probably was not the only designer but at the very
least the responsible patriarch. Starting out from the RCA-application-notes, the circuit of the
Twin-Reverb – as it is presented here – was developed over the years into an internationally
recognized standard that inspired competitors, as well. Once the era of the octal-tubes had
passed, noval-tubes entered service at Fender in the mid-1950’s, and in particular the high-
gain 7025 (and its colleagues, the 12AX7 and the ECC83) won the pole-position that they
never relinquished again.
The Fender-circuit starts with a triode-

input-stage of a gain factor of about 40.
Tone-filter, volume control and inter-
mediate amplifier follow. The Music-Man
circuit shows considerable similarity
although the input OP-amp has only a gain
factor of 23 – a concession to its lower
supply voltage that leads to earlier
Fig. 10.9.1a: Fender AA763 (Twin-Reverb). limiting. Despite small differences, the
effects of the tone-filters are comparable.
The volume pot, however, is connected
not directly after the filter but inserted
after the intermediate amp. First, these
circuits will be compared regarding their
linear behavior.
Fig. 10.9.1b: Music-Man 2100
Differences show up already at the input: the MM has lower impedance than the TR: the
pickup-resonance receives a stronger dampening. On the other hand, the input capacity is
lower in the MM (do consider the Miller effect!). The series-capacitance in the MM-input has
barely any effect on the signal, and no shift of the operating point need to be feared, either
(due to the symmetrical limiting in the OP-amp input). The 50-pF-capacitor is there to reduce
the gain at high frequencies, it has an effect from about 10 kHz The 1-µF-cap reduces the gain
at very low frequencies (below 20 Hz). Compared to the TR, the impedance level in the MM-
tone-filter is lowered by a factor of 10 to normal OP-amp-typical values. Disregarding this
change, the two tone-filters are indeed very similar, despite one missing capacitor in the MM
– however such variants with only two capacitors did exist at Fender, as well (e.g. the Super-
Amp, see Chapter 10.3). These circuits were modified again and again.
The small-signal transmission factors of both circuits are shown in Fig. 10.9.2 (referenced to
1 kHz for both). This similarity is not likely to have been an accident; rather the Fender circuit
will have been the given objective. The only significant difference in the small-signal
behavior is the different input impedance; we can only surmise that the design process was
possibly checked with a low-impedance generator such that this aspect did not become
apparent. Large differences are apparent, however in the behavior for strong signals i.e. at
high drive-levels.

Fig. 10.9.2: Frequency responses referenced to 1 kHz. Bright-switch in “off”-position for both cases.
The drive limit of the amplifying element (tube, OP-amp) is the significant factor for the
behavior at high drive-levels. Differences pop-up right at the input: in the TR we have grid-
current distortion that is not there in the MM. From about 5 kHz, the MM shows slew-rate
distortion but the TR does not. Highly significant: the TR has the volume control positioned
after the first tube while the MM has it only after the second OP-amp. With the treble-control
turned up fully, the gain factor in the MM from input of the first OP-amp to output of the
second OP-amp is about 126. To maintain distortion-free operation, the input voltage must not
increase beyond 70 mV. For the TR, this is quite different: assuming 35 V as limit of the first
tube for hard clipping, the permissible input voltage would be about 900 mV. Having said
that: as already mentioned in Chapter 10.1.4, it is difficult to compare tube- and OP-amp-
distortion. Below the clipping-limit, the OP-amp works practically distortion-free, while for a
tube, distortion rises continuously across the drive-level-range. Fig. 10.9.3 shows, for the
MM, the maximum input level for undistorted operation. Especially in the brilliance-range (3
– 5 kHz) that is so important for Fender guitars, distortion can very easily occur even though
the volume pot may turned up only a bit. In Fender amps, the Bright-switch most often is
located at the volume control, but for some MM-amps this is included into the negative
feedback of the first OP-amp – possibly to reduce OP-amp-noise. Switching-on the Bright
switch in the latter case further decreases the treble headroom (right picture), independently of
the position of all tone pots and of the volume control. This marks a difference to the Twin-
Reverb and to similar Fender amps. The MM-amps therefore do show clear differences in
their behavior at high drive-levels compared to typical tube amps.
Fig. 10.9.3: Music-Man: maximum input level for undistorted operation. Left: solid line = tone controls as in
Fig. 10.9.2, dashed line = treble control turned up fully. Right: amplifier in which the Bright-switch changes the
gain of the input-OP-amp. Hatched area: tube input-stage.

10.9.2 Tube-Watt vs. Transistor-Watt
Allegedly, tube amps are louder than transistor amps rated at the same output power. Before
we discuss this topic on a scientific level, we first need to establish what exactly is meant with
“this is a 50-W-amp”. Not meant is the power consumption i.e. the power drawn from the
mains. Rather, such a specification always refers to the output power fed to the loudspeaker.
If the speaker impedance were frequency-independent and real, this power could be stated
without any issue. However, the loudspeaker impedance is frequency-dependent and complex,
despite the simple 8-Ω-label. In order to still be able to specify a number of watts, an ohmic
resistor replaces the speaker, and it is for this resistor that the indicated number of watts holds.
In other words, the manufacturer specifies that an amp generates e.g. 50 W at 8 Ω. This does
not in fact tell us how much power this amplifier can deliver into an 8-Ω-loudpseaker,
because an 8-Ω-speaker does not have 8 Ω at all frequencies (Chapter 11.2).
In an 8-Ω-resistor, an alternating current of the RMS-value of I = 2 A generates the RMS-

voltage U = 16 V; the product of current and voltage yields the power: P = 32 W. In order to
clarify that these are RMS-values, a tilde is often put over the formula symbols:
; ;
For an RMS-current of 2 A, the matching current amplitude is 2 A (i.e. 2,8 A), and
correspondingly the voltage amplitude of 22,6 V matches an RMS-voltage of 16 V.
Multiplying the amplitude-values rather than the RMS-values yields double the power: a 32-
W-amp turns into a 64-W-amp. This (higher) wattage specification is not in use in the
professional audio technology – rather, the nominal power calculated from the RMS values is
specified; in the example this is P = 32 W. What does this power depend on? Its factors are
e.g. the squared RMS-voltage, and a more or less arbitrarily defined nominal resistor R that
initially replaces the loudspeaker. The resistor is defined as fixed quantity in the data sheet;
the voltage is, however, variable. So, for which drive-levels do we specify the nominal
power? For studio- or HiFi-equipment, the largest voltage just below distortion level is used,
or the voltage at which a certain total harmonic distortion (THD, to be specified) occurs: e.g.
32 W at 8 Ω and for k = 1%. For a guitar amplifier, such a THD-specification is not possible,
and therefore – for a sine-shaped drive signal – the output voltage is visually judged to specify
at which level clipping occurs. Again: for the calculation this limiting voltage may not be
substituted into the formula because it represents the amplitude (i.e. the peak voltage). Rather,
this limiting voltage needs to be divided by . Alternatively, the amplitude is used, and the
calculated power is then divided by 2. As an example: for an 8-Ω-resistor, clipping occurs at
40 V. The resulting RMS-voltage is 28,3 V, and the power is calculated to P = 100 W.
Alternatively: .
Incidentally, it is not sufficient that a loudspeaker box connected to a 100-W-amplifier can

withstand merely 100 W. Since guitar amps are typically overdriven, they generate more than
the specified nominal power. Given that the nominal power mentioned in the above example
is independent of the load, for a square-shaped signal the power would be double, i.e. 200 W!
This is because square- and sine-shaped signals of the same peak-value differ by a factor of
in their RMS-values.

The limiting-voltage (i.e. the voltage at which the output voltage starts to clip) is, however,
not entirely independent of the load because the internal impedance of the power supply is not
zero. The power supply furnishes the operating voltage to the power amplifier – for a tube
amp e.g. 450 V. This is a dc voltage the value of which depends on several parameters: on the
mains voltage, on the power transformer, on the rectifier, and on the load. In the unloaded
state, the operating voltage has its maximum value but buckles (“sags”) under load, i.e. as the
amplifier feeds power to the loudspeaker. This has a very simple reason: the current flowing
to the power amp first needs to pass through the mains-transformer and the rectifier – and
either of them causes a voltage drop. The exact voltage and current time-curves are anything
but simple to describe (these are coupled non-linear systems), but we do not need to examine
this very precisely here. With a load connected, the operating voltage buckles and decreases,
e.g. from 450 V down to 400 V, or even down to as low as 360 V. Given a large mains-
transformer and an efficient silicon-rectifier, the voltage drops only little; with a small
transformer and a tube-rectifier the drop is larger – this is another genre-typical difference.
Massive 100-µF-capacitors make the “sagging” (as well as the subsequent recovery) slower
than the (from today’s perspective) puny little 16-µF-caps. Here we actually may have a
difference between tube-Watts and transistor-Watts: modern transistor amps often have very
“stiff” power supplies, i.e. power supplies with a small internal impedance the voltage of
which decreases only little as a load is connected. Tube amps (especially if they are from back
in the day and carry a tube rectifier) have power supplies with comparatively larger internal
impedance (see Chapter 10.1.6). Of course the two aspects are not necessarily connected to
each other: a tube power amp could just as well include a power supply with low internal
impedance – but in particular the legendary amps do not. For a guitar-note played after pause,
the full charge of the power-supply-cap is available during the first instant. The limiting
voltage may e.g. be 40 V yielding 100 W of nominal power into 8 Ω. However, the voltage
buckles after a few milliseconds and the limiting voltage drops to e.g. 35 V. With the power
being in a square-dependency to the voltage, the power decreases to 77 W. Measuring the
nominal power with a continuous sine-tone yields the second value, i.e. 77 W. For a transistor
amp fitted with a “stiff” power supply, the limiting voltage may decrease e.g. only from 37 V
to 35 V, so that both amps have the same nominal power. For an impulse, i.e. as a string is
struck, the tube amp does however have a higher power; in the example it is 100 W rather
than 85 W. In case the limiting voltage of a tube amplifier does not only decrease by 12,5%
but by 15% or 20%, these differences become substantially larger.
Thus, one difference in the power yielded by tube- and transistor-amps relates to the temporal
behavior: the “attack” is delivered with more power in a tube amp. This holds for the generic
circuits – of course it could be designed exactly the other way round. Consequently, the
theorist is of course right as he states: “there is no difference between tube-watts and
transistor-watts; watt as the unit for power is universally standardized”. However, in just the
same way the musician is correct in perceiving his or her tube amp as louder. It is not the unit
of measurement that is different but the measurement process. A second difference is found in
the resistance of the loudspeaker that is not constant, but frequency-dependent and complex.
The magnitude of this complex resistance, the impedance, may easily reach 20 Ω or 30 Ω at
certain frequencies although the loudspeaker is specified at 8 Ω. Not only the copper-
resistance of the voice coil contributes to the impedance but also the inductance of the voice
coil and the moving mechanic component as they are transformed into the electrical domain
(Chapter 11). At the resonance frequency, the loudspeaker assumes high impedance, and the
same happens at high frequencies.

Fig. 10.9.4 shows the frequency responses of some guitar speaker boxes: the changes with
frequency are most obvious. It is rather up to the manufacturer which impedance value he
specifies for the respective box. There are indeed standards for this, however the musician and
the manufacturer do no actually shake hands over a sales-deal based on specific DIN- or
ANSI-norms. For the following consideration, we simply assume the loudspeaker impedance
to be 8 Ω at one frequency, and 16 Ω at another frequency. If the amplifier has a transistor-
typical “stiff” power supply and features an also transistor-typical strong negative feedback,
the output voltage will be impressed i.e. almost independent of the load. With a 16-Ω-load,
the amp will merely be able to feed half the power that it can generate in an 8-Ω-load. The
situation is very different for a tube amplifier: operating it without a speaker could even cause
flashover at the power tubes – the voltages that may occur are that high. The tube amp is not
actually a true current source, but it does feature higher internal impedance compared to a
transistor amp. This has consequences on the power delivery. For example: an amplifier with
8 Ω internal impedance feeds P1 = 50 W into 8 Ω and P2 = 44 W into 16 Ω. A (transistor-)
amp with 0 Ω internal impedance would generate 50 W and 25 W, respectively. As the
loudspeaker impedance increases, the power delivered by a transistor amp will decrease more
strongly than for a tube amp. Again, the exact calculation is rather complicated because linear
behavior (internal impedance) and non-linear behavior (limiting voltage) interact, and also
because not a sine-tone but a guitar-signal drives the amp. Still, the statement remains: your
typical tube amplifier will generate on average more power into a loudspeaker than a
transistor amp having the same nominal power rating.
Fig. 10.9.4: Frequency responses (impedance) of typical guitar speakers; measured in a reflecting environment.
As an example we will look more closely at the frequency response of the speaker impedance
of a Marshall 1960 AX speaker. It is specified at 16 Ω, its minimum impedance is 15 Ω.

Z reaches its maximum (70 Ω) at 130 Hz. A transistor amplifier rated for 16 Ω and fitted with
a (ideally) stiff power supply will feed into 70 Ω merely 23% of the power that it could feed
into 16 Ω. In reality, the power reduction will not be as pronounced because the supply
voltage will sag less at increasing load-impedance – a reduction to “only” 30% is nevertheless
quite drastic. A tube amplifier will behave quite differently: if it is specific for operation with
16-Ω-load, as well, we could expect 60% of the power at a 70-Ω-load, after all – double of
what the transistor amp could generate. With a tube amp, the Marshall box will emphasize the
frequency range around the resonance frequency, and it will reproduce the treble range more
strongly. This tendency cannot be compensated for in the transistor amp by increasing the
gain at high frequencies (e.g. by turning up the treble control) because that measure does not
influence the maximum power delivery.
Fig. 10.9.5: Maximum available power dependent on the (ohmic) load resistance. Nominal impedance: 8 Ω. Left:
typical transistor power-amp. Right: typical tube power-amp. Dashed: model-calculation.
In Fig. 10.9.5, the maximum available power is shown for a typical transistor power-amp and
for three tube power-amps, respectively. “Maximum power” means total overdrive. The
transistor amp is specified for 8 Ω; for lower loads the amp shuts down. The tube amps are
also specified for 8 Ω but can deal with lower as well as with higher load impedances. For the
transistor amp, a load-independent imprinted voltage was used as idealized model, while for
the tube amps a constant internal impedance of Ri = 8 Ω was assumed. When discussing the
internal impedance of a power amplifier, we need to distinguish between linear and non-linear
behavior: during linear operation (no overdrive), the typical transistor amplifier features a
very small internal impedance (e.g. 0,1 Ω or even less), while a tube power amp without any
negative feedback (such as the VOX AC30) possesses e.g. 80 Ω (there are several variants).
The AC30 therefore emphasizes already in its linear operational mode those frequency ranges
where the loudspeaker features high impedance. In non-linear operation, the internal
impedance can only be defined using special model laws; the dashed line in Fig. 10.9.5 was
calculated for tube amplifiers and Ri = 8 Ω. Again, the frequencies of higher speaker
impedance are emphasized although not as much.
The VOX AD60-VT realizes an interesting concept: this guitar amp uses a weak double
triode (ECC83) as push-pull power amp and supplements the missing power via transistor-
support. The peculiarity here is that the speaker impedance influences the power that the
power amp is able to muster. The lone tube is not included as an alibi, as both power-
measurement and listening tests prove (Fig. 10.9.6). What is not advertised as loudly, is that
in the AC-30-power-amp, pentodes (EL84) do the work while in the AD-50 VT, triodes are
on the job. They do, however, this job with very good success.

Fig. 10.9.6: VOX AD60-VT: standardized maximum power (left), frequency response of the impedance (right).
Fig. 10.9.6 depicts the maximum offered power dependent on the load resistor. The power-
range selectable via a switch is the parameter. The power-characteristic is not really identical
to that of a tube amp but the result is quite easy on the ears. Relative to the 8-Ω-reference, a
boost can be seen and heard in the frequency ranges with higher speaker impedance – this
boost is even stronger than that found with a “true” tube amp. In the 60-W-mode, the power
maximum is at 16 Ω – this supports an operation with a (serially connected) second 8-Ω-
speaker: more power is available although the treble boost effect is now getting a raw deal.
To summarize: both the impulse power (also termed peak power) and the power delivered in
the higher-impedance frequency ranges of the speaker is higher for a typical tube amplifier
than for a typical transistor amplifier – with are both rated at the same nominal power for the
same nominal load resistance. A percentage value of the difference that would be generally
valid can, however, not be given, since the individual circuit concepts are too different.
In closing we should quickly also visit the issue of loudness – which in the end is the main
aspect of interest to the musician. It is well known that doubling the amplifier power will not
always double the loudness. On the other hand, the rule taught in psychoacoustics that for
doubling the loudness the 10-fold amplifier power is required, only holds for a 1-kHz-tone.
The guitar generates a broadband sound that does not share much with a pure tone, and this
naturally needs to be considered. However, of even more practical importance is the fact that
the musician judges the loudness of his or her instrument based on how well it can (sonically)
hold its own relative to other instruments. In this scenario, the absolute loudness is not as
important as the so-called partial masked loudness [12]. For example, we may think of a
keyboard player sounding a loud chord, and of a guitar player remaining unheard1 although
his amp generates 10 W into the loudspeaker. The latter is not broken at all, but the sound it
radiates is fully masked by the sound of the keyboard. As the power of the guitar is increased
(e.g. to 20 W), the guitar becomes audible. However, as long as the keyboard is sounded, the
loudness of the guitar remains a partial masked loudness and the guitar will be perceived
softer compared the loudness it would have when played on its own. For the increase of the
partial masked loudness the simple 10-dB-per-doubling-of-loudness law does not hold; a
smaller dB-value is valid, e.g. merely 3-dB per loudness doubling. That way, relatively small
power-differences gain a bit more significance than basic psycho-acoustical know-how would
acknowledge. We shouldn’t overdo it, though. The difference between a 50-W-amp and a 55-
W-amp remains insignificant. The exact location of the perception-threshold can only be
established for each case individually because the masking effects are dependent on the
temporal and spectral structure of the involved sounds.
1
Translator’s comment: the guitar not loud enough - as if that ever happened! Not a realistic example, it seems.
Maybe the other way ‘round ..... the Leslie for the Hammond won’t ever match the Marshall stack, anyway …

10.9.3 Coupling capacitors
Coupling capacitors? For some just a cheap run-of-the-mill product about as portentous as a
tapping screw, for others a true object of desire worth investing the occasional 50 €. The
coupling capacitor separates the AC-part of the plate voltage from the DC-part – it “couples”
the AC-signal to the next stage in the circuit. A plate voltage varying between e.g. 200 V and
300 V may alternatively be seen as 250 V DC voltage with a superimposed AC voltage (of an
amplitude of 50 V) – the coupling capacitor passes the ac component and blocks the dc
component. Let’s ignore, to start with, that the coupling cap in fact doesn’t let anything
through because there’s an insulator inside, and let’s also disregard that this insulator, on
closer inspection, does not perfectly insulate, either. A capacitor lets through AC and
blocks DC – that is a good first working hypothesis. We may – or may need to – modify it
when necessary. If it is that simple, why do “those in the know”, the self-proclaimed amp-
gurus, report with a conspiratorial vibe that the long-desired sound presented itself only after
swapping the coupling caps? And why is it that – depending on which camp one seeks
association with – the original Fender-sound allegedly can only be achieved with the ABC-
Orange-Drops, while the opposing camp warns of using exactly these ABC-Orange-Drops
(mid-rangy sound, totally unsuitable), recommending rather the yellow Mustard-caps? Wait –
not the ABC-Mustards: these are inadequate copies (clones, so to say), you should use the
others, the original copies. Even better: use silver-foil capacitors, or, if that much budget can
be committed, copper-foil-caps. … “committed” …, no let’s not go there, and rather focus,
with the naïve curiosity of the researcher/scientist, on the task at hand: trying to find for a
grain of truth in that pile of rubbish.
In the framework of the present considerations, a capacitor is a component in an electrical

circuit, and as such is subject to the rules and standards of electrical engineering. Whether it
has an aura, whether it holds spiritual energy or ethereal psi – that will not be investigated
here. The very powerful instrument of Maxwell’s equations describes, to general satisfaction,
the processes in electromagnetic fields. It has made wireless transmission, space exploration
and EMP computable; so why not use it as well on the triode-preamp-stage of a guitar
amplifier? These equations are the big guns pressed into action whenever starting-from-
scratch is called for – but fear not, they will serve here as a mere launch pad that is left behind
as quickly as possible. With the limitation to the audible frequency range and to concentrated
components (i.e. components smaller than about 1 km in size), Maxwell’s second equation
may be simplified, resulting in Kirchhoff’s second law (the loop rule):
Maxwell’s and Kirchhoff’s 2nd law, resp.
The line integral of the field strength along the closed curve K is zero, as is the sum of the
n branch-voltages Ui along the closed loop. Correspondingly, the (slightly modified) 1st
equation of Maxwell is:
  n
∫∫ S ⋅ da = 0 ⇒ ∑I i =0 Maxwell’s and Kirchhoff’s 1st law, resp.

A k=1
The enveloping-surface-integral over the current density is zero, as is the sum of all n node
currents Ii (nodal rule, rule of charge conservation).
€

In an electromagnetic field, there are just three quantities characterizing the material: the
specific resistance ρ, the permittivity (dielectricity) ε, and the permeability µ. Limiting
ourselves to the audible frequency range, we may disregard the magnetic properties of typical
materials used for capacitors. Thus, for the formal description merely the loop rule, the nodal
rule, and two equations relating to the material remain. Despite this simple analysis, a number
of misunderstandings exist that have their roots in the inappropriate application of actually
appropriate insights. The area of electrical engineering was in fact not developed to build
guitar amplifier but it is, with all due respect, quite a bit older. Besides other important
technical fields such as power engineering, one main focus was communication technology,
and here in particular the question how to achieve long-distance wireless transmission of
speech or Morse-code. Heinrich Hertz [1886] and Guglielmo Marconi [1901] are to be
mentioned, amongst many other pioneers. The first radio transmission across the Atlantic
succeeded a generation before the start of the Rickenbacker/Gibson/Fender-age, and already
then capacitors were in use. The so-called “Leyden jar” was invented even much earlier in
1745. Radio transmission, however, does expressly not work in the audio frequency range.
High frequencies (or radio frequencies – sic) are required: 1 MHz for the AM-range and
about 10 MHz for the short-wave range. Considering this, and also the fact that the “bibles”
from back in the day were titled "The Radio Engineers' Handbook", and not "The Guitar Amp
Designers' Handbook", it is easy to imagine what can happen: someone (from the guitar
community) reads that caps may have their problems at higher frequencies, and immediately
fears for the silvery highs of the famous Fender-sound. As if they had been built for just that
situation: there are indeed Silver-Mica-Caps! They will make that Fender-treble bounce right
back, won’t they! The echoism “silvery highs” should not be criticized – that can very well
remain here as a term of art. It is also correct that capacitors become inductive, but at which
frequency does that occur? Even for a wound capacitor (very remotely related to a coil),
effects of this inductance happen only above 1 MHz, i.e. at “higher frequencies”. However,
that does not mean the “higher audio frequencies” are affected – there is a factor of about 50
separating these ranges! In just the same way, the loss-factors known to “rise towards higher
frequencies” will worry only the RF-technician or HF-filter-designer. Typical guitar
amplifiers, however, include neither a short-wave-input nor a 12th-order elliptic filter, and the
low-loss styroflex capacitor is rather misplaced here (it’s too hot an environment for it!).
On top of the difficulties to interpret books on RF-technology in the correct way, we also
encounter the problems found in psychometrics: how do we measure audio-perceptions? In
Chapter 10.9.4 (sound event, auditory event), some suggestions were made, and in particular
the need for blind-tests was emphasized. Daily practice, however, looks quite different. For
example, a manufacturer states that his capacitors require a run-in time of 100 hours until
they sound good. The guitarist having fitted his (sound-wise not convincing) Marshall with
these caps hears exactly that, and reports to all colleagues: “it was only after 100 h that the
treble was how it should be.” Indeed, it is generally known that technical devices require a
run-in time: that relates to the car-engine (1000 km run-in) as it does to the charcoal on your
grill (no smoke should remain after 10 minutes. And if a newly elected politician gets a 100-
day grace period: why not the capacitor, as well? Let’s consider this: an amateur musician
playing for 10 h a week will after 2,5 months arrive at the point where he remembers how the
caps sounded when they were just freshly soldered-in: it was atrocious, the treble just wasn’t
there. Now, however, after almost a quarter of a year, it suddenly has appeared. And the
reason for this improvement must not be attributed to the tube-aging♣, not to the (possibly
new) loudspeaker and certainly not to the strings changed multiple times. No, it must be the
caps – in fact the guitar mag has found the same, and even recommends 200 h of run-in time
♣
the manufacturer says (1952 A.D.) that a run-in time is required. Why are we not surprised: it’s 100 h.

(i.e. … wait another quarter of a year). And why wouldn’t it be the caps – for tubes, nobody
doubts time-variant sound changes, either, do they? Whether copper-foil capacitors sound
better than aluminum-foil capacitors (‘cause copper has lower resistance) – somebody must
be able to verify that beyond reasonably doubt! Skeptics pointing to missing data regarding
the thickness of the foil could possibly be convinced, if indeed the listening experiments
would not show such dramatic deficits. If such an experiment includes the announcement: “ I
will now solder-in the copper-foil-caps; that will give a much more forceful sound”, it
immediately becomes futile since all participants will be biased. A much better approach can
be found with Tone-Lizard.com: here it may happen that the same capacitor is connected to
5 taps of a 6-position switch – not 5 equal caps but in fact one and the same capacitor
connected to the 5 positions, and without the judging guitarist knowing this. The latter
declares: ‘the orange-drop in position 1 sounds much better than the one in position 4’. No
further comment necessary. Obviously, this is just a single case, but just as obviously, the
judgment ‘ I cannot for the love of it hear any difference’ is not likely to occur a lot because
no sound-expert will want to damage his/her reputation. And so capacitors all have their own
sound, as does every bolt and every rivet.
It is not possible to find out who first introduced the myth of the 100-h-run-in-time for
capacitors. The following might be an explanation, though. In data sheets for capacitors we
find a time-constant of 100 000 s for high-grade builds – this equals just under 28 h. If now
someone has some lingering remains of memory that for a full charging process, 5 time-
constants waiting time is a good measure, then we arrive at 140 h … there it is – the run-in
time? No, that’s not it, at all, of course – these are all data relating to the insulation (!!)
behavior (which is not unimportant but an entirely different issue, so let’s postpone the
discussion of it a bit). The above idle-state time-constant would only play any role for the
capacitor by itself i.e. completely disconnected (which you wouldn’t want in a guitar amp,
would you!?). Connected in its regular habitat, the capacitor is loaded on one end with the
grid resistor (e.g. 1 MΩ), and at its other end with the tube and the plate-resistor (e.g. 50 kΩ ).
Multiplying these resistances with the capacity yields the actual time constant in operation,
and that is, at e.g. 22nF ⋅ 1,05 MΩ = 0,023 s, significantly smaller than the idle-state time-
constant. If you really could talk about a “run-in time” here, and multiplied (according to the
above approach) that time by a factor of 5, you would arrive at a “run-in time” of a full 0,12 s.
Alternatively, there is reasoning based on material changes that start only after the onset of a
voltage and need to stabilize first. It appears that someone has looked too long over the
shoulder of the colleagues dealing with electrolytic capacitors. This type of cap uses,
similar to the foil-capacitor, a rolled-up aluminum foil – however the dielectric is not formed
by plastic foil between the aluminum layers but by a thin Al2O3-layer (aluminum-oxide) that
grows onto the anode of the capacitor during a formation process. It is only this oxide-layer
that has the insulating effect, and therefore an unformed electrolytic cap must never be
subjected to a voltage. However: what does this have to do with the coupling capacitors that
are never, ever, of the electrolytic sort in a tube guitar amp? In the latter, polyester or
polypropylene capacitors are used that do not have to – and cannot – be ‘formed’ in the first
place. Still, couldn’t there be any other slow-moving and possibly unknown processes at work
on the metal surface or in the dielectric? Yes, actually, that is highly likely – the field-strength
is high and the temperature, too. Let’s assume that a capacitor somehow changes during the
first 100 d. What parameters could in fact change? In any case the impedance, i.e. the
complex ac-resistance, would have to change. Only variations that would have any effect at
all on the impedance could change the grid-bias-voltage – hopefully nobody will claim that
the sound might change while all tube voltages remain the same. Now, here’s the thing about
impedance: it is only defined in the linear model; there are no impedances in the non-linear

model. Could a coupling capacitor be non-linear? Sure, every capacitor is non-linear – it

increases its capacity when a voltage is applied (charged electrodes attract each other,
decreasing the distance between them and increasing the capacitance). However, already
simple math shows that the resulting pressure (some kPa) remains (together with the elasticity
module of around 1.000.000 kPa) below the critical deformation limits by several orders of
magnitude. Measuring the THD confirms: as long as a HDK-capacitor (highly uncommon in
guitar amps) is avoided, the THD remains below 0,01 %. Thus: we do have linearity in caps –
with good approximation.
So: which parameters could change? To start with the insulation resistance: it could e.g.
increase from 50 GΩ to 100 GΩ (), or it could decrease to e.g. 25 GΩ (?). Taking 1 MΩ
as the input impedance of the following stage, a leakage current of 4 nA (or 2 nA, or 8 nA)
would flow at an operating voltage of 200 V, yielding a shift in the operating point of 4 mV
(or 2 mV or 8 mV). None of these values represents any reason for concern, plus (most
importantly): at an insulation resistance of 25 GΩ, the manufacturer would have pulled the
plug because this would be outside of the specification at least for branded products.
Incidentally: will you wait for 100 h for the insulation resistance to deteriorate? That could be
achieved in a much simpler way. In any case, the insulation resistance may be a shambles (as
it actually is for the very long serving caps in vintage amps – sic!), but for new name products
it will be good enough, and will not have any audible effects on the sound.
And on to the capacitance. Does that change during the first 100 h, right? Of course it does -
panta rhei – everything changes all the time. Data books for high-quality capacitors indeed
specify a change: e.g. <± 1%, at 70°C , for the first two years. No cigar, then, either. And
again we need to note: if the sound were indeed better only after the capacitance has in- or
decreased by x% - the improvement could much simpler be achieved than by waiting for all
that time.
The loss factor remains. It could drop – that sounds desirable even without any exact
knowledge (lower loss of whatever) – or it could rise. Hold on … the amp sounds better after
100 h, so presumably it will be lower losses? From "smeared sound to clear sound", as the
guru elucidates. Has anyone actually posed the question why we would solder a capacitor –
no: several capacitors for serious money into our amp, if it sounds “smeared” for 100 or 200
or even more hours? Someone probably did ask, and the stunning answer would be: “because
afterwards a undreamt-of sound experience will set in that off-the-shelf products can never
impart.” Per Aspera ad Astra – we know this, as well. If the loss-factor were crucial to the
sound (and indeed who wants any “loss” in their sound), then the lowest-loss dielectric would
have to be used in coupling capacitors, correct? Consequently, and after excluding the les
temperature-resistant polystyrene, polypropylene would be the right choice. And indeed, it is
exactly this material that is found in the Orange-Drop-caps already mentioned. However: the
guru from the other side of the fence strongly advises not to include those, but only the ones
with polyester as dielectric, just as in the original in the 1960’s. The manufacturer adds:
“because of its deeper tonal quality" – whatever that may be. Polyester, however, has a loss-
factor about 100 times higher than of that of polypropylene – so now the going gets tough.
Could the solution be: more losses for a better sound? A short calculation comes to the rescue:
polyester will give a loss-factor of a about 1% at 10 kHz which – in the high-frequency
equivalent circuit (to be discussed later) corresponds to a 7-Ω-series-resistor (for C = 22 nF ).
Imagine this connected in series with the source impedance of the preceding tube stage (e.g.
50000 Ω): it will be 50007 Ω instead of 50000 Ω – so much then for that approach. Clear? Or
smeared?

No, the losses do not help us here. In short-wave resonant circuits, there would possibly have
been trouble, but nobody will include a 22-nF-capacitor is such a design. Nope, losses do not
explain any sound difference in coupling-C’s.
Don’t stop now – here comes the slew-rate: the cap-manufacturer still has some aces up his
sleeves. The slew-rate is the differential quotient over time of the voltage across the capacitor;
it is also called the speed of voltage-change, given in V/µs. Again, a glance into the data
books is helpful and indeed they show limit-values that must not be exceeded. Examples are
500 V/µs for a polyester capacitor, or a mere 30 V/µs; or 750 V/µs for the polypropylene cap,
or even more than 1000 V/µs. Not to forget: the mica capacitors that considerably improve
the sound excelling at 100.000 V/µs. These are considerable differences, so what is the secret?
Let’s do the math: a tube that is required to generate an AC-voltage of 30Veff at the plate
reaches 0,8 V/µs. (N.B.: we work with 3 kHz here since the pickup will never manage full
drive-levels at 20 kHz). Or maybe a bit more which brings the slew-rate to just about 1 V/µs.
Given the above data, this would not cause any problems. However, the tube may be
dramatically overdriven, e.g. by a factor of 30 – and now we would arrive in the range of the
above limit-values. Strike!!?? No, still not: the slew-rate happens at the plate but not across
the capacitor! The voltage across the capacitor is in fact much smaller, and it even decreases
with increasing frequency. That’s why the slew-rate is not a meaningful parameter – the
maximum current load that can equivalently defined for the cap would be more suitable.
The data books specify values of 0,1 – 1 A, and even more than 1000 A for the mica capacitor
– is that good enough? Well, yes – your typical noval-triode is able to supply only a few mA.
So again: no issue. But here’s a hunch based on this line of thought: the originator of the buzz
around the cap might have built a loudspeaker crossover at some point. That scenario provides
an entirely different picture – we are confronted with the big-boy-currents: 100 W = (5A)2 ⋅
4Ω, i.e. 7 A peak current. Could it be that someone has put to (mis-) use the frequency-
crossover design-rules in the context of coupling capacitors? Sure: the manufacturer of the
expensive copper-foil-caps is also known for his crossovers, isn’t he? Directly quoting the
guru: “it’s an unbelievably fat sound …” Unbelievable?
An on we go: the inductance of a coupling capacitor now surfaces. It may today be reduced
to the point where 0,8 nH/mm CS can be reached. CS stands for “contact spacing” and not for
“compact size” … but the latter would anyway not be appropriate given the sheer size of these
super-coupling-caps (45 mm x 26 mm). In the Roederstein (ERO) data sheets we find much
larger Nanohenry-values (12 nH) – but STOP: that brochure dates back to 1980. What was the
size of that super-cap again? 45 mm in length? So it’s 45x0,8 nH = 36 nH? And ERO was
also per CS? No, that was the overall inductance. My, so much has changed since the 80’s!
Just to mention it: even 36 nH would be o.k. – at 10 kHz a reactance of a full 0,0023 Ω would
result. That would have to be added to the 50000 Ω mentioned before. Pythagorean addition
to be used, of course.
So, what remains? Not much … maybe the skin effect: that is also an object of adverts. “At
high frequencies (ah, the old story …) the current flows merely along the outside of the wire,
such that the conducting cross-section is reduced and the resistance grows.” That is entirely
correct: at high frequencies. H. H. Meinke always started his famous lectures with just this
issue, and noted: “… and that can start already at 10 kHz”. It is from around that frequency
that the resistance of connecting wires increases noticeably – maybe from 0.002 Ω to 0,004 Ω.
Nothing could be added here that was not already covered above when we looked at the loss
factor. No, the skin effect doesn’t contribute anything either, at audio frequencies. And
another thing: someone advertises that the connecting wires of his caps are made of tin-coated
copper, and not of copper-coated steel reaching merely 30% of the conductivity of copper.

Here’s a suggestion for improvement: “our replica caps come including the connecting wires
and not without – the corresponding enhancement in terms of conductivity is beyond anything
that could be expressed as a percentage.”
Lest our commentary turns into mere satire, let’s try to arrive at a summary. First, however, a
look into the internet: word is that there are rationales that are so misguided that it is not
possible to even get close to them on the basis of normal training in electrical engineering.
Our capacitor-guru provides some inspiration: every capacitor brings unwanted effects in the
form of additional inductances and resistances (ESR = equivalent series resistance). While
the inductance practically has next to no influence, the ESR determines – in conjunction with
frequency and capacitance - the loss-angle tan(∂). O.k. – it is possible to express the issue
that way – all text-books on electrical engineering do it in a similarly. Let’s move on: tan(δ)
does not remain the same for all frequencies but is frequency-dependent (that is correct). If
tan(δ) rises strongly with frequency, the frequencies are not all treated in the same way; if
there is a smaller change between 1 kHz and 100 kHz at least the frequency spectrum of an
impulse mixture is transmitted more time-correctly. The highest frequencies are shifted in
terms of amplitude and phase = differential phase-error. For the ideal capacitor, tan(δ) needs
to be as frequency-independent as possible implying an ESR that drops towards high
frequency. That way, time-distorted frequencies are shifted further towards the MHz-Range
(=smallest possible error within the audio-range). Phew – now we have arrived at a hodge-
podge of desire & reality, of science & sales – this is now really incorrect.
The above “sales-supporting comment” could be shortened to: every signal-carrying

capacitor deteriorates the reproduction of impulses, and it does this the stronger, the more
the loss-factor depends on the frequency. Whether there are actually any time-distorted
frequencies does not need to be investigated yet, but the reproduction of impulses needs to be
looked into. Let’s start with an equivalent circuit for a simple high-pass, as it is also presented
by our capacitor-guru for a coupling capacitor (Fig. 10.9.7):
Fig. 10.9.7: The coupling capacitor in the high-pass circuit.
The source (this could be the plate-circuit of the preceding tube) is drawn as a voltage source
with a series resistance RG, RV = ESR, the load (this could be the grid circuit of the subsequent
tube) is shown as a simple real resistor RL. If needed, this simple schematic could be
extended, but for basic considerations, it is adequate. RV is frequency-dependent – our guru
states this, as well. He does not state that in this equivalent circuit diagram, C needs to be
frequency-dependent, too – however, since this dependency is small, it may be ignored for
now (just as L may be ignored for now). We assume that RV is a simple, real, frequency-
independent resistor, for example a short length of copper wire. Copper? No, why don’t we
better turn to silver wire as it is offered likewise by the capacitor manufacturer: with a choice
of Sterling silver (92,5%), or pure silver (99,97%). For our thought experiment, we naturally
take only the purest silver because of the higher conductivity (silver conducts about 5 – 10%
better than ordinary copper). Such a short piece of wire will measure maybe 2 mΩ, and this
we have to think as connected in series with our source impedance. The two resistors maybe
added up yielding a full 50.000,002 Ω. This is one of the undesired side-effects: the source
resistance is increased.

What does the loss-factor of this model-capacitor look like? Since the ESR was assumed to be
frequency-independent, tan(δ) rises proportionally with frequency. Big mistake, because: for
the ideal capacitor tan(δ) needs to be as frequency-independent as possible, i.e. the ESR
should decrease with 1/f: RV ∼ 1/f. The guru states: the smaller the increase of tan(δ), the
higher the voltage steepness dV/dt and the smaller the differential phase-errors in the audio
range – especially for higher voltages. Voltage-dependent errors? The model of our guru is a
linear one and thus cannot include any voltage dependent components. It isn’t that capacitors
would show no non-linear effects at all, but if you want to describe non-linearity, you have to
have a non-linear model. Let’s stick to the linear model, though, in order not to further
increase the already considerable confusion; for small voltages things are linear, anyway.
Let’s go back: we have the allegation that highest frequencies are amplitude- and time-
shifted. Just to be precise: a frequency cannot be shifted – neither in amplitude nor in time.
Frequency is the inverse of the cycle duration: a cycle duration of 10 ms results in a frequency
of 100 Hz. What is presumably meant: highest-frequency signals are time-shifted and changed
in their amplitude. We could also say: in the highest frequency range, amplitude- and phase-
changes occur. That, indeed, is how this would be expressed in systems theory. Differential
phase-errors – no, you wouldn’t really say that. What could be meant here? Maybe it is the
spectral derivative, dϕ / df, that usually is supplemented with -2π and designated the group
delay τg = -dϕ / d(2πf). This is in systems theory, usually. It is not problematic to use
uncommon terms, but for misinterpretations, the liability needs to go to the party responsible.
Again, in a nutshell: the ideal capacitor does not generate any amplitude- or delay-
distortion♣. Is that what was meant? O.k. then – let’s move on.
Not any amplitude-distortion. At another passage it is even more drastic: the ideal capacitor
should not influence the audio signal at all. So obviously, there must be signals that are not
audio signals, and these are clobbered by the capacitor? Correct? No – wrong! In a
loudspeaker-crossover, the capacitor connected ahead of the tweeter is supposed to attenuate
low- and mid-frequency signals i.e. it is there to reduce their amplitude. Are these signal also
audio signals? Of course they are! But lets set aside the crossover – it is indeed possible to
talk, in the context of a coupling capacitor, of small influences on the audio signal, or none at
all. The capacitor operates (together with the resistors) as a high-pass filter and therefore
attenuates the (very-) low-frequency components. Our guru, however, seems not interested at
all in the low-frequency range, since he localizes – even several times – the undesired effects
in the range above 1 kHz: a capacitor would let high frequencies pass without limitation if it
weren’t for the losses. Herein lies a grain of truth: the impedance of a capacitor would
continuously drop with increasing frequency if there would be no losses. But wait a moment –
is this really about the impedance? At 1 kHz, the losses (10-4) increase the impedance of a
22-nF-cap (polypropylene) by 0,0000005%. That’s no joke, it’s covered in the basic course in
material science and components => ! Even at 10 kHz (and further up in the
audio range), the impedance is not an issue. So, then, on to the phase, or rather its spectral
derivative. We read that it would be ideal if voltage and current would occur at the capacitor
with an exact 90°-phase-shift between them. Indeed, that is correct: only the ideal capacitor
can achieve that.
♣
In systems theory, the term distortion is employed in two ways: non-linear distortion (harmonic distortion) and
linear distortion (amplitude-, phase- and delay-distortion).

The real, lossy capacitor shows a phase-shift that is different from 90°. Is that bad?
Apparently so, because differential phase errors in the audio range result. If the phase shift is
frequency-independent, then “all frequencies arrive at the same time” – to stick with the
terminology of our capacitor-guru. In other words: signals of different frequencies all are
subject to the same delay. However, if the phase-shift is frequency-dependent, the “high
frequencies are time-shifted”. Or better: higher-frequency signals are time-shifted (or delay-
shifted, or, alas, phase-shifted) relative to lower-frequency signals. Possibly, an impulse will
be stretched out, or “smeared”, due to the phase-shift. We do recall that term: “smeared
sound”. The absolute phase difference between current and voltage (which for the ideal
capacitor will be 90°) can be accepted; the differential phase-error (that, in the absence of any
explanations, needs to be interpreted as the group-delay) causes impulse distortion. So, our
guru had a convoluted and sibylline way of saying it – but what he actually wanted to express
is this: "The loss-factor of a capacitor causes group-delay distortions that smear the
sound (at high frequencies)”. As a remedy – direct quote guru – the loss factor needs to be
as independent of frequency as possible, or, in other words, the group-delay needs to be as
frequency-independent as possible. At first glance, this sounds familiar: systems theory says
such a system is a linear-phase system, and certifies a distortion-free behavior.
And with this, we have arrived at the core of this grandiose misunderstanding: the group delay
is a transmission quantity (a quadripole-quantity), while the phase-shift between current and
voltage is a two-pole quantity. Differentiating the wrong phase will yield a wrong result.
More specific: a quadripole is a system with four terminals (also called two-port network),
with a two-terminal input and a two-terminal output – the high-pass shown in Fig. 10.9.7
would be an example. That input and output in this example have the same ground connection
does not make the system a tri-pole – it still is designated a quadripole. Between the input
signal (input voltage) and the output signal (output voltage), a complex transmission function
is defined from which the frequency response of the phase and of the group delay can be
derived. A capacitor, on the other hand, is a two-pole because it has merely two terminals. A
complex impedance is defined between the voltage and the current, and a phase frequency-
response can be derived from this impedance. But try and deduce a group-delay frequency-
response from this – that is nonsense. There are merely two special scenarios in which it is
purposeful to see a two-pole as a quadripole: if voltage is the input quantity and current is the
output quantity, or vice versa. Now, one could argue that every quadripole is on fact
constructed from two-poles and that therefore any deficiencies of these two-poles must also
be a deficiency of the quadripole. This, however, is not the case. In the present framework, we
cannot present the systems theory in its full scope but have to refer the reader to special
literature [5, 6, 7]. Very briefly: in the high-pass mentioned above (and equally for an RC-
low-pass), the input voltage is divided between R and C – it does not fully span across the
capacitor.
The simple formula yields the loss-factor d as tangent of the loss-

angle ∂. In the ideal capacitor, a sine-shaped voltage precedes the current by exactly 90°; for
the real capacitor this angle is smaller than 90°. For a loss-angle of δ = 0,01° (polypropylene
at 1 kHz) the phase-shift between current and voltage therefore does not amount to 90° but to
89,99°. If the loss-angle were frequency-independent, as curiously demanded by the
statements of our guru, then the phase-shift differentiated with respect to the frequency would
result in a constant value of zero (the derivative of a constant is indeed zero). That would
seem to be the ideal case: no smearing of impulses.

Now, the phase-shift between the current through and the voltage across a capacitor is one
thing, and the phase-shift between the input and output of the coupling circuit is something
different – in fact something entirely different.
The phase-shift ϕ appearing in Fig. 10.9.7 between the generator voltage and the output
voltage (across RL) can easily be calculated:
[6, 7, 17, 18, 20].
Fig. 10.9.8 shows the frequency response of the phase in a presentation with both axes in
logarithmic scaling – for the ideal, loss-less capacitor! In reality, the group delay♣ depends on
the frequency, and dispersive impulse-distortion results. Again: the case calculated here is the
best-possible one with the phase-shift between voltage across and current through the
capacitor being exactly 90° at all frequencies. However, for a real polyester-capacitor,
practically the same figures would emerge – the difference would be entirely insignificant:
e.g. for the group delay it would be as little as 0,0004% (1 kHz). Using a capacitor with a
constant loss-angle across the frequency range would deliver differences of a similar
magnitude. Relative to a load resistor of 1000000 Ω it does indeed not make any difference
whether the ESR is 7 Ω or 0 Ω. What does make a difference is a change in the capacitance –
shown in the figure for a 30%-increase. It will be discussed later whether such huge
tolerances can occur at all in a high-grade capacitor, and, if yes, whether they are significant.
Fig. 10.9.8: Phase frequency-response and group-delay frequency-response of the circuit acc. to Fig. 10.9.7
RG = 50kΩ, RV = 0, RL = 1MΩ, C = 22 nF. The dashed line is valid C = 28,6 nF (i.e. +30% tolerance).
In summary: the conjecture that the loss-angle would have to be as frequency-independent as

possible leads to incorrect conclusions, since it is derived from an entirely unsuitable two-pole
phase-angle. For a typical coupling circuit with tubes, all capacitors (including theoretical,
ideal capacitors) generate practically the same group-delay distortion (“impulse smearing”).
This delay distortion is, however, so small, that it remains far below the threshold of
audibility.
♣
In case the group-delay is to be derived from the phase frequency response via graphical differentiation
(gradient), a representation with linear scale on both axes needs to be used, not with double logarithmic scale.

The last statement needs to be commented on. First, we give the word to our guru: the value
of tangent ∂ does not always tell the full story. If it rises drastically, not all frequencies are
treated in the same way; if it changes less between 1 kHz and 100 kHz (as it is usually the
case for metal-foil capacitors), the frequency spectrum of an impulse-mixture should be
reproduced more accurately with respect to time. In this context, a shallow rise of tangent-∂
between 1 kHz and 100 kHz is desirable, combined with a simultaneously stronger dropping
ESR. For the same reasons, developers aim for rise-times of below 1 µs (about 1 MHz
bandwidth). That is plain wrong. The term “rise-time” is used in circuit technology to define
the time period during which an impulse response rises from 10% to 90%. The rise-time is 2,2
times the length of the time-constant that in turn is the inverse of the angular cutoff-
frequency. From this, the cutoff-frequency calculates as 2,2 / (2π⋅1µs) = 350 kHz, i.e. not 1
MHz. Even if not rise-time but settling-time were meant, the bandwidth-specification is still
incorrect: for 1 µs settling-time, a bandwidth of 500 kHz results for steep-slope bandwidth
limiting (since the term is not really used much for a first-order low-pass). Anyway, rise-times
of below 1 µs, since: the superposition of room-reverb onto the original signal needs to be
correct to the microsecond, so the ear can pick out the exact location in the space.
This we need to think about – that is not entirely wrong. The threshold for detecting an inter-
aural delay (localization blurring) can be determined to be as low as 10 µs under laboratory
conditions, and specialist literature reports even lower values than that. And if you really want
to achieve a “safety zone” of a factor of 10, the result is indeed 1 µs. However, to conclude
from this the requirement of a bandwidth of 350, or 500, or even 1000 kHz – that would be
nonsense. A pure 1-kHz-Tone can perfectly be shifted by 1 µs without tapping into the RF-
range. The hearing system can perform the delay-resolution of 10 µs (as mentioned above and
if indeed it does that well at all) in the mid-frequency-range, i.e. at around 1 or 2 kHz. At 10
kHz, this just noticeable difference has grown quite a bit (100 µs as a rough guideline – the
data depend highly on the experimental conditions), and beyond 20 kHz there is no hearing.
Or is there?
Now, every audiophile has gathered (from wherever) that the stimulation with pure sine-
signals is something quite different than real sounds because the latter contain tons of
impulses. And so one of our gurus manages to demand, on his webpage, on the one hand a
bandwidth of 1 MHz, and to refer, on the other hand, to a thesis that very accurately takes the
upper limit of hearing to be at around 19 kHz. How does that fit together? We are not talking
about 19 or 20 or, even better, 22 kHz – here very casually a factor of 50 is built in, as a
reserve. To voice, in one and the same sentence, an opinion and simultaneously the counter-
opinion – that is normally only achieved by certain politicians (or showbiz-people).
This mixing-of-what-must-not-be-mixed-up is done – for audio signals – in the following

way: every signal – and that means indeed EVERY signal – is in fact the sum of an infinite
number of sine signals. Yessss!! You can’t maneuver around good ol’ Baron Fourier. In
principle, this statement is correct but we must not take the “in fact” too literally. Mind you:
the Fourier analysis is a model consideration, and every signal could just as well be
segmented into many other (not even necessarily orthogonal) functions rather than sine
functions. What is valid for signals is also valid for systems (as long as they are liner and
time-invariant): the consideration of processes in the spectral domain is equivalent to the
consideration of the processes in the time-domain [6, 7, 17, 18, 20]. If the hearing system
cannot hear continuous tones with a frequency of above about 20 kHz (and moderate SPL), it
cannot hear, for impulses, any of their spectral components that lie above about 20 kHz,
either.

Now, this finding must intricately be reformulated in such a way that the capacitor
manufacturer will get a sales boost. That is done in the following way (under the same web-
address): music is not just composed from pure sine-tones but from a very broad spectrum of
different impulses that in part lie very far outside the hearing range but which strongly
influence the hearing perception⊗. For this reason we use – for audio amplifiers – a capacitor
that falters only above 100 MHz. In the coupling branch of a pre-amplifier, this capacitor
gives us an impression of how a piece of wire would sound. No, this is not a printing error – it
indeed is supposed to read not 100 kHz but 100 MHz. And a few lines further we find: mica is
the ideal dielectric for capacitors, yielding the following properties: … applicable up to very
high frequencies into the GHz-range. Mica capacitors are highly favored for filter circuits
where, due to their properties, they can bring a considerable increase in sound-quality.
Gigahertz! Is there no end to this?! Is the Terahertz range next? And yet Schöne et al. have
already proven in 1979, that a reproduction of the ultrasound range adds nothing whatsoever
to the perception♣. That was an investigation carried out by the Institut für Rundfunktechnik
(the internationally renowned German broadcast technology institute), though, and in some
audiophile circles the preference is not to take note of research done there. Any self-appointed
guru who pushes the requirement a further few MHz into the RF range is seen as the new
messiah. Skeptics, however, are branded as “infidel physicists whom one should give a wide
berth”. "C'est la gare" is the only congenial answer to that.
Let us revisit the example used in Chapter 8: a bed of a length of 1,5 m will be judged as too
short for most grown-ups, while a length of 2 m is quite comfortable. Now, there are a few
people who are taller than 2 m, and to accommodate these cubo-philes, a bed should, for good
measure, be a bit longer. Taking the above approach used by our capacitor manufacturer for
the 100-MHz-capacitor, the bed should be about 10 km long, just to stay on the safe side. Has
anybody thought of “indulging” our other senses that way? Our visual sense would lend itself
as a candidate: the limitation to the frequency range generally as “visible” (380 – 770 THz)
seems overly restrictive, and why not give the TV a correspondingly enlarged bandwidth (i.e.
X-ray radiation)? And, of course, that should extend into the lower range, as well: the
microwave oven would stand ready to be a splendid “optical subwoofer”.
But back to the audio amplifier: the frequency range up to 20 kHz needs to be reproduced
precisely, and since no amplifier will shut down abruptly above this limit, a few more tens of
kHz are purposeful to let the amp taper off. In case listening experiments result in other
numbers, any conclusions may be put under scrutiny on the test bench. Ill-considered phase
responses and listening experiments with biased subjects are, however, not conducive. And
one more thing about the 10-µs-delay-distortion mentioned above: even smaller values may
be audible inter-aurally. Given that, and the fact that most people have evolved beyond mono
into stereo-territory, wouldn’t it be desirable that the capacitance-tolerance of the wonder caps
would be of matching dwarfishness? From this point of view, it is peculiar that one of the
manufacturers specifies tolerances of -20/+30%. Sure, hand-made, every capacitor is one of a
kind. Or maybe the manufacturer is aware how strongly the group delays of any two
headphone systems or of two loudspeakers (of the same type, respectively) can differ? Maybe
he knows all this and just doesn’t tell? And continues to jumble and confuse things while
feverishly searching for the ideal capacitor that blocks and at the same time passes DC.
⊗
That is why they are called “lying outside of the hearing range“ (sic).
♣
P. Schöne et al.: Genügt eine Bandbreite von 15kHz... (Rundfunktechnische Mitteilungen, 1/1979).

From the world of the space-worthy 100-MHz-capacitors that can deal with 1000 A now back
to the guitar amp. Don’t panic: here we don’t find either. Even the 10 µs that may be crucial
of the inter-aural delay do not give us a headache. For all guitar players that do not carry a
stereo-system to the stage: the threshold for diotically♣ perceivable group delays is about 2 ms
[3, 12]; one may get to somewhat smaller values via special training, but this is not an issue in
practice. So, for sure, there is no "smeared sound" due to the coupling cap with its group delay
of τg = 0,0001 ms. Things may be entirely different in loudspeaker crossovers where we have
large currents on the way – maybe not 1000 A and 100 MHz but still: this is the power-
engineering-league. The coupling capacitor plays in Little League: it’s communication
engineering and a few microamperes on this playing field.
Let’s acknowledge the difference (whatever it may be) between power- and communication-
engineering, and between research and marketing. After we have (full-monty?-) scientifically
shown that coupling capacitors cannot contribute anything – really not anything – at all to the
sound, we could conclude with a real bombshell and note that these caps do in practice
influence the treble, after all. In fact, that is easily explained and we will get to its in a bit.
First, the relation to the equivalent circuit needs to be covered, in more detail. Gotta do it.
In the daily routine in the lab, a coupling capacitor is described via two quantities: capacitance
(e.g. 22 nF) and dielectric strength (i.e. proof voltage, e.g. 400 V). The third parameter (the
loss factor, is of significance only if the capacitor is connected to inductances. This would be
the official position, and according to it all capacitors of equal capacitance would have to
sound the same. The teachings of electrical engineering do however also state that the
function of a capacitor is of such infinite complexity that only rigorous simplification makes
the above analytical description possible. The series connection of an ideal capacitor and an
ideal resistor is just about the simplest approximation: more sophisticated models consider
special polarization effects, as well, and they arrive at more complex equivalent circuits
(Chapter 9.4). However, at middle and high frequencies coupling capacitors are of such low
impedance that only a very small AC-voltage is created across them. Of the 30 Veff plate-
voltage, a mere 0,2 Veff are found across a 22nF-high-pass capacitor (22nF/1MΩ) at 1 kHz,
and therefore the divergence of that cap from the ideal cap is not that significant. At low
frequencies, however, the AC-voltage across the capacitor rises: half the frequency – double
the cap-ac-voltage, until the cutoff frequency is reached at 7 Hz. Somehow, though, the lows
never turn up in reports about the sound of coupling caps; it’s always the highs that are
smeared, that sound “mushy” or “hollow”, and only “open up” after 100 h. Still, we could –
for once – consider the lows as well … the real deep lows:
Let us consider once more the high DC-voltage across coupling capacitor: depending on the
circuit this will be 150 – 300 V, in special cases even more (beware: mortal danger!). If the
insulation resistance of a coupling capacitor is e.g. 1 GΩ, about 200 mV are measured across
the following 1-MΩ-resistor (at 200 V plate voltage) – for an ECC83, this is already quite a
lot (Fig. 10.1.14) and may cause audible effects. However, whether the sound is improved or
damaged by this offset-shift cannot be generally predicted. There is always the same
rationale: we encounter too many sound-determining parameters. This lack of a general
prediction may not be really necessary, any way: for new high-grade capacitors, the insulation
resistance is far higher than the one used in the above example, but for decade-old capacitors
it many be much lower.
♣
diotic presentation: both ears receive the same signal (mono, both ears listening)

In data sheets we find, for polypropylene-caps, insulation resistances specified to 20000 GΩ,
and of 1000 GΩ for MKT-capacitors (each for 22 nF). These values are given for room
temperature – which is not the normal internal condition in a tube amplifier where we easily
find 70°C. This reduces the insulation resistance by a factor of 5. Still, even then, for MKT-
caps the insulation resistance will remain as high as 200 GΩ; for our above example that
would lead to an offset shift of 1 mV. That is a value that will not by any stretch of the
imagination have an impact on the sound. Measuring various mostly old (but unused i.e.
N.O.S. – new old stock) capacitors yielded values between 5 and 100 GΩ – that is clearly
worse and could possibly classified as borderline regarding audible effects. However, really
bad were the 0,15-µF-caps taken from an old VOX AC30 (of 1965 vintage): one still
measured 2 GΩ but the other had dropped off to merely a 100 MΩ insulation resistance. The
capacitance-values were still within the 20%-tolerance, but the leakage current shifted the
operating point to an extent that in fact should have been designated a catastrophic failure.
The amp, however, still worked, and whether the sound generated with this capacitor is
judged as good or bad, as broken or as vintage, and the caps therefore are judged as junk or
holy grails – that must be dealt with in the subjective domain.
Looking at things in a very fundamental way, it is possible that besides the purely electrical
parameters, electro-mechanical parameters may also play a role. Indeed the coupling capacitor
is charged via a high-impedance resistor, and if the capacitance changes over time, the
capacitor acts as an AC-voltage-source – even without a guitar connected to the amp. The
same principle as the one for a condenser microphone holds [3]: the high-impedance resistor
(of e.g. 1 MΩ) prevents a quick charge transfer, and for an approximately constant charge,
any small relative change in charge superimposed on top of this approximately constant
charge corresponds to the change in voltage. Specifically: as the capacitance changes by 1‰,
an AC-voltage of UDC/1000 results. With a capacitor charged to 200 V, this would be 200
mV. Whether the capacitance can really vary by 1‰ is a different question. In a combo amp
with speaker and amplifier in one and the same enclosure, we do find high sound pressure
levels reaching 100 Pa and more. The resulting forces acting onto the capacitor housing will
change the capacitance – but not normally by as much as 1‰. A simple consideration will
help to estimate the order of magnitude: as a solid object is submerged in water, it is subject to
a water pressure mounting with the submerge-depth. This pressure will crush even submarines
made of steel if they dare to dive too deep – a capacitor however is much more fragile than a
submarine. The higher the pressure, the more the capacitor electrodes will be pressed
together. So, which dive depth might be equivalent to the above mentioned 100 Pa? Which
special laboratory could be entrusted with finding this out? 100 Pa makes for 100 N per
square meter … that corresponds to merely 1 cm dive depth! So: no special lab – the bathtub
is good enough. Although: 200 V in the bathtub … no, better not. Dear music magazine
journalists (if you at all accept advice from a scientist): do not try to do this at home! Danger
to life! Only as a model experiment: the SPL generated in a combo is about as big as the water
pressure at a depth of 1 cm. That should not deform a foil-capacitor to any substantial degree.
For an orientating measurement, some brand-new 22-nF-capacitors were charged to 200 V
and checked for microphonics: for SPL-values of 130 dB, the AC-voltage generated remained
below 0,03 mV. Assuming 30 V to be actual ac-voltage at the plate, this microphonics-
induced voltage would be smaller by factor of one million – for sure fully insignificant. Given
the multitude of capacitor constructions that have found their way into guitar amps we cannot
generally exclude that some capacitors would be among this crowd that exhibit much stronger
microphonics – but the likelihood has to be seen as extremely small.

What remains to be looked at? How about reports such as: “foil capacitors sound somewhat
different than polypropylene”? This statement opens up similar dimensions as “at night it is
colder than outside”. Very basically, capacitors may construction-wise be categorized into
foil-, electrolytic-, sinter- and air-capacitors; for the dielectric, polystrol, polyester, or
polycarbonate are in use, or the cited polypropylene. The typical polypropylene capacitor is a
foil-capacitor consisting – in the KP build – of two foils on top of each other (metal and
polypropylene, respectively), or – in the MKP build – of a metalized polypropylene foil. “If
you hold two fingernails at a distance of one mm, you get a capacitance of 1 pF” – H.H.
Meinke, unforgotten, r.i.p. To keep the in-between space in shape and enhance the insulation,
we insert a thin foil in there, e.g. a foil of polypropylene. This also increases the capacitance
by the relative permittivity (the relative dielectric constant) which for polypropylene amounts
to about εr = 2,2, while for polyester it is 3,3. Both plastics belong to the group of dielectrics
and therefore are insulating materials. The term “insulating” does however not imply that
there are no charge carriers within them – the difference is that they are not as easily
relocated. Current is nothing else than relocated charge: I = dQ / dt; i.e. no movement of
charge, no current. In a copper wire the electrons can be very easily moved around (at an
astonishingly low speed but in huge quantities), while in a dielectric there are next to no freely
movable charge carriers present. Still, there are charges: positive atom cores, negative
electrons, positive cations and negative anions. As a voltage is applied to the capacitor
electrodes, forces are exerted onto the charge carriers, trying to shift and bend them; this is
called the polarization. Since there are different kinds of charge carriers, there are also
different kinds of polarization mechanisms. They are the cause of the capacitor-losses.
All materials are “built” of atom cores and atomic shells (model of the atom according to
Bohr), and as an electric voltage is applied, an electron-polarization will occur in every
material: the electric filed-strength shifts the electron shell relative to the atom core. This
happens very quickly and is effective up into the THz region. In polar materials (e.g.
polyester), the permanent molecular dipoles rotate under the influence of the external
electrical field – this is called orientation-polarization. In materials containing ions, a
counter-shifting of anions and cations occurs: this is the ion-polarization. Finally, it can
happen in highly inhomogeneous materials that free charge carriers accumulate at insulating
grain boundaries – here we have the space-charge polarization. All these polarization effects
draw their actuation-energy from the electrical field and since none of these processes is
reversible, part of the electrical energy is irreversible converted into heat. This caloric energy
is not available to the electric circuit anymore (i.e. it is lost) – this is why we have “losses”.
Fig. 10.9.9 shows typical values.
Fig. 10.9.9: loss-factors in typical coupling capacitors. Data-book information (left), measurements (right).

Borrowing from macroscopic effects, we could say: the microscopic polarization movements
generate friction and corresponding losses; the latter are modeled as resistor(s) in the
equivalent circuit. It was noted already at the beginning of this chapter that dielectric material
properties are described ‘merely’ via ε and ρ. These material parameters are, however,
frequency- and temperature-dependent (to bring up the most important influences), and in the
general case also defined as direction-dependent tensors. Even using a simplified approach,
conclusions based on the infinitesimal small cube and applied to the volume of the real
capacitor do not result in merely one single resistor and one single capacitor but in a
complicated network with actually an infinite number of components. It is, however, possible
to recalculate this structure with good approximation into an impedance-equivalent (or
impedance-like) circuit. This equivalent circuit has a big advantage over the capacitor model
consisting of the two frequency-dependent components R(f) and C(f): it can be used to
describe processes in time. The latter would be not easily handled with frequency-dependent
components.
Fig. 10.9.10: Impedance-equivalent-circuit for a 22nF

polyester capacitor (continued-fraction expansion).
Such an equivalent circuit is shown in Fig. 10.9.10. It is not the only possible one – depending
on the desired accuracy, there are in fact myriad variants. In the diagram, we can see the
slightly rippled approximation that could easily be improved at the expense of the number of
components used. The chosen continued-fraction series expansion includes series-resistors the
value of which rises, from left to right, by a factor of 10 each, and parallel-capacitors the
value of which decreases, from left to right, by a factor of 1,78 each. (For a reduction of the
ripple, both these factors need to be reduced). To the right, the “ladder” continues until we
arrive at resistor values that correspond to the insulation resistance (lower cutoff frequency).
The continuation to the left determines the high-frequency trend of the loss-factor. For the
frequency range shown in the figure, the ladder does not need to be continued to the right at
all if the given component values are used. To the left, the continuation needs to happen up to
60 Ω / 1 nF; a parallel capacitor (20 nF) and a series resistor (0,17 Ω) conclude the circuit. We
can imagine that, for an insulation resistance of e.g. 1TΩ, the ladder is to be elongated further
to the right, but it then becomes also clear how small the additionally included capacitances
are (relative to 22 nF). For a capacitor terminated with extremely high impedance, this
extension might be required, but for a typical tube circuit (1 MΩ), an equivalent circuit with a
largest-resistor-value of 60 MΩ suffices as a good compromise.
Granted, this equivalent circuit is not that simple, either, but with today’s computer-support,
“impulse smearing” (group-delay distortion) can easily be determined. However, since a
change in dielectric (e.g. to polypropylene) has no audible effect for the typical tube-coupling
(as elaborated above at length), we will do without further explanations towards this.

Let us now turn to the question whether there could not be another reason why guitar
magazines again and again report of sound-changes due to capacitors. Of course, we
immediately remember the -20/+30% capacitance tolerances of the super-capacitors as
mentioned above. However, first there are components boasting narrower tolerance ranges
(the 60-€-mica-caps, for example, have a tolerance of merely 10%, at the most ), and
second, the reports related mostly to changes in the high-frequency range. If we insinuate that
this judgment does not relate to the MHz-range, but to the upper audio-range, what could be a
technically supported reason? The size of the capacitors! Not the capacitance but the physical
geometric dimensions. A coupling capacitor may come in axial build of the size ∅6x14, or the
size ∅22x35 (mm each). Does size matter? Depending on circumstance, maybe yes. Since this
capacitor is (relative to the resistors surrounding it) of low impedance at higher audio
frequencies, the plate-ac-voltage is connected across it – independently of the capacitor’s
polarity. Between this electrode-surface of several square-centimeters and all conducting
amplifier components, stray-capacitances result. In many guitar amps, the wire connected to
the grid of the tube in question is of the un-shielded kind, and this will create a small
capacitance between the coupling capacitor (plate) and the grid. This will not be a big
capacitance, maybe 1 pF or 2 pF. Although every amplifier is put together a bit differently,
with a big likelihood this capacitance will increase if a larger-volume-cap is incorporated. A
mere 2 pF – that doesn’t sound like much. However, we now need to consider the Miller-
effect that increases (e.g. for an ECC83) the grid-input-capacitance by 100 pF (or even more)
for any added 2-pF-grid-anode-capacitance. The tube itself has, according to the data sheet,
Cga = 1,6 pF, which yields (subject to the voltage gain) about CE = 80 pF. Since the circuit
build will not be totally free of capacitances, let us assume in the example CE = 120 pF. This
value would now be increased by the coupling capacitor to 220 pF. In conjunction with the
source impedance we now arrive at a
low-pass with 7,2 kHz cutoff frequency.
Do compare this number with the Megahertzes cited in the capacitor adverts and do consider
how big the reactance values could be here. Sure, not every amp has to be like that, indeed
there are countless variants: Fender- and VOX-amps the insides of which deservedly have
been called “birds-nests” of “cable jumble” already by other authors. Then there are boutique
amps with wires bent at exactly 90° angles, fiber-boards, turret-boards, PC- and PTP-boards,
source impedances of only 50 kΩ, but also of 250 kΩ, plus many more anomalies and
peculiarities. And, indeed, stray-capacitances. So, as our guru introduces a hand-wound cap
into the circuit with his heated iron while the circle of disciples holds devout silence, and as
he calls for a listening test: maybe the sound of the amp has actually changed. This is because
some wires were bent, because the plate-capacitors are moved closer to the grid wire by 1 cm,
because the performing guitarist doesn’t dare to dig into the strings as much in view of the
horrendous price, or because the loss factor at 100 MHz has suddenly been reduced. There are
even more possibilities, more things between heaven and earth, more knowledge, and more
BS (not meant as abbreviation for Bachelor of Science). Science is not always welcome in
this vicious cycle, and especially not the science of the electric current. Some authors in
musician’s magazines generally dismiss their perceived enemy (‘studied physicists’) and
advise to ‘give scientists a wide berth’. The latter will reciprocate right away, generally
disqualifying every non-technician (or non-scientist ) as not having any ability to do
scientific work.

Science requires reproducibility; the audiophile realm requires reproduction – that is

something different. If an admirer of the arts buys a multi-million-€ painting not as an
investment but just because he loves it, why would anybody carry out research to find a
technical reasoning? Why would the aesthete want to know whether the green used by Gaugin
might be greener by 0,221 nm than the green used by Dali? Somehow, this seems to be
different in the area of audio-technology. Here, the Strat-player argues with the particularly
high (or particularly low) weight of alder/ash, and the owner of a Plexi reasons with the
particular group delay distortion of the Yellow Mustards. A 330B-fan will replace the
polypropylene-capacitors by oil-paper caps because – as all the enlightened know –
polypropylene is a synthetic, or, in lay-man’s terms a plastic and so of course these caps will
have that horrible, synthetic plastic-sound (advertisement). Do not ask whether an oily sound
is actually preferable, because the 300B has entirely different problems: these oily comrades
come either with aluminum foil, or with copper foil. Copper has better conductivity, and
therefore – says the ad – the copper-sound will be better. A hand-wound aluminum cap will
set you back 12 € a piece, but that is anyway more within the low-cost segment in these
circles, and does not really match the matched triode-pair (at 250 €). And so copper-foil it is,
because: the conductivity is 60% better, and the price is 100% higher – that works as a
beginners-set. For the next birthday, we will nevertheless rather reward ourselves with the real
deal: with silver-foil capacitors, because: silver has still better conductivity, says the ad, and
who but the webpage of the manufacturer would know better. So: silver. There’s a lingering
memory from that dreaded latin class: silver – argentum – Argentarius? Sin-offering … no:
money business! That fits: big money business, because: there’s not just one coupling
capacitor in that radio – er: guitar amp, but there are two … no: three. Per channel! O.K.,
there’s the little box on the on-line order-form: enter “6”. And stay strong, as in the box on the
bottom the sum appears: 1101,00 €. ‘Tis the birthday – off into the shopping basket, done.
Well … just to be safe, enter “resubmission” for the next but one birthday – at the latest,
replacements should be acquired then, because: for Ag-caps, the manufacturer explicitly
mentions the minimum life-time: 2 years. That’s not difficult: acquire, solder in, wait for the
burn-in time to pass, listen, buy replacements, solder in, wait for the burn-in time to pass, and
so on. And in case anybody has any doubts at all about these mod(ification)s: data tables from
electrical engineering: indeed, the conductivity of silver is better than that of copper by 6%.
Though this be mod-ness yet there is method, or so Shakespeare notes.
The capacitance-tolerance of these money-capacity-robbing darlings is specified to +30% …

o.k., it is what it is, don’t get wound up, they are hand-wound. “Quality has its price (sic)”
You should not take too narrow a view on the fact that the auditory system can muster the
cited µs-resolution – if at all – only inter-aurally i.e. “between the channels”. The audiophile
writes in an internet chat room: hopefully this tolerance will not have a big impact in front of
a tweeter? No, no worries – tweeters are generally known to be very tolerant towards
minorities. Plus, if indeed any uneasiness remains, for sure there will be someone offering –
for something like 2022,00 € – a selected version with smaller tolerance. Don’t you even
think about the 1%-filter-caps! They are down the cheap end, and there’s no way they can
sound at all. If only the best is good enough: selected Ag-caps. Grab them every other year,
or every 10 000 km, whatever comes first.
By the way, what would the synthesis of idiographic♠ und diotic♠ be? Audiophile??
♠
idiographic = describing the very special; diotic = listening with the same signal at both ears.

Dielectrics for capacitors
Mica
Up to 125°C (max. 155°C). Relative permittivity εr = 5,5 … 7.
Also: Phlogopite mica, Micanite, Micalex, mica foil, Samikanite (with different data).
Mostly electron- and ion-polarization. Losses are frequency-independent (≈GHz).
Highest stability of capacitance over time; smallest temperature-coefficient.
Polystyrene (KS, Styroflex)

Thermoplastic, mostly electron-polarization.
Since 1936 up to 60°C, since 1953 up to 70°C (max. 85°C). Relative permittivity εr = 2,5.
Very high insulation resistance, very small losses.
Polypropylene (KP)
Thermoplastic, mostly electron-polarization.
Available since 1960; up to 85°C. Relative permittivity εr = 2,3.
Polycarbonate (KC)
Thermoplastic, mostly electron- and orientation-polarization.
Available since 1961; up to 100°C, max. 125°C. Relative permittivity εr = 2,8 … 3.
Polyethylene terephtalate (KT, Polyester)

Thermoplastic, mostly electron- and orientation-polarization.
Available since 1957; up to 100°C, max. 125°C. Relative permittivity εr = 3,3.
High insulation resistance, small losses.
Paper, impregnated (P, MP)

Sulfate cellulose, mostly electron- and orientation-polarization.
Characteristics depend strongly on density, water content and impurities.
Depending on the situation only moderate insulation resistance, small losses. Max. 100°C.
Capacitor-oil (Naphthenic oil etc.)

mostly electron-polarization; however: oxidization products (acids) are polar.
Relative permittivity εr = 2,2. Copper will accelerate the oxidization of the oil.
Depending on the situation very limited life-time.
Al2O3, Ta2O5
For electrolytic capacitors, not used in coupling capacitors.
Ceramics, e.g. TiO2

For ceramic capacitors; not used in coupling capacitors.

10.9.4 Sound event vs. auditory event
On the one hand, it is possible to document the operational behavior of a guitar amplifier via
formula and results of measurements; on the other hand, it may happen by verbal description
of sensory perceptions. “Smells like a goat” would be a genre-typical choice of words, or “has
one hell of an oomph and creates just the right sizzle”, so stick with auditory perception. If
everybody knows what an oomph is, this description indeed does help. However, because
scientists often do not know what an oomph is, and because they like to quantify things into
interval- and relational scales, there are also numerical specifications such as “cutoff
frequency at 5238 Hz”. So, we have, on one side, physics with its objective sound-event data:
100 W, 8 Ω, 5238 Hz, 10 ms. On the other side we find the auditory event with verbal,
subjective judgments; louder, much more authentic, vintage-like, throaty sound, too short
sustain, etc. In between there is the magnitude estimation: twice as loud as … , just noticeable
reverb amount, 50% longer sustain.
Guitar amps mostly do not play for measuring equipment but for people. Okay, they also play
for tables, chairs, the dogs of innkeepers and their fleas, but predominantly for people, after
all. Whether a measuring device certifies an increase of the effects-mix from 1% to 2% is
insignificant if this remains inaudible in both cases. The physical sound event leads – if it is
audible – to an auditory event, and it is only the latter that is judged by the listeners. The
assessment is anything but objective: whether an amp-sound is judged as being good or bad is
a matter of taste and depends on subjective criteria and also on environmental conditions.
Everybody knows optical illusions, and there is no surprise in the fact that there may also be
auditory illusions. Nobody will assume that a car speeding away on a straight road actually
decreases in size although the optical angle that it occupies in our visual perception indeed
becomes smaller. The brain will correct for the shrinking image on the retina and, in a way,
creates an illusion. Is it actually an illusion? The car has not shrunk, after all, just the picture
on the retina! Anyway, the term “optical illusion” found its way into everyday language.
What is the reason for such illusions? Is a lion that only then a lion when we see it in full, or is
it a lion already as it steps out of the bushes, only half visible? This is a clear-cut case of
evolution and/or selection. It was conducive to survival to supplement fragmentarily arriving
perceptions, and to correct distorted sensory impressions. The immense flood of data arriving
from our sensory receptors needs to be reduced momentarily by many orders of magnitude:
the data flow taken from a stereo CD amounts to about 1,4 Mbit/s but at best only 50 bit/s of
that arrives at our consciousness. However, the synapses working on our internal signal-
processing do not just throw, without discretion, 99,996% of the incoming information into
the bin; there are rules – but rules that may change from one second to the next, with our
cooperation but also without. Since we perceive our environment exclusively through this
information-reducing filter, the philosopher arrives at the conclusion: nothing is as it seems –
and he seems to be right. The “seems” is attributed to the realm of the perceptions (auditory
event), the “is” to the realm of physics (sound event). It must not surprise us if a guitarist
perceives sound changes if he is being told that a coupling capacitor has been swapped –
although the amp remained in fact untouched, and merely the judgment criteria have
undergone a change. The opposite may also happen: a capacitor is indeed swapped but
nobody hears a difference. And of course there is the third variant: the swap is clearly audible.
There are countless guitar amps, if not more – for the individual case no remote diagnosis can
be established. The following explanations can therefore only impart basic knowledge but not
offer retrofitting plans for specific amplifiers.

Fig. 10.9.11 shows some optical objects. In the first picture we see two crossing straight lines,
in the second two overlapping circles. Or are these in fact other objects? Aren’t there two
angles with meeting apex in the first picture? We could just as well assume that – but the
crossing straight lines are simply more obvious. Our brain always chooses the interpret-
tation of reality that is more likely. In this case, this is the crossing of two lines (or two tree
branches that have fallen on top of each other). For the same reason we do not recognize, in
the second picture, a crescent and a waning moon with a convex lens-type area in between,
but two circles. In the third picture, we see two triangles on top of each other that do not at all
exist in the drawing. In particular, the “upper”, white triangle is predominantly “make-
believe” rather than “actually being”. The right-hand picture conveys a depth in space that is
not at all present in reality. And although this picture does not change, it can “jump” in our
perception: one moment we see a cube on the floor, the next we see a cube hanging (fastened
with its rear surface to wall) towards the left … or towards the right. Visual perceptions seem
not to correlate perfectly with the optical stimuli.
Fig. 10.9.11: Examples regarding the visual perceptions of optical objects. For more examples see D. Picon 2005.
Consequently, we should not be puzzled if auditory perceptions change as well, without any
alteration in the acoustical sound event. A special experimental methodology is necessary to
establish whether or not there is in fact a causal correspondence between a change in our
auditory perception and a change in the physical sound event. How would a guitarist who has
just swapped a capacitor in his amp (and now plays to check out the result) judge whether any
perceived difference in sound is due to the changed capacitor, or due to the (unconsciously)
changed way of playing, or due to the (unintended) change in the listening position, or due to
changed judgment standards (autosuggestion)? Psychometrics has a few hints here: for
example, the sounds to be judged should be presented such that the test person does not know
which sound is presented at the given time (“blind”-test). The sounds should have a duration
of only a few seconds, and the interval between sounds should be short (about 0,5 s). In a
comparison of pairs (A-B-A-B) only a single parameter should be changed at a time. How
much does a demo-CD for replacement pickups tell us if there is a different guitar-riff for
each pickup, and if possibly different players have recorded the riffs? Not much!!
The first run-through of a listening experiment could, for example, contain simple nominal
verdicts: the perceived sounds sound the same or different. To increase the certainty of the
statements, it is necessary to have the subject judge identical sounds without the subject
knowing that this pairing is included. A subject that repeatedly hears differences when
identical sounds are presented (perceived A-B-A-B is in reality A-A-A-A) will either uncover
faults in the experimental setup, or he/she is unsuitable as a test subject. If two sounds are,
objectively, not significantly distinguishable, the question about which sounds better is moot.

If A and B are judged as sounding different in the auditory experiment, the second experi-
mental stage can serve to ask about comparative ranking characteristics (ordinal character-
istics): “I like B better than A.”, or “B sounds more distorted than A”, or certainly even “B has
more oomph that A”. In the last stage♣, quantitative cardinal characteristics are addressed: “I
would spend 100 € more for B.”
In order to judge the subjective difference in sounds between A and B, the exact objective
difference between these sound needs to be known – that should be a matter of course. For a
listening test on the sound of capacitors (Chapter 10.9.3), this implies that the amplifier is
always driven by the same signal, i.e. not by a guitarist (with a guitar) playing this now and
that then. Rather, the guitar is recorded once in an appropriate way, and this recording is fed
to the amp in an identical manner for the listening test. Specialist knowledge is indeed
required in order not to destroy the sound already by the experimental setup. As a result, the
following could be obtained: “Of 20 subjects only 3 could hear a difference between A and
B.” Or something like: “15 of 20 subjects judge A as sounding better but would on average
accept no more than 10 € additional cost.” could be the result. Still, even such tests leave
questions unanswered: anybody who has not personally participated will not know whether
he/ she would belong to a) the 15 or to b) the remaining 5, and if a), then the pecuniary
equivalent might be as much as 500 €, as well. In general: if I am asking for the opinion of
someone else, then I will receive the opinion of someone else – that is highly trivial. If I want
to rely solely on my own opinion, then I need to test everything myself (and why not?). If I do
ask another person, I might be i.a. interested in how reliable this person’s opinion is. In such a
case this approach holds: for a prejudice-free subjective judgment of objective issues, blind
tests provide a powerful tool.
But what about those instances when the sound of an amp changes without identifiable
objective reason? Those cases when an amp has lost its unique sound after a repair job,
although it was – embarrassment city! – accidentally shipped back without having been
opened up? The case of the guitar that never sounded right again after it had been kidnapped
for a stage-quickie by a pal. Or the case of the capacitor-swap that led to a sound miracle
although everybody (or rather all “studied physicists”) tirelessly continues to emphasize this
to be impossibly? There could be physical reasons (transport, shift of a slightly loose guitar
neck, stray capacitances), but we might also see in such cases the impact of judgment
benchmarks that are easily influenced. Most people fancy themselves to be superior to the
average in many areas, and prefer that their equipment to stand out from the mainstream:
alloy wheels … or copper caps. No sooner than a prejudice takes hold, it is pampered and
cultivated – the smallest confirming hint is scraped up and blown out of proportion while
every counterargument is conveniently ignored. As a rule, every confirmation is trustworthy
while every disagreement is questionable. No one is spared this kind of delusion: 94% of all
scientists at university deem their research to be above average! The deeper reason for our
biased dealings with information stems from a conflict between the search for truth, and the
search for harmony and for agreement with ourselves. To admit that one has been wrong can,
after all, chip away at one’s self-esteem and one’s image. [R. Degen, Lexikon der Psycho-
Irrtümer – lexicon of psycho-errors]. This is why an assumed change can lead to a change in
perception. If, after 100 h of playing the new capacitors, suddenly the treble comes to life, the
underlying mechanism is not necessarily an objective reason – the belief is already sufficient.
It is a rather big paradox that training can render our hearing more precise but at the same
time more susceptible to influence.
♣
These results may be achieved as well in a single run-through, if a matching evaluation-statistic is employed.

That the brain can be trained is without a doubt. Practicing for many years fine-tunes the
auditory performance, makes small differences stand out, allows for more comparison
patterns to be available, and enlarges the sensory areas in the cortex. From the awareness of
above-average hearing-prowess, the idea can easily arise that “the whole hearing” is now
perfected and has become the unswayable calibration-standard. In this, it is easily overlooked
that numerous auditory functions are not (or only to a very small degree) trainable, after all –
they function just as they do for the untrained and are therefore – relatively seen – worse off
than at the beginning of the training process.
An example from optical processing: for the cube in Fig. 10.9.11, we can decide whether we
want to see it as one whole object (the cube), or as individual lines. Everybody with normal
vision can do that; it does not require special training. For acoustical objects, however,
different rules apply: in a complex sound made up from partials (harmonics) it is much harder
to hear individual partials; often it is even entirely impossible. A simple trick may help: a
special (non-masked) partial is switched off (filtered out) for a short time and then switched
on again. At the switching-off instant we hear, as expected, a change in sound (thinner, more
hollow). As the partial is switched back on, there is a surprising effect: first, the thinner, more
hollow remaining sound is joined by an individually audible sine-tone that “melts” into the
remaining sound within a few seconds to eventually form the original sound. Something new,
especially when appearing abruptly, is deemed important, and the brains switches to “make
individual object audible”-mode. After some seconds, the new additional object is categorized
as a kind of prodigal son perfectly fitting in with all other objects, and the precedence circuit
is switched off again: the partial is not audible per se anymore. No training can change this
effect. The auditory perception changes although the sound remains static! On top of such
autonomous (endogenous) signal-processing algorithms, other external (exogenous) signals
affect the perception process: directional hearing is influenced by visual clues, as well, as is
the impression of reverberation and even speech intelligibility. Nothing is, as it appears, and
everything appears different
A real-life example shows how difficult listening test can be: in a pretty hefty pickup
comparison test (Gitarre&Bass 2/05), there are 10 pages of verbal assessments: "In
comparison almost mushy … the picking attack substantially softer and brittle … surprisingly
glassy and rich in harmonics … an entirely different spectrum in the mids … far less richly
colored … acutely transparent and translucent … a sound beautifully soft and compressed …
a very creamy tone that however seems a bit dull and lackluster … although completely
covered in wax, the pickups sound open and as airy as un-potted ones.” These short excerpts
indicated that clearly audible differences must exist between the judged pickups. Some 2
years later, the same magazine publishes a flash-back to the same test. This flash-back arrives
at the conclusion that “in fact all models sounded almost the same.” (Gitarre&Bass 5/07) The
difference between “entirely different” and “almost the same” has to be seen – according to
the flash-back – in the different recording situation. Mind you: for “almost the same”, the
recordings were not done in a garage but again in the recording studio, and getting “good and
professional results.” Based on this, every reader can pamper his personal prejudice: one will
shell out 400 € for a pair of PAF-clones and enjoy the exclusivity, the other will (because of
“almost the same”) stick with the equipment he already owns, and prefer to perfect his finger
vibrato – chacun à son goût. Another one may comment on the published sound examples
from the above test with “You must all be mad! There’s nothing to hear but one and the same
pickup again and again!“ (Gitarre&Bass 4/08). It does dignify the author of the article that he
has not withheld this comment from his readers.

10.10 Comparative Analyses 10-271
10.10 Comparative Analyses
For the most part, the chapters so far dealt with the analysis of special partial circuits. From
now on we will look at guitar amps as a whole. To begin with, the large variance in old
amplifiers should again be stressed: passive components could have tolerances of up to 20%
or even 30%; same-type tubes vary in their transmission parameters, circuits were modified
by the manufacturer without notice. Thus, Tweed Deluxe amps, for example, come within a
considerable scatter range – even if they carry one and the same 5E3 designation.
10.10.1 … for they knew what they did?
An old Princeton is comprised of no more than 3 tubes, 2 transformers, 11 resistors, 10

capacitors, 2 potentiometers – and many see it as ingenious miracle work of a brilliant circuit
designer whose genius-ideas to this day deny any analysis. The same holds for old Voxes,
rare Parks, original JTMs, or whatever else is called up as a precious gem. Well, while it may
not have been entirely trivial to develop a power amplifier, and to run a series production for
it, during war- or post-war-times, this did not require superhuman ingenuity, either. In most
cases, the basis was probably not much more math than U = RI und P = UI, supplemented by
the knowledge that a capacitor conducts the better the higher the frequency. Isn’t this what
distinguishes the circuit expert, oh dear editors of musicians’ magazines? During the war the
following could easily happen: a chap more or less enthusiastically joined the Royal Air
Force, was really annoyed with the wireless constantly breaking down, enjoyed a surprising
success after replacing a blown capacitor, was as a result promoted to technician (or even
engineer) – and had laid down all the groundwork for a later career as circuit designer. Not
that the actual theory was unknown: in the Langford-Smith compendium – published for the
first time in 1934 – there are hundreds of pages of the basics of circuit design that to this day
is worthy of being taught at university. But times were difficult and not everybody who
wanted to go to college could do so. Back then, that is. That in 2007 a circuit-“expert” at a
well known German musicians’ magazine makes a statement along the lines of “more than
400 V flow through such plastic stuff” … shows a kind of congeniality, somehow ...
For the old circuits, it is impossible today to know what was the result of an intentional
development towards a clear aim, and what “just happened”. Presumably, the designers back
in the day were not entirely sure themselves what exactly they soldered together. For one, the
technical education probably left something to be desired in many cases, and the same was
most likely true for the available equipment, as well. There were no PCs in 1950, and neither
had “electronic calculators” been developed yet. Transistors were available merely as
prototypes in R&D – the lab equipment was exclusively tube-powered. That did work pretty
well for a tone-generator and an oscilloscope, but already distortion measurements posed a
serious challenge. It wasn’t impossible – HP (from 1939) and B&K (from 1942) offered audio
measurement equipment – but it was expensive. For a small Brüel&Kjaer audio measuring
station, even as late as 1987 one had to shell out (in €-equivalents) 13 grand for a level
recorder, 8 grand for a sine-generator, 13 grand for an FFT-analyzer, 30 grand for a distortion
analyzer and another 60 grand for a 1/3-octave analyzer … summed-up € 124,000.-. The two-
channel version would have set you back another 21 grand, and had you decided to go for a
printer … that luxury (color? Dream on, my friend: black only) would have added 14 grand
more. The printer alone would have been the equivalent of three brand-new, fully gassed-up
VW Beetles. At the time when the famous amp-forefathers were put together, their designers
were mostly ham-radio “amateurs”, just barely beyond their teens. In no way could they have
afforded a full set of the wonderful light-green B&K-equipment. At best, they operated a
tone-generator, an oscilloscope, and one or two “MaVoMeters” - plus a soldering iron.
© M. Zollner 2011-2013 Translation into English by Tilmann Zwicker

10-272 10. Guitar-Amplifiers
Indeed it was possible to build an amp with only little equipment; the circuits were known. In
his book about VOX, Jim Elyea relates that it was customary to nick left and right from the
competitors. Well, he doesn’t actually write “nick”: "JMI, like everyone else, borrowed
literally wherever appropriate... It was not uncommon for the engineers at JMI to bring in the
equipment of other manufacturers, take them apart for ideas, put them back together, and sell
them in the shop". Ideas were “borrowed” … an approach in practice as late as 1984: just
before the Frankfurt music fair, a CEO (who shall remain unnamed, as shall his company) had
“his own” face plate mounted on a Japanese competitor’s device, and proudly presented it at
the fair as the newly developed reverb. Just to be safe, and to avoid that somebody else would
nick it in turn, he took it to his hotel room every evening. 'Knowhow-Transfer' was and is
common – and not just in the Far East. Marshall’s JTM is a copy of a Bassman, the tremolo-
effect for guitars previously was used in organs, the VOX tone control is derived from the
Gibson GA-70 (in turn inspired by the Fender Pro 5E5-A), Marshalls 18-Watt amp previously
was already successful on the market as Watkins Dominator. Gibson 'disassembled every
Fender-Amp' [Elyea /Smith]. Of course these were mostly not actual 100-%-copies: one’s
own ingenuity has to come out somewhere. (Fig. 10.10.1).
Fig. 10.10.1: Input circuits of various guitar amplifiers.
In Elyea’s VOX book we repeatedly find hints that Dick Denney’s prototypes were “bird’s
nests” – heaps of components artfully soldered together. 'Dick's working was a more organic
approach, involving endless fiddling with individual parts until he got the sound he was after.
He didn't care what the value of a part was; all that mattered was if it sounded right.' Denney
had a severe hearing loss but that did not get in the way. No, not because that would have put
him on the same “ear-level” as his customers, but because despite the damage in his ears he
was aware of what the marked demanded. He was not always aware of the inner workings of
his circuits. Only when he dropped his screwdriver into the circuit (shorting two wires) did he
discover that his Wurlitzer-inspired phase modulator also could accomplish amplitude
modulation [Elyea]. Indeed, that Vib/Trem-channel ... it includes a 500-Hz-highpass followed
by a further high-pass of a cutoff frequency of 8 Hz (0,8 Hz for the bass version). That’s how
it’s done and that’s how it is passed down from generation to generation. Or the JTM-45, Jim
Marshall’s holiest cow: nowadays available as reissue, but with a changed electrolytic
capacitor at the cathode. Our musicians’ magazine recommends: “the 330-µF-cap should be
replaced by one with 250 or 220 µF. This minimizes the bass a bit.” The explanation with a
better match would have been: since it was that way in the original circuit. Let’s do a simple
estimate: with an internal resistance (1/S) and the cathode resistor of 820 Ω we get a pole-
Translation into English by Tilmann Zwicker © M. Zollner 2011 - 2013

frequency of 2 Hz (330-µF), and 2,7 Hz (220 µF) respectively. For a more exact calculation
we would have to consider the plate resistor, as well … however: is that really necessary?
How relevant is the operation at 2 Hz in a guitar amp? We might look into the issue of
transient phenomena – had not our author in the expert journal written in another passage that
jumper wire would sounds differently compared to stranded wire, and that silver wire had a
“cruel” sound. Here we touch the world of HiFi, where wire with blue insulation sounds
more airy, while wire with brown insulation sounds somehow … shitty. Side-note: for
electrolytic capacitors, the exact capacitance was never really of much importance – it is not
uncommon to find tolerances of e.g. +50/-20% printed on the housing.
So, how did that 250-µF-cathode-cap arrive? That’s truly difficult to assess and we can only
speculate. The circuit of Leo Fender’s very first Bassman (5A6) is shrouded in time and
mystery – it seems nobody has actually seen a drawing of this circuit⊗. In any case, the
closely related Pro Amp (allegedly delivered with the very first P-Basses since the Bassman
Amp was not ready yet) had double-triodes (6SC7) with joint cathodes in the input circuits,
and it included normal 25-µF-caps. The second variant Bassman (of which the circuit diagram
5B6 is available) also featured the same-type double-triode with joint cathodes, but sported
the infamous 250-µF-biggie – almost as big as the power-supply filter cap. Why was that?
Some thoughts about that:
1. While in the active channel (of the two-channel input configuration) the signal from grid
to plate is inverted, it also reaches the anode via the other channel (2nd half of the tube). This
common-grid signal path is non-inverting such that in the plate two out-of-phase signal are
summed and thus there will be an attenuating effect. However this happens only at very low
frequencies since the route via the cathode is a low-pass. With 25 µF a loss of 3 dB would
have occurred at 2 Hz – more than adequate even for a bass guitar, and not really any reason
to up the cap by a factor of 10.
2. The big cap was supposed to eliminate hum induced by the tube heater. That may actually
be a possible reason – however the Pro Amp did quite well with 25 µF at the cathode.
3. A bass amp needs to operate at low frequencies. O.k. – but as much as 250µF? Both
output transformer and speaker are far from able to carry such infrasound to any extent.
4. Someone in logistics misread Leo’s handwriting and accidentally ordered 1000 pieces of
250-µF caps instead of 25 µF. Hm … maybe not.
The mystery remains – from other angles as well: why does Leo keep the 250 µF as he
switches to the 12AY7? Now he’s got a modern double-triode with totally separate systems,
but still he maintains the big 250-µF cap in the Bassman. He holds on to it for years – until
the completely redesigned 6G6-Bassman, when suddenly the “small” 25-µF cap suffices. Just
as it suffices in the Deluxe, but there it had held its own from the very start (5D3). Same as
for the Pro (5D5), as mentioned above, and for the Super (5D4). They all got by with 25 µF.
Only the Bassman features the 250 µF. It’s a bass amp, after all, so let’s accept it. Next,
however, Marshall’s Ken Bran copies the Bassman and it becomes a guitar amp – and
naturally keeps the 250-µF-cathode cap. Since then, all Bluesbreaker imitators adamantly
insist on that cap … most likely because, now as it was then, they may not always exactly
know what they are doing. Because the cutoff frequency is so excessively low, we could look
for other criteria: for example for transient phenomena that play a role as the tube is
overdriven. Still, no find – Marshall will use the 820-Ω-resistor for the two cathodes
⊗
The circuit „Old Bassman“ or even „5A6“ found on the internet at the time of this writing can NOT be a
Bassman but is highly likely to be a Dual Professional in view of the two speakers, the dual output transformers,
the three inputs, and two volume controls (contrasting a single speaker, one output transformer, two inputs and
one volume control on the Bassman).

connected in parallel, and then again also for single-cathode operation. Shouldn’t they have
brought in 1600 Ω for the latter? No, they didn’t. Not that Fender looks much better: they
change from the 12AY7 to the 12AX7 without matching the cathode resistor although these
are quite different tubes. Who cares, as long as the contraption doesn’t go up in flames. The
RCA-Receiving-Tube-Manual recommends 1,5 kΩ as cathode resistor for the 12AX7 (at
300V/100kΩ) – that may have been the starting point for it all, and that was then somehow
copied, and copied again, and again …
With respect to the design approach of early guitar amplifiers, the VOX book gives
interesting insights: 1935 for the first time an effort was made to do more than amplify the
signal of an electric guitar. Rather, the idea was to alter the tone, both making the electric
guitar a different instrument, not just a louder guitar, and also making the amplifier itself an
important part of the sound [Elyea]. Some manufacturers, however, arrived at this realization
only much later: In late 1957, it was a natural to apply the Hi-Fi designation to the new
amplifier (VOX AC2/30). Similarly, Dave Funk reports about the early Bassman: Everything
was very technical, hi-fi, and by the book. The first guitar amps either included no possibility
at all to influence the sound, or merely included a primitive tone control to attenuate the treble
range. Dick Denney’s VOX AC15 followed this design approach, as well, and was supposed
to reproduce as “HiFi-like” as possible. At first, the “Normal” channel sported merely a
control to diminish the treble. The lower cutoff frequency of lower than 20 Hz was
determined by the values of the coupling capacitors, while the upper cutoff frequency of
about 17 kHz resulted from unavoidable stray capacitances. This configuration would have
done a good job in a music box, as well. It was only the power amp that refused the trend – to
include negative feedback. Apparently the amp worked better without it, as had the amps of
Fender, Gibson, and many more. Indeed, dispensing with negative feedback was not an
invention that VOX came up with. To have a 500-Hz-highpass in the Vib/Trem-channel be
followed by a further 0,8 Hz high-pass – well, that actually may be “VOX-typical”.
In Fig. 10.10.2 we see the frequency responses of two Bassman amps (are these then
‘Bassmen’?) from input to the second stage. Compared to the fundamental of the (regular)
lowest string of an electric bass (E1 = 41.2 Hz), the 5B6 appears quite a bit ‘oversized’. For
the later 5F6-A, the lower cutoff frequency even depends on the position of the volume
control of the “other” channel, which would appear to push the significance of the lower
frequency limit even further into the background.
Fig. 10.10.2: Frequency responses of two Fender Bassman amps. String-fundamental: Bass (E1), guitar (E2).
Thus we shouldn’t look for reasons that never existed. Much resulted from circumstances that
cannot clearly be seen anymore today, or happened due to pure chance and by accident.
(Translator’s note: at this point Manfred Zollner makes a comparison to processes which may have influenced
how literature was written. As an example, he relates to the well-known poem “Der Erlkönig” written in 1782
by famous German poet Johann Wolfgang von Goethe. Since this passage “works” only in German, it was not
translated and is not included here – please see the German version of this book if you are interested.)

10.10.2 Stage-topology, level-plan
A guitar amplifier contains several consecutive amplification stages, in a sequence typical for
the respective amp type. The total frequency response results from the commutative
multiplication of the individual transmission-functions. As such the order of the sequence
would be irrelevant; however this only holds for linear operation – and that is not the only
operational state of a guitar amp. Whether preamp- or power-tubes are overdriven makes a
difference, and in which frequency range this happens, plays a role, as well. When comparing
different amps, we therefore need to consider the sequence of the stages.
General differences were already highlighted in Chapter 10.2; now special amplifiers are at
the center of attention. When comparing, we run into a huge number of parameters, and we
need to simplify rigorously. Because 3 to 5 tube stages follow each other in a typical guitar
amp, a multitude of combined nonlinearities may exist. In addition, filtering in and between
the stages happens – the effect of which we may not be able to account for at the first glance.
For example, a simple volume control may also have the effect of a treble filter with the
frequency response depending on the position of the wiper, and also on the input capacitance
of the subsequent tube. If at that point there is also a summation of two channels, the volume
control of one channel may influence the frequency response of the other channel, as well. To
limit the number of representations, we decided to measure all amplifiers with a standard
setting. The volume control was positioned such that for an input voltage of 90 mV (at 500
Hz) the power stage was just starting to clip. Why 500 Hz? Well, a choice needs to be made –
673 Hz or 1000 Hz would also have been o.k., as is 500 Hz. Why 90 mV? Your run-of-the-
mill singlecoil pickup will confidently reach that voltage: Telecaster, neck pickup, normal
picking strength – 90 mV. Maybe a bit more or a bit less, but – again – we need to pick a
value. Some arbitrariness is unavoidable here. The same holds for the terms “maximum level”
and “clipping”. For an operational amplifier, the clipping limit is clearly definable, but not for
a tube featuring a continuous increase of distortion. Since for a guitar amp, HiFi-standards are
out of place, we chose as the limit the level at which the total harmonic distortion (THD)
products are 25 dB below the primary signal.
The tone controls were adjusted to generate a treble boost typical for the genre. The general
frequency response was dictated by the amplifiers that offered only few possibilities of
control (Tweed Deluxe, AC15). The other amps had to comply as far as possible. You may
ask: “why would I want to adjust a VOX such that it sounds like a Fender?” While that is a
legitimate question, it also tempts to go the second step before the first. Not to have to
evaluate at the same time different distortion sounds and different frequency responses is
highly conducive for a comparison. It is helpful to be able to concentrate on the non-linearity
while keeping the linear behavior similar. As mentioned before: there are myriads of
possibilities, and other priorities may be purposeful, after all – but they would push beyond
the present scope.
Fig 10.10.3 shows the block diagrams of some amplifiers; the differences in the sequence of
stages are striking. The tone filter (the oval with arrows) is located after the first tube in one
amp, after the second tube in another, and in some cases it is driven by a cathode-follower
(two overlapping circles). In some amplifiers, the volume potentiometer is bridged with a
capacitor that is switchable in some cases (Bright Switch). Coupling capacitors were only
included here if they caused a very high lower cutoff frequency (VOX). Additional second
channels are indicated via a resistor with a free end. The last stage included in the diagrams is
the phase inverter (PI). The respective gain of each stage is indicated in dB and given at the
standard setting (f = 500 Hz).

Fig. 10.10.3: Block diagrams

Fender’s Tweed Deluxe (5E3) has only a single
tone control as sound filter boosting or cutting the
treble. The phase inverter is a cathodyne-circuit
with a gain of 1. Fender’s Deluxe Reverb (AB763)
already sports the widespread Treble-Bass-filter,
and uses the differential amplifier as phase inverter
(as do all the following amps). In both Deluxes, the
push-pull power stage employs the 6V6-GT, with a
cathode resistor in the 5E3, and with negative grid
voltage in the AB763. The Super-Reverb (AB763)
is similar to the Deluxe-Reverb in many details, but
has two 6L6-GC working in the power stage. We
will not expand on the fact that in all these amps,
the loudspeakers are different as well. The Bassman
(5F6-A) – in fact intended to be a bass amp – is
highly regarded by guitar players. It is the only
Fender amp considered here that includes a cathode-
follower, and it distinguishes itself in other ways, as
well, over its colleagues. The Treble-Bass-Middle-
filter, for example, is located towards the end of the
signal chain; it is supplemented with a presence
filter integrated into the negative feedback loop. Jim
Marshall’s JTM-45 looks very similar – no surprise
there since it is a Bassman copy. Only the tubes are
different: instead of the 6L6-GC we find the KT66
as power tubes, and in the input amp the slightly
more “gainy” ECC83. Several developmental stages
are documented for the VOX AC30: the four-input
AC30/4 features merely a switchable high-pass as a
tone control, plus a low-pass within the phase
inverter. In the input stage there is a high-gain
pentode that is however replaced already in the
AC30/6 with the ECC83. The AC30/6 looses the
high-pass-switch but adds a “Normal” and a
“Brilliant” channel (on top of its “Vib/Trem”-
channel. Finally, the AC30-TB adds a Bass-Treble-
filter to the Cut-filter, and also includes the
distortion-promoting cathode-follower. Similarly
simple as the AC30/4 was the AC15 (with only the
power section being different), and Marshall’s 18-
Watt shares this approach: there is only single tone
control offering a choice of treble- or bass-cut.
For all these amps, the sequence of the sections in the system determines their (over-) drive
levels and thus determines the sound. Even if the behavior (i.e. the frequency response) at low
signal levels (with a single coil pickup) is similar, connecting a humbucker (i.e. a higher drive
level) will make differences in the sound audible. The same may happen with a single-coil as
the volume is turned up. Here, some amplifiers offer a surprising, even incomprehensible
reserve: fully cranked, 3 mV at the input of the AC30/4 is sufficient to fully drive the power
stage. No, there was no Heavy Metal at the time of the debut of the AC30 (around 1960). But
back then they may have used the amp as all-around PA system, i.e. for microphones, as well.

The so-called level plan offers a possibility to depict the voltage levels as they pass through
the amplifier, but unfortunately it has a distinctive disadvantage: showing only one frequency
is insufficient, and showing multiple frequencies is confusing. The approach may be adequate
in studio technology where clean (and often similar) equalization stages (such as the
Baxandall tone control) abound – for the multitude of filters we meet in guitar amps
supplementary representations are required. For the AC30-TB and the Super Reverb, the level
plans are shown in Fig. 10.10.4. From these we can see that differences occur particularly in
the second amplifier stage: while both power stage and cathode-follower start limiting at the
same input drive level in the AC30, the second stage in the Super Reverb still has a reserve of
17 dB when the power amp goes into saturation. The result is that in the VOX both power and
intermediate stage significantly contribute to distortion while in the Fender the distortion is
predominantly generated by the power stage. At 500 Hz, and with the chosen setting, that is
… because as we turn the knobs, the level plan changes, as well.
Fig. 10.10.4: Level plans for the VOX AC30-TB and the Fender Super-Reverb, f = 500 Hz.
There are only a few amplification stages but many frequencies, and therefore we will not set
up a level plan for every frequency. It is more conducive to present the frequency dependence
of the drive-limit every stage has (headroom chart, chapter 10.10.3), and to include only the
tube stages since passive RC-circuits do not show any distortion in the context of the present
investigations. Fig. 10.10.5 reveals that in fact one drive-limit is not sufficient: the drive-
dependency of the HD (harmonic distortion) has many variants (more on this in Chapter
10.10.4). The right-hand section of the figure shows the frequency response from input to
output (loaded with a speaker). The small resonance spikes and part of the treble boost are
caused by the speaker-impedance. The two frequency responses are not identical but at least
they are similar, something that cannot be said of the HD: as the Super-Reverb goes into
overdrive, it generates strong 2nd order distortion (on top of the k3 not shown here) while for
the VOX, the k2 may be neglected in comparison to the k3. By the way: so much for the
statement “compared to transistors, 2nd order distortion is dominant in tubes”. Again: more
on that in Chapter 10.10.4.
Fig. 10.10.5: 2nd order harmonic distortion ak2 (500 Hz) from amplifier input to power-amp output (left),
frequency responses in standard setting from amplifier input to power-amp output (right). ak = 20lg(1/k)dB.

10.10.3 Headroom-chart
“Headroom” means drive margin i.e. “how much more gain until overdrive”. The headroom-
chart is the graphic representation of the frequency response of the headroom. This chart
shows a frequency response for each amplifier stage – it is not “transmission frequency
response” but the frequency response of the headroom relative to the drive limit (clipping) of
the power-amplifier. Since this clipping is the reference, it is represented by a horizontal line
at 0 dB. If, for example, the curve for the 1st stage at a specific frequency is indicated to be at
12 dB, then this1st stage can be driven with 12 dB more until clipping than the power amp. In
other words, as the power amp goes into overdrive at this frequency, the 1st stage still has a
reserve of 12 dB, or, as the 1st stage goes into overdrive, the power amp has already been
pushed into overdrive by 12 dB. In Fig. 10.10.6 we see four headroom charts. For the Super-
Reverb (normal channel) the curve for the 2nd stage runs almost constant at -17 dB indicating
that this stage starts distorting only as the power amp is already overdriven by 17 dB.
Conversely, the 2nd stage of the VOX (AC30-TB, brilliant channel) at 100 Hz has a mere 4
dB margin, and at 1 kHz, the 2nd stage and the power amp go into overdrive at approximately
the same input signal level. The drive margin of the 1st VOX-stage decreases towards low
frequencies because a high-pass between 1st and 2nd stage attenuates the bass transmission.
Fig. 10.10.6: Headroom-Chart for Fender Super-Reverb (AB763, upper left), VOX AC30-TB (upper right),
Fender Tweed Deluxe (5E3, lower left), and Fender Deluxe-Reverb (AB763, lower right).
The higher the curve is located in the chart, the smaller the drive margin is relative of the power-amp clipping.
The 2nd stage of Tweed Deluxe directly feeds the phase inverter (via a capacitor), and the
headroom chart therefore runs in parallel to the horizontal power-amp-line. However, with the
cathodyne-circuit of the Tweed Deluxe not having any voltage gain, there is much less margin
compared to e.g. the Super-Reverb. The Deluxe-Reverb, on the other hand, is much closer in
circuit design to the Super-Reverb than it is to its ancestor Tweed Deluxe. It does not quite
reach the high margins of the Super-Reverb due to its lower supply voltage. As we change the
setting of the volume potentiometer, it is only the curve for the 1st stage that also changes:
the larger the amplification, the more this curves sinks to the bottom (= larger drive margin re
the power amp). With no control located between the 2nd tube and the phase inverter, the
curve for the 2nd stage cannot be changed. This is in contrast to the Fender Bassman and its
Marshall-clone, the JTM-45.

For most Fender amplifiers, the tone control is located ahead of the 2nd tube stage – but from
1954 to 1956 some amplifiers were designed with the (in-) famous cathode-follower as 2nd
stage, and the tone control positioned behind it. We may surmise that the RCA Receiving-
Tube-Manual in its 1954 issue (RC 17) is the source; it introduces for the first time a 12AU7
with cathode-follower and "Bass and Treble Tone-Control". Incidentally, the Twin (5D8)
receives a very similar circuit that year, and designers at Gibson take up the same idea and
include a cathode-follower into the GA70/77 (although they do change the tone control
circuit). VOX, however, does not really bother with alterations and simply adopts the Gibson
tone control 1:1. The cathode-follower driving the tone control is deployed in the 5D6-
Bassman, as well – that amp spawned the inspiration of Marshalls Ken Bran and his JTM-45.
The respective tubes are all configured in the common-plate-circuit (= cathode-follower) but
the tubes themselves and the details of the circuits vary. RCA shows the 12AU7, Fender
initially includes the 12AY7, with Gibson, VOX and Marshall, the 12AX7 is found. All amps
use a double-triode, i.e. a tube containing two independent triodes within one glass container
– independent but equivalent. This is actually not that advantageous because the first tube
system operates on common-cathode mode (no AC at the cathode) but the second tube system
operates in common-plate mode. The first tube system is to amplify the voltage, and the
second should amplify the current (Fig. 10.10.7).
Fig. 10.10.7: Double-triode w/cathode-follower (Chapter. 10.2.2). On the right: circuit of the AC30-TB.
Fender first deploys the 12AY7 (Fig. 10.10.8) but then changes over to the 12AX7, the
amplification of which is somewhat larger but which features less drive margin. The reason:
the first tube can reduce its plate voltage (and correspondingly the output voltage) only down
to about 120 V. This is what drives the second tube as it conducts. At the output we thus have
available no more than about ±35 V (for modest distortion). The subsequent tone control
circuit attenuates the signal by about 15 dB, and now there might not be enough signal
strength left to fully drive the power amplifier.
Fig. 10.10.8: Output characteristics of the double triodes used for the cathode-follower (Fender).

In the Fender amp, the cathode-follower therefore generates merely just about the voltage
required to fully drive the power amplifier – that may be the reason why it is not used
anymore from ca. 1960. Not so at VOX, where the cathode-follower enters the picture at the
time when it is shed at Fender. VOX, however, does not “borrow” the circuit from Fender but
from Gibson where the cathode-follower is first included in the GA70 and GA77♣.. It receives
a rather astonishing dimensioning, too: with the changeover from the 12AY7 to the 12AX7, it
is not the first cathode resistor that is halved in value but the second – for whatever reason.
With this resistor, the quiescent current of the second triode (Fig. 10.10.9) becomes large
enough to cause a considerable grid current to flow, which again has consequences on the
drive situation and on non-linearity (chapter 10.2.2).
Fig. 10.10.9: Output characteristics of the double triode used in the cathode-follower (VOX AC30-TB).
Of course, we may surmise that it is exactly this non-linearity that is required for a good
guitar-sound. But then: why do Fender and Gibson not continue with the approach, why does
Leo Fender try, shortly after the debut of the cathode-follower, to decrease this non-linearity
via negative feedback (e.g. Super 5E4 – 5F4)? Why does it disappear from all Fender amps
after about 1960? Mind you: that was still pre-CBS! In retrospect, many decisions are
glorified into strokes of genius – which they probably weren’t. Elyea’s book on VOX can
easily live with such discrepancies: on one hand Dick Denny designed that AC30 exactly
according to his own ideas, on the other hand the TB-circuit (cathode-follower and tone-filter)
is an exact copy of the Gibson amp. On one hand it is the EL84-power-amp that creates the
sound, on the other hand the originally used EL34 was discarded not because of the sound but
because it would have made the amp "two inches too tall". On one hand Dick’s amp had
"more clean headroom than most other amplifiers", on the other hand it featured "high
harmonic content" and "plenty of even numbered harmonics". Measuring the output voltage
reveals something else altogether: lots of odd-numbered harmonics (chapter 10.10.4). Besides
all speculation, there is an objective reason: for full drive levels, the phase-inverter of the
AC30 requires less than 10% of the voltage necessary in a Fender (EL84 vs. 6L6GC), and
consequently the inferior drive situation created by the cathode-follower could be more easily
tolerated compared to a Fender amp. So what about Marshall? Ken Bran does not copy the
VOX approach but adopts the Fender circuit. The situation here is rather tight with regard to
maintaining sufficient level so the tone stack fed by the cathode-follower is optimized to have
low basic attenuation (Fig. 10.3.12). Marshall’s PA-amplifiers document the fact that the
cathode-follower was not regarded as a special guitar distortion device: all microphone
signals – not something you would want to distort - had to pass through the cathode-follower,
as well. Gibson advertised their amp (with cathode-follower) as having "unusual clear bell-
like treble". What else indeed – it was 1958! Distortion was called for only later.
.♣
A variant of the Gibson GA30 temporarily featured a cathode-follower, as well.

Let’s speculate some more. Possibly, some designers believed that a tone filter would only
work properly if driven by a (actually or only supposedly) low-impedance cathode-follower.
That would explain why in the AC-30TB it is only the “brilliant”-channel that has the c-
follower but not the “normal” – or the “Vib/Trem”-channels – the latter do not feature such a
filter. Would the cathode-follower have been considered important to the sound, surely all
channels would have been fed to it. Only with the AC50 both the “brilliant”- and “normal”-
channels each receive their own c-follower – because each channel has its own tone filter. In
the JTM-45, Marshall’s first amplifier, the power amp includes strong negative feedback and
therefore requires a relatively high voltage to be fully driven. The preceding cathode-follower
therefore had to be strongly driven and might have caused distortion. Was this intended?
Apparently not, because very soon the negative feedback is reduced♣, and the c-follower
distortion decreases. Is this why all over the planet the very old Marshalls are sought after
most? Maybe. Or maybe not.
Both triodes (12AX7) in the c-follower of a JTM-45 (Fig. 10.10.7) need to be driven strongly
and may distort. This distortion, however, is highly dependent on the individual tube, as can
be seen from Fig. 10.10.10. In particular, the 2nd order distortion may change by a factor of
more than 10 as one 12AX7 is replaced by another12AX7. Thus “the first tube is the most
important”-rule (as it can be read here and there) is not correct here – it is the second tube
that’s important. At the same time, we must not make a connection to particularly old tubes
since while these may be great, they may also be bad just as well and do not justify any
surcharge. As has been shown already in Chapter 10.1, tube characteristics show different
curvature and therefore give different distortion. It would be helpful if some of the “expert”
writers in various magazines would for once support their monthly elaborations (“for
Marshalls from early 64 to late 65 use only Brimar tubes in the input stage”) with a
measurement of the tube characteristics or distortion. It may be that in a particular specimen
of a Marshall the individual Brimar 12AX7 makes for a super sound. It shall also not be
questioned that a guitar player who has been writing tests and other reports eventually can
judge what a good sound is. What needs to be criticized, however, is the approach to turn such
insights into undocumented sweeping judgments that are incorrect in this generalization.
Fig. 10.10.10: JTM-45, harmonic distortion of the cathode-follower; four individual 12AX7. RQ = 200 kΩ.
At this point, we will not continue discussing harmonic distortion of the individual amplifier
stages. Details on this will be included in Chapter 10.10.4. First, the headroom-charts of a few
more amplifiers need to be analyzed – these are amps in which the tone filter is not located
after the input valve but immediately ahead of the phase inverter (Fender, Marshall).
♣
First taken from the 16-Ω-tap via 27 kΩ, then via 47 kΩ, finally via 100 kΩ from the 8-Ω-tap. Tubes (KT66,
EL34) and primary impedance (8 kΩ, 3.4 kΩ) varied, as well.

An objective analysis of Marshall-distortion is hampered by not only four tone controls that
need to be considered, but also by the fact that Marshalls came with two different output
tubes, two different output transformers, different negative feedback, various shunt-capacitors
– just to name the most important versions … there were additional issues for short time
periods. Fig. 10.10.11 shows some selected examples: on the upper left there is a standard
setting matching Fig. 10.10.6. On the right we see the ancestor, on the lower left the setting
for forgetful guitarists (all on 10). On the lower right there is a variant deriving its treble boost
mainly from the power amp (Presence control set to 8).
Fig. 10.10.11: JTM-45, headroom-chart. In this amp, the volume-control was not bridged by a capacitor.
As a comparison, measurements of the Fender Bassman (5F6-A) are given at the upper right.
The weak dynamic range of the second amplifier stage is striking. As the tone controls are
turned up, the filter attenuation drops and the second stage is given a larger dynamic range.
With increasing amplification the first tube reaches a larger range (N.B.: re the power amp!).
Fig. 10.10.12 clarifies the step from the JTM-45 (KT-66) to the JTM-50 (EL34): swapping
the output tubes (with bias adjustment) and the output transformer slightly reduces the gain
margin for the 2nd stage. Additionally decreasing the negative feedback in the power amp cuts
back drive levels to the 2nd stage and improves the dynamic range. (Supplemental info on this
in chapter 10.10.4).
Fig. 10.10.12: JTM-50 (EL34), power-amp feedback 27 kΩ / 16 Ω (left), 100 kΩ / 8 Ω (right). Over the years the
negative feedback was reduced and thus the gain margin of the 2nd stage increased.

To conclude, let us have a look at a few amplifiers without intermediate stage: in Marshall’s
18-Watt amp (built from 1965 – 1967), the plates of the two input triodes are simply
connected together which is the source of considerable preamp-distortion (L. Fender had tried
this already 13 years earlier in the 5B6-Bassman). Apparently, that was not desirable (at least
then!) since the 20-W-successor sums in the conventional manner. The VOX AC15 sports a
pentode in the input circuit, just like the successor AC30/4 with 4 inputs; it is said to have
been microphonic and unreliable. For this reason, there is a swap to the ECC83 in the AC30/6
(extended to 6 inputs). There were 3 versions of this amp: Normal, Bass, Treble, and it is not
yet the actual AC30-TB – that then finally received the distorting cathode-follower as the 2nd
amplifier stage.
Fig. 10.10.13: Comparison Marshall JTM-18, VOX AC15_1960, VOX AC30/6_Normal, VOX AC30/6_Treble.
Fig. 10.10.14:
Circuit diagrams of
the input stages of
several amplifiers.

The exact identification of VOX-amplifier or their channels in not entirely easy: first, there is
a Brilliance-Switch allowing for attenuating the bass. The AC30/6 separates this switching-
option into 2 channels (with two input jacks each): the Normal-Channel and the Brilliant
Channel (also called Bright-Channel). Consequently, the sound-characteristics available
merely as alternative switching-options in the AC15 are permanently both available in the two
parallel channels. For a good measure of confusion, the AC30/6 was issues in three different
model variants: Bass/Normal/Treble. “Normal” may therefore indicate the channel
(as opposed to “Brilliant” or “Vib/Trem), or it may designate the model (as opposed to “Bass”
or “Treble”). That “Bass” and “Treble” moreover characterize the tone filter controls of the
AC-30TB feels somehow almost normal, again.
The conclusion of the headroom analysis is somewhat ambivalent: on one hand the charts
reveal characteristic differences between the drive margins of various amplifiers, but then
again, they do not – because the diversity of the parameters is simply too large even when
setting aside the diversity of models. The unmanageable hodgepodge starts with the tubes,
continues with the settings of the controls and the definition of a reference condition, and ends
with the will (or lack thereof) to add another 100 diagrams to the 50 already cluttering the
table. While the frequency response curves show delightfully little change when swapping
one tube against another of the same type, the harmonic distortion can change drastically.
This is true not only as we plug in a well-kept Siemens ECC83 but also as we change from
one 12AX7-AC to another 12AX7-AC. The much-lauded carbon film resistors join in with a
zest: some do not even fall into the 10%-tolerance range (which in itself is quite intolerable).
It is annoying that a 100-kΩ-resistor in cosmetically fine condition found in a 50-yeal-old
VOX measures a full 300 kΩ - but it is understandable. However, the brand-new replacement
(“absolutely high-end”) had 117 kΩ rather than 100 kΩ, and this caused a few not-printable
eruptions. After a successful chill-down, and after arriving at the assumption that this might
simply be a single out-of-the-tolerance-range case, the realization followed: all 10 carbon-
composition resistors of this “High-End” batch read similarly far away from their nominal
value. It is thus recommended not to interpret the diagrams shown here to the 10th of a dB, but
use them as an “orientation”. The significant result we can retain is that the cathode-follower
creates considerable distortion. Was this the reason why the designer of the famous AC-30TB
told Jim Elyea that he in fact preferred the AC30/6 [Elyea, Section 4]?

10.10.4 Comparison: harmonic distortion
Transistors generate 3rd order distortion (= bad) while tubes generate 2nd order distortion (=
good) – that’s correct, isn’t it? Nonsense. Some things may be square (aka 2nd order), and
others hip, but tubes are … well, they are cylindrical. Tube as well as bipolar transistor as
well as field-effect transistors have progressively curved characteristics and therefore generate
both 2nd and 3rd order distortions (and many more). A big difference is that for the tiny
transistor, very soon extended circuits established themselves that included negative feedback
across several stages. Meanwhile, for tubes, the single stage with little negative feedback (or
none at all) continued to dominate. There are exceptions (e.g. power stages), but in input
stages we almost always find single tubes – mostly with a cathode resistor bridged by a
capacitor i.e. without any substantial negative feedback. The contrary happens in an
operational amplifier (OP): here there are 20 or more transistors concentrated in a tiny space
– something entirely impossible with tubes but doable with transistors in an “integrated
amplifier” on a chip of 1 mm2. The strong negative feedback in typical OP-circuits results in
symmetrical signal clipping i.e. in strong odd-order distortion (k3, k5, k7, ...). Thus, it is the
circuit that determines how an amplifier distorts, and not primarily its amplifying elements.
The transfer characteristic of a bipolar transistor from base-emitter-voltage (UBE) to collector-

current (IC) may be approximated by an exponential function:
Simplified transistor characteristic
The constant K is the value on which the blocking behavior of the transistor depends. The
collector current rises progressively with increasing base-emitter-voltage, and since this
function is not point-symmetrical to any strong degree, the dominant distortion is the 2nd order
one and not the 3rd order one (Fig. 10.10.15).
Fig. 10.10.15: Harmonic distortion for a bipolar transistor (left); transfer characteristic (right) .
The figure shows that the 2nd order harmonic distortion increases proportionally to the drive
level while the 3rd order distortion rises with the 2nd order of the drive level. At UBE ≈ 2.5 mV
the 2nd order distortion exceeds 3%; the 3rd order distortion amounts to merely 0,1%. It needs
to be considered, however, that the above equation holds for the small-signal behavior that
reaches its limit at the latest when the collector voltage approaches the residual voltage (when
the transistor conducts best). The collector current cannot increase indefinitely and as it
reaches its limit, the characteristic (initially arched to the left) turns to the right. As a
consequence of this change in the direction of the arch, the collector-current receives a
limitation in both directions and odd-order sections of the function gain in weight, and with
them the odd-order distortion products. For strong overdrive, the dominant harmonic
distortion will generally not be the 2nd order distortion but the 3rd order distortion.

A triode distorts similarly, although the functional relations are of a different kind (Fig.
10.1.12), and the following analysis will be dedicated not to the transistor but to this triode.
The basic behavior has already been presented in Chapter 10.1.4; now special guitar
amplifiers will be targeted. Fender’s Super-Reverb (AB 763) features a 7025 (ECC83) at the
input in a typical wiring – at low drive levels the 2nd order distortions dominate (Fig.
10.10.16). Fig. 10.1.13 has already demonstrated that the drive-level-dependency of the
harmonic distortion varies with the individual tube-specimen but at low signal levels (e.g. -20
dBV, equivalent to 0,1,V) the 2nd order distortion always is stronger. “Typical for tubes”, one
could think, however this holds true only for the first tube stage. The right hand picture shows
the distortion measured at the second plate, and here the 3rd order distortion dominates that –
according to some gazettes for musicians – is reserved for the transistor. Taken individually,
each triode generates predominantly 2nd order distortion at low drive levels. However, since
the signal phase is inverted from grid to plate, the 2nd order distortions compensate each
other to a large degree for two tube stages. In other words: the first triode generates a concave
downward characteristic while the second triode generates a convex upward characteristic,
and the result in a series connection is an S-shaped overall characteristic that predominantly
generates 3rd order distortion products (odd functions result in odd-order distortion).
Fig. 10.10.16: Super-Reverb, harmonic distortion: input to 1st plate (left), input to 2nd plate (right).
Of course, the details of this k2-compensation depends on the network located between first
and second tube (in this case the tone stack and the volume control); the measurement was
done at the not untypical setting of B = 2, T = 7, V = 7.
Fig. 10.10.17: VOX AC30-TB, harmonic distortion: input to 1st plate (left), input to 2nd plate (right).
An entirely different harmonic distortion situation is seen in the VOX AC30-TB (Fig.
10.10.17): although the first tube stage behaves similarly to the Super-Reverb especially at
low drive levels, the distortion rises dramatically in the second stage (cathode-follower,
Chapter 10.2.2). These are the effects of a very unusual choice of component values that leads
to a nonlinear operation with strong grid-current (control setting: V = 12:00 h, B = 10:00 h, T
= 12:00 h).

Another again different situation is found at the loudspeaker output (Fig. 10.10.18). For the
VOX, 3rd order distortion dominates for strong drive levels (“… it’s a tube amp so it has to
be k3.” ), for the Fender we find k2 and k3 to be of a similar magnitude (“... strange, are there
any transistors in the Super Reverb?” ). The details of this behavior depend on the specific
individual tubes used and, for the Fender, additionally on the quiescent current and the degree
of asymmetry in the phase-inverter. As the latter’s plate resistors are changed (100 kΩ and 82
kΩ, respectively), the k2 changes, as well. Altogether we see a rather “multivariant” scenario.
Fig. 10.10.18: Harmonic distortion, input to loudspeaker: Super-Reverb (left), AC30-TB (right).
This figure is reserved for the printed edition of this book.
What is the reason for the basic difference? The Fender uses the 6L6-GC while the EL84 is
deployed in the VOX. The offset voltage of the grid is about -10 V for the EL84 and -45 … -
50 V for the 6L6GC. In the Fender, the phase-inverter thus needs to deliver five times the
voltage and, for high drive levels, is not able to do this as well compared to the VOX.
Consequently, the operating points shift (chapter 10.4.3, 10.4.4, 10.5.12), the duty cycle
changes, and the 2nd order distortions differ. In summary: with a typical singlecoil pickup,
the Fender generates pure power-amp distortion with a dominant k3. Conversely, in the VOX
both the cathode-follower (k2) and the power amp (k3) distort. The distortion rises somewhat
more steeply in the Fender but still more gentle compare to the clipping of a transistor power
amp with strong negative feedback (Fig. 10.10.19, lower left).
Fig. 10.10.19: harmonic distortion, input to loudspeaker: Marshall JTM-45 (KT-66, Raa = 8 kΩ). Measurements
with different tubes in the impedance converter (cathode-follower).

It was already noted in Chapter 10.5.12 that the JTM-45 power amplifier has strong negative
feedback and that the drive-level-dependent rise of the 3rd order harmonic distortion (k3)
consequently has similarities to a transistor power amp. Fig. 10.10.19 shows the “overall”
measurement from preamp input to loudspeaker output. Symmetric limiting should generate
exclusively odd-order distortion but the measurement reveals even-order distortion (in
particular 2nd order components ak2), as well. These k2-distortion-products are generated by
the power amp but also in particular by the preceding tubes – and here the cathode-follower
enters the picture. Its strange operating point with an uncommonly high grid current can make
for strong distortion. “Can” – doesn’t “have to”, though. Swapping the cathode-follower-
ECC83 for another ECC83 may change the 2nd order harmonic distortion by a factor of as
much as 10 (or even more). We are not talking about damaged tubes here – no, these are
brand new. Or they may have 100 h of “burn-in” under their belt, or be switched on in
accordance with the moon-cycle, whatever. Take out one tube, put in another: 10 times the
distortion. Or 10 times less if it’s the other way round. Weird, ain’t it? One might think that
the developer was clobbered with this circuit botchery, but no, countless “expert”-journalists
around the globe rave about it. Yes, it may indeed sound damn good. It may ……
Here a little story from way back in the day: at the Siemens R&D lab there was an infamous
head of department who – as a tube circuit design was completed – took from his closet two
borderline specimen for each tube type. He plugged them in and personally took
measurements. If the great new circuit did now not perform so great anymore, the designer
received a great talking-to and was sent back to rework the circuit. Well, Marshall & Son was
not Siemens, apparently. Thank God, many will say: otherwise these distorting, screaming
monsters would never have seen the light of day. Also, it is only fair to spread some blessing
of early birth over 50-year-old developments – however why are there still no tubes in this
century that are selected for just this strange c-follower? Rather, the “experts” elaborate about
changing a transformer (RS vs. Drake), or whether yellow rather than orange capacitors
should be used, or metal-film rather than carbon resistors, or 250 µF rather than 330 µF, even
whether solid wire or stranded wire sounds better. No one ever thinks of better specifying the
nonlinearities of the c-follower-tube that may actually make a real difference, for a change.
Finally, let us look at two amplifiers that do not include the cathode-follower: Fender’s
Tweed Deluxe (cathodyne, 6V6-GT), and the Deluxe Reverb (differential amplifier, 6V6-
GT). Die Tweed power-amp has no negative feedback, and therefore the k3 is stronger at low
drive levels compared to the Deluxe Reverb (AB763) that does have feedback.
Fig. 10.10.20: harmonic distortion, input to loudspeaker: Tweed Deluxe (5E3), Deluxe Reverb (AB763).
Conclusion: clipping on both sides will generate odd-order distortion. With increasing
negative feedback the k3–rise will be steeper, but the really big differences are in the k2: there
is compensation of pre-amp-tube distortion as well as extreme dependency on individual
tubes in the c-follower. Plus, of course, the individual push-pull-anti-symmetry plays a role.

The filter circuit in VOX amps known as Cut-Circuit merits special consideration. It was
already an established custom to connect a small capacitor between the plates of the
differential amplifier used as phase-splitter (Chapter 10.4.3); this reduces the gain in the
highest frequency region. As this capacitance is increased into the nF-range, the treble is
rigorously “cut”! However, in contrast to the treble controls used otherwise, this is a non-
linear low-pass!
Fig. 10.10.21: Cut-circuit.

With the pot turned down, the
remaining capacitance may be
interpreted as series circuit with
intermediate grounding (Fig.
10.4.8).
Fig. 10.10.21 shows how the capacitance connected between the plates may be seen as series
circuit (this works the same way with an RC-two-terminal-network, if the pot is not fully
turned down). Both plate voltages are approximately equal in amount but out-of-phase so that
“between them” we find zero volts. The large plate loading dramatically reduces the slew-
rate, and therefore this low-pass has a non-linear effect. Another consequence is that the
treble-loss cannot be compensated for in any further intermediate stage: the power amp
generates less treble even when overdriven (!). ‘Turning down Cut’ therefore is different from
‘turning down Treble’.
Fig. 10.10.22: Effect of the Cut-circuit (at 1 kHz).

In Fig. 10.10.22 we see the results of measurements taken from an AC30-TB (from Normal-
input to power amp). Even with strong overdrive, the power amp cannot do any “hard
clipping”: the shape of the curve is round and the high frequencies are attenuated. Conversely,
if the Treble knob were turned down on e.g. a Fender amp, and the power amp strongly
overdriven at the same time, the result would be a square output wave-shape. Here, the VOX
offers an interesting alternative.

10.10.5 At which strength is harmonic distortion audible?
This is a difficult topic because there are so many details influencing it that a single number is
not even close to doing the job. We can merely state: ”somewhere between 0,03% and 10%.”
For synthetic test signals it will be more towards the lower value while for guitar sounds, it
will be more towards the upper.
Nonlinear distortion of a sine-tone can be detected only at strong distortion levels because the
new (higher-frequency) partials generated by the nonlinearity are masked to a large extent
[12]. Two-tone signals are more critical since their nonlinear transmission generates (on top
of the masked summation tones) low-frequency difference tones, as well – and these can
relatively easily be detected. Webers writes in his book "Tonstudiotechnik" (recording studio
technology) that tones of flutes are seen as particularly problematic. He notes a threshold of
detection of k2 = 1% for 2nd order distortion and of k3 = 0.3% for 3rd order distortion. Rossi
lists even smaller limits of audible distortion but feels that 1% intermodulation-distortion is
acceptable. Our guitar amps? No, they do not fit at all into this system of (mostly purposeful)
rigid values of audibility thresholds found in recoding studio technology. Still, it would be
helpful to have an understanding of the distortion levels at which clean becomes crunch, of
the characterization of strong and ultra distortion, respectively, and of what even and odd
distortions are.
Nonlinear distortions happen at curved transmission characteristics, i.e. predominantly in

tubes and semi-conductors. Curved characteristics may be developed into mathematical series
expansions, and if these expansions include odd powers only (x, x3, x5, ...), they generate odd
distortions. If only even powers occur (x2, x4, ...) on top of the linear term (x), even distortions
result [Taylor/MacLaurin, Fourier-series, communication technology]. To start with a simple
signal (even if it barely shows any similarities to a guitar tone): in Fig. 10.10.23 a sine-tone
receives nonlinear distortion via the characteristic y = x – x3.
Fig. 10.10.23: Nonlinear (3rd order) distortion of a sine-tone; time function, transfer characteristic, spectrum.
Inserting for x = sin(ω t) into the characteristic and calculating the equation immediately
shows the result as seen in the spectrum: we obtain a new spectral line at three times the
fundamental frequency with a level-distance of 22 dB re the level of the fundamental. In the
following formula, the index i stands for the order of the partial tones (i = 1 marks the
fundamental), u is the distorted voltage, and ui is the voltage of the i-th partial (all voltages are
RMS-values). Consequently, k3 is the 3rd order “Harmonic-Distortion”-factor (HD), and ak3 is
the difference level between the fundamental and the distortion products. This approximation
works the better, the smaller the HD is.
Harmonic Distortion

Fig. 10.10.24 shows the corresponding results for purely 2nd order distortion; the chosen
characteristic was y = x – 0.3x2. The new partials generated are now at twice the fundamental
frequency and at 0 Hz. The DC-component is usually blocked using coupling capacitors
because it may disadvantageously shift the operating point depending on the signal.
Fig. 10.10.24: Nonlinear (2nd order) distortion of a sine-tone; time function, transfer characteristic, spectrum.
Using two-tone signals we achieve a step towards more natural signals, but we also increase
the number of degrees of freedom: we may now choose the frequency relation between the
two primary tomes, the difference in their level and the difference in their phase. For Fig.
10.10.25, a frequency relationship of 6/5 is chosen, with the levels of the primaries being
equal. For 3rd order distortion, new lines are generated at the frequencies 2f1 – f2, 2f2 – f1,
3f1, 2f1 + f2, 2f2 + f1, 3f2. At 2f1 – f2 we find the 3rd order difference tone.
Fig. 10.10.25: Nonlinear (3rd order) distortion of a two-tone signal; time function, transfer function, spectrum.
With 2nd order distortion (Fig. 10.10.26), a DC-component results, as well as new lines at
f2 – f1, 2f1, f1 + f2, 2f2. At f2 – f1 we find the 2nd order difference tone.
Fig. 10.10.26: Nonlinear (2nd order) distortion of a tow-tone signal; time function, transfer function, spectrum.
The distortion does not only generate lines at new frequencies but also at the frequency of the
primary tones. The level and phase of the latter is correspondingly changed.

A tone from a guitar is much more complex than the signals just looked at, and therefore the
multitude of parameters explodes. The HD is not a fixed value but dependent on the drive
level. Doubling the input signal makes the 2nd order HD grow by a factor two and the 3rd
order HD by a factor of four; k2 is proportional to the drive level while k3 is proportional to
the square of the drive level. Changing the phases of the partials changes the crest-factor
(peak-value/RMS-value) and thus the HD even if the drive level remains constant. For a
guitar signal, this drive level is of course not constant but decreases quickly after a strong
attack. So, what should we reference the HD to? To the maximum value that lasts only a few
milliseconds? Or some kind of average value defined one way or another? For sine-shape
drive signals it is easy to specify the HD but driving a system with a guitar signal creates a
problem.
Fig. 10.10.27: Changes in the 3rd order distortion spectrum as the phase of the partials in a three-tone signal is
changed. The RMS-value, and thus the level of the primary signal is identical for both cases.
For Fig. 10.10.27, a signal consisting of 3 partial tones is distorted. Changes of the phases of
the partials do change the level of the strongest distortion product by no less than 6 dB. This
does not mean that measuring of (T)HD (or intermodulation- or difference-tone-distortion) is
not purposeful – in fact these measurements are highly suitable to describe the nonlinear
behavior of a system. An approximate estimation of how strongly a specific signal is distorted
by this system is possible, but does not really indicate how the resulting distortion in fact
sounds.
After this introduction we will now look at real guitar signals, using the pickup voltage of a
Telecaster. As a string is plucked with little force, the levels of the partials decay
approximately linearly over time, as it has been shown in Chapter 7.7. For strong plucking
(with the string hitting the frets - Chapter 7.12.2) we find a strong level-decay of up to 10 dB
during the first 20 – 50 ms, and a slow decay afterwards, similar to weak plucking. In a
simple model generating merely 3rd order distortion, the HD would change by a factor of 10
during the first 50 ms. For such time-variant signal a single HD-limit is not very purposeful.

In Fig. 10.10.28 we see the time functions of two pickup voltages. A non-linear amplitude
limiting to e.g. ±0.5 V would have very different implications for the two signals. This
example clarifies that for HD-limit-values not only amplitude limits are significant, but
durations in time, as well.
Fig. 10.10.28: Two different pickup voltages normalized to the same maximum value. In the upper section the
string was weakly plucked, and strongly on the lower section. Telecaster, bridge pickup, E3 on D-string.
Before we subject these pickup voltages to distortion, we

first return to the series expansion of the characteristic curve.
For small HD it is purposeful to study the behavior of purely
2nd order and purely 3rd order distortion. In guitar
amplifiers, however, strong distortion occurs, as well, and
therefore the model using purely 2nd order and purely 3rd
order distortion is incomplete. Tubes (as well as semi-
conductors) limit on both sides for strong drive levels – this
is the domain of odd distortions. A straight, symmetric
characteristic (such as y = x2) cannot generate limiting to
both sides. A 3rd order characteristic can do this – however
only within a small range, as shown in Fig. 10.10.29. The
blue line approximates the characteristic of a tube close to
the origin, but it turns off in the opposite direction as it
moves away. And it continues to grow without any limiting.
Fig. 10.10.29: Tube characteristic (ECC83); third-order parabola.

To better adhere to the tube characteristic depicted in Fig. 10.20.29, the approximation-
polynomial would require further odd-order members in the series (x5, x7, ...), and in addition
series-members of even order would be necessary, because the amounts of the limit-values
differ (tube-characteristics are not exactly point-symmetrical). Therefore, the distortion in the
following is done not by a polynomial characteristic but by a real tube characteristic (ECC83).
Fig.10.10.30: Pickup voltage, without (top) and with non-linear tube-distortion (bottom). String strongly
plucked, Telecaster, E3 on D-string, bridge pickup.
In Fig. 10.10.30 we again see the signal from Fig. 10.10.28, with and without tube-distortion.
It may be hard to believe, but these two guitar signals do not in fact sound that different. One
does hear differences but not in terms of “undistorted/distorted”. The attack is louder for the
undistorted signal, but afterwards there is no audible difference. This may be due to post-
masking [12], and/or due to the fact that any limiting in the subsequent development affects
merely very short signal peaks. Another reason: for a strongly plucked string, contacts
between string and frets occur for a relatively long time period, and these sound similar to
slight overdrive and hamper the recognition of actual tube distortion. A value for the HD in
the signal shown in Fig. 10.10.30 cannot be established since there is no definition of a HD
for such a multi-tone-signal. It is however possible to create a sine-tone with the same
envelope, and to distort it in the same way (i.e. feed it through the same tube characteristic).
The result is that at first 3rd order distortion dominates with k3 reaching 28%. From 50 ms the
2nd order distortion starts to dominate with k2 ≈ 5%. It is noted again, however, that despite
these large HD’s the guitar does not actually sound distorted but is limited in its dynamic
range. The “thud” at the beginning is softer – and that’s it.
We obtain an entirely different result as the string is merely lightly plucked. Without
distortion, it sounds weaker in the treble range than the strongly plucked string. Therefore,
and also because the level does not decrease as fast, distortion can be heard clearly as the
signal is fed to the same characteristic as the strongly plucked string (with both signals
normalized – pre-distortion – to the same maximum drive.

For the lightly plucked string, Fig. 10.10.31 shows the time function of the undistorted and
distorted pickup voltage. Despite the same maximum drive level and the same distortion
characteristic, the subjective degree of distortion is different. This is because the hearing
system does not exclusively evaluate the attack, after all. It is well known from experiments
on loudness scaling [12] that the loudness of short bursts decreases: the hearing integrates
over 100 ms.
Fig.10.10.31: Pickup voltage, without (top) and with non-linear tube-distortion (bottom). String lightly plucked,
Telecaster, E3 on D-string, bridge pickup.
It has already been elaborated repeatedly that the inharmonicity of the guitar signal plays a
role, as well (Chapter 1.3, 8.2.5, 10.8.5). For a strictly harmonic sound, all spectral lines
generated by the distortion fall onto already existing lines, and it is merely level and phase of
the frequency component that changes. However, for an inharmonic sound the non-linearity
will cause new spectral lines at frequencies where no partial was present in the undistorted
signal. The subjectively perceived sound may change considerably due to this, depending on
the circumstances. It will obtain a more stochastic character and sound as if noise had been
added (Fig. 10.8.23). Because the inharmonicity depends on the type of string, on
characteristic of the circuitry, and on the individual tubes, and on the guitarist, it is not
possible to give a single threshold value for the audible HD. It may be noted as an orientation,
though, that we are not talking about values in the range of or even below 0,1% here. There
are investigations comparing capacitors with a THD of below 0,0001%. This is extremely
sophisticated metrology but entirely without meaning for auditory acoustics.
Well then – despite all constraints, the reader will expect a number here, and now. And so, to
the best of our knowledge: k = 3%. This would be the orientation value – and surely a basis
for splendid discussions. Guitar-distortion becomes just audible as a sine-tone of the same
level distorts with a THD of 3%. “The same level” should be interpreted such that not the
level at the attack of the guitar tone is measured, but the level of a purposeful section of tone
following the attack region. This puts the responsibility back to the esteemed reader and
hopefully helps to avoid a discussion in internet fora (e.g. dedicated to the question of
whether the threshold of audibility is not at 2,6%, after all).

10.10.6 Comparison: frequency responses
Let’s now go for the full enchilada: the mapping of the input voltage onto the sound pressure.
We will not check the transfer function of individual parts to the circuit (as in Chapter 10.3)
anymore, but the transfer behavior of the whole “amp-plus-speaker”-system. For the
associated measurements, the speaker enclosure (for combos including the amplifier) was set
up in an anechoic chamber (AEC), i.e. a room with fibrous wedges of 80 cm length mounted
to all six boundary surfaces to substantially suppress any reflections. The sound pressure was
picked up axially in front of the speaker using a precision microphone (B&K 4190), and
analyzed with a workstation (Cortex CF 100). Beaming effects were not captured here –
Chapter 11.4 is dedicated to the associated effects.
The ancestors of modern guitar amps did not differ much from other audio amps of the time.
The design objective was apparently a reproduction as broadband and as frequency-
independent as possible. Simple amps did not have any tone filter at all (e.g. early Fender
Champs), or they sported – what luxury – a single tone knob (Fender Deluxe). Later, two-,
three and even four-band tone filters were included, as well as tremolo and reverb – but,
again, that came later. These old amps did not sound bad because the transmission was in fact
not frequency-independent, after all – due to the frequency dependency of the loudspeaker
impedance (Chapter 11.2), whether the designers were aware of this or not. Early power
amps did not have any negative feedback (e.g. Champ 5C1, Princeton 5D2, Deluxe 5B3,
Super 5B4, Pro Amp 5C5, VOX AC15, Gibson GA-20, Gibson GA-40, Rickenbacker M11,
Epiphone EA-50, and many more), giving the pentode-power-stages a high-impedance output
that leads, in combination with the speaker impedance, to a characteristic frequency-response.
Fig. 10.10.32 shows this exemplarily for the AC30 – this amp is not that old but it never had
any negative power-amp-feedback in any of its incarnations. With a 16-Ω-resistor serving as
load, the transmission is independent of frequency. However, as the speaker replaces the
resistor, the speaker resonance appears at 65 Hz, an enclosure resonance shows up at 170 Hz,
and towards the high frequencies we see the contribution of the voice coil. This situation is
quite different for power amps including strong negative feedback such as the JTM-45:
unless the presence control is turned up, the voltage levels for resistor- and speaker-load do
not differ by more than ±1 dB. For the overall frequency response, three main sources can be
identified: the tone filter (as far as it is present), the speaker-impedance, and the frequency
response of the speaker. In addition, there are high-pass filters (the coupling capacitors) and
low-pass filters (the Miller-capacitance), as already described earlier. The overall frequency
response depicted in Fig. 10.10.32 shows a pronounced treble boost although there is no
special filter for this – it is the result of the high-impedance power-amp-output + speaker
impedance + frequency-response of the speaker. In the AC30, the treble could be attenuated
with the Cut-filter but that is not in fact desirable. For many users, the Normal-channel
featured too little treble – that is why the Treble-version and the TB-channel were developed.
Fig. 10.10.32 left: VOX AC30-TB, transmission from phase-inverter-input to loudspeaker-output; dashed: with
resistor; solid: with speaker; right: transmission from Normal-input to SPL in the AEC, volume = 12:00 h.

The frequency responses of some power amps (with corresponding loudspeaker) are shown in
Fig. 10.10.33. For these measurements, the sweep-generator was directly connected to the
input of the phase-inverter, and the microphone (B&K 4190) was at 2 m distance from the
speaker. The lower curve shows the level of the speaker-voltage, referenced to 500 Hz (for the
frequency dependency of the speaker impedance see Chapter 11.2). For power amps with
strong negative feedback (e.g. the JTM-45) there is only little mapping of the speaker
impedance maps onto the voltage level, while for power amps with weak or no negative
feedback (Super Reverb, AC30), the speaker impedance strongly influences the voltage level.
Moreover, the speaker itself and the enclosure construction (open or closed) of course
influence the transmission behavior (Chapter 11).
Fig. 10.10.33: SPL (d = 2m) and voltage level (lower curve). Reference: 1 W at 500 Hz.
SPL measured in the AEC on axis, sine sweep impressed onto the phase-inverter.
(N.B.: the parts of the figure not shown are reserved for the printed edition of this book.)

Taking the measurement not starting with the PI-input (as in Fig. 10.10.33) but starting with
the input jack, the tone filter and other parts of the circuitry determine the transfer
characteristic, as well. For the following measurements (Fig. 10.10.34), the tone filters were
adjusted such that all amps had a similar, treble-heavy transmission; due to the limitations of
some filters this was only possible as a rough approximation for several cases.
Fig. 10.10.34: SPL (d = 2m) and voltage level (lower curve).

SPL measured in the AEC on axis, sine sweep impressed onto the amplifier input.
(N.B.: the parts of the figure not shown are reserved for the printed edition of this book.)
As expected we find differences between the individual measurements. However, a

comparison to the headroom-charts (Chapter 10.10.3) shows that the differences in the non-
linear behavior are at least as big. As soon as an amplifier reaches substantial distortion, it is
not sufficient anymore to merely determine the frequency-response (which, as stipulated by
theory, is then anyway not defined anymore, either).

10.10.7 Special amplifiers: VOX, Fender, Marshall
VOX amplifiers AC15, AC30
The character of VOX-amplifiers is most readily understood starting the analysis with the
output stage. This part of the amp shows extensive similarities for the AC30/4, the AC30/6
and the AC30-TB, and the AC15 from 1960 is based on this circuit, as well, albeit with only
two power tubes. Fig. 10.10.32 depicts the gain measured from the phase-inverter input to the
16-Ω-loudspeaker output: once with a resistive load (16 Ω), and a second time with the VOX-
loudspeakers (Celestion Blue). Due to the high source-impedance, the loudspeaker impedance
maps onto the output voltage, and local maxima appear in the overall transmission
characteristic: around 70 Hz (speaker-resonance), at 180 Hz (Helmholtz-resonance of the
enclosure) and in a broad band towards the high frequencies (speaker inductance, details in
Chapter 11). With the resistive load, the power amp shows very little frequency dependence –
there is merely a tiny bass-boost resulting from the Cut-filter. The characteristic frequency-
response in the SPL is therefore generated not by the circuit per se but by the interaction
between power amp (sans negative feedback), the speaker impedance and the radiation
characteristic of the speaker. For all frequency responses shown here it is important to
consider that all resistors in the circuits may have tolerances⊗ of up to 10%, and the capacitors
occasionally up to 20%. Since the total frequency response is due to the interaction of many
components, substantial “overall”-deviations are possible.
Fig. 10.10.35: VOX AC15, Normal-channel; "second circuit" (1959, top), "third circuit" (1960, bottom).
The built-in vibrato channel is not shown in the figure, supplements: Chapter. 10.8.2. Pictures: Elyea.
In Fig. 10.10.35 we see two variants of the AC-15-circuitry. Rather outlandish in the ’59
circuit: the 100-Ω-series-resistors in the plate-connections of the power tubes. Did somebody
confuse plate and screen grid here? This was not an error in the drawing – this did go into
production, as photos in Elyea’s book show. In the 1960-successor, the resistors show up
where they belong: in the screen-grid-connections. The 60’s-circuit was issued as Normal-
and as Bass-model, with corresponding coupling caps. Opening the Brilliance-switch
attenuated the low frequencies, and the Cut-control decreased the treble (Fig. 10.10.36).
⊗
Two 100-kΩ-resistors bought from a tube-distributor each had 117 kΩ although specified with 10% – probably
a concession to the black carbon-soul that supposedly ensures „absolute high-end in the signal path“.

The treble could not really shine in the ’59 because the input circuit was inappropriate. In the
series branch it had a high impedance (noise!), and in the parallel branch it was a bit too low-
impedance due to the second 220-kΩ-resistor connected to ground. The modulator (not shown
here) necessitated the ECF82 (a combination of triode and pentode). This RF-tube (oscillator,
mixing stage) was a bit out of place in the given environment and its gain is rather moderate.
There was, however, not much choice if a triode was required for the Normal-channel and
only a single pentode was foreseen for the whole of the modulator. Only in 1960 does the
AC15 receive the deluxe-modulator (Chapter 10.8.2) and, in the Normal channel, the high-
gain EF86. The latter had to yield to the ECC83 in the same year for the AC30/6.
Fig. 10.10.36: Coupling preamp/phase-inverter (Brilliance switch, left); Cut-Filter ("Normal", right)
As we have established, the special transmission characteristic of the AC15, AC30/4 and
AC30/6 amps results from the frequency response of the power-amp/loudspeaker interaction
– tone filters are present to only a very modest degree in these amps. With increasing drive,
the compression of the power stage comes into play (Chapter 10.5.12), plus the dominance of
the 3rd order distortion (Chapter 10.10.4). Whether an EF86 or an ECC83 is placed in the
preamp should only very indirectly affect the frequency response. Both tubes work from 0 Hz
up into the MHz-range. Still, Elyea notes: "The 12AX7 had a narrower frequency range, with
a bit more treble, but less bass response than the EF86. The EF86 gave a wider frequency
range". “More treble”, but less bandwidth? Well, of course that depends how you define
treble … but in any case: if there is any effect at all, then this is not due to the tubes
themselves but the result of the circuits around the tubes. By the way: regarding the
comparison AC30/4 vs. AC30/6, Petersen/Denny opine: "The AC30/4 seemed to have a
clearer tone". And they add "An EF86 has five elements as opposed to the three of a triode,
so it can have up to 25% more gain". One is tempted to comment: but 5/3 is 67%! Of course
the number of the electrodes is correct – it is the word “so” that rubs the wrong way because it
implies the gain depends on the number of electrodes. Both percent-quotations are nonsense;
the increase in amplification (EF86 vs. ECC83) is more than 100% (vU = 140 to 180 vs. 70).
The “direct” influence of the input tube relates to the input capacitance, the amplification and
the channel linkage. The pentode features a smaller grid-to-plate-capacitance resulting in a
measureable difference to the ECC83 (Miller-effect). That is no reason to go into drama-
mode, however, because a similar influence would result from shortening (or increasing) the
length of the cable between guitar and amp by ½ a meter or so. On the other hand, the
difference in gain is considerable: : +43...45dB for the EF86 (tube-specimen dependent)
compared to +37dB for the ECC83 (each in VOX-typical environment). A further 6-dB-loss
is due to the channel addition, and consequently an AC30/4 will yield the four- to five-fold
amplification compared to an AC30/6. Furthermore, in the AC30/6, the frequency response of
the Normal channel depends on the position of the volume control of the Bright channel, plus
the coupling capacitors are different. The same for the loudspeaker, by the way: the change
from Goodmans to Celestions in 1960 happens in the same year when the AC30/4 and the

AC30/6 were both concurrently on the market. There are, in summary, many reasons why one
may hear differences in the sound of the amps. Not to forget: the microphonics of the EF86
which was the main reason for its retirement, and for a change in the circuit: many EF86 were
configured as triode via the circuitry (to reduce microphonics; also lowered were the gain and
the frequency response [Elyea]). In some AC15, the EF86 was swapped with a ECC83 in the
factory. While the AC15 was not subject to radical redesigns that other amps had to undergo
(e.g. the Bassman), there still were changes. Of the 17 versions listed by Elyea many differ
only in cosmetics or minor details; there are, however, three documented circuit variants. The
obscure EL34-AC30 existed in 5 versions, and the bestseller AC30-Twin in 15 Versions
during the JMI-period (1960 – 1967): there was the AC30/4 in the Normal- and Bass-variants,
the AC30/6 in Normal, Bass-, and Treble-versions, pre and post the so-called 'List of
changes', with 80-Ω- or 50-Ω-cathode-resistor, or with included Top-Boost-circuit. After that
we see semiconductor diodes arriving replacing the GZ34, ceramic-magnet-speakers, and
even pure transistor amps … but that was after the golden era that the JMI-period is sees as
today. Fig. 10.10.47 documents the change from the AC30/6 to the AC30-TB: originally
installed as a retrofit, it was included ex-factory from 1963/64. Thus, the most important
representatives of the VOX-flagship were the AC30/4, the AC30/6 and the AC30-TB, each as
“Twin” since fitted with two loudspeakers, and occasionally as “Super-Twin” if the amp and
speaker resided in separate enclosures. The AC30/4 sported 2x2 inputs; the AC30/6 and the
AC30-TB had 3x2 inputs. The AC30/4-circuit largely corresponds to the “third” circuit of the
AC15 shown in Fig. 10.10.35 but boasted, on top of four instead of only two output tubes,
other transformers and two speakers♣ instead of just one. In the AC30/6, the EF86 is replaced
by an ECC83 – resulting in an additional channel with two inputs (connected in parallel). The
two Normal- and Brilliant-channels differ in the coupling-capacitors in the input-stage: 47 nF
vs. 500 pF, i.e. a bass-attenuation in the Brilliance-channel (Fig. 10.10.37). The AC30/6
emerges into the AC30-TB by the addition of the Brilliance-Unit. The cathode resistor of the
latter was first bridged with a capacitor – this was later omitted.
Fig. 10.10.37: VOX AC30/6. Of the in total 4 power tubes (2 each in parallel) only 2 are shown. The circuit at
the right was inserted into the Brilliant-channel behind the volume pot at the marked position; in addition, this
pot had a bright-C (100 pF) shunting it. Result: the AC30-TB.
First, the AC15 and AC30 were available as a “Normal” model and also as “Bass” model. The
“Bass” model included enlarged coupling-C’s: in the AC30/6, for example, 100 nF instead of
47 nF, and 1000 pF instead of 500 pF, respectively, were used. Moreover, the Cut-capacitor
was doubled in value. The “Treble” model experienced further changes, as exemplified in Fig.
10.10.38. On top of the separation of the cathodes and Bright-C’s, the coupling-C’s feeding
the output tubes were decreased to 47 nF, and the Cut-C reduced to 2,2 nF. The separation of
the cathode circuits in the input tube, however, shifts the operating point of this tube! While in
the “Normal” version the currents of both triodes run through the 1.5-kΩ-resistor, only one of
these currents remains in the “Treble” version. To maintain the operating point, a 3-kΩ-
resistor should have been included into the cathode connection. It was not done …
♣
The AC15 was also available as Twin, fitted with two (low-cost) Goodman loudspeakers [Elyea].

The high-frequency boost in the “Treble”-model is predominantly caused by the 220pF-

capacitor while the smaller cathode capacitor generates a mere 3-dB-treble-increase. The 330-
kΩ-resistor ensures that the Bright-C does not become entirely ineffective as the volume
control is turned up fully, but the possible maximum gain is reduced by 7 dB.
Fig. 10.10.38: VOX AC30/6 "Treble". Frequency response from input to the first plate (-----), and to the phase-
inverter-input respectively; Brilliant-channel (middle), Normal-channel (right).
Fig. 10.10.39 shows the frequency response from the input all the way to the power-amp
output (loaded with speaker), and to the resulting SPL in the anechoic chamber – a simple
“sound scale”, perfectly balanced. The more elaborate filtering in the AC30-TB was already
introduced in Chapter 10.3.1.
Fig. 10.10.39: VOX AC30/4. Frequency response up to the power output (left), and including the resulting SPL
(right).
A specialty of the early days that is rarely used today is the Vibrato-channel. Already the
second AC-15 version included it in its deluxe-incarnation, as did all AC30. The function is
discussed in Chapter 10.8.2. Brilliant- and Normal-channel require one single triode each, but
the Vibrato-channel needs no fewer than six. Six sells here, as well – it was a powerful sales
argument. The only problem was that the low-frequency modulation signal could not be fully
suppressed – despite the carefully designed bridge circuit. This is why already the Gibson
GA70 included a multi-stage high-pass, that VOX “borrowed”. The frequency response of the
high-pass is shown in Fig. 10.10.40.
Fig. 10.10.40: High-pass in the Vibrato-channel (VOX AC15, AC30/4 five-stage, VOX AC30/6 four-stage).

The attenuation of the bass that this high-pass caused as well in the guitar-signal had to be
accepted; this channel corresponded approximately to the Brilliance-Switch. Why, however,
this 500-Hz-high-pass is followed by yet another high-pass (at the PI-input) with a cutoff-
frequency of 8 Hz (in the Bass-AC30/4 even as low as 0,8 Hz) probably only Dick Denny
would have been able to explain that. Or not. Doesn’t matter – these are the myths from the
past, emanating from the billowing mists of the dionysiac 1960’s, and having found a new
home in the thicket of the WWW, the world-wide-wilderness.
Around 1967 the golden times of the original VOXes comes to an end. Turnovers come
crashing down, and chucking company founder Tom Jennings does not help. In March 1969
it’s almost curtains – VOX is "in preparation for its liquidation [Elyea]". From then on, one
owner follows the next: 'Corinthian Securities', 'Birch Stolec', 'Dallas Arbiter', 'Rose Morris';
they all buy and sell the remains … and at the very end Korg takes over. And they do revive
the production of the AC30 (from 1993) – at Marshall, of all places, thanks to good relations.
The re-launch is successful and VOX is back (Chapter 10.10.7), drumming up business via
advertizing the glory-days back then.
In view of all the different variants of the AC30 it is clear that “the”AC30-sound does not
exist. Just as there is not “the” Fender-sound – although the EL84-power-stage missing any
negative feedback, and the speaker/enclosure-construction do create commonalities. Too
simple: the equation Beatles-Sound = VOX-Sound even if advertising does go down that path.
But then, just as valid would be the derivatives of the equation: Beatles-Sound = Stones-
Sound, or – in the extreme – Shadows-Sound = Queen-Sound. No, that doesn’t work. Jim
Elyea dedicates 20 pages to the question: when did the Shadows receive which amplifier, and
what was recorded when using what? And it becomes even more extensive (and confusing)
for the Beatles. That was not “the” VOX – the next larger amp was grabbed and used as soon
as it hit the market. Verifiably, Lennon did play an AC15 … but he also played through
AC30’s, AC50’s and AC100’s. And even though the 7120 and the Conqueror, although the
latter were – dare we write it – hybrid- or transistor amps. Even THAT is the VOX-sound,
however.
Fig. 10.10.41: Various AC30 [Jim Elyea: VOX Amplifiers, The JMI Years].

Fender-Amplifiers
"In the 20's, Leo Fender was a bookkeeper who got into ham radio as a hobby". That’s how
Dave Funk, in his TUBE AMP WORKBOOK, starts the description of an extremely influential
bookkeeper whose amplifiers and instruments were to write history. After a short
collaboration with Doc Kaufman, Leo Fender started his own Fender Electric Instrument
Company in 1946, located in Fullerton, California. First, he built amplifiers based on circuits
from the "Radiotron Designer's Handbook", and from 1950 also electric guitars and basses. A
plethora of different amplifiers originated on his workbench – Dave Funk requires no less
than 250 pages for the circuit diagrams alone, and doesn’t even go beyond the 1970’s.
Skipping the uncalled-for question “were there actually any Fender amps worth considering
after 1964?” throwing in a concise “yes”, we will try to bring some order to the diverse range.
Fender amps of the early period used a number system the first character of which denotes
the decade: 5 for the 50’s, 6 for the 60’s. The second character is a letter indicating the change
variant, and the third position specifies the model. A 5B3 is a Deluxe from 1952; its
successor is the 5C3. The Bassman of 1952 is the 5B6, the Twin of that year is the 5B8. It is
assumed that the letter was supposed to code the year, but this system broke down in 1955
because it was not possible to revise every amp every year. For some it is of the utmost
importance to be able to date the production to the respective month – we shall not go into
that here, but approximately: A = ’51, B = ’52, C = ’53, D = ’54, and from E = ’55 we loose
coherence, until the G-variants arrive around 1960. From 1963, a simplification spanning
across the models arrives with the AA763-circuit. It receives a revision in the AB763.
Model Name Start of production; typical power tubes Power class
1 Champ 1946/47, 1x6V6-GT *

2 Princeton 1946/47, 1x6V6-GT *
3 Deluxe 1947, 2x6V6-GT (Model 26) **
4 Super 1950, 2x6L6-GC (Dual Professional) ***
5 Pro 1950, 2x6L6-GC ***
6 Bassman 1951, 2x6L6-GC ***
7 Bandmaster 1952, 2x6L6-GC ***
8 Twin 1952, 2x6L6-GC ⇒ 4x6L6-GC *****
9 Tremolux 1955, 2x6V6-GT ⇒ 2x6L6-GC **
10 Harvard 1956, 2x6V6-GT ⇒ 2x6L6-GC **
11 Vibrolux 1955, 2x6V6-GT ⇒ 2x6L6-GC **
12 Concert 1960, 2x6L6-GC ***
13 Vibrasonic 1959, 2x6L6-GC ***
14 Showman 1961, 4x6L6-GC *****
15 Reverb Unit 1961, spring reverb, no power amp -
16 Vibroverb 1963, 2x6L6-GC ***
Table: Fender amplifiers; N.B.: the available sources are incomplete and to some degree contradictory.
In terms of cosmetics, distinguished are: the very early K&F amps (1945-46), the 'Woodies'
with their wooden look (from 1947), and subsequently the 'Twotone-Vinyl-Amps'. After that
we get to the famous 'Tweed'-Fenders (from 1948), named after their lacquered cloth-
covering. Then there’s light and dark brown for the 'Brownface' amps (1959 – 63), various
white tones 'Blonde', 'Cream' (1960 – 64), 'Blackface' (1964 – 67), and finally 'Silverface'
(1967 – 81). That’s with some leeway in the dating – the source situation is kinda dubious.

For today’s used-goods-commerce, establishing the production date to the day may be of
importance. From the technical point of view, however, the circuits, components, enclosures
and loudspeakers are more important. While there are some guidelines there, we also
encounter many exceptions. It is understandable that not all amps could receive new tone
filters at the same time, and that is was important to use up existing stock first before the new
model was allowed to leave the factory. The amp versions are so extremely manifold that it is
impossible to list them all even only approximately: a capacitor is deleted but rematerializes
two years later again, capacitance values change without recognizable rules, negative
feedback is incorporated but discarded again shortly afterwards, various tremolo-concepts are
tested, and much more. No criticism here: this is how products evolve – but it makes
documentation difficult. The old octal tubes give way to new noval tubes, a mercury-rectifier
steps up – and steps down again right away, the phase-inverter stage mutates from the
paraphase circuit (1946, from 1951 with negative feedback) to the cathodyne-circuit (about
1955) and on to the differential amplifier (about 1956, Chapter 10.4). The output power grows
(e.g. for the Twin from 18 W to 185 W), and the speakers of course need to keep up: from the
weak Alnico to the high-resilience ceramics. However, not everything intended as an
improvement is seen as such by the guitar players, and consequently old concepts are reheated
as “Reissues”, and “Historic-” or “Vintage-Models” are revived.
On our search, we do find commonalities♣ in all Fender amps but then again hit exceptions
right away. Indeed, Leo Fender liked Country music, so the assumption is probably correct
that his amps were to do well in that music scene. And yes indeed, distortion presumably was
a fault to his ears. Brilliant treble was desired and easily achieved in combination with the
typical Fender single-coils. However, to attribute to all Fender amplifiers a common sound
character – no, that would push it too far. Not just between models but also within a single
development line (e.g. from the 5B3-Deluxe to the AB868-Deluxe) there are large sonic
differences everywhere. And therefore there isn’t even “the” characteristic Deluxe-sound.
It is not necessary to include the very old Deluxe from 1947 for comparisons because it exists
today only in homeopathic doses. But it does get interesting from 1954: as the 5D3, the amp
receives the modern noval-tubes (12AY7, 12AX7, 2x6V6GT, 5Y3GT), a stable input circuit,
and the paraphase circuit including negative feedback. Apparently, it works so well that the
power-amp can dispense with any negative feedback. The biggest change in the 5E3 is the
introduction of the cathodyne circuit, accompanied by small capacitance changes and other
modifications. It is controversial whether there ever was a 5F3 – a schematic has not turned
up. The 1960 6G3 has an additional 12AX7 for the Vibrato-effect, and includes the change
from the cathodyne PI to the differential amplifier. Moreover, the cathodes of the power tubes
are now connected to ground (fixed bias) and the vibrato signal is superimposed onto the
negative grid voltage. That ain’t optimal ‘cause this power amp does feature negative
feedback. In the power supply, the 5Y3GZ has to yield to a GZ34, and in the pre-amp, both
channels now include separate tone-filters. In the AA763 from 1963 the LDR-modulator is
deployed for the first time, each channel receives its own Treble/Bass-filter – and from the old
10-Watt-amp (5 tube stages, 3 knobs), a 21-Watt-amp (11 tube stage, 8 knobs) has now
sprung. Is this the end of the line? No way: the Deluxe-Reverb trumps this and offers (as the
name suggests) in addition a spring reverb: 15 tube stages, 9 knobs. It is understood that all
these modifications will have an impact on the linear and (in particular) on the non-linear
behavior, and thus onto the sound.
♣
no, just the shared Fender-logo is not sufficient ...

Fig. 10.10.42 shows the topological variations between two Deluxe amps. Already the input
tubes differ (Chapter 10.11), as does the plate circuitry. In the 4E3, the volume pot is
“reverse”-connected – a feature found in many very early amplifiers. Changes in the control
setting directly change the amplification of the tube, and with the volume turned down fully,
the tube operates into a short. It being a current source, this does not do any harm to the tube.
The simple tone pot has backwards-effects on the plate, as well, and on top of this, both
channels are coupled. This scenario is easy to analyze but very difficult to describe because
everything depends on everything else. The AB763, on the other hand, sums the two channels
only directly ahead of the phase-inverter (PI) and makes a much better decoupling possible
(Fig. 10.10.43). The effect of the tone control is depicted in Fig. 10.10.44: it is a wide-band
treble-filter dependent on the setting of the volume control, and the mid-range attenuation
(Chapter 10.3) so characteristic of the later versions is absent. The reverse-connected volume
pot is impractical because in its middle turn-range the amplification changes little (by merely
10 dB between settings 2 and 8). Plus the two volume pots influence each other in their effect.
Fig. 10.10.42: Comparison 5E3-Deluxe (Tweed) vs. Deluxe-Reverb (Blackface). The respective given gain
values relate to the reference condition from Chapter 10.10.2 (90 mV / 500 Hz for full drive (not overdrive)
Fig. 10.10.43: Deluxe-input-circuits: 5D3 (1954), 5E3 (Tweed, 1955), AB763 (Blackface, 1963).
Fig. 10.10.44: 5E3-Deluxe, first-stage-transmission, Tone-pot; volume-pot of “the other channel” turned down.

Only from the 6G3-Deluxe produced in 1960 the two volume pots are normally connected,
and from 1963 both channels receive a bass- and a treble-control each. The HD-ak2 of the 1st
stage is shown in Fig. 10.10.45: for the 5E3-Deluxe, the plate-load decreases as the volume
control is turned down – the gain drops and at the same time the distortion rises. Wide-open,
the distortion is less than in the Deluxe Reverb, due to the lower gain of the 12AY7. It has
already been noted (Chapter 10.1.4) that the distortion depends on the individual tube as well.
Fig. 10.10.45: 2nd order harmonic distortion from amplifier input up to the first plate. The reverse-connected
volume-pot is also found in other Fender amplifiers, e.g. in the Pro, the Princeton, and in the Super.
We need to consider the distortion of the 1st tube only if the guitar pickup delivers a high
output. Feeding 100 mVeff to the input of a 3E5 (with its volume set to 8) generates merely k2
= 0.5%, while the power amp is already pushed far into overdrive, as also documented by the
headroom charts (Chapter 10.10.3). The phase-inverter is always part of the power amplifier,
in its respective variant (paraphase, cathodyne, differential amp; Chapter 10.4). The 5D3 had
a paraphase with negative feedback while the 5E3 included the cathodyne-circuit, and the
AB763 had the differential amp. The signal symmetry resulting from the cathodyne circuit is
acceptable: for the 5E3, the overall k2 is smaller than the k3 across the whole dynamic range
(Fig. 10.10.20). In the differential amplifier deployed from 1956, the symmetry depends i.a.
on the plate resistors. Simply trusting the carbon resistors to be “absolute high-end” involves
risking large tolerances, and thus a large scatter range in the k2. In the 6G3-Deluxe, the plate
resistors differ in value, in the AA763 they are equal, and in the AB763 again different.
Equal means: both have 100 kΩ, different means: they have 82 kΩ and 100 kΩ respectively.
For the Super Amp, the evolution is similar: in the G64 different, in the AA763 equal, and in
the AB763 ??? According to the schematic, the resistors are equal, but the layout shows them
to be different. Indeed, it may happen on top of everything that the documents include errors.
Fig. 10.10.46: Differential amplifier with and without overall-negative-feedback; 2nd order HD.
Because the right-hand triode in Fig. 10.10.46 is driven by the cathode (common grid circuit),
its gain is a little smaller than that of the left-hand triode – this may be compensated by
increasing the right plate resistor a bit (e.g. 82 kΩ /100 kΩ). However, for the negative
feedback signal arriving from the output transformer, the left triode operates in common grid
configuration – equal resistors may serve as the compromise.

Do these small asymmetries actually play a role? Fig. 10.10.46 shows related measurements
for a Super Reverb (AB763). According to their color-coding, the plate resistors in the phase
inverter should have 82 kΩ / 100 kΩ, but in fact the values were 92 kΩ / 98 kΩ; the 82-kΩ-
resistance was too big by 12%. It was replaced by a resistor of the correct nominal value
which reduced the k2 at small drive levels considerably. Several opinions on this are possible:
1) The distortion in a guitar amplifier should be small and thus such a high degree of
symmetry is purposeful. N.B.: power tubes and output transformer can cause asymmetry, as
well!
2) Only with the 2nd order-distortion a guitar amplifier sounds typical for the genre. N.B.: as
above.
3) A THD around 2% is not really that important (Chapter 10.10.5). N.B.: other pairings of
resistors make larger distortion possible, too.
Fender schematics allow for a tolerance of 5% for the plate-resistors of the differential
amplifier, but they also document different philosophies: 82 kΩ / 100 kΩ, 100 kΩ / 100 kΩ,
47 kΩ / 47 kΩ, both with the 7025 and the 12AT7. This is a considerable spread in the phase-
inverter alone and exemplifies that a specific model (the Pro, the Deluxe) was built in very
different variants. Some amplifiers (such as the Pro) at least kept their power tubes (6L6, later
the non-identical 6L6GC), the operating point, however, changed from ‘automatic’ (cathode-
resistor) to ‘fixed’. The plate voltage changes, as well, from 350 V to 440 V. The Princeton
started out with a single power tube (6V6) and later received a second one. The Twin had two
6L6 to begin with and changes to four 6L6GC (or four 5881, respectively). The Tremolux
sports two 6V6GT first, and two 6L6GC later. The often yearly variations in the tone-filters
has been documented in Chapter 10.3 already; that coupling capacitors and small blocking-
C’s were different from one year to the next goes beyond the scope of the present concise
presentation.
There are, however, also similarities: the Pro corresponds to the Super with only the speakers
being different: the Super had 2x10", the Pro 1x15". At first, that is – later this changes and
the Super receives 4x10", the Pro 2x12". Both amps are again similar to the Bandmaster and
the Concert; merely the speaker configuration (and therefore sometimes the output
transducer) is different. The Tremolux is a Deluxe modified by the inclusion of tremolo (or
vibrato – Fender uses both terms synonymously), the Vibrolux is a toned-down version of the
Tremolux. The Showman is a head-only and the piggyback variant of the Twin. That’s one
side of the medal that after 70 years still shines brightly. The other side: not every difference
is audible.
Fig. 10.10.47: Fender amplifiers [www.Fender.com]

Marshall-Amplifiers
The variety among the types of Marshalls is not as huge as it is for Fender, but still
sufficiently big. Particularly confusing is the numbering: there is, for example, the type 1963
issued in 1966, the type 1987 from 1966, or the type 1964 produced in the year 1973.
Marshall’s type-numbering is not connected to the year of production at all! Doyle’s book
seeks to shed some light onto this and the special “Marshall”-edition of the German
“Gitarre&Bass”-magazine dedicates itself to the topic, as well. One criterion allowing for a
coarse classification is the output power: there are the small amps with 18 W (later 20 W), the
medium line with 50 W and the big ones with 100 W. Additionally, we find some exotic birds
such as the 200-W-behemoth, or the 1-W-dwarf, as well some odd mutants from the other
side of the tracks such as transistor- and hybrid-amps. On top of all this, there were often
guitar-, bass, organ, and PA-variants of each amp.
For many guitarists the “Marshall Stack” with its 100-W-amp sitting on top of two 4x12-
boxes is the prototype per se. It came into existence as derivative of a Fender-clone put
together around 1962 by Dudley Craven and Ken Bran: Bran copied the 5F6A-Bassman in
almost every detail and labeled it MARSHALL, adding JTM 45 shortly afterwards. There were
some differences in the pre-amp-tube (ECC83 instead of the 12AY7), in the power tubes
(KT66 instead of 5881), in the transformers and in the loudspeakers incl. the enclosure. The
change in the pre-amp-tube increases the input gain by 4 – 5 dB, the different power tubes
necessitate a change in the transformer (8 kΩ rather than the 4 kΩ customary for Fender)
which also brings a change in the negative feedback (Chapter 10.5.2), and swapping the
4x10"-Jensen for 4x12"-Celestion influences radiation patterns and sound. Also, the guitar
version of the amp received another capacitor to increase the treble. No circuit endured for
long: in 1964 we see the changeover to two EL34, and shortly thereafter (or even
concurrently) the power output explodes to 100 W. Any rumors that this was sponsored by the
hearing-aid-industry could, however, never be verified.
Fig. 10.10.48: Marshall JTM 45. The capacitors across the Volume pot depend on the type (100pF – 500pF). In
Doyle's Marshall-book, a choke (20 H) is included in the power supply: for some schematics this choke is
connected between plate and the remaining circuitry; there is a large difference with respect to the current.
Abb. 10.10.48 shows the schematic of the JTM 45 – it is nothing out of the ordinary. Old
documents indicate a plate voltage of 440 V (idle), and with this the amp will give about 30
W (with Raa = 8 kΩ). Allegedly the KT66 soon turned out to be too expensive, or too weak, or
both, and from 1966 the EL34 was used. A higher output power would in fact have been
possible with the KT66, as well, if Raa had been reduced, but … that was done only as the
EL34 was phased-in. At this point, the opportunity arose to swap the GZ34 rectifier tube for
silicon diodes, and to adapt the name to JTM 50. Now name and power did match – the JTM
45 had failed in that discipline.

The JTM 50 was in production for just shy of one year when the era of the JMP-amplifiers
began. Not wanting to list all type numbers in detail (this can be found e.g. with Doyle), a
coarse classification would be: JTM (1962 – 1967), JMP (1967 – 1981), JCM800 (1981 –
1989), JCM900 (1990 – 1999), from then JCM2000. However, this rough structuring does
not seek to imply that all JMP-amps would be similar. For exact specification the type number
is required but not sufficient because even within one single type-number there were
modifications.
The gold-colored plexi-glass screen of the early Marshall amps gave another grouping its
moniker; the “Plexi-Marshalls”, built until 1969, represent the pinnacle of Marshall-dom for
many. (For many but not for all: others find this zenith in the JCM800-, or in the JCM900-
amps, or in …). But we do not need to get into that here. Fig. 10.10.49 shows the circuit of
the 1987 with EL34 in the power-tube-position. The 1987-designation has surfaced already
for the first JTM 45 and is not unambiguous. Compared to Fig. 10.10.48, some differences
catch the eye: there are larger summation resistors, larger smoothing capacitors, a higher
supply voltage from the Si-diode power supply (not shown here), and other power tubes with
a changed negative feedback. The screen grid of the power tubes is connected directly –
without resistor – to a big 50 µF electrolytic capacitor, leading to scary-big grid-currents.
The changes in the capacitor values (22 nF rather than 20 nF) are due to the standardization
starting around the time (e.g. DIN41426) and allowing merely for the values of 10, 15, 22, 33,
47 and 68 nF in the E6-series but not 20 µF (these would only be elements of the E24-series).
Fig. 10.10.49: Type 1987, a 50-W-amp from the golden era, with shared pre-amp cathode.
In Fig. 10.10.50 we see a 1987-variant built from about 1969. The bass response in its “High
Treble”-channel as radically thinned out: smaller cathode- and coupling-C’s made for a more
aggressive sound, along with an extremely large 5-nF-cpacitor across the volume pot, an
altered tone-stack and a reduced negative feedback in the power amp. Corresponding details
may be found in Bernd Meiser’s article in the German Gitarre&Bass-magazine (07/2006).
Fig. 10.10.50: Marshall 1987, the development with the split-cathode in the input stage.

The big brother of the 1987 was the 1959. Rather than 50 W from 2xEL34, it generated 100
W from 4xEL34. Undistorted! At k = 10%, it had a remarkable 170 W up its sleeve, as noted
in the Marshall data sheet. In the 100-W-versions, the output tubes were given a grid-resistor
(1 kΩ) each; the 50-W-versions had to do without that for years and only received relief when
the high-gain “Two-Inputers” were issued in 1975. In the latter, the two halves of the input-
ECC83 could be connected in series, enabling them to offer an absurdly high overall gain.
The 2204 data-sheet notes an input sensitivity of 0.1 mV – that should be enough for any
pickup (which can generate – depending on the type – up to more than 1 V). The 50-W-
variant of this Heavy-Rock-amp was designated 2204 (Fig. 10.10.51); the 2203 is the 100-W-
version. To achieve strong overdrive at moderate loudness, the amps received a Master-
potentiometer. With this, however, the power amps did not contribute to the distortion
anymore, resulting in a different sound.
Fig. 10.10.51: Marshall 2204, 50 W Heavy-Rock with Master-Volume. The 2203 is the 100-W-variant.
What can be done to obtain strong distortion even at low loudness?

1) use an amplifier with small output power,
2) reduce the level ahead of the power amp (pre-amp-distortion),
3) reduce the signal between power amp and loudspeaker,
4) include a diode-distortion circuit.
The first variant (in the form of e.g. the 18-W-models) was thinkable for Marshall users but
what do you do in case you suddenly do need more loudness, after all? Variant 2 was
practiced e.g. in the 2204, the third option could establish itself only over the course of
decades (impedance-emulation), and the forth variant? Solid-state-distortion? Yes, indeed,
and for Marshall this era begins in the 1980’s. First, switching-transistors and 1N4007-diodes
find their way into the signal path (2205, 2210), later we see whole diode arrays (2250, 2255),
and then the OP-Amp-models with the alibi-tube (Valvestate) enter the picture. From time to
time, there were experiments trying to include a bit of the power-amp-distortion. An example
is the 4001 (as are the 4140, 4145, and 2150). Here, the master-volume is a dual-pot located
after the phase-inverter that can now still have an imprint on the sound. This approach never
enjoyed a wide-scale break-through, though. In the 2150 another detail is notable: it is a 100-
W-amp with a single 12”-speaker, a Celestion Powercell 12-150. This speaker can withstand
the 100 W – however: for it a white-noise-sensitivity of 89.8 dB (1W/1m) is listed, while for a
G12-50 the corresponding number is 97.4 dB. It’s the same old story: high-power speakers do
not necessarily have a high efficiency. 7.6-dB-difference corresponds to a factor of 5.75 in
terms of power. Recalculating: the noise produced by the Powercell with 100 W can be
generated by the G12-50 with merely 17 W i.e. all that power is wasted! Marshall should have
used the Powercell 15-250 specified with 95.5 dB. That would have been the true "Rock'n-
Roll-Baby", especially if the 200-W-power-amp would have been included into this combo.
The weight? Yeah, it would have been around 45 kg – the speaker alone weighs in with 14 kg.

An example for diode-distortion is shown in Fig. 10.10.52. The 2205 features two channels
selectable via a footswitch: a ‘Normal-Channel’ and a ‘Boost-Channel’. The switching is
performed by a transistor array (CA 3046) that however concentrates on that function and
does not contribute to any amplification. The Normal-channel stands out due to a special,
rather Marshall-un-typical, tone-filter. Its position (directly behind the input tube) and its filter
characteristic are indeed extraordinary. The Boost channel is where things get really exciting:
a diode in the cathode connection of the second tube stage increases the 2nd order distortion,
a bridge-rectifier takes care of signal-limiting on both sides. As a first approach, the rectifier
may be interpreted as an anti-parallel-arrangement of three diodes connected in series. The
most astonishing aspect is, however, the explanation published by Marshall: Critics have
wrongly alleged that this amp creates “transistor distortion”. Fact: in the channel-switching
of the head there is merely a voltage limiting via diodes – this however in no way works as a
distortion device but only limits the signal level and thus prohibits unwanted overdrive in the
following amplification stages. [http://www.marshallamps.de/equipment/2210-%28milestones%29—289;
available at the time; deleted from the Marshall website since]. Here, the statement “no transistor
distortion” would have sufficed – indeed, there are no transistors in the signal path. However,
why should a voltage-limiting not cause any distortion … no matter, there is good to report, as
well: the amp had send/return-jacks for connecting external effects, and a built-in spring
reverb. All in all a really advanced Marshall that receives special praise from Doyle: “Over
the years ... the 2210 had become one of the all-time great distortion amplifiers, and was
consequently even outselling the classic 2203.”
Fig. 10.10.52: Marshall 2205 (simplified). The 2210 has a 100-W-power-amp instead of 50 W. [Marshall.com]
In 1987 Jim Marshall celebrated 50 years in music and 25 years in amplification, Doyle
introduces the Silver-Jubilee-chapter, and one is tempted to add: “… and then they discovered
LED’s and parallel-negative feedback”. It is indeed possible to achieve distortion with two
anti-parallel-connected diodes, but the resulting voltage is rather small in a tube environment.
With a red light emitting diode (LED), a voltage approx. three times that of a Si-diode results
in the flow-direction – this saves components. So what about the parallel-negative feedback?
That has – in its entirety – the name parallel/parallel-negative feedback or I/U-feedback, and it
has several effects: input- and output-impedance as well as amplification are reduced but also
stabilized at the same time. And since it’s jubilee-time, here’s another feature: a switch
brought the output tubes from pentode- into triode-mode, halving the output power.

Fig. 10.10.53 shows the circuitry of the 2550. It already makes three basic settings available:
a lead-channel, and a rhythm-channel that can be put into distortion mode by a “Rhythm-
Switch”. The first picture shows the Normal-mode with the anti-parallel diodes having little
effect. This is because – watch out, here it comes – the subsequent tube operates in parallel-
negative-feedback-mode reducing the input impedance to about 50 kΩ. The diodes are not
entirely without effect but somewhat decoupled by the 47-kΩ-resistor (weakly distorting
compressor). This is very different in the distortion mode (right-hand picture). Now the
diodes are connected across the signal path and contribute hard limiting.
Fig. 10.10.53 Marshall 2550 (simplified). The 2555 has a 100-W-power amp instead of 50 W.
In the lead-channel two LED’s and three diodes (1N4007) take care of limiting. Since the
effective number of the diodes depends on the direction of the current, an asymmetric limiting
is achieved that somewhat prefers even-order distortion. The effect of this asymmetry is not
very strong and is only present for low drive levels just as the limiting starts. Compared with
the anti-parallel diodes (that limit at about 1.2 VPP), the LED/diode-combination has a
limiting voltage at about 5.5 VPP. The enables the lead channel to be louder than the rhythm
channel.
At the lower right in Fig.10.10.53 we see the power-amplifier switching. For pentode-
operation the screen-grids of the power tubes are connected to the supply-voltage (via a 1-kΩ-
resistor to limit the grid-current). The high, almost constant Ug2-voltage accelerates the
electrons nicely, and the cathode current can take on large values. In the triode-mode, the
screen-grid is at the plate potential. As the plate voltage drops with increasing drive levels, the
cathode current cannot increase to the same degree as in the pentode-mode. Gain and
maximum power drop to about half. However, there is a further effect because the power-
amp-impedance is reduced, as well (Chapter 10.5.14) – this is why amplifiers in the triode-
mode are not only less loud, but also differ somewhat in sound. Still, this is a good alternative
available at the discretion of the user. Inevitable, however, are the Si- and GaAs-diodes – is
this the new, typical Marshall sound?

Tube purists will turn up their noses at such grossness: semiconductors in a Marshall! Others,
however, buy and play and are happy. Doyle writes about the silver amps: and many people –
notably Slash, of Guns N' Roses – won't play anything else. Of course, such statements will
not remain valid for all eternity; rock musicians change their commitments as often as they
change their shirts (is that every other year?) … but they do not entirely miss the point, these
diode-Marshalls. The production numbers speak for themselves … or rather for the sound.
Fig. 10.10.54: Characteristic, time-function and normalized spectrum of the diode distortion circuit. Dashed in
blue: the spectral envelope of a square-wave signal (1/f-spectrum).
In Fig. 10.10.54, the diagrams relating to the anti-parallel diodes are shown. For strong
overdrive an almost square limiting results – for two LED’s a bit more strongly pronounced
than for two Si-diodes. The spectra diverge only little from the 1/f-envelope as long as one
stays with the lower partials. The corresponding curves for the Si/GaAs-combination used in
the lead-channel are depicted in Fig. 10.10.55. For strong overdrive (red), the main difference
re. Fig. 10.10.54 is the DC-component appearing at 0 Hz. For smaller drive levels (blue)
even-order distortion becomes visible, as well.
Fig. 10.10.55: Diagrams for the combination of 2xLED and 3x1N4007 (Marshall 2550).
After the generation of distortion had been successfully transferred to semiconductors, the
latter now also had to amplify. The 8040, for example, is a purebred transistor amplifier …
uh-oh … almost overlooked that alibi-tube. It almost drowns in that sea of OP-amps. The
circuits become more extensive, the model-variety, as well – too extensive for the present
overview. In short: after the JCM800-series the JCM900-series followed, having even higher
gain, and then the 2000-models. If it continues that way, the 3000’s should be in sight, soon.

www.Marshallamps.com
www.mylespaul.com

10.10.8 Modeling Amps
Modeling amplifiers are guitar amps with a large variability of transmission parameters
allowing for an approximate emulation of the sound of many well-known amplifiers. The
linear and non-linear signal processing is usually done in a digital signal processor (DSP); the
musician can call up different amplifier models from the program memory. First on the
market were the Roland and Line6 companies, and by now many others have followed suit.
The recipe: take a good AD/DA-converter, a low-cost switching-power-amp and a DSP-board
– and you get 12 (or 24) of the most famous guitar amps in a little box. Is it actually that
simple? No, it ain’t! It is not sufficient to emulate the frequency responses of the famous
predecessors; it is also their non-linear distortion and their operating-point-shifts that need to
be modeled. It is here where the difficulties really begin: while it is possible to combine the
linear characteristics of cascaded stages, the non-linear stages need to be emulated
individually. It has already been mentioned repeatedly that the interface between tube power-
amp and loudspeaker needs special consideration. To simulate every detail in the software is
not helpful, either, since this increases the calculation time in the processor (i.e. the
responsiveness of the amplifier becomes sluggish). The constant development of the
algorithms has by now led to useful concepts which – in direct comparison to the original –
still leave a bit to be desired, but which due to their unbeatable variability are preferred by
musicians who need to cover a wide range of styles and sounds.
The following investigations were carried out on a VOX AD60VT, an amplifier that not only
practices digital signal processing but also filters using an interesting power-amplifier circuit.
The block-diagram is shown in Fig. 10.10.56.
Fig. 10.10.56: Signal-processing in the VOX AD60VT (simplified).
The guitar signal reaches the digital signal processor via an impedance converter and a treble
pre-emphasis, and is then fed via a complementary treble de-emphasis to the power stage.
Immediately striking; the input impedance is not the 1000 kΩ typical for VOX but merely 560
kΩ, and the non-linearity of the grid found in tubes is not emulated. These characteristics are
not the main reason why tube amps are much beloved, but this lack is not “perfect modeling”,
either. On the other hand, this VOX amp (as well as the more powerful AD120VT) scores due
to its very special power-stage. The output impedances of tube amplifiers are high, even with
the output transformers (Chapter 10.5). Due to this, the loudspeaker impedance influences the
frequency-response of the transmission and thus the sound. The VOX does account for this
scenario using an actual tube power-amplifier (incl. output transformer). No, that’s not a high-
power output-stage but a modest 1-W-power-amp making do with the two triodes of an
ECC83. The resulting output voltage is not simply further amplified and fed to the speaker
via a low-impedance transistor-amp; rather, the output of the midget tube amp is connected to
a high-impedance power-amp. For the tube amp to catch something of the speaker behavior,
the speaker voltage is fed back to the tube amp. This way the output transformer senses a
load-impedance as it is typical for a loudspeaker, and the linear and non-linear characteristics
of the push-pull output-stage take effect in the usual manner.

Fig. 10.10.57 shows details of the VOX power-amp. A very familiar phase-inverter is present
as is even the 82k/100k-pair; there are two tubes in push-pull configuration, there is a
transformer … and now it gets really interesting. Via a power-selector-switch we arrive at the
power-amp that is best described by its conductance S (just like an OTA♣), and then we are
guided to the loudspeaker and via a second power-selector back to the transformer. The
power-OTA works in a substantially linear fashion, any overdrive happens in the triodes. A
feedback circuit may be placed between the connectors designated with NFB (Fig. 10.10.58)
but this is deactivated in the typical VOX-circuit. Because of the opposed effect of the two
power-selection-switches, the loop gain (and thus the transformer load) is not (or only
negligibly) dependent on the position of the switches – but the power fed to the loudspeaker
is. With the dimensioning chosen in the AD60VT, the secondary winding of the output
transformer “sees” approximately the 50-fold speaker-impedance, including the
corresponding frequency dependency. And this, my friends, is indeed typical for a tube amp.
Fig. 10.10.57: Power-stage circuit of the VOX AD60VT Valvetronix (simplified).
Tube-amp-typical, however, does not generally imply guitar-amp-typical. In this VOX, two
triodes are at work, while in the famed forefathers we had two or four pentodes doing that
job. Nevertheless: it’s a speaker-loaded tube power-amp. The basic principle of the load
transformation is shown in Fig. 10.10.58: the input impedance Z1 calculates (in an idealized
way) to Z1 = R⋅(kSZL+1), and therefore is approximately proportional to the loudspeaker
impedance ZL, as long as kSZL remains large relative to 1. This requirement is pretty nicely
fulfilled: for 8 Ω speaker impedance, Z1 is 380 Ω, and R with about 30 Ω does not get in the
way. The secondary resistance of the transformer (180 Ω) has a somewhat stronger effect, but
the real culprit here is the rather high copper resistance of the primary winding that drastically
reduced the model consistency. This is the result of the relatively small transformer (EI-42).
And since we are looking closely now: the feedback network seeks to be a compromise
between authenticity and effort, and e.g. fails to offer the continuous control possibility of a
presence-pot. For modeling the AC15 or AC30, this is o.k., but with respect to emulating the
Bassman or Marshall amps it is an issue. The grid circuit of the triodes, on the other hand,
deserves praise with its switchable resistors, as does the switchable cathode-resistor (not
included in Fig. 10.10.58).
Fig. 10.10.58: Negative feedback circuit (left), "Vari-Amp-circuit" (middle), transformer (right).
♣
Operational Transconductance Amplifier

The switchable common cathode resistor creates the possibility to operate this power-amp in
either A- or AB-mode. To emulate the AC30-power-amp, the triodes work with about 2.2 mA
plate current, which is about the middle of the characteristic, and thus A-mode. The primary
impedance is in total about Raa = 50 kΩ for an 8-Ω-load, i.e. 25 kΩ per triode (cf. Chapter
10.5.5). Although pentodes are at work in the AC30, although especially in this amp the plate
resistors are equal (100 kΩ / 100kΩ, not 100kΩ / 82 kΩ), although the non-linear Cut-filter
is not modeled correctly even to begin with, and although the transformer is terribly high-
impedance … that’s an approach one can live with. The circuit of the transistor power-amp is
shown in Fig. 10.10.59. The input-transistor operates in common-base-configuration and
feeds a complementary Darlington-circuit. The emitter output could be interpreted as low-
impedance – but that would not be correct. The driver transistor approximately works as
current source and the output transistor as current amplifier, the current through R and
through the loudspeaker being almost equal. Thus, this circuit has a high-impedance output
just like a tube amplifier. It is only at very high frequencies that output impedance drops off
due to the voltage feedback via the RC-circuit – and that effect is in fact rather purposeful.
Fig. 10.10.59: AD60VT-power-amp (left), effect of the complementary-Darlington-circuit (right).
So, the AD60VT-ouput-stage has received considerable tube-like-qualities – but what about
the digital modeling? Unfortunately, that is as inadequate as it is found in other DSP-amps:
there’s some filtering, some distortion, and that’s it. It might be understandable that the input
stage does not emulate a tube-input – the effort must not too big, and the DSP-board found in
the Amp seems to be a rather universal one. It has already been mentioned that the input
impedance is not 1 MΩ. One could get over the quite small input capacitance of a mere 75 pF,
but that the Lo-input also features 560 kΩ, that wouldn’t have to be: for almost all amplifiers,
this input is – at usually 136 kΩ - of clearly of lower impedance compared to the Hi-input. Be
sure: this has significant effects on the dampening of the pickup resonance.

Even more problematic: as a floor-pedal-model is chosen, the 560-kΩ-input-impedance still

remains. For all 16 amp models and for all 10 “Effect Pedals”: always 560 kΩ. Conversely,
especially distortion pedals and treble boosters often have very low input impedances (some
down to as low as 10 kΩ), but this was apparently not grasped by the VOX-people or whoever
came up with this korg-promize. The uPC4072 used for the input does not feature any tube-
like clipping, either – all non-linearity happens in the DSP. Alright then, let’s analyze the
distortion that the latter provides and let’s see how it models different amps, for example the
AC30TB compared to the AC30. Fig. 10.10.60 shows the corresponding HD. Big surprise:
VOX apparently did not catch that these two amps are distinguished by the infamous cathode-
follower (in the AC30TB). Or maybe they sought to emulate the Normal channel in the
AC30TB? No, that would have been a laughing matter, and the manual does specify the
“Brilliance unit”. Apparently an additional treble boost was thought to be sufficient. A
measurement of the AC15-model can be seen as the third curve in Fig. 10.10.60, and it is
barely different from the two “colleague”-models. These are not untypical distortion
characteristics, and one can get by working with them – it is however not an actual distortion
model of the famous ancestors.
Fig. 10.10.60: Harmonic distortion (AD60VT-DSP) for the AC15, AC30 & AC30TB-models. 2 gain settings.
Apparently, the differences between the amplifier models are limited to modifications of the
frequency responses, as they are documented in Fig. 10.10.61. A few ripples, more gain and
more treble for the AC30TB-model – that’s it. One criterion that apparently was seen as
deserving some more attention: the sequence of filtering and limiting. For some models the
treble content of a distorted 500-Hz-tone can be strongly altered (i.e. the filter is located post-
clipping) while for others almost no effect is present. Model-specific characteristics are
recognizable in the time-functions of the distorted sine-wave, as well, and there are large
model-specific differences in the behavior of the tone controls. However, there are
inconsistencies, too: the VOX-manual states that “Presence” is a “feature in the power-
amplifier”, but there is no Presence-potentiometer anywhere in the VOX-power-amp – the
effect is calculated in the DSP.
Fig. 10.10.61: Transmission characteristic (AD60VT-DSP) from input to an 8-Ω-load; (B = min, M = 12:00, T =
13:30, Pr = 12:00). On the right is the dampening of the power-amp feedback (Fig. 10.0.58, AB-models only).

The negative feedback in the power-amp of the AD60VT is simply not variable as it is the
case in amplifiers with Presence-function. It has only 3 variants (Fig. 10.10.58): off, slight
treble cut and strong treble boost. The results of the measurements♣ depicted in Fig. 10.10.61
were taken with a resistive 8-Ω-load – connecting the speaker results in a (desirable) treble
emphasis. And again, as we look closer: the two variants in the grid-circuit are a well-meant,
nice try to start; however, the recharging of the coupling capacitors (Chapter 10.10.4) happens
in real life (i.e. with the EL84, 6V6, 6L6 or EL34) with more than two variants.
So, what does remain of the promise, that each and every one of our models is as tonally
authentic as possible - as opposed to the usual “close but definitely no cigar” norm of digital
modeling [VOX-Manual]? A definitely useful, versatile amp with purposeful control concept
(let’s not talk about the VC-4, though). The AD60VT certainly is not an amplifier in which its
16 different amp-models perfectly imitate the corresponding real amplifiers. That simply
cannot work since – for a start - the loudspeaker cannot emulate all of the sound radiation
patterns of an 8x12”-Marshall stack, a 4x10”-Bassman, a 2x12”-Twin and a 1x12”-Deluxe.
And because the amp (for economic reasons?) does without certain special circuits (Cut,
Presence). And because the nonlinear distortion is emulated in a rather simple fashion in the
DSP. And because the speaker is a typical Celestion, and not a Jensen or Eminence or JBL.
Still: useful. That the distributor resolutely shoots down any inquiry regarding schematics:
forget it – the manual for the AD120VT can be found on the internet, and the printed circuit
for the power-amp is single-layer and thus easily analyzed .
Model presumably Tubes AD60VT-pwr-amp sequence Pwr-amp-FB
Boutique CL Dumble 4x6L6GC A CF no

Black 2x12 Twin-Reverb 4x6L6GC AB FC
Tweed 1x12 Deluxe 2x6V6GT A FC no
Tweed 4x10 Bassman 2x5881 AB FC
AC15 AC15 2xEL84 A FC no
AC15TB AC15TB 2xEL84 A FC no
AC30 AC30-6 4xEL84 A FC no
AC30TB AC30TB 4xEL84 A FC no
UK Blues JTM-45 2xKT66 A FC
UK '70s Marshall Plexi 4xEL34 AB FC
UK '80s 80’s Marshall 4xEL34 AB CF
UK '90s 90’s Marshall 4xEL34 AB CF
UK modern Marshall 4xEL34 AB CF
Recto Mesa Tri-Rectifier 6x6L6GC AB CF
US HiGain Soldano 4x6L6GC AB CF
Boutique OD Dumble Overdrive 4xEL34 A FC no
Table of the amp models in the VOX AD60VT (to the best of our knowledge). FC = Filter -> Clipping, CF =
Clipping -> Filter.
♣
Attenuation in the negative-feedback branch results in gain in the forward branch.

Let us move now from "Valvetronix" on to "Valvestate", i.e. from VOX to Marshall.
Because here, as well, there are (besides the famed all-tube amplifiers) – watch out! –
transistor amplifiers in existence. Ouch!! Fear not, though, dear guitarists: they do include an
alibi-tube. “The new Advanced Valvestate Technology (AVT) is the fruit of years of
development and innovation since the birth of the original Valvestate amplifiers. The
resulting new hybrid technology outclasses in one stroke all 'virtual' and 'modeling' amp
concepts, and is therefore today the best possible alternative to all-tube amplifiers"
[Marshall]. This is because: "All AVT-preamps work with a ECC83 (12AX7) preamp tube.
This tube makes for authentic bell-like clean-sounds and harmonically rich overdrive that
cuts through". Indeed, that had to be said – finally. What is rather not said is that the actual
non-linearity is generated by two anti-parallel LED’s. And it is only the IC-data-sheet that
tells us that the power-amp – previously the undisputed territory of the EL34 – is now
dominated by a solid-state power circuit that was developed "for use as audio class AB
amplifier in HiFi field applications (Home Stereo, Top Class TV)". Marshall only writes that
the AVT-power-amp is unique. We happily take their word for it. Marshall also could have
written in the brochure that the typical THD of this power-amp-IC is a possibly record-
breaking 0.005% (IC data sheet), but this remains unmentioned – maybe they took this as a
given. That, on the other hand, the boucherot-resistor tends to throw in the towel – this info is
obtainable via the internet. It seems not that easy to exorcise RF-oscillations from this
Marshall power-amplifier in the framework of series production … sounds familiar, many a
service-technician will think to him- or herself.
If we trust Marshall, a single ECC83 is sufficient for an authentic tube sound. An ECC83
preamplifier tube, but that doesn’t mean that it is employed in the first amplifier stage. In the
latter an NMJ072 takes care of business, followed by an M5201. The M5201 is a so-called
"switching-OP" i.e. a switchable operational amplifier that activates either the Clean- or the
Overdrive-channel. This NMJ072 has a rather modest 2V/µs-slew-rate, but the fan base
thankfully offers advice: a swap for an NJM2121 boasting 4V/µs. That’s actually not a lot,
either … whatever – the priorities seem to be shifted elsewhere. They might lie, for example,
with the anti-parallel light-emitting diodes connected in the negative-feedback-branch: 3mm,
red. Would yellow sound different? Affirmative! And let’s not even talk about green or blue.
Marshall: only genuine sporting the red distortion LED’s. At last, the ECC83 is called up for
service now, fired up using a terrifying 109V supply voltage and a series-heating-voltage of
13.5 V. 13.5 V? Yep – in Marshalls, tubes always had to suffer. However, the circuit of the
ECC83 in not the infamously distorting cathode-follower pressed into service since the
JTM45-days (i.e. it is impossible to imagine a true Marshall without it). No, both triodes
operate in common-cathode-configuration, without the cathode-capacitor – they are really
well-behaved. “Authentic”, as the brochure notes. They feed their signal to a further OP-amp
(i.e. a whole lotta transistors – and why not?); then another two anti-parallel LED’s (3mm,
red) spring to action, and two more OP-amps (the reverb needs to connected with befitting
style, as well), and off the signal goes to the power-module (the output IC). It must not have
been easy to get a handle on all this with respect to RF-stability: there is a C93 so it probably
wasn’t doable with only 92 capacitors. And in the middle of the whole shebang: the tube –
with its photo in the top position in the brochure. Tube amp, advanced technology! Now if we
would replace this ECC83 by two FET’s … ÿÿÿÆ¿¿ÐþĄĦĶŦǿǿ .... no, even WORD cannot
deal with this anymore. Marshall with transistors only … shudder ….,
Over the years many manufacturers have tried to emulate that sought after all-valve sound using solid state tech-
nology. All such attempts failed miserably up to now. Enter Marshall's Valvestate technology [Marshall].

Translator’s note: the following 3 pages contain an ironic send-up of the interaction
between “experts” in the media and resulting legal issues. The included specifics and
terms used work best in German and are therefore left un-translated.
Anwälte, oder: Wenn die Sau läuft
So schnell kann's gehen. Es gab ja Zeiten, da wünschte man Thorben ein paar Probleme an den Hals,
aber Anwälte – so bösartig war keiner der Kollegen. Anwälte! Wir betreten gerade die Kanzleiräume
von Winklhofer & Winklhofer, herrlich am See gelegen, und während Thorben die 5 cm dicken
Acrylglastüren zu öffnen versucht, murmelt er nur ein verzweifeltes "so eine Scheiße." Dies betraf
weniger die Türen, die ein dienstbarer Motor mit einem leisen "ffft" aufschwingen lässt, kaum dass
man die Hand in Griffnähe bringt, nein, das betraf die Sache an sich. Die Sache (resp. die Causa) war
Thorbens erster Versuch als Autor. Er, der durchaus begabte Fotograf, hatte sich in den Kopf gesetzt,
zu seinen Bildern auch die Story zu fabrizieren, also nicht nur Pics, sondern auch noch Docs. Und
hatte sich gleich als erstes eine Kolumne der bekannten Zeitschrift "Gitarre 4 U" vorgenommen, in der
ein sehr von sich eingenommener (aber technisch noch nicht so ganz kundiger) Kollege Vermutungen
über "Vintage Guitar Amps" unter die Leser brachte. Thorbens Meinung hierzu schlug ein wie das
Ding, das man in Zeiten weltweiter Krisen nicht mehr gern beim Namen nennt, und rief zuerst den
erbosten G4U-Autor, und dann dessen Anwälte auf den Plan. Winklhofer & Winklhofer,
Wessling/Moskau/Tokio/NewYork.
Lektion 1: Tritt dir jemand auf die Füße, hol dir den größten Bruder, den du kennst. Das anwaltliche
Schreiben, 1½-zeilig auf handgeschöpftem Papier, war einer dieser Binnenbriefe: Wennse nicht
binnen 2 Wochen eine Erklärung abgeben, könnense gleich mit der Zahnbürste in SantaFu antreten.
Natürlich besser formuliert, und 5 Seiten lang, aber im Prinzip: Entweder jetzt viel zahlen, oder später
viel mehr zahlen. Denn mit den 5 handgeschöpften Seiten war's ja nicht getan, als Anlage fanden sich
auch noch ein paar weniger-wertige DIN-A4-Blätter, die ganz unverhohlen dazu aufforderten, die
Traumlage am See mit 3591,- Euro zzgl. MWSt. mitzufinanzieren. Immerhin wurde ohne direkt
ersichtliche Mehrkosten ein Gesprächstermin mit Herrn RA Gerhard O. Winklhofer angeboten, dem
Juniorpartner der Sozietät (der Senior kam wohl erst bei Fünfstelligem in die Puschen). So schnell
kann's also gehen.
Wer weiß, was dezent getöntes Acrylglas kostet, kann verstehen, warum die Herren Anwälte die
Hände ihrer geschätzten Mandantschaft hiervon gern fernhalten und Servomotoren einbauen lassen.
Aber mal ehrlich: Man könnte sich auch ans Harlachinger Krankenhaus erinnert fühlen, da gehen die
Türen nämlich auch von selbst auf, kaum dass man in ihre Nähe kommt. Doch der Boden beseitigt
jeden Zweifel: Carrara – also nicht maroder Münchner 60er-Jahre-Charm, sondern: Wessling!
(Moskau/Tokio/NewYork war's allerdings auch nicht ganz). Eine Blonde in perfekt sitzender
Business-Kombi hob ihr entzückendes Lockenköpfchen und fragte strahlend: "Ja bitte?" 'Guitar-Lix-
and-Trix-zu-Herrn-Winklhofer-Termin' schnarrte Thorben herunter, der sowas von angefressen war,
dass man Tätlichkeiten nicht ausschließen wollte. "Wenn Sie bitte einen Moment Platz nehmen
möchten? Darf ich Ihnen einen Kaffe anbieten?" Thorben unterdrückte ein 'wir nehmen alles was nix
kostet' und nickte nur, und dann versanken wir einen kurzen Moment in den Besuchersesseln.
Den wollte ich nutzen, um Thorben etwas einzubremsen, aber schon kam eine weitere Blondine in
Business-Kombi mit zwei Tassen Kaffee (und 2cm2 großem Keks) sowie ein perfekt gestylter
Endzwanziger mit affenscharfer Designerbrille. "Von Greiffenklau – wir gehen ins B8, Herr
Winklhofer kommt dann hinzu". Da konnte man nicht mäkeln, das war schon irgendwie perfekt, wenn
auch nicht ganz billig. Das Acryl zwischen Foyer und Arbeitstrakt schwang auf, und Thorben folgte
dem Rat, nicht gleich in medias res zu gehen. Doch, das verstand er schon, er hatte nämlich früher
einmal 2 Semester Philosophie studiert. In Passau ... bevor's dort zum Eklat kam.

Thorben, bitte, wenn schon einmal ein Bischof zu Besuch in der Uni weilt, sollte man halt nicht gleich
loskrakeelen, dass ja wohl nicht Gott den Menschen, sondern wohl eher der Mensch sich einen Gott
geschaffen hat. Mensch Thorben, Passau! Und dann noch nachzufragen, ob sich Luthers in der Woche
zwier auch auf die Missbrauchsfälle ... nein, Thorben, Philosophie wäre sicher nicht dein Ding
geworden. Also nicht gleich in medias res, Thorben – Smalltalk!! 'Schönen Boden hammse da, kommt
gut zu der Art Deco.' "Ja, das sagen viele. Schiffseiche!" Erklärend wandte sich Thorben mir zu: 'In
Bayern spricht man von schiffen und von Seiche, wenn...' Es war nicht zum Aushalten, nein – ich hätte
es wissen müssen! Von Greiffenklau, eindeutig außerbayerischer Provenienz, guckte sehr verwirrt,
und wollte schon nachfragen, als von rechts unerwartet eine Acrylgläserne in den Gang schwang und
mir fast das Kaffeetässchen aus der Hand geschlagen hätte. "Bitte nicht zu nahe rangehen, die
Electronic ist hier sehr sensitiv" warnte von Greiffenklau, und man spürte förmlich, dass er Electronic
mit "c" aussprach. "B8, wir sind da, ich darf mal vorgehen." Diesmal blickdichtes Acryl, achtsitziger
Teak-Tisch, Leder, Kunst, Seepanorama: B8. Achtsitzig! Das war uns schon früher aufgefallen, sie
kommen selten allein, die Herren Advokaten. Meist in Rudeln, wie die Waschmaschinenmonteure. Ein
kurzes Telefonat (die Herren von der Zeitschrift sind jetzt da,... ja,... B8, nein, danke), und Tröstliches:
"Herr Winklhofer kommt sofort." Man hätte nun eigentlich erwartet, dass ein Sensor unsere An-
wesenheit automatisch entdeckt und der Herr W im nächsten Moment von unten mitsamt seinem Sitz
durch den Boden geschossen kommt, aber das wäre wohl doch zu James-Bond-mäßg gedacht.
Stattdessen schwang die Blickdichte auf, und Herr W trat selbstlaufend herein. Begleitet von einer
weiteren Blonden in der Kombi (das war jetzt schon die dritte), die sich mit einem Stuhl Abstand
neben dem Herrn W niederließ und ermunternd lächelnd den Laptop aufklappte. Beeindruckend.
"Wir wurden von unserer geschätzten Mandantschaft beauftragt, gegen Sie eine Abmahnung
vorzubringen und eine Erklärung aufzusetzen, gemäß der Sie es künftig bei Androhung einer
Konventionalstrafe in Höhe von 250.000,- Euro unterlassen, das Ansehen unserer Mandantschaft
verächtlich zu machen. Sie haben unseren Entwurf erhalten?" '250 Mille für G4U? Das sind bei denen
doch drei Jahresumsätze?' Thorben war schon wieder auf 180, doch sein Gegenüber blieb
geschäftsmäßig kühl: "Ich bitte Sie, gleich in Ihrem ersten Absatz schreiben Sie, der Autor sei dumm
wie ein Stück Scheiße ... das ist doch kein elaborierter Code, das ist schlicht ein Insult." "Was für ein
Kot soll das sein?" maulte Thorben, dem der Adelige mit der Affenbrille schon mächtig auf den Geist
ging. "Meine Herren, versuchen wir doch, die Angelegenheit schnell und professionell hinter uns zu
bringen." Auch der Herr W verteilte nun ein paar seiner teuren Worte. "Ich will ja nicht das
Sprachniveau Ihrer Zeitschrift im Allgemeinen kritisieren, aber ihr Artikel gegen unsere
Mandantschaft ist definitiv beleidigend. § 185 StGB – um konkret zu sein."
Nun musste ich auch mal ein paar Worte sagen, Thorben ging die Sache einfach zu emotional an: "Ich
finde nicht, dass sich aus diesem Satz eine beleidigende Absicht herauslesen lässt. Mein Kollege
schreibt ja: 'Wüsste man es nicht besser, man könnte glauben, der Autor sei dumm wie ein Stück
Scheiße'. Daraus ergibt sich nach den Regeln der Logik doch zweifelsfrei, dass Ihre geschätzte
Mandantschaft, eben gerade kein Stück Scheiße ist. Also zumindest diese eine, geschätzte..." Nun gut,
der Halbsatz war auch nicht ganz emotionsfrei, aber das mit der Logik, das hatte doch was, oder? Und
gleich nachgesetzt: "Zur Juristenausbildung gehört doch noch immer Fausts Collegium Logicum,
oder?" "Florian Faust, BGB?" wollte von Greiffenklau wissen, und war damit endgültig raus, wie ein
entsetzter Blick seines Herrn W verriet. Auch die Blonde, die gelegentlich Notizen machte, verdrehte
kurz die Augen. Sie war gar nicht ohne, und bei jedem "Scheiße" musste sie an sich halten, um nicht
unschicklich loszukichern. Gleichzeitig schien sie sehr fasziniert von Thorben, der immer, wenn
keiner der Beanzugten herschaute, minimalmimische Signale an sie sandte, was jedes Mal ein ganz
kurzes Lächeln auf ihre Lippen zauberte. Thorben gab den Proll, ja schon, aber mit zwei Semestern
Philosophie im Rücken. "Sie wissen ja" begann er gerade zu erläutern, "dass die Nichtbeleidigung, die
nicht unabsichtlich vermieden wird, schwerer zählt als die vorsätzlich vermiedene Unterlassung einer
Beleidigung. Schon Kant hatte ja in seiner Kritik der ..." "Würden Sie bitte nochmals Ihre..." fiel ihm

von Greiffenklau ins Wort, womit die Situation endgültig aus dem Ruder lief. Dem Herrn W war
dieses Tamtam die ganze Zeit über schon zuwider, seine Farbe wechselte schlagartig in ein dunkles
rot, und im Business-Kombi entlud sich ein unkontrolliertes "mmpff", sofort gefolgt von einem sehr
verlegenen Händchen-vors-Mündchen-Halten.
Doch – wär's nicht so teuer, man könnte sich von den Herren Advocaten und ihren Maskottchen schon
ganz elaboriert unterhalten lassen. Und von Thorben, der gerade nachsetzte: "Und das mit dem
Aufhängen war ja auch von einer Art Mentalreservation begleitet." Das-mit-dem-Aufhängen, das hatte
so richtig Schwung aufs Pleuel gebracht, und den G4U-Autor dazu veranlasst, zwei Hefte später zu
schreiben, er ließe sich nicht von diesem neofaschistischen Gehirnamputierten seine Amp-Clinic-
Kolumne kaputtschießen. Da konnte man geneigt sein, schon wieder ein Intelligenzdefizit
festzustellen: Denn während Thorben (ganz im Sinne Kants?) meinte, früher hätte man derartige
Schriften auf der Straße verbrannt und den Autor an der nächsten Laterne aufgehängt, schrieb der
G4U-Autor 'von diesem neofaschistischen Gehirnamputierten'. Ohne jede Einschränkung,
Direttissima, dant genommen. Nun ist es ja eigentlich sinnlos, einem Juristen etwas erklären zu
wollen, ein Oxymoron sozusagen, (wie z.B. auch 'Frankenschnellweg'), aber Thorben versuchte es
trotzdem: "Das Temporaladverb FRÜHER ist ja in seiner Tempusspezifikation eher unbestimmt, und
schließt auch sehr frühe Zeitvorstellungen ein, sodass ohne ergänzende Konkretisierung 'nach erstem
Anschein' auch Verjährung nicht auszuschließen wäre". "Mord verjährt nicht", versuchte der
Hochwohlgeborene die Situation zu retten, kassierte aber postwendend ein "aber Totschlag schon,
gelle?" Was man in Philosophie so alles lernt – und noch dazu in Passau – faszinierend. Nun war
Thorben auf Betriebstemperatur, hatte vielleicht Lektion 2♣ im Kopf, als er nachsetzte: "Da mich Ihre
sehr geschätzte Mandantschaft in Heft 07 öffentlich als 'Gehirnamputierten' bezeichnet hat, werde ich
wohl als nächste Maßnahme von meinen Anwälten Gegenklage einreichen lassen. Konkret: § 185
StGB – sie erinnern sich?"
Dann ging alles ganz schnell: Man werde sich mit Herrn Winkladvokat Senior besprechen, und die
geschätzte Meinung der geschätzten Mandantschaft im nächsten Schriftsatz mitteilen. Falls sich die
Geschätzten tatsächlich diesem Vorschlag anschließen könnten, würden sich die geschätzten Kosten
dann lediglich um die Vergleichsgebühr erhöhen. Und Tschüss.
Wieder draußen, blickte Thorben gedankenverloren auf die gerade erbeuteten Visitenkarten. "FLORA
GARLEITNER – geiler Name. Ich wollte sie ja noch fragen, ob ich ihr behilflich sein kann, falls sie sich
mal in Deflora umbenennen will." Thorben sui generis, wie immer halt.
♣
Konfuzius sagt: wenn die Sau läuft, lass sie laufen

10.11 Tube Data 10-325
10.11 Tube-Data
10.11.1 Nomenclature for tubes
Amplifier tubes (or valves, as they are called in British English) are designated with letters
and/or numbers, e.g. ECC83. While this system helps as a coarse classification, we must not
expect precise statements regarding the function of the respective tube. Even within one and
the same batch of tubes manufactured in one go, there will be variations due to production
aspects. Often, developments led to changes in characteristics that were not necessarily
reflected in the designation, plus competitors sought to win over customers with
improvements. Today, the designation system has gone completely astray because almost
every “supplier” tends to invent fancy designations to make “his” tubes stand out more. The
classic system for European and US tubes, respectively, specifies them as follows:
The first character stands for the heater voltage: G and 5, respectively, for 5.0 V, E and 6 for
6.3 V, 12 for 12.6 V. A 12AX7 (and many similar double triodes) may be operated with 6.3
V, as well, by connecting the heaters in parallel rather than in series.
The second character designates the type of tube/valve in the European system: C for triode,
F for pentode, L for power pentode, Y for half-wave rectifier, Z for full-wave rectifier. In the
US-system the letters are consecutively allocated to this character position.
In the European system, the third character stands for the tube socket: 3 = octal (8-pin)
socket, 8 = 9-pin socket. In the US-system, the corresponding character designates the type of
tube: 4 = triode, 7 = double triode, 6 = tetrode or pentode in an octal socket.
In the US-system, the letters following the dash stand for the type of glass container: G for
the large, bulbous container, GT for the small cylindrical one. A prefix “W“ indicates MIL-
specs. The last letter designates the production version starting with an “A”.
Examples:
ECC81, ECC82, ECC83: double-triode, 9-pin socket, 6.3 V (or 12.6 V) heater.
EF86: pentode for preamplifier, 9-pin socket, 6.3 V heater.
EL34: power-pentode, octal socket, 6.3 V heater.
EL84: power-pentode, 9-pin socket, 6.3 V heater.
GZ34: full-wave-rectifier valve, octal socket, 5.0 V heater.
EZ81: full-wave-rectifier valve, octal socket, 6.3 V heater.
12AT7, 12AU7, 12AX7, 12AY7: double-triode, 9-pin socket, 12.6 V (or 6.3 V) heater.
6L6-G: power-pentode, octal socket, 6.3 V heater, bulbous glass container.
6L6-GA, 6L6-GB, 6L6-GC: continued development of the 6L6-G.
6L6-WGC: 6L6 with military specifications.
6V6-GT: power-pentode, octal sockets, 6.3 V heater, glass container.
5Y3-GT: full-wave rectifier tube, octal socket, 5.0 V heater, glass container.

10-326 10. Gitarrenverstärker
10.11.2 Double-triodes
Which idle current, or which internal impedance does an ECC83 have? Such data may be
found in the tables of tube manuals. These tables, however, are likely to have been compiled
50 or more years ago when manufacturers such as RCA, General Electric, Telefunken,
Sylvania, and many others, were still producing tubes. Today, only few manufacturers
remain: they each manufacture using several labels. As a consequence, a tube labeled
“Valvotron” may well stem from the same Chinese manufacturer as another tube labeled
“Tubitronics”. Their data may correspond to values found in the tables – or not. As an
orientation, these tables and manuals are certainly helpful: they specify mostly average
characteristics applicable for typical operating points in the HiFi context … which may not be
at fit for guitar amplifiers. More information is given by data sheets featuring characteristic
curves, but these may have been subjected to an averaging process (to make them “look
better”), or could be a third-generation copy of dubious quality.
The above may be the reason why our freshly unpacked tube does not match the data sheet.
Another reason may be the tube itself because, depending on availability, suppliers will often
offer similar (but not identical) tubes under the same designation. For example, the packaging
may specify “5751 = ECC83 = 7025” although the 5751 has slightly different data compared
to the ECC83. We might be reminded of the chocolate Santa the brown body of which, after
removing the wrapping, looks suspiciously like an Easter Bunny. It’s all marketing driven …
To say that an ECC83 is the same as an E83CC is not quite correct, either, since in fact the
E83CC is a special tube⊕ (long-life-tube, long-distance-communication-tube). Such tubes
often have gold-plated grids or zirconized electrodes with a highly special cathode build in
order to avoid the development of a disruptive intermediate layer. There’s magic in the
numbers: the 7025 supposedly is a special version of the 12AX7 that in turn is an equivalent
to the ECC83. The E83CC is a special version of the latter … but according to the data sheets,
it does not correspond to the 7025 but to the 6681 …
European designation ECC 81 ECC 82 ECC 83 - -

Alternate designation 12AT7 12AU7 12AX7 12AY7
6201⊕ 6189⊕ 7025⊕ 5751⊕ 6072⊕
Plate voltage V 250 250 250 250 250
Grid/cathode-voltage V -2,0 -8,5 -2,0 -3,0 -4,0
Plate current mA 10 10,5 1,2 1,1 3,0
Transconductance mA/V 5,5 2,2 1,6 1,2 1,75
Open-loop gain - 60 17 100 70 44
Internal plate resistance kΩ 11 7,7 62,5 58 25
Grid/plate capacitance pF 1,6 1,5 1,6 1,4 1,3
Max. plate voltage V 300 300 300 300 300
Max. plate power dissipation W 2,5 2,75 1,0 1,0 1,5
Table: Tube-data (typical standard values). All operational values depend on the corresponding operating point,
and are subject to individual scatter. Heater voltage: 6,3 V, heater current: 0,30...0,37A. 9-pin socket.
ECC81 ≈ 6060 ≈ 6201 ≈ 6679 ≈ 7492 ≈ 7728. ECC82 ≈ 5814 ≈ 6189 ≈ 6680 ≈ 7489 ≈ 7730.
ECC83 ≈ 5721 ≈ 6057 ≈ 6681 ≈ 7025 ≈ 7729.
⊕
This designation is supposed to indicate special tubes, see the tube manuals.

10.11 Tube Data 10-327
The following table lists old octal tubes that were in service in the amplifiers in pioneering
times (up to about the mid-1950’s).
Octal tubes 6 SC7 6 SJ7 6 SL7 6 SN7

System(s) 3+3 5 3+3 3+3
Plate voltage V 250 250 250 250
Grid/cathode voltage V -2 -3 -2 -8
Plate current mA 2 3 2.3 9
Transconductance mA/V 1.33 1.65 1.6 2.6
Open-loop gain - 70 70 20
Internal plate resistance kΩ 53 1M 44 7.7
Grid/plate capacitance pF 2 0.005 2.8 4
Max. plate voltage V 250 300 300 300
Max. plate power dissipation W 2.5 1 3.5
Socket connections:
left 9-pin socket, mid
and right octal socket
(seen from below).
Fig. 10.11.1: Output characteristics of triodes (according to data sheets).

The data books recommend the ECC83 as preamplifier tube when high amplification is
required. The equivalent to the European ECC83 is the US-made 12AX7, replaced around the
beginning of the 1960’s by the slightly improved 12AX7A. The changes related to the
maximum load (1,2 W instead of 1,0 W), to the maximum plate voltage (330 V instead of 300
V), and to the typical noise voltage referenced to the input (equivalent noise). For the first
time this is limited: (1.8 µVeff, 25 Hz – 10 kHz). The RCA Receiving Tube Manual writes
about the equivalent 7025: The 7025 is identical with 12AX7A except that it has a controlled
equivalent noise and hum characteristic; the data sheet in addition limits the maximum
interference voltage. If less amplification is needed, the recommended tube is the 5751 and
for even less gain it is the 12AY7.
The ECC81 (12AT7) and the ECC82 (12AU7) are specified to 10 mA plate current, and are
used not for preamps but for driver- and reverb-circuits. This does not exclude that an ECC81
is operated with a plate current of 1 mA – however in this case we must not expect the low
output-impedance specified for 10 mA plate current. The below table lists tube data for the
following operating conditions: plate connected to 300 V via 100 kΩ, cathode connected to
ground via 1,5 kΩ, cathode resistor bridged with a capacitor; grid at 0 V.
European designation ECC 83 - ECC 81 - ECC 82

Alternate designation 12AX7 5751 12AT7 12AY7 12AU7
Plate voltage V 195 176 140 140 85
Grid/cathode-voltage V -1,6 -1.8 -2,4 -2,4 -3,3
Plate current mA 1,05 1,2 1,6 1,6 2,2
Transconductance mA/V 1,6 1,4 1.5 1,3 0,8
Open-loop gain - 100 72 42 41 15
Internal plate resistance kΩ 70 52 27 30 18
Table: tube data; benchmarks rounded off for small plate current.
The Barkhausen-relationship should connect the transconductance S, the open-loop gain µ

and the internal resistance Ri (see chapter 10.11.4): . Checking the tables and
data sheets provide by the manufacturers shows that this relationship is often not complied
with. Deficient theory is not likely to be the reason; rather, we can surmise that this is due to
rounded-off or inaccurate values. The tube parameters given here and in the following are
those provided by the manufacturer – they are not corrected even though they may give rise
to small errors.
In Fig. 10.11.2 we see the output characteristics of commonly used double triodes. For all
diagrams, the ordinate range is 0 – 4 mA to obtain operating conditions typical for preamp
applications. For Ra = 100 kΩ, a load characteristic is included as the dashed line; it crosses
the abscissa at 300 V (operating voltage). Positive grid voltages will not occur for high-
impedance drive-signals (that are typical for guitar amplifiers); therefore the minimum plate
voltage is quite high in some cases (e.g. 90V for the ECC83).

10.11 Tube Data 10-329
The missing content in some figures (w/red

text) is reserved for the print version of this
book.
Fig. 10.11.2a: Triode-characteristics for small plate currents (taken from data sheets provided by manufacturers).
Fig. 10.11.2b: Pentode-characteristics for comparison: left Telefunken, right Svetlana.

10.11.3 Power tubes
Power (or output) tubes deliver the high power required for operation with loudspeakers.
They are larger than preamp tubes, need more heater power, and dissipate more heat. In guitar
amps, it is normal to use exclusively pentodes and tetrodes; however, in HiFi-power-
amplifiers, triodes may indeed be found.
As already elaborated in chapter 10.5.1, the maximum allowable power dissipation must not
be confused with the output power. Data sheets give information as to how much load a tube
may take; however, the approach is rather inconsistent: the various rating systems make
comparisons difficult. We find design center values, design maximum, absolute maximum,
système des limites hybrides, or simply limit data. There are hints as to what these definitions
are supposed to imply (chapter 10.5.9), and approximate conversion factors may be
discovered – but it’s still kind of flaky. For example, the Siemens data book of 1972 specifies
the “limit data” as “average limit data” which can be exceeded by a maximum of 10%. For
the table, we interpreted this as “design centre” to be able to compare to the US-nomenclature.
The limit data may also be exceeded at the expense of the operating life, and in fact this does
happen in guitar amplifiers. However, no specification can be found as to what effect e.g. an
overload of 50% will have on the durability. Consequently, the table below may only be
regarded as a rough guideline – the professional user is well advised to request data with a
binding commitment from his or her supplier.
Design center Design max. Absolute max. For example similar
EL-84 12W / 2W Philips 6BQ5
6V6-GT(A) 12W / 2W 14W / 2.2W General Electric 7184
EL-34 25W / 8W Siemens KT-77
5881 23W / 3W Tung-Sol 6L6-WGB
6L6-G (B) 19W / 2.5W 22W / 2.8W Tung-Sol
6L6-WGB 26W / 3.5W Tung-Sol 5881
6L6-GC 30W / 5W General Electric
KT-66 25W / 3.5W 30W / 4.5W Marconi, MOV
KT-88 35W / 6W 42W / 8W Genalex 6550
6L6, 6V6, 5881 EL-34 EL-84 Pin connections of the

socket (seen from below)
6550, KT-66, KT-88 KT-77 6BQ5

10.11 Tube Data 10-331
EL 34
Pentode, octal socket, 6.3 V, 1.5 A. Pa,max = 25 W, Pg2,max = 8 W.
Data sheet diagrams:
Measurement results:
The missing content in the figures (w/red text) is reserved for the print version of this book.

6L6-GC
Beam-tetrode, octal socket, 6.3 V, 0.9 A; Pa,max = 30 W, Pg2,max = 5 W.

Similar transmission values are obtained using the 5881.

10.11 Tube Data 10-333
KT-66
Beam-tetrode, octal socket, 6.3 V, 1.3 A. Pa,max = 25 W, Pg2,max = 3.5 W.

6V6-GT
Beam-tetrode, octal socket, 6.3 V, 0.45 A. Pa,max = 14 W, Pg2,max = 2.2 W.

10.11 Tube Data 10-335
EL 84
Pentode, 9-pin socket, 6.3 V, 0.76 A. Pa,max = 12 W, Pg2,max = 2 W.

KT-88
Beam-tetrode, octal socket, 6.3 V, 1.6 A. Pa,max = 35 W, Pg2,max = 6 W.

Similar transmission values are obtained using the 6550.

10.11 Tube Data 10-337
Comments
As the pictures below show, diagrams for tubes can turn out very different. Although all 4
bundles of curves supposedly characterize the KT-88, they correspond only moderately. This
may be due to the diligence (or lack thereof) when drawing the curves (there were no PC’s
back in 1956), or due to developmental progress (between 1956 and 1974). Today, we cannot
really find out how well the tubes produced 50 years ago in fact met the data sheet
specifications. Even using “new old stock” – i.e. tubes manufactured back then but becoming
operational for the first time today – is not conclusive since the long storage time might well
have changed the tube. It would be possible to determine how well tubes offered today meet
the historic specs … but normally this is not done. Inquiring with the chief technician of a big
tube supplier about a much too high residual voltage led to the counter-question: “what do
you mean by residual voltage?” – this clarified that the good man had never done any power
measurements himself.
The diagrams shown on the previous pages get their bearings from the old data sheets,
including all associated uncertainties. The measurements were done with new tubes some of
which showed considerable variance – indicating that the term “selected” tube is based on
quite inconsistent selection processes. Sometimes it is difficult not to conclude that, with
some tube suppliers, the term “selected” means little more than “glass container still in one
piece”. Indeed, that is important, as well, after all …
Four different data sheets for the KT-88 (output characteristics).

10.11.4 Tube-parameters
In triodes, the plate current Ia depends on the grid voltage Ug and the plate voltage Ua . This
bi-variant correspondence can be depicted via a pseudo-3D-graph. Due to the perspective-
related warping, additional sectional views are also required. Fig. 10.11.3 shows the context
using an idealized performance map. The sectional views of the bent “working area” are
derived first for constant plate voltage (Ua = const), second for constant grid voltage (Ug =
const), and third for constant plate current (Ia = const). In the sectional views the slope of the
curves (i.e. the partial derivative) yields the three tube parameters transconductance (S),
internal resistance (Ri) and gain (µ):
The transconductance increases (for Ua = const) with growing grid voltage; the internal
resistance decreases (for Ug = const) with growing plate voltage; the gain remains (in this
idealized example) independent of the grid voltage (for Ia = const).
Fig. 10.11.3: Tube-parameters: pseudo-3D-picture (top) with sectional views (bottom).
The theory of quadripoles would – given the two input and two output terminals – in fact
require 4 quadripole parameters. However, due to the normally negligible input current, three
are sufficient♣. Moreover, these three parameters (S, Ri, µ) are interdependent such that in the
end only 2 of them are required to describe the transmission behavior. Additionally, for some
triodes the gain µ is almost independent of the plate current, plus it is possible to calculate e.g.
the internal impedance from the transconductance:
Barkhausen formula, D = “Durchgriff”
♣
It was already shown in Chapters 10.1.3 and 10.2.2 that the gird current must not generally be ignored.

10.11 Tube Data 10-339
If we do not consider the grid current, three static parameters (Ua, Ia, Ug) and three dynamic
(or differential) parameters (S, Ri, µ) remain. The static parameters describe the behavior at
the operating point and the dynamic parameters describe the behavior at small drive levels.
Only with linearization (i.e. replacing the curved transmission characteristic by the tangent),
we can obtain a linear equivalent circuit with signal-independent components. In it, the tube is
replaced by a controlled source with internal resistance (Fig. 10.11.4):
Fig.10.11.4: Two tube-equivalent-circuits for small drive levels (small-signal-EC) with equivalent behavior.
Source voltage UQ und source current IQ are interdependent via the internal resistance (UQ =
IQ ⋅ Ri), RL is the load resistance at the plate. RL combines the total external plate load i.e. plate
resistance (from plate to supply voltage) plus in parallel the input impedance of the
subsequent stage. The AC-values UQ and IQ are “controlled” by the alternating voltage at the
grid :
The ratio of the alternating voltage at the plate and the alternating voltage at the grid
yields the alternating voltage gain vU:
Alternating voltage gain
The alternating voltage gain vU, (also called operational gain) needs to be distinguished over
the gain µ; µ is also called the open-loop gain (see the above tables). Under regular operating
conditions (i.e. with a plate load RL) the gain is smaller than the open-loop gain. Of course,
both formulas given for the calculation of vU lead to the same result. For tubes featuring a µ
almost independent of current (the ECC83 belongs to this group), the first formula would be
more conducive because with it only the internal impedance remains as current-dependent
(i.e. operation-point-dependent) variable. The larger the plate current Ia, the smaller Ri, gets
and the larger the amplification vU becomes. On the other hand, the (static) plate voltage drops
with increasing voltage, and so does the maximum possible alternating voltage at the plate.
Let us quickly repeat, just to be clear: without any drive signal we obtain the static values for
the operating point (Ua, Ia, Ug). With a drive signal, the small dynamic alternating values
are superimposed on top of (i.e. added to) the static values of the operating point.
“Plate voltage” always signifies the voltage between plate and cathode, and correspondingly
the “grid voltage” always is the voltage between grid and cathode. Nonlinear behavior
(distortion) cannot be covered via the small-signal equivalent circuit. Often, tube data sheets
merely give the three dynamic tube parameters for a single operating point that may or may
not fit. For the ECC83 we find, for example, data at Ia = 1.2 mA: a reasonable fit for typical
input stages. For the ECC81, however, the parameters in the data sheet are specified at 10
mA; this value is normally not a good match at all because preamps and intermediate stages
mostly operate with smaller currents.

Fig. 10.11.5 compares tube parameters within the range of plate currents typical for
amplifiers. Special consideration is required because: 1) the data sheets on which the
comparison is based are most often of a very small format and not precisely drawn, 2) the
nominal curves for the same types of tubes from different manufacturers are not an exact
match, and 3) there is significant production scatter. In the top right figure, the dependency of
the open-loop-gain on the current is shown, below that we see the gain at 91 kΩ (a value
resulting from connecting a 1-MΩ-pot via a coupling capacitor to a 100-kΩ-plate resistor).
The largest gain is obtained by the ECC83 (12AX7, 7025) – therefore this type is often found
in the input stages of amplifiers. Since this tube can be overdriven with more sensitive
pickups, the 12AY7 may also occasionally be found – however compared to the ECC83 the
12AY7 requires a more negative grid voltage at the same plate current. Given its parameters,
the ECC81 (12AT7) would be a suitable replacement of the ECC83; however the data sheet
does not feature the small hum and noise values as they would be necessary for input stages.
Fig.10.11.5: Comparison of tube-parameters, for 250 V plate voltage. Taken from manufacturer data sheets.
Other than from the plate current, the tube parameters also depend on the plate voltage, but
this effect is relatively weak (Fig. 10.11.6).
Fig.10.11.6: Tube parameters dependent on the plate voltage. Taken from manufacturer data sheets.

10.11 Tube Data 10-341
The two transmission-parameters transconductance S and open-loop gain µ are defined for
short circuit and open loop, respectively, at the plate. These are operating conditions that do
not appear in practice. In real circuits the plate resistor Ra interconnects the static values
plate voltage Ua, plate current Ia, and supply voltage UB: Ua = UB – Ra ⋅ Ia (for the dynamic
values see Fig. 10.11.4). In Fig. 10.11.7, the load plane – a slanted area in the 3D-
representation – intersects the characteristic area in a line (dashed in the figure), the projection
of which onto the Ug/Ua-plane below shows the dependency on the grid voltage. We do not
achieve a distinction between static and dynamic plate load yet (there’s only Ra, no coupling
capacitor, no additional load) – still: we get the whole range and thus real large-signal-
behavior. Almost, that is, since the grid current remains not considered. It is here where the
beautiful tube models find their limitations, because no data sheet tells us anything reliable
about the grid current. The latter is subject to too much scatter to be specified in the data
sheets. That is why it is not possible to reliably describe the Ug/Ua-development in the right
corner (Ug > -0.5V), why distortion models always remain limited to idealized characteristics,
and why every individual tube can sound different when overdriven. Less emphatically: that
is why tubes of the same type differ in particular in their non-linear behavior. The differences
can be very large: grid currents of tubes of the same type can vary by a factor of 20! We may
neglect the grid currents only as we drive the tube with a low-impedance signal generator in
the lab. With a high impedance source (such as a guitar pickup or a preceding tube in
common-cathode-configuration driving the tube), the individual grid current is significant.
Fig.10.11.7: Load plane for Ra = 100 kΩ (left); projection onto the Ug/Ua-plane (right).
Based on measurements, Fig. 10.11.8 shows how much the grid currents can vary. However,
the figure must not be interpreted such that e.g. tubes manufactured by Siemens would
generally have a strong grid current; another ECC83 by Siemens may well have a much
smaller grid current.
Fig.10.11.8: Grid currents for five different ECC83.

Data sheets do specify an operating point with associated transconductance. This does not
help much, however, if the tube is deployed using a different operating point, and thus there
are supplementary diagrams. For the triode, the grid and cathode define the input port, and
plate and cathode define the output port. Grid voltage and grid current are the input signals,
while plate voltage and plate current form the output signals. Fig. 10.11.09 shows a
characteristic area, selected characteristic curves (for Ug = const), and the projection of these
curves onto the right-hand boundary plane. The axes of this boundary plane represent the
output signals of the tube, and thus the curves are called “output characteristic curves”.
Fig. 10.11.09: Characteristic area and batch of

characteristic curves for a triode. The lines of constant
grid voltage are projected onto the left boundary surface
(upper right); this results in the output characteristic
curve diagram (left)
Alternatively, the curves for constant plate voltage may be projected onto the boundary area
towards the back (Fig. 10.11.10). Since in this case one of the axes belongs to the input values
while the other belongs to the output values, these characteristic curves are designated
“transmission characteristic curves”, or transfer characteristic curves. As a supplement,
further characteristic curve diagrams are customary, for example for a special plate load (see
Fig. 10-11-7).
Fig. 10.11.10: Transfer characteristic curves for constant plate voltage.

11. Loudspeakers
If you wanna play music, you gotta move some air. For the operation of the acoustic guitar, it
is predominantly the vibrating body that generates this air-movement (commonly called sound
wave), while in the framework of the electric guitar, that job is done by the loudspeaker.
That’s the dynamic loudspeaker, specifically, because other transducer types [3] are not called
into action as guitar loudspeakers. The diameters of these speakers are specified in inches (1”
= 2.54 cm). Most guitar loudspeakers sport 10” or 12”, and occasionally also 15”; in small
practice amplifiers, 8”-speakers are also common. The guitar loudspeaker is part of the overall
instrument – it is supposed to contribute to forming the sound. To put it another way: the
guitar speaker should have an atrocious frequency response, and it should distort dreadfully.
Okay, maybe not dreadfully – but at least it should distort “adequately”. Playing an electric
guitar using a HiFi-system will result in a very special sound that is not entirely unusable but
not at all reminiscent of Hendrix, Clapton, Beck and Page, either. In the typical sound of an
electric guitar that we are accustomed to, not only the guitar player takes part (indeed, that
role should never be underestimated), and not only guitar and amplifier contribute – but the
loudspeaker, as well. While this book has concentrated so far on guitar and amp, some room
shall now be also given to the loudspeaker and its cabinet.
11.1 Build and function
The principle of the dynamic transducer finds its scientific essentials in two simple linear
mappings: 1) In a magnetic field, the force acting on a wire conducting a current is F = B⋅l⋅I,
with B = magnetic flux density (induction), I = strength of the current, and
l = length of the wire. 2) Moving this wire (in the magnetic field) generates an electric voltage
across it: U = B⋅l⋅v, with v = speed of the movement. The force is termed Lorentz-force after
the Dutch physicist HENDRIK ANTOON LORENTZ (1853 – 1928), the induction voltage usually
is linked to the British scientist MICHAEL FARADAY (1791 – 1867). However, not forgotten
should be the American physicist JOSEPH HENRY (1797 – 1878) who – independently of
Faraday – described the mechanisms of induction, too.
The above-mentioned mapping between electrical quantities (U, I) and mechanical quantities
(v, F) is a linear mapping – at least as long as the system parameters B and l remain signal-
independent. The latter will of course not be the case anymore for large drive levels. Still, a
linear and time-invariant model proves a useful entry point into the description of the
transmission behavior of dynamic loudspeakers. That especially for the guitar loudspeaker
non-linearity will be essential, that the transmission not only needs to reach a single point in
space but an infinite number of these, that in the end time-invariance will not hold – all this
foreshadows how complex a model for a speaker can become if we seek to describe “all”
characteristics. So let’s not go there – the extent of a profound literature search alone would
go beyond the scope intended here. The theory presented in the following therefore is limited
to the basics, and the examples and measurement protocols given are judiciously selected but
not statistically conclusive.
© M. Zollner 2008 – 2014 Translated by Tilmann Zwicker

11-2 11. Loudspeakers
Fig. 11.1 represents a cross-section through a membrane-loudspeaker. The build variant

shown on the right is deployed for Alnico magnets (very high flux density) while the one on
the left is conducive when ceramic magnets are used – they require a larger cross-sectional
area of flux due to their not-quite-so-high flux density.
1 = air gap w/voice coil and its carrier, 2 = pole core,

3 = pole plate, 4 = centering (spider), 5 = dust cap,
6 = membrane, 7 = suspension, 8 = basket. N/S = magnet.
Fig. 11.1: Cross-section through a membrane-loudspeaker. Left: ceramic magnet; right: Alnico magnet. The
shape is largely rotationally symmetric, the ceramic magnet is disc-shaped; the Alnico magnet is cylinder-shaped.
The permanent magnet generates a radial magnetic field in the air gap, and the ring-shaped
current-flow in the voice coil has the effect of an axial drive-force on the membrane. The flux
density achievable in the air gap is rather high: typically 1 – 1.6 Tesla and occasionally just
above that. Both the law of induction and the Lorentz-force require, as a system parameter,
the product of flux density B and wire length l; this is the transducer coefficient Bl. For an 8-
Ω-loudspeaker, Bl often has a value between 10 and 20 N/A indicating that a direct current of
I = 3 A is transformed into a force of F = 30 – 60 N. 60 N will hold up a weight
corresponding to 6 kg – quite surprising given the fragility of the materials used: the
membrane is made of paper, the voice coil of thin copper wire. The geometric data of this
voice coil are: its diameter D” (usually given in inches), its axial length H, its turns number N,
its wire length l, and its wire diameter d (often termed conductor diameter). If the insulation is
included in the consideration, d increases by about 10%. The electrical coil parameter is the
resistance R, at least as long as only low frequencies are discussed. Fig. 11.2 shows, for a
two-layer winding, the dependency of wire diameter d, wire length l, and turns number N on
the voice coil diameter D” and the voice coil length H – given that the copper resistance
remains always at R = 6 Ω. For a 1.5”-coil of 10 mm length, 11 m of wire (∅ = 0.22 mm) are
required; with B = 1.5 T, this yields a transducer coefficient of Bl = 16 N/A.
Fig. 11.2: Wire diameter d (left), turns number N (middle), and wire length l (right) depending on voice coil
diameter D". DC resistance R = 6 Ω. Parameter in the family of curves: H = axial voice coil length.
Translated by Tilmann Zwicker © M. Zollner 2008 - 2014

11.1 Build and functionality 11-3
The wire-length may easily be calculated from the winding diameter and the number of turns;
however, it is the magnetically effective wire-length that is of significance to the Bl-product,
and not the geometric length. Fig. 11.3 depicts three different cases: coil-length = air-gap-
length, as well as a relatively longer and a relatively shorter variant. The magnetic field is
focused in the air gap and grows weaker towards the outside. A coil of a length equal to the
air gap (formed by the upper pole-plate) will start to leave the (reasonably) homogenous range
of the field as soon as the flowing current deflects the coil. This could formally be considered
by defining either the flux density or the coil-length as dependent on the displacement. In the
second case, the coil is longer than the air gap – here, the length of the air gap would
approximately have to serve as the magnetic coil-length. In the third example, the geometric
and the magnetic coil-length correspond. For linear operation, the cases b) and c) would have
to be chosen because they feature a coil-penetrating flux that remains approximately constant
when displacement occurs. With regard to the efficiency, a disadvantage makes itself felt in
case b) in that a part of the coil mass needs to be moved that can contribute only little force
because it is located in the weak fringe-field. For c), the whole coil is always positioned
within the strong field, but additional magnetic energy is required to generate the – little used
– fringe-field. Case a) appears to be the efficiency-optimal, as long the non-linear distortion is
not under scrutiny. Since minimizing this distortion does not get top billing for guitar
loudspeakers, the latter often feature coil-lengths that approximately correspond to the air-
gap-length. Conversely, case b) is commonly found in HiFi-speakers.
Fig. 11.3: Different voice-coils

in the air gap.
In order to obtain a large transducer coefficient Bl, flux density and wire-length need to be
large. However, because the flux-guiding pole pieces will saturate, it is not possible to
indefinitely increase the flux density. A simple solution appears to present itself for the wire-
length: large diameter of the voice-coil and/or large (effective) voice-coil-length seems
attractive. However, both these approaches cause an increase in the vibrating mass, and thus a
decrease in efficiency. On the other hand, a large transducer coefficient will increase the
motive force and therefore also the efficiency. The latter is important, but not the one single
criterion: power capacity and high-frequency behavior need to be up the desired overall
performance. The manufacturers have found their own ways to develop marketable speakers.
There is the British philosophy that guitar loudspeakers should have a membrane diameter of
12” and maximum voice-coil diameter of 2”. And then there is the approach found on the
other side of the Atlantic that demands (among other things) that nobody – and especially not
the Brits – will tell an American how to do things. And so – with a sneer of superiority – 12”-
speakers with 4”-voice-coils are produced. Nowadays, there is some restraint to dump Brit-
ware into the Boston harbor, but the stuff still somehow feels trashy. Or so advertising tells
us. Still, despite the 600-W-behemoths with the loud-n-proud 4”-voice-coil fabricated (or at
least designed) under the Stars & Stripes, Yanks (and Rebels – and those from the West-
Coast, as well) – as far as they play guitar – scour the Internet for that legendary blue British
Celestion that will take no more than a measly 15 W. Well, eight of those standing united in a
Marshall stack will easily deal with 120 W, after all. Also, if the real original blue ones are
not available anymore: allegedly, Celestion has unearthed the olde machinery and produces
original-replicas on it. In A.D. 2000, those replicants were offered at the steal of 584 Euro.
Per unit, that is. 8 x 584 = 4672 Euro … you should be able to beat that down to 4500. Then,
only be careful that your roadie – after a particularly smoky night – does not solder a mains-
cable to the newly-acquired treasure …
© M. Zollner 2008 - 2014
12"-loudspeakers are manufactured with very different voice coils: customary are diameters
between 1” and 4”, with a resulting moving mass of 25 to 75 g. Indeed, a larger voice coils is
naturally heavier – but it allows for a larger transducer coefficient, as well, and it can dissipate
more heat. In the low-frequency domain, these are already the essential parameters, while in
the higher frequency range, the voice coil will influence the partial oscillations of the
membrane (Chladni♣).
The involved quantities will be exemplified in the following: a 12”-speaker is operated at 200
Hz – this is above the resonance frequency and therefore we have mass-control, and it is
below the cutoff-frequency of the radiation – thus there is mass-loading [3]. Simplifying the
loudspeaker impedance to 8 Ω, a current of 0.35 A is required for an operation at 1 W. With a
transducer coefficient of Bl = 14 N/A we get a motive force of 5 N. This force generates, in
conjunction with the moving mass (e.g. 28 g), a membrane acceleration of a = 177 m/s2 –
mind you, that’s no less than the 18-fold gravitational pull of the earth! In fact, this is not
unusual for a loudspeaker; at full power, these values will be much higher. From the
acceleration we calculate (via integration) the membrane velocity (0.14 m/s), and another
integration yields the displacement: 0.11 mm. Since we have been using RMS-values so far,
the displacement needs to be multiplied by 1.4 to obtain the maximum displacement of 0.16
mm. Increasing the current 10-fold (to 3.5 A), the power rises from 1 W to 100 W, and the
displacement grows to 1.6 mm (given linearity). Now, before we classify the displacement as
an unproblematic quantity, let’s quickly recall that the displacement has a low-pass
characteristic (with the speaker driven from a stiff current source): reducing the frequency
will increase the displacement. With the power of two, that is! At 100 Hz we already have 6.3
mm, and at 20 Hz that would make … 16 cm. No, not really, because here the resonance
enters the game: if the loudspeaker would have its main resonance at 100 Hz, it would operate
stiffness-controlled below that frequency, with proportionality between force and spring
stiffness. But back to 200 Hz: with the membrane velocity as calculated above, we can call in
the effective membrane area (530 cm2) and the real part of the radiation impedance, and
compute the effective power radiated onto a half-space: Pak = 48 mW. Distributing this
acoustic power over a hemisphere of a radius of 1 m, a sound intensity of 7.8 mW/m2 results,
which yields a sound pressure level of L = 99 dB. This value applies to a non-beaming
radiation into a half-space.
Fig. 11.2 has already shown that, for a given DC-resistance (e.g. 6 Ω), the wire-length, the
wire-diameter and the turns-number may not be chosen independently from each other. One
of the parameters is the length of the voice coil, another is the number of layers. Fig. 11.2 was
calculated for a two-layer winding, but a four-layer winding would be possible, as well,
resulting in an increase of the wire-length and –diameter. The transducer coefficient, and
correspondingly the efficiency, would profit from the greater length. At the same time,
however, the mass that needs to be moved would increase, and a wider air-gap would be
required to contain the double-thickness winding. Increasing the width of the air-gab reduces
the magnetic flux density i.e. the transducer coefficient. To compensate for the B-decrease,
the magnet – the most expensive component of the loudspeaker – would have to be made
larger. For the power capacity, the relations are not entirely trivial, either. The power fed to
the voice coil needs to be dissipated for the most part via convection (= heat transfer) through
the coil surface. However, a four-layer winding has almost the same surface as a same-length
two-layer winding –the corresponding gain would be insubstantial. Every manufacturer needs
to find their own strategy of optimization; there are two- and four-layer coils on the market,
and even coils with rectangular wire, all in order to push for that last further bit of efficiency.
♣
Ernst Chladni (1756 – 1827), pioneer in experimental acoustics.

11.1 Build and functionality 11-5
In fact, it is quite astonishing that a wire area of 25 cm2 can withstand 200 W, and that 5 A
can flow through a thin enameled copper wire without melting it. The current capacity of
corresponding wires in a transformer amounts to 3 - 5 A/mm2 – in a loudspeaker, this value is
easily exceeded by a factor of ten. It is the current density that usually is seen as the load-
limit: current per cross-sectional surface – apparently, there is a line that should not be
crossed. If too many Amperes flow through one square-centimeter, the wire goes kaput? No,
that’s not the case. Across the wire-resistance, the current causes a voltage drop that, when
multiplied by the current, represents the absorbed power. 2.83 V ⋅ 0.35 A = 1 W, for example
(without any phase shift between U und I). Instead of the unit Watt, we may also use the unit
kilo-calory as customary in thermodynamics: 1 W = 0.86 kcal/h. If an electrical resistor is fed
with 1 W for an hour, this corresponds to an energy supply of 0,86 kcal. This energy cannot
disappear; part of it is transferred to other objects, and part of it leads to a temperature-
increase in the resistor. To enable the resistor to dissipate any caloric energy, its temperature
needs to be increased. From the temperature difference relative to the surrounding air, the
caloric energy dissipated via convection is calculated, and from the temperature difference
relative to surrounding objects the energy transferred via radiation can be determined. The
former is more important than the latter. A resistor (or in the present case: an enameled copper
wire) that cannot dissipate heat well enough will heat up strongly, and it is here where the
danger lies: if it gets too hot, it will go kaput, after all. First, the insulating lacquer and the
glue will burn, and at too high a temperature the copper will even melt (melting point is 1083
°C). Therefore, it is not the cross-sectional area of the wire that is of importance but rather the
surface of the heated object (together with further parameters). The value of the current
density thus is not an adequate parameter to estimate the power capacity. Copper traces in
printed circuit boards bear testimony to this, too: here, 200 A/mm2 are not a rarity.
The voice coil needs to pass the energy fed to it predominantly as heat; indeed the share
converted into oscillation energy (and sound) may almost be disregarded in comparison. The
flowing current heats up the voice coil which heats up the surrounding air; the latter in turn
needs to pass its caloric energy as well as at all possible to the field-focusing pole-plates. For
that reason, too (i.e. not only in order to achieve a high flux density), a narrow air gap is
advantageous. If the voice coil is longer than the air gap, the protruding part is in particular
danger to overheat, because the distance to the cooling-providing pole plates is larger. An
added extension (necessarily made of non-magnetic material, e.g. aluminum) serves well in
this case (Fig. 11.4). This extender has no bearing on the static magnetic field but it does on
the heat transfer. The dynamic magnetic field will be affected – however, this may indeed be
desirable: the eddy current induced in the extender pushes the AC-field out of the magnetic
circuit (low-pass), and decreases the non-linearity caused by the field’s modulation. Whether
a pole piece vent is helpful can only be determined in the individual case: given an airtight
dust-cap (calotte), a pump results that pumps cooling air into the air gap. However, the effect
of a non-linear spring is created also. The vent decreases the non-linearity, and the cooling
effect, as well [Klippel W., JAES Vol 52, 2004].
Fig. 11.4: Pole plate with non-

magnetic cooling extension (left),
pole-core with vent for ventilation
(right).
© M. Zollner 2008 - 2014

In operation, the voice coil gets very hot, but its material (usually copper, sometimes
aluminum) can deal with this issue quite well. Not so insulating material, glue and bobbin.
Early on in the era of loudspeakers, the voice-coil carrier was made of paper: thin and
lightweight – but not very temperature-resistant, with about 100 – 120°C being the limit for
continuous operation. Accordingly, the first 12”-speakers were specified at a power capacity
of merely 15W. As new plastics were developed, materials with higher resilience appeared,
for example Nomex (meta-aramid) consisting of polyamide fibers and enduring up to 220°C.
Kapton can withstand even higher temperatures: the manufacturer (DuPont) specifies 230°C,
but loudspeaker manufacturers readily rely on the short-term specification of up to 400°C. If
that is still not good enough: bobbins made from aluminum would take even higher
temperature loads. They did not catch on for guitar loudspeakers, however.
Kapton has proven itself as standard material in more recent loudspeakers, but Nomex and
even paper are still deployed, as well. The main reason is the sound. Manufacturers such as
Eminence attest the paper-bobbin a slightly warmer sound while Kapton allegedly produces a
somewhat more brilliant sound. Nomex supposedly gives an intermediate result. In any case,
these would not be big differences – shape and build of the membrane have a much more
considerable effect here. Eminence offers a 12”-speaker (L-122) optionally with paper- or
Kapton-bobbin, with – of course – different power capacity: 20 W and 35 W, respectively,
which is a common value for 1”-voice-coils. At the same time, Eminence also offers five
further 12”- guitar speakers, among them a 100-W-speaker with a 2”-voice-coil on a Kapton
bobbin. Options include “British” membranes, on paper- or Kapton-bobbins.
Temperature-resilience and efficiency are without doubt important features of a loudspeaker,

but the main criterion is the sound. Even if the voice-coil may have a small share in this, the
membrane (also termed diaphragm) is what takes care of the sound radiation, and it is the
component most crucial to the sound. Following simple piston-membrane theory, we have
frequency-independent power radiation between the resonance- and the cutoff-frequencies
(e.g. between 90 and 600 Hz). Above this, the radiated power drops off with 1/f 2. At low
frequencies, the speaker radiates the sound power into a half-room; from about 600 Hz,
beaming sets in, and the power decreasing with 1/f 2 is increasingly focused onto a smaller
section of the room. This piston-membrane theory holds, however, only for a rigidly
oscillating membrane not changing its shape at all. At middle and high frequencies, the real
membrane vibrates not rigidly but it “breaks up”, i.e. it vibrates in eigenmodes (standing
waves, partial oscillations). This “life of its own” of the membrane (not initially covered by
the simple theory) is undesirable for HiFi-speakers but positively welcome in guitar
loudspeakers: it does enrich the guitar sound with invigorating high-frequency interferences.
As already noted: color-free, neutral reproduction is not the objective in a guitar loudspeaker.
And so the loudspeaker designer batters up the
membrane with many a corrugation – such that it may
generate as many partial oscillations as possible up to
about 5 kHz. In Fig. 11.5, one of these circumferential
corrugations is shown. Loudspeakers made by
Celestion (a brand often used in guitar amplifiers) in
most cases include 8 corrugations; in speakers by
Jensen (another highly popular brand) we find up to 12 Fig. 11.5: Membrane with corrugation
corrugations. More details regarding membrane
oscillations are to follow in Chapter 11.3.

11.2 Electrical two-pole characteristic
By definition, every loudspeaker is an electro-acoustic transducer i.e. a two-port device with

an electrical and an acoustical port [3]. The electrical port (the two connectors) represents a
relatively complicated electrical resistor that may be described by its impedance Z. As a
rough approximation, the complex impedance Z consists of a series connection of a resistor R
(real part) and a coil-impedance pL (imaginary-part), with the inductance L and the complex
frequency p = jω. Both components result from the voice coil, a cylindrically wound copper-
or aluminum-wire positioned in the air-gap of a strong magnet and taking care of the motive
force acting on the membrane. The movement of the membrane has the effect that an
(additional) voltage is induced into the voice coil, and for this reason it is necessary to
consider, within the framework of a more precise model, the mechanical elements
transformed onto the electrical side as well. In fact, membrane-movement and –displacement
are factors of mechanical energy that cannot appear out of nowhere but have to have their
source on the electrical side of the transducer – which is why these quantities need to factor in
the electrical impedance [3].
On the mechanical side, the simplest equivalent circuit diagram (ECD) of the transducer
considers a mass (membrane incl. suspension and voice coil), a spring (membrane-
suspension), and also a friction resistance modeling the energy losses due to deformation of
membrane and suspension. The loading by the radiation impedance may be neglected in the
simple model. In Fig. 11.6, the frequency responses of the impedance of two typical 12”-
speakers (not mounted in any cabinet) are shown.
Fig. 11.6: Frequency response (magnitude) of the electr. impedance; left: Celestion Blue, right Eminence L122.
Both frequency responses include a characteristic maximum at low frequencies: together with
the spring stiffness s, the mass m forms a velocity-resonance that generates a large counter-
active voltage via the transducer-coupling (U = α⋅v, [3]): the current decreases, the
loudspeaker is of high impedance at this frequency. For most guitar loudspeakers, this
resonance is in the range of about 70 - 100 Hz; for bass speakers it will be somewhat lower.
In the impedance-increase at high frequencies, we can recognize the inductive component of
the voice coil; however, it is not a simple, frequency-proportional increase but a flatter one.
This is due to the fact that it is the magnetic circuit that causes a considerable share of the
voice-coil inductance, and in this circuit we find induced eddy-currents that cause a -
characteristic. For this reason it is not possible to model (in a more exact approach) the
inductive increase with a single inductance; rather, we require an RL-network. Given less
requirements, a single inductance will suffice; this is often set to 1 mH. The small impedance
fluctuations around 1 kHz result from partial oscillations of the membrane, i.e. standing
waves that preclude the membrane from maintaining its shape. In HiFi-speakers, designers
seek to suppress this kind of behavior – conversely, it is not undesired in guitar loudspeakers.
© M. Zollner 2008 - 2014

Fig. 11.7 depicts an equivalent circuit for a loudspeaker-impedance. The resistor designated
with RCu represents the ohmic voice-coil resistance while the LR-array generates the high-
frequency increase of the impedance. The parallel-circuit models the three mechanical
elements of the membrane. If needed, this circuit may be extended or modified without great
effort. At resonance, the impedance of the mechanical membrane-resonator is purely ohmic
(W), and it is mapped with (Bl)2 onto the corresponding (ohmic) resistor of the parallel circuit:
RW = (Bl)2 / W. Herein, Bl is the transducer coefficient based on the magnetic flow density B
and the length of the voice-coil wire l. Therefore, the resonance-maximum of the loudspeaker
impedance is determined mainly be two parameters: the membrane dampening and the
transducer coefficient. For this reason, high-value resistances at resonance are often found in
speakers with strong magnets.
Fig. 11.7: Frequency response of the impedance, and schematic of an equivalent circuit for a loudspeaker [3].
As already mentioned, the membrane movement induces a counteracting voltage, and

therefore in a more exact model, special attention needs to be paid to the radiation impedance.
At low frequencies, the membrane is predominantly loaded by the co-vibrating mass of the
air – this will amount to about 7 g for a 12”-speaker (operated without baffle). In absolute
terms, that is not much, but it is of considerable magnitude relative to the membrane mass (20
– 50 g). Changing the mounting conditions (baffle, enclosure), this air mass will also vary and
detune the resonance (Fig. 11.8) Merely adding a baffle will have not much of an effect (the
air-mass approx. doubles), but mounting the speaker in an enclosure considerably modifies
the impedance. Of course, not only the impedance changes – the behavior of the radiation
will vary drastically, too. In principle, every change in the electro-acoustical efficiency needs
to find its match in the frequency response of the electrical impedance. However, in practice
this will, especially in the high-frequency range, not be noticeable because the corresponding
changes in the radiation impedance become small compared the mass of the membrane.
Moreover, the ohmic resistance of the voice coil will see to it that these small load-variations
are practically invisible in the frequency response of the impedance.
Fig. 11.8: Impedance: loudspeaker without (----) and with baffle (–––); right: with and without open housing.

We can easily derive the most important membrane parameters from the electrical frequency
response of the impedance – this is without any mechanical measurements. With s and m, the
resonator has two degrees of freedom but only a single known quantity: fRes. However,
detuning of the resonance by applying a small additional mass to the membrane yields two
further known quantities and only one additional unknown variable. The system therefore has
a solution [3]. In practice, difficulties may be encountered, though: for example if – due to a
large dust-cap – a relatively big mass-ring needs to be laid onto the membrane. In this case, it
may be that the membrane-stiffness between voice coil and additional mass already has
disturbing effect such that the frequency response of the impedance does not merely show a
detuned maximum but two maxima. This scenario requires an extension of the equivalent
circuit diagram. It may also help to work with two additional masses. The typical membrane
mass of a 12”-speaker will be in the order of 20 – 50 g, typical stiffness will be about 5 - 10
kN/m (without the stiffness of the air inside an enclosure) – in singular cases a bit more.
To determine the transducer coefficient (Bl), measuring a transmission-quantity is necessary.

The membrane-acceleration can be ascertained relatively easily: if is only even
slightly above the earth’s gravitational pull, small particles (e.g. sand) set on top of the
membrane will start to dance. Typical transducer coefficients are found to be in the range of
Bl = 10 – 20 N/A.
As the figures presented so far show, the DC-resistance of an 8-Ω.speaker is not actually 8 Ω
but less: about 6 – 7 Ω may be seen as customary. This is at room temperature! In operation,
the voice coil heats up to above 200°C under certain conditions, and the resistance rises
correspondingly by up to 80% (for example from 6.5 Ω to 12 Ω). If the speaker is operated
from a stiff voltage source, the power taken in by the loudspeaker decreases be a third, as does
the radiated sound♣! Likewise, with a tube amplifier having no negative feedback (that in
principle is similar to a current source) the received power will drop, as well, if the amplifier
is pushed to the drive limit. This volume-drop caused by the heating-up of high-power
loudspeakers is system-immanent – undesirable but unavoidable. For ceramic magnets, a
further effect may manifest itself: their flux density may noticeable drop off with rising
temperature. Alnico magnets show this behavior only at temperatures that considerably higher
than the operating range of guitar loudspeakers; the flux density of these magnets is
practically independent of temperature.
It is understood that an amplifier needs to feature stable operation (i.e. no RF-oscillations) not
just with an ohmic nominal resistance but with a complex speaker load, as well. Therefore,
measurements with a real loudspeaker loading need to be taken in fact not just because
otherwise any instability would not be noticed, but because only that way the typical output
signals occur. Irrespective of whether we have operation with a stiff voltage source or a stiff
current source, the electrical impedance of a loudspeaker is crucial for its transmission
behavior. The power fed from an amplifier is dependent on the actual loudspeaker impedance,
and the nominal value (e.g. 8 Ω) only offers an orientation value. Combined with tube
amplifiers with their transformer coupling at the output, we get a particularly complicated
system with non-linear source- and load-impedances. Swapping the loudspeaker may cause
considerable changes in the transmission behavior especially around 100 Hz – these changes
are caused already at the interface output-transformer/loudspeaker. Further contributions are
made by the radiation characteristics of the individual loudspeaker.
♣
The exact value will depend on the internal impedance of the power supply.
© M. Zollner 2008 - 2014

Fig. 11.9: Frequency response of the impedance of 8-Ω-speakers. Upper left: Celestion, upper right: Jensen.
Center left: Eminence, center right: 12"-loudspeakers with resonance frequencies below 70 Hz.
Lower left: 2-way-speaker (Canton, 8 Ω), lower right: 3-way speakers (Canton, 4Ω).
In Fig. 11.9 the frequency responses of the impedances of a number of 12”-speakers are
shown. All measurements were taken in the anechoic chamber and with un-mounted speakers
(i.e. without enclosure). The curves are in principle similar but differences show up in the
details. The lower two diagrams show a comparison to HiFi-speakers. All impedance curves
were taken with low voltage i.e. in the linear range. Chapter 11.6 will discuss that the
voltage/current correspondence may be non-linear, as well.

11.3 Frequency response of sound pressure level 11-11
11.3 Frequency response of sound pressure level
A linear and time-invariant system can unambiguously be described by its magnitude- and
phase-characteristics. However, if we take the transmission behavior of a loudspeaker to be
approximately linear and time-invariant (for reasonable drive levels this assumption is
certainly allowable), one single magnitude/phase characteristic is completely inadequate. This
is because the loudspeaker is not an electrical two-port! While it does include an electrical
input-port (the connectors), at its output it radiates a special field formed by sound pressure (a
scalar) and particle-velocity (a vector). Both these quantities are location-dependent within
the three-dimensional space, and thus an indefinite number of transmission functions exists.
In order to still handle this issue in a reasonably manageable way, we limit transmission
behavior to special cases (subsets): the analysis of the frequency response in a single
direction, and/or analysis of directionality at a single frequency. In particular, measurements
of the frequency response “on axis” (i.e. with the microphone positioned centrally ahead of
the speaker) belong to the former group; the latter group includes directional (polar) diagrams.
Trying to appreciate all details will render the frequency response of a loudspeaker infinitely
complicated; therefore a rigorous simplification is called for. Starting point for many
observations is a loudspeaker mounted in a very large baffle, and with a membrane that is
simplified to a flush plate (a so-called piston-diaphragm) [3] to begin with. Assuming linear
behavior, the current is proportionally mapped into a force acting onto the membrane and
moving it. The spring-like membrane-suspension and the mass of the membrane in
conjunction form a resonator with a pole frequency at 70 – 110 Hz. Below this pole- (or
resonance-) frequency, the membrane acts approximately like a spring, and above it acts like a
mass. Alternatively, we may say that below resonance, the membrane is spring-controlled,
and above resonance, it is mass-controlled. Given a sinusoidal current, the three movement-
quantities displacement, velocity and acceleration are generated; they can be converted into
each other via differentiation or integration. Since the membrane is mass-controlled above the
resonance frequency, a stiff current source will imprint the acceleration in the corresponding
frequency range (Newton: F = m⋅a). With the linear model, it is no problem that loudspeakers
are not always driven from a stiff-current source: the electrical impedance links voltage and
current.
Integration of the membrane acceleration yields the membrane velocity from which – using
the real part of the radiation impedance – the radiated effective sound power may be
calculated [3]. In the simple model, this effective sound power is frequency-independent
between the resonance frequency and the upper cutoff frequency. The latter is at about 600
Hz for a 12”-speaker; above that, the radiated power decreases with 1/f 2. Or so the simple
theory says. The frequency responses measured on-axis do show that your typical guitar
speaker will radiate frequencies up to 5 kHz with a rather decent level – only above this limit,
the frequency response drops off quite abruptly. This is, however, no contradiction to the
theory, because sound-level and sound-power are not equivalent: upwards of 600 Hz, the
radiated power decreases, but beaming-effects focus it increasingly to the area in front of the
membrane. In fact, power-decrease and beaming compensate each other in the simple model
such that on-axis there is no high-frequency drop-off at all. Still, this is where grave
differences between theory and practice become visible: the real membrane deviates
particularly in the high-frequency range from the idealizing theory. While the theory of the
axially oscillating piston-diaphragm requires a rigid-shape membrane, the real membrane
shows partial oscillations changing the shape: it “breaks up” and forms nodal lines with
partial areas radiating in opposite phase.
© M. Zollner 2008 - 2014

Fig. 11.10 shows measurements taken with a loudspeaker installed in a baffle. This was not
an infinite baffle as required by the theory of piston diaphragms, but a square baffle of 3m by
3m, or a circular baffle of 1 m diameter. Its finite size has the effect of a diffraction wave
generated at the rim that reaches the microphone and superimposes itself with the direct sound
wave radiated by the loudspeaker. As a result, interferences appear i.e. frequency-dependent
amplifications (same-phase superposition) and cancellations (opposite-phase superposition) in
the sound pressure. For the circular baffle the distance of all points on the rim to the
membrane center is equal – a pronounced comb filtering occurs. For the square baffle, the
path-lengths of the sound wave (around the baffle) are dependent on the direction, and also
the wave diffracted around the baffle needs to pass a longer distance: its amplitude therefore
is much smaller than that of the direct sound wave, and the interferences are much less
distinct.
Fig. 11.10: Frequency response of a 12”-loudspeaker installed in a baffle. Microphone position: 0.5 m from the
speaker (axially). Left: baffle of 3m x 3m. Right: circular baffle ∅ = 1m. Theoretical interference (----).
In the simple model, two opposite-phase half-spherical waves are radiated on the two sides of
the baffle (Fig. 11.11). As the wave front reaches the rim of the baffle, its shape changes
because now a diffraction wave enters the space behind the baffle. This diffraction has the
character of a low-pass: low-frequency sound runs around the baffle without significant
attenuation but with increasing frequency the amplitude of the diffraction wave diminishes
such that in the high-frequency range only the primary sound dominates – no interference
effect remains.
Fig. 11.11: Generation of an opposite-phase diffraction wave at the rim of the baffle. The dot above the baffle
designates the position of the microphone; the two opposite-phase diffraction waves follow the primary wave.
In Fig. 11.11 we see the wave at four subsequent points in time. In the second picture, the
primary wave just reaches the microphone. In the third picture the wave has reached a bit
beyond the baffle, and in the fourth picture the opposite-phase diffraction wave reaches the
microphone.

As Fig. 11.10 has shown, the large baffle prevents an acoustic short between the opposite-
phase sound waves radiated by the front and the rear of the membrane – however, this
approach is not actually stage-worthy. Alternatively, the propagation of the wave radiated
from the rear may also be stopped via mounting the loudspeaker in an airtight enclosure. That
will have three main effects: 1) The radiation of the wave from the rear is stopped, 2) in the
enclosure, resonances occur that influence the membrane oscillation and thus the sound
radiated from the front, 3) the stiffness of the air encased in the enclosure increases the
frequency of the main-resonance. Before we look into the specifics of enclosures, we first still
need to investigate the frequency response measure with baffle-mounting in more detail.
According to the theory of piston-diaphragms, the SPL measured on axis rises with a slope of
40 dB/decade up to the resonance (e.g. 100 Hz), and remains at a constant level above the
resonance frequency. We have already seen from Fig. 11.10 that reality does not reflect this:
from 1.5 kHz, ripples cannot be overlooked anymore, and from 5 kHz, the curve takes a
nosedive. The reason for these deviations from the idealizing theory are partial oscillations
of the membrane; the latter indeed does not manage to rigidly keep its shape but develops a
position-dependent pattern of oscillation. Fig. 11.12 depicts a cutaway view of a typical
loudspeaker membrane. From the cylindrical voice-coil bobbin (in the picture at the bottom),
the slightly curved membrane extends, with the dust-cap glued to it a few millimeters out. The
upper half of the membrane includes circumferential corrugations representing a mechanical
filter designed to decouple the peripheral parts of the membrane at high frequencies. At the
positions indicated by numbers, the axial membrane velocity was measured dependent on the
frequency using a laser-vibrometer – see Fig. 11.13.
The analysis of the velocity shows that only in the frequency range up to about 300 Hz, the
membrane manages to maintain its shape rigidly. In this frequency range, the frequency
response of the velocity follows the theoretical band-pass curve. At higher frequencies, a vast
variety of eigen-oscillations of the membrane show up.
Fig. 11.13: Frequency response of the membrane-velocity at various locations; radial (left), circular (right).
© M. Zollner 2008 - 2014

Particularly striking, however, is the fact that the membrane corrugations actually do not form
a band-pass, after all! In the fringe areas, the membrane does not at all vibrate less compared
to close to the centre – rather contrary is the case: the rim vibrates more strongly. The low-
pass theory is quite old and stems from a time when it was not possibly to do an on-the-fly
quickie-scanning of the membrane with a laser vibrometer. It is easily imaginable that the
loudspeakers investigated back in the day with simple methods had such efficient
corrugations that the effective membrane-diameter indeed became smaller with rising
frequency – as it was desirable in order to optimize beaming and efficiency of the speaker.
For the loudspeaker investigated here, however, a multitude of relatively weakly dampened
eigen-oscillations are created, the amplitude of which is larger than that of the actuation. Two
each of the frequency responses of the velocity from Fig. 11.13 are shown in Fig. 11.14: one
for a measuring point at the glue-seam of the dust cap (----), and another one for a measuring
point close to the rim. Neither for the Celestion speaker (with 8 corrugations) nor for the Fane
speaker (smooth membrane), a low-pass filtering is evident.
Fig. 11.14: Comparison of the membrane velocities: close to the center (----), close to the rim (––––).
Comparing the two loudspeakers, it seems not far-fetched to assume that in fact the
corrugated membrane may even be resonance-happier than the even one. That would not
actually be a big surprise: every movement actuated by the voice coil (or the magnetic force)
starts at the inner rim of the membrane and propagates across the latter as a bending wave.
Any change in the wave-impedance – as it is introduced by the corrugations or at the rim –
creates reflections. In the end, a multitude of primary and reflected waves run across the
membrane. In specific membrane-areas, many waves superimpose with the same phase
leading to particularly strong oscillations (anti-node), while in other areas the waves cancel
each other out to a large degree, resulting in nodes in the vibration (nodal lines). These nodal
lines may have the form of concentric circles – as we measure along a radial line, this would
be captured as a minimum (Fig. 11.13, left section). However, the nodal lines may also run on
a radial course, which would require a circular measuring path (Fig. 11.13, left section).
Fig. 11.15:
Vibration nodes
of the membrane
In Fig. 11.15 we see a few typical patters of the membrane vibration. The left-hand picture
stands for a membrane rigidly maintaining its shape: all points move in the same direction. In
the second picture, a nodal line separates the right and left halves: while the point on one half
vibrate in one directions, the points on the other half move in the opposite direction. This
standing wave does not need to be fully distinct – an additional traveling wave may well be
superimposed. The other pictures show vibration nodes of increasing complexity as it may
well occur already at frequencies as low as 1 kHz.

The multitude of maxima and minima shown in Fig. 11.13, and also their extreme dependency
on the location, proves that in the middle- and high-frequency ranges many different modes
come into existence the exact calculation and verification of which was not intended as the
subject of the present investigations. A more precise analysis was done only for the G12-M –
in this speaker, a location dependent membrane movement occurs already at 300 – 400 Hz.
The reasons are probably two 21-modes. Fig. 11.16 illustrates the vibrations that occur at two
relatively close frequencies. As outlined clearly by Fleischer♣ in 1994, this behavior is often
found in approximately rotation-symmetrical structures. The eigen-values of anti-symmetric
vibration always occur in pairs for the ideally rotation-symmetric shape (e.g. 21-mode). For
approximate rotation-symmetry they break down into two different values with two
corresponding, slightly different eigen-frequencies. The corresponding eigen-shapes are of
equal type but differ in the angular position of their node-diameter, as shown by Fig. 11.16:
the eigen-shapes occurring at 350 Hz and 374 Hz are shifted relative to each other by 45°.
Fig. 11.16: Vibration modes of the Celestion-membrane. The mode shown on the left occurs at 350 Hz, the one
shown on the right at 374 Hz. These two modes are the ones of lowest frequency for this 12”-membrane.
Theory and practice concur in that the membrane vibrates – however, it vibrates in such
diverse fashions that refining the theoretical models could not be a subject of the presently
planned investigations. Therefore, practical measurements were conducted in the anechoic
chamber (AEC), generally at a distance of 3 m, with 2.83 V (for an 8-Ω-loudspeaker) fed
from a stiff voltage source, or in the reverberation chamber (RC), also using a stiff voltage
source (pink noise, 2.83 V per third-octave for an 8-Ω-speaker). For the first measurements, a
12”-speaker was mounted in a small wooden enclosure (39x39x25 cm3) and a somewhat
larger wooden enclosure (39x75x25 cm3). Fig. 11.17 shows the corresponding frequency
responses of the impedance: as expected, the additional stiffness of the enclosed air increases
the frequency of the main resonance. The corresponding effect is relatively strong for the
small enclosure and less pronounced fort the larger enclosure.
Fig. 11.17: G12-M, impedance-frequency-response, sealed enclosure (“mit Gehäuse”) 39x39x25 cm3 (left),
39x75x25 cm3 (right). “Ohne Gehäuse”: without enclosure.
♣
H. Fleischer: Spinning Modes. Research report UniBW Munich, ISSN 0944-6001.
© M. Zollner 2008 - 2014

The increase in the resonance frequency amounts to slightly more than 41% for the smaller
enclosure and somewhat less in the larger one. Consequently, the stiffness of the air is a little
larger than the membrane-stiffness♣ for the former case and a little less in the latter. For
adiabatic change, the stiffness of the air is sL = 1.4⋅105 Pa ⋅ S2 / V. In this formula, S is the
effective membrane area, and V stands for the net-volume of the enclosure. From the effective
membrane mass m and the overall stiffness s´ = sL + sM, the resonance frequency is
calculated: . Mounting the speaker in a sealed enclosure will, however, not
only shift the resonance frequency towards higher values but also generate a secondary
maximum at about 45 Hz that can be traced to leaks. This "leakage-resonance" (as the
secondary maximum is often called) stems from the mass of the air moving in the fissures,
and the air-stiffness s. And of course, the co-vibrating membrane will – strictly speaking –
also contribute. In a completely airtight enclosure, the leakage resonance should disappear.
Should it, really? Not necessarily, because that would take an airtight loudspeaker, as well.
Any ventilation hole will change the leakage-resonance, too.
Further maxima in the impedance curve are visible above the main resonance, for example at
250 Hz for the larger enclosure. They may be attributed to cavity resonances appearing due to
reflections occurring within the enclosure (standing waves, Chapter 11.8). In the higher
frequency range (above about 1 kHz), the enclosure loses any influence on the electrical
impedance; the rising value of the latter has its main source in the voice-coil inductance.
Fig. 11.7 clearly showed an impact of the enclosure (a sealed box) on the frequency response
of the impedance; however, essential for the sound is the frequency response of the sound
pressure level (SPL). In this context, Fig. 11.8 shows the differences between mounting the
speaker to a baffle, and mounting it in an enclosure. Two characteristics stand out: the
enclosure is unable to easily radiate sound in the bass range, and it generates a series of
resonance-peaks in the range of 200 – 2000 Hz that can be traced to standing waves. This is
particularly evident in the right-hand picture: 240 Hz matches the wavelength of 1.43 m, thus
half the wavelength fits exactly into the enclosure (internal length is 72 cm). The peaks at 800
Hz found for both enclosures fits the depth of the enclosure (internal spacing is 21 cm).
Fig. 11.18: 12"-speaker. Left: baffle vs. 39x39x25-box; right: baffle vs. 39x75x25-box.
Cavity resonances can be fought with a tried and tested remedy that no HiFi-box can do
without: dampening material, e.g. quilting cotton, or glass wool, or mineral wool. Standing
waves are effectively dampened by loosely filling the enclosure with it, and the frequency
response becomes more even. There is, however, a loss of efficiency that is undesirable for
guitar-speakers, and any padding is usually dispensed with here. In contrast to its acoustic
cousin, the electric guitar has no adequate body that would take care of introducing cavity
resonances, and therefore loudspeaker resonances are indeed rather welcome.
♣
This term always actually refers stiffness of the membrane suspension.

Fig. 11.19 shows frequency responses of enclosures with dampening fitted in the form of
porous absorbers. The latter represent a real load-impedance to the membrane and transform
effective power into heat. This is not desired, since corresponding dampening of the
membrane from the rear reduces its movement and thus also the radiation of sound.
Fig. 11.19: Two different loudspeakers mounted in enclosures with (–––) and without (––––) absorber.
Besides absorption, the opening up the box is another possibility to reduce resonance-effects.
In the un-dampened enclosure (closed box), the sound waves generated by the rear of the
membrane are efficiently reflected back and forth; standing waves of high Q-factor can
manifest themselves. In a box open towards the rear a large part of the sound energy
generated by the rear of the membrane leaves the box after only a few reflections. Desirable
side effect: both sides of the membrane contribute to the sound arriving at the listener’s
location. Undesirable side effect: ditto. That is because of course the two involved sound
waves will not generally superimpose on each other with the same phase, and destructive
interference (cancellation) is bound to occur, as well. The membrane acts as a dipole: as one
side generates a positive pressure, the other side will generate a negative pressure. Still, the
same happens for the vented enclosure (bass-reflex box), and that does work quite well. The
reason is that phase shifts [e.g. 3] are introduced via acoustical filters and different-length
travel paths of the sound wave. For sound reproduction of very low frequencies, the open box
certainly is sub-optimal – in this frequency range, the sound waves generated by the font and
the rear of the membrane, respectively, will cancel each other out to a large degree. In the
closed box, this cancellation is prevented – but problems in the very low frequency range still
appear due to the high air-stiffness increasing the resonance frequency. Luckily, very low
frequencies are not that important for the guitar and often even unwelcome. The first guitar
amplifiers thus were of the open-box “combo” design – to this day a tried and trusted variant.
Fig. 11.20 shows frequency responses with an open box. The way the sound is guided
increases the rear mass loading of the membrane and the resonance frequency decreases
slightly. Cavity resonances are still present but more strongly attenuated than in Fig. 11.19.
Fig. 11.20: G12-M, frequency responses of the impedance; open box (“offenes Gehäuse”) 39x39x25 cm3 (left),
39x75x25 cm3 (right). “Ohne Gehäuse” means “without enclosure”.
© M. Zollner 2008 - 2014

Fig. 11.21 depicts the SPL frequency response relating to Fig. 11.20. Compared to the
reproduction using a baffle, the ripples clearly increase but with a different characteristic
compared to a sealed enclosure: they are less narrow-band but more global and come in
broader arches. The figures in the second row hold information on the sound power radiated
in the diffuse sound field: from 200 Hz and 160 Hz, respectively, the open cabinets radiate
more sound; only in the frequency range below, selective attenuation occurs. Conclusion:
compared to the closed cabinet, the open-cabinet design is louder but also somewhat weaker
in the bass. Again, it remains a matter of taste, which one you prefer.
Fig. 11.21: Top: baffle vs. 39x39x25-open-cabinet (left) and 39x75x25-open-cabinet (right).
The lower row shows frequ. responses in the diffuse field, w/rear panel of the cabinet (––) and w/out (---).
One could argue that in modern times with super-powerful signal processors, the frequency
response of the loudspeaker is insignificant because any desirable frequency response may be
“designed” with a few rows of program code. Again, the guitar amplifier breaks rule: if
power-amp distortion is favored (as it is by many guitarists), digital filtering is not possibly
anymore. The loudspeaker directly follows the power amplifier, and – as irrevocably
postulated by systems theory – the sequence of circuit sections may not be changed in non-
linear systems. Only the loudspeaker and its cabinet can filter the signals generated by the
power amp, after the speaker there is only the space … the final frontier. The speaker, or
rather the membrane, filters mechanically, and the cabinet acoustically – and not
insignificantly, either. In the dimensions of the loudspeaker cabinet, the designer has effective
parameters at his/her disposal to kick the frequency response into shape one last time – after
that the sound leaves the production area. Presumably, the size of the loudspeaker that had to
be accommodated was the main criterion for the dimensions of the first guitar combos, and
even for Jim Marshall’s 4x12-cabinet, that was no different: the cabinet primarily served as
mount and protection. Acoustic filter design came later – if at all. Maybe it was a happy
chance that the dimensions of the now legendary small combos were not far from the
dimension of an acoustic guitar. The shape of a cavity determines the cavity resonances, and
what sounds good in a guitar may help to arrive at the right sound color in a speaker cabinet,
as well.

The lowest body resonance (the so-called Helmholtz-resonance♣) of the acoustic guitar is
located between the notes of F#2 and A2 i.e. at 92 – 110 Hz. Incidentally, that is exactly the
range where most guitar loudspeakers have their main resonance – unless you mount them
into a small, sealed enclosure. The latter may push the resonance up to 160 Hz (see Fig.
11.17) corresponding already almost to an E – not the E2 of the low E-string but the E3 one
octave higher. If we now would combine such a cabinet (“tuned high”) with one of the
legendary amp-forefathers (e.g. a Tweed Deluxe or an AC-15), we would obtain entirely
different frequency responses than those shown on Fig. 11.18. These early amps had no
negative feedback (NFB) in their tube power amps, and therefore they featured a rather
special internal impedance: within the small-signal range, the terms “stiff current-source” is
almost appropriate, while in overdrive conditions (clipping), they form almost a stiff voltage-
source. All SPL frequency responses presented so far in this chapter had been measured using
a stiff voltage-source; switching to a stiff current-source (imprinted current), the frequency
response of the impedance multiplies onto the transmission factor. For example: if the
impedance at 160 Hz rises from 7 Ω to 50 Ω, the SPL will increase by 17 dB! Not all (tube)
power amps dispense with negative feedback: in Fender amplifiers, for example, NFB
becomes a standard circuit feature from the 1960’s. VOX, however, does not follow that
route, and to this day the AC-30 does not have NFB. Power amps without NFB feature high
output impedance with a value of 200 Ω easily reached. Introducing NFB will decrease the
internal impedance – but not down to zero. For one, a high NFB-factor will decrease the gain
(which is a precious commodity in tube amps), and second, phase-shifts may quickly lead to
instability. Therefore even a tube amp with NFB may easily have an internal impedance of 20
Ω – which would, in the case of the above example, not lead to a resonance boost of 17 dB
but still to one of 9 dB.
Abb. 11.22: SPL frequency response with imprinted current, 39x39x25-box; closed (left), open (right).
Fig. 11.22 shows, for the small speaker box, frequency responses resulting from driving it via
a stiff current-source (imprinted current). In this mode of operation, any trace of a weak
bass-response has disappeared in the open cabinet; the resonance frequency (lower than with
the closed box) takes care of the required low-frequency-boost. In the linear range, that is,
since internal impedance of the power amp becomes lower as the drive level increases. In
addition, non-linearity in the output transformer makes for a rather complicated signal-
shaping. Here, there is room for the developer to design – based on the combination power-
amp/transformer/speaker/cabinet – a convincing product the characteristics of which surely
are not describable with a few diagrams. Measuring frequency responses will help to
document transmission functions – no more, no less. The final decision happens with
listening/playing tests – and those are not done in the anechoic chamber. Not to forget: the
eyes “listen”, as well! Not unheard of is the combo that did not pass the final test in the music
store because it had the “wrong” name on the front cover …
♣
This resonance is not only defined by the cavity, but also by any co-vibrating walls of the enclosure.
© M. Zollner 2008 - 2014

The loudspeaker housing alone offers many design possibilities – an impression of the
diversity is found in Fig. 11.23. A Celestion G12-M was mounted in 5 different typical
speaker boxes with the rear wall being either open, or half-open, or closed. The various peaks
occurring at different frequencies (depending on the cabinet) are the result of geometry-
specific cavity resonances. In comparison, the type of wood used for the cabinet does not play
any role as long as the construction is not untypically fragile.
Fig. 11.23: SPL frequency response using a stiff voltage source, AEC, on axis; various boxes, 1W/1m.
On the right the corresponding frequency responses of the impedance are depicted.
To discuss DSP-filtering again: of course, it would be possible to approximate the shown

frequency response via software. However, the amplifier/loudspeaker-interface connects two
non-linear, interacting systems – a simple pole-zero design will not get you far in that context.
And not to forget: the loudspeaker filters direction-dependent – something a modeling amp
fitted with a DSP is not able to simulate. The filtering calculated in the DSP effects all
radiation directions in the same way while every speaker cabinet will have its geometry-
specific directionality (Chapter 11.4).
In order to achieve clear resonance effects, the two speaker cabinets used for Fig. 11.21 were
deliberately built with special dimensions – they are, however, not entirely typical for the
genre. For this reason, the following measurement results were taken with a VOX-cabinet. No
the one of an AC-30 because there, two loudspeakers cause interference, but the cabinet of an
AD60-VT – the modern housing for a VOX-typical 12”-Celestion. Celestion has been the
purveyor to the court of VOX since the late 1950’s, despite all attempts by Goodmans and
Fane. In this AD60-VT-housing, the following speakers were mounted: G12-80, G12-M,
G12-H, G12-S, Vintage-30, G12-Century, Celestion Blue, and the original speaker of that
amp. Fig. 11.24 shows the measured frequency responses – again referenced to 1W/1m.
Fig. 11.24: SPL frequency responses, various Celestion-12"-speakers in the AD60-VT-cabinet; 1W / 1m.

The curves shown in Fig. 11.24 share a lot of commonalities. There are, however, also
selective divergences that, in the relevant frequency range, do exceed 5 dB here and there.
These are very different loudspeakers, after all, with a power capacity of between 15 W and
80 W, and a price range from 127 Euro to 584 Euro (this was in A.D. 2000, and apparently
they were serious regarding the latter price). Details are shown in Fig. 11.25 – and suddenly
we are not quite sure anymore whether the same speaker is not erroneously included twice.
But no, these are all different speakers, and closely inspecting the peaks reveal the deviations.
The latter justify the whole effort – there must be a reason why Celestion builds so many
different 12”-speakers. The term many might by misleading here, because this small excerpt
represents merely a fraction of the allegedly much more than 100 different variants. If the
Vintage-30 is not to your liking, just get yourself the Celestion Blue: the price has dropped to
a yummy 349 Euro by now Translator’s note: that’s in 2008, in 2018 it was 279 Euro street price .
Fig. 11.25: Comparison of different Celestion 12”-speakers. AEC, AD60-VT-cabinet, 1W/1m.
In the price list from A.D. 2000 mentioned above, the “Blue” sets you back four times the
financial damage the Vintage-30 would do. That makes sense somehow, since the Vintage-30
has four times the power capacity of the “Blue”. Sales-math – it can be so simple: 155 Euro
for 60 W, and 584 Euro for 15 W. That justifies a closer look at these two candidates: indeed,
there are differences besides many similarities (Fig. 11.26) – but stop: there is (in the right-
hand picture) another competitor in the race that features a similar response curve.
Fig. 11.26: Celestion "Blue" (–––), compared to Celestion Vintage-30. AEC, AD60-VT-cabinet, 1W /1m.
© M. Zollner 2008 - 2014

We can see from the lower-most curve in the left-hand picture that the magnitude differences
between the “Blue” and the Vintage-30 are mostly smaller than 2 dB; larger deviations are
found at only one single place. In the curves shown in the right-hand picture, the maximum
differences are smaller although the average square deviation is in fact even a bit bigger than
in the picture on the left. Which speaker is that? O.k. – here we go … From the point of view
of the manufacturer, it may seem outrageous that, despite the deterringly high price,
somebody goes out and buys no less that two specimen of that blue Celestion … and
compares them to one another. Well, it was simply too appealing to miss. Right: 2 specimen
are of course not the quantity that you would need for a reliable variance-analysis, but lets still
cut to the chase (without safe statistical base): according to the present measurements, the
differences between a Celestion “Blue” and a Vintage-30 lie in the same range as the
differences between two Celestion “Blue”. The differences between the “Blue” and the
Vintage-30 are just about noticeable – but the same holds for the differences between two
“Blue”. If the sound pressure levels of two randomly selected Celestion “Blue” differ already
by ±3 dB, it must be assumed that there will be even larger tolerances across the whole “hand-
built series”. With this, the statement “the Vintage-30 sound more mid-rangy than the blue
Celestion” becomes untenable. Broadening the term intra-individual from the individual to
the same-type group (all the Blues), the rationale is: given such large intra-individual
tolerances, the inter-individual tolerances are not significant; the Vintage-30 on average
sounds just like the Celestion Blue does. Sure, that is speculation at this point – the sample
was much too small, and it might be that one of the two acquired Blues is different from all
the rest of the family. In any case: showing bottomless impudence, this author has carried out
more comparative measurements with further speaker-pairs: see Fig. 11.27. To preempt any
wrong conjecture: all speakers were bought in pairs, none was re-coned, and none had been
subjected to excess power.
Fig. 11.27: Comparison of two same-type Celestion speakers: 2 x G12-80 (left), 2 x Vintage-30 (right).
What can happen if a loudspeaker is re-coned (i.e. if has received a replacement membrane),
is shown in Fig. 11.28: someone has re-coned an old AC-30-speaker … with the wrong
membrane, however! So much for the legendary vintage-sound …
Fig. 11.28: Left: frequency responses of the two Celestion speakers of an AC-30 from the 1960’s.
Right: frequ. response of the impedance of the two Celestion Blue from Fig. 11.26 (measured w/out cabinet).

The measurement results of loudspeakers of the same type or build advise caution: even if we
insinuate that speakers from modern production have negligible tolerances, it would be quite
appropriate to have some doubts regarding the holy cows from the 1960’s or even from back
in the 1950’s. That AC-30 (copper panel) offered for a whopping $ 4000 – does it sound so
good because its speakers have been “played in” for so long? Or because they were re-coned
at some point in time with no-name membranes … which the always-helpful Mr. Ly-Ing has
discretely stamped with “T530”? Or maybe the amp features yet un-played NOS-Types♣??
Word is the latter are unearthed more and more often these days. It is also easily possible that
new replicas are mounted: lovingly wound by British hand using old original tooling re-
discovered in the back of the basement. Well, that would not be cool, though, ‘cause even if
they’d been “aged” by Mr. Murphy himself personally – nothing beats the real stuff. This
nagging question remains: what’s real, if two original G12-80 differ by ±5 dB? An answer
cannot be given as long as the vita of most of the old speakers remains shrouded in the mists
of time, and artificially inflated prices impede statistically relevant investigations.
So, let us dwell some more on the loudspeakers at hand, and think not just about life in
general but the frequency response in particular. A measurement in the AEC is a required
criterion, but not a sufficient one. Of course, beaming-effects need to be considered (we’ll get
to those in Chapter 11.4), and non-linearity (Chapter 11.6). In order to be able to give at least
a general statement on directionality, we find measurements in the reverberation chamber in
Fig. 11.29. No surprise there: differences of a few dB across all measured Celestion speakers,
and small deviations between the Vintage-30 and the “Blue” (they do not stand out
significantly beyond the – assumed – production tolerances). It certainly would be an
exaggeration to attribute the same sound to all (measured) Celestions: there are differences,
and they are audible. However, despite all appreciation of the odd decibel that distinguishes
the frequency responses here and there, we must not overlook one fact: if you remove the
combo from its stand (in Bavaria, that would be two beer-crates …) and place it directly on
the floor, level-changes of the same order of magnitude will occur. And that’s free of charge!
Fig. 11.29: Measurement in the reverberation chamber: overlapping 1/3rd-octave analysis, pink noise,
rotating microphone. Right: Celestion "Blue" (–––) vs. Vintage-30 (----).
And with that, enough space has been dedicated to Celestion, manufacturer of "the finest
guitar loudspeakers that money can buy" – there are others, after all. No, not Goodmans, "the
largest UK manufacturer of loudspeakers". And not Fane, "Home of the greatest high power
speakers in the world", either. And neither JBL, "the leading loudspeaker manufacturer in
the world". Rather: Eminence, "the world's largest loudspeaker manufacturing company",
shall be checked out, and Jensen, simply “the inventor of the loudspeaker”. What Celestion
represents for VOX and Marshall, Jensen was for Fender. From the 1940’s to the 1960’s,
Fender mounted Jensen Alnico-speakers, and until about 1967 Jensen ceramics-speakers.
Optionally, the JBL D-series was available, but Jensen was the standard.
♣
NOS = New Old Stock = unused stock. Allegedly stowed away for decades.
© M. Zollner 2008 - 2014

Already at first glance, the P12-R-membrane reveals a different build, distinctly deviating
from the Celestion-standard: a smaller dust-cap, and more (and differently formed)
corrugations. Fig 11.30 clarifies the differences: the Jensen is a bit less loud but puts more
emphasis on the treble. The latter characteristic, at the very least, would suit the Fender
community fine – “silvery treble” is expected there.
Fig. 11.30: Comparison Celestion Vintage-30 (–––) vs. Jensen P12-R (–––) in the AD60-VT-cabinet.
You may naturally ask right away what sense there would be in installing a typical Fender-
speaker into a VOX-cabinet. Indeed … but how else would you do a comparison? Both in a
Fender-cabinet? That would not work either, for the same reason. Each speaker in its own
proper cabinet? In that case you would not only compare two loudspeakers but also two
different enclosures. Each speaker in a baffle? That would be absolutely not stage-typical.
From the almost indefinite number of possible enclosures, we very arbitrarily picked the
AD60-VT – a choice had to be made, eventually. Also, in order to enable us to compare to the
other measurements presented so far, all further speakers were analyzed mounted in this
cabinet.
Fig. 11.31: Three Jensen-speakers in comparison: P12-R, P12-N, C12-N in the AD60-VT-cabinet.
Fig. 11.31 depicts the comparison of three Fender-typical Jensen 12”-speakers. Unlike the
Celestions – which gave very similar measurement curves – the Jensens show pronounced
differences. It is not only the power capacity that is distinct but in fact diverging sound-
philosophies are realized: we have the treble-emphasizing P12-R, the Celestion-like P12-N
with the marked 1.5-kHz-dip, and the balanced C12-N … as far as you actually want to use
the word “balanced” in the face of ±6-dB-fluctuations. These peaks in the frequency response
are, however, typical for the genre – none of these loudspeakers could be termed “better” or
“worse”. Yes, we could wish for a little better efficiency in the P12-R, but that’s it.
Everything else is a matter of taste. Guitarists that appreciate a treble-laden sound with little
distortion often opt for the Jensen. Distortion-rockers tend to go for the Celestions. And then
there are those players that seek a not-quite-so-trebly sound without much distortion – it takes
all sorts to make a world, doesn’t it …?

And on to Eminence, Fender’s choice in speakers after 1967. As shown by Fig. 11.32, these
loudspeakers prove to have their own character, too – both in the direct sound (AEC) and in
for the diffuse sound field (RC). Jensen and Eminence each offer about a dozen guitar-
suitable speakers; only three from each manufacturer were selected and analyzed.
Fig. 11.32: Three Eminence-speakers in comparison: L-122, L-125, L-M12 in the AD60-VT-cabinet.
As a supplement, we will now call in the Fender-cabinet, after all, in order to at least once
operate Jensens and Eminence-speakers on their home turf: a Tweed Deluxe (Fig. 11.33)
shall now serve. The small 14-W-amp in the 5E3-Deluxe would not ask too much of any of
these speakers; the authentic choice would be a Jensen P12-R. The measurements (as always
not with the guitar amp, but with a stiff voltage source) reveal differences that occurred in a
very similarly manner with the AD60-VT-cabinet, as well – no surprise there. Up to about 2
kHz we see significant, enclosure-typical divergences that – depending on your mentality and
sense of mission – could be called “huge” or “marginal”. Some significance should be
attributed to at least the 190-Hz-peak in Fig. 11.32 that is followed by a 320-Hz-dip: that’s
quite typical for VOX alright.
Fig. 11.33: Measurements in the Tweed-Deluxe-cabinet: Jensen and Eminence 12"-speakers.
We have given a relatively large amount of space to the 12”speakers. Before we – more
concisely – get to their 10”- and 15”-colleagues, let us try to come to some kind of evaluation
– a classification of significant differences (Fig. 11.34):
© M. Zollner 2008 - 2014

In Fig. 11.34, three different characteristics are emphasized. They were not elaborated using
sophisticated factor-analysis but constructed (hopefully not all too arbitrarily) on the basis of
visual criteria from the frequency responses. The first criterion is found in the efficiency
achieved at 1 kHz, i.e. level (dB-value) of the frequency response. The investigated 12”-
speakers exhibit differences up to 4 dB, and that is indeed noteworthy: a level difference of 4
dB corresponds to a power-increase of 150% i.e. for example form 10 W to 25 W. Relating
that to loudness, as it is readily done from the side of psychoacoustics, is permitted but
requires some special caution: the simple rule of “double loudness necessitates 10 dB level
increase” is valid for (sufficiently loud) 1-kHz-tones that are not subject to masking! A guitar-
tone having to assert itself against competing sounds is not of that category! (For more on this
see “masked loudness” in the psychoacoustics-textbooks). 4 dB – in everyday life on stage,
that is the difference between “always a bit too soft” and “that’s it!”.
As a second criterion, we picked the range of the mids that was defined for this comparison
from about 600 Hz to about 4 kHz (i.e. incl. the so-called “presence”-range). Here we have
speakers with and without a middle-dip (or “mid-scoop”): it is generally strongly pronounced
in the Celestions and rather less in the Eminence L-125. As the last criterion, a distinction is
made between a more even level-curve at middle and high frequencies, and a more resonant
curve. The corresponding first group includes e.g. the G12-H or the P12-N, while the G12-80
and the original speaker of the AD60-VT are to be counted in with the second group. The
main differences between all measured Celestions manifest themselves in these resonance
peaks: their markedness (damping, or Q-factor) shapes the sound – but it is subject to
pronounced manufacturing tolerances, as we have seen in Figs. 11.26 and 11.27.
The question “which is the best loudspeaker, then?” has to remain unanswered for two
reasons: if manufacturing-induced variances of speakers of the same type are larger that type-
specific differences, classifying becomes rather problematic. And then: beyond the efficiency,
sound evaluations are subjective. There are more speakers between …
Fig. 11.34: Attributes for distinguishing loudspeakers: 0.1 – 1 kHz (upper left), 0.6 – 4 kHz (upper right),
relatively even treble-range (lower left), resonant treble-range (lower right).

At last, we now turn to the 10"-loudspeakers, as they are found mounted in single units in
small combos (e.g. Princeton), but also installed as 4x10”-quartetts in rather grown-up amps
(Super-Reverb, Bassman). Compared to the 12”-speaker, the membrane-surface of the smaller
10”-cousin is reduced by 30%; at the same vibration-amplitude, the membrane can thus set
less air in motion. More precisely: with the same membrane-movement, the smaller
membrane radiates less sound power. Given the equal input power, a 10”-speaker does not
need to be less loud than a 12”-speaker but it is in many cases. The “golden rule” of: the
larger the loudspeaker, the louder it is has a seductive rationale: as the surface of the
membrane approaches zero, the efficiency also needs to go down to zero. In reality, however,
the surface of a membrane is never close to zero, and the reasoning is misleading. In fact, the
efficiency depends not only on the membrane surface but also on the membrane mass and the
force-factor (transducer coefficient Bl), and these normally differ from speaker to speaker.
Eminence, for example, specifies the L-B102 (10") with 101 dB (1W @ 1m) and the KAPPA-
18 (18") with 97 dB (1W @ 1m). And a counter-example from the same manufacturer: the L-
102 (10") has a 97-dB-spec while the L-151 (15") lists 100 dB.
There is only one safe statement for the difference between 10”- and 12”-speakers. The 10”-
loudspeaker is smaller. The 10”-spealer is not generally of lower power capacity, not
generally less loud, not generally lighter and not generally brighter in its sound. Regarding the
power capacity: both Jensens P12-R (12") and P10-R (10") are specified at 25 W, the C10-Q
(10") is listed with 35 W, the C12-R (12") with 25 W, and the NEO-10 (10") with 100 W. As
to the weigh: L-122 (12") = 2.5 kg, L-B102 (10") = 5.5 kg. Things become more complicated
concerning the treble response, because the upper cutoff frequency of the power-radiation
indeed depends on the diameter of the membrane, and on the other hand an ideally form-rigid
membrane would be the corresponding pre-requisite to make a corresponding simple
statement. Completely wrong is the often-voiced justification that the larger membrane would
be too heavy to vibrate at high frequencies. Very basically: as the mass of the membrane is
enlarged, the efficiency drops in the whole range above the resonance frequency (e.g. 100 Hz)
and not merely at high frequencies [3]. There are other reasons for the fact that in many cases
the larger membrane does not sound as trebly as the smaller one: the former has more
beaming at high frequencies and generates less diffuse sound in the treble range. In the end, it
is always the membrane that plays the pivotal role: its shape and thickness, its corrugations,
its damping and its dust-cap determine the transmission characteristics. Eminence specifies
3.5 kHz as the upper cutoff frequency of the DELTA-10, but a whopping 4.5 kHz for the larger
GAMMA-15. The basis for this info is, however, an on-axis measurement – presumably the
power bandwidth is larger for the DELTA-10 (the datasheet is silent about that).
In Fig. 11.35 we see the frequency responses (measured in the AEC) of a 10"- and a 12"-
speaker. Both were mounted for the measurement in a sealed 39x39x25-enclosure. The P10-R
generates, on average, a smaller level but relatively more treble than the P12-R.
Fig. 11.35: Comparison between a 10"-loudspeaker (P10-R, left), and a 12"-loudspeaker (P12-R).
© M. Zollner 2008 - 2014

Fig. 11.36 compares 15"-loudspeakers, all mounted in the 106-l-enclosure (1 l = 1 liter =

0.264 US-gallons) and measured in the AEC at 1W/1m, with no damping introduced to the
enclosure so that the cavity resonances at 230 Hz and 500 Hz are clearly visible. An enclosure
of the given size is relatively small for a 15”-speaker. But then almost everything is possible
for guitar setups: the Vibroverb, for example, only makes a scant 88 l (gross) available to its
15”-speaker in an open-back configuration, while the Showman is much more generous at
163 l in a ported box. We shall not concentrate on the bass-range here, however – rather the
focus shall be directed to the range upwards of about 300 Hz: the G15-100 and the Fane
display an even level-response (save for the enclosure resonance) but differ by more than 6
dB. Reminder: in order to increase the level by 6 dB, the input power needs to be quadrupled.
The Powercell, on the other hand, is not designed with an even frequency response in mind
but shows the typical S-curve of instrument-loudspeakers. The measurements prove a 15”-
speaker does not generally generate a higher SPL than a smaller loudspeaker, and document
that the upper cutoff frequency (measured on axis) can readily by at 5 kHz – just like for a
12”-speaker. The differences in the beaming-behavior will be examined in Chapter 11.4.
Fig. 11.36: Frequency responses of SPL and impedance, 15"-loudspeaker in a sealed 106-l-enclosure,
AEC, 1W @ 1m. All three speakers are manufacturer-specified at 8 Ω.
Maximum power input (manufacturer specs): Celestion Powercell = 250 W, G15-100 = 100 W, Fane = 200 W.
If two loudspeakers are mounted in an enclosure instead of one, the on-axis sound pressure
theoretically doubles. Compared to the doubled power input, this implies a gain of 3 dB (Fig.
11.37). The efficiency does, however, not simply continue to rise proportionally with the
number of speakers but depends on the individual geometry.
Fig. 11.37: Comparison 1x12" (---) vs. 2x12" (–––); ordinate values are referenced to the same power input.

11.4 Directional characteristics 11-29
11.4 Directional characteristics
As a loudspeaker radiates sound, it gives rise to a sound field around it, i.e. a section in space
to which physical quantities can be assigned as a function of location and time. In a sound
field, these quantities are the sound pressure (p, a scalar), and the (sound particle-) velocity (v,
a vector). Both quantities are not only dependent on time but also on the location. For
frequency responses of loudspeakers, usually the SPL measured “on axis” is given, i.e. the
SPL that occurs e.g. 1 m ahead of the membrane. Deviating from this measurement point by a
specific angle (e.g. 30°) from the axis, the result is a different frequency response. The
reasons for these differences in the frequency responses are differences in travel time (and
corresponding interferences) between the sound waves emitted from different sections of the
membrane. These are effects summarized with the term beaming, or directionality.
As a first simplification, the loudspeaker membrane is described as a circular plate (piston

diaphragm) vibrating without changing its shape. To explain the directionality, Huygen’s
principle (well known from optics) is called into action: every differentially small part of the
membrane emits a spherical wave, and all these spherical waves superimpose in the free
sound field resulting in the radiated sound wave [3]. At a measurement point located axially,
all sound waves will have to travel approximately the same distance, and arrive at the same
time (with the same phase). However, as we move the measurement point off-axis, the sound
paths will differ, and phase shifts – and thus cancellations and beaming – will occur. At low
frequencies (= long wave-length), the travel path differences are relatively small and the
beaming is less pronounced. However, as the wavelength becomes smaller with rising
frequency (λ = c / f), already small differences in path-length (e.g. 5 cm) give rise to a
noticeable phase-shift (elaborated in [3]). Consequently, the loudspeaker will radiate without
beaming (spherically) in the low-frequency range, but as the frequency rises, so will the
beaming effect. Usually, the frequency with a wavelength just fitting into the circumference
of the loudspeaker is taken as limit from which beaming occurs. For an effective diameter of
27 cm, this results in fg = 400 Hz. A 12"-speaker therefore features approximately (!) two
different radiation characteristics: without beaming below 400 Hz, and above 400 Hz a
frequency-proportional beaming. So much for the simple piston diaphragm theory, anyway.
Measurements with lasers (Chapter 11.3), however, show that the membrane already “breaks
up” (i.e. it fails to keep its shape) upwards of 350 Hz. Therefore the piston-diaphragm theory
also breaks: it breaks down, though. To formulate this more obligingly: from 350 Hz, we
leave the range of validity of the piston-diaphragm theory. Now, it is simple to shoot down a
theory but much harder to present a better theory instead. Of course, there are powerful
formula the global significance of which can hardly be shaken, e.g. rot(v) = 0. Given the
(location-dependent) membrane velocity, we may – now already more specifically –
formulate the radiated wave as an integral that can be solved at least numerically. In
approximation, that is, without saying. However, one differential equation won’t do the job
because the pattern of partial vibrations on the membrane may strongly change already with
small frequency variations (e.g. +5 Hz). Also, to put together a directional diagram, the
solution is required not only for one point in space. Because numerical algorithms for
calculating the sound radiation are effortful (and require even more effort in corresponding
measurements), the approach using purely metrology can still hold its own next to analytical
descriptions. So let’s go ahead, and let’s measure frequency responses in various directions,
put together polar diagrams for various frequencies, and determine frequency dependent
directional indices in the AEC or the RC. The following characterizations use the piston
diaphragm theory as a basis and compare its teachings with measurement results.
© M. Zollner 2008 - 2014

The directional gain Γ of the piston diaphragm is calculated from the Bessel-function J1:
Directional gain [3]
Γ is dependent on the wave-number k = ω /c, on the effective membrane radius a, and on the
angle Θ defined relative to the loudspeaker axis. The logarithm (with the base 20) of the
directional gain is the directional index D. For low frequencies, D is approximately zero, as
the frequency rises or as the angle Θ increases, D becomes negative. The left-hand section of
Fig. 11.38 shows the directional index, the right-hand section shows the directivity.
Directional indices are bi-variant quantities; they depend on frequency and angle. To obtain
the directivity, the envelope integral is calculated (“averaged”) across all angles – only a
frequency-dependency remains. Since the theory of the piston diaphragm is based on an
infinite baffle, sound is only radiated into one half-space – and thus d = 3 dB at low
frequencies.
Fig.11.38: Directional index D of the piston diaphragm, a = 13.5 cm, Θ = 15°, 30°, 45°. Right: directivity d.
So much for our (simple) theory – how do measurements in the anechoic chamber compare?
For that assessment, a 12”-Celestion-speaker (G12-M) was mounted in a small sealed
enclosure (39x39x25 cm3), and measurements of the SPL were taken at 0° and 35° (Fig.
11.39). Easily recognizable is how nicely the curves run in sync up to about 150 Hz – from
then on the 35°-curve increasingly deviates from the 0°-curve. However, it is also clearly
evident that this deviation corresponds only with a very coarse approximation to the piston-
diaphragm theory. In the right-hand section of the figure, the calculated directivity for 35° is
included (dashed line) – the curves do take a rather different course. Particularly evident: the
figure holds three measurements taken with the enclosure turned by ±90° around the speaker
axis (as indicated by the small sketch). We would expect rotationally symmetrical behavior
from a single speaker, requiring the membrane to vibrate exclusively in rotationally
symmetrical fashion. Which in fact it does – but not exclusively, as shown in Fig. 11.6. In
particular in the high-frequency range, a multitude of complex modes occurs that certainly are
not all rotationally symmetric. The radiation behavior is correspondingly complex.
Fig. 11.39: 12"-loudspeker measured in the AEC, with 0° (–––) and 35° (---). Right: directivity.

The directional characteristic found in Fig. 11.39 clearly deviates from that of an ideal piston
diaphragm. Still, it should not be inferred that only the sound radiated off-axis is going to
buck the theory; some effects (e.g. at 1.5 kHz, but also in the high treble range) originate in
the axial SPL that for the ideal piston diaphragm should be frequency-independent. Evidently,
this is not the case, and for that reason alone the directivity deviates from the nominal curve.
Why in fact is the directional characteristic of the radiation behavior that important? An often-
heard comment is that most of the listeners are seated in front of the loudspeaker, and
therefore the sound radiated to the side would be insignificant. Well, it is significant, because
in the listening room (or hall), the sound radiated off-axis will be reflected by floor, ceiling
and walls, and it will reach – as room sound – the ears of the listener with only little delay. It
is impossible to exactly describe all individual reflections in a real room because already
simple objects (chairs, lamps) feature a highly complex reflection behavior. That’s why we
make do with the directivity. It is quite useful as an approximation: a high directivity means
much direct sound and little room sound. Sure: that room … its special absorber-distribution
... the position of the listener … and much more. Still, we need to simplify in order to push
forward to the essentials. When operating two loudspeakers with significantly different
directivity, the above statement holds as a simplification: more beaming = less room sound.
Like the directional index, the directivity d is calculated using the first-order Bessel-function
(J1). Approximately, d rises at a rate of 20dB/dec above the cutoff frequency, with the latter
being defined by its wavelength λ effective membrane-circumference (12" → 400Hz).
Directivity [3]
The larger the membrane is, the lower the frequency where the beaming starts: a 15”-speaker
has stronger beaming than a 10”-spekaer, but four 10”-speakers have a more pronounced
beaming than a 15”-speaker because the effective membrane area of the former quartet is
larger than the membrane area of the latter. Fig. 11.40 juxtaposes theory and measurement
results. As already mentioned, measuring the directivity is difficult because the “artifacts”
encountered in reverberation chamber and anechoic chamber can add up. However, if we do
not regard the directivity as a system-immanent quantity (which in fact it is not, anyway) but
as relating to the environment, then the measurements become sufficiently reliable, and even a
negative directivity appears purposeful: at low frequencies, the loudspeaker positioned in the
reverberation chamber has a higher efficiency compared to the positioning in the anechoic
full-space (Chapter 11.5). If we do not attribute any significance to differences as small as up
to 1 dB, the basic curve can be interpreted nicely, especially when comparing several
loudspeakers measured in the same room.
Fig. 11.40: Beaming for an 8cm- and an 8"-loudspeaker. Measurement (––), simple model-calculation (––).
© M. Zollner 2008 - 2014

The rather bad match between measurement and theory seen in Fig. 11.40 for the 8”-speaker
is not likely to be a result of unsuitable instrumentation: the theory simply does not fit the
speaker – in the higher frequency range, the membrane is not vibrating anymore without a
change in shape. The beaming-minimum at 7 kHz has its basis in a destructive interference
that leads to a minimum in the axial radiation. Half the wavelength amounts to a mere 2.5 cm
at this frequency, and cancellations are easily conceivable. The off-axis radiation is not
subject to this interference, and that leads to the effect of a minimum in the directivity. The
latter does depend on two quantities: on the direct sound, and on the diffuse (room) sound.
Consequently, a minimum in the directivity may be obtained via two ways: by efficient
radiation of diffuse sound, or by inefficient radiation of the direct sound.
With Fig. 11.41, we return to the 12”-speaker that was already used for most of the previously
presented measurements: the Celestion G12-M. The left-hand picture shows measurements
with a small sealed enclosure. Up to 1 kHz, the beaming is somewhat stronger than calculated
using the simple theory – that may be due to the enclosure: at 39cm x 39cm, the front panel is
not actually infinite but already larger than the effective membrane diameter (27 cm). The
curve above 1 kHz cannot be clearly attributed anymore to anomalies of a single sound field:
both direct- and diffuse-sound deviate significantly from the simple piston diaphragm theory.
Fig. 11.41: Frequency response of the directivity: G12-M, mounted in two different enclosures.
For the right-hand graph in Fig. 11.41, the G12-M was mounted in the open-back VOX-
cabinet already used in Chapter 11.3. This cabinet shows everything but a dipole-
characteristic! As main effect, we recognize two beaming-minima (350 Hz, 1.2 kHz) on the
one hand, and on the other hand a global widening of the treble-reproduction (reduction of
beaming). Given the high-frequency beaming of the loudspeaker, it will not make a difference
for the on-axis AEC-measurement whether the rear panel is open or closed. For measurements
in the RC, however, a difference will show because the same amount of power is radiated
from the rear of the speaker (in idealized thinking: level of diffuse sound +3dB). At low and
middle frequencies the superposition of the sound waves radiated from the front and from the
rear leads to comb-filter-like ripples in the directivity. Again, it is predominantly the sound
radiated to the front that forces the shape of the frequency response in the beaming: the
minima at 350 Hz and 1.2 kHz are found with axial AEC-measurements, as well – as e.g. Fig.
11.24 shows for all measured Celestion-speakers
Shape and type of the cabinet contribute significantly to the loudspeaker-sound. That also
holds for HiFi speaker arrangements, but here the direct SPL should be as much as possible
frequency-independent, and the directivity should rise evenly across the frequency such that
in the end the speaker will sound good (i.e. neutrally) despite the enclosure-specifics.
Conversely, for the guitar speaker the cabinet provides a distinct filter; its directionality
cannot be changed electronically.

How dominant the influence of the cabinet is in comparison to variations of the loudspeaker
may be seen from Fig. 11.42 – it includes the directivities of several Celestion speakers all
mounted in the VOX AD60-VT-cabinet. At first glance all the curves are of very similar
shape – at second glance one speaker is strikingly different at 8 kHz: it is the Celestion
“Blue”. After what has been stated above regarding that speaker, at last we have an
objectifiable rationale in favor of this speaker … possibly a late satisfaction for all those still
paying off the debts caused by that speaker. Of course, we shall not even start questioning
how significant the frequency range in question actually is ( Abb. 11.25).
Fig. 11.42: Frequency responses of the directivity for various 12”-loudspeakers: Celestion, Jensen, Eminence.
In the right-hand section of the figure we find a few loudspeakers with a more strongly
differing directivity: Jensen and Eminence. The general reason is quickly identified: their
membranes show more diversity than those of the Celestion speakers: size of dust cap,
corrugations, depth of membrane, diameter of the voice coil. Still, the main effect is caused by
the enclosure; the opening in the rear takes care of characteristic beaming-minima. A
directivity of 0 dB is often interpreted as spherical radiation although this is not always
applicable. The degree of beaming (or beaming factor) relates the intensity radiated in the
axial direction to the averaged intensity radiated in all directions [3]. If – due to an
interference-cancellation (pole) – no sound is radiated axially, the beaming-factor is zero and
the directivity is – ∞. If axially only little sound is radiated but in all other directions beaming
occurs, d = 0 dB may result – despite the fact that there is no spherical characteristic.
Directional diagrams give clues regarding the direction-dependency of sound radiation. In

corresponding measurement setups, the object to be measured rotates by 360° on a revolving
table, and the SPL is registered dependent on the rotation angle. The resulting diagram is
usually laid out using polar coordinates. Fig.11.43 exemplifies 3 directional diagrams
measured with the AD60-VT. None of the diagrams shows the dipole-typical radiation pattern
– this is due to the cabinet acting as a phase-shifting filter for the wave emitted to the rear.
Fig. 11.43: Horizontal directional diagrams measured with third-octave-noise. Loudspeaker in AD60-VT-cabinet.
© M. Zollner 2008 - 2014

Directional diagrams have descriptive qualities but can only exemplify one single plane – as
such they have limitations. For a circular membrane, often a rotation-symmetric radiation is
implied, coupled to the hope that a single measurement (per frequency!) will be sufficient.
Often, this is a reasonable approach, but just to be safe we should take additional
measurements. Fig. 11.44 shows horizontal directional diagrams – in contrast to Fig. 11.43, a
sinusoidal test-signal was used, though. At 400 Hz a perfect symmetry exists, while at higher
frequencies, any asymmetric shape may occur due to membrane resonances. Since these
shapes are highly dependent on frequency, the information contained in directional diagrams
needs to be drastically reduced in order to remain clear. Therefore, noise (of octave- or 1/3rd-
octave bandwidth) is often employed as test signal – this has the effect of an averaging across
the corresponding frequency interval. Using that approach, the small variations contained in
directional diagrams are not an expression of high directional selectivity but the result of
stochastic processes. Given an optimized averaging time-constant, misinterpretations are not
to be expected – if necessary, fluctuations can be reduced via averaging over several turns of
the rotational table.
Fig. 11.44: Horizontal directional diagrams, measured with a sinusoidal signal. 12"-speaker, AD60-VT-cabinet.
Installed in a sealed cabinet, a loudspeaker will operate as a spherical source; with a rear
opening in the cabinet, a dipole will result. However, the stiffness of the air contained in the
cabinet forms, in conjunction with the inert (mass-dominated) radiation impedance of the
opening, an acoustic filter creating phase-shifts, and therefore the directional diagrams have
the shape of a (logarithmized!) eight only at very low frequencies. Already at 200 Hz, this
dipole-behavior is all but gone, and the horizontal directional diagram approaches a circular
shape. In Fig. 11.45 we find a comparison between the original VOX and a variant where the
rear was closed off with a board. The latter does not provide a complete seal, however: the
slits foreseen to provide ventilation for the amplifier section let sound pass through.
Horizontal directional diagrams for the AD60-VT cabinet with open and closed rear wall are
juxtaposed in Fig. 11.46 (measured in the AEC using 1/3rd-octave noise).
Fig. 11.45: VOX AD60-VT: rear wall closed with board (–––) vs. original condition (–––).

Fig. 11.46: Horizontal directional diagrams with sealed (1st and 3rd line) and open rear wall of the AF60-VT.
The dipole-characteristic begins to become visible at 80 Hz, while at 200 Hz barely any differences are visible. In
same ranges, more sound is radiated to the rear than to the front (e.g. at 315 Hz); this is caused by an impedance-
transformation (an effect of the cabinet cavity).
© M. Zollner 2008 - 2014

Fig. 11.47 shows how strongly the cabinet influences the sound-beaming. All measurements
were done using the same cabinet with or without a rear panel (i.e. closed resp. open). For
both variants, the 10”-speaker features less beaming in the high-frequency range – however,
the cabinet-specific differences far outweigh the loudspeaker/diameter-specific differences.
The directivity is negative at 315 Hz – this again is due to the direct sound radiated to the
front and showing an interference minimum (rear-ward diffraction wave) at that frequency.
Fig. 11.47: Comparison of the directivity of the closed and the open cabinet; G12-M vs. P10-R.
If two or more loudspeakers are mounted in a cabinet, the beaming increases because the
membrane area grows. Corresponding measurements that support this general statement are
shown in Fig. 11.48. Differences are visible in the details, though: first, the enclosure shapes
are different, and second, the sound power radiated to the rear is loudspeaker-specific. Finally,
Fig. 11.49 presents the directivity of loudspeaker cabinets that are designed to reproduce the
whole frequency range relevant for music transmission (so-called “full-range” speakers).
Their directivity should increase as evenly as possible – this is achieved quite well in the
Quinto.
Fig. 11.48: Left: Fender Super-Reverb, 4x10", Jensen P10-R; right: typical 2x12"-Box, Celestion G12-M.
Grey curve = VOX AD60-VT (1x12") for comparison. The directivity is given as a function of frequency.
Fig. 11.49: Directivity of a HiFi-Box (left) and a small full-range-box for stage use (right).
Grey curve = VOX AD60-VT (1x12") for comparison. The directivity is given as a function of frequency.

The area-rich 15”-loudspeakers should actually show particularly strong beaming effects –
but measurements support this hypothesis only in part (Fig. 11.50). For both the Fane and the
Powercell, the directivity decreases again in the highest frequency range. Possibly, this is
connected to the large dust-caps of these speakers: both sport air-tight dust-caps acting as
high-frequency emitters with a diameter of naturally not 38 cm but merely 10 cm. However,
no further measurements regarding this hypothesis were conducted.
Fig. 11.50: Directivity of 15"-loudspeakers. Lower right: comparison 15” vs. 10” speaker.
The last section in Fig. 11.50 depicts a comparison between 15”- and 10”-loudspeakers.
Below 1 kHz, differences are rather limited; only above this limit, things become more
specific. According to the simple piston-diaphragm theory, we should see a difference of 3.5
dB. Prerequisite would, however, be an infinite baffle – but the measured loudspeakers were
installed in airtight boxes of different sizes (10" ⇒ 39x39x25 cm3, 15" ⇒ 40x74x36 cm3).
As a last point, a special characteristic of 2x12”-combos shall be considered: almost always,

the loudspeakers in the corresponding cabinets are mounted horizontally next to each other,
conversely to the speakers in public address systems where the speaker chassis are mounted
vertically above each other. The vertically aligned column has the advantage that the vertical
beaming is increased (less sound to floor and ceiling), while horizontally a wide-angle
radiation is retained. That the 2x12”-combo is realized exactly the other way ‘round may be
the result of a desired visual look and feel, but also a necessity required by the amplifier-
chassis: you need quite a bit of space to line up 11 knobs (like in the Twin-Reverb). Fender’s
first foray into tall cabinets and two rows of control knobs (1967, in the first Solid-State
amplifier series) pretty much was a disaster (possibly not merely due to the geometric
configuration…): comments included the terms “ugly refrigerators” or “TV-trays”. With
regard to beaming and sound dispersion, the tall, slender design would certainly have had
advantages. (Translator’s note: maybe the more vertically oriented sound dispersion is something favoured by
the musician standing in front of the amp? Especially when using the “tilt-back” legs customary in these amps,
the vertical spreading out of the sound makes it less crucial at what distance the guitarist stands relative to the
amp, and how tall he/she is. Also, more sound is aimed at the guitarist when playing – too? – loud …)
© M. Zollner 2008 - 2014

11.5 Efficiency and maximum sound pressure level
Is a 100-W-speaker twice as loud as a 50-W-speaker? That question is asked a lot, and has at
its basis a common misunderstanding. The Watt-specification of a loudspeaker only tells us
about the maximum power the speaker can take, but includes no statement at all about the
acoustical power-yield. Even if you just mount four 100-W-lightbulbs into an enclosure, you
may label it with “400-W-Box” – but you won’t get any sound out of it.
Strictly speaking, we would have to distinguish the power fed to a speaker into effective
power, reactive power, and apparent power. However, in practice that is simplified: every
speaker is to be assigned, by the manufacturer, a nominal impedance R (e.g. 16 Ω), and,
together with the maximum power P, the maximum voltage is derived: . A 16-Ω-
speaker with a maximum power rating of 100 W may be driven by an RMS-voltage of 40 V.
Some limitations need to be observed here: a DC-voltage of 40 V may not be connected to the
speaker, although again 100 W would be the result – however the speaker would be destroyed
by this “drive signal” (the manufacturers do not specify any maximum DC-voltage at all). A
typical source material would be guitar-tones, but this signal definition is too general. As a
compromise, specially filtered noise-signals are often chosen, e.g. the EIA-noise (RS-426-A,
RS-426-B), or the IEC-268-1-noise, or the AES-2-1984-noise, or the DIN-45573-noise, or
other specifically defined signals. These are then (depending on the specification) to impact
for 8 or 100 or 300 hours on the speaker without destroying it. If the loudspeaker can take e.g.
100 W according to such a standard, the sales department labels it with “100 W”. Or with
“200 W”, because there may be further considerations: since, allegedly, the power load is
much smaller in practice, a "CONTINUOUS PROGRAM POWER" was defined. This is a power
specification 100% above the limit-power data determined with the noise. We can see: power-
data are manufacturer-specific; they may not simply be grasped via U=RI und P=UI. That’s
similar to the area of power amps: at the Frankfurt music fair, a French manufacturer
answered – slightly irritated – to the question why his 90-W-specified amplifier would deliver
no more than 55 W: “that’s French Watts”. Ah oui, monsieur, bien sur.
The nominal impedance is not something the knowledge-seeking person will readily
understand at first glance, either. Is it the DC resistance, or the minimum- or the maximum-
impedance? It’s none of these three. The impedance Z(f), i.e. the magnitude of the complex
resistance, for a loudspeaker depends strongly on the frequency: at 0 Hz it may e.g. amount
to 6.5 Ω, at resonance (at 110 Hz) it may rise to e.g. 75 Ω, at 300 Hz, it may almost be back to
6.5 Ω again, and it will rise continuously towards higher frequencies♣ (Fig. 11.51). This curve
cannot be specified via a single value, and so the manufacturers choose a (another?) method
to arrive at one value. For example, the value of the impedance at 1 kHz is measured. Why is
that 1 kHz? Because that’s an often-used standard-frequency. Or 800 Hz may be employed …
because the recommended crossover frequency is here. Or 400 Hz: you may want to set
yourself apart from the competition that way. Or the speaker is labeled right away with
“Impedance: 4 - 8 Ω”. No, that doesn’t mean that the speaker features an impedance of
between 4 and 8 Ω. Rather, the speaker is recommended for amplifiers the manufacturers of
which on their part recommend using speakers of 4 or 8 impedance. Well then. Given all this,
it comes as no surprise that the guys at Just Barely Loud frankly admit: "The JBL 2215B Pro-
fessional Series Loudspeaker is rated at 16 Ω, while the LE15A Home Loudspeaker, which is
the same unit, carries an 8-Ω-rating". Thanks a lot, then: both allowable maximum power and
impedance are now precisely defined, and everybody can calculate from these values the
allowable maximum voltage. In case the speaker starts to communicate via smoke signals,
JBL recommends: Turn it down!
♣
For enclosure- and membrane-resonances, see Chapters 11.3 and 11.8.

11.5 Efficiency and maximum sound-pressure-level 11-39
When does a loudspeaker actually cross the River Styx? The most frequent reasons for
malfunction are too high a voice-coil temperature (excessive effective power), or too wide a
membrane displacement. Both these effects may influence each other: a strong membrane
displacement may increase the cooling of the voice coil and push the power limit a bit further.
Since, for a drive signal from a high impedance source (stiff current source), the excursion
drops off with above the resonance frequency, large displacements are only found at low
frequencies – that is one reason why the resonance of guitar speaker is not located at 20 Hz
but at 80 – 110 Hz. Another reason is the fact that as guitar player, you do not want to get in
the way of your bass player – that’s the guy who owns the low end (not necessarily implying
that guitar players are generally to be seen as belonging to the High-End-range).
For the musician, it will generally not make any difference why exactly the speaker died after
the volume was cranked up from “5” to “10”. Had to be cranked because otherwise the guitar
would have been drowned out (by the keys that just went to “10”, as well). Now the speaker is
kaput – overloaded, as the roadie knowingly attests. That happens if the amplifier delivers
more power than the speaker can take. So how much power can the speaker take? We’ve been
there – see above. Other question: how much power does the amp in fact deliver? We should
be able to at least measure that value with adequate accuracy, shouldn’t we? In principle: yes
… but: guitar amplifiers often dispense with (strong) negative feedback, and a power
specification at e.g. 1%-THD does not make much sense. Rather, the amplification is turned
up until visible clipping sets in, and from this a power-value is calculated. Maybe happily
using 1 kHz, and gladly at the nominal impedance. The power that the amp can feed to a real
loudspeaker, and in particular what it can generate under overdrive conditions – that remains
unknown. And so we read statements from the service technician testifying that he never saw
a Marshall 1959 that had a mere 100 W: it always was 140 W, or even 160 W. On the other
hand, the question does pop up how an AC-30 with its continuously-under-overload power
amplifier can generate 30 W if a quartet of EL-84’s is specified at no more than 24 W. Let’s
jot this down: both the generated amplifier power, and the power capacity of a loudspeaker
could be measured with good accuracy – but the market has found its own standards that “not
always” coincide with the norms in metrology.
Ah - the market: that is the key to understanding. Fender’s Pro-Reverb sported 40 W, so that’s
5 W more than the Vibrolux. At the end of the 1960’s, Celestion’s G-12-speaker received the
urgently expected power-upgrade from 25 to 30 W. Grown up, at last! You will recognize
similarities to the car-market: isn’t the 220 something entirely different compared to the 219?!
On the one hand, there are classifying power-ratings that portray a 10%-difference as relevant
– but on the other hand differences of 50% or more seem to be subject to pure arbitrariness. It
is difficult to avoid the impression that the head of sales – just before the big music fair –
quickly checks repair-statistics, and if the 12-50 has next to no failures, that speaker receives a
red cover and mutates into the 12-65-S. To cite Cicero: O tempora, o mores (liberally
translated: where there’s a market, there’s a way). No, this is not meant to say that power-
upgrades happen only in the sales brochures: from the 12 W of the first 1,25”-voice-coil-
carrier made of paper to the 200-W-3”-polyimid-carrier, there has been indeed a mighty
development. Individual cases need to be scrutinized, however: the Vintage-30 (12", 60 W) is
specified at 100 dB "average sensitivity", the Powercell 12-150 (12", 150 W) at 94 dB.
Attention: "6 dB less" means that at the same power input, only ¼ of the sound power is
generated. For the same sound power, the Powercell would require an input of 240 W. That is
beyond its power limit – so better buy two of the guys. Powercell? Rather, it’s Powersell!
© M. Zollner 2008 - 2014

Let us remain for a moment with the term "average sensitivity". There is – and that is not the
norm for the business – consensus that this specifies the SPL that can be generated at a
distance of 1 m with 1 W electrical power. However: this one Watt is not actually generated,
rather a voltage is applied to the loudspeaker that would create 1 W at the real nominal
impedance (for 16 Ω that would be 4 Veff). If the speaker actually has 12 Ω rather than 16 Ω,
that alone will result in the gift of another 1.25 dB for the specification listing – in the
brochure, a measly 99 dB is turned into some stately 100 dB that way. Also, across which
frequency range the averaging happens has, in case of doubt, a company-specific definition.
Let’s let a manufacturer have a say: The Sensitivity represents one of the most useful
specifications published for any transducer. It is a representation of the efficiency and volume
you can expect from a device relative to the input power. Well said – that had to be defined
for once. However, the text continues with: Loudspeaker manufacturers follow different rules
when obtaining this information – there is not an exact standard accepted by the industry.
Okay then… We can leave the world of datasheets for a bit and look into what theoretical
electro-acoustics have to offer. A spherical source generating a sound pressure of 100 dB at a
distance of 1m produces a sound power of about 126 mW [3]. Guitar loudspeakers reach these
100 dB @ 1m already with about 1 W power input; the efficiency therefore would be 12.6% –
if indeed the radiation were spherical. In the relevant frequency range, however, on the one
hand the beaming effects need to be counted in, but on the other hand many loudspeakers
exceed 100 dB @ 1m, so that overall we find efficiencies of about 10% to be the approximate
limit for the single membrane-loudspeaker. HiFi-speakers often reach only 0.1% whereas
horn-speakers can achieve more than 25% efficiency♣. Thus only the smaller part of the input
power is converted into sound, the larger part ends up as heat. No wonder that voice coils can
be destroyed if from the 100 W input power, more than 90 W dedicated themselves to heat up
the thin wire. As is generally known, a soldering iron of a mere 30 W generates a lot of heat;
the voice coil therefore needs to be able to bear substantial strain. At full power, 200°C or
more will occur; only special materials can withstand that. To decrease the temperature, there
are only two possibilities: turn it down, or increase the heat-dissipation. The former approach
would be up to the musician, the latter is the manufacturer’s area (constructional build of the
pole-pieces carrying the magnetic field, broadening of the pole-plate, pole-piece vents, etc.).
We carried out measurements with a number of guitar loudspeakers to obtain more precise
date regarding efficiency. The instrumentation used was of high precision while the
measuring rooms were somewhat more limited in that respect. The fiberglass wedges of 80
cm length in the available anechoic chamber (AEC) will absorb 100 Hz to a sufficient degree;
disturbing room resonances will occur below this limit. With 220 m3, the reverberation
chamber (RC) is large enough but still sub-optimal (due to a lack of diffusers and because of
unsuitable installations). The results presented in the following therefore may not generally
claim an accuracy of ± 1dB, but they are still well usable to arrive at statements for
orientation. Measurements in the AEC (B&K 4190) were done at a distance of 3 m to the
baffle but were re-calculated for 1 m distance to make them better comparable: L1m = L3m +
9.5 dB. For sweep-measurements, the input voltage was 2.83 Veff from a stiff voltage source,
for 1/3rd-octave measurements, the voltage per 1/3rd-octave was kept constant (pink noise +
1/3rd-octave-filtering). Polar diagrams were taken in the AEC with 1/3rd-octave noise,
revolving table B&K 3922, d = 3 m. In the RC (B&K 4135), measurements were carried out
following a skewed circular path (∅ = 3 m), along which energy-averaging was performed.
Most RC-measurements were done with 50%-overlapping 1/3rd -octave pink noise (IEC 1260
class 0); U1/3rd-octave = 0.5 Veff. Employed as analysis-software: CORTEX-Viper and Matlab.
♣
H. Fleischer: Hörner endlicher Länge (horns of finite length), research report from the Institute for Mechanics,
HSBw, 1994.

Fig. 11.51: Magnitude of impedance (left), SPL in the AEC (speaker mounted in VOX AD60-VT enclosure).
The loudspeaker analyzed in Fig. 11.51 is a Celestion Blue, commonly termed “the legend”
because it served in the famous early VOX-amplifiers, and the just as famous early Marshall
cabinets. This speaker is said to have a fabulous efficiency that is – if you believe in
statements on the Internet – due to the magnetic material (Alnico) deployed back in the day.
And indeed: with 1 W as input, and at a distance of 1 m, this speaker generates up to 108 dB!
Given far-field conditions, this results in an intensity of 63 mW/m2, giving (with a sphere
surface 12.6 m2) 0.79 W sound-power and 79% efficiency. Indeed? Can that be?
Without question this is a fine loudspeaker, and it does have a high efficiency, but never 79%.
At 2.5 kHz, we must not assume spherical radiation any more, so that the “efficiency”
mentioned above needs to be multiplied by the beaming factor [3]. And while we are doing
corrections: the real input power is not P = U2/Rnominal, but results from the actual real part of
the electrical impedance.
Let us first look at the directional characteristic (directional index, [3]): loudspeaker
manufacturers publish (if they publish anything at all) the transmission frequency response
“on axis”. However, the loudspeaker radiates sound not only to the front but in all directions.
This behavior is captured either via direction-dependent transmission factor, or via frequency-
dependent directivities. That means: level plotted over frequency for various directions, or
level plotted over direction for various frequencies (Chapter 11.4). If we insinuate rotation-
symmetric sound-radiation, beaming measurements in one plane will suffice. In Fig. 11.52 we
see two directional diagrams from measurements of a combo cabinet, the rear wall of which
has an opening of 49 cm x 21 cm. Against all expectations, an almost circular radiation
pattern shows, and not the “eight” of a dipole (for details see below). At 2.5 kHz, however,
we find typical high-frequency beaming – despite the opening on the rear.
Fig. 11.52: Directional diagram in the horizontal at two different frequencies; VOX AD-60VT-cabinet.
© M. Zollner 2008 - 2014

The efficiency η is a quantity relative to the power: sound power / electrical power, or – more
precisely – effective acoustical power Pak / effective electrical power Pel. Operating the
loudspeaker from a stiff voltage source, Pel is obtained via Pel = U2 / Re(Z). It is not very
difficult to determine the real part of the electrical impedance, but establishing the effective
power Pak emitted by the speaker does become complicated – and doubly so! The
metrological investigation requires a substantial effort to begin with, and in addition this
power Pak is dependent on the environment of the loudspeaker, i.e. is not a constant. It’s a bit
like with a car: the engine may have a power output of 400 hp … but not on an icy road. The
acoustic source-impedance of the membrane (defined as the quotient of sound pressure over
particle velocity) is relatively high: the membrane could generate a high pressure, but only at
a relatively small membrane velocity. The real part of the radiation impedance is, however,
more on the low side: even for relatively high membrane velocity the forces transmitted to the
air remain relatively small, and a considerable mismatch at the membrane results. High/large
and low/small need to be seen task-specific; literature [e.g. 3] delivers supplementary data.
The loudspeaker membrane is highly unchallenged in the typical mode of operation – just like
a pitcher throwing a very small ball: whether that ball weighs 10 or 20 grams is immaterial,
with the speed being approximately the same for both cases. The energy of the heavier ball
will be twice that of the smaller one, the efficiency will be load-dependent. Applied to the
loudspeaker: could we increase the load-impedance, the efficiency would increase, as well.
The load-impedance can actually be increased by positioning the speaker enclosure directly
on the floor, or even right away into a corner of the room – that increases the efficiency. Not
without limit, of course, the velocity will drop with too high a load. Again there are parallels
to the pitcher: a 5-kg-ball will not be able to have a higher speed than the 20-gram-ball.
Apparently it is not easy to determine the loudspeaker-efficiency – that may be the reason
why the industry rarely publishes corresponding data. According to established theory, η may
change by a factor of 8 (!) if the loudspeaker is taken out of the AEC and placed into a corner
of a reflective room. Even if in practice the limits of the corresponding range are not reached
– already a factor of 2 would represent considerable uncertainty. A way out of this dilemma is
linked to comparative measurements in a special room: for example, 2 loudspeakers are
measured in the AEC – however, the desired results are not so much their absolute
efficiencies but the relation between the two. If we find, for example, a relationship of 5% to
3% in the AEC, a similar difference can be expected to occur in the real room. Measurements
in the AEC deliver pretty accurate results but require considerable effort because of the non-
spherical sound radiation that necessitates a high number of measuring points (or measuring
paths). Moreover, imperfect absorption of the absorber-wedges in an AEC even at
frequencies above 100 Hz needs to be considered. Therefore, there is still no perfect free-
space field if we limit the measuring range to f > 100 Hz. In the available AEC, we measured
level differences of ±1 dB up to 300 Hz as the positions of loudspeaker and measurement
microphone were changed (axial measurement at d = 3 m). For the efficiency, a difference of
only 2 dB represents a relative deviation of 58%, i.e. e.g. 8% instead of 5%. In addition, the
instrumentation devices will have some tolerances; they may be still connected in spirit to
Messrs. Brüel and Kjaer, and be of exemplary precision – but they will deviate a bit from the
reference value, anyway. This author does have a bit of a queasy feeling when, after just
mildly ridiculing the 35/40-W-differences in Fender amps, suddenly a measuring uncertainty
of an ample 58% pops up. What the heck … other measuring rooms are not available, and
things become even more inaccurate in the reverberation chamber. Seriously: of all the
examined AEC-positions, the best possible was retained for all further measurements.
Comparative statements can quite well made based on this situation, and above 300 Hz, the
deviations already remain below ±0.5 dB. Also, this holds in general: any more precise
measurement result is most welcome.

For measurements in the AEC, we assume that the sound wave emitted by the loudspeaker
is not (or almost not) reflected anywhere; as it hits the glass-fiber wedges that constitute the
borders of the room, the energy of the wave is (almost) completely transformed into heat. In
this mode of operation, the radiation impedance (= the impedance loading the loudspeaker)
may be calculated for a few idealizing cases [Beranek, Olsen, Zollner/Zwicker]. However, the
loudspeaker is rarely used in such an environment – there are not that many occasions when
the guitarist plays in an anechoic chamber. That does not mean that measurements in the AEC
are without purpose; it’s just that supplementary measurements in other rooms and, of course,
listening experiments are desirable. In contrast to the walls of the AEC, regular walls do
reflect the sound to a considerable extent. Sound waves (in fact an infinite number of them)
return to the loudspeaker, and the membrane does not radiate anymore into a free sound field
but has to work against the sound pressure of the reflections. Still, due to the fact that the
membrane is not challenged anyway (see above), its movement is not weakened much by the
returning sound but – if the involved phase shifts are advantageous – the efficiency is
increased. In a real room, the loudspeaker can thus generate more sound power than in the
AEC – but it may also be less depending on the circumstances, for example if the speaker is
position at a pressure node.
At this point it is recommended to also take a look at the electrical impedance. The
loudspeaker is a passive two-port, and changes in the load impedance should also change the
input impedance. Fig. 11.53 confirms that this indeed is the case – but only to a rather small
degree♣. The straightforward reason: the efficiency of course influences the impedance
transformation, as well. Or, more elementary: relative to the ohmic voice-coil impedance, the
load impedance plays only a minor role. The relationship between magnitude and real part of
the electrical impedance is depicted in the right-hand picture. The two curves more or less
correspond at the extremes (the impedance is approximately real here), in between the real
part is smaller than the magnitude – just as it need be with impedance functions.
Fig. 11.53: Left: magnitude of the electrical loudspeaker impedance (AEC ––––, RC ––––). On the right, the
comparison between magnitude (––––) and real part (-----) of the electrical impedance is shown (AEC).
From these results, we may take the following approximation: as the acoustical environment
of a loudspeaker changes, its input power remains almost unchanged; its power emission may,
however, drastically change (this needs to be looked at some more).
♣
Again, a difference of 10% can easily occur here, but the focus shall remain with the main effects. Moreover,
the differences are limited to the range below 200 Hz; above this limit, both curves coincide.
© M. Zollner 2008 - 2014

Now, on to the reverberation chamber (RC). In the ideal case, this is a room with strongly
reflecting walls that lead to a diffuse sound field in the room (except for the space in close
proximity of the sound source). This is a sound field in which the sound arrives at the
measuring point from all directions with the same probability and in which the (averaged)
sound pressure is independent of the location. The exception is the close-up range around the
sound source, this range being delimited by the effective reverberation radius [3]. A typical
reverberation radius would be 0.5 m (or less); the effective reverberation radius is calculated
from it via a multiplication with the square root of the beaming factor (e.g. 0.5 m x 3 = 1.5 m).
To be a bit more precise: the free and the diffuse sound field superimpose within the whole of
the reverberation chamber (which forms an LTI-system); close to the source, the free sound
field is more dominant while further away the diffuse field takes over. Given spherical (non-
beaming) radiation, the beaming factor is γ = 1; the boundary between free field and diffuse
field is defined by the reverberation radius. If beaming occurs, we need to use the effective
reverberation radius instead: rH* = . For broadband excitation, the low-loss sound
reflections lead to the creation of countless♣ standing waves, with the density of the
eigenmodes rising with the square of the frequency. Exciting the reverberation chamber with
a (very slow) sine-sweep, the individual resonances clearly emerge in the low-frequency
range whereas for high frequencies, there is merely a tangle of smaller and larger peaks (Fig.
11.54). And here we have the fundamental issue of measurements in the RC: these maxima
and minimal are strongly dependent on the location – they do not represent room-related
constants. While the eigen-frequencies of the room indeed need to be seen as constants (given
constant room temperature, humidity and air pressure), it depends on the loudspeaker- and
microphone-positions whether the matching oscillation modes are excited and measured.
Fig. 11.54: Sweep-measurements in the RC, 2 microphone positions; •–•–• mean values across a 1/3rd octave.
Since the resonance-peaks found in the reverberation chamber via a sine-sweep may vary by
the odd 30 dB or so when changing the microphone position, it is customary not to use sine-
tones for the measurement but noise of a width an octave or of 1/3rd of an octave. This noise,
however, is a stochastic signal and thus requires a special measurement approach. Each noise
measurement performed over a period of time represents an average over samples that must
be interpreted merely as an estimate of the true value of the basic collective. Therefore two
subsequent measurements will not yield the same but merely similar results. For normally
distributed noise (as mostly used in room acoustics) the squares of the sound pressure
(required to calculate the RMS-value) will show a χ2-scatter. Extending the averaging time of
the bandwidth reduces the standard deviation of the measurement errors. [Bendat / Piersol].
♣
Strictly speaking, the reflections may be counted, after all, so: “a lot, a real whole lot“.

The lower the center frequency of the 1/3rd-octave to be analyzed, the smaller the absolute
bandwidth; the lowest 1/3rd-octave therefore requires the longest averaging time. A 1/3rd-
octave bandwidth of 23 Hz corresponds to fm = 100 Hz, with a standard deviation of the
normalized measuring error of about 2%. At a stately 30s averaging time, that is! If we now
position the borders of the confidence interval at µ ± 3σ, then 99.7% of all measuring results
differ by less than ± 0.5 dB from the true value. Thus, the 1/3rd-octave level spectrum of the
loudspeaker voltage may be measured with sufficient accuracy with 30 s averaging time. The
1/3rd-octave wide sound pressure spectrum of the reverberation chamber could also be
determined with this approach, but the fact the SPL (stochastically) depends on time and
additionally on the location♣ needs to be considered. A level that is representative for the
diffuse field only results when the number of room resonances per 1/3rd-octave is high
enough. Without going into detail too much: that will surely not be the case below 100 Hz
(Fig. 11.54), and even above 100 Hz, pronounced level differences are still visible (Fig.
11.55). The level measurement was therefore not done at one point in the reverberation
chamber but via a rotating microphone.
Fig. 11.55: 1/3rd-octave level in the RC, measured at 4 positions with a stationary microphone (left). On the right,
an averaging along a circle (not oriented in parallel with the walls) of a diameter of 3 m is shown.
The measurement microphone takes 80 s for one orbit on the 9.4 m long circular track. This
makes for an adequate averaging accuracy both in terms of time and location – at least within
the framework of the chosen task definition. The sound pressure level L derived from energy-
related averaging along the circular orbit first results in the intensity I = 10-12 W/m2 ⋅10L/10dB;
from the latter, the sound power Pac may be calculated:
Sound power
In this formula, S is the surface area of the room, λ is the wavelength, V is the volume of the
room and TN is the reverberation time. The term within brackets represents the so-called
Waterhouse-correction♥ which considers the energy concentration close to walls.
As an example: 100 dB sound pressure level yields (with V = 220 m3 and TN = 2 s) a sound
power of 42 mW in the high frequency range. The small difference between the intensity level
LI and the sound pressure level Lp (LI = Lp – 0.2 dB) is considered in the pre-factor of 0.038.
♣
The propagation and reflection of each individual wave is subject to a determined process,
♥
Waterhouse R.V., JASA Vol. 27, March 1955.
© M. Zollner 2008 - 2014

With the instrumentation for determining the sound pressure levels in both the anechoic
chamber (AEC) and the reverberation chamber (RC) ready to go, measurements of the
radiated power could start. Two objects came first:
• 8"-loudspeaker (Eminence α-8), mounted in an airtight enclosure (22x30x18),

• 12"-loudspeaker (Celestion Blue) in the open VOX AD60-VT (Fig. 11.52).
Fig. 11.56 shows the results. The AEC-measurements were taken at a distance of 3 m but
recalculated for 1 m (L + 9.5 dB). The RC-measurements were obtained from averaging over
a circular path as described above; the level measured in the diffuse field was recalculated for
1 m. Pink noise served as test signal, it was filtered to a width of a 1/3rd-octave (IEC 1260
class 0), with the 1/3rd-ovtave-voltage fed to the loudspeaker amounting to 2.8 Veff for both
measurements.
Fig. 11.56: Comparison of AEC- and RC-measurements. AEC: 2.8 V per 1/3rd-oct., 1m. RC: 2.8 V per 1/3rd-oct.,
rH → 1m. In the frequency range below 125 Hz, the sound fields in both rooms show “acceptable” artifacts.
“RAR” = AEC, “HR” = RC
For both loudspeakers we see clear differences between the frequency responses measured in
the AEC and the RC. The main reason of the deviations is the beaming increasing with rising
frequency, but the different radiation impedance also plays a role. The Eminence-speaker
mounted in a relatively small, airtight enclosure acts, at low frequencies, approximately as a
spherical source – in the AEC, its radiation impedance is mainly formed by a mass [3]. In the
RC, we find a much more complicated radiation impedance depending on the individual RC-
data and the position of the loudspeaker. The small number of room-modes per 1/3rd-octave
has the effect that the speaker can feed sound energy only into a few narrow frequency bands
with a relatively high efficiency, and therefore the RC-level (recalculated to 1 m) is somewhat
smaller than the AEC-level. The VOX-enclosure has a rear opening of 49x21 cm2 and
consequently beaming may be expected already in the low-frequency range (rising with
increasing frequency) – but in a different manner than with the Eminence-speaker. The VOX
was measured freestanding in the AEC, and set on a 50-cm-high stool in the RC. The latter,
stage-typical mode of operation causes differences in the radiation impedance up to about 600
Hz – these will have to be discussed below in connection to Fig. 11.61. In addition, there is
the special location- and mode-dependent loading in the RC. The question regarding the
efficiency therefore needs to be discussed specifically for the given room – there are
systematic differences between the efficiencies determined in the AEC and the RC. These
differences are on the one hand typical for the respective sound field, but on the other hand
represent effects of the individual room parameters.
To be able to more precisely quantify the beaming behavior, horizontal directional diagrams
(i.e. the directional gain) were taken for both loudspeakers in the AEC using 1/3rd-octave-
noise (Fig. 11.57).

In Fig. 11.57 we see, in the left column,

the directional diagrams of the Eminence
speaker, and on the right those for the
VOX-Celestion. The Eminence was
mounted in an airtight enclosure, and the
Celestion in the AD60-VT-housing open
to the rear (Fig. 11.52).
In the Eminence, we find almost

textbook-grade beaming increasing with
rising frequency, while the VOX shows
a much more complex behavior. There is
no frequency range in which the latter
acts as a pure dipole because the air
within the enclosure forms, in
cooperation with the complex impedance
of the “compensation opening”, a phase
shifting filter. The characteristic of this
filter is remotely reminiscent of a bass-
reflex box with a rather special tuning –
certainly not one following the
Thiele/Small-approach. That is not
required anyway: this enclosure is
supposed to radiate the tone of the guitar
optimally and may (or even should) be
shaping the sound – something rather not
desired in a HiFi-loudspeaker.
Not all guitar loudspeakers are mounted

in open enclosures: the probably most
famous representative of the closed box
may be the one by Marshall. However,
Fender – more known for open
enclosures in their smaller combos –
early on offered a closed speaker
housing for the Showman and
Bandmaster setups. These included
classical bass-reflex enclosures with
sometimes quite ingenious co-axial bass-
reflex openings. It appears that in the
upper power-range, the 2- or 3-part
“piggyback”-solutions are a bit more
dominant compared to combos reigning
in the lower-power range – but that must
not be seen as a dogma. In the end, each
guitarist decides according to sonority
and radiation characteristic – or simply
Fig. 11.57: Horizontal directional diagrams. grabs "same as Jimi had".
Eminence Alpha-8 (left), VOX AD60-VT (right).
All directional diagrams are normalized to the maximum.
© M. Zollner 2008 - 2014

Fig. 11.56 already reveals much about the radiation but does not directly represent the
efficiency. The latter may be determined in the AEC via integrating over the squared sound
pressure along the enveloping surface, or in the RC using intensity and spherical surface of
the reverberation radius. For the AEC-measurement, either a large number of measuring
points (or paths) are required, or a rotationally symmetric radiation; for the measurement in
the RC we need merely the SPL in the diffuse field, volume (cubic capacity) and
reverberation time. In order to limit the effort, the efficiency was determined not in the AEC
but in the RC – starting with nominal conditions, i.e. Pel = U2 / RN. This specification is
physically still not entirely correct but does deliver purposeful comparative values for the
operation from a stiff voltage source. Guitar amplifiers do not generally feature low output
impedance but approach this mode of operation as the rather typical clipping occurs.
Supplementary measurements regarding the physically exactly defined efficiency will follow.
Fig. 11.58 shows the nominal efficiency of the Celestion “Blue”, established in the RC and
with the speaker mounted in the VOX AD60-VT enclosure. Certainly impressive but not at all
unique: the thin lines in the figure belong to the competition issued by the same manufacturer
and behaving similarly efficient. The new neodymium speaker (“Neodog”, uppermost curve)
even steps up the game. The figure on the right, however, shoes that the efficiency may be
smaller, as well: only the JBL-box with its 12”-speaker weighing in at 9 kg can reasonably
keep up – the other two speaker boxes were obviously were optimized using other criteria.
Fig. 11.58: Left: Nominal efficiency of the Celestion “Blue”. Thin lines: 4 further Celestion 12” speakers for
comparison: Neodog, Vintage-30, G12-80, G12-30H. Right: Full-range speaker-boxes. The “nominal efficiency”
was established for the specified nominal impedance, irrespective of the actual speaker impedance.
Let us quickly discuss, using two examples (Fig. 11.59), the question whether speakers using
Alnico-magnets are “louder” or “deliver more treble” compared to speakers deploying
ceramic magnets. P12-R and L-122 (both featuring Alnico magnets) have a smaller efficiency
than the Vitage-30 (ceramic magnet). The Celestin “Blue” (Alnico), however, shows a higher
efficiency than its ceramic-fitter competitor Eminence L-125. Besides the magnet material,
mainly the magnet size and the membrane are of importance – the “inspired Alnico sound
characteristics” are nothing but vapid advertisement.
Fig. 11.59: Nominal efficiency as in Fig. 11.58, comparison Alnico- vs. ceramic-magnets.

We now turn to the correctly defined efficiency, i.e. the ratio between emitted and received
active power. Again the RC is used, with its special characteristics. As depicted in Fig. 11.60,
the real part of the electrical impedance differs from the nominal impedance in particular at
the resonances points 95 Hz and 190 Hz, and in the high-frequency region. Hence in these
areas the loudspeaker efficiency is higher than the “nominal efficiency” determined relative to
the nominal impedance (8 Ω). The differences are clearly visible but may be ignored when
aiming for a rough orientation. This approach may be allowable even more so because all
12”-speakers investigated here showed similar frequency responses of the impedance. Merely
at the main resonance (around 95 Hz), the behavior may be substantially different. If this
range is of particular interest, exact impedance measurements are required.
Fig. 11.60: Real part of the electrical impedance (left), comparison between nominal efficiency (---) and actual
efficiency (1/3rd-octave average). VOX AD60-VT with original loudspeaker.
It has already been mentioned that the loudspeaker efficiency is not a constant but depends on
the acoustical environment. The VOX AD60-VT, a small combo, finds itself often placed on a
stool in its everyday stage work. The controls are better accessible that way, and the guitarist
can better hear him/herself. On the other hand, one could leave the VOX on the floor, as well
– the stored sound settings could be called up via a footswitch. How does the sound radiation
of these two modes of operation compare? Since the load impedance rises as the speaker
approaches a reflecting (floor-) surface, the level radiated at low frequencies will increase up
to 3 dB (Fig. 11.61). Closing the rear of the amp will have the opposite effect: the level
decreases across a wide frequency range, and only at very low frequencies there is a gain. The
latter is not generally desirable, because many guitarists will rather leave this frequency range
to the electric bass.
Fig. 11.61: Left: level gain when placing the VOX AD60-VT on the floor (compared to placement on a stool).
Right: level loss when closing up the rear of amplifier. Both measurements taken in the reverberation chamber.
The following page compares measurements in the AEC and the RC for several loudspeakers.
All 12”-speakers were measured mounted in the AD60-VT-enclosure.
© M. Zollner 2008 - 2014

Fig. 11.62: Comparison between measurements in the AEC (––––) and the RC (----), recalculated for 1W / 1m.
1/3rd-oct. analysis w/50% overlap (main/side 1/3rd octave), pink noise. Ordinate: sound pressure level dBSPL.
The measurements for the first 5 lines of figures were done using the AD60-VT-enclosure.
The thin angled lines in the figures are not target-curves but serve for orientation only.

The frequency responses shown in Fig. 11.62 indicate common characteristics due to the
enclosure (and possibly due to similar constructional details in the speakers), and also show
differences that have their reason in the different membrane designs. The difference between
the AEC- and RC-measurements is of particular interest since here the directivity manifests
itself (Chapter 11.4). Two peculiarities need to be considered: 1) the speaker remained at the
same location in the RC, its radiation impedance therefore is highly room-specific, 2) the
reference direction in the AEC was always 0° even if more sound power was radiated in other
directions – therefore, negative directivity is possible. It has already been elaborated that a
directivity of d = 0 dB does not always imply spherical radiation.
Given the measurement curves in Fig. 11.62, it should be mentioned once again that for guitar
loudspeakers, different optimization-guidelines are valid compared to e.g. a studio monitor.
With slight exaggeration, we could state: if the efficiency is high enough, the rest comes
together by itself. A Canton Quinto will not make the hard-rocking player happy at all – it just
ain’t no box for guitar. A single Vintage.30 will generate at its maximum permitted power (60
W) up to 123 dB at 1 m distance, while the Quinto will manage only 102 dB at maximum
power. In absolute terms, we would have to feed the Quinto with 100 times its maximum
power to make it compete with the Vintage-30. Conversely, the Vintage-30 would be utterly
out of its depths as a studio monitor, with a way-too-unbalanced frequency response. None of
the instrument speakers analyzed in Fig. 11.62 could be attributed a “bad” frequency response
– the peaks and dips are typical for the genre, with one guitar player preferring this and the
other player preferring that.
Measuring frequency responses aids in objectively determining differences and similarities –

but it can not replace a listening test. From the measured curves we can derive general
statements about efficiency and therefore about loudness; and we may obtain some very
general ideas about the sound: the pronounced 1.5-kHz-dip combined with the 3-kHz-peak of
a G12-H will clearly shape the sound. However, whether the Celestion “Blue” also entered in
the diagram for comparison will sound better or worse – the diagrams cannot say anything
about that. The trade business has masterfully understood how to fuel the flames of “tuning”
and “retrofitting”: the guitarist who is unhappy with sound of his rig will find so many clever
statements suggesting that changing the pickups or the pots or the loudspeaker will push
him/her into the professional realm. The swapping of components may be purposeful, it the
original parts were truly substandard. On the other hand, whether swapping a G12–H (at 119
€) for a “Heritage” G12-H (195 €) will transform scrap into Hendrix-like sound – that is more
than just doubtful. This author had the exciting pleasure and privilege of ear-witnessing (in
the front row at a perceived 150 dB) the Guv’nor JH letting loose heaven and hell with two
Marshall stacks in the Congress Hall in Munich – but had the master decided to present the
encore via a wall of AC-30’s … that would have been (very) fair enough, as well. It’s in the
fingers – we need to be reminded of that fact again and again.
© M. Zollner 2008 - 2014

11.6 Non-linear Distortion
In communication engineering we very carefully distinguish between linear and non-linear

signal distortions: a linear system generates linear distortion exclusively, and a non-linear
system (as far as it is free of memory) generates only non-linear distortion. Generally, one
seeks to avoid mixing up linear and non-linear effects by defining sub-systems that
individually generate purely linear or purely non-linear distortion.
In a linear system (e.g. an amplifier) the principles of proportionality, superposition and

absence of sources hold. The latter characteristic is easily explained: where there’s no input
signal, there’s no output signal. For a (non-zero) output signal , a matching (non-zero) input
signal must exist. If is doubled, must double, as well – that’s proportionality. We
quickly realize that the “linear function”, as used in mathematics and defined by the linear
equation y = k⋅x + m, will fulfill the requirements of proportionality and freedom of sources
only if m is zero. The law of superposition requires that the mapping of a sum must equal the
sum of the mapped summands. Thus: y = T{x} represents the mapping of the input signal x
onto the output signal y. If the sum of two signals is fed to the system input, the following
must hold in a linear system:
Superposition in the linear system
Proportionality and absence of sources alone do not suffice to specify linear behavior, as
shown by the example of an ideal full-wave rectifier (that reverses the sign of a negative
input signal): this device is source-free, and an n-fold input signal is matched with an n-fold
output signal – but as soon as a further signal (e.g. a DC voltage) is added at the input, the
waveform of the output changes … the rectifier is non-linear.
It is tempting to go and reduce the linear system to the matching-formula y = k⋅x; however,
this would unduly exclude the group of differential equations. A system that maps the speed
of a mass onto its acceleration is a (time-related) differentiator♣ . This system meets the
requirements of absence of sources [d/dt(0) = 0], of proportionality [d/dt(k⋅x) = k⋅dx/dt], and
of superposition: d/dt(ξ + µ) = dξ/dt + dµ/dt. The differentiator is a linear system in spite of
the fact that its sinus-transfer characteristic is not a straight line but an ellipse. Typically, the
linear distortion generated by a linear system is specified for sinusoidal drive-signals as
amplitude- and phase-distortion (or delay-time distortion), and is graphically represented as
amplitude-frequency-response and phase-frequency-response. The bass-cut generated by an
RC-highpass is a linear signal distortion, as is the presence boost of an equalizer (that of
course must not be overdriven). Reacting to an impulse-like excitation, the resonant circuit of
an equalizer will ring (theoretically for an indefinite time). Without a doubt, this is a signal
distortion – but a linear one. Unfortunately, there is often a lack of distinction between linear
and non-linear distortion, especially when it comes to loudspeaker characteristics discussed in
popular “science”. Non-linear distortion results if a system fails to fulfill one of the above
mentioned linearity criteria – this system is then non-linear. Whenever possible, we try to
separate linear and non-linear distortion into subsystem (possibly only existing as a model): a
linear subsystem described by its “straight” characteristic, and a (memory-free) non-linear
subsystem defined by its “curved” transmission characteristic.
♣
The formula-representation is meant to save space: it may not meet the expectations of all mathematicians.

11.6 Non-linear distortion 11-53
It is, however, in many cases not possible to divide a real system into one linear and one non-
linear subsystem: since in non-linear systems the (commutative) exchangeability is not there
anymore, it may be that a plurality of subsystems is required, and the corresponding
description may become highly complicated. The dynamic loudspeaker, as well, includes
several non-linearities that do not allow themselves to be modeled in one and the same
subsystem: the displacement-dependent force-factor (aka. transducer coefficient Bl), the
displacement-dependent stiffness of the membrane-suspension, and the inductance that is also
displacement-dependent. If the loudspeaker is mounted in an airtight enclosure, the non-liner
stiffness of the air-suspension caused by the enclosure weighs in, as well. With the speaker
mounted in a ported enclosure, the airflow within the port-tunnel introduces non-linearity. All
these non-linearities generate a non-linear transmission characteristic but also cause a reaction
on the electrical side and make for a strongly non-linear loudspeaker-impedance. In addition,
we have non-linearity generated in the amplifier and the output transformer (if one is present).
All in all we get a complex system with coupled non-linearities – and one that produces
pronounced linear distortion to top it all.
For a loudspeaker mounted in a cabinet that is open to the rear, we may neglect the non-
linearity of the air. At low frequencies, we may – to start with – dispense with a consideration
of the inductance so that as a first approximation, a non-linear mechanical subsystem and a
non-linear magnetic subsystem remain. The mechanical non-linearity is found in the stiffness
of the membrane-suspension, i.e. the inner centering (spider) and the outer fastening
(surround). As the membrane is deflected slowly, force is directed against a progressive
spring with its stiffness increasing as the displacement increases. The stiffness is a system-
variable while the displacement is a signal-variable. If a system-variable is dependent on the
signal, we always have a non-linear system. The non-linearity in the magnetic system clearly
is the transducer coefficient (the force-factor): as system-variable Bl, it takes care of the
proportionality between current and Lorentz force: F = Bl ⋅ I. However, this proportionality
requires that the system variable Bl is independent of the signal – specifically independent of
the displacement. That is not the case here: with increasing displacement, the coil moves out
of the magnetic field and therefore Bl decreases. A further effect may play a role in this
scenario: two magnetic fields superimpose as current flows. One is constituted by the
permanent field generated by the permanent magnet, the other is the AC field surrounding the
voice-coil wire. Because the ferro-magnetic parts located in the magnetic circuit all show a
non-linear characteristic (the specific magnetic conductance µ is field-dependent),
“modulations of the magnetic field” may result. Some manufacturers seek to decrease the
effect via short-circuit rings while others do not do anything about it, regarding it as typical. It
is here where the peculiarities of the guitar loudspeaker begin: while for HiFi-speakers there
is consensus that non-linearity must be as small as possible, opinions diverge considerably
when it comes to guitar speakers. You may hear (or read) on the one hand that the guitar
speaker is, after all, a loudspeaker too (correct), and thus what has been taught in the HiFi-
domain for decades cannot be wrong (??). On the other hand, (positive) reviews including
evaluations such as “dirty midrange” give rise to some hope that at least a few designers have
recognized that sound-shaping function of the guitar loudspeaker.
We must not fail to mention here though, that not only among the manufacturers, but also
among the players multiple opinions abound. You get the Jazz-dude who brutally chokes the
hard-won brilliance of the guitar by bottoming out the tone control, the Country-picker with
his piercing treble, the crunching-along Blues-man, the chainsaw-ing Metal-ist, the Jack-of-
all-Trades cover-guy, and the folksy oom-pah-strummer. A consolidated drive towards
standardized loudspeaker distortion may not be expected given such a heterogeneous
population and mix of opinions.
© M. Zollner 2008 - 2014

Here’s an example taken from loudspeaker history: JBL, the renowned American speaker
manufacturer, looks back on a long tradition as supplier for cinemas, recording studios, and
living rooms. Nothing but High-Fidelity – non-linear distortion is marginal as a matter of
course: ... "low distortion which has been always associated with JBL products". In the early
60’s the demand for instrument speakers grows, and at JBL, a tried and trusted workhorse, the
D-130, is modified to become the D-130F. The changes mostly relate to the air gap that was –
according to statements by the designer (Harvey Gerst) – slightly enlarged in order to obviate
damage. And then there’s the designation: F is for Fender, the largest customer. Years later,
the K-130 follows with double the power capacity compared to its predecessor but still "clean
at any volume level" (a quality that probably would not have always applied to the associated
musicians). Both the D-130F and the K-130 were fitted with Alnico magnets, but the next
generation – the E-Series – received ceramic magnets. This prompted JBL-mastermind John
Eargle to state that Alnico was known for its "inherently low distortion performance".
However, according to him, the new E-Series is even better: "The improvement has been in
reducing second harmonic distortion", obtained with the "symmetrical field geometry". Given
this upgrade, the loudspeaker is eminently suitable "for vocals – and guitar". Presumably, this
further added to the already long list of JBL-users shown in the adverts. Due to space-
restrictions, this list cannot be commemorated in its full extent here, but the following may
serve as an excerpt: Count Basie, Harry Belafonte, Tony Curtis, Sammy Davis jr., Doris Day,
the unforgotten Carmen Dragon, Duke Ellington, Ella Fitzgerald, Hugh Hefner (!), Dean
Martin, Frank Sinatra, not to forget Richard Nixon and “The Duke” John Wayne♣. Global
super-stars, all of them – and all of them JBL-users. Such a feat of course calls the
competition into the arena. And thus it was that Electro-Voice retaliates with a big swing,
proclaiming: "Symmetrical magnet gap structures have been promoted as desirable in a
guitar speaker. We have found this to be a fallacy". Because: "A coil moving in an
asymmetrical magnetic gap will generate a mixture of odd and even harmonics, resulting in a
more complex, richer sound." To each his own … there’s no accounting for taste.
So, let’s not begrudge The Duke the undistorted JBL-sound of his electric guitar (hm … still
thinking about that one …), and Joe Bonamassa his EV-sound chirping from the 4x12’s –
beauty is in the eye of the beholder. What can be said about magnetic non-linearity from a
scientific angle? If you leave the pole-core (the cylinder in the interior of the voice coil)
formed as a cylinder over its full length, as shown in Fig. 11.63, then an asymmetric scatter-
field will result: the shape of the field above the coil is different from the shape below it.
Reducing the core-diameter in the lower section, though – as it is shown in exaggerated
fashion in the figure – will render the two stray-fields more symmetric. The result is that a
symmetric Lorentz-force acts on the voice coil for both positive and negative displacement.
As already mentioned, this force depends on current and displacement. While the current-
dependency is desired, the displacement-dependency is not, because it generates non-linear
distortion. For a symmetric field, the distortion is of even-order (even function) – given
asymmetry distortion, odd-order also weighs in.
Fig. 11.63: Different designs of the pole-

core. On the left it is purely cylindrical, on
the right offset. The scatter-field generated
outside of the air gap depends on the
geometry of the pole-core.
♣
From JBL's 1968 loudspeaker brochure.

The field-limiting at both ends leads to a degressive, clipped current/force-characteristic that

leads some people to conclude that the course of the oscillation would now also be limited –
similar to an overdriven amplifier. This assumption, however, overlooks that the force is
mapped onto the displacement (via Hooke’s law) only in the range below the resonance. At
resonance the membrane acts (in the simple model) as a damper, and above the resonance it
acts as a mass [3]. If, in the range above the resonance, the motive force becomes weaker at
strong displacement, this reduces the acceleration primarily. Of course there will be an effect
on the displacement, as well, but: displacement and acceleration are in opposite phase. Or, to
be somewhat more precise: the acceleration is the second derivative of the displacement.
Analytically, this leads to a non-linear differential equation that can be solved approximately
– with a big effort, however (the system is not just weakly but extremely non-linear).
The qualitative effects of the inhomogeneity may be studied rather well using the simple
membrane model [3]. If we reduce the membrane to spring, mass and damper, and the
electric side to the ohmic voice-coil resistance, we have a 2nd-order system that may be
described via the frequency of the pole (resonance frequency) and the Q-factor of the pole.
The pole-frequency unambiguously results from the stiffness and the mass, while two limiting
cases are of importance for the Q-factor of the pole: open circuit and shot circuit on the
electrical side. Given the open circuit, no current can flow and consequently the magnetic
field will exert no force onto the membrane. The Q-factor of the pole depends solely on the
mechanical parameters: . The purely mechanic dampening of the membrane is
relatively small such that the Q-factor of the pole is considerably larger than 1 (5 is not
uncommon). As the voice coil is shorted (or as an amplifier with a very small internal
impedance is connected), the voice-coil resistance that is transformed from the electrical to
the mechanical side acts as an additional dampening♣, and the Q-factor of the pole drops
below 1. Since the electromechanical coupling becomes smaller at large displacement (due to
the inhomogenities in the field), the membrane-dampening decreases with increasing drive
levels – the displacement tends to become too large, and not too small as it would with
degressive limiting [3, Chapter 6.2.3].
In the asymmetrical magnetic field, the reset-forces acting at the extremes of the membrane-
displacement are unequal (in terms of magnitude), and therefore the average force is not zero.
A steady force of the frequency 0 Hz results that pushes the membrane out of its neutral
position in the direction of the weaker of the two fringe-fields. Because in reality the two
fringe-fields are never exactly identical, this effect always occurs: a small asymmetry suffices
to make the membrane wander slightly from the idle position. This enhances the lack of
symmetry, and the membrane continues to wander off – it is only stabilized by the onset of
the counter-force exerted by the membrane-suspension. Therefore, 2nd-order distortion is to be
expected in the range above the resonance – even if there is a symmetrical layout of the
magnetic field. “Field-modulation” is a further source of 2nd-order distortion: part of the
magnetic field generated by the flowing current superimposes onto the steady field of the
permanent magnet, i.e. the flux density therefore fluctuates in sync to the excitation current.
Because of this, the force obtains a share that is dependent on the square of the current – and
that implies 2nd-order distortion. Another way of explaining the effect: the flowing current
generates attracting forces between neighboring ferromagnetic parts (through which the field
flows). These attracting forces are independent of the sign and therefore generate even-order
distortion (just like a rectifier). Relief, if at all sought, could come in the form of a short-
circuit ring. It forces the AC field out of the magnetic circuit, and the 2nd-order distortion
decreases.
♣
The eddy-current brake in lorries and trains works based on a similar principle.
© M. Zollner 2008 - 2014

Analyzing the current fed from a stiff voltage source will give a first impression regarding the
linearity (or non-linearity) of the loudspeaker. The mechanical membrane-impedance F/v is
transformed to the electrical side with the square of the transducer coefficient. If we disregard
the inductance, the electrical impedance consists of two components: the voice-coil resistance
(e.g. 6 Ω), and the transformed mechanical impedance. Any non-linearity in the electrical
impedance (given a stiff voltage source these would be current distortions) will consist of
two components: a non-linear transducer coefficient (Bl), and/or a non-linear membrane
impedance. In a loudspeaker, both these components are present: both the stiffness of the
membrane-suspension and the transducer coefficient are displacement-dependent. The
current-curves for the operation close to resonance are shown in Fig. 11.64, with the
distortion being very significant. It should be noted that the voltage amounts to merely 10
Veff, i.e. nominally only 12.5 W for this 8-Ω-speaker (deployed in a 60-W-amplifier).
Moreover, since the impedance will be at its maximum at resonance, the power taken by the
speaker will be even (much) less – we are far away from any undue overload situations. The
curves reveal a strong 2nd-order distortion. The amplitude of the 2nd harmonic rises to up to
67% of the 1st harmonic; this would correspond to a harmonic distortion of k2 = 56% (the
approximation U2 / U1 should not be used anymore at such high distortion levels). It is beyond
the aim of this chapter to localize or separate the individual roots of these distortions – the
effort would grow too big. Rather, we will present comparative distortion measurements;
these will consistently show that all investigated loudspeakers are strongly non-linear systems
even at very moderate drive levels. Not that guitar players would generally be adverse to such
characteristics …
Fig. 11.64: Time-functions of the loudspeaker-current fed from a stiff voltage source, U = 10V; VOX AD60-VT.
The frequency responses of the distortion (Fig. 11.65) show that the maximum distortion of
the current happens at the main resonance (116 Hz); it is here that the displacement is at its
maximum. There are two reasons that 2nd-order distortion can rise to such heights: at
resonance, the 2nd-order harmonic of the current is highest (due to the mentioned non-
linearity), and at the same time the overall current becomes smallest (due to the rise in the
impedance. The difference of the two levels (the distortion dampening) therefore has a
pronounced maximum here. However, the current-distortion describes predominantly the
electrical behavior – non-linearity in the sound radiation needs to be analyzed separately.
Fig. 11.65: Frequency-dependence of harmonic distortion in the current when fed from a stiff voltage source (as
in Fig.. 11.64). The drive-level-dependent shift of the maximum is a result of the strong non-linearity.

In order to measure the (non-linear) distortion in the sound pressure, the speaker was operated
from a stiff voltage source and mounted in the VOX AD-60VT enclosure. Measurements
were taken in the AEC, with the microphone on axis at a distance of 3 m (Fig. 11.66). The
results are not untypical for a dynamic woofer: at low frequencies very strong distortion is
generated. Then, from 70 Hz up, the 3rd-order distortion drops off faster than the 2nd-order
distortion, and above 150 Hz, the THD remains below about 1%. Compared to the analysis of
the current (Fig. 11.65), the distortion has mostly increased.
Fig. 11.66: Frequency responses of SPL (left) und distortion level. VOX AD60-VT, U = 10V.
However, a THD of 1% is not the professed aim for a guitar loudspeaker – some speakers
easily reach ten times that distortion: Fig. 11.67 illustrates measurements with another
Celestion loudspeaker, operated with the same voltage and in the AD-60VT-enclosure: the
Celestion “Blue”. This speaker is not broken – far from it; it’s just that the input voltage of 10
V already pushes it close to the borders of its power capacity (15 W). On the other hand, this
should not be taken as evidence that a THD of 10% would be typical for getting near to the
power limits: the Vintage-30 speaker (depicted below the “Blue”) is specified at 60 W and, at
10 V, distorts similarly to the “Blue”. Since the concept behind the Vintage-30 is that it
should be a descendant of the “Blue”, it is only sequacious that it should distort like the latter.
Fig. 11.67: 2nd- and 3rd-order distortion. Enclosure: VOX AD-60VT, U = 10V, d = 3m.
Upper row: Celestion "Blue" (Pmax = 15W); lower row: Celestion Vintage-30 (Pmax = 60W).
© M. Zollner 2008 - 2014

Documentations on loudspeaker non-linearity are often limited to measurements of the

harmonic distortion; this may have to do with the fact that such measurements are a standard
tool in systems analysis. Since Brüel&Kjaer has released its legendary instrumentation-
combination of the 2010/1902-devices, difference-frequency measurements are also common
for band-limited systems – but there is a further distortion-mechanism that is found especially
in loudspeakers: sub-harmonics. This term means to describe the generation of distortion
tones that have a frequency lower than the excitation frequency, e.g. f/2 or f/4. Fig. 11.68
depicts, accordingly, two spectra derived from the sound pressure. A sinusoidal voltage (10
V) was imprinted at the loudspeaker connectors, with f = 1.6 and 1.5 kHz. Highlighted in grey
are those spectral lines (broadened by leakage) that could be expected as regular “harmonic
distortion”; in addition, however, we see a sub-harmonic developing, and frequency-multiples
of it. The double-peaks in the right-hand diagram point to fast time-variant processes: the
spectrum resulted from a sweep, and the “sub-harmonic distortions” change their level very
fast.
Fig. 11.68: Sub-harmonics at half (left), and a quarter (right) of the excitation frequency.
Such sub-harmonics appear – if at all – only in small frequency ranges. Fig. 11.69 shows two
spectrograms representing the level as grey value across the f/t-plane. The bottom rising
curve belongs to the level of the first harmonic, the levels of the higher harmonics follow
above. The grey dots or groups of dots appearing in the right half of the diagram point to sub-
harmonics (or their frequency-multiples). The speaker analyzed on the left (Jensen P12-N)
shows sub-harmonic distortion only at an excitation frequency of about 1760 Hz, while the
speaker on the right (Celestion G12-Century) features it in several ranges from 920 Hz up.
Both speakers have an impedance of 8 Ω and both were measured at 10 V. The C12-N is
specified with a power capacity of 50 W, and the G12 at 80 W – neither speaker is therefore
operated close to any power limit.
Fig. 11.69: Sweep-spectrograms f = 50 – 5000 Hz, U = 10V. The speaker analyzed on the left generates sub-
harmonics only at about 1760 Hz while the speaker on the right does so in several frequency ranges.
Abscissa-scaling: sweep-time = 0 – 30 s; ordinate-scaling: frequency = 0 – 7 kHz.

The generation of sub-harmonic distortion cannot be explained merely via a curved

transmission characteristic. As Fig. 11.70 shows, a tilting vibration of half the frequency is
superimposed. Mathematics laconically (and correctly) explains such phenomena with
“solution of a non-linear/time-variant differential equation”, physics offer “parametrically
excited eigen-oscillations of a system with time-variant system-variables”. Time-variant
quantities are indeed easily imaginable: the membrane deforms, and the location-dependent
stiffness of the membrane is certain to be dependent on load – and therefore on time.
Moreover, the oscillation of the membrane by no means needs to be one-dimensional: tilting-
and tumbling-movements are possible, and the overall system is of a complicated, non-linear
nature. We may expect sections of the membrane oscillating with the same phase – but of
course not the whole area; there will be phase shifts, and since the system parameters are
time-variant, most probably there will be time-variant phase shifts, as well. Simple models
fail here, e.g. since already the superposition principle may not be applied anymore.
Fig. 11.70: Time function of sound pressure in a sub-harmonically distorted sinusoidal tone: f = 1,6 kHz.
Fig. 11.71 is targeted at showing at showing another example of the complexity of sub-
harmonic distortion: above 1.5 kHz, this loudspeaker generates sub-harmonics not only at half
the excitation frequency, but – among others – also at f/4 and f/5 (and at the multiples).
Fig. 11.71: Jensen P12-R, sweep-spectrogram, f = 800 – 2500 Hz, U = 10V.
© M. Zollner 2008 - 2014

The levels of the sub-harmonics follow their own laws, not even showing the power laws to
be expected for standard models. IN Fig. 11.72, the level of sub-harmonic (f/2) is shown
dependent on the level of the primary tone (f). For primary tones below a certain threshold
(here at just under 108 dB), there is no sub-harmonic at all. Going across that threshold, the
sub-harmonic builds up. As we reduce the level of the primary tone below the threshold value,
the level of the sub-harmonic remains constant first – only as the primary tone falls below
about 104 dB, the sub-harmonic disappears again.
Fig. 11.72: Level-hysteresis (left), distortion spectrum (right). Celestion Vintage-30, f = 844 Hz.
Fig. 11.73 more precisely depicts the evolution of the level for three frequencies. The
generator level rises by 25 dB during 30 s: the corresponding measured sound pressure levels
are shown. At 1081 Hz, no sub-harmonic is created and the levels grow monotonously. At
about 1.3 kHz, however, a sub-harmonic appears around -11 dB (50W / 12.5 = 4 W), and this
has effects on all measured sub-harmonics. From -9 dB, there are audible beats.
Fig. 11.73: Sum SPL L and distortion level, Eminence L-105; 0dB = maximum power.
Lower left: for f = 1303 Hz, the level development of the sub-harmonic (f/2) is shown.
The following Fig. 11.74 depicts the non-linear behavior of several loudspeakers in an
overview; supplementary measurement data are added in the last of the three diagrams.

Fig. 11.74a: Distortion suppression (harmonic distortion attenuation) ak2, ak3 of various loudspeakers. The
distortion suppression of the sub-harmonic is, respectively, included at the upper right in the right-hand.
© M. Zollner 2008 - 2014

Fig. 11.74b: Distortion suppression (harmonic distortion attenuation) ak2, ak3 of various loudspeakers. The
distortion suppression of the sub-harmonic is, respectively, included at the upper right in the right-hand.

Fig. 11.74c: Distortion suppression (harmonic distortion attenuation) ak2, ak3 of various loudspeakers. The
distortion suppression of the sub-harmonic is, respectively, included at the upper right in the right-hand. Given
the nominal impedance of 8 Ω, the voltage (10 Veff from a stiff voltage source) results in a power of 12.5 W. All
measurements were taken in the AEC, with the 12”-speakers mounted in the VOX AD-60VT enclosure, the 15”-
speakers mounted in an air-tight enclosure measuring 36x74x40 cm3, and the 10”-speakers in an air-tight
enclosure measuring 39x39x25 cm3.
© M. Zollner 2008 - 2014

11.7 Alnico vs. ceramic magnets
Alnico! Guitar players feel that this word connects them to the innermost circle of magic.
Pickups? Only those fitted with Alnico. Loudspeakers? Same! For ceramic magnets „don’t
sound right“ … sometime, somewhere, an enlightened shaman has expressed this, and his
disciples keep repeating it all over the world. The Celestion Blue, "the world's first dedicated
guitar loudspeaker", sported – of course! – an Alnico-magnet. As the manufacturer that
produces "the finest guitar loudspeakers that money can buy", you owe that much to yourself.
Give me Alnico or give me death! However, as we take off the rose-colored glasses of the ad-
writer, things become less euphemistic: Alnico-magnets were the industry-standard to
generate strong magnetic fields. Carbon-steel magnets were produced until 1910 [21], from
1917 there were cobalt-steel magnets, and from the mid-1930’s we see magnetic alloys that
contain, besides steel, also aluminum (Al), nickel (Ni) and cobalt (Co): AlNiCo-magnets.
They appear in many compositions designated with numbers and letters, and even according
to prescription, if more precision is required: 8% Al, 14% Ni, 24% Co, 3% Cu, the rest Fe.
However, the effect of a magnet is not only the result of the formula – it’s the crystalline
structure that does it. So if the label says Alnico-5 on two magnets, the impact may still be
different. For this reason, there are subgroups such as e.g. Alnico 5-A, or 5-B, or 5-C, 5-7, 5-
BDG, 5-ABDG, or whatever their designation may be. To trust the conjecture that there
would be a magnet-material named Alnico-5 that generates that wonderful “vintage sound” –
that’s believing in a fairytale. In fact, there is a multitude of Alnico-5 materials featuring
rather different characteristics. We must also not forget that, because of competition amongst
manufacturers, we also have Ticonal, Nialco, and Coalnimax. All these materials have a very
high remanent flux density of between 1.2 – 1.35 T, and therefore are of excellent suitability
for loudspeakers. However, as a side effect of WW II, supply bottlenecks and restrictions on
“metals needed for the war effort” with corresponding cost-explosions happened, and so the
manufacturers were very happy that low-cost ceramic magnets became available as
replacements. Guitarists were less happy, because “ceramic does not have the sound of
Alnico”. Well then, what makes ceramic magnets so distinct over their ceramic imitators?
Assuming the same volume, Alnico-magnets are stronger than ceramic magnets. That is no
knock-out criterion, though, because it only pushes up the weight of ceramic-magnet
loudspeakers. The flux density in the air-gap is not limited by the magnet (that could be
enlarged almost at will) but by the saturation of the field-guiding pole-plates. Considering
that, for operation in the optimum operating point, Alnico magnets need to be oblong while
ceramic magnets need to be disc-shaped, both materials can serve equally well to reach
similarly high flux densities (and flux). Hearsay has it, however, that, as the material became
warm during operation of the first speakers fitted with ceramic magnets, the flux density
dropped. Indeed, the flux density decreases with almost 0.2% per °C, and depending on the
material, 100 °C are not out of the question when pushing the speaker. For voice-coil carriers
made of paper, this was kind of a maximum allowable temperature, anyway. However, as
high-temperature-resistant materials (Nomex, Kapton, glass fibre) were introduced the
maximum temperature for the coil rose to above 250 °C, and at that point it is conceivable
that some ceramic magnets had problems. Corresponding difficulties have been largely
overcome by now, and industry offers ceramic magnets that tolerate loudspeaker-typical
temperatures. Also, the heat generated in the voice-coil does not directly flow fully into the
magnet material, and the magnet does not become as hot as the voice-coil. And incidentally,
the main allegation towards the ceramic-faction is not that it’s weak-kneed but that there’s a
sound-deficit, somehow, kinda. Alnico has that "vintage" sound, and thus sounds good.
Vintage, that’s more treble, or (depending on the source) less treble; either way: simply better.

11.7 Alnico vs. ceramic magnets 11-65
Eminence, the world's largest loudspeaker manufacturing company, with the finest voice
coils in the industry, explains the Alnico-sound as: "warm, bluesy tone". Jensen, on the other
hand, the inventor of the loudspeaker, sees "their sparkling trebles" as the Alnico-
characteristic. JBL, world’s leading loudspeaker manufacturer, defines Alnico via "it's low
distortion performance", Jensen does so via "their dirty midrange". Everybody can find
his/her own thing with Alnico.
Fig. 11.75: B/H-characteristics of various Alnico-5 magnets [22, 23]. 1Oe = 80A/m, 10kG = 1T.
In Fig. 11.75 we see hysteresis curves of various Anico-5 magnets in comparison to three
ceramic magnets. The remanent flux density of regular Alnico-5 magnets is just short of four
times that of anisotropic ceramic magnets – conversely, the coercitive field strength of the
latter is 3 times larger that that of the Alnico-5 magnets. Again, just to be clear: there is no
one Alnico-5 magnet nor is there the ceramic magnet, and moreover the data on remanence or
coercitivity allow merely for approximate conclusions regarding the operating point. A
comparison can highlight the fundamental differences of the two material groups. A l00-W-
lamp (10V/10A) is to be lit up; batteries of 1V/10A and 10V/1A are at our disposal. Whether
we connect 10 of the 10-V-batteries in parallel, or 10 of the 1-V-batteries in series does not
make any difference to start with – both variants enable the lamp to receive 10V/10A. That
does not mean that there are no differences at all anymore: the 10-V-batteries might be a bit
more expensive, or larger, or come from a country to which (despite unbelievable successes in
sports) economic relations are for the time being uncalled for … anyway, the normative force
of facts will call for the 1-V-batteries. The clever businessman will however still try to give
the 10-V-batteries a chance, and may for example advertise the source of their energy as
“directly from the sun” (sustainability is “in”). He might give 0,5% of the unfortunately 50%-
higher sales price back to the estranged country (with the imperative condition that an
English-language crèche is financed with the money). This diversification increases the
market share, makes for a nicer company-car, and enriches the world by another crèche.
© M. Zollner 2008 - 2014

To transfer this scenario to loudspeakers: the volume-specific energy of the magnetic field
(the energy density) w corresponds to half the product of flux density B and field strength H.
Alnico facilitates higher B-values than ceramic but does not reach the high field strengths of
the latter. To compensate, Alnico magnets need to be long and slender, ceramic magnets need
to be short and wide. That’s just like a series circuit compares to a parallel circuit. Both
magnet materials enable the realization of the required specific energy of the magnetic field:
ceramic corresponds to the standard, and Alnico sort of corresponds the crèche in the above
example.
Why exactly does the BH-product of a magnetic material have that kind of importance when
the definition of the transducer coefficient includes only B but not H? It is true, indeed: the
Lorentz-force depends – except for the length of the wire – only on the flux density B.
However, in air (as in air gap) the flux density is connected with H via µ0 such that inevitably
a specific H is connected to a correspondingly specific B. The Cu- or Al-winding also located
within the air gap does practically not change anything about that – because these materials
are not ferromagnetic. The product of the field strength in the air gap and the flux density in
the air gap happens to correspond exactly to double the energy density wL of the field within
the air gap. With the air-gap volume VL, the energy in the air gap computes to WL = wL⋅VL.
The energy must be made available by the magnet; for the ideal magnetic circuit, this holds:
WL = wL⋅VL = WM = wM⋅VM. Spelled out: magnet-energy = air-gap-energy. Within the
magnet, the formula wM = 0.5 ⋅ BM⋅HM holds; consequently for a small volume of the magnet,
the BH-product of the magnet needs to be as large as possible. As an example: for an air gap
of an area of 10 cm2 and a width of 1 mm, the air gap volume is to 1 cm3. Given B = 1.5 T, the
energy in the air gap is 0.9 J = 0.9 Ws. This value is not directly connected to the sound power
to be generated: one may imagine the magnetic field as a kind of catalyst that is necessary but
will not be used up. The radiated sound energy is not sourced from the magnetic field but
from the electrical energy (fed from the power amplifier). Assuming the BH-product
characterizing the magnet to be 45 kJ/m3 (not untypical for Alnico-5), a volume of the magnet
of 40 cm3 (or a magnet mass of 286 g) results. A ferrite magnet generating only as little as
BHmax = 22 kJ/m3 would require 81 cm3 (or 390 g). This would be the situation for the ideal
(i.e. loss-free) magnetic circuit. Alas, this idealization is not even approximately realistic, and
so the magnet needs to be bigger: for Alnico 2 – 3 times, for ferrite 3 - 4 times … or still
bigger, depending on the individual realization. To achieve comparable air-gap energy,
ceramic magnets are therefore larger and heavier than Alnico magnets. Still, any differences
in sound or efficiency cannot be substantiated that way.
The energy within the air gap is, however, only a first parameter in the electro-acoustical
transducer process. As already shown by Fig. 11.1, the shape of the magnet (long/slender vs.
short/wide) causes different geometries in the magnetic circuit, and from this shape, two
different behaviors may result in dynamic operation (i.e. given current-flow and
displacement). It is therefore not sufficient to merely check the static magnitudes in the air-
gap – the membrane is to move, after all. Indeed, there is a dynamic magnet parameter
showing differences: the so-called permanent permeability. In a permanent magnet, it
characterizes the B/H-relationship for small shifts in the operating point. For a field-change
forced by a current, the operating point does not move along the limit-curve of the hysteresis
but within it on a smaller slope. This slope is the permanent permeability, also called
reversible permeability. It is about 5 in Alnico-5 and about 1 in ceramic. These data are
relative permeabilities, i.e. for small field changes the ceramic magnet behaves like air while
Alnico is already perceivably ferromagnetic. Globally seen (for large changes in the field),
both magnets are of course ferromagnetic, but for differential considerations material-specific
differences emerge.

It is, however, insufficient to regard merely a differential magnet-parameter (the permanent

permeability) und predict differences in the operational behavior merely based on this. In the
respective operating point, not only the slopes of the hysteresis characteristics differ, but the
coordinate values, as well. Since, compared to the Alnico magnet, the ceramic magnet
features smaller B and larger H, a kind of transformation needs to be done via an area-
reduction: from the wide magnet cross-section to the comparably small air-gap cross-section.
This transformation will not only adapt B and H correspondingly, but also the slope of the
hysteresis such that the effective permanent permeabilities become closer to each other.
Whether they in fact become equal or whether differences still remain, depends on the
individual design, and on the all-decisive stray-flux.
This holds for both magnet materials: parameter-variations that may already result from
smallish construction-changes are so considerable that is of no purpose to generally speculate
about type-specific idiosyncrasies. Rather, measurements are to reveal typical differences –
if such exist to begin with. One quantity that is easy to measure and that gives indications
about differential field changes is the electrical impedance. Its high-frequency increase is
determined by the loudspeaker-inductance, and thus by the magnetic field. In Fig. 11.76 we
see the frequency responses of the impedances of several 12”-loudspeakers. On the left, only
marginal differences show up – although two loudspeakers with different magnets were
measured (Celestion “Blue” vs. G12-H). In the right-hand section, however, three Alnico-
loudspeakers were analyzed – and specifically here clear differences emerge. Conclusion:
there is no special “Alnico-impedance”.
Fig. 11.76: Frequency responses of the impedances of various 12”-loudspeakers (w/out enclosure).
Left: Alnico (–––) vs. ceramic (----). Right: 3 different Alnico-speakers (Celestion "Blue", Jensen P12-R, P12-N).
Impedance-measurements will only expose a small-signal characteristic. Loudspeakers will,

however, predominantly be operated at large signal levels, with high currents and often close
to the power limit. As already shown in Chapter 11.6, non-linear processes step into the
foreground here: the voice coil pushes into the fringe-regions of the magnetic field rendering
the transducer constant (the force-factor) dependent on the displacement. The membrane-
stiffness becomes displacement-dependent, as well, and the inductance shows non-linearity,
too. It is certainly possible that the secret of the dearly bought Alnicos lies in the specific non-
linearity, and that their harmonic distortion shows magnet-typical idiosyncrasies. The
magnet should however not be held responsible for any non-linearity of the membrane: that
the centering (the spider) becomes progressively stiffer with increasing displacement really
has nothing at all to do with the magnetic material. The displacement-dependent inductance,
on the other hand, is connected to the magnet, and the signal-dependent transducer constant is,
too. Both these non-linearities result from the magnetic field penetrating the voice coil, and
because this field is displacement-dependent, the transducer constant becomes signal-
dependent. The component of the electrical impedance that stems from the mechanics (that
would be everything except the Cu-resistance), consequently becomes non-linear.
© M. Zollner 2008 - 2014

A non-linear impedance can be measured by either feeding a sinusoidal current to it from a

stiff current source (“imprinting” the current) and measuring the voltage, or by feeding it with
a sinusoidal voltage from a stiff-voltage source (“imprinting” the voltage), and measuring the
current. The two principles lead to different results because there is no proportionality
anymore for non-linear systems. For the following measurements, the voltage was imprinted.
Mostly, 10 V were applied, corresponding to a nominal 12.5-W-load for an 8-Ω-speaker. The
loudspeakers were not mounted in any enclosure, this leading to larger membrane
displacements compared to installation within an enclosure. The harmonic distortion of the
loudspeaker current was analyzed, in particular the 2nd- and 3rd- order distortion. It is shown
as distortion dampening ak2 and ak3 (Fig. 11.77). 60 dB 0.1%, 40 dB 1%, 20 dB 10%.
Fig. 11.77: Non-linear distortion of the loudspeaker current for sinusoidal imprinted voltage (10V).
Alnico = Celestion "Blue", ceramic (“Keramik”) = Celestion G12-H.
In this figure we clearly see significant differences: the 2nd-order harmonic distortion differs
by a factor of 3, the 3rd-order distortion even up to a factor of 10! The 2nd-order distortion
generally dominates over the 3rd-order distortion, but their frequency dependency differs
specifically depending on the loudspeaker. In the range of the main resonance, the Alnico-
speaker distorts more than the ceramic-speaker; however, at higher frequencies the
differences should be treated with caution. Also, distortion generated by the guitar amplifier –
as a rule rather significant – should be considered.
It is only a small step from the measurements shown in Fig. 11.77 to statements such as:
Alnico-loudspeakers distort more than ceramic-loudspeakers. That is, however, not really
entirely correct since from 124 Hz upwards we see the 2nd-order distortion dominating in the
ceramic speaker. So, what catchy message should we take home from these measurements?
Best would be none – the comparison between two loudspeakers cannot be taken as a
significant sample. Fig. 11.78 offers supplemental analyses: a Jensen C12-N (Alnico) is
compared to a Jensen C12-N (ceramic). Now, suddenly, the situation is reversed: the k2 of the
Alnico-speaker is smaller than that of the ceramic speaker.
Fig. 11.78: Non-linear distortion of the loudspeaker-current given sinusoidal imprinted voltage (10V).
Alnico = Jensen P12-N, ceramic (“Keramik”) = Jensen C12-N.

In fact, the two Jensen-loudspeakers (Fig. 11.78) represent an ideal pair: both are sourced
from the same manufacturer, both have a 12”-diameter, both play in the same power-league:
50 W, 1,5"-voice-coil. Only the material of the magnets is different: ceramic (C12-N) vs.
Alnico (P12-N). O.k. – the price is also different … we understand: the expensive cobalt.
What some of us do not understand: why do the resonance frequencies of these two speakers
(bought at the same time) differ, too? By 56%, after all – specifically 120 Hz (C12-N) vs. 77
Hz (P12-N). Don’t start with “the magnet change might retune the resonance” – the
mechanics do not require any magnetic field for that. At least the stiffness of the membranes
is very different, as a simple push with the fingertip confirms. So, there’s not just another
magnet included, but the membranes are entirely different, as well! One can imagine the kind
of “wisdom” that results if, after comparing these two loudspeakers, musicians post their
findings on the internet. Without a doubt, there are type-specific differences in the non-linear
behavior of the loudspeakers, but it is not possible to derive any Alnico-specific characteristic
from these.
Checking out two Celestion loudspeakers may serve as a counter-example to the above
comparison: Vintage-30 (ceramic magnet) vs. “Blue” (Alnico-magnet). Fig. 11.79 indicates
the corresponding comparison-analyses. Up to 250 Hz, we in fact recognize merely a slightly
different resonance frequency in the k2, and in the frequency range above the effects of the
modes of partial oscillations can be seen. In the k3, the differences are somewhat larger but by
no means classifiable as a characteristic. But now it gets really interesting: in the second row
of the figure, two Alnico loudspeakers, specifically two Celestion “Blue” bought at the same
time, are compared. The differences between these two speaker-specimen (both Alnico, both
of the same construction!) are, as a whole, larger than the differences between the differences
found between the Alnico- and the ceramic-speaker shown in the upper row in the figure!
Fig. 11.79: Non-linear distortion of the loudspeaker-current given sinusoidal imprinted voltage (10 V).
Alnico = Celestion "Blue", ceramic (“Keramik”) = Celestion Vintage-30.
Lower row: two Celestion "Blue" specimen.
To conclude these measurements, Fig. 11.80 shows comparisons across 4 Alnico- and 5
ceramic-loudspeakers. Again, the effects of different membrane-suspensions dominate, while
an “Alnico-characteristic” is nowhere to be found.
© M. Zollner 2008 - 2014

Fig. 11.80: Non-linear distortion of the loudspeaker-current given sinusoidal imprinted voltage (10 V).
After this analysis of the electrical two-pole parameters we of course need to pay tribute to the
transmission parameters. After all, the meaning of life for the loudspeaker is not just to offer a
load to the amplifier – it is supposed to radiate sound. Still, the trend found in the distortion
measurements continues here (Fig. 11.81). The differences between Alnico-speakers of the
same type are similar to the differences between Alnico-and ceramic-speakers – there is
nothing whatsoever to be found that could be interpreted as a magnet-specific sound. That
does not mean at all that using Alnico-speakers is pointless. Jensen and Eminence, for
example, do not offer an immediate ceramic-alternative to the P12-N and the "Legend 122",
respectively. If you want to have the sound of these legends, you will have to buy them – the
C-12N and the “Legend 125”, respectively, differ in more than just the magnet. With
Celestion, the situation is different: a serious alternative to the Celestion “Blue” stands ready
in the form of the Vintage-30, with the latter having four times the power capacity but still
costing only one third of the former – or even only one twelfth, if you calculate per watt.
However, the flair surrounding “the Blue one” is so attractive that there is no cure for its lure.
And so there will always be true-Blue devotees to the brilliant (or soft) and the dirty-distorted
(or distortion-free) sound of the Alnicos.
Fig. 11.81: Differences in SPL between two specimens of the Celestion “Blue”, and between the “Blue” and the
Vintage-30. Measured in the AEC; 1W @ 1m; speaker mounted in the enclosure of an AD60-VT.

11.8 Loudspeaker enclosures 11-71
11.8 Loudspeaker enclosures
11.8.1 Basics
Often, cases guitar-amplifier and -speaker are mounted within the same enclosure (combo);
alternatively, there is also the two-part piggy-back or stack design. From the multitude of
sizes available on the market, Fig. 11.82 shows a small selection: predominantly, 10”- and
12”-speakers are found, occasionally also 15” (with 1” = 2.54 cm). The small combos almost
always have a large opening in the rear while the larger enclosures are either of closed design
or realized as ported box (bass-reflex). In the widest sense of the word, the enclosures open to
the rear also represent a kind of bass-reflex system – albeit a very special one.
Fig. 11.82: Loudspeaker enclosures;

Membrane-diameters in inches
The enclosure (or cabinet) makes a significant contribution to the sound generation. If it is
airtight, it predominantly has the effect of an air-suspension to the membrane that increases
the resonance frequency. Since this air-stiffness grows invers to the volume, a small enclosure
would strongly increase the resonance frequency – it is presumably for this reason that small
cabinets mostly have an open back. The stiffness of air is sL = 1.4⋅105 Pa ⋅ S 2 / V for adiabatic
changes. In this formula, S is the effective membrane surface, and V is the net volume of the
enclosure. For a 12”-speaker and a 50-litre-box, we calculate 9179 N/m – this approximately
corresponds to the stiffness of the membrane. As an example: with such a mounting, the
resonance frequency for a Celestion Blue would rise by 50%. However, the enclosure acts as
an air-suspension only at low frequencies; from about 300 Hz, standing waves establish
themselves in the interior – these represent a complex, frequency-dependent load for the
membrane. The effects of such cavity-resonances could easily be mitigated via a porous
absorber loosely placed into the enclosure, but this approach is not normally taken with
instrument loudspeakers. For one, these resonances liven up the sound, if they are the correct
ones, and second, because any absorption kills off sound energy (transforming it into heat).
Since guitar-speakers by their nature need to be loud, absorbers are normally eschewed.
© M. Zollner 2008 - 2014

Very fundamentally, loudspeaker enclosures may be divided into open and closed cabinets.
The closed cabinet acts as an air-(suspension)-spring towards the backside of the membrane
for low frequencies: as the membrane moves into the interior, the pressure therein rises.
Taking adiabatic changes as a base, holds. An example with numbers: as the
12”-membrane moves 2.5 mm into the interior of a 38-liter-cabinet, the volume decreases by
0.5%, resulting in an excess pressure of 670 N/m2. This causes a rear wall with an area of 0.18
m2 (38cm x 48cm) to receive a force of 121 N. In layman’s terms: about 12 kg push against
the rear wall. That corresponds to the weight (or the mass) of no less than three Celestion
“Blue”! It is immediately clear that such a wall needs to be attached firmly and must not be
too thin. If, however, the rear wall has an opening (or consist of two sections with a gap in
between), as is the normal case for small cabinets, then the “excess pressure can be vented”,
and the forces acting onto the cabinet walls are significantly smaller (at most a tenth). The
open cabinet is barely strained by the sound pressure and therefore the material used makes
(acoustically) no difference.
Sure: there are Leo Fender’s pine-crates and their unique sound. Though this be madness, yet
there is method in it … so teaches us Shakespeare. The probably impossible-to-silence legend
tells us that that a guitar combo needs to by crafted from finger-jointed pinewood with glued-
on “tweed”. No, it needn’t. Of course, there are sound sources the sound quality of which
depends on the utilized wood – the acoustic guitar is a good example. However: would you
use pinewood? Never. The HD-28 made of pine, or the big Guild? No way, definitely NO
way, at all! The Stradivari? Come on! Maple, that’s a tone-wood, spruce as well – cedar, too.
Not pine, though. Pinewood was available on location, it was inexpensive, it was easy to
process. Moreover, Leo Fender was not a luthier – he was trained as bookkeeper. In an
acoustic guitar, the body needs to vibrate in order to radiate sound. That may be another
reason why it is not plastered with tweed or Tolex. Also, the walls of an acoustic guitar are
not half an inch thick –indeed there seem to be fundamental differences. In a guitar combo, it
is the loudspeaker membrane that vibrates – it does generate the sound. Without a doubt, the
cabinet acts as an acoustic filter; that, however, is due to the dimensions and not due to the
material. While the cabinet is made to vibrate by the sound the speaker generates, the
corresponding effects are, for the most part, entirely negligible relative to the membrane
vibrations.
To list the most important impacts of the cabinet: it operates as a conduit to the sound, it
makes for the formation of cavity resonances, and it (mechanically) supports the loudspeaker.
To the latter characteristic we may attribute a mechanical impedance against which the
loudspeaker braces itself. If this impedance is infinitely high (huge mass), the loudspeaker
frame mounted to it cannot vibrate. Of course, the cabinet does not have an infinitely high
mass, and therefore there will be a small movement at the interface between baffle and
speaker-frame. Mechanics teach us: Actio = Reactio: the force acting on the membrane is just
as big as the counter-force acting on the speaker frame. But let’s think for a moment: the
membrane has a weight of maybe around 30 g, while the speaker weighs in 3 kg – or up to 10
kg for some US-made muscle. Doesn’t the tail wag the dog here? Even if the speaker remains
un-mounted, the sheer mass of it will prohibit any significant movement of the speaker frame.
Okay, there may be some resonances where a small cause may escalate to have a big effect.
As is so often the case, measurements clarify the situation: with a laser-vibrometer, it is easy
to target a point on the loudspeaker frame or on the cabinet, and to measure the speed of
oscillation – also termed (particle-) velocity. Reference for these measurements is the velocity
of the membrane: we find it to move with 1 m/s, while the cabinet wall shows 0.01 m/s. We
have thus verified that the cabinet does vibrate – but there is not relevance to this vibration
when considering the sound of the amp.

Fig. 11.83 depicts cabinet-vibrations (----) in comparison to membrane-vibrations. The

sidewall of a Tweed Deluxe (fitted with P12-R speaker) vibrates with considerably smaller
amplitude compared to the membrane; merely at 440 Hz there is a noteworthy maximum –
but even here, the vibration of the wall is merely one tenth of the membrane-vibration. It
therefore remains negligible. The baffle (right hand part of the figure) vibrates more strongly
than the sidewall – which is not surprising given its spartan mounting and a thickness of
merely 9 mm. Still, big effects may not be expected: the radiated sound power is based on the
square of the velocity. Assuming an equal area, a difference in the velocity-level of 20 dB
results in a power difference of 1:100 in favor of the membrane.
Fig. 11.83: Membrane-velocity in comparison to the velocities of cabinet sidewall (left), and baffle (right).
These results are supplemented via the velocity levels for membrane and loudspeaker frame
as shown in Fig. 11.84; the measurement point was at the mounting ring between two screw-
holes. The speaker frame does vibrate, no contest there, and is influenced by the equipment-
feet of the combo. For the measurement shown on the left, the combo was set onto a stone-
table without feet, while for the one on the right, it sat on its factory-fitted rubber-feet. The
impact on the speaker-frame is clear, while that on the membrane is just about visible. If we
would attribute any significance to such small effects, we would also need to specify the
mechanical point-impedance of the combo-base. For the musician’s everyday life, however, it
does not play a big role whether the combo is placed on a stool or on a beer-crate. In case we
would consider that, we first would have to specify the height above ground: whether it is 45
or 50 cm makes a huge difference due to the resulting comb filtering. In theory, anyway: most
guitarists don’t really care as long as the thing won’t topple over. Still, knee-deep in all this
scientific stuff we almost forgot: any combo made of pine can’t be kept from sounding fine …
Fig. 11.84: Membrane-velocity in comparison to the frame-velocity for two different cabinet foundations.
Conclusion: Only the super-thin baffles of early Tweed amps could be seen as having a
marginal influence on the sound, if any. For enclosures with regular wall-thickness (2 cm),
only the dimensions and the resulting cavity resonances are of significance. It is irrelevant
which wood is used – the cabinet-vibrations are of secondary importance compared to the
membrane vibrations (see also Chapter 11.8.2).
© M. Zollner 2008 - 2014

Let us now consider the closed cabinets, with their probably most prominent representative
being Marshall’s 4x12"-box. The possible amount of the sound pressure inside a closed
cabinet has already been elaborated, and also which forces can act onto the cabinet walls. It is
good practice to build such enclosures to be very stable and maintain their shape, and to bolt
down the back panel using a substantial number of screws. It is also wise to insert one or even
two internal braces. A thin, strongly vibrating back panel will deprive the loudspeaker of
vibration energy, without re-radiating much of that energy but converting most of it into heat.
Such a back panel is, indeed, not a membrane suspended in a flexible surround, but needs to
bend to vibrate – this generates much inner friction i.e. useless heat. Not so much that the box
would burst into flames – the heat energy does not reach that kind of level. However, it is
energy that is lost to the generated sound. Of course, it is conceivable to design special
cabinets with a back panel that will dissipate exactly those sound energies that would lead to
atrocious sound ... but that would lead us astray from the beaten track that the sacred cows
travel on …
A small detail that keeps on being discussed when it comes to sealed enclosures is the air-
tightness (or lack of it). How leak-proof is the cabinet without a leak, actually? You get the
full bandwidth from “seal it all off with silicone” to “ leave a clearance of 1 mm all around –
otherwise it gets jammed”. In the simple model, a leak (a gap) is an acoustic filter: in
conjunction with the radiation impedance, the air within the gap forms a mass, and the air
within the enclosure acts as spring. That’s your ready-made 2nd-order low-pass: spring and
mass combined generate a resonance; for excitations below the resonance frequency, the gap
is open, and for frequencies above the resonance it is closed because here the inertia of the air
prohibits stronger movements. The ported-box (bass-reflex) enclosure takes advantage of
the same principle; it belongs to the “leaky” cabinets [3]. As an example: connecting a 1.5-V-
battery to the loudspeaker will (almost) abruptly change the air pressure in the enclosure.
However, depending on the polarity, air immediately starts to flow through the gap into or out
of the enclosure, and the pressure balances itself out. When exciting the membrane with
higher frequencies (e.g. with 1 kHz), the pressure cannot even out quickly enough due to the
mass inertia, and the enclosure operates as if no gap at all were present. If a cabinet has little
leakage, this can manifest itself as an effect only at low frequencies. The smaller the area of
the gap, the lower is the frequency range in question. Since speaker-boxes for guitar do not
have to reproduce frequencies down to 20 Hz, the requirements regarding their air-tightness
are not very stringent. We get some orientation-values from Fig. 11.85. This diagram does not
consider that the stiffness of the membrane-suspension and of the enclosure-walls can have an
effect on the resonance, and that a considerable flow-resistance occurs in particular in narrow
gaps (slits). Still, the figure is useful to approximately estimate the effects – in practice, a
simple impedance measurement will deliver data about the actual resonance. An example is
shown in Fig. 11.86: a weak leakage resonance occurs at 33 Hz.
Fig. 11.85: correspondence between volume of the

3 2
enclosure (abscissa, dm = liter), area of gap (cm )
and resulting resonance frequency.
Example: a 100-liter-cabinet has gaps with a total
2
surface of 5 cm – this results in a leakage-
resonance at about 20 Hz.

The shift in the resonance caused by the air-stiffness can be clearly seen in Fig. 11.86. The
impedance plot furthermore reveals the primary cavity resonance (260 Hz) that forms as
longitudinal λ/2-oscillation within the enclosure (70 cm length). From this frequency, the
enclosure volume does not act anymore as concentrated stiffness but as a continuum. A mass
(about 12 g) and two springs of approximately equal stiffness can form a simplified model of
this lowest cavity resonance. Transformed to the electrical side, the impedance of this
analogy corresponds well with the measurement. The differences occurring above 300 Hz are
mainly due to membrane-resonances (partial oscillations).
Fig. 11.86: Frequency response of the G12-M impedance (see also Fig. 11.17); measurement (left), model (right).
Middle: simplified longitudinal-oscillation model of the λ/2-resonance.
The cavity resonance makes itself felt in the transmission frequency-response via an S-curve
(Fig. 11.87). Just below this resonance, the efficiency deteriorates, and just above it there is
an improvement. The loading from the rear of the membrane has effects onto the radiation
occurring on the front, as well – this is easily explained by the relationship between source-
and load-impedance. Driving the speaker from a stiff voltage-source, the electrical source
impedance of the FI-transducer [3] is the ohmic voice-coil resistance (below 1 kHz we may
disregard the inductance). The source impedance of the mechanical side of the transducer
therefore is a purely ohmic resistor. This resistive source is loaded by several mechanical
components: the membrane, the (inner) cavity resonance, and the (outer) radiation impedance.
The three impedances need to be added up resulting in an overall impedance, and therefore
each of the three will influence the matching. Bold and simple: even if you secure the
membrane only from the inside, it still cannot radiate any sound on the outside anymore. The
frequency-selective change of the matching lies at the core of the function of any reactance-
filter – as such the S-curve is no surprise. That the effect onto the electrical impedance
frequency response partially is rather small may be traced to the relatively low efficiency: the
ohmic voice-coil resistance dominates (possibly together with the voice coil inductance) the
electrical impedance. Of course, resonances pronounced to that degree are indeed audible;
whether they sound good or bad is – as always – a matter of personal taste.
Fig. 11.87: Transmission frequ. response: enclosure with cavity resonances. 39x75x25 (l.) and 40x74x39 (r.).
© M. Zollner 2008 - 2014

As the preceding chapters have shown, loudspeaker and cabinet act as two filters connected to
the output of the amplifier. A signal tapped from the input of the power amplifier misses just
that filtering. That is why we most often find microphones in front of the loudspeakers – to
record or mix instrument-specifically. However, in the near field of a sound source with a
relatively large area there may be transfer functions that depart significantly from the far-field
characteristic. Strictly speaking, the border between near-field and far-field is assessed
according to the size of the loudspeaker enclosure, but as a simplification to start with, we
may use the size of the membrane as a criterion: if the distance between microphone and
speaker is only about as big (or even smaller) than the diameter of the membrane, then the
microphone is located within the near-field. If more than one speaker is mounted in the
enclosure, the diameter of the equivalent membrane must be considered. For a typical 4x12”-
box that would mean not merely 28 cm but already almost 1 m. Customarily, microphones are
positioned more closely to the box, i.e. within the near-field.
To model sound-radiating surfaces (e.g. membranes), they are divided up into small partial
areas each radiating spherical waves (according to Huygens’ principle). Fig. 11.88 points this
out using the example of a plane membrane (on the left side of the figure). For a point
infinitely far away, the outgoing sound rays are travelling in parallel and the individual sound-
paths are of equal length. The closer the measuring point gets to the membrane, the more
unequal the sound-paths become, resulting in different delay times between the sound rays.
This has no bearing for low frequencies, but for higher frequencies the different path-lengths
may be equal to half a wavelength, and interference cancellations will then happen.
Fig. 11.88: The closer the microphone gets

to the membrane, the more the individual
sound paths differ in length.
Fig. 11.89 shows the effects of such interferences, measured in the AEC with a Tweed
Deluxe. As the microphone approaches the centre of the membrane axially from a larger
distance, the SPL increases. This does not happen for all frequencies in the same way,
however! Since the absolute sound pressure levels are not as relevant here, a constant was
subtracted in the diagrams such that values floating around 0 dB result. As the speaker
approaches, predominantly the low frequencies are emphasized, and furthermore other
frequency-selective filtering occurs. A condenser microphone with an omni-directional
characteristic was employed; for directional microphones, an additional proximity effect
needs to be considered [3]. If the loudspeaker is not positioned in the AEC but on a reflecting
surface, environment-dependent comb-filter effects weigh in as well.
Fig. 11.89: Tweed Deluxe, axial near-field measurements; normalized level changes relative to d = 1m.

Even when keeping the distance between microphone and loudspeaker constant and “merely”
adjusting the lateral location in front of the speaker, the transmission frequency response
changes, as Fig.11.90 strikingly proves – it shows the difference in SPL as the microphone is
repositioned by the given offset. The left-hand diagram was recorded from a Tweed Deluxe,
the right-hand one from a 2x12”-box. It is understood that such significant differences have a
dramatic effect on the sound. Consequently, the choice of the microphone position may
possibly by more crucial than the choice of loudspeaker! The directionality of the microphone
will also influence the sound: if the mike is positioned directly at the cloth protecting the
speaker, sound waves from different membrane areas arrive from different directions.
However, if the microphone is located (in the recording studio) at a greater distance from the
loudspeaker, it will record sound reflected by the room in addition to the direct sound.
Fig. 11.90: Level changes with repositioned microphone; const. distance to the baffle. Left: 1x12”; right: 2x12”
Given the results of these measurements it is understandable that musicians and studio-experts
pay the utmost attention to the choice of the microphone and to its position. Often even two or
three microphones are deployed to record a guitar speaker, with small markers on the speaker
cloth supposedly guaranteeing the retrieval of the magic spot. Global rules such as
“microphone distance = membrane diameter”, or “microphone distance = 3 x membrane
diameter”, or “don’t point the mike to the center of the speaker but to a point half the distance
between dust cap and rim”, are well meant but must never be generalized. What sounds good
with one speaker can be utterly unsatisfactory with another – individual tuning is required.
In order to avoid the not insignificant effort of bringing (besides the guitar) the whole
amplification equipment, setting it up, and painstakingly finding the right microphone
position, many musicians (and producers ) often opt for a radically simpler approach. The
guitar is plugged into a “modeling amp” that takes care of all necessary linear and non-linear
filtering. By now word has spread that this also includes the filtering contributed by the
loudspeaker. Daily studio-practice shows that this route makes it possible to generate
nightmarishly artificial guitar sounds, but it also proves that impressively wonderful results
can be achieved – which afterwards need to be camouflaged with fake-evidence (“even in the
tile-covered bathroom, we had a ’64 Blackface fitted with NOS-tubes and miked up with three
condensers”) to be able to survive in a world of vintage-craziness. That such a modeling amp,
good or bad, cannot emulate the directional characteristics of its paragon has been already
noted on Chapter 11.4. Also, if an amp with a high-gain-sound is to be emulated: the feedback
onto the guitar (that can support the ringing of the strings or even generate ringing by itself) is
missing if the guitarist only uses headphones to hear himself play. Of the multitude of
modeling amplifiers (whether with or without power amplifier), we selected the POD 2.0
made by Line-6. This is not meant to give a rating in terms of particularly good or
particularly bad – the device simply was easily available.
© M. Zollner 2008 - 2014

Fig. 11.91 shows the frequency responses of the loudspeaker emulation of the modeling amp
(Line-6 POD 2.0) Unfortunately, the manual does not give any information about the virtual
microphone position.
Fig. 11.91: Loudspeaker-emulation in the Line-6 "POD". The absolute ordinate-scaling is arbitrary.

The VOX AC-30 will serve to find out which effects small details in the enclosure can have
on the sound. In this combo, the amp is located together with the two 12”-loudspeaker in one
cabinet; merely a lateral board separates the two sections (Fig. 11.92). However, there is a
special variant, the “AC-30 Super Twin” in which the amp is afforded a separate cabinet (i.e.
it’s a piggy-back design), and therefore the separation-board in the loudspeaker enclosure
(with otherwise identical dimensions) is omitted. At first glance, there are thus two different
enclosures: with and without amplifier. However, the separation-board in the regular AC-30
features a rather big opening in order to allow, via a kind of chimney effect, cooling air to get
to the amplifier positioned above the board. This air escapes through vents on top of the
enclosure – accompanied by sound, of course: because where there is an airflow, sound will
also pass. The dimensions of the vents have definitely changed over the years – whether the
opening in the separation-board is subjected to the same “time-variance” was not investigated.
For the transmission behavior this implies that not only do we need to pay attention to the
fitted loudspeakers but also to the cabinet-design. The electrical loudspeaker–impedance
changes as the separation-board is removed, or as the air vents are changed. Fig. 11.92 shows
that in the range of 100 – 300 Hz the impedance may vary by a factor of 2. And since the tube
output stage of the AC-30 lacks any negative feedback and therefore has a high output
impedance, this impedance change makes itself felt in practically the same magnitude in the
transmission frequency response. The impedance maximum at 170 Hz, for example, has the
same effect as if we had boosted a narrow band around this frequency by 6 dB with an
equalizer. Here, we find an interesting parallel in the area of acoustic guitars: measurements
performed by Fletcher and Rossing [1] with a Martin D-28 show a strongly pronounced
resonance at 200 Hz in the sound spectrum. Possibly, the selective emphasis of this frequency
range positively influences the sound of acoustic and electric guitars.
Fig. 11.92: Left: 2-part rear-panel of the VOX AC-30 (“Verstärker” = amplifier). Right: frequency responses of
the impedance (series connection of 2 speakers). The different impedance curves correspond to modifications in
separation-board and air vents.
© M. Zollner 2008 - 2014

11.8.2 Comparison of various enclosure-materials

An often-asked question: what is the contribution that the type of wood used for the enclosure
makes to the sound (i.e. to the transmission function) of the loudspeaker? Dealers point to
historic models, and attribute to the wood a significance similar to that it would have for
Italian master-built violins – and the musician believes it and shells out the money. To go
beyond assumptions and obtain some objective data, we analyzed a number of cabinets of
identical dimensions but made out of different woods: pinewood (18 mm), poplar (14 mm),
and medium-density fiberboard (MDF, 14 mm). The enclosures were carefully assembled by
Tube-Town (www.Tube-Town.de) and were measured with the same loudspeaker installed
(Eminence MOD-12). The external dimensions were 50 cm x 41 cm x 30 cm. The sealed
enclosures were closed off to the rear with a non-reinforced panel while the open cabinets
featured two boards to the rear that had a gap of 13 cm between them.
All measurements were done in the anechoic chamber at 3 m distance on axis. The speaker
was fed from a stiff voltage-source (2.83 V at first, later more); a B&K 4190 served as
measurement microphone,. The resulting frequency responses of the SPL (recalculated for 1
m distance) are shown in Fig. 11.93. There are visible differences between the wood-types,
but they are so small that they will be insignificant for everyday stage-use. In fact, our hearing
does not recognize such small sound differences in music performances. Moreover,
production tolerances will have a similar magnitude. However, the differences caused by
changing the back panel (open vs. closed) are of significance – the sound does change.
Fig. 11.93: SPL (1W/1m); enclosures: pine (black), poplar (red), MDF (blue).
Left: open rear-panel (gap of 13 cm). Right: closed rear-panel.
Fig. 11.94 shows the corresponding frequency responses of the impedance; again there are no
peculiarities. The pronounced similarities guarantee practically the same behavior when
driving the speaker from a high-impedance source (tube amplifier) – independent of the wood
type. However, the changes in the rear-wall have in a particularly strong effect for operation
from a high-impedance amp because transmission behavior and voltage at the speaker change.
Abb. 11.94: Frequency responses of the impedance, loudspeaker and enclosure as in Fig. 11.93.

In Fig. 11.95 we find supplementary measurements with pinewood-enclosures that either had
no back-panel at all, or a partial one consisting of one board of a given size, or a partial one
consisting of two boards of the same (given) size each (i.e. the latter corresponded to the open
cabinet of the previous measurement). Again, there are no unexpected peculiarities.
Fig. 11.95: SPL and impedance; completely w/out back panel (magenta), 1 board (cyan), 2 boards (black).
The operation with 2.83 V (resulting in 1 W at 8 Ω) is typical for loudspeaker measurements

but does not correspond to customary power loading. As long as the speaker works in a
reasonably linear fashion, the transfer function may be taken at any voltage. However, since
loudspeakers can generate significant non-linear distortion, we opted to include measurements
at a higher power level: at 2.83 V, 8.94 V, and 17.9 V, corresponding (at the nominal 8-Ω-
impedance) to a power of 1 W, 10 W, and 40 W, respectively. Upping the power from 1 W
to 10 W and 40 W, the level rises by 10 and 16 dB, respectively. This is shown in Fig. 11.96 –
merely in the bass-range we see deviations due to very strong distortion. To facilitate
comparing the curves, Fig. 11.97 depicts a representation normalized to 1 W: the 10-W-
curves was lowered by 10 dB, and the 40-W-curve was lowered by 16 dB. Overall, the 40-W-
curve is low visibly at too low a level; this is, however, not wood-specific, but simply caused
by the heating up of the voice coil (all measurements were done with stiff voltage-source).
Fig. 11.96: SPL at 1 W (blue), 10 W (black), 40 W (red). Enclosures: pine (left), poplar (right).
Fig. 11.97: as in Fig. 11.96 but representation normalized to 1 W.
© M. Zollner 2008 - 2014

The wood of the enclosure might influence the transmission function in two ways: via
changes in the mechanical impedance of the bearing of the loudspeaker, and via sound
radiation of co-vibrating enclosure walls. In order to obtain quantitative data relating to
enclosure vibrations, we carried out measurements with a laser-vibrometer (Polytec). The
loudspeaker was again fed from a stiff voltage source (2.83 V, 8.94 V, and 17.9 V). The laser-
vibrometer measures the velocity; from this the displacement can be derived via integration,
and the acceleration via differentiation. For the ideal loudspeaker (given a stiff current source)
the acceleration is imprinted at f > fRes; in the real speaker, resonances of the membrane cause
selective frequency dependencies. Acceleration values corresponding to up to the 100-fold of
the gravitational acceleration may be expected: 30 N at 0.03 kg yields 102 g (1 g = 9.81 m/s2).
Only the membrane experiences such strong acceleration, however; the side-panel vibrations
are markedly weaker relative to the membrane-vibrations (Fig. 11.98).
Abb. 11.98: Displacement (left) and acceleration (right) of the middle of the side panel; poplar;
1 W, 10 W, 40 W. Curves were normalized to 17.9 V (40 W), i.e. the 1-W-curve was elevated by 16 dB.
As is well known, the radiated sound power depends on the square of the velocity, on the size
of the vibrating area, and on the radiation impedance [3]. The latter, and the effectively
radiating area as well, can only be determined with much effort; therefore here just an
approximate estimate: if the velocity of the membrane is, at 460 Hz, about 7 times as high as
the velocity of the side panel (ΔL = 17 dB), the membrane will radiate about the 49-fold
sound power at this frequency compared to the side panel. The other ‘round: the side-panel
contributes merely 2% to the sound radiation. Even if it were 5%: that’s still rather
insignificant. The contribution of the baffle is similarly small; only the back-panel weighs in
with two relatively strong vibration-maxima (Fig. 11.99).
Abb. 11.99: Acceleration, poplar-enclosure, P = 10 W. Left: baffle, right: edge of back-panel.

Each of the two rear-panels is bolted on only three of its sides so it’s understandable that
larger vibration amplitudes are possible. Of course, the one-point-measurements presented
here cannot provide exact data of the sound radiation – due to the lack of a scanning
vibrometer, an exact sampling of all enclosure surfaces was not possible, and the selected
measuring points can only give a first impression. The comparison of the three enclosures
shows that all their walls vibrate in a similar manner (Fig. 11.100). The maxima of the rear-
panel vibrations are a bit stronger in the poplar-made cabinet. For the sidewalls and the baffle,
we see clear differences in the resonance frequencies but the maximum levels are similar.
There are several reasons why the measured enclosure vibrations contribute so little to the
SPL. The vibration amplitude of the cabinet walls, for example, is never larger than that of the
membrane vibration. At low frequencies, the whole membrane surface vibrates with the
same amplitude, which is not possible for a board bolted down at its rims even at its
resonance. As regards the radiation impedance: the rear panel vibrates (at 200 Hz) strongest
with its free rim, similar to a dipole. With an outward movement, the outer surface of the
rear-panel generates excess pressure while the inner surface generates low pressure – both
balance themselves out momentarily around the rim of the panel. Regarding the sound
radiation, this is a most inefficient movement that is termed “operation with acoustical
shortcut”. At higher frequencies, lines of nodes appear in all enclosure walls, separating areas
of the panels that vibrate in opposite phase: as one point of the panel moves outward, a
neighboring point moves inward at the same moment. With the two movements being in
opposite phase, only little sound is radiated. In Fig. 11.100, the SPL-measurement is again
included for comparison: as different as the enclosure vibrations may be, they all have very
little bearing on the sound pressure level.
Abb. 11.100: Left: acceleration of the enclosure wall. P = 10 W; poplar (red), pine (black), MDF (blue).
Right: SPL-measurement axially, 3 m in front of the enclosure, P = 1 W; color-coding as above.
© M. Zollner 2008 - 2014

Abb. 11.100 depicts the frequency response of the SPL in front of the membrane; however the
loudspeaker radiates in all directions. Fig. 11.101 shows the SPL frequency-responses for two
further measuring points: 3 m behind the enclosure and 0.5 m above it.
Fig. 11.101: Left: SPL 3 m behind the enclosure, P = 1 W; poplar (red), pine (black), MDF (blue).
Right: SPL 0.5 m above the leading edge of the enclosure, P = 1 W; color-coding as above.
All SPL- and vibration measurements were done with one and the same loudspeaker, an
Eminence MOD-12. To mount it, the rear panels had to be disassembled and reassembled
each time. Repeat-measurements carried out to investigate the reproducibility showed SPL-
differences that can be traced to the mounting of the rear panels (Fig. 11.102). Measuring the
rear-panel acceleration showed a very strong dependency on the torque with which the
mounting screws were tightened. This torque had not been checked when re-mounting the
loudspeaker♣; consequently it can be assumed that enclosure-specific differences found in the
SPL are in part due to differences in the attachment of the rear panel. Therefore the
differences purely due to the wood turn out to be even smaller.
Fig. 11.102: Rear-panel screws tightened with different torque. Left: SPL; right: rear panel acceleration
Given these measurement results, the question poses itself why the dealers put so much
emphasis on the wood used for the construction of instrument-loudspeaker cabinets, i.e. why
it is imperative that the guitar box is made of “Baltic birch” or “solid pine”. Simple answer:
because it has always been that way – there’s no connection to vibration-engineering. You
can’t build a Fender “Woody” using MDF-panels, because you will want to see only the most
♣
The screws fastening the rear-panels had been tightened „strongly“ by hand each time.

strikingly beautiful wood grain. With a Tweed Deluxe, you could use MDF – as long as
nobody looks inside. Leaving aside cosmetics, we have sound, weight, and durability
remaining. Without further testimony, let’s believe in the higher on-the-road resilience of the
precise finger-joint construction. The question regarding the maximum weight we shall
delegate to those tattooed knights-of-the-long-braid who will willingly schlep all that stuff
back and forth every night. That leaves us with the sound. Is the wood in fact supposed to
vibrate or not? Fortunately, our much-shepherded problem-child cannot be bothered about
that question and just vibrates, as soon as it receives the invitation to do so from the
membrane – irrespective of whether it is pine, poplar, birch, particle-board or MDF. Not in
the identical manner for each of those materials, but so little that any influence on the sound
radiated by the membrane remains marginal. We would not be adverse to the wood
contributing some resonances (this being a sharp contrast to the world of HiFi) since the
electric guitar has to offer little in that area. However, such contributions would have to be
product-specific, and that would require a disproportionate effort – in a number of ways, not
just in the tightening-torque for the mounting-bolts. The dimensions are crucial, as well: if the
rear-panel rests on a slightly convex bar, it will vibrate differently compared to it sitting on a
concave bar. Minute tolerances would be of importance here – one reason why acoustic
guitars are not bolted together from planks. Speaker boxes, on the other hand, receive just that
treatment – there appear to be differences to your D-28 or J-200, after all.
Fig. 11.103: Left: SPL (1W/1m), cabinets made of pine (black), poplar (red), MDF (blue).
Right: SPL (1W/1m), pine cabinet; 60x40x29 cm3 (red), 50x41x30 cm3 (black).
In Fig. 11.103, the differences caused by the type of wood are contrasted with those caused by
changing the enclosure dimensions. The latter are varied by only a few centimeters – but hat
is enough to result in greater differences that swapping poplar for pine.
Whether the loudspeaker is front- or rear-mounted onto the baffle-board also makes for
small differences in the frequency response: The rear-mounting results in slight advantages: a
gain of 1 – 2 dB in the range between 0.2 and 1 kHz, and a loss of about 2 dB around 3 kHz.
The exact values depend on the given chassis and the dimensions of the enclosure.
The speaker-cloth on the baffle can have a two-fold effect: comb-filtering because sound is
reflected back to the membrane, and – in particular at high frequencies – absorption. Some
cloths, for example the material used in Fenders “Silverface”-amps, have next to no effect at
all. Others, such as e.g. the thick material used by Marshall, cause an attenuation of about 1
dB at 1 to 5 kHz … which can certainly be measured but will be audible only when listening
VERY closely.
© M. Zollner 2008 - 2014

11.9 Beam blockers, diffusers, and such
In a loudspeaker, beaming effects increase with increasing frequency (Chapter 11.4). The
treble, i.e. frequency range upwards of about 1 kHz, is predominantly radiated on-axis, while
the lows propagate spherically in all directions. If the loudspeaker (e.g. a 1x12”) is set on the
floor, the guitarist standing right in front of it or next to it gets to hear too little treble. If the
guitar player positions the speaker at the level of his head, the treble will be unbearably shrill
(and dangerously loud, potentially damaging the hearing system). Therefore, beam blockers
are available that are supposed to distribute the treble within the room, working similar to a
diffuser lens.
The concept of the acoustical lens has in fact been around for quite a while – it is already
mentioned by Olson [1957]. Similarly to an optical lens, the peripheral sections of a wave
need to be delayed if divergence is called for (Fig. 11.104). To achieve that, the peripheral
sound rays are run through an array of slanted sheets bent in serpentine fashion, creating a
longer, indirect path and therefore a phase-shift. JBL has introduced these acoustical lenses in
the early 1970’s, but they vanished again from the market as horns were further developed.
Fig. 11.104: Acoustical diffuser lenses; pictures from: www.jblpro.com
Today, not lenses but massive scattering bodies are deployed in order to reduce beaming
effects in guitar loudspeakers. The Weber Beam Blocker (Fig. 11.105) is supposed to scatter
the treble coming from the speaker-center via a spherical cap of convex shape. However,
theoretical acoustics teach that beaming will occur the stronger, the larger the (uniformly)
radiating source is – a ring-shaped emitter therefore does not have less beaming compared to
the membrane centre thought to be the source of the treble. Reality is even more complex
because it’s not only the centre of the membrane that can radiate treble but the fringe areas as
well, and because the beam blocker will reflect sound back to the membrane, too.
Fig. 11.105: Weber Beam Blocker, www.webervst.com/blocker.html

11.9 Beam-blockers, Diffusers 11-87
To obtain quantitative data, a Tubetown-diffuser was fitted to a 12"-loudspeaker mounted in

the cabinet of an 18-W-Marshall. Measurements were done in the anechoic chamber with the
loudspeaker box positioned on a turntable (B&K 3922), and the microphone (B&K 4165)
located at the elevation of the speaker axis. Distance was 3 m. To obtain the colored
directional spectrograms, pink noise was analyzed using overlapping 1/3rd-octave filters. The
results are shown in Fig. 11.107 (level dynamic = 40 dB) as a function of turn-angle
(abscissa) and frequency (ordinate). The directional diagrams pictured below this are
horizontal cuts through the color-diagrams.
As can be seen without much difficulty, this diffuser has practically no effect at low and
middle frequencies – this indeed being purposeful. Around 3.5 kHz for the G12H, and around
5 kHz for the P12R, a slight broadening of the radiation is achieved. The effect is moderate –
as is the price. “I’d rather invest those 15 Euro in a few beers – that will change my sound,
too” … this assessment would not seem unreasonable. For those who want to experiment
themselves (with the diffuser, not with beer): fasten a cardboard disc (∅ 8 cm) to the outside
of the speaker cloth, and if you like what you hear, then buy the professional diffuser and
mount inside of the cloth to the loudspeaker frame. Or make one yourself from cardboard.
Jay Mitchell proposes another solution in the "Manufacturers' and Retailers' Forum": a
doughnut of foamed plastic is positioned within the circular cutout in the baffle board that
however must not touch the membrane. The thickness of the doughnut is just under the
thickness of the baffle board (about 15 mm), its outer diameter corresponds to the speaker-
cutout in the baffle (about 28 cm for a 12”-speaker). The hole in the centre of the foam
doughnut measures about 7 cm. Supposedly this arrangement will also distribute the treble
better within the room. Our measurements cannot confirm this assumption: the main effect is
a dampening of the treble. Which may in fact be a solution for the original problem, too.
Hoovi offers a rather more expensive solution: a handsomely styled reflector panel that is
intended to deflect the sound to the side and to the top. Indeed, this works, and you can join
the fun for the stately sum of around 350 Euro per speaker. Don’t stumble over the thing,
though, and make sure you don’t leave it behind during the load-out. That would be rather
aggravating considering the price. Also, you will not want the precious device to be scratched
– but then you won’t let your roadie throw your prewar-Adirondack on the truck without a
case, either; so: take along a tailor-made transport case. And don’t you dare set, instead of the
Hoovi, a slantwise cut detergent-drum in front of the amp! That does work as well – but looks
decidedly less noble♣.
Fig. 11.106: Deeflexx,

Donar's missile.
[www.hoovi.at]
Particularly interesting is the
solution for 2 speakers: the
sound deflected towards the
right from the speaker on the
left … where does that in fact
go? Yep, exactly – that’s
where it goes!
♣
Cited from the depths of the www: “if the guitar player doesn’t cut it, at least his rig should look cool…”
Opposing view: “no way I’m going to let such a shitty-looking thing ruin my vintage AC-30-appeal”.

The following measurements were done, at a distance of 3 m in the AEC, with a Tube-Town
diffuser attached to a G12H that was mounted in a Marshall-18-W-cabinet.
Level-dynamic: 40 dB
Fig. 11.107a: Celestion G12H; without (left) and with (right) Tubetown diffuser.

11.9 Beam-blockers, Diffusers 11-89
The effect of the diffuser shifts to higher frequencies for the P12R (which radiates somewhat
more treble than the G12H).
Fig. 11.107b: Jensen P12R; without (left) and with (right) Tubetown diffuser (left).

For the angled diffuser (construction similar to the Deeflexx), the effect is more brute, the
distribution is broader, and there is a total treble-loss on axis.
Fig. 11.107c: Angled diffuser in front of a VOX AD60-VT, Celestion G12 Century. It was not the DeeFlexx that
was measured but a replica of equal dimensions.

11.10 Horn-loudspeakers 11-91
11.10 Horn-loudspeakers
The classic guitar speaker arrangement does not feature a horn – but in a PA, or in the control-
room of a recording studio, horn-loudspeakers may well be deployed. This is to increase the
scarily low efficiency, and to modify the directionality. In a studio-speaker at most only about
3% of the power generated by the amplifier is converted to sound (Chapter 11.5), so
experiments to improve the matching were done early on [Olsen]. The source impedance of
the membrane is relatively large, and that of the load is small: in such a scenario, we would
call for a transformer in electrical engineering; in acoustics, we would call for – right: a horn.
, J1 = Bessel-function
, H1 = Struve’s function
, a = membrane-radius
Fig. 11.108: Normalized radiation-impedance for a piston-membrane [3].
Fig. 11.108 shows the complex radiation impedance R + jX. Multiplied by Z0 = 414 Ns/m3, it
gives us the sound-field impedance p/v for a circular, plane membrane: a first approximation
for the loudspeaker-loading by the adjacent air [3]. Below 450 Hz, the membrane loading is
predominantly imaginary; the membrane shoves air back and forth without actually sending
off a lot of effective power in the form of a wave. Above 450 Hz, the real part does dominate,
but at the same time, the membrane starts to have beaming effects. Positioning a horn in front
of the membrane increases the real part of the loading at low frequency, and therefore
improves the efficiency. However, in the bass-range this solution would require horns of
enormous size, and therefore horn-systems operate mainly in the middle and treble range.
For first considerations it is purposeful to assume the cross-section of the horn to be circular
(calculations may be done using cylinder coordinates). Hyperbolic horns give advantageous
dimensions, with the radius r(z) of the cross-section growing with z from the horn-“throat”:
Here, z is the distance to the throat (radius rTH), M is a form-factor, and the horn-constant ε
represents how fast the radius grows with increasing z. Given M = 1, the area increases
according to an exponential function (exponential horn); given M = 0, the increase happens
along a chain-line (catenary horn). For an exponential horn, the (plane) cross-sectional area
grows exponentially: S(z) = STH ⋅ exp(ε z), with the area of the throat being STH. Towards
lower frequencies, a cutoff-frequency f > ε c / 4π (for the wave-propagation within the horn)
results from the flare-rate ε. The “mouth” of the horn (mouth-radius R) yields a further cutoff-
frequency f > c /πR for optimal matching. If the cutoff-frequency of the mouth is too high,
disruptive reflections may occur within the horn.

Fig. 11.109 shows, for three different horns, the cross-section as it develops with the
increasing value of z, and also the logarithm of the real part of the acoustical load impedance
(0 dB = Z0); on the right, the impedance without the horn (red curve) is included for
comparison. For calculating the load, the length of the horn was assumed to be infinite so that
reflections and standing waves could be ruled out. In horns of finite length, part of the wave
running towards the mouth is reflected; the smaller the opening at the mouth is, the stronger
the reflection. Fig. 11.110 depicts two cases of identical wave-cutoff frequency but different
mouth-cutoff frequency. The optimum angle at the mouth is about 90°.
Fig. 11.109: Cross section (links) and membrane-loading for various horns. Throat-radius = 5 cm.
Fig. 11.110: Logarithm of the real part of the membrane loading for two different horn lengths; equal ε .
The circular cross-sectional area is a first approach towards calculation. In reality the cross-
section develops from a round throat-area to a rectangular mouth-area, allowing for different
directionality in the vertical plane compared to the horizontal plane. The beam-width Φ is a
measure for the radiation but still remains rather limited in its meaningfulness, as it is seen in
Fig. 11.111: despite equal angle the directivity of two loudspeakers may differ significantly.
Fig. 11.111: Directionality; differing

directionality factor despite equal
aperture angle.

Besides the beam-width that gives a single number for the beaming behavior in one plane, the
directivity yields an average value for all directions. Formally, the squared directional gain Γ
needs to be integrated in space along an enveloping surface surrounding the source, and
reciprocally be referring to this surface [3]. The logarithm of the resulting beaming index γ
becomes the directivity d. In the simplest case, a sphere with the surface area of HF = 4π r2
serves as the enveloping surface.
HF = enveloping area
Let us assume as an example that the source radiates conically into the sector of a sphere, with
a center-angle ψ (Fig. 11.112) and a spherical-cap surface S. For ψ = 180° (half-space) we get
from this a directivity of d = 3 dB, and ψ = 60° would yield d = 11.7 dB.
Fig. 11.112: Beaming for a conically radiating emitter.
The first rectangular horns produced in large quantities were radial horns. For this type, the
horizontal dimension grows in linear fashion such that the vertical dimension needs to take
care of the progressive increase required for the exponential growth of the area (Fig. 11.113).
This geometry achieves a reasonably frequency-independent aperture angle – at the expense
of the vertical directionality. Later developments (such as the so-called Mantaray horn by
Altec Lansing, Fig. 11.113) allowed for a frequency-independent patterning of the directivity
index (rather than of the horizontal aperture angle). The result was not perfect nor did it
extend over the whole frequency range, but worked to a passable extent from a recommended
cutoff frequency. Behind Altec Lansing, other manufacturers (JBL, Electro Voice, et al.)
followed suit and developed horns with an approximately frequency-independent beaming
index. At low frequencies (where the wavelength is large relative to the dimensions of the
horn), all horns exhibit little beaming – only in the mid/high frequency-range, the specified
beaming occurs.
Fig. 11.113: Radial horn and Constant-Q-horn;

vertical (top) und horizontal lateral dimensions.

Fig. 11.114 shows the beaming of a radial horn [D.B. Keele, AES Prep. 1083] compared to a
constant-Q-horn [JBL 2356A]. The directivity index (i.e. DI and d, respectively) increases
between 500 Hz and 15 kHz by 10 dB for the radial horn. On one hand, this is helpful because
the power-frequency-response of typical horn-drivers decreases from about 2 kHz up, but on
the other hand it is unattractive: only the direct sound, but not the diffuse sound, profits [3].
Fig. 11.114: Aperture angle and directivity index (DI) for two different horns. The radial horn was marketed as
60°x40°-horn (according to the datasheet) – rather courageous given the vertical beaming.
In midrange- and treble-horns, the horn does not directly sit on the membrane but connects to
it via a compression chamber (Fig. 11.115). Assuming a location-independent pressure, the
continuity requirement (q1 = q2) yields the relationship between membrane (Index 1) and the
starting point of the horn (throat, Index 2): the load impedance rises by the ratio of the areas.
In practice, the compression chamber is not of cylindrical shape, though, but forms a so-called
phase plug that supports avoiding path-dependent interferences. Drive and membrane
combine into the driver, to which horns with varying beaming behaviors can be fitted. In
order to be able to specify driver-data independently of the horn, the former is mounted to a
plane wave tube (PTW) – a tube with a length up to 6 m in which the waves can travel
without reflections. The input impedance of the tube is approximately real: p/v = 414 Ns/m3.
Fig. 11.115: Compression chamber. Photo: Lansing-Heritage

In Fig. 11.116 we see a typical frequency response of a PWT. For the measurement, the driver
(JBL 2451) was mounted to a 1.5”-PWT (although in the datasheet a 1"-PWT is noted). At
middle frequencies, this driver reaches an electro-acoustic efficiency of 30%. From 3 kHz,
the coupling deteriorates such that at 10 kHz, only 3% remain – which is not bad either. As
this driver is mounted to a radial horn, the beaming of the latter (increasing with frequency)
makes for a partial compensation of the treble loss, at least for the direct sound in front of the
speaker. According to the rules of simple room acoustics, the beaming has no effect on the
diffuse sound. The manufacturers recommend compensating the weak treble via filters
(equalizer, EQ), but that only works up to a point: a 10-dB-boost requires the 10-fold power!
Fig. 11.116: Power-frequency-response and impedance of a driver [JBL 2451] mounted to a PWT. The ordinate
specifies the SPL (in dB) obtainable with Pak = 1mW. Since the acoustical loading is real, the sound pressure can
directly be recalculated into the sound power: P = p2 ⋅ S/Z0.
Fig. 11.117 shows two further horns: with acoustical lens, and with separating strips within
the horn. The extreme vertical beaming of the lens is probably not entirely unrelated to its
becoming extinct. The Smith-horn is a kind of multicell-horn but includes a closed bottom
and a closed lid.
Fig. 11.117: Beaming for a JBL-lens (left) and for a so-called Smith-horn (right). “vertikal” = vertical

11.11 Studio-monitors
In the control room of a recording studio, high-grade multipath loudspeakers are deployed.
Especially during the 1950’s to 1970’s, they were often fitted with mid- and treble-range
horns. Contrary to widespread opinion, the frequency response measured on-axis is not the
most important criterion. The frequency dependency of the “free-field transfer function” is not
unimportant, but premium loudspeakers handle this aspect so well that other criteria move to
the focus, for example the beaming, or (at high monitoring volume) the distortion (THD,
difference tones, sub-harmonics). Because satisfactorily handling the whole audible frequency
range with a single loudspeaker is not possible, filters (crossovers) take care of a subdivision
into several frequency bands fed to corresponding speakers. Shown in Fig. 11.118 is a simple
circuit, as it is found (with slight modifications) in many DIY-guides. For the corresponding
calculation it is assumed that the loudspeaker impedance is real, and that impedance and
transmission-factors are frequency-independent. These assumptions are far from reality: the
impedance is complex and dependent on frequency (Fig. 11.9), as are the transmission factors
(in particular the phase). However, let us follow for a moment the idealized train of thought:
the 2nd-order low-pass shifts the phase from 0° to -180°, and the 2nd-order high-pass generates
a shift from 180° to 0° – such that across the whole frequency range the speaker voltages are
in opposite phase. Only connecting the speakers ‘out-of-phase’ will avoid a complete
cancellation at the crossover frequency (600 Hz in our example). That an all-pass filter results
is, on the other hand, not critical: out hearing system does not take notice of that [3].
Fig. 11.118: 2nd-order two-way crossover: circuit (left), frequency response of the phase (right).
The big problem is created at the crossover frequency, if both speakers radiate the sound with
the same amplitude. Even if the two partial sounds sum up perfectly on-axis – the radiation
towards the sides always involves a phase shift, creating a destructive interference. If the
difference in the path-length corresponds to half the wavelength (λ = c/f ), the partial sounds
cancel each other out (Fig. 11.119). An improvement is possible via so-called coaxial systems
with the woofer being positioned behind the mid-range-speaker on the same axis; however
here the speakers may get in each other’s way. The argument that we should simply listen
only exactly in front of the speaker does not hold water, either: the reflections arriving from
the side do influence the hearing perception, as well.
Fig. 11.119: Interference in the crossover-frequency-range: cancellation for radiation towards the side.

11.11 Studio-monitors 11-97
The following examples show the beaming behavior of different 3-way-speakers. We see
from Fig. 11.120 how even well known manufacturers struggle: the often requested
“monotonous increase” of the directivity is nowhere in sight.
Fig. 11.120: Aperture angle and directivity of two 3-way-speakers (according to the manufacturer’s datasheet).
“Vertikal” = vertical
The data of Sentry III are shown in Fig. 11.121; this speaker already enjoys a cult-status, and
not undeservedly, as the graphs indicate. Still, we need to note that the two frequency
responses of the aperture angle are always only simplified representations of a highly complex
beaming behavior (Fig. 11.111). Also, hearsay states that there may be manufacturers who
will “lend some help” to a less optimal curve and conjure up a characteristic favored by the
sales department.

The following figures (Fig. 11.122) belong to two-way speakers. The JBL and the Altec can
easily be imagined placed in the studio while the EV-speaker is more intended for PA-use.
The 604-8L combines a 15"-woofer with a Mantaray-horn mounted coaxially with the
woofer; the two JBL’s employ so-called bi-radial horns (100°x100°), and the EV-box sports a
90°x40°-horn. None of the directional characteristics could be designated as particularly good
or particularly bad – the quality always depends on the individual deployment-location. In the
studio, this will be a relatively strongly absorbent control room where the reverberation time
is between 0.2 and 0.4 s resulting in a reverberation radius of about 1.5 m. The effective
reverberation radius [3] will then be around 2 – 6 m, and that will give the diffuse sound some
significance, after all.

11.11 Studio-monitors 11-99
Fig. 11.123 shows the reverberation time T60(f ) of two professional control rooms. For one of
them (black curve), the right-hand graph indicates the effective reverberation radii, i.e. the
reverberation radii increased by the square root of the directivity factor [3]. An engineer
listening back at a distance of 3 – 4 m from the speakers is therefore predominantly located in
the diffuse field for low frequencies. Depending on the beaming of the speakers, a very
special mixture of direct and diffuse sound results that turns out to be rather … shall we say:
“characteristic” for the JBL 4425.
Fig. 11.123: Frequency responses of the reverberation times in two control rooms (left).
The black curve in the right-hand graph indicates the effective reverberation radii for 6 different studio monitors.
As a conclusion, let us look at a few measurements regarding non-linear distortion (Fig.

11.124). The requirement to be able to generate an SPL of 80 dB at distance of 2 m is not a
very challenging one. However, if the maximum harmonic distortion needs to be kept below
0.1%, a few speakers fail, after all. Your classical studio-monitor will be able to rather reach
around 1% – that is not all that bad, but more modern, newly developed types are able to
remain clearly below the 1%-mark. Of course, the 80-dB-@-2-m is not the maximum required
SPL – that would be about 110 dB / 2m. But even at that level, the non-linear distortion
should remain “inconspicuous”.
Fig. 11.124: Harmonic distortion suppression of different studio monitors

(according to the manufacturer’s datasheet).

11.12 Loudspeaker cables
If the cable connecting guitar amplifier and loudspeaker is merely a few meters long, any
cable with adequately thick cross-section may be used. “Adequate” is a conductor cross-
section of 0.75 mm2, beyond reproach would be 1.5 mm2. Regular conductor-copper is
perfectly suitable, low-oxygen special copper – or even silver – is not required. It is entirely
irrelevant whether 49.4 W or 49.5 W, of an amplifier power of 50 W, arrive at the speaker,
and possible sound changes are certainly inaudible at ΔL < 0.05 dB. However, conventional
guitar cables are unsuitable because the inner conductor will, as a rule, be too thin. The
following table specifies the percentile power loss for a loudspeaker cable of a length of 2 m
and for a load impedance of 8 Ω:
Cu Cu! Ag Al
2
2x0.75 mm 2,33 2,24 2,10 3,76
2
2x1.5 mm >> 1,18 << 1,13 1,06 1,91
2
2x2.5 mm 0,71 0,68 0,64 1,15
Cu = regular cable copper, Cu! = high-purity copper, Ag = silver, Al = aluminum♣.
Example: with a 2-m-long 2x1,5mm2-cable you will experience, given a load of 8 Ω, a power loss of 1.18%;
instead of e.g. 50 W, only 49.4 W arrive at the speaker; with a pure-silver cable, that power rises to 49.5 W. For a
16-Ω-load, the losses are even smaller: 49,71 W and 49,74 W, respectively.
In A.D. 2014, robust cables with high-quality collets-plugs are available for less than 10 Euro
– these should be good even for professional use. Yes, cables do have a capacitance, and an
inductance, and a skin effect may also be observed – but all associated effects are completely
irrelevant compared of the loudspeaker impedance. Minimizing such aspects will increase the
price but not the relevant quality. As soon as cable comparison tests are carried out under
blind-test conditions, the inexpensive cable sounds just as good as the costly designer-cable.
“Costly” may indicate 100 Euro – but possibly much more: for 5 m loudspeaker wire, the
asking is 21.000.- €! Silver, braided. Here’s a recommendation: cut up in pieces, it makes for
a nice (alternative) necklace for the groupies.
Fig. 11.125: Loudspeaker- and cable-impedance- in comparison (left), calculation vs. measurement (right).
Fig. 11.125 indicates that the measured rise of the impedance of a relatively thick speaker-
cable is only in part due to the skin effect. The main share results from the inductance of the
two-wire line. Even this increase of the impedance is insignificant because it makes for only
less than 1% of the loudspeaker impedance.
♣
A CCA-cable (copper clad aluminum) is an aluminum cable with a thin copper coating!

11.Appendix: Metrology & instrumentation 11-101
11.A Appendix: metrology & instrumentation
In the following we will give a short overview regarding some of the devices and methods
necessary for loudspeaker measurements. More extensive information is available from
publications of the instrumentation-manufacturers, in particular from Brüel&Kjaer (Technical
Reviews).
11.A.1 Measuring microphones

In contrast to microphones used in the recording studio, the transmission coefficient TUp of
measuring microphones should be frequency-independent [3]. GUp, the transmission index of
a typical free-field microphone (½" B&K 4190), for example, varies by less that ±0.7 dB in
the range from 10 Hz to 15 kHz. This range of tolerance is, however, valid only for axial
sound incidence; as soon as the direction of the sound deviates from the microphone axis,
beaming effects that increase with rising frequency make themselves felt. In the anechoic
chamber (AEC), such beaming is of no effect since the microphone is directly pointed
towards the source. However, in the reverberation chamber (RC) with its diffuse sound field,
attenuation towards the higher frequencies occurs that can easily amount to 5 dB at 15 kHz.
For this reason, ¼"-microphones are preferred in the RC, and the fact that they are noisier
compared to the ½"-microphone is accepted in exchange. The B&K 4135♣ used for our
measurements has a beaming error of 0.5 dB at 5 kHz and of 1 dB at 10 kHz – this we
deemed acceptable.
Non-linear distortion (harmonic distortion) is far below any relevance in the microphones
used, and at the sound pressure levels that occurred. The intrinsic noise is insignificant for
the 4190 (15 dBA), and marginal for the 4135 (45 dBA). Not insignificant are the effects of the
microphone mounting: clamps and stands reflect waves and lead to comb-filter-like
interferences♥. With suitable set-ups, such errors could however be kept below ±0.2 dB.
11.A.2 Reverberation time

The time it takes for the diffuse-field SPL to drop by 60 dB in the reverberation chamber after
the sound source is switched off is specified as the reverberation time. In order to mainly
measure diffuse sound, the microphone must not be located too close to the sound source, and
to catch as many room modes as possible, the microphone should move a long a (slanted)
circular path. All measurements in the reverberation chamber were done with 50%-
overlapping 1/3rd-octave analysis (IEC 1260 class 0), with the microphone moving along a
circle (∅ = 3m) within 80 s. The microphone boom was mounted to a turntable (B&K 3299).
The latter transmitted a lot of structure-borne sound to begin with (equivalent air-borne SPL
84 dB); however, suitable decoupling reduced this value to some just-about-acceptable 45 dB.
Customarily, the reverberation chamber is excited via broadband noise to determine the
reverberation time. After switching off the sound, the slope (dB/s) of the level decay is
identified, and the reverberation time TN results from it. Typical values are 2 – 5 s, and up to
10 s in the low frequency range. Since noise processes are of a stochastic nature, it is
necessary to average over several decay processes.
♣
Brüel&Kjaer does offer a special pressure microphone (4136) that would be even more suitable.
♥
M. Zollner: Einfluss von Stativen und Halterungen..., Acoustica, Vol. 51 (1982), 268-272.

Even before this averaging (over several level-decay curves), an RMS-averaging is required
to make the transition from the sound pressure to the sound pressure level (SPL). Fig. 11.A1
clarifies this process, first via a decaying sine tone.
Fig. 11.A1: RMS-averaging. Left: exponentially decaying sine-tone (----), the square of this result (–––).
Right: exponentially averaged decay process; two different time-constants. Not yet subjected to the square root.
To get to the RMS (Root-Mean-Square) value, the signal needs to be squared (S) in a first
step, subsequently averaged (M = mean), and last the square root (R) needs to be applied.
While the squaring is an unambiguous step, taking the average is not. In metrology, two
averaging methods are predominantly used: the so-called exponential averaging, and the so-
called linear averaging – the latter should more appropriately be termed arithmetical
averaging. Averaging devices in the above sense are linear low-pass filters that are described
unambiguously by their impulse response. To achieve exponential averaging, a straight-
forward (1st-order) RC-lowpass is used; its impulse response is a decaying e-function. The
linear averaging happens in the gap-lowpass that features the unipolar rectangular pulse as its
impulse response. Both approaches to averaging may be described by one parameter each: by
the time constant τ for the exponential averaging, and by the duration of the rectangular pulse
(block length) T. Even for τ = T, the results are not the same.
The general problem of every averaging process becomes clear from Fig. 11.A1: with too
short a time constant, the smoothing is insufficient, and with a long time constant, the decay-
process representation holds errors. Fig. 11.A2 shows corresponding level-graphs: it is
evident that for the exponential averaging, the slope of the flanks depends on the time
constant. With the linear averaging, the decaying slope is merely delayed but its slope
remains. For determining the reverberation time, this implies that linear averaging has
advantages. It still is not without problems: in particular during the early decay phase, the
slope is too shallow – this could lead to the calculation of too long a reverberation time. This
early decay range is important, since real reverberation curves do not decay as ideally as
given in this example but show a degressive decay (i.e. an increasingly shallow curve).
Fig. 11.A2: Level-decay for exponential averaging (left, two different time-constants), and for linear averaging
(right, two different time-constants).

11.Appendix: Metrology & instrumentation 11-103
From a systems-theory point-of-view, averaging is filtering: the signal to be averaged is

convolved with the impulse response of the averager. In other words: every averaging is a
weighted (convolution-) integral over a range. For the linear averaging (= block-averaging or
arithmetic averaging), this integration happens over the block-length T ahead of the center-
point (in time) of the averaging. For T = 0.4 s, the linear average indicated at the point in time
of t = 2 s specifies the integral over 1.6 … 2 s. Therefore, the linear average value measured at
the time when the sound source is switched off (t = 0) does not constitute an averaging over
the decay process. Staying with T = 0.4 s: 50 % of the linear average measured at t = 0.2 s is
determined from the steady-state process, and the remaining 50% are captured via the decay
process. Only the linear average measured at t = T captures 100% of the decay process♣. It is
exactly from that point on that the slope of the level-decay is correctly shown when linear
averaging is used (Fig. 11.A2, right). Every averaging (over time) needs to happen over a
range (as elaborated above). If the duration of this range is set too short, the averaging cannot
serve its purpose: the convolution with a Dirac pulse results in the unchanged signal. Since
the averaging needs to happen over a range (the duration of which needs to be larger than
zero), all averages arrive delayed after the signal to be averaged.
If the decay process were an exponentially decaying sinusoidal oscillation as presented in the
first figure, and if the travel time in the averager were known, the slope of the process could
be precisely determined. However, already an envelope composed of two e-functions (see
Fig. 11.A3) renders determining the slope problematic: if the slope changes significantly
within the block-length T, a linear averaging will not be adequate to reliably detect this.
Fig. 11.A3: Level decay for exponential averaging (left, three different time-constants) and for linear averaging
(right, three different block-lengths). Degressive decay-curve.
As consequence, the averaging time T, or the time-constant τ, respectively, needs to be short

enough not to too much distort the shape of the space-impulse-response (or the space-step-
response, respectively), but on the other hand it needs to be long enough to average out the
stochastic fluctuations of the noise. This is because, contrary to the figures presented so far,
decay curves of reverb are not determined using sine-tones, but with noise (of a bandwidth of
an octave or of 1/3rd of an octave). If a linear averaging over e.g. 50 ms is not good enough to
average out the noise sufficiently, and if a longer averaging time would prohibit measuring
the early decay of the curved decay curve, then a further dimension remains as a way out: the
ensemble-averaging across different realizations of the noise process. This simply entails
averaging over several decay curves – however not with identical noise-excitation but using
different excerpts from a noise signal (e.g. 1/3rd-octave noise).
♣
Strictly speaking, the travel time of the sound needs to be considered, as well.

In Fig. 11.A4 we see, in the left hand section, 4 decay curves that were determined from the
squared SPL-signal via linear averaging over a block-length of 50 ms. For this measurement,
microphone and loudspeaker were in fixed locations such that the signal fluctuations are to be
attributed predominantly to the stochastic of the noise. In the right-hand graph, a multitude of
such curves is included – as is the averaging curve derived from them. For orientation, the
dashed line represents the decay for a reverberation time of 1.6 s. We can see that the latter
approximately corresponds to the early decay, while the remaining curve is shallower. The
fluctuations that still remain in the decay curve are not primarily associated with the noise
stochastic with the room. The superposition of many decay processes (with many frequencies
and different dampening) does not result in a single decay-time-constant; rather we get a
curve of any arbitrary complexity that can normally only be approximated to a straight line in
sections. For power measurements in the reverberation chamber, it is not the level range
between -5 and -36 dB that is to be captured, but rather the initial slope♣.
Fig. 11.A4: Decay curves; composed from 1/3rd-octave noise (fm = 200 Hz) via linear averaging (50 ms).
In the graph on the right, the white line represents the mean of the ensemble.
In conclusion a short comment regarding the Hilbert transform, since it occasionally is

accredited the capability to do ideal averaging. For a decaying sine-tone it is indeed possible
to derive, from the sound pressure and using the Hilbert transform, the analytical signal, and
from this a smooth decay curve. However, given the narrow-band noise commonly used for
reverberation measurements, the Hilbert transform is not an option – at least as long as it
alone is put to use (Fig. 11.A5).
Fig. 11.A5: Left: decay curved filtered with a bandwidth of 1/3rd-octave (fm = 200 Hz), linear average, T = 50 ms
and 500 ms, respectively.
Right: Level of the analytical signal belonging to the signal on the left (also termed “magnitude”).
♣
H. Larsen, Technical Review Nr. 4, Brüel&Kjaer, !978.

A. Appendix: Vibrations, Waves, and the Cryo-Schlock
A.1 Vibrations and waves
A vibration (or oscillation) is a process in which the vibration quantity (e.g. displacement,
force, current) is non-monotonic and has at least two extreme values (DIN 1311). In
mechanics, oscillations occur when masses move around their resting position. Masses are
elements of mechanical systems, as are springs and dampers. If a system contains no dampers,
it is lossless or undamped, otherwise it is lossy or damped. A mass can store kinetic energy; a
spring can store potential energy. A damper is not one of the elements that store energy; the
energy that is fed to it is irreversibly converted into heat (caloric energy). The number of all
independent energy-storage elements determines the order n of the system.
A mass that can swing up and down hanging from a spring is a second order system. In order
to fully describe the state of the vibration of the system at one single point in time, n state
variables (field quantities) are necessary. In the case of the spring-mass system, this can be,
for example, displacement and velocity. In systems-theory, the state variables are also called
signal quantities; they are to be distinguished from system quantitys (mass, stiffness,
impedance...). Physical laws of force/displacement, and structural laws (topology, Kirchhoff’s
laws for nodes and meshes) determine the interactions between the system elements, and lead
to a differential equation (DE), the solution of which yields the oscillation equation. An nth
order system is defined by a DE of n-th order. For mechanical systems with discrete (locally
concentrated) elements, the independent variable is usually the time t; the signal variables are
represented as being dependent on t in a time function.
Vibrations that occur (exclusively) under the influence of an external excitation source are
called forced vibrations. If the source oscillates mono-frequently with f0, and the system is
linear, then all state variables are sinusoidal and have the frequency f0. The term sinusoidal
allows for any phase angle, including a cosine curve. With external excitation, the descriptive
DE is inhomogeneous, i.e. it contains a so called ‘constant term’ (which need not to be
constant). Without any excitation at a given time (= homogeneous DE), the system is either
constantly at rest, or it 'responds' to previous excitations (it continues to oscillate: it 'rings').
The ability of a vibrating system to ring is due to the ability to store energy in its mass(es) and
spring(s). The ringing occurs at an Eigen-frequency (“natural frequency” being an equivalent
term) of the system, and here the terminologies of mechanics and systems-theory differ: in
mechanics, the damped oscillation is described as a product of exponential function and sine
(or cosine) to which one frequency is assigned. Systems-theory recognizes a period in the
decay process, but no periodicity (!), and assigns an infinitely wide continuous spectrum via
the Fourier integral. This is because spectral analysis is decomposition into summands, not
into factors [see literature on systems theory, e.g., 6].
If the properties of an oscillatory system are continuous functions of the place, then the
system is called a continuum. Its state variables are functions of time and place (space); the
composite of individual particle vibrations is called a wave. The dependence on two variables
leads to partial DE’s, which in turn can be homogeneous (for the free wave) or
inhomogeneous (for the forced wave).
© M. Zollner 1999 – 2014 Translation provided by Volker Eichhorst

A-2 Appendix
A.1.1 Forced vibration

In the forced (externally excited, source-excited) oscillation, energy is supplied to the
oscillation system from the outside. For example, the displacement ξ(t) of a mass suspended
from a spring can be impressed sinusoidally with the frequency f0. This simultaneously
defines the velocity, acceleration, and other temporal differentials and integrals. The force
acting between mass and spring can be calculated both from Newton's law of inertia, and from
Hooke's suspension law.
The vibrations are not easy to grasp if the displacement is not impressed on the mass end of
the spring but on the opposite (upper) end of the spring. The mass now performs oscillations
with f0, but with initially unknown amplitude and phase. In terms of systems theory, there is a
transmission – a mapping – from an input variable (displacement of the spring) to an output
variable (displacement of the mass). In the time domain, the problem is solved by the DE or
by convolution with the impulse response, and in the frequency domain by the transfer
function [6]. Considerations along the lines of the analogies between electrical and
mechanical systems allow electrical signal and systems theory to be applied to mechanical
systems as well [3]. External excitation does not mean that only the forced vibration must
exist. Despite externally forced excitement, a free oscillation can exist at the same time. This
free oscillation represents a response to previous excitations. In the case of the linear system
(linear DE), both oscillations overlap without influencing each other.
A.1.2 Free vibration

After external excitation, a free oscillation can form at a natural frequency (Eigen-frequency)
of the system. First-order systems do not generate self-oscillation♣ but exponential transient
processes exclusively. Second-order systems show exactly one natural frequency; systems of
higher order usually have several natural frequencies. In the case of the linear system (linear
DE), all possibly existing natural oscillations (equivalent term: “Eigen-oscillation”) are
superimposed undisturbed. Each Eigen-oscillation is characterized by its natural frequency
(system quantity), its initial amplitude and its phase (signal quantity), and its damping (system
quantity). In the undamped system, the damping is zero (infinite Q-factor): the vibration does
not decay. In the damped system, each natural vibration decays exponentially. Special cases
arise in case of coincident natural frequencies (multiple poles).
EXAMPLE: A second order system has the natural frequency fE. A source with a frequency
delivers an excitation signal, with the source being switched on for the duration
, and being switched off at t = 0. How does the system react? It should be noted
that the switched sine (burst) contains not only the frequency f1, but rather all frequencies (→
Fourier-Integral). During -T ... 0, a forced oscillation is generated, afterwards we have a free
oscillation. The frequency fE and damping of the latter are determined by the system
characteristics, the oscillation amplitude and phase result from the values of two state
variables at the time t = 0. If between T2 ... T3 >0, the source is switched on again, but now
operates with the frequency f2, the second forced oscillation is superimposed on the first
decaying oscillation. After T3, two free oscillations (with the same frequency fE) are
superimposed. This example assumes that the switching does not change the system
characteristics, i.e. that in particular the source impedance and the structure remain the same.
◊
♣
DIN 1311 calls creep processes "vibration in a broader sense". More precise definitions → systems theory.
Translation provided by Volker Eichhorst (with Tilmann Zwicker) © M. Zollner 1999 - 2014
A.1 Vibrations and waves A-3
The solutions of the differential equations set out in the preceding chapters do not yet give
any special indications of the actual string vibrations. For their detailed description of the
latter, string geometry and string excitation must also be known. String geometry includes
string length and bearing data; with excitation, a forced wave is generated; without (or after)
excitation, we get a free wave. With guitars, a forced wave occurs only in the short moment
when the moving plectrum makes contact with the string (attack). The following decay
process is a free wave - at least as long as the string does not hit the frets or is plucked again.
The differential equation (DE) required to describe the string vibration is initially a partial one
because the dependent variable (e.g., ξ) depends on location z and time t. The DE is linear,
because the variables occur only in the first power. Without consideration of the bending
stiffness, a 2nd-order DE results, with consideration of the bending stiffness we get a 4th-order
DE. If the system quantities (e.g., stiffness, density, geometry) maintain their values over time
– as it is assumed here - the DE contains constant coefficients. Of particular importance are
linear DE’s with constant coefficients (linear/time-invariant systems or LTI systems),
because for them the principle of superposition holds. According to this principle, any
complicated space- or time-function may be seen as the sum of individual partial oscillations,
with these partial oscillations not influencing each other. Free and forced vibrations or waves
may exist at the same time – they then overlap in their effect.
In the general case, forced and free oscillations (or waves) superimpose on the string.
Mathematics express this as follows: The general solution of an inhomogeneous linear DE is
given as the sum (superposition) of a particular solution of the inhomogeneous DE and the
general solution of the homogeneous DE.
A.1.3 Forced Wave

In the forced wave, energy is supplied to a continuum (e.g. a string) from the outside. Since
the string vibration depends on space and time, complicated excitations may be formulated.
The following simplification is important in practice: the time-dependent excitation happens
at one place. For plucking/picking with a pointed plectrum, this approximation represents a
first step, with Chapter 1.5 showing detailed results.
Further simplification is required in the definition of the excitation quantity. Rigid strings are
described by force, moment, displacement and angular velocity. Each of these quantities can
act (singly or in combination) as an excitation variable. The description may be simplified if
the excitation quantity can be defined as the direct string input quantity. We would have
indirect excitation if, for example, the defined movement of a point (finger) acts on the string
via a spring (plectrum). In the case of direct excitation, the variable acting directly on the
string (e.g. displacement) is defined via its temporal progress.
Systems theory assigns a source impedance (internal impedance) of value zero or infinity to
the direct excitation; the indirect excitation receives an intermediate value. Direct excitation
(impression of a signal quantity) is easier to describe. Particularly descriptive is the
impressing of the transverse displacement (at the constant location z0).
© M. Zollner 1999 – 2014 Translation provided by Volker Eichhorst (with Tilmann Zwicker)
A-4 Appendix
Suppose, for example, that the transverse displacement , as function of t, is impressed

at a given location z0. Given this, several questions can be formulated:
a) According to which function does the displacement develop at another location z?

This is a standard problem of systems theory, solved via impulse response and convolution
(see below).
b) Which function holds for the other signal variables (F, M, etc.)?
At one and the same location, power-generating quantities (F, v or M, w) are linked by means
of the string-impedances; the cross-linking is defined by coupling terms. The mapping to
another location is solved as in a).
c) Which oscillation occurs after the excitation has ended?
The result is a free wave, with its starting conditions defined by the final conditions of the
preceding excitation.
With respect to a): The Bernoullian approach takes as a solution the harmonic exponential,
which yields sinusoidal space- and time-functions. Since in the previous calculations the
string could be modeled as being lossless, the amplitude of a mono-frequency oscillation
remains the same during the propagation, and only the phase changes. Thus, at the location z,
again only a sinusoidal oscillation of the same frequency and the same amplitude can arise,
the phase of which is rotated relative to the excitation phase. Since the propagation velocity
for the homogeneous string (with location-independent string parameters) is independent of z,
the phase rotation is proportional to the distance traveled. Moreover, in the case of a string
without bending stiffness, the propagation velocity is frequency-independent (linear-phase
system), and thus the phase rotation is proportional to the frequency and the distance. In the
rigid string, dispersion occurs: the phase grows over-proportionately with increasing
frequency.
Arbitrary non-sinusoidal signals need to be decomposed into their sinusoidal components

(Fourier analysis), which are then mapped individually from z0 to z; subsequently the
transformed components are reassembled (Fourier synthesis). Alternatively, the entire
mapping can also be performed in one step with the convolution integral: for this purpose,
the excitation time signal is to be convoluted with the impulse response of the string. For the
string without bending stiffness, the impulse response is a time-shifted Dirac impulse
(dispersion-free delay-time system, delay line); for the stiff string, it is an all-pass function
(Chapter 2, or [6]). Reflections can be modeled as system responses to additional mirror
sources, and can then be superimposed.
As Fig. 1.19 shows, the pick/string contact can be a few milliseconds long for normal
plucking/picking. This is the length of time during which a forced vibration exists. The
excitation spreads over the entire string during this time (Fig. 1.10), so that in a strict sense we
may no longer speak of an impulse- or step-stimulus.
In most cases, the sound generated by the guitar-amp/loudspeaker can feed back to the guitar.
If high gain is employed, a significant excitation of the guitar string happens via the airborne
sound generated by the speaker. This can lead to self-excitation (howling feedback). The
overall system is active and nonlinear in this case; the oscillation can no longer be termed free
but rather self-exciting.
A.1 Vibrations and waves A-5
A.1.4 Free wave

Free waves are solutions of the homogeneous differential equation (Chapter 2). Looking at the
matter with a cursory glance, we might suspect that free waves can only exist if there are no
external forces. Indeed, external forces cause forced waves to arise – however, free waves can
always exist in addition. Since these waves contain kinetic and potential energy, they cannot
emerge out of nowhere, though.
The partial wave-DE has two variables: time and space. If both the periodicities with respect
to space and time are imposed from the outside, we speak of a space-time forced wave –
abbreviated: forced wave. If the external excitation occurs only within a limited space, so-
called spatially free waves build up outside of this space. If the external excitation is limited
in time, then a space-time free wave is created – in short a free wave.
In the previous chapters the differential equations were set up without external forces
(homogeneous DE), their solutions describe free oscillations. As an example, we will examine
a transverse wave (Fig. A3.1):
DE and solution
The solution is a harmonic oscillation with amplitude initial phase ϕ, and angular
frequency ω. To prove that the solution fits the DE, we differentiate ξ twice with respect to
space and time, and insert these derivatives into the DE:
The equation on the left may be truncated for each ξ by ξ ( is the trivial case and not of
interest). The characteristic equation yields the wave number k as a function of the angular
frequency, and of the string parameters. If k is inserted into the solution, the latter still
contains the three parameters amplitude , initial phase ϕ, and angular frequency ω. This
result indicates that transversal waves of any amplitude, initial phase and angular frequency
can propagate on the (infinitely long) string. They can – but do not have to! The actually
existing vibrations depend on the preceding excitation – the latter defines the initial
conditions, and thus the above-mentioned parameters.
A.1.5 Standing waves

A standing wave results from two waves of equal frequency but opposed propagation
direction being superimposed in one area at the same time. The complex notation of the wave
equation gives us:
Here, the cosine can be interpreted as a place-dependent amplitude term: at the zeros of the
cosine function, nodes of the standing wave arise; at their maxima, antinodes arise. The
distance between two adjacent nodes or antinodes, respectively, is one half (!) wavelength.
A-6 Appendix
If the amplitudes of the waves propagating in opposite directions are exactly equal, the result
is a pure standing wave; otherwise there is a combination of standing waves and progressive
waves (the latter also called travelling waves). A progressive wave carries energy in the
direction of its propagation, while a standing wave does not.
There are forced standing waves and free standing waves, depending on whether or not an
excitation source is present. With the guitar, the plucking/picking takes only a few
milliseconds; after that, the progressive waves reflected again and again at the bearings
generate a free standing wave, the amplitude of which slowly fades away. The Eigen-
frequencies (natural frequencies) of this standing wave are calculated from the cosine function
given above; for the lowest Eigen-frequency, the nodal distance is precisely the string length.
Fig. A.1.1 below shows for comparison a progressive transverse wave and a standing
transverse wave.
Fig. A.1.1: Progressive transverse wave (left), standing transverse wave (right); phase increment = π/ 4.
See also: https://gitec-forum.de/wp/collection-of-the-animations/ or https://www.gitec-forum-eng.de/knowledge-
base-2/collection-of-animations/.
A.2 Longitudinal waves
In a longitudinal wave, oscillation direction and propagation direction coincide. (The

transverse wave that is more important for the guitar is discussed below in A.3). For the
calculation, we divide the transmission medium into thin disks via flat, transverse, equidistant
separation surfaces. Given the wave propagation, these discs will change both their position in
the propagation direction, and their thickness. The separation surfaces are perpendicular to the
propagation direction, forming planes of equal normal stress, and equal displacement,
respectively. The change in thickness is connected to the engaging force via Hooke's Law,
which in turn is linked to the moving mass of the respective disk via Newton's law of inertia.
A.2.1 Pure longitudinal waves

The signal quantities (field quantities) are force F and longitudinal velocity v, as well as
displacement and acceleration derived from these quantities. The system quantities are the
material data sL and ρ. The longitudinal stiffness sL characterizes material deformations in
the case of large transverse dimensions, i.e. with inhibited transverse contraction. With the
guitar string, transverse dimensions are very small so that this load case does not occur. In the
string, tension in the longitudinal direction rather leads to a length-wise extension of the string
while reducing the diameter. Therefore, in addition to longitudinal oscillations, coupled
thickness-oscillations also occur. The combination of both vibrations is called dilatational
wave, contraction wave, or quasi-longitudinal wave.
A.2.2 Dilatational waves in strings
Dilatational waves (quasi-longitudinal waves) occur in transmission media that feature small
transverse dimensions with respect to the wavelength, e.g. in plates, rods, or instrument
strings. The primary forces and movements are parallel to the longitudinal axis of the string.
However, secondary effects occur in the transverse direction perpendicular to the string axis:
elongation in the longitudinal direction reduces the string diameter, compression increases it.
While the percentile changes of the transverse dimensions are very small, they are still
essential. In the purely longitudinal wave, the transverse dimensions remain constant while
longitudinal forces act; this is only possible because lateral forces act at the same time (three-
axis stress state). In the case of dilatational waves, only longitudinal forces (or longitudinal
stresses) occur, and it is precisely for this state that the elastic modulus E was defined as a
constant of proportionality. Relative change in length Δz/z, and the longitudinal stress
(= longitudinal force / cross-sectional area) are proportional:
E = Modulus of elasticity = Young's modulus
The change in diameter caused as a secondary effect in the x- and y-directions depends, via
the relative lateral contraction µ (Poisson's ratio), on the longitudinal strain:
Lateral contraction
The minus-sign is required because longitudinal increase in the dimensions results in a

transversal decrease. The dimensionless Poisson’s ratio µ is material-dependent – for steel it
amounts to approximately 0.3.
A-8 Appendix
The oscillation DE of the (lossless) dilatational wave can be established from Hooke's law and
Newton's law:
; Differential equation
This differential equation is of the same type for the longitudinal force Fz, and for the
longitudinal velocity vz or for its integral (displacement), or its differential (acceleration). For
the solution of the DE’s, the separation approach according to DANIEL BERNOULLI is suitable.
In it, a time-dependent factor and a place-dependent factor are separated (using the example
of the longitudinal force Fz below):
Proposed solution
Herein, Fz is the time- and place-dependent longitudinal force of the wave – it must not be
confused with the tensioning force Ψ in the string. As usual in signal theory, Fz is formulated
as a harmonic exponential, i.e. as a circulating pointer (phasor). Its projection onto the real
axis (the real part of the complex quantity) corresponds to the actual force; the imaginary part
(complementing the quantity to be of a complex magnitude) does not appear in practice. The
complex representation nevertheless is not more of an effort, but rather makes for easier and
shorter handling e.g. in integration / differentiation.
is the complex amplitude that contains the initial phase angle ϕ (at t = 0 and z = 0). The
angular frequency ω is connected to the time period T via 2π, just as the wave number k is
connected to the spatial periodicity (wavelength λ) via 2π. Both quantities are related via the
phase velocity cP (= propagation velocity):
Wave quantities
For the temporal partial differential, the location z is a constant – for the spatial partial
differential the time t is a constant. For a fixed time t (flash-recording), the local force is of
sinusoidal shape, as is the temporal force for a fixed location (force sensor). The term
sinusoidal allows for any initial phase; the specific value is determined by the excitation
signal. As long as the system is considered to be linear and time-invariant (LTI-system), any
signal can be synthesized by superposition. This solution approach is therefore valid not only
for sinusoidal vibrations, but for all waveforms. Fig. A.2.1 shows a snapshot of a sinusoidal
(mono-frequency) dilatational wave, in Fig. A.2.2 different phase positions are shown for this
purpose.
Fig. A2.1: Mono-frequency dilatational wave
Fig. A.2.2: Progressive longitudinal wave (left), progressive dilatational wave (right). The phase increment is
π/4. The densifications or dilutions run through the image from left to right, the partial volumes (in two cases
colored) swing around their rest position. In the longitudinal wave, the displacement is indicated by oblique lines
below the wave. The transverse constrictions of the dilatational wave are greatly exaggerated. See also:
https://gitec-forum.de/wp/collection-of-the-animations/ or https://www.gitec-forum-eng.de/knowledge-base-
2/collection-of-animations/.
A-10 Appendix
A.3 Transverse waves
In the transverse wave, all particles of the medium move transversely to the propagation
direction of the wave. To calculate the wave, the medium is divided into thin discs, as it was
for the longitudinal wave (see above). However, these slices now do not shift in parallel to the
direction of propagation, but (with an additional change in shape) perpendicular to it. In the
simplest case, the center of gravity of each disc oscillates in the same plane of vibration
(plane or linear polarization). In the general case, the center of gravity of each disc oscillates
on a space curve (e.g., z = constant, circular or elliptical polarization). In contrast to the
bending wave (see A.4), however, the flat separating surfaces of the discs always remain in
parallel. (Fig. A.3.2, Fig. A.4.2).
A.3.1 Pure transverse waves

If all particles of the medium move at the same velocity in the same direction, no wave
propagation will occur – this case is not interesting in the context of the present
considerations. In the case of a transverse wave, the transverse movement is place-dependent
and time-dependent, with non-constant dependencies in each case. This requires a change in
shape in the medium, which in the case of the pure transverse wave will deform a cuboid into
a parallelepiped (oblique prism) via the acting shear stresses. The system quantities are the
material data density ρ and shear modulus G. For a string, this state of tension would at best
be taken into account if the string were moved without the presence of a tensioning force.
Since this operating case is untypical, it will not be pursued any further.
A.3.2 Transverse waves in strings

Newton's axiom of inertia states that the velocity of a mass can only be changed by the action
of a force. Each of the sections of string (cut into slices – see above) has a mass resulting from
its density and its volume. At the disc-separating surfaces, mechanical stresses are engaging
that (averaged and multiplied by the cross-sectional area) result in one single external force
per separating surface. The separation surfaces are perpendicular to the string axis (z-axis);
each external force can be decomposed – with respect to the separation surface – into a
normal and tangential component. In the resting state of the string, the tangential component
is zero, while the normal component corresponds to the tensioning force Ψ . In a transverse
oscillation, the sections of the string move only in the transverse direction, and the normal
force thus contains no alternating component – it remains at a constant value Ψ. The tangen-
tial forces acting on both sides of the sections result – as a vector sum – in the lateral force
which is responsible for the lateral acceleration (with any losses being neglected here).
The mass of each piece of string is thought to be concentrated in the respective center of
gravity, and the directions of the external forces follow the connecting lines between the
centers of gravity (Fig. A.3.1). If ξ(z,t) describes the place- and time-dependent string
displacement (plane polarization), then represents the slope of the string, or the forces.
For a non-zero shear force to arise, the left and right slopes must be different; otherwise –
because of the same normal forces – the tangential forces would be the same and the resultant
would be zero. Consequently, a non-zero transverse force can only occur at locations where
the change in the slope (i.e. ) is non-zero, i.e. at locations featuring a non-zero
curvature.
A.3 Transverse waves A-11
For small amplitudes, the length-specific lateral force difference Fq(z+dz) – Fq(z) is
approximately: ; it corresponds to the product of length-specific mass
and lateral acceleration . The effects of shear stress are neglected here. The resulting
2nd-order partial DE is solved – like the DE for the longitudinal wave was solved – with the
Bernoulli approach:
Fig. A.3.1: Curved section of string with the two engaging external forces. The surface-normal z-component of
the external force corresponds to the constant (!) tensional force Ψ. The shear forces are tangential to the
interfaces (that are oriented normal to the z-direction). The transverse wave has two wave quantities (also called
signal quantities): the velocity, and the lateral force. The shear-force difference corresponds to the inertial force.
For both the longitudinal and the transverse wave, the solution contains a complex e-function,
with the time-dependent term ω t and the place-dependent term k z. The signal quantities that
can be described for the transversal wave (longitudinal wave) are: lateral force (longitudinal
force), transverse velocity (longitudinal velocity), and their respective temporal
integral/differential. Regarding the forced oscillation, the time-dependency of the string
oscillation (ω t) is given by the external excitation; for the free oscillation it is determined by
the geometry and the phase velocity (see Chapter A.1). The location-dependency (k z) is given
by the phase velocity cP in both cases.
It should in particular be noted that the rigidity of the transverse movement described here is
not caused by the material properties (Young’s modulus E), but solely by the string tensioning
force Ψ. For simple wave propagation over short distances, as well as for low-frequency
considerations, this description is sufficient. However, a closer analysis reveals that in
addition to the tensioning rigidity, the (material- and geometry-dependent) bending stiffness
B must also be taken into account. It causes the propagation speed of the signal to be not
constant: it rather increases with increasing frequency. In the formal description, we would
have to consider not only the forces, but also the torques and, in addition to the translational
motion quantities, the rotational motion quantities (Chapter A.4).
Fig. A.3.2: Mono-frequency transverse wave
A-12 Appendix
A.4 Bending waves (flexural waves)
The longitudinal and transverse waves discussed in the preceding chapters can be described
by 2nd-order differential equations. However, more detailed investigations on vibrating strings
show that in addition to the translational variables force and velocity, the rotational variables
bending moment and angular velocity should also be taken into account. From this, a 4th-
order differential equation results, which can now also describe dispersive (frequency-
dependent) wave propagation.
A.4.1 Tension-free beam, pure bending wave

In the case of the pure transverse wave (A.3.1), the shear stiffness (described by the material
quantity shear modulus G) counteracts the engaging shear forces. However, practical tests
show that the effective stiffness is curvature-dependent, and this behavior is better described
by a stress-loaded particle (of the medium) with nonparallel interfacing surfaces (Fig. A4.1).
Since the effective transverse string dimensions (0.2-0.5 mm) are small compared to the
wavelength (1-130 cm), shear deformations and rotational moments of inertia may be
neglected (Euler-Bernoulli theory of the bending rod). The bending stiffness B results as the
medium-characterizing quantity, being dependent on Young’s modulus E, and on the axial
geometrical moment (of inertia) I, the latter determined by of the diameter D of the
cylindrical string:
Flexural stiffness
For the solid string, D is the outer diameter; for the wound string D is, as a first
approximation, the core diameter. For more detailed descriptions, the stiffness of the
wrapping will also have to be taken into account to some degree, essentially depending on the
tension under which the wrapping was applied.
To put together the differential equation, we divide the string into differentially small slices of
the width dz. The circular separation surfaces are, in the rest state, perpendicular to the
longitudinal axis (z-axis, Fig. A.4.1). Given excitation, they can change their position and
direction, but they remain always flat and always perpendicular to the local (curved) axis of
the string. The center of each separation surface can move in the direction perpendicular to z-
axis ξ; in addition rotation in the image plane is permitted. The lateral displacement is
denoted ξ (z,t), the rotation β(z,t). The mass dm of the discs enclosed between two adjacent
separating surfaces always remains the same.
The direction β of each interface vector is at the same time the direction of the local string
axis – the latter corresponding to the local derivative (slope) of the displacement ξ:
Motion quantities
Here β is the rotation angle, is the (particle) velocity in the ξ-direction, and w is the
angular velocity of the rotation. The angular velocity w must not be confused with the angular
frequency ω! w is amplitude-dependent while ω is not.
The place dependence of the angle of rotation β causes a deformation of the disc-shaped slice
of the string. From the science of strength of materials, we know that in the case of a straight
bend, the curvature ∂2ξ /∂z2 is linked to the moment M via the bending stiffness♣ B:
Straight bend
The minus sign corresponds to a convention chosen in [11], textbooks in mechanics tend to
write a plus sign here. If the sign definition is adhered to in the same way for all calculations,
both considerations are equivalent.
The relationship between the lateral force and the moment M is directly derived from
its fundamental definition: a force F generates the moment dM =Fdz with respect to an
orthogonal axis spaced at a distance dz. In differential notation, with the sign convention
according to [11]:
Transverse force, moment
The axiom of inertia provides the relationship between transverse force, transverse
acceleration and mass of the disc. Local differentiation with respect to z yields:
Inertia axiom
For solid strings, ρ is the material density, D is the diameter, S is the cross-sectional area. For
wound strings, the outer diameter can be used for D, but we must then use an average density
for ρ, taking into account the proportion of air in the winding and, if necessary, the
difference in density between the core and the winding (Chapter 1.2).
Fig. A.4.1: Curved piece of string, deflected in the direction of ξ.
♣
The symbol M is also used for the scale length of the string, but not in this chapter.
A-14 Appendix
If the two equations above for the straight bend and the transverse force are inserted into the
inertia axiom, the differential equation for the pure bending wave results:
Differential equation
Instead of force F we could also use (on both sides of the equation) the moment M, the angle
β, the angular velocity w, the angular acceleration the lateral displacement ξ, the lateral
velocity v, or the lateral acceleration . The solution is again based on the Bernoullian
approach, assuming time- and location-functions of sinusoidal shape:
Solution
Substituting the two derivatives in the differential equation, we obtain for each F:
This 4th-order equation has 4 solutions. The positive real wave number k1 describes a
sinusoidal bending wave propagating in the positive z-direction, the negative real wave
number k2 describes a sinusoidal bending wave proceeding in the negative z-direction.
Wave number
Solving the last equation for the phase velocity yields:
Phase velocity
indicates the propagation velocity of a certain wave phase, e.g., a wave crest (maximum).
As we can see, depends on the frequency. Higher-frequency signal components run faster
than low-frequency ones (dispersion). If, however, not a particular phase is of interest but the
envelope maximum of a poly-frequency signal, it is not the phase velocity that is decisive,
but the group velocity . It is twice as large as the phase velocity in the pure bending wave.
The third and fourth solutions of the characteristic equation are imaginary. It is expedient to
define a real fringe field number k' with k'1,2 = k3,4 /j . With this, indexing can be simplified,
so that only k and k' occur. For the pure bending wave, we simply have k = ± k'; in the case of
the rigid, tensioned string, the differences are greater (Chapter A.4.2).
; ;
= Wave = fringe field (z ≥ 0).
In practice, only the fringe-field solution with a negative exponent can be used; it describes
(with increasing distance to the bearing) a locally exponentially decaying fringe field.
Fig. A.4.2: Progressive transverse wave (left), progressive bending wave (right). The phase increment is π/4 in
each case. The transverse displacements run from left to right through the image, the partial volumes (colored in
two cases) oscillate in transverse direction around their rest position. In the case of the transverse wave, the
dividing lines (dividing surfaces) remain parallel; in the case of the bending wave, the angle between them
varies. See also: https://gitec-forum.de/wp/collection-of-the-animations/ or https://www.gitec-forum-
eng.de/knowledge-base-2/collection-of-animations/.
A-16 Appendix
A.4.2 Rigid string with the tensioning force Ψ
The tensioning force generates the predominant part of the string stiffness (Chapter A.3.2). If
this tensioning rigidity is supplemented by the bending stiffness (Chapter A.4.1), we already
obtain a good approximation for the string vibration. At low frequencies, the tensioning
rigidity very clearly dominates; at middle and high frequencies the additional bending
stiffness becomes noticeable. At very high frequencies the model is no longer usable because
the Euler-Bernoulli approximation no longer applies (→ Timoshenko [11]).
For the rigid, tensioned string, the stiffness formula presented in Chapter A.4.1 is to be
supplemented to the tensioning stiffness:
Transverse force
The differential equation thus receives an additional 2nd-order term:
Differential equation
As with the bending wave (Chapter A.4.1), M, β, w, ξ, v, or could also be used (instead
of F). The Bernoullian approach provides the same solution as for the bending wave;
however, the wave numbers k are different from the fringe field numbers k':
Solution, characteristic equation
The equation – biquadratic in k – yields two real and two imaginary solutions. The two real
solutions (k1, k2) describe waves propagating to the left or right, respectively; the two
imaginary solutions (k3, k4) describe the (spatially!) exponentially increasing or decaying
fringe fields (cf. Chapter A.4.1):
As B approaches zero (flexible string, A.3.2), k converges to (limit value, from

l'Hospital, or via a series expansion of the root); k' approaches infinity – this corresponds to a
fringe field approaching zero. On the other hand, if Ψ converges to zero (bending beam,
A.4.1), we obtain (same as with the pure bending wave):
Pure flexural wave
With simplified indexing, the result for the tensioned rigid string is:
A.4.3 Flexural Eigen-oscillations (natural oscillations)

The plucking of the string causes the progressive waves to propagate in both directions. A
large number of resulting reflections overlap to form a standing wave. Since the string is a
localized continuum, DIN 1311 nevertheless recommends that the corresponding term should
not be called self-oscillation, but Eigen-oscillation or natural oscillation. Alrighty then…
After plucking, each point of mass of the string can perform a free Eigen-oscillation with an
Eigen-frequency of the string. The Eigen-frequencies (natural frequencies) are system
quantities, and as such remain independent of the excitation. The oscillation amplitude,
however, is a signal quantity with a value is determined by the excitation-signal and -location.
Each oscillation with a natural frequency is a mono-frequency process based on sinusoidal

time functions. However, the DE of the flexural wave contains a 4th-order place dependency,
and therefore non-sinusoidal location functions can arise. Eigen-vibration modes and Eigen-
frequencies result from four boundary conditions (two per bearing each). Amplitude and
phase of the vibration result from the excitation signal.
At the string bearing (bridge, nut/fret), lateral force F and lateral velocity v are linked by the
translation impedance ZFL. For example, the translation impedance caused by a mass m
amounts to ZFL = F/v = jωm. Independent of this, the bearing for rotary motion has the
rotational impedance ZML . For example, an inertia moment Θ causes a rotational impedance
ZML = M/w = jωΘ, linking the signal quantities of moment M and of angular velocity w. In the
general case, a coupling between translational and rotational motion is to be expected.
Simple boundary conditions arise when the bearing impedances converge to zero or to
infinity. An infinite bearing mass leads to transverse-displacement, -velocity, and –
acceleration jointly approaching zero. In the case when a translational impedance approaches
zero, a transverse movement may, but there is no transverse force. Analogous relationships
exist for the rotational movements. The combination of translational and rotational boundary
conditions results in the following special cases:
• Free end: lateral force F and moment M are zero.
• Clamped end: transverse velocity v and angular velocity w are zero.
• Guided end: angular velocity w and shear force F are zero.
• Supported end: transverse velocity v and moment M are zero.
A closer analysis reveals that the bearing itself is a continuum capable of Eigen-oscillations;
the bearing impedances are therefore not constants, but frequency-dependent. The different
natural oscillations of a particular string can therefore encounter different (i.e. frequency-
dependent) boundary conditions.
In the following, the flexural Eigen-vibration of a string clamped at both ends is examined.
The tensioning force is initially assumed to be zero (pure flexural wave). The solution of the
DE holds four components (in the range z > 0): a wave (jkz) running towards the bearing
(assumed to be located at z = 0), another wave (-jkz) running away from the bearing, a fringe
field (-k'z) decreasing with increasing x, and another fringe field (k'z) increasing with
increasing x. The latter is meaningless in practice. The other three components of the
particular solution derive their complex amplitude from excitation and bearing impedances.
A-18 Appendix
From the DE, as well as from the associated characteristic equation and its four solutions
, the general solution results as a superposition:
General solution
The place-dependent and time-dependent transverse displacement ξ depends on two waves

and one fringe field. The time dependency results from the complex amplitudes ξi, the place
dependence results from the e-functions. We can regard the wave ξ1 running towards the
bearing as an excitation that the other two components are dependent on:
With and γ , the last equation contains two parameters the value of which is determined by
the two bearing impedances. To calculate, the DE of the transverse displacement needs to be
converted onto the other state variables ( ):
Velocity
Angular velocity
Force
Moment
The bearing conditions for the location z = 0 can now be used in these four equations. For
example, a free bearing requires: . The left side of the equation thus is
zero; it can be reduced by ξ 1, and the signal-independent system quantities and γ can now
be determined. An analogous approach provides the reflection- and boundary-field-
parameters for the other special cases:
Free end:
Clamped end:
Guided end:
Supported end:
Here, constitutes the complex reflection factor. For = 1, the reflection is in phase, for
it is in opposite phase; at –j it is phase-shifted by -90°. γ is a fringe field factor. For
the last two of the above wave terminations, no fringe field is generated.
The fringe field is strongest at z = 0, but decays exponentially with growing z. A limit
distance can be defined as the distance from the fringe at which a drop to 1/e occurs; with the
guitar string, this is only a few millimeters. The amount of the reflection factor is always 1
for the compiled special cases; this is to be expected because the reflection happens without
damping.
At the supported end, transverse motion and moment are zero; rotational motion and lateral
force may or may not be zero. Since there is no fringe field, the place-function is of sinusoidal
shape. The oscillation looks like a standing transversal wave; the Eigen- Frequencies are,
however, not an integer multiple of the fundamental frequency. Fig. A.4.3 shows the
displacement ξ(z) for different Eigen-frequencies:
Fig. A.4.3: Standing flexural wave, supported ends, order n = 1, 2, 3.
At the guided end, the rotational movement and the lateral force are equal to zero, the lateral
movement and the moment may or may not be zero. Since there is no fringe field, the place-
function is of sinusoidal shape. The oscillation looks like a standing transversal wave, but the
Eigen-frequencies are not an integer multiple to the fundamental frequency. Fig. A.4.4 shows
the displacement ξ(z) for different Eigen-frequencies:
Fig. A.4.4: Standing flexural wave, guided ends, order n = 1, 2, 3.
At the free end, moment and lateral force are equal to zero, lateral movement and rotational
movement may or may not be zero. Since there is a fringe field, the place-function is non-
sinusoidal. Fig. A.4.5 shows the displacement ξ(z) for different Eigen-frequencies:
Fig. A.4.5: Standing flexural wave, free ends, order n = 2, 3, 4. The thin curves show the vibration without
fringe field. In reality, the case n = 1 cannot occur as Eigen-oscillation.
A-20 Appendix
At the clamped end, transverse motion and rotational motion are zero, moment and lateral
force may or may not be zero. Since there is a fringe field, the place-function is non-
sinusoidal. Fig. A.4.6 shows ξ(z) for different natural frequencies:
Fig. A.4.6: Standing flexural wave, clamped ends, order n = 1, 2, 3. The thin curves show the vibration without
fringe field.
Due to the frequency-dependent propagation velocity, the Eigen-frequencies of the flexural

vibration deviate considerably from those of the transversal oscillation. For the bar supported
on both sides or clamped on both sides (i.e. where there are no fringe fields), we obtain [11]:
n-th order Eigen-frequency
For the bar free on both sides or clamped on both sides, the Eigen-frequencies are calculated
approximately (with only one-sided consideration of each fringe field [11]) as follows:
n-th order Eigen-frequency
With guitar strings, the bending stiffness is to be considered mainly for large diameters and in
the higher frequency range – and even there still only to a small extent. The predominant
portion of the rigidity is generated by the clamping force; Figs. A.4.5 and A.4.6 correctly
show the general influence of the fringe fields – but greatly exaggerated if this were a guitar-
string scenario. Formally, the difference between the pure flexural wave and the rigid string
with regard to the force is taken into account via the following:
Beam String
However, if the bearing conditions are not as ideal as in the special cases mentioned above,
differences with regard to the flexible string may occur in the low-frequency range, as well
(blocking mass, total passage, Chapter 2.5.2).
A.5 Characteristic impedances A-21
A.5 Wave impedances (characteristic impedances)
For the progressive transversal wave, the characteristic impedance ZW (equivalent term:
wave impedance) connects the transverse force F to the transverse velocity v. For the
idealized (rigidity-free) string, ZW depends on the length-specific mass m' and the length-
specific compliance n':
Wave impedance
In this case, we have the following correspondences: Ψ = clamping force, ρ = density, D =

diameter, M = scale♣, fG = fundamental frequency, c = phase velocity. Common ZW-values are
between 0.1 and 0.4 Ns/m for solid strings. For wound strings, the flexural stiffness may be
ignored in the low frequencies-range; may simply be replaced by . The associated wave
impedances are then in the range between 0.3 and 1.2 Ns / m.
Fig. A.5.1: Wave impedance for different outer diameters. The dashed curves belong to strings with a relatively
thin core. In the E2 string (right), the influence of bending stiffness can be seen in the upper frequency range.
With decreasing string diameter, the bending stiffness loses its importance; the flexural wave becomes a
transversal wave. “Aussendurchmesser” = outer diameter, “Frequenz” = frequency; “h” (string) = B (string)
Taking into account the flexural rigidity, we encounter more complicated relationships. The
wave equation now contains, in addition to the second spatial derivative, a fourth spatial
derivative. Because of this, not only progressive waves (in both directions) occur, but also
exponentially decreasing fringe fields (in the vicinity of the bearings). The progressive waves
need to be classified into bending waves and bending-moment waves, and therefore it is
necessary to define an M/w-wave-impedance in addition to an F/v-wave-impedance. The
M/w-wave-impedance similarly connects the bending moment to the angular velocity. Both
resistances are real, but frequency dependent. In a string, the transverse dimensions are small
compared to the wavelength of the flexural wave ("thin rod"), and a simplification may be
applied: outside of the fringe fields that extend merely a few millimeters, the description for
one single wave type is sufficient. The four wave quantities are: F, v, M, w (Chapters A.4.1 &
A.4.2); given two quantities, the other two may be calculated. Fig. A.5.1 shows the F/v-wave-
impedance for the fundamental frequencies of the strings. The ratio of core-to-outer-diameter
(κ) has little effect for f = fG; for heavy strings, and high frequencies, the bending stiffness
needs to be considered, after all.
♣
In chapter A.5, M stands for a mechanical moment, and M for the scale length of the strings.
A-22 Appendix
Taking into account the bending stiffness, the F/v-wave-impedance is calculated as:
The sign of the wave impedance depends on the direction of propagation of the wave: for
waves travelling to the right (increasing z), ZWF is positive; it is negative for waves travelling
to the left. The flexural stiffness changes the F/v-wave-impedance in two ways: the summand
Bk2/c is added, and the first term is also changed because the phase velocity (c = ω/k)
increases with increasing frequency. Fig. A.5.2 illustrates, for an E2-string, the influence of
the bending stiffness on the F/v-wave-impedance. In the range of low frequencies, Bk2 can be
neglected with respect to Ψ, with the wave impedance approximately depending only on Ψ/c
(for conversions see above). At higher frequencies, the influence of bending stiffness on the
treble strings is small (G-string in Fig. A.5.2), but for the bass strings it is pronounced (E2-
string in Fig. A.5.2) – in particular given a relatively thick core (i.e. large κ).
Fig. A.5.2: Influence of bending stiffness on the

F/v-wave-impedance. E2-string (46 mil, κ = 0.42),
G-string (17 mil, plain). The solid line indicates the
total impedance, the dashed line indicates the first
term (Ψk /ω = Ψ / c).
“Wellenwiderstand” = wave impedance,
“Frequenz” = frequency
Taking into account the bending stiffness B, we have a 4th-order differential equation:
independently of transverse force and transverse string displacement, excitation with a
moment or a rotational movement is possible, as well, and the string end may be (at least
theoretically) free, supported, clamped, or guided. The idealized boundary conditions for the
force F, the velocity v, the angular velocity w, and the moment M result in:
free: F = 0, M = 0; supported: v = 0, M = 0; clamped: v = 0, w = 0; guided: F = 0, w = 0.
The free end of the string will be called into question immediately: it cannot exert any
clamping force. Even in the theoretical literature, a guided mounting is listed only for the sake
of completeness. However, it must not be overlooked here that these bearing conditions
(bearing impedances) are frequency-dependent. At f = 0 Hz a tensioning force is
indispensable, but at f ≠ 0 entirely different conditions can occur, as the following example
shows: a spring-loaded mass is defined as the string bearing; the bearing impedance thus
calculates as: Z = jωm + s/jω = (s – ω2m) / jω. For f = 0, this bearing acts like a spring – it can
absorb static tensioning forces. At resonance, however, the impedance is zero - which
implies: no force, despite movement.
A.5 Characteristic impedances A-23
In addition to the F/v-wave-impedance, the M/w-wave-impedance must also be taken into

account for the rigid string. The M/w-wave-impedance connects the moment M to the angular
velocity w. This angular velocity is not the angular frequency ω with which the string
vibrates, but the plucking-attack-dependent rotational velocity of the individual string
particles (Fig. A 4.1).
M/w-wave-impedance
In the range of low frequencies, the phase-velocity c can be calculated in good approximation
from the basic string frequency fG and twice the scale length M: c = 2M ⋅ fG. Since the bending
stiffness depends on the string diameter to the power of four, ZWM also increases with the
string diameter to the fourth (Fig. A.5.3). Given increasing frequency, c may however no
longer be taken to be constant; rather, an increase over f must be considered, especially in the
case of the bass strings (Fig. A.5.4):
Phase velocity
Fig. A.5.3: M/w-wave-impedance (left), bending stiffness (right). Solid strings (---), wound strings with thick
core (-----),wound strings with thin-core (.......). “Wellenwiderstand” = wave impedance; “Biegesteifigkeit” =
bending stiffness; “Aussendurchmesser” = outer diameter. “h” (string) = “B” (-string).
Fig. A.5.4: Phase velocity of bending wave:

solid strings (---),
wound strings with thick core (-----),
wound strings with thin core (.......).
“Phasengeschwindigkeit” = phase velocity
“Frequenz” = frequency;
“h” (string) = B (-string)
A-24 Appendix
For Dilatational waves, the wave impedance ZW is also calculated from the product of length-
specific mass m' and phase velocity c, but now the phase velocity of the Dilatational wave is
to be assumed as follows:
Wave impedance (Dilatational wave)
Here, S stands for the cross-sectional area, E for Young’s modulus, and ρ for the density.
Since no dispersion occurs, the wave impedance is frequency-independent.
For wound strings, again the ratio of core diameter / outside diameter needs to be considered:
κ = DK / DA. The winding increases the mass without significantly increasing the longitudinal
stiffness (in approximation). The characteristic impedance results in:
; Dilatational-wave-impedance for wound strings
The impedance of the Dilatational-wave-resistance is about twenty times as large as the

impedance of the transverse wave.
Fig. A.5.5: Wave-impedances for Dilatational-wave propagation. “Dehnwellenwiderstand” = Dilatational-

wave-impedance; “Aussendurchmessen” = outer diameter.
A.6 stiffness A-25
A.6 Stiffnesses
The rigid string is described by the tension-force-dependent transverse stiffness sQ and the
bending stiffness B. In the case of static displacement in the transverse direction, the string
acts approximately like a spring with the spring stiffness sQ,, whereas for transverse-wave
propagation (without dispersion) the length-related compliance n' applies:
Transverse load
Herein: F = transverse force, ξ = transverse displacement, Ψ = tensioning force, M = scale

length, R = distance between force-engagement point and bridge, L = distance between force-
engagement point and nut. Fig. A.6.1 offers an alternative to the string model as shown in
Fig. 2.5 - advantageously of being able to process a tension force running in the z-direction,
but with slight deficiencies with respect to the algebraic sign. Both sQ and n' are a function of
the tensioning force - they are independent of the elastic modulus E !
; ; .
Fig. A.6.1: Decomposition of forces in the string model. The

length-specific compliance n' is reciprocal to the tensioning
force Ψ.
Fig. A.6.2 shows typical transverse stiffnesses of the strings (compare to Fig. A.1.2).
Fig. A.6.2: Transverse stiffness; Scale M = 64.8 cm, the plucking-engagement point is 9 cm from the bridge.
“h” (string) = “B” (-string).
A-26 Appendix
While flexural stiffness may be neglected for static loading (f = 0), it should be considered for
thicker strings and high frequencies. As the string diameter increases, B grows in proportion
to D4, which would have audible effects on solid strings (chap. 1.3.1). To reduce the bending
stiffness, the heavy strings are not solid, but wound; for them B is determined practically only
by the core.
Flexural (bending) stiffness
I = area moment of inertia, E = modulus of elasticity, D = string diameter (solid strings) or

core diameter (wound strings). The bending stiffness B = E⋅I can be thought of as a length-
specific torsional rigidity, similar to how the product of modulus of elasticity E and cross-
sectional area S can be interpreted as a length-specific spring stiffness s' in the uni-axial strain
state. However, for the analogy, the reciprocal (the compliance n) is of advantage:
uniaxial strain state bending load♣
The length-specific compliance n' is compliance n per length Δz. For the reciprocal of the
compliance (i.e. the stiffness s), the reference to the length is unfamiliar at first: s = 1 /n, and
s'= 1 /n' = s⋅Δz. To get from stiffness to length-specific stiffness, we have to multiply by the
length Δz! Unfamiliar, but need be – because with the length approaching zero, the
compliance converges to zero, while the rigidity approaches infinity.
Fig. A.6.3 shows the bending stiffness for customary guitar

strings. The modulus of elasticity of all strings was assumed to be
E = 2⋅1011 N/m2; for the wound strings, only the bending stiffness
of the core was taken into account.
E4: DA = 8 ... 13 mils, solid.

B3: DA = 11 ... 17 mil, solid.
G3: DA = 14 ... 20 mil, solid.
G3: DA = 20 ... 26 mil, κ = 0.48 and 0.60.
D3: DA = 22 ... 36 mil, κ = 0.40 and 0.60.
A2: DA = 30 ... 46 mil, κ = 0.33 and 0.50.
E2: DA = 38 ... 52 mil, κ = 0.33 and 0.42.
DA = outer diameter, κ = core-diameter/outer-diameter.

Fig. A.6.3: Bending stiffness
(= “Biegesteifigkeit”)
“H” (string) = B (-string)
The bending stiffness B connects the local change of the string direction (i.e. dβ /dz) to the
bending moment M (Fig. A 4.1). The minus-sign corresponds to a sign convention.
♣
nD is also called rotational compliance.
A.7 pulses A-27
A.7 Impulses
Mechanics defines the impulse (momentum) p as the product of mass and velocity: p = mv⋅.
This definition is not used in the present book - instead, the definition of systems theory is
used: an impulse is a signal of basically short-duration, the signal being either non-zero for a
short time, or zero for the remainder of the time, or converging so fast to zero that the signal
is practically zero during the predominant part of the observation period. The terms
"basically", "predominant" and "practically" can only be specified in the individual case. The
magnitude of causal signals is identical to zero on the negative part of the time axis, while
acausal signals are non-zero even for such “negative locations on the time-axis”.
In systems theory, the symbol p stands for the complex angular frequency:
The magnitude of a bipolar impulse changes polarity (its sign) at least once, while a
unipolar impulse does not cross the zero line.
The Dirac impulse δ (t) is a theoretical signal that cannot occur in practice but is still used as
a standard signal for (theoretical) system excitation. δ(t) is always zero except at t = 0; here
δ(t) is infinite [Marko, Küpfmüller]. If, without further specification, only the "impulse" is
mentioned, either a Dirac impulse is meant, or - usually occurring in measurements - a
unipolar approximation for it.
A signal in which short-duration pulses occur in periodic repetition is called a pulse (e.g.
square-wave pulse, Dirac pulse).
Fig. A.7.1: Pulse shapes: causal rectangular pulse, acausal rectangular pulse, square pulse; (top, left to right).
Causal sine half-wave pulse, causal exponential pulses; (bottom, left to right).
A-28 Appendix
A.8 … and the Ultimate End: Cryo-Schlock
It may have been sometime around 1990. I was already somewhat known as an acoustician
and in the mail was a parcel containing a "sound improver" for me to test. I’ve since forgotten
its designation: let's just call it “Schlock”. The letter accompanying the parcel revealed that I
was holding in my hands a miracle retailing at 350 Deutschmarks (DM). It was suggested that
I please dignify it properly, and - if at all possible - recommend it to others. We scientists are
not generally adverse to anything new, so I analyzed it more closely. It looked like a plastic
clothespin with a naked semiconductor chip (called a 'die' in the trade) mounted to it. And that
was it: a 'die', not connected anywhere to anything.
Since such a clothespin cost 15 German pennies at the time, the 'die' obviously had to be quite
a terrific thingy; a lot of MIPS crossed my mind … or even MFLOPS? We had just bought
some outrageously expensive Motorola 56001 prototypes – at a grand apiece, and so those
350 Deutschmarks didn't really scare us much. Still: while we could not peek into the interior
of the Motorola ICs, all my co-workers had a sure sense that most ICs would somehow have
to be contacted somewhere. “Bonded”, my hardware engineer told me. However, that "die"
sat there on the Schlock - no contacts whatsoever. Anyway, I first read what it was all about:
"Clip the Schlock across the speaker cable – then play your music – wait for the Schlock to
calibrated itself … and you’ll notice an extraordinary improvement in sound." Any Doubts?
The scientific justification was included: "We all know from our physics course that a
magnetic field is created around each wire". Well … yeah – assuming that a current flows
through that wire. "Magnetic fields generate electrical voltages in wire turns." Okay, now I
understood. Although … wire turns?? OK, there would be some conductor loop somewhere in
the chip so that the law of induction could generate its euphorigenic effect. "Physics also
teaches that the effect of the magnetic field is bilateral: the chip reacts back onto the speaker
cable." I’ll be damned! Why didn't I come up with that myself? Of course - this was a super-
DSP that autonomously detects and corrects the deficiencies of the speaker system. Y’all
know: that ain’t easy – we indeed should give the thing some time to tune itself to the system.
Round about 20 minutes, the enclosed instructions explained, and then everything would be
set. If not: you may want to clip the Schlock to another section of the cable, and wait …
(repeat ad lib) … and after a few iterations, you'll then have found the sweet spot, for sure.
For a moment I wondered whether the sender of the parcel saw me still living in a mono-
world because he had sent me just one single Schlock. But again, the instructions had some
good advice: for standard stereo systems you buy two, and for your quadraphonic setup, you
buy four of the guys. Four Schlocks – that would’ve set you back 1400 DM. (Special note to
my fellow German senior citizens: please do not simply divide that number by two. The loss
of purchasing power brings the price up to roughly 1,400 Euros today, subject to favorable
development in certain countries run by EU partisans. Why is the semantics checker sending me a
warning now? Uh-oh – sorry, I should have said EU participants.) But back to the Schlock:
Now the system sounds perfect? Wait, there's more! Because there's current not only flowing
down the speaker cable, there's also the power cord! It benefits from a Schlock, as well, and
of course all interconnecting cables. The antenna cable was not explicitly mentioned in the
instructions, but well-established physics teach that inside ... that's right: electric currents do
flow. Um … could we ask for a discount if we buy a full complement of 8 of the Schlocks?
This episode ain’t made up, I actually had the thing on my desk. It was gray … can’t
remember much more. Sent it back – didn't want to spend any money on it. A Schlock on the
cable? Total nonsense, technically speaking. A psychologist would arrive at quite a different
assessment, though, for an auditory event is something entirely different than a sound event.
A.8 Ultimate End A-29
Why does one person just loooove Justin Bieber, but the next person doesn’t so much? And
why might those who find him super-hot today, may only look away in sheer embarrassment
in 5 years time? Why do some Stradivarius violins sound less than special when rated in a
blind test? Because the human being is a most subjective measuring device - enthusiasm will
shift the given scale pronto (You married? Brazilian? 25 years old, prospective doctor?  ).
What we hear depends on the sound – no question there. But it also depends on our subjective
judgment, which in turn depends on our experience - and on our current mood. That's where
advertising comes in: if we only believe that this super-cable sounds special, it will sound
very special to us - and we will buy it. Anyone who believes in the Schlock will actually hear
a sound improvement. It's that simple. To assist the faith, some physics come in handy, as will
a cool name (admittedly, "Schlock" would be suboptimal here), and an exorbitantly high price
helps, as well. Because those who spend a 100 grand on a stereo won't be satisfied with a run-
of-the-mill DIY-store cable. And if you can shell out merely 2 grand for the amplifier, you
might (for your next birthday) let the dealer squeeze 400 Euros out of you for the cable - and
consider yourself bit closer to the 100-grand-system. The genuine high-end customer will,
however, only turn away in disgust – ‘cause his speaker cable costs more than 10,000 Euros.
Yep! Witnessed by some trustworthy people in a Munich specialty store: Over € 10,000 – per
channel! By contrast, the "Infinite Fidelity Speaker Cable Reference-XXL" offered in the
electronics shop round the corner is kinda too cheap at € 6.90. Oh shit, that's just shipping …
the 3m braided Infinite Fidelity cost 6400 Euro. Is it even possible that they could slap on any
markup given that price dumping? How about we let an expert speak [www.highend-anlage.de]:
Mornin’.
So, what I read makes me, a longtime hi-fi distributor, blush with shame. Those who in all seriousness want to
assert that there is no difference between cables (materials, cross-section, shielding, etc.) should first learn to
listen. I've been successfully doing hi-fi and high-end distribution for more than 14 years now, and I can say
without exaggeration that with the cabeling, I can sound-wise take full advantage of a component chain, or I can
ruin it.
On the topic DIY-store cable vs. speaker cable: physically proven and therefore indisputable is that a cable is a
waveguide. Accordingly, 90% of the total electrical information happens on the outside of the cable strands.
Electrician’s cables and cheap LS-cables from the DIY store have a cross-section of 1.5 mm2, and between 8 and
25 strands. An inexpensive hi-fi loudspeaker cable with the same cross-section (Oehlbach #1040, 1.5 mm2
copper, for example) has 5 separate sub-assy’s of stranding of 40 strands each, i.e. 200 individual strands! You
don't have to be an engineer to calculate that with such a large surface maximization, considerably more signal
and thus more information reaches the loudspeaker.
It is also a fact re. DIY-store cables that they are not pulled under a protective atmosphere, and thus the outer
layer of the strands is oxidized throughout. Copper oxide has a much worse conductivity value compared to
oxygen-free copper. Thus, the negative effect in the topic “DIY-store cable vs. hi-fi cables” further increases
(fewer strands, and those few oxidized in addition).
Capacitive and inductive interactions cannot be denied, either. Every 1st-semester student of electrical
engineering can calculate the capacitive effect of a loudspeaker cable (the LS cable is similar in cross section to
a capacitor; capacitors are fundamental building blocks in each crossover for loudspeakers; thus the cable
capacitance should DIRECTLY by connected to sound-altering characteristics). The ratio of the strengths of
insulator and conductor thus plays just as large a role as the electrical properties of the insulating material. In
their Rattlesnake series, for example, Oehlbach even inserts a dummy core made of insulating material centrally
into the cable assy – in order to keep the capacity of the cable as neutral as possible. With a cable length of 3m,
a capacitance of 10 to 20 pF can easily occur. Considering that capacitors that are used for crossovers are
rejected at such a tolerance, one should really give some thought to whether the speaker sounds at home the
same way as it does in the showroom, if one does not use a decent speaker cable.
It only remains to mention that hi-fi speaker cables are largely immune to self-induction of the strands, since the
latter are twisted and thus shield each other. The cable from the DIY-store has a much simpler structure and is
merely pulled straight into the insulation without twisting.
As regards the cable cross-section, the following can be said: a circuit is very often comparable to a water circuit.
A multi-purpose, 10-cm-wide C-hose (used as a standard by the fire department) discharges will – with the
same water pressure made available by your local water provider – shoot many times the amount of water that
will trickle from of the Gardena garden hose when the neighbor is washing his car.
A-30 Appendix
The fire department taps into the same water pressure as John, your next-door neighbor, the difference being that
the FD crew have 10 times as wide a pressure line, and much shorter hose distances at their disposal (neighbor
John has another 100m on the drum, while the boys from the brigade switch to the 15-cm-wide B-hose upwards
of 20 m hose-length). Let's translate these empirical observations into physics: logically, the resistance of the
cable increases proportionally to the length, just like the resistance of the water pipe. The electrical conductivity
of the cable is defined by the reciprocal of the resistance (conductance = 1/resistance). As the resistance
increases, the conductance decreases. That much for the mathematical logic. In order to obtain a constant
optimum conductance value in the hi-fi area, experience has shown that the following rule of thumb can be
applied: use 1 mm2 cross-section per (started) 2 m cable length. As I said, this is not a law, but only an optimal
guideline, and is more applicable for full-range operation, i.e. for stereo- and main-speakers. The surround-
channels may be supplied up to 15-20m cable length via 2.5 mm2, and will incur only very minimal perceptible
losses, since much less power flows and also a completely different frequency spectrum is pushed through then,
compared to the front channels.
All these fakes are just the tip of the iceberg. Anyone who still says that he would hear, in the face of irrefutable
physical evidence, no difference between DIY-store-cables and hi-fi cables, should seriously consider whether
the components on his hi-fi chain are a good match (I would happily be available for advice), or whether he
wants to be messing around only with MP3s and pirated copies. Because THAT is not hearing/listening, and has
nothing to do with hi-fi ! March 17, 2008, Norman
Doesn’t that make for a nice point of attack: it’s a guy from the sales department! He certainly
has learned his lesson well, and brings on science and physics right away. Sure: a cable is a
waveguide (as hollow as the sales dept., then?). But why is 90% of "the information" on the
outside? That’s not a fixed percentage – it’s frequency dependent! And – begging your
pardon, dear Norman – the outside of the waveguide is a lateral surface. It does take,
however, a cross-sectional area to judge resistance. Anyway, Norman opines: at high
frequencies, the electricity flows mainly in the outer layers, and he's right, in principle. With a
1.5 mm2 cross-section, however, effects become noticeable only beyond 20 kHz – interesting
perhaps to the hi-fi fetishist, but not for guitarists. The corollary is where it get’s interesting:
the maximization of the surface. 200 individual strands! Actually, these are termed not
individual strands, but individual wires, but no matter. However, the wires would have to be
individually enamel-lacquered, as is the case with the actual RF-strands, otherwise the surface
enlargement just won’t happen. And if they are indeed individually lacquered, then the wires
also need to be individually stripped for contacting! Whether loudspeaker cables are produced
under inert gas … we don’t know that, and we'll leave it, for now. However: if the expensive
high-end cable actually has individually lacquered wires, and if the cheap DIY-store-cable
allegedly has (slightly less) single wires that oxidize on their outside (i.e. a thin insulating
layer forms), then ... hmm ... well .... And then: when actually did we last encounter a cable
that was heavily oxidized? Right, at grandma's, over 70 years old (the cable, not grandma!).
Now we turn to the student of electrical engineering ... and thus again to the physics.
Capacitance – yes, absolutely correct. But what is neutral capacitance, then? The one that is
supposedly is, at a length of 3 m, 10 - 20 pF? Or does Norman mean neutral = 0 pF, and 10 -
20 pF = rejects? Both interpretations would in any case be real nonsense: a regular cable has
70 - 200 pF per meter). Moreover, capacitors used in crossovers have some umpteen µF, thus
10 - 20 pF would correspond a tolerance of about 0.0001% - this is 10,000 times better than
the tolerance of good capacitors. Only total duds would reject a capacitor for such a tiny
tolerance. The matter of twisting strands or wires, on the other hand, cannot be dismissed:
while the cable inductance is not really bothersome at the relevant frequencies, nothing speaks
against twisting. Next comes the fire department: allright, that’s well-meant … large cross-
section, OK. For 3 m cable length, a cross-section of 1.5 mm² is a decent fit – we have already
seen entirely different, monster-ous numbers. And Norman even offers advice at the end - but
hey, what can you do: he's a salesman. Right. So now let's now take a look at the numbers, it's
1st-semester electrical engineering stuff, after all. Two-wire or coax cable, copper or silver,
Ohm, Henry, Farad, let us calculate, then.
A copper cable (2 x 1.5 mm2) of a length of 3m has a series resistance of 71 mΩ (0.071 Ω).
This is very little compared to an 8-Ω-speaker and does not even incur 1% power loss. And
that’s connected to a perfect voltage source! Tube amplifiers, however, rather are current
sources. More specifically: With the AC-30, the source impedance that the speaker 'sees'
increases, using this cable, from about 200 Ω to 200,071 Ω. Forget it. Skin effect? At 10 kHz,
it increases the cable resistance by 6% - but at this frequency, the speaker impedance
increases by over 300%, so again: forget it. Cable capacity? That will only start to play any
role in the MHz-range, compared to those few loudspeaker-Ω’s. Cable inductance? That’s
about 1 µH, and thus about 1000 times smaller than the speaker inductance. Conclusion: it’s
all insignificant. Completely, utterly insignificant – just forget it.
Anything else? Yes! One manufacturer has another argument up his sleeve: cables carrying a
current will attract. Forces are at work here, and only a particularly robust cable can with-
stand these (namely the cable of that particular manufacturer). Forces, right! Referenced to the
area and thus also referred to as pressure. Given your typical currents, that will amount to, by
and large, 1 Pa (1 Pascal). Is that a lot? Yes and no, depending on the point of view – every-
one has to decide for themselves. To give a guideline: the atmospheric pressure on our earth is
100000 Pa. The pressure (force) between two wires will be a 100000th of that. Oh wow!
What have we learnt? Physics mumbo-jumbo, cool names, and ridiculous prices will always
find their victims. If you want to connect a speaker, you don’t really have to be trained in
electrical engineering. It does help to blow the cover of dubious bullshitters, though. All the
above physical effects are relevant for TV cables, but not for speaker cables. Forces are
generally irrelevant, mechanical strength, shielding, and appearance may be important.
APPEARANCE? Of course, that’s the only justification for that 10-grand-cable. Who puts
steel rims on a Testa Rossa? Exactly, there you go! Once you've graduated from Wall-Mart to
Fender … as soon as you can play not Just in C-major but also in B-locrian, you really do
need a professional cable. And a Schlock. It audibly improves intonation and timing!
Now, after that … there’s one more height we can climb: "cryo tuning". We all know from
physics (wicked, ain’ it!?) that metals are of crystalline structure. Unfortunately, the Lord has
not seen to it that these crystal structures are always perfect, they include dislocations and
other lattice defects. However, if you cool it all down a lot, and heat it back up to room
temperature, the crystal grid has sorted itself out. The metal is somehow jivin’ with a cooler
goove, if ya know what I mean. It’s true! Even patented! Word! So plunge your cables, your
jacks … what the heck: your whole guitar and plectrum into liquid nitrogen (or at least bring
it close) … and Hendrix would've gone pale with envy had he heard you afterwards. Now,
let’s take a trip to the southern climes – Sicily, for example. Compared to the Norway-
vacation, something catches our eye: "Jeez, these be wee folks here!" A tad more
demographically: the average height of southern European residents is significantly smaller
than that of northern Europeans (who are thus vertically less challenged). And another thing is
striking: though the Sicilians often pursue an activity that does not reveal itself at first glance,
they go to Catholic church every Sunday. In the North, however, Protestantism dominates. So
what do we learn from this? Catholicism causes dwarfism! Not convinced? You think
someone has confused correlation with causality? Never mind: in sales, that’s neither here,
nor there. If Cryo-Tuning helps butcher knives to be sharper, it also has to improve the guitar
jack. We also get to learn how it works: the resistance is reduced, due to crystal lattice order.
Cleaning up has never done any harm, not even in the crystal lattice. So: send in a jack for
tuning, or buy a socket tuned that way (now at triple the price), install it, and ... yeah, what
really? The pickup has e.g. 6000 ohms, the amplifier e.g. 1000000 ohms, and the socket
sports a mere 0.009 ohms instead of 0.01 ohms. And is now referred to as a Cryo-jack.
A-32 Appendix
We do not want to allege that cryo-treatment doesn't change anything. Let’s believe the
advertisement that "the goal of any heat treatment is to turn as much austenite into martensite
as possible" [www.cryo-tuning.de]⊕. That's about carbon steel, though, not Telecaster bodies
– those ain’t made of steel. And if that jack is indeed made of hardened steel, and – after the
cryo-treatment – has become "considerably more conductive" because of the now denser
molecular structure … that couldn’t be any more immaterial. Really. It will so completely
NOT matter. Not everyone believes this, however: "The Cryo-treatment results in a denser
molecular structure, and thus electrical components such as cables, pickups, pots are
considerably more conductive" [www.georgeforester.de]. Pots – aha! Does the 500-kΩ-
potentiometer then only boast 400 kΩ after treatment? Would that really be what is desired?
Or a cable dealer: "What exactly does the cryo-treatment do? If materials are exposed to these
extreme temperatures for a defined time-temperature span, a kind of pressure builds up in the
material. During "relaxation", a change in the molecular structure occurs. The molecules
align themselves more uniformly, bind more firmly, and form a much more stable structure.
This considerably improves the electrical conductivity and thus the transmission of energy
and signals. For example, copper or silver, or other electrical conductors and connectors,
have a freer flow of electrons. The sound enhancement is just enormous." Ouch! That really
hurts: the freer flow of electrons in copper or silver? How the heck do people come up with
such nonsense? Presumably via inferences of analogies. An example: a well-known author of
a well-known magazine offers a workshop in which the barbecuing of a guitar is celebrated.
A large stove, a large pan, in it some oil - and a Tele-body. Roast on each side for 3 minutes,
because: science has found that this procedure leads – for beef – to a huge increase in flavor.
For the guitar, it’s similar …
But seriously: Deep-freezing (DCT, deep cryogenic treatment) generally speaking is not
nonsensical. The method is known and it’s used, e.g., for tool steels. Or for racing engines, or
other highly stressed metal parts. The benefit: increased hardness, retention of sharpness,
ductility, abrasion resistance, service life, in short: everything that's important for drills,
milling cutters, and the like. When it comes to the electric guitar, only those who completely
wear down their Strat vibrato every 6 months should think about cryo-treatment. For steel
strings ... not completely nonsensical … might be worth a try. If the tensile strength (breaking
load) is increased, and nothing else gets worse ... better ask some metallurgists. With wound
strings, however, fat (tallow, dust ...) will settle in the grooves of the wrapping despite the
cryogenic treatment, and that is in essence the sound killer, since a "more homogeneous
molecular structure" won’t help at all.
Of course, charging € 600.- for cooling down complete electric guitars is lucrative business -
purely on the basis of the promise that the sound would somehow be better: "High-frequency
and harmonic vibrations are significantly reduced or eliminated by de-stressing the
workpiece". If harmonic vibrations are significantly reduced, are then inharmonic vibrations
amplified? Just how nonsensical all this is can be seen from the controversial statement:
"Guitars that have received a cryo-treatment sound much more harmonious". Here, all and
any nonsense that the ad writer can come up with is thrown in: "Cryogenizing hi-fi and audio
devices, as well as amplifiers for electric guitars and electric basses, makes for a richer
sound. On CDs treated that way, bass frequencies were more contoured and comparatively
louder". However: "Due to the structural change of the conductor, you have to expect a new
playing-in time (i.e. a time until the instrument or device has its optimum sound), which is
different depending on the cable." That’s very much like the "Schlock". Don't forget: "Holes
in the internal structure of the material prevent the oranised movement of electrons (sic)."
⊕
More precisely e.g. in Rösler, Harders, Bäker: Mechanisches Verhalten der Werkstoffe, Teubner, 2003.
The improvements found in cutting and bearing metals apparently translates 1:1 to guitar
woods, pickups, switches, pots, jacks, cables, just about anything that can make somebody
some money. One violin-maker says that treated wood is much easier to work with because it
is absolutely relaxed after treatment. The next one says that after the treatment, the service
life would be extended, and another one confirms that the tone sustains longer. An enterprise
located in the neighborhood (dealing in screws, perhaps?) has identified further areas of
application: "Deep relaxation through cryo-screwing! Weekly changing experts (m/f)
guarantee longer service life!" The fact that occasionally lacquer or bindings tear open can
also be seen as a positive: "The sound gain through the broken up paint is extremely clear to
hear and feel." That’s for a solid-body guitar, mind you. Brass instruments, too, benefit: "less
blowing resistance, rounder resonance-tone." A counter-opinion: "I was unable to detect any
differences in the instruments after cryogenic treatment that I could attribute to the treatment
[trumpet forum]". Let's hear from another fellow contributing to the thread: "I imagine that
somehow it would be interesting to investigate the placebo effect: just go and pay 500 Euros,
and store a guitar for 4 weeks somewhere in some closet. And then: it was really worth it,
that's the sound! [Tsb.olaf]". This last comment at least gives some hope: there's still a bit of
sanity in this world.
Placebo! And many feelings of inferiority that must be concealed via gimmicks. If it can’t be
that guitar for 100,000 Euros, then at least get a cryo-jack. If playing Highway Star isn't
working out all that well, then at least the electrons in the tuned cryo-cable should. Onto
which we should clip a “Schlock”, as we’ve learned. And, finally, a tip for the home stereo
system (or the home studio): we all know from physics that cables lying on the ground
transmit bass sounds not as well (Faraday’s ground effect). Ultramax has recently launched
the Cryno-Cablespacer, which keeps the speaker cable at a constant 9.4 cm above the floor
(recommended spacing: apply a Cablespacer every 10 cm). Cryno is a completely new
development: two different nanometals are alloyed into a bimorph, and then optimized with
cryo-tuning. The fact that these fullerenes have excellent properties has been proven for a
decade – but only now has Ultramax managed to maintain stability in production. 5 year
warrantee! On its way from the amp to the speaker, the cable is supported every 10 cm, so
that a constant capacitance area sets itself up, providing an optimum stationary-wave load to
the speaker current. Audibly better transients, because wave dispersion no longer occurs;
fuller bass because the Faraday effect is compensated. Optimal stereo processing, due to the
matching of the group delay. PPP only €997,00 (excl. VAT). Ask your high-end dealer for it!
If he is not stocking the Cablespacer yet, stop being a customer there! You owe it to your ears,
after all: you only have these two … … and hopefully no hollow waveguide in between!
Actually, Norman, our sales guy (see above) was unintentionally close to hitting on the truth:
"All these fakes (sic) are only the tip of the iceberg". A small spelling mistake – excusable.
No, not facts, indeed it should have spelled fakes ... 1
Despite the absurdity shown on the left: listening tests

with professional guitarists who were initially
convinced by the sound of "their" cables, proved
unambiguously: in a blind test, the professional simply
cannot hear any differences in speaker cables; for
guitar cables, though, differences are easily audible,
but the cause is exclusively the cable capacitance.
Other quality factors are limited to mechanical
Source: www.sg-akustik.de; Free delivery! properties, and possibly to the shielding effect.
1
translator’s note: … a term that since the year 2017 has attracted a lot of attention even in the White House,
for better or worse.
A-34 Appendix: Epilog
Epilogue
Somewhere in Africa, an explorer arrives amongst a group of people who, back in the day at
the time of this encounter, were still called "negroes" in some parts of the world, rather than
"people with maximum skin pigmentation". Anyway, the members of the group were all
sitting in a circle and drumming away: wild, polyrhythmic sounds the transcription of which
would go beyond the framework sought here. Very exotic, fascinating for a Westerner – and
scary at the same time. Come nightfall, they are still drumming fiercely. Far beyond midnight,
the explorer wants to go to sleep, but the drums rise to an ecstatic inferno, and he cautiously
dares to ask: "You guys are really having at it – how much longer is all this going to last?"
There’s no answer, just vacant stares. He asks again, and again no answer … The man
summons up all his courage and asks again … and finally: "We always drum. When the drums
go silent, the HORROR will come". Uh-oh. "... so what is this horrible thing, the HORROR?"
"The HORROR is unspeakable – its name may not be uttered!"
The explorer tries the next guy: "How long are you all going to keep playing tonight?" "We
always play, when the drums are silent, the HORROR will arrive". "But what is the
HORROR?!" "The HORROR is inexpressible!" "And nobody may say what it is?" "Only the
chief may". So the explorer goes to see the chief: "Exalted chief, your guest from distant lands
is very tired and seeks to sleep. When will the drums go silent? " "The drums are never silent,
for if they are silent, the HORROR will come!" "Yes, I have already heard that - what is this
horrible thing, then?"
… "When the drums are silent, the horrible HORROR-thing comes: the bass solo begins!"
This knowledge from distant lands and distant peoples teaches us that for the bass, completely
different, unexpected rules hold. As a guitarist, you do not understand these rules, and
therefore you should never transfer results obtained and valid for the electric guitar to the
electric bass just like that. An electric bass in not simply a tuned-down guitar – it is played
differently, it needn't be 50 years old, its neck resonances interact differently with the strings
than those on a guitar. "Physics of the Electric Guitar" describes exactly the latter - otherwise
it would be "Physics of the Electric Bass".
Time to turn off the computer, and to listen to music again: "Take me down little Susie, take
me down, I know you think you're the queen of the underground. And you can send me dead
flowers every morning. Send me dead flowers by the mail. Say it with dead flowers to my
wedding, and I won't forget to put roses on your grave♣". Take me down ...
♣
Jagger / Richards aka Nanker / Phelge
A-34 Appendix: Epilog
Epilogue
Somewhere in Africa, an explorer arrives amongst a group of people who, back in the day at
the time of this encounter, were still called "negroes" in some parts of the world, rather than
"people with maximum skin pigmentation". Anyway, the members of the group were all
sitting in a circle and drumming away: wild, polyrhythmic sounds the transcription of which
would go beyond the framework sought here. Very exotic, fascinating for a Westerner – and
scary at the same time. Come nightfall, they are still drumming fiercely. Far beyond midnight,
the explorer wants to go to sleep, but the drums rise to an ecstatic inferno, and he cautiously
dares to ask: "You guys are really having at it – how much longer is all this going to last?"
There’s no answer, just vacant stares. He asks again, and again no answer … The man
summons up all his courage and asks again … and finally: "We always drum. When the drums
go silent, the HORROR will come". Uh-oh. "... so what is this horrible thing, the HORROR?"
"The HORROR is unspeakable – its name may not be uttered!"
The explorer tries the next guy: "How long are you all going to keep playing tonight?" "We
always play, when the drums are silent, the HORROR will arrive". "But what is the
HORROR?!" "The HORROR is inexpressible!" "And nobody may say what it is?" "Only the
chief may". So the explorer goes to see the chief: "Exalted chief, your guest from distant lands
is very tired and seeks to sleep. When will the drums go silent? " "The drums are never silent,
for if they are silent, the HORROR will come!" "Yes, I have already heard that - what is this
horrible thing, then?"
… "When the drums are silent, the horrible HORROR-thing comes: the bass solo begins!"
This knowledge from distant lands and distant peoples teaches us that for the bass, completely
different, unexpected rules hold. As a guitarist, you do not understand these rules, and
therefore you should never transfer results obtained and valid for the electric guitar to the
electric bass just like that. An electric bass in not simply a tuned-down guitar – it is played
differently, it needn't be 50 years old, its neck resonances interact differently with the strings
than those on a guitar. "Physics of the Electric Guitar" describes exactly the latter - otherwise
it would be "Physics of the Electric Bass".
Time to turn off the computer, and to listen to music again: "Take me down little Susie, take
me down, I know you think you're the queen of the underground. And you can send me dead
flowers every morning. Send me dead flowers by the mail. Say it with dead flowers to my
wedding, and I won't forget to put roses on your grave♣". Take me down ...
♣
Jagger / Richards aka Nanker / Phelge

Physics of The Electric Guitar - Manfred Zollner

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Physics of The Electric Guitar - Manfred Zollner

Uploaded by

Copyright:

Available Formats

0.

Getting into the mood (Satire) 0-01

1. The fundamentals of string oscillations 1-01

2. The string as a waveguide 2-01

3. String magnetics 3-01

5. Magnetic pickups 5-01

6. Piezoelectric pickups 6-01

9. Guitar Circuits 9-01

10. Guitar amplifiers 10-01

11. Guitar loudspeakers 11-01

Supplement: Vibration and waves A-01

Animations (see www.gitec-forum.de, no download possible)

Some Words from the (main) Translator © Tilmann Zwicker, 2020 2

I hope many readers get something out of this translation.

Be well and take care - live long and prosper,

Some Words from the (main) Translator © Tilmann Zwicker, 2020 3

Hi there … or rather: “Grüß Gott” – as we say in the south of Germany!

Cordially and guitar-istically yours,

Preface – by the author

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker, 2014 & 2019

Chapters 4 and 5 describe magnetic pickups. Extensive and varied measurements

Relatively short, Chapter 6 is dedicated to piezo pickups. Combining quadripole-theory,

© M. Zollner & T. Zwicker, 2014 & 2019 Translation by Tilmann Zwicker

Regensburg, in autumn 2014

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker, 2014 & 2019

0.1 Barking up some (wrong?) tree

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

"So: mellow-ly rocking?" "That’s for ash, of course."

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

And some supplementary opinions1:

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

0.2 Signature guitars

Jeff Beck’s Fender-Signature-model is already marketed in a second edition. He himself (as

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

... and many more ...

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

0.3 How the vibrations of the strings arrived in the Orkus

Fig. 0.1: Vibration threshold

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

0.4 The sound of the unamplified guitar

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

Fig. 0.3: Decaying sine-oscillation, f = 1000 Hz, time constant τ = 0,3 s.

Nay, that's past praying for [Shakespeare].

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

0.5 The Growse-Glowsock Affair

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

Translation by Tilmann Zwicker © M. Zollner & T. Zwicker 2009 & 2019

Counter statement, on behalf of Ms. Glowstock, Phy.D.

1) My name is not at all Growse-Glowsock but Grous-Glowstock.

August 24, 2010, Anna-Katerczyna Grous-Glowstock

Statement by the author:

August 25, 2010, Prof. Dr.-Ing. Manfred Zollner

As a stringed instrument, the guitar belongs to the subgroup of composite chordophones/lute

1.1 Transversal waves

© M. Zollner & T. Zwicker 2002, 2020 Translation by Tilmann Zwicker

Fig. 1.1: Propagation of a transversal wave on a clamped string.

© M. Zollner & T. Zwicker 2002, 2020 Translation by Tilmann Zwicker

Fig. 1.3: Time function of the particle velocity of the

In this model-consideration it is important to distinguish the actual string oscillation

1.2 Wound strings