Professional Documents
Culture Documents
OliverCage ElectronicMeasurementsAndInstrumentationt PDF
OliverCage ElectronicMeasurementsAndInstrumentationt PDF
'**
electronic
and instrumentation
D CD
&9
W,
CD
Si
measurements
"\
MW
TS6
ELECTRONIC
MEASUREMENTS
AND
INSTRUMENTATION
Advisory Committee
Frederick E. Terman, Vice President, Emeritus, Stanford University
Ernst Weber, President, Polytechnic Institute of Brooklyn
of Electrical Engineering,
University of California,
Berkeley
Steering Committee
Edward
E. David,
Jr., Bell
Telephone Laboratories
It.
Rambo, Stanford
University
Books
in Series
Vol. 1: Jamieson
Vol. 2: Bennett
Vol. 3
Vol. 4;
1966
Vol. 5: Luxenberg
and Kuehn.
Vol. 6: Balakrishnan.
Vol. 7:
Vol. 8
Vol. 9: Schwan.
and Milsum.
Huelsman. Active
Vol. 11:
Filters:
12
ELECTRONIC
MEASUREMENTS
AND
INSTRUMENTATION
EDITED BY
Bernard M. Oliver
Vice-President for Research
Hewlett-Packard
and Development
Company
John M. Cage
Hewlett- Packard Laboratories
Hewlett-Packard
Company
Copyright
1971 by McGraw-Hill, Inc. All Rights Reserved.
No part of this publication may be reproduced, stored in a
retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording, or otherwise,
without the prior written permission of the publisher.
Library of Congress Catalog Card Number 71-124141
07-047650-0
When
ordering this
TOSHO PRINTING
CO., LTD.
title
use ISBN
TOKYO JAPAN
0-07-085544-7
CONTENTS
Series
Purpose
Contributors
Preface
xv
Further Preface
One
xvii
BASIC PRINCIPLES.
1-1
1-2
1-3
Two
xi
xiii
SINE-
WAVE
TESTING
11
OF LINEAR SYSTEMS
16
Mathematical Background 17
2-2 Gain or Loss Measurement
19
2-3 The Measurement of Phase
23
2-1
Network Analyzers 29
Measurement of Delay Distortion 30
The Measurement of Loop Gain 35
The Measurement of Nonlinearity 40
2-4 Automatic2-5
2-6
2-7
Three
42
TESTING
OF
LINEAR SYSTEMS
and Techniques 46
Relations between Transient and Sinusoidal Responses 47
Response to Generalized Inputs 49
Effect of Low-end Cutoffs on Square-wave Response
57
Time-domain Re flee torn etry 61
3-1 Tools
3-2
3-3
3-4
3-5
44
vi
Four
Contents
IN
THE
65
Mathematical Background 66
Measurement of Noise 80
87
4-3 Measurements with Noise as a Test Signal
93
4-4 Measurements by Pseudorandom Test Signals
4-1
4-2
4-5
Measurement
97
Fourier Transform
5-2
Statistics
5-3
Signal Analysis
5-4
Summary
FREQUENCY
6-1
6-2
6-3
6-4
6-5
Review
Theory
108
112
114
150
AND
TIME MEASUREMENTS
157
Definitions and Standards
Standard Frequency and Time-Signal Broadcasts
Time and Frequency Standards 162
Frequency Measuring Instruments 172
Frequency Synthesizers 179
Eight
7-2
188
Direct-current Amplifier Considerations
Direct-current Amplifier with Automatic Reset
7-3
Differential Amplifiers
7-4
Chopper Amplifiers
158
186
197
201
204
8-2
8-3
8-4
8-5
8-6
8-7
8-8
8-9
8-10
8-11
8-12
8-13
8-14
8-15
56
Time
106
210
Contents
Nine
IMPEDANCE MEASUREMENT
vii
264
9-1
9-2
269
272
Inductors 275
Resistors
Capacitors
Meter Methods
9-3
to
Measure Impedance
276
Direct-current Meters
276
Capacitance and Inductance Meters
Complex Impedance Meters 280
279
9-4
281
9-5
Classification of
288
Four-arm Bridges
291
9-6
291
303
Networks 305
Resonance Methods 308
The RF Meter Methods 310
311
Precision Measurements
Standardization of Impedance Units
9-7
Methods
Ten
Measurement
311
313
10-2
10-3
10-4
10-5
10-6
10-7
10-8
10-9
10-10
Eleven
of Precision
319
Signal Source
319
The
The
The
The
Phase-shift Oscillator
Ring
Oscillator
337
338
342
350
OSCILLOSCOPES
11-1
The
11-2
Oscilloscope
11-3
Oscilloscope Function
CRTs
351
353
360
Contents
Twelve
11-4
CRT
11-5
General-purpose Oscilloscopes
11-6
Sampling Oscilloscopes
11-7
Special-purpose Oscilloscopes
11-8
Accessories
365
371
398
407
418
RECORDERS
12-1
12-2
12-3
12-4
12-5
12-6
Thirteen
Storage-target Characteristics
427
440
13-2
13-7
13-8
Slew Limiting
13-3
13-4
13-5
13-6
Fourteen
480
MEASUREMENTS
505
ON
TRANSMITTERS
AND
RECEIVERS
14-1
14-2
14-3
14-4
14-5
14-6
14-7
14-8
14-9
14-10
Fifteen
15-2
15-3
507
484
544
Contents
15-4
15-5
15-6
15-7
15-8
Sixteen
584
595
16-4
16-5
Application Considerations
16-6
16-3
16-7
16-8
16-9
16-10
16-11
16-12
16-13
16-14
16-15
Wave
mW
604
608
632
Spectrum Analyzer 633
Multifilter Real-time Spectrum Analyzer
634
The Tracking Generator Counter 635
Techniques and Applications for Analyzers 637
Some Rules of Thumb for Choosing Bandwidth and Other
Control Settings when Viewing Pulsed RF Spectrums 651
Electromagnetic Interference Measurements 653
The
Analyzers
TRF
17-1
Reflection
17-2
Attenuation Measurements
Ferrite Devices
661
17-3
17-4
17-5
17-6
17-7
17-8
Eighteen
16-2
Seventeen
Solid-state
654
655
659
18-2
18-5
18-6
Thermal Voltages
18-7
Scanners
18-3
18-4
18-8
18-9
Index
ix
709
712
713
Automatic Analyzer Systems 714
Automatic Test Systems 716
721
678
704
705
SERIES PURPOSE
expanded in
explosive rate at which knowledge in electronics has
presentastate-of-the-art
unified
recent years has produced the need for
electronics.
of
fields
individual
of
pictures
authoritative
tions that give
The
and
matter
research or in educational programs dealing with the subject
involved.
The volumes are written with a viewpoint and at a level that makes
them suitable for reference use by research and development engineers
and scientists in industry and by workers in governmental and university
They are also suitable for use as textbooks in specialized
laboratories.
The complete series of volumes will provide
courses at graduate levels.
of electronic
a reference library that should serve a wide spectrum
xii
Series Purpose
Advisory Committee.
ticular volume.
Over-all
management
hands
of the
Consulting Editor.
Frederick
Emmons Terman
CONTRIBUTORS
STEPHEN
ADAM
F.
PAUL BAIRD
ALAN
ARNDT
L.
B.
BESSER
BERGH
B.
WILLIAM HEINZ
BILL
KAY
CHARLES KINGSFORD-SMITH
WILLIAM McCULLOUGH
xiv
Contributors
RICHARD
Y.
MOSS,
II
Corporation
E.
TUTTLE
CRAIG WALTER
LARRY
A.
WHATLEY
PREFACE
This book
electronics,
is
and
in other chapters.
had obsoleted
advanced greatly since these works were written, I had personally been
involved with much of that advance, etc. Overcome by this suasion, I
agreed to accept the assignment. My attitude at the time seemed
euphoric, but now appears to me to have been more one of conceit, for as
humbled me
I approached my task closer, it loomed ever larger, and soon
to delude
which
tried
I
during
time
long
a
After
paralysis.
of
into a state
could simultaneously write a book, be president of the
IEEE, and carry out my regular job, a time during which my sense of
his
guilt steadily grew, I had the good fortune to have John Cage offer
myself that
Preface
xvi
schedules.
It is
and
time.
in
many ways
superior to
B.
M.
Oliver
FURTHER PREFACE
As our
complexity of
spectrum analyzers?
Several years ago, McGraw-Hill suggested that B.
M.
Oliver, Vice
Further Preface
xviii
President
for
Research
and
Development
of the people
who
of the Hewlett-Packard
could be logically chosen to write a
and
love.
John M. Cage
ELECTRONIC
MEASUREMENTS
AND
INSTRUMENTATION
CHAPTER ONE
BASIC PRINCIPLES
Bernard M. Oliver
Hewlett-Packard
Company
problems.
Electronic measurements are of two kinds; those
made
of electronic
1
proving
mainly
consider electronic instrumentation as a tool of its own trade.
This
avoids a great deal of repetition and superficiality and at the same time
represents very little loss of generality, since the first step in measuring a
nonelectrical physical variable electronically is to convert the variable
to
an
it is
shall
electrical quantity.
1-1
would
be
no
science.
When Lord
Kelvin warned
that knowledge not expressible in numbers was "of a meager and unsatisfactory kind," he was not expressing a fetish; he was identifying an essen-
laws,
and
fifth
their motions
earth's
own
planets at
all
two degrees.
two
Basic Principles
became
his assistant.
Kepler was assigned the task of computing the orbit of Mars from
Brahe's observations. After four years of arduous work Kepler came to a
painful conclusion.
be allowable.
Kepler then went on to discover his famous laws of planetary motion.
Eighty years later Newton showed that all these laws were a consequence
of his own laws of motion and his theory of universal gravitation, and
thus provided convincing proof of the latter. Shattered forever were
the crystalline spheres that carried the planets in their Ptolemaic orbits.
All the complex motions of the planets, which had puzzled men for ages,
were
distilled into
tion
we yet have
is
ments
which subsumes
of celestial
mechan-
of the atom,
answers.
elements.
Imagine how different human history might have been had Aristarchus
Samos had a telescope and spectroscope, and Hippocrates a microscope!
What Greek could have believed in the quintessence of matter having seen
the mountains of the moon and spectral lines of earthly elements in
sunlight?
Or who could have insisted that all heavenly bodies revolved
around the earth, having beheld the satellites of Jupiter? How could the
deity have been so wasteful as to adorn the sky with stars not even visible
to man's naked eye?
What need for evil spirits if microbes cause disease?
The impact of such discoveries, had they been made by the Greeks, would
surely have greatly accelerated civilization and profoundly
affected
theology.
Indeed the western world might have been spared the dark
ages and the tortures of the Inquisition if only the Greeks had had
better
of
instrumentation.
In recent years both astronomy and biology have taken new leaps
forward, again because of new tools, this time the result of progress in
electronics.
The radio telescope has enabled astronomers to study the
matter between the stars in what was once thought of as simply space.
Quasars, perhaps the most distant objects in the universe, and pulsars,
believed to be star corpses composed almost entirely of neutrons, have
been
discovered with radio telescopes.
Meanwhile, the electron microscope
has revealed single strands of
and many of the fantastic transfer
DNA
Basic Principles
mechanisms
antibodies,
certain,
to
and experience." It took Whitney two years, during which time not a
gun was produced, to develop the machines, tools, and fixtures to
do the job. Washington officials became nervous at the delay, but finally
Whitney appeared before the Secretary of War and other Army officers
with boxes containing all the parts of his musket. While they watched in
amazement, Whitney assembled ten muskets, taking parts indiscriminately from the boxes. Afterward, in a letter to Monroe, Jefferson
wrote: "He (Whitney) has invented molds and machines for making all
the pieces of his locks as exactly equal, that take a hundred locks to
pieces and mingle their parts and the hundred locks may be put together
by taking the pieces that come to hand."
Accurate measurement is needed too for economy of design. A bridge
single
its
to survive this
worst load safely. For millions watching on television, the most dramatic
moment of the Apollo 11 mission occurred when Neil Armstrong first set
But for many of the engineers who designed the
foot on the moon.
vehicles and the computer programs, the most dramatic moment occurred
two hours earlier when the lunar landing module set its feet on the moon.
At that moment, only ten seconds worth of fuel remained. Close timing
costs
pound
of
spare fuel that did not have to be allowed for in the landing module
could be used to increase the payload of the lunar escape module.
Not only
sis
are instrumentation
role
in
process control.
automatic
instruments has vastly increased our ability to measure and thereby our
ability to find nature's answers to new questions.
Where science will take us in the future, no one knows. That is what
makes it such an exciting adventure. But one thing seems certain. If
social or political or ecological catastrophe can be avoided, science will
continue to probe with new and even more sensitive instruments
while the riddles of matter, of the origin of the universe, and of life are
being answered. Perhaps in time we may be able to construct a philosophy in total accord with all knowledge. Or perhaps, as is more likely,
we shall no longer feel the need for philosophy. For what is philosophy
but intellectual speculation turned into belief, and what place is there for
speculation except to develop premises to be tested?
7-2
Set into the stone wall of Saint Stephen's Cathedral in Vienna are two
One is about a yard long and the other
is about a meter long, but they are much older than either of these units
of measure.
In medieval times Vienna was the western terminus of
caravans that carried the trade from the East, and these bars were used to
iron bars with protruding ends.
The church
in those
latter.
Basic Principles
Alexandria at the moment the sun was directly overhead in Syene, and
knowing the north-and-south distance between these cities, computed the
circumference of the earth as 250,000 stadia, we can only admire his
genius, but we cannot check his result with certainty.
The measurement of angles is unique in that the unit is dimensionless
no standard is needed, but only a numerical convention. Perhaps this
accounts for the longevity of the Babylonian system of angle measurement, which is still in use today. In keeping with their sexaguesimal
(base 60) number system, the Babylonians divided the angle of an
The degree was then
equilateral triangle into 60 parts to get the degree.
divided into 60 tiny, or minute, divisions, and these in turn were divided
into 60 second-order minute divisions, today called simply minutes and
seconds.
circle into
24 parts, as we now divide the day, to obtain their basic units for then
these two traditional measures of time and angle, incompatible as their
subdivisions are with the decimal system, would at least be consistent
with each other and astronomers would not have to reckon with two
1
kinds of minutes and two kinds of seconds.
As experimental science developed in the eighteenth and nineteenth
centuries, the need for commonly accepted units of measure began to be
Without such standardization the intercomparison of results by
felt.
workers in different countries was much more difficult. After the
metric system was adopted in France in 1799, its measures of length and
mass were gradually accepted, along with the already established unit of
time, the second, as the units in which scientific findings in mechanics
were reported. Even though many laboratories used different metric
units such as the meter-gram-second (mgs) system, or the millimetermilligram-second system, the integral powers of 10 relating these units
made conversion
gram-second
relatively easy.
(cgs)
everywhere.
Electrical Units.
by
1
A new unit, the neugrad (or grad), equal to one four-hundredth of a circle, has been
introduced in Europe. It is subdivided decimally, but this is also possible with
degrees, and the virtue of dividing a right angle into 100 parts rather than 90 is any30
thing but obvious. Indeed it is more awkward to have common angles such as
expressed as 33M grad. If any change is to be made, let us choose the new unit to
be
15!
In both
q\
and
r =
92, i.e.
(1-2-D
er 2
VML
3
charge the dimensions of
/T 2 = M^L^T' 1 in the electrostatic unit
(esu) system.
Then workers in electromagnetics followed a similar route.
The unit magnetic-pole strength mi and m 2 was that which produced unit
mm
F =
1
fir
(1-2-2
2 M /i/ 2
for the force F per unit length I between two infinitely long parallel
conductors separated by a distance d. Again because m was considered
di mension less, current, defined
VML/T
= M*L*T-\
time, therefore
by Eq.
Charge,
(1-2-3),
(emu) system.
The dimensions of the esu unit of charge were thus L/T times the
dimensions of the emu unit of charge, and the same was true for other
units in the
is
two systems.
we
in
a given quantity
vqesu
Basic Principles
the
emu
units,
The dimensional disagreement between the esu and emu units for the
same quantities made it clear that e in Eq. (1-2-1) and n in Eqs. (1-2-2) and
(1-2-3)
if
not dimensionless,
why give them the value of unity for vacuum? This practice, while it
may have accelerated Maxwell's unification of electric and magnetic
theory, also led to much confusion by concealing the fundamental physical
between the electric field strength E and the electric displaceD and even more between the magnetic field strength H and
the magnetic induction B. As soon as it was realized that there was no
more reason for choosing e and m equal to unity than for choosing y equal
to unity in Newton's law of gravitation
difference
ment density
F = y ^Mi
2
(1 .
M)
(and thereby defining a unit of mass), the way was cleared for reconciling
the esu, emu, and practical systems. This was done by adopting the unit
of length as the meter, rather than the centimeter, and the unit of mass as
the kilogram rather than the gram, to give the present
mks
(meter-
kilogram-second) system.
In the process, one other defect of both the esu and emu systems has
been partially eliminated. Coulomb's law involves spherical geometry.
Each charge experiences a force from the spherical electric field of the
Since by definition
other charge.
F =
qE, the
field
must be assigned
with
1.
F =
-^
?-
(1-2-lr)
4irer
(1-2-3T)
2wd
and
cylindrical
geometry respectively.
Had this
the equation numbers designates the rational form.
been done originally and e and m assigned the value unity for a vacuum, the
The
r in
Heaviesu and emu units of charge would have been v47r times as great.
and Hendrik A. Lorentz proposed such a system and used it in their
works.
Modern practice is to write the equations in the above form,
side
10
use the esu, emu, or practical units, and subsume a compensating factor
of 47r into
eo
and
10 11
4tt
47r's
in the
mks
values of
10- 9
and
(1-2-5)'
36tt
10- 7
of 10
is
4ttc
Mo
This
m-
(1-2-6)
and permeability
As a
remaining powers
from the ratios of the esu and emu
Q 2 T2
coulomb 2
(newton) (meter)
ML*
and equal
to unity.
of light as dimensionless
of units in
which
all
variables in physical equations are pure numerics, and the unit size
is
the
set
identities.
Basic Principles
11
dimension.
It
should be understood that the size of a unit can be chosen independits dimensionality.
We are free to pick any convenient size for
ent of
it is
convenient,
i.e.,
the quantities
we need
to express
The standard
In
fact,
system resides not in the length of the meter, nor the mass of the kilogram,
but rather in the decimal subdivisions and multiples of these units.
The question arises: Is there a "natural" system of units? That is,
can we choose the size of our units of mass, length, time, and charge so as
to cause
many
rational
numbers
or
all
or, at
become
integers or
freedom or the
like?
At
No
Moc
4tt
The point
system of
number
Sir
e2
he
is
137.03602
that a
units.
is
dimensionless, so
its
is
a natural physical
and offered several hypotheses for why this might be true. More accurate
modern measurements have disproved his premise. Until future study
reveals the existence of a natural system of units, there appears to be no
reason to abandon the present mks system.
1-3
12
meter bar are prototype standards. So was the earth itself, since the
second was taken to be 3^6,400 P ar t of the mean solar day.
Prototype standards have several defects: They can change with time,
they can be damaged, and like the stadium and cubit, they can be lost.
Metals abrade and creep from internal stress. The earth slows down from
tidal friction.
The best standards appear to be atomic standards. So far
as we know, all atoms of a given isotope of a given element are absolutely
identical and invariant in their properties.
In recent years two of our
fundamental units, the second and the meter, have been defined in terms
of atomic standards rather than their original prototype standards.
This conversion to atomic standards has been made possible by the
development of instrumentation techniques, specifically the interferometer
and the cesium-beam clock, that enable the atomic properties to be compared with sufficient precision once and for all with the original prototype
standard.
From then on the original prototype need be preserved only
for historical and sentimental reasons.
So far, sufficiently accurate
techniques have not been devised to permit replacing the prototype kilogram with an atomic reference standard, but that day may not be far off.
The present definitions of the base units of the international system of
units (Systeme International d'Unites, abbreviated SI) as adopted at the
1967 (and earlier) general conferences (Conference Generale des Poids et
Mesures, abbreviated CGPM) of the International Committee on
Weights and Measures (Comite International des Poids et Mesures,
abbreviated CIPM) are given below. Only the kilogram involves a
prototype standard. The kelvin and candela, while not exactly atomic
standards, involve only the reproducible properties of matter.
These
two units are less basic than the others, but are included in the basic set
for convenience.
The thermodynamic scale of temperature could in
principle be defined by specifying Boltzmann's constant, while the candela
is a physical standard for a physiological quantity, since luminosity
s
human
eye.
(Eleventh
CGPM,
1960,
Kilogram
(kg).
The kilogram is the unit of mass; it is equal to the mass of
the international prototype of the kilogram.
(First and third CGPM,
Basic Principles
13
water.
Candela
(cd).
CGPM,
The candela
is
direction, of a surface at
of freezing
CGPM,
convenient shape and size can be computed, though even this is not
The discovery by Thompson and Lampard [1] of a class of
necessary.
may
be given
Volt (V).
[2]
The-volt
is
Ohm
tor
electrical units
of
A,
The ohm is the electric resistance between two points of a conducwhen a constant difference of potential of 1 V, applied between these
(0).
by a current
of
A.
Note that although we may have used the henry in determining the ampere, (a)
was not necessary and (6) the reasoning is not circular since the ampere was
not used to determine the inductance in Eq. (1-3-1). Instead we used a computable
capacitor and the fact that o>L = 1/&/7 for some co.
1
this
14
Farad (F). The farad is the capacitance of a capacitor between the plates
of which there appears a difference of potential of 1 V when it is charged
by a quantity of electricity equal to 1 C.
Henry (H). The henry is the inductance of a closed circuit in which an
emf of 1 V is produced when the electric current in the circuit varies
uniformly at a rate of
A/s.
Weber (Wb). The weber is the magnetic flux which, linking a circuit of
one turn, produces in it an emf of 1 V as the flux is reduced to zero at a
uniform rate in 1 sec.
Tesla (T).
The
tesla
is
a flux density of
Wb/m
2
.
abandonment
The
as voltage standards subject to occasional absolute calibration.
hierarchy of standards and the interlocking series of crosschecks by which
TABLE
1 -1
Recommended Values
of Physical Constants
[3]
Units
Symbol
Quantity
Velocity of light
Electron charge
c
e
mks
Value
2.9979250
1.6021917
4 803250
1.6021917
2.4179659
8.065465
1 160485
6.626196
4.135708
6.022169
1.660531
9.109558
5.485930
1 672614
1 00727661
10 8 m/sec
10- 19 C
Equivalent to
Equivalent to
Equivalent to
Planck's constant
h/e
(c)~ l (hc/2e)
Avogadro's number
amu
M*
Ratio of proton mass to
electron mass
p /m e
e/rn.
#
k
k/e
Gravitational constant
io- 19 J
10- 14 Hz
10 6
m-
10 4
10 10 cm/sec
IO" 20 emu
10" 10 esu
IO" 12 erg
IO 3
cm -1
IO" 34 J-sec
10-" J-sec/C
10 -27 erg-sec
10" kmole- 1
IO" 27 kg
10" 31 kg
IO 23 mole" 1
IO
-4
amu
10-" kg
IO -7 erg-sec /emu
10- g
10-28
g
-4 amu
IO
io- 24 g
amu
amu
1,836.109
1 7588028
5 272759
2.067854
1 380622
8.617087
6.6732
cgs
10 11
C/kg
IO 7
emu/g
10 17 esu/g
10~ 16
T-m 2
10-" J/K
10" 6
IO" 7 G-cm 2
IO- 16 erg/K
V/K
IO
-8
dyn-cm 2 /g 2
Basic Principles
15
For the convenience of the reader, Table 1-1, giving the most recently
determined values [3, 4] of some of the universal constants, is used to
conclude the present chapter. Reference 3 is particularly recommended
to the serious reader interested in examining the extremely meticulous
techniques required in the refinement of standards. An excellent bibliog-
in
both references.
CITED REFERENCES
1.
A New
Theorem
in Electrostatics
and
1956.
2.
3.
CHAPTER TWO
SINE-WAVE TESTING
OF LINEAR SYSTEMS
Bernard M. Oliver
Hewlett-Packard Company
Palto Alto, California
The most
Linear-circuit theory
engineer's training, as
is
is still
of sine-wave testing.
Later chapters fully treat the most important instruments and techniques used in testing linear systems, but we feel the need to set the stage
The present chapter defines and brings into focus those properties
first.
of linear systems, the understanding of which is essential in the sound
This chapter also gives a mathematical picture
practice of measurement.
of the basic sine-wave measurement procedures or methods.
Only the fundamental aspects of transmission measurements will be
discussed here.
keep in
16
full
wave
testing.
2-1
Mathematical Background
Of
their shape
is
17
unaffected
by
In response to a
a careless engineer.
A
that
system
is,
to the
be linear
will
if it
is
composed
first
power
of
any
is
directly proportional
where
a>
=
is
i,-r'
the angular frequency of the forcing function (the applied
all nth derivatives of the amplitudes have the form
= (to)^^' =
(iu>)
ai(t)
(2-1-1)
solution.
18
pulse-code
b are constants.
its
response to
g(t)
f(t) is
J~ m
f(r)k(t
We
f' m
(2-1-2)
T ) dr
[f" m
f(r)k(t
t)
dr] dt
(2-1-3)
M=
/_".
/to [/_".
w-
variables, letting
f' m
/(r)e-*
[f' m
~iat
T^ e
(t
dt
t)
dT
(2 " 1_4)
u, to obtain
(2-1-5)
in brackets, denoted below by K(u>), is the Fourier transform of k(u). Since K(u) is not a function of t, we may remove it from
under the first integral, which then becomes F(u), the Fourier transform
The quantity
of f(t).
G()
Thus
K(a)F(u)
(2-1-6)
and we see that for any system having the two properties given above, the
spectrum of the output is the product of the spectrum of the input and a
Sine-wave Testing
of Linear
Systems
19
system filtering function K(u), the latter being the Fourier transform of
the impulse response.
The output time function g(t) is the inverse
transform of G(u).
Finally
we note
that
F(u>)
8(ca
and G(u)
K{u>) 8 (a
wo)
),
which
is
K(ioo).
oidal
if f(t) is
component produces
its
own
sinusoidal response
Thus
ae-^'oFio))
\F(a>)\
One
simple to see
if
distortion (that
is that it is so
Often, as in audio circuits, phase
phase shift not proportional to frequency) is relatively
unimportant.
2-2
Often the range of amplitude encountered in measuring the transmission of a device is extremely large.
Further, when devices are operated in tandem, the overall amplification or attenuation is the
product of
the amplifications or attenuations of the individual units.
For these
reasons it is convenient to express amplification and attenuation in
logarithmic units, which are of convenient size and can be added rather
than
multiplied.
At one time, and particularly in Europe, the neper, which is
the natural logarithm of the ratio of output power to input power,
was
20
used.
Today
The gain
10 log
commonly used
(2-2-1)
input power
loss as
L = -G =
The
is
in decibels is defined as
output power
G =
and the
the decibel 1
input power
10 log
method
classical
(2-2-2)
output power
of
measuring gain or
loss is
shown
in Fig. 2-1.
must be added
nience of having
connections at a
all
common impedance
The conve-
level is evident.
The gain
G =
of the device
is
then
(2-2-3)
012
where on and a 2 are the settings of the two attenuators plus their assoThis procedure is repeated at each new frequency.
ciated pad losses.
The readings obtained do not depend upon knowledge of the absolute
The law of the detector is
level of the test signal or of the output signal.
1
One-tenth of a
bel,
Graham
Device
Attenuator
under
[a
Oscillotor
Detector
8
f
Attenuator
7
FIG
2-1
Classical substitution
method
for
measuring gain or
loss.
Bell.
The accuracy
of the attenuators,
of
21
measurement
in the source
and
nonlinearities in the
FIG 2-2
loss
Voltoge
current
Device
Signal
generator
under
power
meter
or
test
Direct
measurement.
method
for sain
and
is
22
Device
under
Detector
and log
test
amplifier
X- Y
recorder
a
X
or
E--
log
Log
*
FIG 2-3
amplifier
much
less
than,
its
if a frequency range of a
decade or more is to be covered, an exponential sweep is commonly
chosen. If a wide range of output amplitudes is expected or if a direct
reading scale in decibels is required, the output signal is passed through
a logarithmic amplifier.
The output of this amplifier is used to drive the
y axis of an xy recorder, while the x axis is driven by a signal proportional
Some sweepers supply this
to the frequency or the logarithm thereof.
x-axis output.
With others a frequency meter must be used to measure
See
the frequency being generated and to provide the z-axis signal.
of frequency with
Chaps.
13, 15,
time
and 16
is
for elaboration.
rich in detail.
Measurements are
made over
readily measured.
Sine-wave Testing
of Linear
Systems
23
at cardinal frequencies.
2-3
Device
under
test
Signal
generator
Load
'
'
Amplifier
FIG 2-4
Phase measurement.
'
<
Amplifier
24
X
X
=
=
x
y
If
<f>
smcot
sin (wt
0,
to apply the
gains (and,
0)
same
if
good idea
and adjust the
It is a
<f>,
<(>
<f>,
>
minor axis
will
it is
\(f>\
at
90, the
45 and +45
are respectively
u =
V2 sin *
(2-3-2)
V2 cos |
(2-3-3)
Thus scales like those shown in Fig. 2-5 can be constructed to read the
minor axis. The ellipse always crosses these lines at right angles, which
removes one source of reading error, and the minor axis is a sensitive
function of
<
for all
</>.
When
the phase-shifted signal is then applied to the y axis, its positivegive the phase shift in units of 20 per major
Obviously care must be taken to position the signals vertically
division.
going
Sine-wave Testing
of Linear
Systems
25
magnitude.
ring
FIG 2-5
Elliptical
measurement.
cathode-ray
oscilloscope
display
for
phase
26
B >RL
'*i
J^L
FIG 2-6
Phase-detector
where
I,( e vi*T
V =
circuit.
is
(2 _ 3 _ 4)
1}
= charge on an electron
= Boltzmann's constant
T = temperature, K
/, = saturation current when V
and if R = 0, then with small signals
q
k
negative
is
applied,
it is
a simple matter to
R L = 0, the voltages applied to the diodes are specified by the input signals.
If the
reference input is d = Ex sin coif and if the signal input is
solve for the short-circuit output current
e2
Elicit
0)
H = 2EiE
since with
jjM
will
I, {cos [(,
be
ui)t
<f>]
cos [( 2
ui)t
<]}
(2-3-5)
The
allel
and
is
1/Sz'i
lifer
(2 - 3 - 6)
^-iUL-o; =iw.
giving an open-circuit output voltage
2
eo
= -y-
{cos [(
Wl )<
2?i
</>]- cos
and
[(oj 2
+ 0 +
0]}
(2-3-7)
Sine-wave Testing
of Linear
Systems
27
may
iff 2 q
e
kT
cos
(2-3-8)
<}>
and we
and the low output signal levels produced make the unballasted (R = 0)
diode ring modulator a poor choice for phase detection, although in many
applications it makes an excellent mixer, demodulator, or single-sideband
detector.
The diode ring modulator can also be used at high signal levels by
incorporating the ballasting resistors R, whose resistance is much larger
than the forward resistance of the diodes. If the reference carrier is now
made very large compared with the other input or made a square wave
whose amplitude is only slightly larger than the peak signal amplitude,
For example,
the modulator acts like a synchronous reversing switch.
upper
diodes are
and
the
conduct
diodes
lower
the
is
positive,
when ei
back biased. When e x is negative, the reverse is true. Thus, the output
is taken alternately from one-half or the other of the signal transformer
secondary, and the instantaneous output open-circuit voltage is as shown
in Fig. 2-7.
FIG 2-7
= - E2
How
reversing switch.
is
cos
is
very nearly
[(o) 2
the circuit
<*>i)t
in Fig.
</>]
(2-3-9)
28
where we assume
|2
o>i|.
The
| 2
i|
wi and neglect
internal impedance
is
R/2.
With
this
mode
of operation
is
<j>.
two
signals.
The
Phase detectors are commonly used as null detectors to tell when the
two waves have, say, exactly 90 of phase shift. The exact reading of the
true phase is then taken from the scale of a calibrated phase shifter in the
reference (or signal) channel required to establish the 90 condition.
If
from the phase detector can be used to drive the phase
Many
Device
generator
under
29
Load
test
Sampler
Sompler
Ui
Vector
voltmeter
Reference
l_
Sampling pulses
~~f"f
FIG 2-8
by
a constant frequency / as
this
shown
in Fig. 2-8,
of frequency f
and usually f
/.
In
and relative phase are the same as the original waves of frequency/. The
samplers can be regarded in fact as very broadband, unity-gain mixers.
The offset sampling frequency can be obtained by using the output of a
2-4
30
for rejection.
Not only does computer control greatly increase the speed of measurement; it can increase the accuracy as well, as discussed in Chaps. 17
and 18.
Measurement of Delay
2-5
Distortion
of fit)
is
g{t)
~ /.", G()e*'
dco
/""'
~
2ir J
F(co)e^-^
dco
f(t
r)
(2-5-1)
by
t.
<t>io))
and
<(0)
Delay
=
=
0.
<f>i
is
o>)
Also
d<f>/dco
<'(w)
<*>'(
co)
is
an even function.
Sine-wave Testing
the function
<j>(w)
of Linear
Systems
31
series
</>(a>)
</>(
Now
<(a>o)
w) =
0'(cuo)(w
4>(o>o)
a> )
<f>'(ooo)(co
wo)
(2-5-2)
'
/(0
25 sin
cos
- 5<a;<wo +
coo
(2-5-3)
V
'
ajoi
St
<
< wo +
and
or
if
we
25 sin
8(t
xf4
<f>')
_l.
cos &<*
Upon
taking
(2-5-4)
*(o)]
let
r --d<j>
(2-5-5)
25 sin
8(t
8(t
<f>(coo)
t)
C S
t)
~^ +
(2-5-6)
where
<t>o
and
w ==
is
oj
<f>(ooo)
+wo
Ow
amount
shown
d<l>/da}
jo>
= wo
in Fig. 2-9.
<(co)
The function
(2-5-7)
! = <"
at
co
p() is f(t)
of the cosine
intercepts the
delayed by an
term shifted bv
^ an
amount <
In particular, the envelope term
simply delayed by a time t, and for this reason
additional
d0()
is
sin St/ St is
32
FIG 2-9
Unless t(w)
is
shift versus
w.
For a "short" transmission system where both the input and output are
same locale, the phase characteristic can be measured and
the variations in its slopes can be computed to determine the delay
distortion.
For "long" circuits, such as transcontinental telephone or
television circuits,^, direct method is needed for measuring the variation
available in the
pair of circuits.
Oscillator
0)2
Bol.
mod.
2
>
\
^^
Amplifier
Oscillator
Aw
2
Test
FIG 2-10
'
Oscillator
O),
Bol.
'
mod.
1
w,
Aw
~2~
Method
(or
33
= Ex
eA
= Ea
cos
e<i
= E2
cos
(o)2t
+ ai)
cos (wi<
Aco
(2-6-9)
+ a2)
kEiE&
cos
= E
(o>ii
cos
Aco
ai) cos
sine
1 is
<
waves
J +
wi
ai\
and
(2-5-10)
eib
= E
cos
wi
+ -^J + aj
t
e ia
626
= E
cos
= E
W2
LV
2/)
ACO\
COS
cos 2(wi<
by
+ a\)
+a
(2-5-11)
+ J +
,
(c02
wave contains
CK2
sum frequency
The sum frequency can be removed
+ at).
filters.
If
phase characteristic
cos (Acot
cos (Acot
+ A0
A<i)
and
(2-5-12)
2)
where
L<tn
<(<".
+ Aco) -
Atf.^Aco-^
<t>(wi
Aw)
-Acor(co.)
1,
(2-5-13)
(2-5-14)
34
Thus the
receiving end,
At
At the
T2
Tl
receiver,
A<f>i
= -*1_
Ad>2
r_
(2-5-15)
Aw
to the
the measurement
The value
is
basically a comparison of
two
differential
phase
shifts.
/w;+A/2 dd>
fui + Aw/2
= ~
-W2 ^*
L-^2
is
'<> d
(2- 5 " 16 )
of
width Aw.
It is apparent that one difference frequency merely serves as a time
reference against which to compare the phase of the other difference
frequency.
No absolute measure of total delay is obtained or desired.
If one pair of frequencies is swept rapidly and periodically across the
cos
[Aw+
sin
<(&/)]
[<f>[&t)-f]
Phase
detector
i
,
i
VTO
FIG
'
LPF
2-11
Another arrangement
group-delay measurement.
for
d.(fl]
(2-5-17)
where w
Sine-wave Testing
the average value of
sum
is
frequencies)
kept small,
is
is
<f>(ut).
now
The output
of the
form sin
is
35
Systems
of Linear
and
4>]
test circuit,
if
the argument
In this
way
only
obtained.
If
modulation frequency
is
observed as above.
2-6
is
is
Knowledge
We
shall only
36
cies
(Nyquist diagram) and also the loop gain and the loop phase as
is
(2-6-1)
-M/3
where
/*
/x/8
and
=
=
since
the curve of
/x/8|
is
+ jO
to
in the
Sine-wave Testing
of Linear
Systems
37
meter (or a network analyzer) is extremely convenient for this measurement. However, if the loop contains no nonminimal phase structures,
only the magnitudes of E x and E 2 need be measured, since in principle
the phase can be inferred from the loss-phase relationships [3].
t
^1
HZ
FIG 2-12
Elementary method
measurement of loop sain.
When
X \zL
z2
for the
is
if
(as
is
sufficient accuracy.
In systems with high loop gain at dc, the loop gain at moderate and
high frequencies can usually be measured by opening the loop at these
frequencies, but leaving it closed at dc and very low frequencies.
As
shown in Fig. 2-13, this can be done by inserting an additional RC lowpass filter in the feedback path. If the test signal source has negligible
impedance and if c
R/\l m| ver the frequency range measured,
then the filter provides a new stable gain crossover at low frequencies
above which the loop is effectively open. Actually, it is not necessary
that the loop transmission be reduced to less than unity over the fre-
38
\z'-
RZy
R-Z\
frequencies.
filter in
place be stable.
Then the
is
be omitted.
Another, and often a very convenient method of measuring loop gain
is to leave the loop gain unaffected at any frequency and simply inject
a test-signal voltage in series with the loop, as shown in Fig. 2-14. In
order not to disturb the normal loop gain, the source impedance of the
Z 2 \, and the shunt
generator should be small compared with \Z\
1/Z 2
admittance to ground should be small compared with \l/Zi
This is easily accomplished by obtaining the test-signal voltage E from
an isolated secondary winding (often a single turn is enough) of a step-
or signal generator.
will differ
from that
is
I2
= E2/Z1 =
of
is
|.
x,
because the currents taken from the outIn Fig. 2-12 the current supplied to the termination
In Fig. 2-14 the current supplied to the
n(3Ei/Zi.
in Fig. 2-12
I2--
Z2
<3D
(3
FIG 2-14
-"
Loop-gain
measurement by
actual circuit
we have
E%
is
/','
E1/Z1.
nfiEi
-f-
(^2
ju)8i
(m/3
12
-l)
)Zi
Er
2
tf,
Sine-wave Testing
of Linear
Systems
39
and therefore
l*P
E2/E1
1
Z2/Z1
(2-6-2)
+ Zt/Zx
Very often an
10 -3
m/3
it
can be shown
[5]
that
- h/Ii
= Zx/Z*
1
(2-6-3)
+ Zi/Zt
Aside from the sign difference in the numerator, which arises from
choosing the positive directions of Ii and I 2 to be opposite, Eq. (2-6-3)
Z 2 for then
is the dual of (2-6-2) and is especially useful when Z x
-/2//1.
In making loop-gain measurements, it is prudent to scan the frequency
region for at least two decades beyond gain crossover to make sure that
Itfi
known
z,
FIG
2-1 5
Injection of a
measure loop gain.
z,
test current to
40
2-7
is
nonlinearity.
Common
ferroelectric devices.
has a dead zone or a highly nonlinear region near the origin and
fairly linear at higher levels, as is true for an overbiased class-B
amplifier or a mechanical system with static friction (3) those for which
the nonlinearity is "fine grained" and present at all levels, as in a quantized pulse code modulation system.
Most analyses of distortion
treat only the first class and yet the second and third classes are often
encountered.
teristic
becomes
oiit
a 2 cos
a^
of the
(2-7-1)
all
frequencies
form
pan
ga> 2
(2-7-2)
series
will
41
power
may
Common
1.
relative to the
fundamental
(single-frequency input)
2. Amplitudes of individual intermodulation products relative to the
amplitudes of the inputs (often of specified ratio) producing them
3. The total rms distortion, or the ratio of the power "scattered"
into other frequencies to the power at the desired frequency or frequencies
4. The power scattered into a particular frequency band by signals in
other bands
Which of these
upon the
entirely
(or of
many
application.
is "best" depends
In sound reproduction, any harmonics or
42
not necessary to use sinusoidal test signals to exploit this propNoise passed through a band-rejection filter is often used to
measure intermodulation in multichannel communication circuits. At
the receiving end, the power in the frequency band rejected by the input
filter is measured.
This method has the advantage that the test signal
ever,
it is
erty.
simulates the statistical properties of the actual signals quite well, and
this gives a more meaningful measure of distortion in many cases.
Chapter 14 discusses the application of this
and
2-8
approach
in testing transmitters
receivers.
desired signal.
Alternatively, appropriate
Sine-wave Testing
filters in
of Linear
Systems
43
Many
much
greater
or even a
CITED REFERENCES
1.
Graham, R.
E.: Linear Servo Theory, Bell System Tech. J., vol. 25, no. 4, October,
1946.
2.
3.
4.
5.
6.
Measurement, IRE,
vol. 36,
January-June, 1948.
CHAPTER THREE
SQUARE-WAVE AND
PULSE TESTING OF
LINEAR SYSTEMS
Bernard M. Oliver
Hewlett-Packard Company
Palo Alto, California
Although sine-wave measurements can completely and accurately characterize a linear system, a single measurement at one frequency will not
do so. Rather, an array of steady-state measurements or a sweptfrequency measurement must be made. And although these measurements may tell all about a device, they may tell it in a roundabout way.
For example, what is often important about a linear system is its response
to transient signals.
While the transient response may be inferred from
the steady-state measurements, or may be computed by using Fourier
transform techniques, a much more direct approach is to apply a kind
of standardized transient and observe the response.
The "standard transients" most often used are step functions, square waves, and impulses.
Square-wave testing first became widely used in the testing of video
amplifiers and other circuits designed to handle television waveforms.
44
of Linear
Systems
45
common
contain
all
is
in
(which
may
be regarded as a series
and negative-going steps) does not have the continuous smooth spectrum of the isolated event, but rather, being periodic
in nature, contains only frequencies that are harmonics of the repetition
rate.
Such a test signal samples the response of the device at these
frequencies only. Any anomaly lying completely between zero frequency
and the fundamental or between any two harmonics will not be observed.
Just as care must be taken to spot the test frequencies closely enough in
a steady-state measurement, so one must use a low enough repetition
rate in pulse testing if fine detail in the frequency domain is to be properly
of alternate positive-
rate.
The duration
by the overload
of
short compared with the period of the highest frequencies passed by the
system. The wider the frequency response of the system, the smaller
the area of impulses that are suitable test signals.
The ideal zero-rise-
46
3-1
Tools
and Techniques
special instrumentation
for repeated
is
than by using sinusoidal measurements. Further, in some devices, different elements affect distinctly different parts of the transient so that
identification of a defective or misadjusted part is obvious from inspection
This is especially true in testing systems where the delay
of the response.
is large compared with the rise time, as in time-domain reflectometry
(see below).
for
system saturation
may in
When
of Linear Systei
47
whereas
may
3-2
and
(see
Chap.
2).
Sinusoidal Responses
= A
2^ J-"
unit step
may
K^
e<ut
d"
(3-2-1)
of a unit-area
impulse,
(0
=/l
(X)d\
(3_ 2 _2)
co
unit step.
transform of K(u)/iu, or
K(u)
?lwt
*rJ--^r e
2W-
a
du
"
(3-2-3)
48
Unit
System
Step
Unit
impulse
Impulse
Ways
Scope
test
Step
Impulse
response
System
response
Scope
Integrator
under
impulse
generator
3-1
Step
response
under
Integrator
Unit
FIG
System
Step
Impulse
generator
(c)
Scope
test
generator
(b)
Step
response
under
step
(a)
test
Now
filter.
directly (a), or
by
an impulse-test
step response.
(a)
Impulse
response
Impulse
Unit
impulse
Scope
generator
Unit
step
(b)
Impulse
Step
Differentiator
(c)
FIG 3-2
Step
System
under
Impulse
response
under
Scope
test
generator
Unit
step
generator
System
Impulse
response
Step
response
Differentiator
test
Scope
of Linear
Systems
49
Table 3-1 shows the step and impulse responses produced by a variety
commonly encountered frequency response characteristics. The
frequency responses are shown graphically with either linear or logaof
rithmic (decibel versus log frequency) scales as best suits the nature of
the function.
Familiarity with these function transform pairs facilitates
3-3
Time domain
Input time
function
f(t)
Frequency domain
Direct
Fourier
transform
CP-fe
Convolution
with impulse
response of
system: k(t)
(scanning)
3F
FIG 3-3
Spectrum
of input
FfuJ
T
Multiplication
by frequency
characteristic
Output time
function
Inverse
g(t)
transform
Fourier
to an arbitrary input.
=>
of system./fjW
(filtering)
Spectrum
of output
GU=F(uijK(uy)
50
TABLE
Step response
System
freq. characteristic
wo
CASE
--%
Typical network
Cg)
1
This is the typical simple low-frequency cutoff such as might be produced by a
condenser-shunt resistor combination. The step response shows an abrupt rise to
unity followed by an exponential decay.
Usually encountered in amplifier interstages
.
series
response
is
2.
may be
In interstages,
as high as several
is
megacycles
in
Step response
CASE
/<>
System
freq. characteristic
Typical network
Step response
System
freq. characteristic
Typical network
}*4
CASE
3.
deficient.
2.
Here
it is
+s
I
C,R,<CZ RZ
is
3-1
of Linear
Systems
51
Step response
System
freq. characteristic
Typical network
CASE
"peaked")
interstages,
Step response
System
freq. characteristic
A2
resistive source.
Typical network
| 12dB/0ct
%
(log)
AU^-f
iv
*"^W1t
/?<
TC
R<
-TC
(log)
RC
Step response
System
freq. characteristic
Typical network
6dB/0ct
A-i
(log)
(log)
CASE
6.
Phase-compensated low
eventually, but initial slope is zero.
May be produced in
Often found in video amplifiers.
tilt.
end
cutoff.
As a
result
52
System
Step response
Typical network
freq. characteristic
Two simple low frequency cutoffs (case 1 ) in tandem. Typical low frequency
response of single-stage resistance-coupled amplifier with input blocking
Principal differences
capacitor or two-stage amplifier with no input blocking capacitor.
compared with case 1 (1 ) faster initial rate of fall for same w , (2) response goes negative,
CASE
7.
transient
crossing axis at
/o
With each additional low-end cutoff one additional axis crossing is produced. Thus,
n 1 axis
if the low end response falls off (ultimately) at 6ndb-/octave, there will be
They do not occur at regular intervals each successive half cycle takes
crossings.
longer.
Step response
System
Typical network
freq. characteristic
44^^')
CASE 8.
Simple high- and low-frequency cutoff. The step response rises exponentially
determined by high frequency cutoff, then falls exponentially at a rate determined
by low frequency cutoff. Typical complete resistance-coupled interstage response. If
1/ona fast time scale
"i/mi,
1 , then on a slow time scale response looks like case
If o>j = i = "o, we have the case of a critically damped
response looks like case 4.
RLC circuit. The response then becomes wot e-w '.
at a rate
Step response
System
/0--Ju*-a?
j2-e- at
Typical network
freq. characteristic
/?VF
sm&t
(log)
CASE
that the
ft point.
/ in Q/ir cycles.
Thus
Q =
where n
is
Step response
System
freq. characteristic
Typical network
Ti
For
CASE
step
hi
0.
due
to
C, 8=-|-
Small resonance
flat
in
an otherwise
transmission plus
damped
flat
characteristic.
oscillation
of oscillation
figure.
Step response
System
Typical network
freq. characteristic
"P
Forhi0,8=--
V
+
Wo
CASE
1 1
complete
used
in
Similar to case 1
(A
T
is
a resonant dip.
1 ) is
Note
is
Step response
System
freq. characteristic
>
1
it
persists for a
Typical network
z_
Z+Z
f^o
z+z
z+z
AA-1
c!0
CASE
12.
ripple
in
(Linear)
Positive echo.
Associated frequency characteristic has nearly sinusoidal
amplitude and phase. Frequency interval between successive maxima or
minima is reciprocal of echo delay. The longer the delay, the closer the ripples. Commonly encountered in systems having faulty or misterminated delay lines. Also in
measurements where multipath transmissions can exist such as acoustic measurements.
53
"
54
TABLE
System
Step response
3-1
Typical network
freq. characteristic
Z+Z
Z+Z
'"
"""
h-H
-?0
|z<z
CASE
System
Step response
1
1
Negative echo.
1 3.
(Lineor)
in
Typical network
freq. characteristic
~VN
7V~7 \
r
J-JWV-l-
Zo
Z=0
8
1
it
CASE
14.
Minima of frequency
echo.
Can be considered
ripples have
now become
to
nulls.
be a 100% negative
Shape of amplitude
characteristic
end shorted.
Step response
System
Typical network
freq. characteristic
J,
p-z
k'-H
(Linear)
CASE 1 5.
line
is
of Linear
Systems
(Continued)
Step response
1-
System
freq. characteristic
Typical network
~A$*
Matched
constant-
I
'o
CASE
A'
filter
Zt
16.
disappears.
Step response
k*Jr,
1-
System
freq. characteristic
Typical network
i-
Ideal
low-pass
filter
(phase equalized)
t-
f*
CASE 1 7. The "ideal" low pass filter passes all frequencies below / with the same
amplitude and delay while attenuating completely those above /o. Its step response is the
sine integral.
This function differs from zero (except at discrete points) for all t > oo
Hence the ideal filter cannot be realized without infinite delay.
practical approximation will have a finite delay and its step response therefore will execute only a finite number
of wiggles before the main rise.
Here again, the ripples in the step response do not
indicate high frequency enhancement, but are the "Gibb's effect" encountered in Fourier
series, and are properly called band elimination ripples.
Step response
+1-,
2
System
freq. characteristic
Typical network
1-
Ideal
hi
Z~
2
-pass
filter
(
phase equalized
fo
CASE 18. The ideal high pass filter. By superposition the response of this filter is
obtained by subtracting the response of the ideal low pass filter from an equally delayed
unit step.
55
56
and phases
of the
components
9 = pj(r)k(t-r)dr
(In Sec. 5-1, convolution
is
is
(3-3-1)
solved digitally.)
When two
(t
> t) and
displaced
reversed in time
how the
integral
A;(t)
is
integral
The
is then found as a function of the displacement.
shown graphically in Fig. 3-4, where jit) is shown in (a) as a
rectangular pulse and kit) is shown in (6) as a decaying exponential.
In (c), k(t t) is shown with /(t) scanned as t varies. Since /(t) is a
constant for
< t < t and is zero otherwise, the integral of the product
The output wave
to the shaded area in (c).
is
proportional
in this case
is shown in (d).
of their product
process
is
f(t)
(a)
ta
(b)
k(t-T)
(O
-+
g(t)~.
(d)
O
FIG 3-4
to
functions.
of Linear
Systems
57
Convolution may thus be thought of as a genIn the frequency domain a network filters
In the time domain
the input spectrum to produce the output spectrum.
the same network scans the input wave with the filter impulse response
Multiplication in one domain corresponds to convoreversed in time.
lution in the other, and in this sense convolution is as fundamental a
mathematical process as multiplication and should be just as well underslit in
3-4
Effect of
Low-end
Cutoffs on
Square-ware Response
it is
little
value to use a square wave of such low frequency that the response to
top half bright sky and the bottom half dark ground. If the frequency
of the test square wave is so high that little distortion is produced, the
Nevertransients from the successive steps will be highly overlapping.
frequency response is easy to infer because only the fundamental and lowest harmonics of the square wave are altered in amplitude
theless, the
or shifted in phase.
K(p)
(3-4-1)
<>
<
Now let us see if we can infer the shape of the output wave from the
frequency response characteristic well above the low-frequency cutoff.
58
2-a
T/2
(a) Input
FIG 3-5
(b) Output
cutoff.
> b n sin
2wn
as
bn
n odd
=
The amplitude and phase
\K\
(3-4-2)
7T?i
of
n even
K(a) are
VI +
ox,
/" 2
a;
(3-4-3)
and
tan -1
.
COo
Wo
(3-4-4)
CO
=2
\Kn \b n
sin 1 2irn -
co
/co
1.
is
(3-4-5)
/.\
g P {t)
= V
)
bn
/
[
cos e sin
27tm
+
t
2irn
sin 6 cos
2irn
'
2wn
(3-4-6)
fit)
S p (t)
(3-4-7)
Square-wave and
Systems
59
where
SM) =
The
>
\2x/ cos -
r
first
sin
\2im/
Trn
is
triangular
^-^HOI-t
which
is
9a(t)
= f(t) +
first
(3-4-8)
'
term of Eq.
(3-4-6).
We
therefore have
(3-4-9)
8 a (t)
where
y _
1 (^y sin
t-i
irn
Sa(t)
which
is
2irn
\2irnJ
the curvature of
(3-4-10)
term
of
8P
Thus
8a
merely doubles
8 P.
The
shift of the
-Input
Output
Baft J
FIG 3-6
pole.
Component
distortion of square
wave produced by
single
60
(e)
RC
FIG 3-7
cutoff
(a + c)
The value of the preceding analysis lies in the ease with which more
complicated low-frequency cutoffs may be analyzed and equalized.
Suppose the low-end characteristic consists of n poles and m zeros. Then
to the same degree of approximation that we have been using,
\K\
'-=*
r^-
(3-4-11)
"at
tan"
3=1
X=
"Pi
- 2
(3-4-12)
t=l
where u pj is the frequency of the jth. pole and o) zi is the frequency of the
zero, and so long as the fundamental of the square wave is well above
the frequency of the highest-frequency pole or zero, the total phase and
amplitude distortion may be predicted just as simply as for a single pole.
Figure 3-7 shows qualitatively the types of square-wave distortions
produced by various frequency response departures.
If both poles and zeros are present and if
ith.
j=i
t=i
(3-4-13)
of Linear
-1
then the linear term in the expansion of tan
6 disappears,
61
which leaves
Systems
y-i
higher-order terms
to phase shift
is
now
a>
Time-domain Reflectometry
3-5
is
when many
In
Fast
oscilloscope
Highimpedance
Sync
probe
T- connector-^
-*-*o
Fast
^o>
pulse or step
\f5
System
*>,
under
test
generator
-<
FIG 3-8
<
62
with the start of the oscilloscope sweep. If the group velocity of the
cable is v, then at a time t = U/v later, the incident pulse or step is picked
up by the sampling probe of the oscilloscope. The pulse continues on
to the test system, and at a time no less than 21/ v sec later, the first
reflection appears.
and
if
on the oscilloscope
of generator
match there
is
2l /v sec later.
Regardless
reflections to
Mfl in shunt with a 50-ft line. This resolution and sencombine to make time-domain reflectometry a powerful measure-
in series or y
sitivity
ment technique.
In addition to locating reflections, the time-domain reflectometer
display gives a good deal of information about their nature and probable
causes.
of reflection
frequency domain)
in
The voltage
is
~ Z
= Z
= Y o- Y
Z +Z
Y +Y
where Z
and load
(or
and
(3 " 5_1)
impedance
respectively.
Any
discontinuity
may
while
AZ
_
_ AZ
~ 2Z + AZ ~ 2Z~
if
AY
P
The
(3_5-2)
2F + AY
is
AY, we have Y =
AY + Y
and
AY
2Y
(3 ~ 5_3)
Square-wave and Pulse Testing
Termination
>
~ZF
Systems
63
Waveform
Resistance
(a)
of Linear
z/ -
K
r *
fi>Z
Ft-Zr,
R<Z
Jf
J
/?--0
Capacitance
(h)
Inductance
reactive terminations.
(c)
>
test signal
u{t)
is
r-
(3-5-4)
ZTT J
is
(3-5-5)
Id)
Rather than include the added phase factor e~ iuT in the reflection coefwe choose simply to remember that the reflection as calculated
ficient,
must be delayed with respect to the incident impulse or step by the roundtrip delay r
2l/v.
By
shown in Figs. 3-9 and 3-10. The figures assume a zero-rise-time step
generator and oscilloscope. The actual display will be the result of convolving the responses shown with the derivative of the test step signal
as seen in the oscilloscope.
d.
Discontinuity
Waveform
(a)
Shunt capacitance
Z -~ if
~-z
'Y
(ill
* Z
Shunt conductance
(b)
-r
Waveform
Discontinuity
Series resistonce
C'T
\-e
j]
z c
~l>-~
Series capacitance
fr)
Gi
^07
ll
7
Z.Q*
Jj
-f
z-izz^r
Shunt inductance
(c)
FIG 3-10
Z^
Typical reflections produced
(f)
~-z
by simple
l)
discontinuities.
1\c
2t
'
64
If
coefficient is p
echo appears
h(t)
This
is
(R
)/(R
+Z
P (0
2R
(3-5-6)
Fig. 3-11.
If
is
now
ohmmeter and
is
it is
shown
in
2/?
/?+z
FIG
3-1
2-V
same relation
between terminal voltage and resistance
by
unit step.
of variable impedance,
is
tometry do not cause much trouble. If one eliminates the first reflection
first, then the next, and so on, multiple reflections can be ignored since
they always occur later than the echo of interest. As each strong reflection
eliminated,
end a
clear,
CHAPTER FOUR
in
reviewed below.
Noise is present in every electrical circuit except at a temperature of
It is also present in all measuring instruments, since
absolute zero.
they are merely specialized electrical .circuits. From the measurement
point of view, one is interested in noise in three basic ways:
1.
2.
65
66
3.
noise
4-1
Mathematical Background
number
Random
a finite
signals,
number
of
tell
us
a waveform
instance, the
list
monly used
statistics of
random
and
sequences.
Before
measurement procedures
will be described.
Ensemble.
Let us start by considering the continuous function of time
represented by the waveform ^(0, Fig- 4-1. This could for
instance be
the variation in power-line voltage, with respect to
the specified value,
in a particular building in a city.
Other waveforms x 2 (t), x z (t), and so
forth, could represent line voltage recordings
taken at other points in the
Such a family
city.
Ml(il)
w k2
=
Xk{ h)
-
waveforms
it is
is
called
at
some
an ensemble.
instant in time
is
written
(4-1-1)
The value of the ensemble average will, in general, depend upon the
instant of measurement.
This can be visualized in the case of an
Measurements
of, with,
and
in
67
x(t)
X,{t)
FIG 4-1
Ensemble of random
ensemble of
variables.
when
The separate
synchronized to an
artificial
time zero.
by
It
is
now
and
possible to take
The
Hx(t 2)
for
is
invariant
(4-1-2)
any t h h
enclosure would be
members
is
experiments.
68
period of time.
We
average.
n,(k)
lim
x*(0 dt
JQ
(4-1-3)
This average gives us specific information about the fcth recording, but
we cannot deduce anything about the time averages of other similar
recordings except in certain special cases.
For instance, in our previous
example of an ensemble of line-voltage records taken at different points
in a city, the time averages at different points are likely to be quite
of
identical
in
resistors
the
environment.
first
be stationary before
it
is
ergodic.
N
1
77
xm (tj)
lim
xk {t) dt
(4-1-4)
or in words,
Ensemble average
for
fix (k)
tj
at time
or for
for
t,
any recording xk or
any ,/,
kth.
record
in symbols,
(4-1-5)
A;
from a
when
is
one recording
single experiment.
Many
and
Measurements
of, with,
and
in
69
Mean-square Values. We have seen that there are two ways to measure
average amplitudes from an ensemble of random data recordings. The
ensemble averages and time averages will in general be different unless
is ergodic.
Similarly, the mean-square value can also be
measured in two ways, across the ensemble or as a time average.
The mean-square value taken across the ensemble at time t} can be
the system
written as
Wit,)
t=
xk *<ti)
(4-1-6)
W(k) =
is
lim
^J
xk *(t)dt
(4-1-7)
the dc component or
component
waveform.
of a
ac)
by the symbol a 2
^
The
<r
alone, given
(4-1-8)
same
is
for
line of
The
a2
total
is
in watts,
but
it is
common
For
voltage
is
mean-square
This inconsistency can be
70
components
number
of
increases.
A random
period.
infinite
number
Power,
V2
f 2Tq
FIG 4-2
Frequency, Hz
Power spectrum of
signal, period
a periodic
T sec.
it
has so
much power
Measurements
of, with,
and
in
71
White Noise.
in physical systems,
call noise
white
if it
has a
interest.
Xi
and x 2
is
equal
is
signal.
Refer to Fig. 4-46 for a simple circuit to measure the pdf of a signal.
i-to-:E2 gate is a circuit of biased diodes that transmits the highIf x 2 X\ is kept
frequency-clock frequency only when x x < x < x 2
The
FIG 4-3 Typical power density spectrum for a random signal. The total area
under the curve is the mean-square value
of the signal, usually spoken of as power
Shaded area is power
in noise theory.
in the frequency band from /i to f%.
72
Amplitude
Probability
density
p{x
Clock
>>
X\ to
X2
Counter
Gate
galte
1
Generator
[b)
FIG 4-4
density function.
The shaded area is
equal to the proportion of time spent by signal between Xi and z 2
(b) Block diagram for measuring pdf.
(a) Probability
constant as the two quantities are varied in steps over the amplitude
range of x, the number of accumulated counts in T sec is proportional to
the pdf at given values of Xi and x 2
The pdf curve can be plotted from
the measurements.
Accurate measurement of the pdf of a signal requires a long averaging
time at each amplitude level [see Eq. (4-2-8)] so that pdf measurement
by this simple technique is very slow and tedious. More elaborate
instruments having 100 or more parallel channels can give an on-line
display of the complete pdf curve in a very short time.
.
The most
familiar pdf
random
is
is
characteristic of
many
naturally
p(z)
disturbances.
i
i
e -x j2a
rV^n
where a
is
(4-1-9)
Measurements
The curve
many
falls
of, with,
away
practical purposes,
random
in
73
and
+ 3<r.
is
The
probability
increased.
so-called gaussian
1'
*.
74
4 x/a
(a)
/^/^probability that x>x.
P(x)=05
will
exceed
*x^/a
p(x)
=
7T(A 2
=
The Rayleigh
|x|
< A
\x\
> A
X 2 )^
is
(4-1-10)
distribution
is
the
P(x)
and
is
f_ x
p(u) du
(4-1-11)
when
the signal
is
Measurements
of, with,
and
in
the Presence of
Noise
75
p(x)
FIG 4-7
Probability
density
function
TABLE
4-1
*Gx(f)
R.(T)
Lowfrequency
noise
kGx (f)
Wideband
noise
G.(f)
*Rx (t)
Sine wave
*Gx (f)
t
Sine wave
R*M
plus
Gaussian
noise
"
GJf)
n
Low - pass
Rx(r)
white noise
M*^J
Narrow
Gx (f)
V/
^*"~
Rx (x)
bandpass
white noise
Wide
GrffJ
RX (T)
bandpass
white noise
Gx (f)
Rx (z)
Periodic
signals
spectrum
J_L
'
>
76
FIG 4-8
P(xh
distribution function
for a
quan-
tized signal.
p(x)
J2
than some specified
always unity, and the deriva-
The maximum
value.
tive of P(x)
is
ordinate of P{x)
is
p(x).
Mean Values and Mean-square Values from PDFs. The pdf can be used
to calculate the mean value, the mean-square value, or any other statistical
function of the amplitude of a random signal.
For instance, the mean
value Hx
may
be written
xp(x) dx
/_
M*
(4-1-12)
minus
is
is
all
amplitude
W = /_+
where ^ x 2
^x
is
Mx
x 2p(x) dx
(4-1-13)
<Tx
is
(4-1-14)
Measurements
Delay
of, with,
and
in
77
X(t-T)
line
xftj-
delay=Tsec
Averaging
,
xO-T)y(t)
circuit
Multiply
Approximate correlation
function
<(t) (outocorrelotion) or
//'/./(crosscorrelation)
R* (tJ--jfJ'(t-r)y(t)dt
y
yftj
FIG 4-9
signals.
The
their crosscorrelation.
and
y(t) is
two
signals x(t)
denned as
R xy {r) =
4im
7*
> ao
J.
l*
x(t)y(t
(4-1-15)
r) dt
**
may
signals,
may
,(t)
y(t) or
Alternatively
x(t).
write
1
lim
Time delays
fT
(4-1-16)
r)y{t) dt
is
[3].
block diagram of an instrument system that performs this calculais shown in Fig. 4-9.
One signal is multiplied by a
tion approximately
delayed version of the other, and the product is averaged. The result
In physically realizable systems the result
a function of the delay r.
Ideally T should be infinite, but
also depends on the averaging time T.
this would mean that it would take an infinite amount of time to get an
is
function of
x(t),
R xx (r) =
defined as
lim
x(t
fj ~v,
(4-1-17)
t)x()
-,-x-, dt
may be
The
listed:
autocorrelation function
is
an even function of
R xx (t) = R xx (-t)
t.
78
2.
The
is
mean
If
pedestal, amplitude
2
.
The
$x
3.
The
Mx
o-x
is
now
is
sitting
on a
equal to
maximum
at zero
delay, or
The
The
autocorrelation function of a
random
signal vanishes as
t*<x>
r.
8.
is
z{t)
x (t)
y(t)
and
R xy (r) =
for all r
then
Rzz(t)
= R X x{t)
Ryv\T)
The
Relationship between Autocorrelation and Power Density Spectrum.
autocorrelation function gives some indication of the relationship between
signal samples taken t sec apart,
is
depend
of the signal.
and it
some way on the bandwidth
in
is still
For instance, if
good correlation
between samples taken 1 msec apart, then the random signal is unlikely
have significant frequency components above 1 kHz. In general, a
wideband signal has an autocorrelation function confined to small values
of delay, and vice versa.
For example, the autocorrelation function
of white noise is just a single delta function at t = 0; this means that any
two samples of the same (infinite bandwidth) white-noise signal are
uncorrelated as long as there is a nonzero time interval between them.
A precise relationship exists between the autocorrelation function and
the power density spectrum. They are in fact a Fourier transform pair.
to
S(f)
R(t)
= y_ J"
= /_*"
R(t) cos
dr
(4-1-18)
(4-1-19)
2tt/t
where S(f)
real,
negative frequencies.
Measurements
of, with,
and
in
79
co
for the
The
=
=
for/>0
2S(f)
otherwise
(4-1-20)
total
(?(/)
R(t)
fj
f~
R(r) cos
(?(/)
2tt/t
dr
(4-1-21)
cos 2tt/t df
(4-1-22)
where G(J)
autocorrelation functions.
it
is
In particular, there
is
not contained
The
= /_*" R xy {r)
R X y(r) =
We
f*~ S xy (f)
relationships are
exp (-j27r/r) dr
(4-1-23)
exp
(4-1-24)
{j2irfr)
df
Gxy (j)
otherwise
>
0,
(4-1-25)
80
The
real
follows:
where
(4-1-26)
is
and
Q xy (f)
is
spectrum.
4-2
Measurement of Noise
Many measurements
The
of noise
is
used.
(integrals) should
be taken over infinite time to get precise results, but averages taken
over finite intervals can yield results sufficiently accurate for practical
purposes [4]. We must now look at the averaging process and see what
errors are involved when the averaging time is limited to finite values.
This matter is also treated in connection with digital analysis of signals
in Chap. 5.
The result of an experiment designed to measure some parameter of
random
signal
is
estimate.
The
over a period of
An
estimate of the
mean
value of a signal
is
written
sec
p.,
would be written
T
fl X
\j
x{t)dt
(4-2-1)
E(fix)
Mx
(4-2-2)
Measurements
of, with,
and
in
81
Mean-square error
in
measurement
of
mean value
fi x
=
=
=
E[{p. x
y. x
variance
of the signal x
nx )
2
]
(i* x )
(standard deviation) 2
(4-2-3)
confidence limits.
Actual error
<
2 standard deviations
The
Mean-value Estimates.
gaussian random signal
(4-2-4)
mean value of a
be a gaussian random variable, and in
the case of band-limited gaussian noise, the variance is given by
Var
where
(fi x )
is
component
care
<r x
will itself
(standard deviation) 2
(4-2-5)
and T
is
is
assumption that
BT
Example
a
flat
\2fl2
'2B7
(,00nY),
<o.oixiOmV
X lOT
T > 2 X 10 sec = 33 min
V- 2
95 percent certainty?
< 2 standard devia-
82
^f2>
Var( fc)
From
flat
we
this
(4-2-6)
1/T.
frequency of the
If the cutoff
filter is
[5].
Many
Digital Voltmeters.
Integrates
digital voltmeters
(DVMs)
are
available with the ability to measure the true average of the input voltage
may be
it
is
Example
What
maximum amount
the
is
integrating
DVM
not to exceed
if
<^
X
maximum
10~ 6
(7*
- <
jD
sec,
an
is
juV?
deviation of error
<
2BT ~
is
10-"
V /Hz
2
limits,
from Eq.
allowable error
(4-2-4)
we have standard
Measurements
This
is
the
of, with,
and
in
83
in order to achieve a
equivalent to 7
measurement
nV rms
yuV.
bandwidth from
to 100
Hz
is
(e.g.,
resistor at 27C).
digital voltmeter
in
some respects
this
is
time of
integration.
M
1
percent
demands that
For small
(4-2-7)
Jj,
maximum
errors, the
BT
must exceed 4
mean square
mean square
value;
e.g.,
10 4
this
.
values
is
measurement
or
by
digital techniques.
The simple
many
single-pole
RC smoothing filter is
84
the statistics
must be
may
to settle
down
The
/2.
between the amplitude levels x - W/2 and x +
affects the resolution obtainable
window sampling width
A wide window is unable to resolve rapid changes in
in the measurement.
On the other hand, a narrow window
p(x) and introduces a bias error.
of information collected is correamount
the
gives a good resolution, but
by the
signal
choice of the
SD
iP(x)]
where
is
(4 _ 2 _ 8)
VBTW$(x)
p(x)
seconds, and
0.7
is
the sample
window width
is
as a fraction of the
rms value.
Var [R xy (r)}
the expression
is
given
by
(4-2-9)
Measurements
of, with,
following expressions
and
in
85
for
functions
^
^
Var [R(t)] =
Var [R xv (t)] =
\R x J(t)
[R xv *(j)
+ R xx (0)R yy (0)}
ft(0)]
(4-2-10)
(4-2-11)
and
by Eq.
is
the averaging
y(t),
To
(4-1-16)?
R xy (r) =
lim
repeat,
- l* x(t -
r)y(t) dt
(4-1-16)
The
in
R*v(t)
= -
JJ x(t
by a variety
r)y(t) dt
of techniques.
(4-2-12)
t=
x ^ At
r ^0'A<)
(4-2-13)
86
Samples of
y(t)
earlier.
N
1
R xv {t) = - ^
x(t,
T )y(tj)
(4-2-14)
y=i
The sampling
instants
is
Analog-to-digital Conversion.
techniques,
form.
The
tion since
shown
it
it is
of computation.
It
has been
It
Measurements
of, with,
Spectral Measurements.
random
noise
is
and
in
87
subject to a
of
as
image effects that cause components at one frequency to affect measurements at another frequency. The errors depend on the length of record
and on the technique employed.
The power density spectrum of a signal can be measured by analog
techniques with a wave analyzer (narrow-band filter) having a true rms
voltmeter.
When the signal is gaussian random noise, the measurement
will exhibit a statistical error depending on the bandwidth B f of the waveanalyzer filter and on the averaging time T. The variance of the estimate
will
be
Var
[(?,(/)]
-^
(4-2-15)
Note that the statistical error in power spectral measurements is independent of the frequency being analyzed. It depends on the bandwidth
Bf, not the center frequency of the filter.
Digital techniques allow the power density spectrum to be computed
directly from the signal waveform, and the task is greatly simplified by
using the fast Fourier transform algorithm described in Chap. 5. When a
suitable computer is available, this is a very convenient method of analysis.
The third approach to power spectra is by way of correlation functions
and exploiting the fact that they are Fourier transform pairs. In many
practical situations, it is easier to measure the correlation function than
to compute the power spectrum directly.
4-3
as a Test Signal
88
indicators of performance
is
The
objec-
background noise introduced into a telephone channel by intermodulation distortion and conversations in other
tive
is
under
the system.
At the receiving end, a bandpass filter examines the vacant channel for
any indication of background noise. This is an example of a test situation
in which it would be inappropriate to use sine waves to simulate the
practical working conditions.
much
be
air
it is
In all these tests that attempt to simulate real working conditions, the
chosen technique depends on the detailed requirements of the system
being tested, and no general method is applicable to all situations (for
noise figure tests on communications systems, see Chap. 14).
between that test signal and the network output. The relative accuracy,
and convenience of the methods depend upon the problem and
upon the sophistication of the instruments employed.
The characterization of linear networks by steady-state frequency
response and by transient response measurements is well established, but
speed,
Measurements
of, with,
and
in
The main
89
areas of applica-
1.
3.
4.
test runs
x{t)
y(t)
hit)
FIG 4-10
x(t).
unit impulse.
s(t),
h(u)x(t
In the situation
mine
are
h(t),
we
u)
x(t)
du
(4-3-1)
explicitly.
is
to deter-
y(t)
the
yft)
\m
R.JT)
-
Delc, y
-*
t sec
FIG 4-11
correlation
- Average
Multiply
'
*
of a system.
lim
T*<*>
-1
J-
C
/
T
x(t
t +oo
t)
h(u)x(t
u)
du
dt
(4-3-2)
90
x{t) is
is
simplified
and reduces
to
R*v(t)
where
h(r)
=
is
HG
(4-3-3)
(0)Hr)
we
is
Thus we
G x (0)
is
true,
it is
y(t)
x(t)
h(t)
FIG 4-12
z(t)=y(t)
K>"
+ nffj
input
x(i)
Rxm(t)
This
is
and the
visible
output
V G (Q)h(r) + R
2
z(t)
xn (j)
R xu {t),
which
is
i.e.,
If the
R xn (r)
will
Under
of the
increases the variance of the result, but the statistical errors can be
Measurements
and
of, with,
in
91
+
h(u)R xx (i
|_ m
u)
du
input
is flat
If
may
applies,
(4-3-5)
R*u(?)
V G (0)Hr)
2
(4-3-2)
(4-3-6)
with the task of solving Eq. (4-3-5) for the function h(u). Even when
computing facilities are available, the errors in computing h(u) can be
considerable if the input signal does not have favorable statistics.
A far
better solution is to inject an artificial random disturbance having suitable
The total signal entering the item under test
white-noise characteristics.
then has two components, the artificial disturbance s(t) plus the inevitable
^*-mH^45 ft)
"""
R
^
Crosscorrelote
FIG 4-13 Injection of an artificial disturbance s(t) into a system under operating
conditions.
background disturbance x(t), Fig. 4-13. The crosscorrelation measurement is then made between the output y(t) and the random test signal
s(t).
In this way, the statistics of the background disturbance x(t) can
be eliminated, and it can be shown (Appendix 4-C) that
R*v(r)
Statistical
Test Signal.
AGMKr)
x
(4-3-7)
Errors in
The
an element yields the impulse response, provided that the input is white
noise and that the average is computed over a long enough period to
reduce the statistical errors to an acceptable level.
Two situations must be considered. One is the "noise-free" situation
in which the visible system response is due entirely to the clean random
92
The second
of spurious
of the
system.
Var [R xy (r)] =
0.(0)
+ ^ [&(0)]
/_t"
Ht +
u)h(r
way
u)
du
(4-3-8)
the bandwidths of the two signals x(t) and y(t) are not identical.
<
<r,
-^
0,(0)
-^ Gx (0)G y (0)B'
(4-3-9)
is
When the
is
Var
where
The
[R(t)]
z(t)
first
(4-2-9), the
= Var
y(t)
[R xv (t)}
Var [R xn (r)]
right-hand term
is
the
(4-3-10)
system output.
as Eq. (4-3-8), and by using Eq.
same
du
(4-3-11)
Exact evaluation demands precise information about the test signal x(t)
and the background noise n(t). However, three special cases can be
considered
1.
Noise bandwidth much less than the signal bandwidth. Whenever the
is introduced into the measurement at a point before
test, then the observed noise bandwidth at the output
background noise
the system under
Measurements
of, with,
and
in
93
Var [R xn (r)] =
where
Gx (0)
is
(4-3-12)
2T
random
test
Var [R xn (r)] =
where
3.
From Eq.
is
(4-2-11)
we obtain
(4-3-13)
2BT
Var [R zn (r)} =
(4-3-14)
2T
4-4
The spectrum and the pdf are two important statistics of random noise
make it attractive as a test signal. However, the very random-
that often
measuring interval
is
pdf.
periodic
AT
FIG
4-1
signal
js
called
pseudorandom noise;
random
it is
becom-
noise in
many
applications.
94
-A A +A7"
FIG 4-15 Autocorrelation function of
clocked random binary sequence.
The
signal
is
N AT sec
(Fig. 4-176).
Many
of this signal
Log (V 2 / Hz)
FIG 4-16
trum
for
signal.
frequency
^--i/Ar
is
no
Measurements
and
of, with,
in
95
^rrjj-u J~LJJ
10)
A
FIG 4-17
AT
-14
(b)
(a)
-7A7-
AT
14
Ar
auto-
its
correlation function.
One technique,
for generating all possible sequences.
however, has found widespread use and can generate sequences of length
The following
2 n 1 using an n-stage shift register with feedback [6].
example (Fig. 4-19) shows how a three-stage shift register can generate
the repeating pattern of seven bits (Fig. 4-17) having the ideal autoFeedback taken from the second and
correlation-function properties.
known technique
1 if
inputs are
Power
Log
'II
-Log
frequency
fc =\/AT
FIG
4-1
Power spectrum of
96
Initial
contents
After
1st
shift
After 2d shift
After 3d shift
Output repeats
after 7 shift
pulses
FIG 4-19
register
pseudorandom
pattern.
1.
2"
1.
All possible patterns of Is and Os that are n bits long occur just once
complete pattern length, except the "all-zeros" pattern, which never
occurs.
Hence, there are 2" - 1 steps rather than 2", and there is slight
2.
in a
one more
first
Because the pseudorandom waveform is periodic, complete informaany statistic of the signal can be obtained by averaging oyer
a time interval of one fundamental period equal to
AT sec. There will
be no statistical error.
6.
tion about
is
Measurements
of, with,
and
noise.
is
in
97
many
process
particularly useful in
by simple
logical operations.
This
and the pdf of the test signal are important, and in general, the required
pdf will be gaussian. How can we generate a pseudorandom binary signal
having a gaussian pdf? This is easily achieved by passing a pseudorandom binary signal through a low-pass filter with a cutoff frequency at
about one-twentieth of the clock frequency. The resulting waveform has
a pdf that closely approximates a gaussian distribution for all signal
amplitudes up to about 3.5 times the rms value, and could be improved
for even higher signal amplitudes by decreasing the filter cutoff frequency
still
further.
However, a
is
should be pointed out that the length of the pseudorandombinary signal sequence has a profound effect on the distribution, short
sequences giving very poor approximations to gaussian. For a 20: 1 ratio
between clock frequency and filter cutoff, sequences shorter than 8,191
quate.
(2
13
4-5
We
It
1)
Measurement
find
that
all
in
distribution.
essential
random nature
of the quantity
98
Noise.
by correlation.
Measurement of dc in the presence of noise can be achieved by pure
integration or by low-pass filtering.
The variance for both situations
is dealt with in Sec. 4-2.
As a general rule, we could say that pure
frequency-selective filtering, or
integration gives a result to within a specified error in about half the time
required for exponential smoothing. The smoothing filter has the
advantage of giving a continuous reading in a simple manner, which
may
Measurement of Sinusoids
by a
followed
in
linear detector
bandpass
filter
the simplest
if
cos 2-irft
1
Filter
o
\J
n(th
FIG 4-20
v(t)
Linear envelope
/ detector
\
xtt)
JL
in
noise,
by
z(t)+A
^J-P>i-
linear detection.
See Fig. 4-20. If the sine wave is A cos wt and the noise n(t) is gaussian
with a uniform power density spectrum of G n W/Hz, the following
relationships are true:
V(t)
= A
Bf =
x(t)
wit)
=
=
+ n(t)
cos ut
filter
w(i)
bandwidth
+A
cos wt
input
filter
in hertz
filter
(5/2 about
center frequency)
output
then
w
<r
= Gn B f =
noise
power
at filter output
Measurements
will
of, with,
and
in
z(t),
but for
Var (i)
finite
averaging times,
99
(4-5-D
where B t is the effective bandwidth of z(t), and T the true averaging time.
(The effect of exponential smoothing instead of pure integration is dealt
with in Sec. 4-2.) In this particular case, the bandwidth of the detector
noise output B z = 1/2B, and the rms noise output <rz is the same as the
Var(i)=^
Now
<r
filter is
= B f Gn
Var (A) =
.-.
(4-5-2)
(4-5-3)
(4-5-4)
This result seems to show that the variance on the estimate of the sinewave amplitude due to the wideband noise is independent of the bandwidth
The
time.
of the bandpass filter, and depends only on the averaging
sufficiently
result is approximately correct, providing that the filter is
narrow to ensure, say, a sine wave-to-noise ratio of 6 dB at its output,
and
1/B.
Detection of
Sine Wave.
For noisy
ing the noisy signal with a clean version of the sine wave available locally.
Both magnitude and phase of the hidden signal can be determined. In
this
is
known
as coherent detection.
x(t)
Then
= A
sin (a c t
<*>)
(4-5-5)
n(t)
wave
R, x (t)
= A
sin
co c (*
4>)
2i
cos (w ct
-</>)+
sin
coo(i
+
-
gives
n(t)]
(4-5-6)
r)n(t)
(4-5-7)
100
cos
(w ct
between
wave
+ R tn T
(4-5-8)
( )
<f>)
Now, in general, the term R, n (r) will be zero if the noise n(t) is truly
random, so that the crosscorrelation R xa (r) exhibits only the periodic
term of amplitude A 2 /2 and phase
This result is true when averages
are taken over very long times, but for averages taken over a finite time,
the statistical error, from Appendix 4-D, is
<f>.
Var
(R,x (t))
SD
= A*Gn
4T
Gn
[estimate of (A 2 /2)]
A72
where
Gn
is
(4-5-9)
A*f
Detection of a Sine
Wave
in
Noise by Autocorrelation.
One
of the
most
on the
autocorrelation function
If
a signal
noise
n{t),
FIG 4-21
noise.
z(t) is
the
R nn (j)
sum
we have
of
random
z(t)
s(t)
n(t)
r,
whereas the
= A
cos at
in
random
t.
and random
+ R nn (r).
Measurements
However, R(t)
B(r)
for large
of, with,
%A
V A* COS
cos wt,
and
in
and
for large
t,
#(t)
0.
>
Thus,
(4-5-10)
cor
101
t.
We see therefore
tion for large
t, it is
APPENDIX 4-A
Determination of Impulse Response by Crosscorrelation
(r)
lim
lim
x(t
r)y(t) dt
x(t
r)
+X
h(u)x(t
dudt
(4-A-l)
-u)dt\ du
(4-A-2)
u)
R xy (r) =
The second
+
/"_
lim
A(.u)
7
f
xit
r)x(t
integral
is
(r-u)
R ty
T)
[*J
h(u)R xx (r
u)
(4-A-3)
du
Now, if we choose the bandwidth of the noise to be much greater than the
system passband, then in Eq. (4-A-3), h(u) will be a relatively slowly changing
function in comparison with R xx {t - u). The term /i(u)will be almost constant
over the small range of values of u around u = t for which R xz (t u) has
The integral then becomes
significant values.
R xv {r) =
h(r)
/"_+"
R xx (r -
u)
du
RM = V
2 h(r)G x (0)
where
(4-A-5)
G x (j)
(? x (0) is
(4-A-4)
102
APPENDIX 4-B
Effect of Contaminating Noise at
Output of System
R*
= Km
( T)
v
lim
JJ
x(t
+ jQ
+ R xn
T
x(t
r)n{t) dt
{ T)
(4-B-l)
white noise,
If x(t) is
R x,{t) = UGMKt) + R xn
If
r)y{t) dt
R*z{t)
z(t),
T
fT
[ x(t-r)z(t)
background noise
Hence,
{ T)
(4-B-2)
n(t) is uncorrelated
x(t),
R xn (r) =
then
for
all t.
RM = y G
2
That
is,
the result
(0)h(r)
is
(4.B-3)
by the presence
unaffected
of
an uncorrelated background
disturbance.
APPENDIX 4-C
Effect of
The term
x(t) is
to a
s(t)
is
System
an
artificially injected
test signal.
R sv {r) =
y(t)
s(t
Jq
response due to
=
Making
lim
/_
h (u ) x (t
r)y(t) dt
x(t)
u)du
(4-C-l)
response due to
j"^
h(v)s(t
s(t)
v)
dv
(4-C-2)
the substitution for y(t) and using a similar argument to that used in
R
t)
},
v( T )
y_ M
h{u)R sx ( T
- u)du +
[ h(v)R
ss ( T
R.v(t)
VG
2
(0)h(r)
v)
all r,
dv
and
(4-C-3)
if s(t)
is
white
(4_
C -4)
Measurements
of, with,
and
in
103
Crosscorrelation between the output and an injected white-noise test signal yields
the system impulse response, regardless of the presence of an uncorrelated
disturbance at the input.
APPENDIX 4-D
Statistical Errors in the Detection of a Sine
Wave
in
Noise by
Crosscorrelation
When the frequency of the sine wave is known, its amplitude and phase can be
measured by crosscorrelating with a clean reference sine wave, even in the presence
Statistical errors are involved when
of large amounts of background noise.
averages are taken over short intervals, and these may be analyzed as follows:
Let the noisy signal x(t) consist of a clean sine wave s(t) hidden in a background
The
of band-limited gaussian noise n(i) having a flat spectrum up to B Hz.
frequency fc of the sine wave is less than the maximum noise frequency. We can
write
s(t)
x(t)
where
= A sin 2ir
= s(t) + n(t)
(4-D-l)
ct
n(t) is
(4-D-2)
fiat to
Hz,
The
B>
fc
The operations
statistical errors
can bp
s(t)-A
sinco*/-
y(t)
Average
RXS W
Noisy
signal
Deloy
x(t-x)
X(t!
FIG 4-D-1
in crosscorrelation.
y(t)
multiplier,
y(t)
y(t)
If
= s(t)[s(t - r) + n(t - T )]
= A sin 2rf sin 2irf (t - r) +n(t 2
ct
of y(t) over
(4-D-3)
r)A
an interval
sin 2ttfc t
of
components
A mean
product
s(t)s(t
t)
cos 2nrfc T
(4-D-4)
be three
104
Gn(f)
GffJ
Hz
(a)
fr
HZ
S,(f)
V4/,.
a} g
'Hz
jL
(b)
-<B-fc )
B+fc
GjfJ
i i
/^\
Gn
FIG 4-D-2
'Hz
Power density
spectra
(c)
(B- fc)
B+ fc
fc
t)
at frequency 2/c
This
will
be completely
fc
A component due to modulation products between noise n{t) and sine wave s{t).
Noise in the band around the frequency fc contributes to the variance in the
estimate of R sx {t).
Only the first term gives rise to a dc component, and this is equal to the true
value of the correlation function R sx (t) which we are trying to measure.
The problem is to evaluate the variance on the estimate of this mean value
caused by the second and third terms, but since the second term is negligible, we
on the third term.
can be written
shall concentrate
result
Var
[R. x (t)]
var ()
2BT
By
(4-D-5)
where cr yn is the rms value of the random-noise component of y(t) and B is its
equivalent bandwidth, and this can be written in terms of the low-frequency
power spectral density
Var[^(r)] = var(A v )
=^
(4-D-6)
"
Measurements
The magnitude
of, with,
G v (0)
of
We
and
in
105
y(t).
can now compute the variance in the estimate of the mean value of
A Gn
y(t).
Var
where
()
Gn
is
Var[ZL(r)]
The normalized
.
(4-D-7)
(4-D-8)
error
Wl =
f"
RM
J-
(4-D-9)'
v
\a*t
CITED REFERENCES
1.
Bendat,
chap.
3,
2.
Ibid.,
chap.
3.
Ibid., chaps.
4.
Ibid., p. 243.
of
Random Data,"
9.
chap.
5 and
6.
5.
Ibid.,
6.
Golomb, Solomon W.
San Francisco, 1967.
7.
8.
6.
:
1961.
9.
1954.
CHAPTER
FIVE
SIGNAL ANALYSIS BY
DIGITAL TECHNIQUES
Ronald W. Potter
Hewlett-Packard Company,
Santa Clara, California
The
numerous
texts
requirement of a
mixed.
106
It
is
The
signal
is
may
107
emphasize the
drive train?
The use
may
108
Again,
it is
transfer functions
different
excitation signals.
In
causes a "smearing" of the true waveform and hence introduces a resoluIn this case, a deconvolution or inverse-filtering techtion limitation.
nique can often be used to recover some of this "lost" resolution. Thus,
the resolution of devices such as the gas chromatograph, the nuclear
scintillation detector, and the sampling oscilloscope can be improved by
using this technique. A similar scheme can be used to improve seismographic records that are obtained during exploration for oil or gas deposits.
The power spectrum "signature" of a device can be used for identificaFor example, underwater noises can be used to locate
tion purposes.
schools of fish or to identify ships
and submarines.
5-1
Fourier Transform
[1, 2, 3,]
F(s)
f(t)
f' m
f(t)e-
iiTSt
dt
F(s)e i2r,t ds
direct transform
(5-1-1)
inverse transform
(5-1-2)
is
given by
0^)
<->
ber i *'"F(b8)
(5-1-3)
by
Convolution
*(0
Let
fit)
is
defined as follows:
/_" /( x
^ Fis),
m< -
x ) dx
in
5 - 1-4)
git)<r+Gis),
Thus convolution
It is
109
From
can be obtained:
F(0)
f~m
and
/(0 dt
/(0)
f~m
F(s) ds
(5-1-5)
Thus, the area under /(0 is simply the value of its transform at the origin.
From this it can be deduced that the convolution operation multiplies
The area under h(t) is the product of the areas under /(0 and g(t).
areas.
Integration and differentiation operations are
Z)<>F(s) <-> (-i2irt) nf(t)
(n)
is
Df(t)
Integration constants
negative.
By combining
<-
(i2xs)F(s)
Integration
is
(5-1-6)
obtained when n
is
shown
pairs
in
Table
5-1,
it is
often fairly
The transform
symmetry and
TABLE
Abscissa resolution
5-1
List of
Common
is finite.
Transform Pairs
F(s)
fit)
HI (8)
IIKO
spacing
sin xs
n(0
0--7T.2
Vi
-tU
U(t)
ITS
+ 1) - 8(s -D]
**[*( + 1) + Ks -Dl
sin
2irt
COS
2irt
Unit step
110
As derived below,
domain implies
integration in one
finite
Assume
that
is
finite abscissa
f(tl
'
'
r~~\
'0riginal\
FIG
5-1
-i_/
-2a
-3c
2a
function
convolved with an
infinite
3a
train
of delta
functions.
F(m
As)
At
(5-1-7)
n=
N-l
j{n At)
X F m As)e i2rmn/N
= As
(5-1-8)
Now the
operations.
If
indices,
is
Though
the following few paragraphs will not enable the reader to use the algoit will give him a brief introduction to the procedure.
Consult the references [4, 5] for detailed instructions.
t
rithm in computations,
111
matrix
new
R-l
n =
r
where
ra
k r 2*
X=
(5-1-9)
2R
J2 r
(5-1-10)
r="0
mn
given below.
is
R-l R-s-l
^2
~N~
s
The
mn/N
=0
(5-1-11)
jrk.2'+*-*
r=0
may now
F(J0,jl,
,j R -l)
be written as follows:
^ 2 2
= ki=0
io
R-l R-s-l
/( fc o,^,
,fc*-o
s =
*_i=0
n
r
e- i2 - e
'
(5-1-12)
where
Qjk =
jT k,2*+r-
it is
replaced
replaced
by Jr- t -\.
by the most
Thus the
is
summed
significant indices of
ra.
After
are
summations, the
summed
The multiplying
112
summation
The discrete Fourier transform is not the same as the continuous transform, and the resulting frequency functions may be considerably different.
As indicated
cases
the
line shape.
frequency
axis.
chapter.
5-2
[6, 7, 8]
its
Then
is
Consider a random
fb+Ax
Ax.
it is
p(x) dx
Obviously, the
possible to describe
The
F( s ) =
p( z)
f'n
f"^
where x"
is
is
p(x)e- ii "'dx
(5-2-1)
F(s)e< 2 *" ds
(5-2-2)
= f'm
p(x) dx
variable
its
random
characteristic function of a
transform of
113
1.
(-i2irx) np(x) dx
J~m
the nth
moment
= /-^f. =
at s
(-t2ir) n
gives
(5-2-3)
of p(x)
xnp{x) dx
(5-2-4)
Thus, the moments of p(x) can be easily calculated from the derivatives
Similar equations apply for multidimenof the characteristic function.
sional distributions.
The cumulative
P( X) =
probability distribution
is
defined as
(5-2-5)
(x) d\
f*^ p
This expresses the probability that the random variable is less than some
The expression P(x) is used to obtain limits within which the
x.
random variable will be found a prescribed percentage of the time.
By using the characteristic function, it is easy to show that the probability density of the sum of several independent random variables is
simply the convolution of their respective probability densities. Define
value
as the
where
=1
sum
P(Z)
Xi
Z,
of
n random
variables.
Pl(Zl) * P2(Z 2 ) *
Pi(xi) is
Then
Pn(Xn)
random
variable x i}
and the
A random
variable,
it
is
function of time.
114
may
stationary process
random
process.
are the
same
as time averages.
Most
work
in this field
of
power
may be
used in place of an
ensemble average.
5-3
Signal Analysis
The
more
be
of
known
The questions
How many
buried in noise?
signal averaging?
or deconvolution?
What
What
found to be
useful.
the count in a
memory
whose address
tude.
The
cell
is
number
115
important to note that the differential linearity of the analog-tois very critical in histogram applications.
To understand this, denote the voltage resolution of the analog-to-digital converter
by AE. Ideally each AE would be identical, but in practice there is some
variation in AE from one voltage level to another.
A typical probability
It is
digital converter
density
shown
is
in Fig. 5-2.
P(AE)
Define
AE m
Then
of AE).
maximum
deviation from
is
77
is
the
same
as
have an amplitude
For example, 1 percent
percent amplitude variation in the
shown in
produce
value
In histogram
directly proportional to the
AE (mean
= AEm /AE.
will
Fig. 5-2.
histogram distribution.
Assume that
histogram.
Define
pz (m)L(\-e)
+1
p (m)
is
2(1-e)
FIG 5-3
count (see
text).
The
total of
with
one
FIG 5-4
Probability
density
after
text).
116
p(m)
-J
m (l
is
given by
n- m
e)
(5-3-1)
where
\mj
ra!(n
m)\
For ne(l
tribution.
PW ~ w
This equivalence
) ^$> 1,
^r(--)''WN]
7T
is
[9].
= Vnc(l
e).
of the statistical
1, all
a priori, but
recognized.
where
z
So,
signal
is
is
tor
twice as
t.
known
is
given by
(M 3)
-
'
of z
-f- .
Thus
proportional
to
= sin
Thenf(0 = cos =
p(z) = l/[x(l z 2 )^] for 1 <
f(t)
readily be
pW
when the
its
(1
slope.
For example,
define
of 2.
memory channel
are
117
with a
mean x
FIG 5-5
one
Probability
density in
of signal averager.
memory channel
sum
If
n random
together n times.
of these
This
will
Thus the averaging process has reduced the noise voltage by Vw, provided
all
-1
(5-3-4)
f(t')g(t')
n=
=0
where
H(s)
- y
[F(s) * G(s)]e-i2
"
k
n=0
- [F(s)
* G(s)]
e- 2 *"
(5-3-5)
118
*w
-hi
n=0
sin itNkrs
The equivalent
is
(5-3-6)
f>ir(Nl)kT8
N sin -nkrs
sin TiNkrs
l*()|
N sin
(5-3-7)
7i
krs
The function \<f>(s)\ resembles a comb filter with major lobes ats = m/kr,
where m is an integer (see Fig. 5-6). The quantity l/kr is the rate at
which the various records of data are acquired. Thus all harmonics of
this data-acquisition rate are preserved in the signal averaging.
There
are minor lobes at s = (m + }^)/Nkr, with an envelope (1/2V) esc Tikrs.
Further, \<f>(s)\ = 1 for s = m/kr (major lobes), and |<K S )I = 1/N at the
midpoint between major lobes. Thus various frequencies are attenuated
by different amounts depending on their relationship with the basic dataacquisition rate.
There are N I lobes, and the main lobe is twice as
wide as the side lobes. Thus as more records are averaged, the main
lobe narrows, more side lobes appear, and the attenuation between major
lobes increases.
The following relation can be readily derived:
fl/kr
jo
<Hs)4>*(s) ds
rl/kr
jo
i<Ks)l
ds
(5-3-8)
Thus, the power of initially white noise is reduced by l/N by the averaging
process, although the noise is no longer white.
At the end of the chapter (Fig. 5-32) the effective transfer functions of
a signal averager are shown, as determined by a digital computer. Starting with the top photograph, the value for
is progressively increased:
2, 4, 8, 16, 32.
FIG 5-6
Equivalent
filter
5.
Digitizing Errors.
digression will be
made
119
sentation.
T = n
so that
Error Eft)
M-.
.*
r^r
FIG 5-7
-A/
+ AE/2
or less,
and
Fig. 5-7
p(E)
1
FIG 5-8
Probability
density
sam-
of
pling error.
random
variable
is
rectangular.
The variance
of
is
(AE) 2 /12 and represents the total noise power. The autocorrelation
function is simply a train of delta functions spaced T apart, each having
an area of At (AjB) 2 /12. The power density spectrum (Fig. 5-9) is therePower
density
(An
S (frequency)
FIG 5-9
of sampling error.
n samples
of the
120
of
amplitude At
Em =
2
at each frequency
is
2 2m
Thus
The power
signal-to-noise ratio
is 64.98 dB.
If a full-scale dc signal is
sampled, the power density spectrum would be a single sample point of
amplitude n At E m 2
Thus the power signal-to-noise ratio is improved by
.
the
number
sample points
in the record.
It must be emphasized that
the above results are only valid if the errors at the various sample points
are independent.
When sampling coherent waveforms, this condition
may
error
of
not be met. The worst case for coherent sampling occurs when the
is constant over the record with an amplitude of AE/2.
The power
it
is
many
possible to average
signal records.
The
E(t)
P(E)
.^(E-Es)
-E
FIG 5-10
An
arbitrary
FIG 5-11
nonrandom input
voltage at
signal.
zero.
AE
axis
is
Probability density
in Fig. 5-1 0.
for
width
AE
p(m AE) =
is
8(E-m AE)
] (m
_^E
Po(X
E.) d\
(5-3-9)
121
The
oo
Pl (E)
p(m AE)
m=
oo
^ KE -m AE) f"a n
A
A^
x
) P"(
~ E
>)
dx
(5-3-10)
where \1(E) is a rectangle of unit width and height centered at the origin.
Note that H(E) = f~\( E). Thus the integral is represented by the conp!E)
p(E)
p
P (E-ES )
(e-&
-|Ak
Es
5-1 2
with
volution of
III
FIG 5-13
FIG
H(E/AE) and p (E
U^\
= AE
8(E
m=
'
of p().
Define
E,).
- m AE)
oo
Thus,
" ()
The
n
[
P " (B
[,
n
"
is
IU
,)
]
(M
"
U)
given by
sin ns
(s)
AE
e- 2 .,]
(5-3-12)
ITS
where
(s) <-*
(-E)
is
Pn (E)
where
Pl ()*
(5-3-13)
denotes convolution
n()
ni()
(5-3-14)
122
^=
where z m
l m ~Va^T n W^"-
(m/AE).
^_
where
Note thatn^O) =
of E is given by
(5-3-15)
(0)
andn
n<oo =
two derivatives
Thus it
of U(s) at s
is
+ W (w - l^r^n^OO]
^cosa^-sin
sin az m
+ - [n
zm
(5-3-18)
(1)
(zm)
i27T^ f n
(2 m )]
e- i2 -^
(5-3-19)
AE
AE
-n
ni o) (o)
5^ 0,
(5 . 3 . 17)
2
^^^
and
for
(-i) m
m =
11^(0) = no<(0)
Q|)
--2^s/ A
(5 _ 3 _ 20)
0,
i2t#.
-z2,r.
(
The second
-2
for
derivative
Only the
algebra.
is
5 .3_2i)
considerable
result for
= _ (2*^)2
(-D^f^n
/-)
+*-^(s)]
e-i2rfnEJ^E
(5_3_22)
5^ 0.
ni <(0) = -(2,)^.
for
0.
necessary to evaluate
mir'm^is)
ni(1)(s)=
0.
U n ^(s) = nnr'is^^is)
for
(0)
nw(0)
where
(1)
moment
m =
H.<(0)
=nn
= nll^O)
<(0)
(5-3-23)
0.
nn
<
- ^(T A) + n
>(0)
>(0)
(5.3.24)
+ n(n - l^n^O)]
(5-3-25)
n< 2 >(0)
[n B <(0)] 2
after
wiii< >(0)
summing
of the
n[n^>(0)] 2
123
mean from
(5-3-26)
is
summing.
E = E,+
E*
(5-3-27)
n Ll2
- IV
(A) 2
(E)
2)
AE
in
+5
(0)
47T'
(5-3-28)
\
(5-3-29)
,AEj
for
t* 0.
v*
m=
AEXAE
m\
oo
2rmE,lbE
for
for
(5-3-30)
e i2irmEJAE
(5-3-31)
7^ 0.
m^O.
The
all
(A/) 2
_|_
errors caused
practical
by sampling
purposes.
are
Notice that
in the
and
AE
2
.
for
EJAE.
The
errors
depend on the
resolution.
for E.
it is
tribution
is
Vn(E)
given explicitly by
e -\HE-E)IZ*
2(2t)
HI
(S)
(5-3-32)
124
where S = E - (E)
The random variable
actually is a discrete
variable with an interval AE/n (the averaging process has reduced the
interval by n).
However, p n (E) may be considered a continuous function for all practical purposes.
2
A =
where
finding
E
E
is
as follows:
dE
(5-3-33)
a confidence interval.
in the interval
and
specified
}e.-e P(^)
is
E - E <E <E +E
a
calculated.
can be written as
where erf (z) is the error function. Given A, then E can be calculated or
determined from a table of the error function.
Consider a special case where the additive noise before sampling has a
gaussian probability density with zero mean.
where a 2
~^
Po(E)
e-*(*M*
(5-3-35)
is
The
characteristic function
is
n
The
(s)
error in
= -
e -2(*")
(5-3-36)
t J^t
x
e -2(m*iAE)* sin
in
Unm
\
^
AE/
is
(5-3-37)
cance
is
1.
Thus,
AE
e-^'iw
AE
|e|
max
S in
e-"/ A
2*
E \
-^A
(5-3-38)
(5-3-39)
For example, suppose |e| max = 0.01(A#/2). Then a = 0.4587 AE. Thus
two samples within the gaussian noise envelope will give a mean-value
W-
(EY =
^(A) +
2
<r]
52
written as
H^^y + A
125
S2
(5-3-40)
_1 e-2(ax/A)
.,._....
r/A^\*
*_\
I
(
E.
COS
cos
2x^4.
2ir
4(r
.
2
E.
cos 2x
AE Es
sin
2t
(5-3-41)
for 8 2 assumes m = 1.
Note that S 2 is very small because
term e-2(*WA) an(j so w^ ^ e ne gi ec t e d. Thus the mean and second
moment of the final voltage after sampling have been determined. For a
specified number of averages, n, and a specified confidence level A, how
much noise a should be added prior to sampling to reduce the digitizing
error to a minimum, and what is this minimum value?
The optimum a
The expression
of the
zero.
tion
is
coth
. _ 1
(*!)
E
Eo\d2/d<r 2J
\2 2 /
AE T
/l
(*&)
%-]d
(2T)!
(i2
<r
\1
^)r""
1B>
'
(M 42>
-
Define
and
22
A;
=
(2)^2
Then
x tanh
a;
2fc
where
nE
iM[H 2 +
1
(<r
/A#
)]
and
k
#o /n\
AE \2/ [y12
(<r*/AE*)]
(5-3-43)
126
The confidence
A Normally
A ~
level
[erf
\x/2k\
erf(k)
(*
can be expressed as
+ !)+
~2i$yi e
(k
A may
and so
|fc|
erf
(5-3-44)
)]
be approximated as follows:
-h '
(5 " 3 - 45)
Assume that A
erf
(A;).
ArxY
<r
= ~ e-^i^
AE
2k 2
(5-3-47)
ttx
a/AE <
0.4 (n
<
100 or
shown
n = 10 5 then
E /AE =
m =
1,
this
E /AE >
in Fig.
method begins
0.1).
For example,
5-14.
to lose accuracy
Curves of n and
if
E /AE ver-
a/AE =
0.57 and
0.005.
function.
transform
pairs.
so useful
The primary
its
makes
it
in the
is
transform domain.
wise be hidden.
The nature
of a time
window and
127
1,000,000
100,000
10,000
1,000
100
A
FIG 5-14
Optimum
disadvantages. The side lobes only drop 6 dB/octave, and hence adj acent
frequency channels interfere with one another (see Fig. 5- 15a).
138
This
(*J
will
broaden the
FIG 5-16
Hanning
129
2
line shapes
main lobe of the line shape and thereby reduce resolution in the frequency
domain, but this is the price for better separation between adjacent
harmonics. There are many window functions that can be used, and
numerous shapes are described in the literature on this subject. Each one
has advantages and disadvantages depending upon the desired result.
In this chapter, only two will be discussed in detail: (1) the Hanning
window, and (2) the Chebyshev window. Hanning is useful because it is
so easy to implement, and the Chebyshev window is of interest because it
gives the narrowest main lobe for a given side-lobe amplitude.
130
Window
area
FIG 5-17
domain.
2 b{s)
345(s
g(t)
2 [\
shape
line
1)
<
<
T.
is
}i8(s
1)
Separation
magnitude to
is
maximum
as
side-lobe magnitude.
maximum
main-lobe
relationships.
Table 5-2 shows the main-lobe width, separation, and window area
Graphs are included in Fig. 5-16.
various Hanning operations.
Recall that As = 1/T is the spacing between adjacent points on the
frequency axis. The relative window area indicates the reduction in
for
TABLE
5-2
Function
Rectangular window
Hanning once
Hanning twice
Hanning 3 times
Separation,
dB
13.26
31.47
46.74
60.95
74.61
Main-lobe
width
Relative
window
dB
1.626/7
3.743/r
5.782/7
1
7.804/7
9.818/77
-6.02
-8.52
-10.10
11.26
area,
more
(Fig. 5-18) is
G(s)
it
cos [(tsT) 2
_
-
2lH
a2
}
implement than
cosh-1
difficult to
name
The
131
(5-3-48)
aliia)
where
side lobes,
and
Ii(a)
is
1.0
r
I
/,
}
'
\ph-*,x*\
I^(a)i\-4x*
-cosh"
^1
domain
Note:
loc ated
0.7
at x- 1/2
2a/,(a)
is implied.
0.6
40 -dB
ripple
Curve C:
-60-dB
ripple
Curve D:
-80-dB
ripple
\A
\B
0.5
0.4
0.3
0.2
0.1
0.1
0.2
0.4
0.3
FIG
5-1
Chebyshev window.
0.5
0.6
0.7
The
132
12
-1 00
FIG
5-1
Chebyshev
line shapes.
is
- (2*/T)
Vl - (2t/T)
/i(flVl
2a
)1
(5-3-49)
Graphs of this line shape (Fig. 5-19) for various ripple factors are included.
Table 5-3 summarizes the characteristics of this line shape.
TABLE
5-3
Characteristics of
133
Chebyshev Window
Ripple factor,
Separation,
dB
dB
Main-lobe
width
-20
-40
-60
-80
-100
20
40
60
80
100
3.373/r
4.839/T
6. 305/7
7.771/T
1.906/77
7
Relative
window
area,
dB
-1.40
-4.59
-6.38
-7.64
-8.62
The window height is unity and therefore the line-shape area is unity.
The line-shape height and the window area are (cosh a)/[al 1 (a)]. The
peak ripple is l/[ali(a)]. The window function includes delta functions
of amplitude l/[2ali(a)] at each end of the window interval.
The window
amplitude immediately adjacent to these end points is a/[2Ii(a)].
A graph of separation versus main-lobe width (Fig. 5-20) is included for
comparison of various window shapes. Note that the Chebyshev line
shape gives considerably higher resolution for a given separation. The
Parzen window given by g(t) = 1 2\t\ is also shown on this plot. This
window has a relative area of 6.02 dB and a line shape given by
G(s)
pW2>
J'
The rectangular window is fine for signals that have an integer number
window interval T, or for single transients which
decay within the T interval. A weighting function is only needed when
the waveform value at the beginning of the window is different from the
value at the end. This is particularly important for random data and for
of periods within the
134
10
sin its
Rectang
window
jlar
ITS
-20
./sin
ndow:2(
ITS
2
1TS
y)
-30
Hanninc once
-40
Hannir g twice
-50
Chebyshe v\
I'm e
shap e
\\
-60
g 3 times
-70
Hanninc 4 times
80
-90
00
T = window
width, s
line.
135
signals.
Thus it is useful for searching unknown data to determine the
nature of any information that may be present. Since translation of the
time origin has no effect on the power spectrum, the spectrum is often
common
it is
to
As indicated
culated.
number
of averages
may
be
cal-
random
variance by the
mathematics more
random
variables.
variable
is
gaussian.
x2
if
(5-3-50)
where x and y are independent random variables representing the real and
imaginary parts of the signal spectrum, and z is the random variable
representing the power spectrum at each frequency.
hy{x,y)
This
is
= -^- e-
[(*-*+(.,-*)=]
(5-3-51)
same variance
mean
(1/2ff2)
o-
2
,
and
their respective
mean
y.
frequency.
calculated.
Me)
dz
= fxv (x,y) dA
(5-3-52)
136
where fz (z)
is
and dA
z,
r cos
6,
r sin 6, z
r 2,
dz
and dA=rdrdd =
2r dr,
}/^dz dd.
Thus,
f( z )
= -J
47T0-
Some
de
(5-3-53)
of Standards,
w)
- f* e
* <*4
<="
IT
o
2(T
is
)(Jcos+i7sin9)
Bureau
It
tabulated form.
[' e (v^A
-'
t
l
1
e -(.m )U+* +v )
/o
,
2
(x 2
(
2
\<T
(5-3-54)
in closed
+ )^
2
form as
e -(i/2,'KHW)
for 2
>
(5-3-55)
From
fj
f,(z)e-<
dz
(5-3-56)
the same reference as above, page 486, Eq. (11.4.29), the following
Substituting z
z (s)
=
1
(6) do}
= w2
dz
= -^ e b'i* at
2o>
dw,
(5-3-57)
and expressions
for a
and
b,
1
e -(*'+i/ )r.-2T,/<i+.w)]
i4iro-
(5-3-58)
interest, as follows:
Mean
value
f,(0)
= ^4r =
t2n
Second moment
z2
x2
p" (0)
'
47T
(5-3-59)
-{-
(x 2
2er
+
Variance
The coherent
z2
(z)
signal
4<r
power
is
(x 2
x2
(5-3-60)
8<x
(x 2
+a
(5-3-61)
+y +
+ y and the noise power is 2a
2
(5-3-62)
<r )
(since a 2
137
occurs in both real and imaginary parts). Thus the mean of z is the total
power. The variance describes the statistical variation about this mean
value.
m
F
^~
=
=
=
z
z
Z2
4cr
t4:W(T
for z
0,
>
(5-3-63)
(5-3-64)
S
2<r
8<r
mal
(5-3-65)
(5-3-66)
(5-3-67)
Note that the standard deviation (square root of the variance) is equal to
the mean. This is a chi-squared distribution with two degrees of freedom.
The average of n records will normally have a gaussian probability
density (central-limit theorem) and a variance equal to the above value
divided by n. The mean is unchanged.
Zn)
TTZJm
<r(2^)
~H
[(
(5-3-68)
where
<*
X2
**)*
and
z
**.(*)
Tn
=
=
=
=
nM =
2<r
y
e~ 2(ff " )2 e~i2 ">
an
= x2
2
2r(l
n
4o-
(5-3-69)
2a 2
(5-3-70)
(x 2
y*
<r
+ r )^
+ 2r
(5-3-71)
(5-3-72)
where
(x 2
The
2
y )^
of each point in a
Fig. 5-21.
Note that
for large
amounts
138
power
of the
at each frequency
is
139
by Vn.
It is of considerable importance to determine the possible resolution of a
coherent spectral line buried in noise by using the power averaging technique.
Consider a random variable w = z n z no where z n0 is the random
,
2
variable for noise alone (x 2
y
and a = a n 2
a n0 2 = (4<r 2 /n)(x 2
Then a =
0).
2
y
the strength of the coherent spectral line,
while <rw 2 is the variance about this mean
Hence,
value.
gaussian
if
random
is
zn
f(t)
assumed to be a
F|G 5 _
z n0
x2
^h(t)
g(t)
G(s)
?(*)
2<r ).
"
H(s)
Generalized
fi
ter sit .
uation.
noise line
The concept
and Convolution.
domain
measurement situations
known and
is
in
desired.
one at a time.
1. F(s)
Find H(s).
This
Only those spectral lines that are common to F(s) and G(s) will occur in
Here G(s) can be any conventional analog filter function,
the output.
although
many
other
readily implemented
filters
by the
one frequency.
digital approach.
140
F(s)
2.
Find
G(s).
G(s)
3.
and
it will
Find F(s).
This
F(s)
is
waveform.
The problem
is
some values
of
zero of G(s).
It is necessary to differentiate
and
will
denoted by F, G, and H.
The basic problem can now be stated: What is the "best" estimate of
F (s) obtainable from the actual quotient H(s)/G(s)?
The question
what
to a noisy function
debatIn this
discussion, a linear mean-square estimate will be assumed, based on the
Wiener-Kolmogoroff theory. See Ref. 9, pages 400 to 405, for furof
constitutes a best
fit
ther treatment.
F(s)
The technique
H(s)/G(s)
is
to pass the
random
variable
is
through a "matched"
.
<t>(s)
F(s)F
denoted by
filter,
whose shape
<f>(s),
is
141
given by
* (a)
>i
(5-3-73)
\F(s)\*
where
mean and F
the
is
The
F(s).
* (s) is
F(s)
Thus the
|^ *(s)
The construction
is
best estimate of
given by
F(s)4>(s)
matched
(5-3-74)
function F(s).
Essentially
function in the
first
filter
is
place.
Assume that
with care.
it
part N(s),
= F ()
F(s)
N(s)
where N(s)
0.
|F(s)|
l^(s)|
Then
where iV 2 (s)
is
is
is
(5-3-76)
power
is
(5-3-77)
!*(,) |
lP(S){2
,
|F()|'
noise
+ iV'( s
signal
F(s)F*(s) =
*( 1
0(S)
(5-3-75)
(s)
N>(s)
|F(s)|
(5-3-78)
N*(s)
These equations
all
signal
F(s).
The
2
power
(s) will be considered next.
simply the variance of the random variable F, but F = H/G.
Thus it is necessary to calculate the variance of the quotient of two
random variables. For another discussion, see Ref. 9, pages 196 to 197.
The basic formula is
This
is
Pf(F)
where
J*^
and
are
dX
(5-3-79)
variables.
142
Although
F =
if
first
and
Fp F (F) dF =
f~n
|X
^Pg{\)
X
(5-3-80)
However,
\Fp(\F) d(\F)
f"x
Thus,
M
F =
/"
^p G (X)dX
(5-3-81)
In a similar manner,
T*
= H* ["
pG
(X)
dX
(5-3-82)
first
is
of
required.
as follows:
in(^
=-n(
a
\
Pg(G)
The
distribution
G >
a/2.
F =
F =
>
is
of
(5-3-83)
width
a,
r5+a/2 d\
Ja-a/2
aX
2H
>
(5-3-84
y
2G
+ a/2 dX
fG
H /,-
tanh- -=
a
2
^ - <s^W>
(3 - 3 - 85)
4H
= = ^=
(G
F
Assume
centered at G.
Then
a2
- Ha
a\
(
(tanh-i -=)
delta function.
Then
F(s)
constant.
is,
(s)
F2 F
2
.
(5-3-86)
2G/
F, that
f(t)
2
,
so
143
4>(si
\-e-&
FIG 5-23
function
that
The
is
>
(5-3-87)
G(s)
ke-Hi'"**
iovG
-ikl
let g(t)
Define
filter
Then,
noise free.
(*)
Now
matched
</>($).
-{0
is
given by
(5-3-88)
e"
'
for
/c
(5-3-89)
2fc\w
So
-(-!)
(5-3-90)
2fr
(5-3-91)
a
()
The graph
FIG 5-24
e<
s2 -V)/ft
for
of this function
is
|s|
<
(5-3-92)
so
shown
in Fig. 5-23.
(a) Output function: convolution of Gaussian "smearing" function with pair of delta functions, (b) Gaussian "smearing"
function.
144
FIG 5-25
(6) Fourier
The photographs,
Figs. 5-24
tions of
sients.
Many
other types of
FIG 5-26
noise
ratio,
(a)
Matched
filter
for
optimizing
quotient signal-to-
filter.
FIG 5-27
145
ing functions.
if
is
if
There is another type of filter that has some useful properties which will
be described in more detail. Consider the following Fourier transform
pair:
cos [2ts
YA
<->
H[5(s
+ TSoto
1
so)
7![2^
(So
where
\s\
<
8(s
(so
2
- s )]
- s*yn
(5-3-93)
s
)^
FIG 5-28
Imaginary part of
this quotient.
146
amount
of
ringing on transients.
pairs 871.2
and 619.
There are
FIG 5-30
(b)
Recon-
This
is
simply the
147
width of a rectangle whose area and height are the same as the waveform
The equivalent width of the assumed gaussian smearing
in question.
function
is
(5-3-94)
6(2tt)
which
is
sech 2tSqU
or
so
2fc\*
2fc\
cosh"
(5-3-95)
2irto
Ii(2irs
time domain
(5-3-96)
to
R =
is
irta
cosh 2wS
The
12A;
(/c
at the peak)
by the
is
obtained by
is
10 log
and
,
so
/In 10
\>
(5-3-97)
I1
o
IN
<
CO OS
IN IN OS
on
00 t^ IO N co :o as * 05
N * o i^ oo
* * CO
o o CO CO CO CO CD CO t^ t^
Oooo
no Tt< OS o t~ CO IO o r^
iO >o IN 00 (N iO r^ 00 Of)
CO
!N
CO > 00
oo
o on
o IO >o
<N CO co CO Tt< * T<
!-H
l-H
(X)
* CO
IN IO (N co oo 00 CO
ns (N (N i
o
CO * *
CO . as r~ 5
IN on * 00
iO
CO iO CD
iQ 00 o CO IO r^ Oi
<N CO CO
"* IO iO iO IO >o CO CO CO CO CO t^ r~ t^ t^
* * "f
O o oo o o
11
-*
O
^
i 1t
co
(N
r^
<N co CO
l-H
11
*<
t-
Of)
* 1^
<N
>o
OS
iO o * (N 00 lO
m
oo IN "* lO t^ OS
o CO
* *
CO CO l> t^ l> l> t^
o o ooo
OS
CO CO
(-> IN Tt<
r^
IN CO
CO
fl
Tt< 00 I CO o
(N Tt<
os
iO on
CO >o 00
iO iO iO iO CO CO CO
CO CO CO *
.-1
T}<
t-H
f-H
r^.
r^
CO
^-(
">*
>*<
(-ft
l-H
Tt<
_
as
(N * OS (N as T oo
CO
o r
on * 05 CO CD OS (N IN co <N <N
IN (N
o
1^ as
*
CO lO t^ OS
r^ (N iO r^ IN
oo
oo
co co * * iO >o iO CO CO CD CO t> t> l> t> t^ oo oo 00
oo o o
co
(->
ii
11
05
TjH
ft*
00
**
s
>
CO on on IO CO lO CO (N IN * OS r^ j^ as CO oo
IN IO IO < (N Ol o Tf r^ o CO * lO CD CO
CO IO
IN t^
*o 00 (N iO r^ as
Tf
on
* * 1iO *
00 oo 00 oo 00 as as OS
5 IO CO CO CO t^
ooo o
l-~
'S
r~
co
I>-
^
^
O
*
H
<*
^
o o o
*
Tt< CO h- on (N
o r- co CO oo oo
CO IO
00 Tf<
t^ * as Tfl 00 IN lO j^. as
CO CO (N
CO CO CO ai
CO
r^ as
IO OS CO CO
CO CO 05 IN * r^ as o<
oo
as as as
00
oo
00
as
*a lO IO CO CO t> t^ l> l>
rft
fO
fS
&2
CD
1>-
-tf
r-4
* as
(N t^ *
CD CO *
00 * as
* OS
CO CO as (N * i^ as
05 05 as
ooo
IO
00
CO
t>
(N l CO 00 Ol
CO o CD
o
1^
r^l> oo 00 00 Ol
1^ ^H
<** OS
iC
00 C~i l-ft
co >o r^ iO OS os
CO i >o os
>o lO CO CO co
o
*
CO as rH IN
CO i-H on CO OS os Tt< CD CO IO
co r^ r^ r^ CO
<N
IN 1^ 1
h- r^ CO CO 00 f~
t^
no Tt< OS
OS CO t-~ 1 kO 05 (N CO as (N kO 00
T-t
i IN IN (N CO
oo oo OS os OS
lO *o CO CO
o
cc
IN as r^ CO
00 Tt< (N OS OS 1^ CO h- CO "5
OS (N TfH <*< CO
r^ co oo IN CD
00 r^
CO (N
CO tOS co IN 00 CO 00 CO 00 (N 1^ I iO OS CO CD
iO
i <N IN <N CO CO
CO CO
00 00 OS OS
ii
o s
Tt<
(-) .<r>
11
*
* *
* m
o
*
O
ooo
3
IN
^
* *
*
o
oo
CO
lO
o o oo
(-ft
t^ CO 00 CO
kO CO
on t^ CO CO
11 <N
00 os
t^ no 00
IN
00
IN
r^
<N
^ * * o
*
00
t~ 00 i^ lO * CO
OS (N a> "* OS
iO oc
CO
CO
C CO CO Ol (N
iO OS CO 00
CO i-H CO
CO
CO OS o
"3 IO CD CO t^ > oo oo oo OS OS
CO CO
oo
"O
af
'ft
CD
<
148
c
o
oc
c
oo
o
oo
oo
o o IO o IO IO IO IO iO IO lO lO
o
rH <N
<N <N CO CO * -* o to CO CO t^ t^ oo oo as as o
1-i
149
-'^ r(cosh[(In
tabulation of
&>
l/rOAtcosh- 1 1/r)
10/10)
R -
(5-3-98)
In 3]*
and
is
Averager
acts like
comb
filter
for repetitive
150
Note that the width reduction varies slowly with respect to both ripple
and signal-to-noise ratio. Also see Fig. 5-32 for the computed transfer
function of a signal averager.
Summary
5-4
Although the basic theory of Fourier transforms and statistics has been
reviewed, the main emphasis in this chapter has been on various factors
The use
associated with the practical implementation of signal analysis.
of digital techniques allows a considerable
by
in accuracy
and a
improvement
Many
in signal amplitude,
and by
finite
time and abscissa resolution, and the system still has noise. These
limitations are imposed by nature and cannot be circumvented.
In the past few years, enough has been written on this general subject
to fill several volumes, so a single chapter must, by necessity, be rather
There are many important topics concerning signal analysis
sketchy.
which have not been mentioned here. For example, there is much more to be said about time windows and frequency line
Conditions for the stability of a power spectrum estimate, and
shapes.
The coherence functhe variance of this estimate, are very important.
coherence, has not
and
multiple
partial
of
tion, along with the concepts
even been mentioned [12]. The importance of trend removal from the
A technique
original time record prior to processing has been ignored.
of overlapping adjacent time records [13] is of fundamental importance in
gleaning the maximum amount of information from a given time record.
Digital filters are becoming very popular, and considerable literature can
by
digital techniques
considerable
[14].
amount
of
for per-
151
field.
APPENDIX
Amplitude-Spectrum Averaging
Define two independent gaussian random variables x and y, each with the
same variance a 2
Consider x and y as the real and imaginary parts of a frequency
spectrum. Define two new random variables
.
Vx + y*
2
y
W = tan" -
<
7T
< W <
<
oo
7T
where
dA = zdz
dot.
pza (z,0))
where u
/z
P: u (z,u)
= (1/V2
o-)(x
'
du = lot
a2
\
Pt u (z,w) dz
cos
1/
[1
\/x 2
and
is
zm
>
(l
so z
will
erf
w^e-a^x^+P)
and
co
be ignored.
/,,() is
^ p*(z)pM,
it
+ Vtt we"
sin w),
^-w \ e-<
-4-
e-J4(./o
<T
is [2
CO]
o-
is
of order zero.
Pz(z)
gin
w.
/"
P()
The marginal
co
spectral point.
(7r/2)]<r 2
=
Jo"
The moments
zm P*& dz
<j
\Ztt/2, and 2 2
2<r
2
.
152
This integral
'
where
iFi(a,6,)
is
and
function,
[10]
, nw(>.
Me-. ,
' ( --./,
"v
f
Jo
form
of the
is
2p"I>
+ -) (V
+ 1) \2pji
]
lFl
("
f)
4p*J
\2
is
the
gamma
/()
L,
(n!)T(a)
n=U
Thus
= (V2<0r(l
+ H.
i.
^-^r")
= el
= a
,F,(1,1,)
so
x2
is
,F 1 (2,l
>
g)
De
Wi^'d'
so
2<r
not so easy.
'*)
where
x2
2r
For
For
oo
>
xF,
+
2a 2
(|
1,
A-2
(l
-\
and
- V* + y + a
2
0,
iFiQ>l,A =
and
^l
(x 2
n(7<i)
0).
"2
52
x2
+y ~*+
2
(*
|)
'2
kja where A;
153
a coherent
100,000
10,000
1,000
100
FIG 5-33
is r).
154
36
3 2
2.8
1
i r itr;i.C)
f"
2.4
2.0
Coherent signal
/x*+y
1.6
o nly
<x
^A^
1.2
0.4
n
0.4
FIG 5-34
of coherent
0.8
1.2
Amplitude of noisy
component.
1.6
2.0
spectral line
2.4
compared
2.8
3.2
to amplitude
power averaging for the same resolution probability as long as n is large. The
two methods of averaging are equivalent for r = 1.092 (0.764 dB).
The interpretation of a power average is straightforward because the coherent
signal amplitude is equal to the difference between a dominant spectral line and
the noise base line.
The interpretation of an amplitude average is more difficult
and is best described graphically (see Fig. 5-34). The coherent amplitude is
asymptotically proportional to the amplitude of a spectral line for small amounts
of noise, but departs significantly from this proportionality as the noise increases.
Thus it is difficult to compare directly the coherent amplitudes of signals mixed
with noise. For these reasons the power averaging approach is generally preferred.
CITED REFERENCES
1.
2.
4.
5.
155
10.
11.
6.
7.
8.
9.
Papoulis,
1944.
U.S.
12.
13.
14.
15.
16.
17.
vol.
vol.
vol.
CHAPTER
SIX
MEASUREMENTS
Alan S. Bagley
Manager, Santa Clara Division
Hewlett-Packard Company, Santa Clara, California
The frequency
unit of time,
is
of a repetitive signal,
156
is
157
standard seems particularly important in time and frequency measurements, along with the precise comparison of the unknown with the
The development of good standards for both
reference that is chosen.
as well as historical or epochal time instants
interval
frequency and time
For many centuries, the rotation of the earth about its axis, viewed
with respect to the sun, was used to set up a uniform time scale. Astronomers gradually increased the precision of their observations and found
that the rotation of the earth was not really uniform. Even after they
applied every correction known to them, they found unpredictable
However, long after astronomers
irregularities and long-term drifts.
of these variations (until 1956) the second was defined as
part
of
an average rotation of the earth about its axis with the sun
3^ 6 ,4oo
as the reference direction.
During the past two decades, largely through the coordinating efforts of
became aware
The
International
definitions
timekeeping.
determination.
Ephemeris time is based upon the earth's orbital revolution around the
In 1956, The International Committee on Weights and Measures
.the second is the
defined the unit of ephemeris time as follows: ".
fraction 1/31,556,925.9747 of the tropical year for January 0, 1900 at
sun.
Time"
[1].
of January.)
The second
of ephemeris time
previously defined.
Theoretically
158
make
What was needed was an invariable time interval in nature that could
be observed and measured by available electronic instruments with
great precision.
This requirement, coupled with recent complex development in the electronic arts, led to the use of transitions that occur spontaneously and continually in energy states in atoms. In 1964, using the
atomic transitions, the International Committee adopted a new standard
that
is
in cesium 133.
The
and
is
definition
based upon a
the
F =
atom
hertz
is
of
4,
mF =
and F
is:
3,
is
mF =
fields
assigned.'
Atomic time is far more readily observable than ephemeris time, for an
electronic instrument with the capability of counting out the required
number of transitions in cesium vapor can be set up anywhere on earth
(or outer space).
While the atomic definition of the second has served well to give
accurate, immediate time scales and intervals, some procedures, such as
precise navigation and satellite tracking, require correlation with the rotation of the earth.
interval for
all
The mean
is
eccentricity.
for
of the
it still
has some
6-2
159
used.
The high-frequency
classes,
according
MHz, and
produced by it would be at intervals of exactly 1 sec. The comparison instrument can be an oscilloscope, the sweep being triggered by
the local ticks and the master ticks being applied to the vertical amplifier.
ticks
160
Antenna
Master
- sec ticks
station
"
Comparison
instruments
FIG 6-1
Local
frequency
Frequency
standard
and clock
divider
The advantage
of the
in frequency can be
method
measured
if
ticks
separately.
In the VLF range, the fact that very low carrier frequencies cannot be
modulated and demodulated accurately with sharp pulses leads to the
use of phase comparisons with the
employed
VLF
stations.
WWVB
60 kHz
preamplifier
60 kHz
amplifier
Signal level
reor panel
To recorder
(
phase comparison)
Phase comparison
{rear panel)
FIG 6-2
(Hewlett-Packard Co.
Model
1 1
7A)
VLF
comparator system.
161
in microseconds.
Again
WWVB
is
either read
The time derivative of the recorded output (the slope of the trace)
If the two 20-kHz derived signals
proportional to frequency difference.
signals, the full-scale output
60-kHz
than
the
are phase compared, rather
three times the time
denotes
obviously
comparator
phase
linear
of the
both.
is
interval
local standards.
The phase-locked 100-kHz output serves as a convenient local-standardfrequency generator in itself, even when no local oscillator of good
quality
is
available.
WWVB
.
is
WWVB
time.
WWVB
is
162
the day of the year, the hour, the minute, and the correction in
milliseconds to arrive at UT2.
6-3
This section will present some of the theoretical and practical aspects
of
fluctuations in frequency standards and discuss the measurement of
these
fluctuations, or noise.
The noise sources will be identified and evaluated.
While this material may appear to be highly specialized for the electronics
engineer, its comprehension at a technical and quantitative level is
essential if the engineer becomes involved in complex
communications systems,
precise navigation, space systems, or even the intelligent use of
time
and
frequency instrumentation.
Some
also presented.
The
Cutler
is
[5].
The
Analytical Treatment.
f(t)
A(t) cos
[o> t
signal
from an
oscillator
may be described by
*(0]
(6-3-1)
where /(0 represents a voltage or current, A(t) and *(0 are slowly varying
functions of time, and w is a constant,
A (t) is the amplitude of the signal
and is assumed not to contribute to frequency fluctuations. The time
origin and w are chosen so that $(t) has zero time average
and
l*(0l < C < oo for all time t, where C is some positive constant,
These
conditions simplify the mathematics (but will have to be relaxed later).
= d ^
r
dt
*^
""
is
*(')
(6-3-2)
=
=
lim
wo
T- f Z"(t)dt
1
*(T/2)
lim
r-*
=
Therefore,
The
wo
<()
-$(-!T/ 2)
T
(6-3-3)
is
163
where
(*,()>
r- 1
The phase
Now
/^
+
(t
0(0-
The
*'> *'
- *
and (0,(0)
signifies
is
<6 - 3 " 6)
A*T (0 = *
Let 6(0
r--
<*, -
co
is
(t
is
<Ot (0t)
consider that
* - (* ( +
i) *
('
- i))
is the
the autocorrelation function of the phase. Similarly, R Q (j)
Writing these both
autocorrelation function of the frequency departure.
are stationary in the wide
as functions of t only implies that * and
is
sense
[6].
*(co)
/_
7T-
B*(t) exp
/""
^*(T ) = 7T
icor)
dr
(6-3-8)
/-".. >S *^
J*
W exp ^ C0T ^
')
S*(co) cos
rfw
(6-3-9)
cot dco
power spectrum).
is
164
*(X)
is
<r.
(6 .3. 10 )
flto-o (1
[I /;
^ dr'J
(6-3-n)
difference,
{2[fl*(0)
R*{t)]}
(6-3-12)
<r<fi T (0>
r-M2[fl*(0)
/2,(T )]}>*
(6-3-13)
~~ =
(fir(O)
The
last
{2[fl#(0)
*M)) =
Also,
RaV)
[I /;
(l
fl # (r)]}
(6-3-14)
in terms of
R a (r):
d/]*
(6-3-15)
we have:
[fl#(0)]
= [t-
"
S*(co)
A,]"
(6-3-17)
[fl O
(0)]
= [t_1
" 2*(co)
H
da>]
(6-3-18)
two
identical
Osc.
ft J J Frequency
V\ (t)
multiplier
165
L_ *i
Phose
detector
V*(t)
Schmitt
V"z (t)
trigger
(or tunnel
(multiplier)
diode)
Vj(t)
Vlftfa Frequency
Osc.
Computer
multiplier
mult, per
L_*<L_J
overage
Analog
recorder and
digital
printout
FIG 6-3
Ai(t) cos [ Ul t
has output
$!(*)]
7,(0
= A 2 (t)
cos
[o> 2 t
(6-3-19)
*,($)]
In the frequency multipliers the amplitude changes get removed by limit(if we have good design, there can be little conversion of the
amplitude changes to phase changes). After multiplication, the signals
ing processes
V[(t)
2 (t)
A x cos
A 2 cos
n*i(0]
[nu 2 t -f
n$ 2 (t)]
[nu>it
(6-3-20)
It is well known that both the instantaneous phase and the average frequency are multiplied by the factor n (provided the multiplier has sufficient bandwidth to encompass the full spectrum of the nth harmonic).
The phase detector behaves as a multiplier. Its output is
(0
V[(t)V 2 (t)
+ Y2 A A
X
YA
2
is filtered
If the
two
AZ
out,
+ )i + n[*i(0 + *,(<)]
+ n[*i(t) *i(fl]} (6-3-21)
cos {n(i
cos {n(i
m)t
<o 2
which leaves only the difference frehave exactly the same statistics
signal sources
166
and
assumed to be
for $i(/)
Then
W2
<>
in
= Aw
(6-3-22)
= HAiAi
V' (t)
cos [n Aut
is
n*()]
(6-3-23)
where *() = Q^t) $ 2 (t). The Schmitt trigger gives a sharp pulse out
each time the signal crosses zero going in, say, the negative direction.
This occurs for t such that
n Aw*
y + 2rM
n$(<)
2 ir
where
is any integer.
Suppose the counter
and the gate opens at t such that
n Aut
The
n$(t
n Aw(i
t)
Subtracting the
n[Awr
3>(t
set to
count
periods
Y^tc
n$(t
first
is
r to
t)
make
V + 2ttN
2 ir
+ t) -
HU)] = 2irN
(6-3-24)
Let
t
2ttN
At
n Aw
to
to
At
2tN
(6-3-25)
n Aw
Then
*(<o
Since t
t)
*(<)
= Aw At
(6-3-26)
not constant, the time difference At between successive measurenot constant, but if Aw At
1 and &(t ) At
1, then only very
small error is caused by replacing $(t
r) with <i>(*
t ).
The process
of averaging over many measurements helps here.
The multiple-period
ments
is
is
+t
)
$(0 = A*r
(0-
There-
fore,
A*To (0 ~ Aw At
(6-3-27)
167
Aw <t(At)
Aw [m- J
=
rt 2
where
t is
the
ith.
- (m- J
=
1
measurement and
is
t*)
(6-3-28)
the total
number
of measure-
{100
this leads to
--
100
-12(2
100
2
r,-t)
,212
t=i
of the data
(y
VA,
40,602
99.99
The order
100
- [999,900 y n A,
X 10 L
1=1
must be preserved
100
^ Tii) (
t=i
TiV
/
1=1
1=1
y2
(6-3-29)
rt
)]
'
The other
may
and
(6-3-14),
The
(6-3-29) in
an obvious way.
be
esti-
Equation (6-3-29)
The subtraction
is difficult
to apply, but
it is
of the best straight line for the data involves the sub-
stitution of a variable t/
r,-
(a
-\-
hi)
sion
source.
The
From
The
A mechanical force applied properly to a quartz crystal produces an electrical charge; an alternating potential across the crystal
produces mechanical motion. This reciprocal relationship allows a sharp
mechanical resonance in a crystal to be viewed as if it were a high-Q electrical resonance when two electrodes are mounted on or near the crystal
oscillators.
'
168
9,180 MHz
90
X18
Osc.1
5 MHz
MHz
<
20
MHz
Mixer
Osc. 2
100
X20
MHz
X92
20-MHz
20
MHz
amp.
9200 MHz
MHz
20
MHz
Synthesizer
20 MHz
Phase
detector
*&f
<
'
Low-
Counter
frequency
wave
FIG 6-4
Af
mult, per
avg.
output and
analysis
printer
ana
Schmitt
analyzer
for spectral
trigger
log
surfaces.
[8, 9, 10].
in
The
vibration.
its
holder.
Figure 6-6
is
beyond
a representation
employed.
and
shunted by
(the
slightly,
169
oscillating frequency is simply the one at which the phase shift around the
feedback loop is zero. Even though the phase changes extremely
rapidly with frequency in the vicinity of mechanical resonance, varia-
tions in Co,
d, C
it
p.W,
and
It
is
variations in driving
circuits.
To be
complete, the circuit in Fig. 6-6 should show the output circuitry
factors.
The design
of
modern
electronic instruments
10- 9
System only
107 vs 107 ~
f~2~
""5
>
10"
>x\
f- Contribul
ion
System^
X
\
'
\N
'
\
X
10"
I0" 4
I0" 3
I0" 2
I0"1
10
t ,sec
FIG 6-5
I0
I0 2
is
as
demanding
170
of creativity, care,
in electronics.
Atomic Frequency Standards. The reader may not have studied quantum
mechanics, upon which atomic frequency standards are technically based,
but the presentation here will require only a few elementary principles.
The chapter by Alan Bagley on Frequency and Time Measurements in
"Handbuch
Vol. 23 of the
gives a
by S. Fltigge-Freiburg [11],
not requiring advanced knowl-
still
The frequency
where
is
constant and
v
v is
the frequency.
applied to an
is
is
In other words,
if
is
E =
hv,
Planck's
energy at frequency
is satisfied, one
relationship
Zi
Co
^i
C,
FIG 6-6
c4
c,
I
Equivalent circuit of a
in an oscillator.
quartz crystal
reason of
[11].
its
This transition
is
171
microwave
range,
Actually only a portion of the atoms are converted, but enough so that
they can be sorted by the magnet E and made to impinge upon a hot wire
that ionizes them.
The resulting free electrons strike the first dynode of
polarity of this error signal indicates whether the drive frequency is too
low or too high. After high-frequency components in the error signal are
removed by an integrating circuit, the signal is used to control and correct
been found to limit the atomic level transitions to the correct ones. Of
course, the time constant of the integrating circuit also determines the
Vacuum chamber
State
selector
magnet
Collimator
C field,---'
KV1
Cesium beam
L
Beam 4
current
Injection frequency
FIG 6-7
in
detector
172
Ampl. and
phase detector
It
Cesium-beam
resonator
z
Integrator
rf
Multiplier
and
Audio-frequency
oscillator
freq. modulator
Crystal
oscillator
FIG 6-8
beam
resonator.
its
good short-term
stability, is controlled.
The cesium-beam
A good portable
6-4
11)
and
digital instru-
connected
-LTLTLT
_L_r
FF
UT
FIG 6-9
f-n
1
,
UT
Ff
FF
UT
Out
FF
T input. Assuming that the statemost positive when "on," it follows that
toggled when the state of the previous one flips from 1 to 0.
applied to the
negative-going pulse
is
indicating outputs
and
each
7-
flip-flop is
initially,
are
173
results.
By means
flip-flops.
is
called a decade-counting
counter
is
ment panel
is
Decimal counters
4-0*-2
Main gate
Ampl.and
Input"
Freq.
counted
Schmitt
Sensitivity
Start-stop signal
(manual control)
FIG 6-10
Manually controlled
totalizing counter.
174
The
voltage, but generally a quartz-crystal oscillator is used.
required quality of the quartz resonator and the sophistication of the
constant-temperature oven and associated electronic circuitry depend
line
Short-term stabilities
specifications of the instrument.
10
than 5 parts in 10 11 and aging rates better than 5 parts in 10
per day are available in commercial counters. The exact frequency of
time-base oscillators can be set to an external house-frequency standard
in
most instruments
of quality.
figure,
if
precisely lO" 5
of
10~ 4 lO" 3
,
lO""
2
,
is
10-\
MHz,
1,
or 10 sec.
quency
The input
input signal
signal
(CW
is first
is
by the
and
is
retained until a
new sample
is
taken.
The sample-rate
control
and
The time-base
Main gate
Ampl. and
Schmitt
Input"
Freq. counted
Sensitivity
Open
Close
Internal
Schmitt
time base
xtal osc.
trigger
FIG 6-1
made
cycle.
175
Decimal counters
2
14
+-
Mom
Freq
counted
Ampl a
Input"
gate
Schmitt
Sensitivity
_J Open
Close
Internal
Schmitt
time base
trigger
xtal osc
6
FIG 6-12
The input-shaping circuit commonly selects the positivegoing zero axis crossing of successive cycles as trigger points for opening
assemblies.
gate.
As in frequency counting, the measurement is
automatically repeated at a rate that is manually variable in commercial
instruments, or the measurement may be made on a one-shot basis.
measurement
of 100
Hz on
kHz.
single period
MHz
as the
counted frequency, would be displayed as 0010000.0 jusec. Thus, resolution is increased by a factor of 100.
The accuracy here is also affected by
176
the
1-count ambiguity,
time-base accuracy,
the
the
trigger
error.
Multiple-period
and the
The
Averaging.
effects of the
1-count
ambiguity
by using multiple-
/xsec for
a 100-period average.
The
function
show the
The
Ratio Measurements.
ratio of
two frequencies
is
determined by
using the lower-frequency signal for gate control while the higher-frequency signal is counted, as shown in Fig. 6-14. With proper transducers,
ratio
represented
Decimal counters
5r*-|7r*-|l
|H6r*H9
3 *i
Main gate
Freq. counted
Ampl. and
Schmitt
Input,
-<^o
Sensitivity
Open
Close
time base
xtalosc.
FIG 6-1 3
P U
A*
-**
*-+
|)0|
|lp|
Arrangement
|I0|
for multiple
! J
1*
-L*
|IO|
|IO|
[lO|
|IO|
period averaging
Schmitt
trigger
Deci
177
4-
*
|
Main gate
Higher
*
frequency ^>
,n P u+
Sensitivity
yvv
Ampl. ond
Schmitt
Freq counted
.
a^c
1
Open
Close
Lower
Ampl. and
Schmitt
input
FIG 6-14
quencies.
well as frequency divider or multiplier operation, are some of the measurements which can be made by using this technique.
Accuracy is + 1 count + trigger error. The accuracy may be improved
by using the multiple-period averaging technique by counting the higher
frequency for 10" cycles of the lower frequency.
With a preset counter or a counter with a preset
Rate Measurements.
plug-in, frequency measurements can be normalized automatically to
be set to a gate time of 600 msec to cause an input from a 100 pulse per
revolution tachometer to be displayed directly in revolutions per minute.
High-frequency Measurements. Accurate high-frequency measurements
can be made above the normal range of an electronic counter by using
heterodyne converters, transfer oscillators, or automatic dividers, and for
frequencies up to 500
bilities of
each
will
unknown
frequency.
Transfer oscillators, on the other hand, are more versatile. They can
measure
or pulsed signals, as well as
signals, over a very wide
frequency range and can produce AT-hertz resolution in 1-sec counter gate
FM
CW
178
time, where
pulsed signals.
In operation, the transfer oscillator generates a variable frequency,
which is adjusted so a harmonic of that frequency's zero beats with the
unknown
CW signal
The
is
Measurements to
the nearest subharmonics has been manually tuned.
15 GHz are possible with the 2590B model, and to 40 GHz with the
what misleading unless
it is
The
CW
Unknown
input 3-12.4
GHz-
Balanced
Tuning
cavity
mixer
Selected
harmonic
All
harmonics
200 MHz
of
200 MHz
from 3
steps
to I2.4
1-212
GHz
MHz
Video
Harmonic
generator
amplifier
Tuning
meter
1-212
'
MHz
'
0.25-53MHZ
lOMHz
signal
Prescaler
Multiplier
U1
(X20)
Gate time
extender
(X4)
FIG 6-1 5
Heterodyne scheme
to counter
input
From time
-To
base
counter gate
'
179
HP 540B
Unknown
input
Mixer
Video
amplifier
~~sr
1
i
1
Harmonic
Zero beat
generator
detector
(scope)
HP 2590B
1
To electronic
counter input
MHz
Amplifier
1
"
Variable
freq. osc.
(
local
osc.)
Voltage
Phase
control
detector
i
Unknown
Unknown
/Vx
/j
for
2590
X3
540B
30 MHz
:l
1
10
FIG 6-16
quencies.
Use of
M Hz
fro m coun- er
upon the
6-5
scale factor.
Frequency Synthesizers
A great deal of the information in this section was excerpted, with permission from the publisher, from a paper by V. E. Van Duzer [15]. It
could be argued that a frequency synthesizer should be classified as a
signal source rather than as a time-and-frequency instrument, but the
emphasis in synthesizers is on precision and stability as much as on variability.
The instrument, in the best designs, is essentially a variable
frequency standard.
180
numerous timing
many
situations,
spectrum analysis,
stability studies,
and
other areas.
dard
[16].
The
mitting fine resolution and fast switching in the same instrument and
fail-safe
This
is
and all integral multiples of 1 MHz from 1 to 10. The 0.1-MHz signal
and the 10-MHz signal were also multiplied by integers from 1 to 10.
Now, if 23.6 MHz were desired, the 20-, the 3-, and the 0.6-MHz signals
could be applied to suitable modulators to give the sideband
23.6
sum
of
MHz.
is
pure,
and
unwanted sidebands
besides,
simple, but none of the signals menmodulators (or mixers) always produce
lem, especially
spaced closely
unwanted ones
is
a serious prob-
impractical to
filter
181
to describe a modern synthewith the capability of providing an extremely large number of frequencies and high output quality.
sizer
The
design objective
is
to
number
of switchable
a total
is
is
Hz
to 50
MHz
in digital incre-
At
Any nonharmonic
remote electronic control. When the frequencies are electrically programmed, switching can be accomplished in less than 1 msec.
A simplified block diagram of the overall instrument is shown in Fig.
6-18.
The driver contains a frequency standard, a spectrum generator,
and appropriate selection networks to provide a series of fixed frequencies
FIG 6-17
MHz
in
182
Driver unit
Frequency
standard
1
"
Int.
Comb
55oT
3
fiz
i'38
24
3.0
MHz
MHz
133
Switch
Switch
24 MHz
f~;i~"l
350-390
30-39
MHz
MHz
2700-
330
MHz
27.09
MHz
0-50
390-400
MHz
MHz
Output
360-361
MHz
3.0+
3.9
MHz
.._J
3009MHz
MHz
30.000,000,0030.999,999,99
MHz
30.0.
30.91
30.00
3.000,000,00-
3.009,999,99
Unit8
Units
3.000-/
3.099 MHz
for the
Un it 2
(\
)
/
Unit
3003 09 MHz
frequency synthesizer
between
frequencies.
The
is
particularly interesting
As shown in
also serves to illustrate
the right-hand portion of Fig. 6-18, there are seven identical mixerdivider units, each of which corresponds to a place, or decimal position,
In each of these units, and in the
in the final output frequency number.
and
the method
of synthesis used.
MHz
is
shown.
In the right-hand unit, which produces what ultimately becomes the
highest resolution digit (lO^ 2 Hz), the 24-MHz carrier is added to a
3.0-MHz frequency in frequency adder A to produce 27.0 MHz. In B
183
repeats.
If
the process
is
followed through,
it
will
fre-
quencies noted in the block diagram are obtained at the outputs of the
various adders and dividers.
In essence, each mixer-divider unit, through
a frequency-division process, moves a given digit one place to the right
methods are
some
extent,
of frequencies
and mixing
possible.
Applications of Synthesizers.
If in
we
184
The
installation.
arrangement
can
provide
for
phase-locking
the
sequencing cycle.
The
effective
known and
selectable increments.
is
specific frequencies
CITED REFERENCES
1.
U.S. National Bureau of Standards: Proc. IRE, vol. 48, pp. 105-106, January,
1960.
2.
3.
Hartke, D.: A VLF Comparator for Relating Local Frequency to U.S. Standards,
Hewlett-Packard J., vol. 16. no. 2, October, 1964.
S. Fliigge-Freiburg, ed.: Encyclopedia of Physics, vol. 23, pp. 289-372, SpringerVerlag
4.
OHG,
Berlin, 1966.
52,
Some Aspects
of the
of
Frequency
5.
Cutler, L. S.:
6.
7. Ibid.,
8.
4 and
chap.
5,
New York,
1958.
6.
Cady, Walter G.: "Piezoelectricity; an Introduction to the Theory and Applications of Electromechanical
New York,
1964.
Phenomena
in Crystals,"
Dover Publications,
Inc.,
Bottom,
Virgil E.:
of
of
10.
Hammond, D. C,
C.
Adams, and
185
111.,
1968.
July-August, 1963.
Flugge-Freiburg, op. cit., p. 321.
12. Davidovits, P.: An Optically Pumped
p. 29,
11. S.
Rb 87 Maser Oscillator, Proc. Symp. Definiand Measurement of Short-term Frequency Stability, Goddard Space Flight
Center, Greenbelt, Md., Nov. 23-24, 1964, NASA SP-80, p. 171, 1965.
Barnes, J. A., D. W. Allan, and A. E. Wainwright: IRE Trans. Inslr., I-II, vol.
tion
13.
26, 1962.
14.
15.
16.
For example, Operating and Service Manual Electronic Counter 5233L, HewlettPackard Co., Palo Alto, Calif.
Van Duzer, Victor E. A 0-50 Mc Frequency Synthesizer with Excellent Stability,
Fast Switching, and Time Resolution, Hewlett-Packard J., vol. 15, no. 9, May, 1964.
Hughes, R. J., and R. J. Sacha: The LOHAP Frequency Synthesizer, Frequency,
:
August, 1968.
of the Theory and Measurement of Frequency
Fluctuations in Frequency Standards, Proc. IEEE, vol. 54, no. 2, February, 1966.
L.
S.
Some Aspects
CHAPTER SEVEN
DIRECT-CURRENT
INSTRUMENT AMPLIFIERS
From Notes by
Paul Baird,
Bill
and Richard
Y.
Moss,
II
ments that
this
for
making measurements on
amplifiers
Chap.
by means
along with a
short discussion of the characteristics that need to be measured.
Whether they are in instruments or not, amplifiers are employed for
13,
one or more of the following reasons: (1) to increase the power available
an electrical signal, (2) to amplify voltage or current levels where
in
power per
se
is
(3)
186
187
source impedance for a load, and (6) to attenuate or reject the commonmode component of voltage on a pair of conductors (the common-mode
voltage is the average of the voltages on the two conductors at each instant
item 4.
In instruments, dc amplifiers are most commonly used for reasons 2
and 5 above. When a signal is very small, as a voltage, as a current, or
Whether one needs a high voltage gain
as both, one needs amplification.
E L /E a high current gain I L /L, a high transconductance I L /E 8 or a high
,
transimpedance E L /I, depends upon the natures of the signal source and
load to be driven. The simple Thevenin and Norton equivalent circuits
in Fig. 7-1, using an amplifier with a common ground terminal, will help
The symbols are defined on the figure. If a
to clarify this situation.
signal is most accurately described by its Thevenin equivalent voltage E 8
even when Z, is variable, then Zi of the amplifier should be high compared
with Z,.
If
the signal
is
Z
determine the desirability of Z ^> Z L or Zl Z
less of variations in
Z then one
desires Zi
as a current / regard-
Similar considerations
t.
So far, the situation seems very simple, for shunt or series feedback can
be used to modify Zi and Z over wide limits while stabilizing the desired
The common problem, however, is that dc signal sources
transfer gain.
can be at such low levels that amplifier noise and drift in operating points
degrade the accuracy of measurement. This chapter describes some of
the methods used to deal with drift and noise while simultaneously achievThe
ing the desired gains, bandwidths, and impedance relationships.
important
amplifiers
is
instrument
of
dc
design
in
the
of
bandwidth
role
the bandwidth is of course never zero, since a signal must vary to contain
intelligence.
Zs
r^ +-s
El
(a)
z,-+
FIG
7-1
f
(b)
188
amplifier extend
of
an
The
resulting amplifier is said to be direct coupled and besides being straightforward in concept, it can have essentially the same upper frequency
range as one with capacitor coupling. An example of a direct-coupled
amplifier
be
will
given
in
Sec.
7-1
to
show some
of the
design
considerations.
random
applications
it
is
of direct coupling.
7-1
The most common means of amplifying dc voltages, excluding amplification of extremely low level signals of the order of 1 ^V, is the direct-coupled
amplifier.
The amplifier can be either inverting (operational) or
noninverting.
If, in addition to the requirement of accurate low-level amplification,
the amplifier is required to have moderate bandwidth (20 kHz at a gain of
40 dB), wide dynamic range (0 to + 15V) at the input, extremely high
input resistance (10 10 fi), very low offset voltage and current (< 1 /xV and
1 pA at the amplifier input) and remain relatively insensitive to its environ-
means
means
The
ously listed.
189
this case
was actually a
Basic Configuration.
The
analog-to-digital converter in
digital voltmeter.
Differentiol
A",
^^sff
amplifier
171
Az
El
Eo
FIG 7-2
direct-coupled amplifier.
benefit
voltage gain,
The
definition
common
is
inputs
Ki
Input Stages.
Ki
A
Ki
Ki
matched pair
of field-effect transistors
(FETs) was
chosen as the differential input stage. The FET offered both the high
input resistance and the small leakage current required. Bipolar devices,
though not necessarily limited by their lower input resistance since their
input resistance is boosted by the amplifier feedback, have considerably
more leakage current the transistor base current is typically 100 to 1,000
times more than the FET gate leakage current.
The FETs are operated in a balanced common-drain configuration to
achieve minimum sensitivities of the input offset voltage to power supply,
190
device parameter, and temperature variation (the input offset AVGS is the
required difference in voltages at the input terminals required to give zero
differential output).
Figure 7-3 is a simplified diagram of the circuit
To
arrangement.
by temperature
temperature controlled at a temperature higher than the maximum expected ambient. Although this
does increase the gate leakage current, it also keeps it constant.
Minimum sensitivity of gain to device parameter variation is also
further reduce offset variations caused
FET
fluctuations, the
environment
is
achieved.
to track or bootstrap the input voltconstant with input level. If this is not done, the
offset voltage A Fes will change as a function of level because of the changing FET bias.
The back-biased breakdown (Zener) diode in Fig. 7-3 acts
V DS
ages to maintain
as
if
and
Q2B and
ages are very small, VDS for both FETs is kept nearly constant, even
though a large common mode input may exist.
Current sources for the active devices are used to reduce the effects
of varying power-supply voltages and bias the FETs in such a way as to
keep V DS and I D s fixed while VG s varies. Actually, the current sources
in the amplifier
of resistance could
Noninver+ing
Ni
input
S,7*
c
,
FIG 7-3
v6 ,.
s,
191
is
of course,
If
These
is
made to
The
is
must be allowed
Kx
K' = Eo/Ei =
1
where K\
K<l
j8
The gain
its
+ Kd
(7-1-1)
Krf\
= noninverting gain
= inverting gain
= R/(Ri + R t )
is
set accurately
by adjusting
/3.
192
If
AK
Ktf
and
^ AKi/Kj
to change,
K-S
Ktf AK!
AK
Krf
(7-1-2)
2)
AK /K
2
_ AK^
(7-1-3)
2,
AKJ_
K
K
(Aid
AK!
Ki
2 (i
AKt/Ki =
AK
where
and
1,
AK'
if
+ K#)
Xi(l
is
+ K $) AK - Ktf AK
(1 + KtfY
(1
K
If
(7-1-4)
K K0
=
2.
VRL
*l
^
.HI
>*<
vN
C\
*!
Noninverting U--
I
I
/?,
ij-
Q2A
input
Inverting
input
Q2BN
-U
i03
_i
^OV^
*
<>
Q4
:<T/? 2
FIG 7-4
Gain
stages.
To output
drivers
than unity.
less
shown
193
Observe that
clarity.
make
fashion to
complementary
at
___
a/X
^V
Y"
'
F|Q ? 5
Simp| fied
at
common
papers [1].
Noise considerations
(and
the
age of
Q2 and
limits
on power-supply voltages)
books and
and assuming no
correlation,
given
EN
by
E N = E NQ2 +
2
>
(^f^)
RL
r'eQ2
(Wr'eQ2) 2
(Fig. 7-4),
and where
= " 2 Q*
(7-1-5)
r'e
and
r'c
are the
2?"eQ3 r Q2
where
RL
<3C 2/S Q3 r^ Q3
and
= a/ (a
1).
r'eQ2 is
V* L
CRl
7^Q2.Rl
(7-1-7)
r Q2
and
r Q2
where
= CRl = C Rl
~v7L
C =
lEQireQ i-
(7-1-8)
194
Therefore,
En 2 = ENQ2
2
(I NQ3 C'R L
(71_ 9 )
and
(7-1-10)
The
RL
decreases,
r'eQ2
The RC pairs (R x and C h R 2 and C 2 ) shown in Fig. 7-4 are used to shape
the amplifier frequency response.
The size of capacitor used is primarily
limited by the slewing rate required.
The slew limit is the maximum rate
of
change of signal de/dt ma which the amplifier can follow linearly. Slew
when the current required to produce a voltage across a
limiting occurs
2,
and
The transition from a balanced or differential configuration to a singleended one is effected through Q4. Such a stage is necessary to achieve
both the high gain and the power-supply rejection required. Diode CR
2
Overvoltage Protection. In any measuring environment in which measurements are not rigorously controlled and monitored, the possibility
exists of damaging the amplifier through input overvoltage.
To prevent
damage to the amplifier (and perhaps to the signal source), dynamic
protection must be provided. See Fig. 7-7.
195
Signal from
gain stages
Load
FIG 7-6
Output
stases.
will
the dynamic range of the amplifier is exceeded, the diodes
the
with
series
in
resistor
the
by
provided
is
limiting
conduct. Current
The top pair of diodes do not significantly degrade the input
When
input.
may
necessary
the amplifier output.) Such open-loop gain measurements are
response and the shaping
if accurate determination of the frequency
needed to ensure
Protection
resistor
Overload
current
stability are to
be made.
Noninverting
'
,n Put
Invert
input
Breakdown
FIG 7-7
diodes
196
FIG 7-8
Within the assumptions noted, the method shown above (Fig. 7-8)
IS
both accurate and relatively effortless.
Let Ki =
The transfer from the amplifier input to its output
2 - K.
is
given
by
Z4
E = -KE = -K
3
Zz
+ Zt
(7-1-11)
-E
Z2
then
it
and
E ~E
Z\
(7-1-12)
2
(7-1-13)
Adding Eqs.
(7-1-11)
= +
E
Z
2
Let
Z = Z
x
2}
Ej _
E ~
100
(Z,
and
+
^ KZ
(7-1-13),
ZlZ *
2
(Z 3
+ Z )/Z
4
(7-1-14)
+ ZJ
100,
and
200.
Then
(7-1-15)
and
= 100'
(7-1-16)
197
is
is
reduced in
effect
by a
factor of
+ PK by the feedback.
Typical operating specifications for the amplifier as previously
Results.
XI
to
"is
shown
in Fig. 7-10.
X100
15
TTtiS
BANDWIDTH, 3 dB
to 20
to 1
7-2
very low
drift,
overall response
when feedback
is
used.
(a)
FIG 7-9
Illustration of
198
+30V-
Noninverting
sinput
h> Output
Current comp.
network
FIG 7-10
Complete
Q2 current
source
FET
current
sources
Single-ended
Output drivers
transition
amplifier diagram.
narrow bandwidth and a rather complicated transfer funccan be paralleled by an ac amplifier to extend the bandwidth,
but when feedback is then used to obtain a flat frequency response, some
severe problems appear in stability, transient and overload response, and
intermodulation with the chopping frequency.
tion
To
It
leakage currents, and (3) that the voltage offsets of the amplifier are
determined by the chopper itself plus the offsets generated by thermal
junctions in the input circuitry.
However, the development of FETs
with little low-frequency noise (10 nV/Hz w at 10 Hz) and also low leakage
current (10 pA) has changed this situation to some extent.
The Basic Circuits for Automatic Reset. United States patent number
2,994,825 describes means for automatically resetting the operating point
Direct-current Instrument Amplifiers
199
Photoconductor
^Light pulses
FIG 7-11
Basic
automatic-reset
amplifier.
measurement
is
on a time-sharing
The
basis.
t
sec while
the amplifier operating conditions are quickly reset to make the output
when the input is zero. In other words, the drift is balanced
voltage zero
out.
In a sense, the amplifier is no longer truly direct coupled, since the signal
briefly interrupted, but in some applications this is no disadvantage.
The original circuit, simplified, is shown in Fig. 7-11. While the input
voltage is being amplified, with switch S, in the upper position, the photois
amplifier.
The
reset, the conditions shown in Fig. 7-12 are approached.
the
input
across
required
be
would
that
(the
voltage
voltage
offset
Eos
terminals to bring the output to zero) is represented by an equivalent
battery, and the photoconductor is illuminated to have a resistance R of
During
about 1,000
If
ft.
Eos+
5,
K
Eo
+
Eos'Ee
'
T
FIG 7-12
state
conditions) during
interval.
the reset
initial
charge on
are
200
BcW =
(1
as a function of
s,
is
which gives
ec(t)
= Eos t^-zz
(1
e u+*>'/s<0
(7-2-2)
as a function of time.
If
}> 1, the voltage across C approaches E s
exponentially with a time constant of RC/K. After a few reset cycles,
the offset voltage is almost perfectly balanced out by the voltage across
C,
which
Since
is
sudden
high,
amplifier saturation.
Now C responds only to the low-frequency components of input offset if it responds to input noise and the switch opens
at times that are arbitrary with respect to the noise variations, then the
voltage to which C charges has a random variation.
Instead of the
photoconductor, a fast reed switch S3 is commonly used to give a very
low switch resistance and lower switch offset voltage. During the
amplifying phase of operation, S 3 is open and Si is closed, which estab;
feedback through
with the purpose of C.
lishes
The
and
Fig. 7-1,
FIG 7-13
fier in reset
Modified
mode.
reset ampli-
201
characteristics
Noise
by slew
and small
1,
Vi
signal levels)
10 V) (limited
rate)
Offset drift
<
Leakage currents
Dynamic range
<
25
<
0.2
mV/C (-20
to
+55C)
pA
(input voltage
= 15 V)
(gain
1)
of level
signal,
amplifier
is
7-3
Differential Amplifiers
Figure 7-14 shows an amplifier with two input terminals, neither one of
which is connected to the system ground. If this is a differential amplifier,
the useful (or differential) gain is
K=
__^-
A(Ei
(7.3.1)
Ei)
202
FIG 7-14
Elementary
differential
am-
plifier.
is
is
\.
shown
However, the
in
common-mode
signal.
mode
voltages.
differential configuration is
One
possible approach
is
shown
in Fig. 7-15.
In this
circuit,
Ah A
2,
and
Kh
shown by the minus signs, and the corresponding voltK 2 and K which are all assumed to be of the
,
with
It is
KEi
a (l
203
K)
(7-3-3)
+R L
Ei
(7-3-4)
Rn
balanced.
is
K ~
=
Ei
The
of
= IRf, and
(7-3-5)
Rg
is
shown
in Fig. 7-16.
This con-
figuration uses a total of four amplifier blocks rather than three, but the
requirements on each are not as severe, making the circuit more compatible
FIG 7-16
Improved
circuit blocks.
204
(2fl t
*x
+ R )Ej
(7-3-6)
(2R 1
+R
)R t
While gain and bandwidth now depend upon a larger number of components, reasonable switching ease and stability are not hard to achieve.
The function of A 4 is to reconstruct the common-mode voltage by means
summing resistors R 5 and to drive the power common and guard shields
Ai and A 2 to prevent deterioration of the intrinsic CMR (commonmode rejection) by stray capacitances. In the configuration of Fig. 7-15,
the CMR depends primarily upon the minimum values of the open-loop
of
of
achieve lower
7-15 circuit.
CMR
CMR
Remember
CMR
is
Kcm/K
d.
amplifiers
chapter.
differential
10 V
mode
voltage to about
7-4
Chopper Amplifiers
in differential amplifiers.
(3)
a slow variation of the output when the input is constant (drift), and
an unpredictable current in the input leads (leakage current). One
205
or else just fc
understanding,
Chopper switches,
Eo Wc)
FIG
7-1 7
basic
chopper
amplifier.
Chopper Switches.
switch
or modulating
However, a
a very highly developed device that minimizes several
severe problems.
First, consider the choice of chopping frequency fc
For two reasons it is desirable for/c to be as high as possible: (1) The bandwidth of the dc system is seldom greater than 0.1/c because of demodulais
good chopper
is
tion
and
and
many
filtering,
^/
can
exist
(2)
[4].
appear as offset and noise after demodulation and can even cause currents
to flow back into the source of the dc signal.
Jitter in the switching cycle
and bouncing
Some
thermal
drift
about 0.2
/xV
ohms
in
of
of
internal resistance.
shown in
Fig. 7-18.
driven at line
206
H(/
~E
1
1
tion.
'
Chopper
ence and harmonics thereof, in either the input circuit or the ac amplifier,
to produce an unpredictable dc output.
One improvement in the chopper amplifier for use with low-impedance
system
sources and high common-mode voltages is to employ a full-wave
rather than the half-wave design described above. Such a configuration
this function;
(Fig. 7-19) commonly makes use of a transformer to perform
cycle, the
one-half
for
ground
to
signal
instead of connecting the input
results in
The disadvantages
former design and shielding problems; frequency limitations and if the isolosses as a
lation feature is used, the dependence upon the transformer
to
limiting factor in accuracy since there cannot be any overall feedback
and
windings
transformer
of
set
another
the isolated input except through
;
choppers.
used as
Other Choppers. Several nonmechanical devices are widely
amplichopper
in
signals
ac
into
signals
modulating switches to convert dc
in
successfully
used
first
were
(photoconductors)
Photoresistors
fiers.
S07
early devices
had
and were
chopper circuits at the microvolt
level.
carrier or
chopping frequency of 50
Hz
is
described above.
Direct-current amplifiers using higher chopping frequencies have been designed since 1958, but faster photoconductors having
lower resistance
improvement
is
with illumination
in these amplifiers.
positive or negative.
On
is
driven
negative, a current flows from emitter to base, and the emitter becomes a
fraction of a volt positive with respect to base.
Now, a collector current
will flow
if
the input
collector current
is
made
As long
as the
is
low.
in the
FIG 7-20
versus
curve,
Transistor used as a
and the
offset drifts
with temperature.
chopper
switch.
The
offset error is
Even
transistors in a circuit in
which their
offsets
is
tend to
208
*Lk Eo
FIG 7-21
Bilateral transistor
chopper.
The FET has virtue as a chopping switch because of its low offset voltage and low driving power. Some of the characteristics of chopping
devices are compared in Table 7-1.
Another modulating device that is used in amplifiers requiring extremely
high input impedance is the vibrating capacitor, the capacitance of which
is made to vary periodically by mechanically oscillating the position of one
The input voltage is applied to the capacitor through a
of its plates [5].
high resistance so that the charge remains nearly constant over a cycle of
The ac output voltage is proportional to charge multiplied by
-16
A can be resolved.
variation in capacitance, and currents as low as 10
choppers.
low-impedance
However, drift and noise are greater than in most
vibration.
Although
currents,
TABLE
this
it is
7-1
Characteristics of
Device
Mechanical
3
Photoresistive
Inverted Qt
Integrated double
Qt
FET
t
Inverted transistor.
Two
transistors
Jmt
Vos,
Ios,
ROFF,
VoFFt
Hz
10 s
10 s
10 s
io-
io-
10 -io
10" 1
10
10""
10*
10 6
30
50
io-
10
i<r 4
io-
10~ 8
25
50
10'
io-
io-
10 l
on one integrated-circuit
chip.
10*
10 8
10 8
20
'20
209
CITED REFERENCES
1.
2.
New York,
1956.
3.
Cage,
4.
J.
April, 1960.
5.
Caldecourt, V.
J.:
Apr.
6,
1962.
CHAPTER EIGHT
MEASUREMENTS
From Notes by
Paul Baird, Arndt B. Bergh, Robert
Dudley,
William McCullough
Hewlett-Packard Company
and by
Charles O. Forge
Durrum Instruments
Palo Alto, California
Along with impedance and power, the basic variables measured electronChapter 1 has discussed the system of
ically are voltage and current.
It was shown that
units for these quantities and their standardization.
standardization requires extreme attention to accuracy and therefore
While these techniques
special measurement and comparison techniques.
are basic and essential to the electronic arts, they will not be treated
further in this book.
This chapter
and current
dc
211
They
Introduction to
DVMs
Digital voltmeters are measuring instruments that convert analog voltage signals into digital presentations. The digital presentation can take
DVM
many
others.
The common element in all these signal conditioners is a
dc output voltage proportional to the level of the unknown quantity being
measured. This dc output is then measured by the DVM, and appropriate annunciation in the digital presentation indicates the quantity being
measured.
Digital voltmeters have made significant contributions in the field of
electronic instrumentation since their introduction more than 15 years ago
and are enjoying a growing popularity as bench instruments and systems
components. Among the many factors that distinguish DVMs from
other voltage measuring instruments are speed, automatic operation, and
programability. Although there are instruments such as precisian differential voltmeters that can be more accurate than most DVMs, few
instruments bring to bear on the measurement problem the same combina-
DVM.
DVM
212
2.
Number
Number
of
They
differ in
the following
measurement ranges
of digits
4.
Accuracy
Speed of reading
3.
DVMs.
5.
Normal-mode
6.
Common-mode
7.
noise rejection
noise rejection
DVM
Noise rejection
this section.
is
Normal-mode
input filtering or through the use of the integration technique. Common-mode rejection is achieved through the use of guarding or other
ways of obtaining differential response. Digital outputs are usually in the
form of
four-line
BCD
serial outputs.
single distinction
between types of
DVMs is the
method used in converting the analog dc signal into a digital presentaDuring the short time that DVMs have existed, many different
tion.
techniques have been developed by which this conversion takes place.
213
In general, these techniques can be divided into two main types integratand nonintegrating. The main nonintegrating types of
:
DVM
ing
techniques are:
Potentiometric
Servo
Successive approximation
Null balance
Ramp
Linear
Staircase
in nature are
Voltage-to-frequency converter
Potentiometric to integrating
Dual slope
8-2
Nonintegrating Types of
Stripped of
Potentiometric.
potentiometric technique
is
DVMs
all
Figure 8-1 shows this arrangement in its simplest form. The linear
is adjusted until the null indicator shows equality of the input
voltage and the output voltage of the divider. Assume that the internal
divider
reference
Vt
is
equal to
+10 V and
balance.
the divider
is
at the input
equal to
The range
(8-2-1)
measured
in this fashion is
dependent
Linear
divider
Vi
J?
'k
Inte
refe
N ull
in dicoti on
"
FIG
8-1
Simplified differentialvoltmeter block diagram.
214
Servo
Ampl.
9
Internal
reference
Linear
divider
FIG 8-2
Simplified
gram of servo
block
dia-
DVM.
+7.41
'|
Readout
of the
measurement
If it
and the
linear divider.
The
ity of the
and the
drift
divider
is
three digits.
DVM
215
converter and the servo motor is replaced with electronic logic. The
readout in the successive-approximation technique is electronic and is
usually driven by the logic of the system.
Figure 8-3 shows a simplified block diagram of a successive-approximation technique.
This system
is
fol-
10.00 V.
example
is
The sequence
8 4 2
1,
reset to
3.
parison amplifier
If the
now
input voltage
is
The comits
input.
analog converter, the current will flow into the amplifier. However, if
is smaller than that being generated by the converter,
the current will flow in the opposite direction.
4. In the example at hand, the comparison amplifier senses that the
the input voltage
is reset.
Comparison
amplifier
High
A^
1
Low
Clock
Digital/
ana log
converter
Logic
control
and
sequencer
TTTTTTTTT
Internal
x
reference
Readout
FIG 8-3
216
8.00
6.00
End of reading
o>
4.00
K-
"
<
Q
7924
2.00
00
Time
FIG 8-4
mation measurement.
6.
+4.000
since
is still
is
made,
logic.
8. Upon receipt of a low decision from the comparison amplifier, the
voltage which was being generated by the digital-to-analog converter is
is
this voltage
how
217
The components
in
this technique that determine the basic accuracy are the internal reference
supply and the digital-to-analog converter. These two items are common
to many different types of DVMs and will be covered in detail in a later
section.
The successive-approximation,
used nonautomatically for
many
and yet it has stood the test of time in providing the ultimate in
dc measurement accuracy. There are some practical considerations,
however, which must be taken into account when applying this technique
Any noise that is in series with the input signal
to the design of a DVM.
can cause incorrect decisions to be made by the comparison amplifier,
which results in an incorrect measurement of the average value of the
unknown. The usual solution to this problem is the addition of a low-
simple,
filter at the input to pass dc information to the converter and attenuate ac components. Attendant on the use of an input filter is an increased
response time of the instrument. Thus, in many practical situations, one
may not be able to take advantage of the inherent high speed of the
pass
filter
The
null-balance technique
is
successive-approximation technique except for the logic. In the nullbalance technique, once a new voltage has been connected to its input,
this
However, once
balance.
In
many
either the
instances, the
successive-approximation
218
Input
comparator
Vi
+ 12
Ramp
l/-v
generator
\r
Ground
comparator
Reset
Local
osci llator
FIG 8-5
Simplified linear-ramp
One advantage
DVM.
of null balance
much
is
smaller time
is
required to achieve a
new
balance.
DVM
converter.
to convert
by the
local oscillator.
is
oscillator.
slope of the
219
of the local
is,
'
10,000 pulses
_
~
=
As Eq.
0.05 sec
200
kHz
(8-2-2)
readout
is
directly propor-
As an example, assume
that 3.792
ramp
the
With a local oscillator frequency of 200 kHz and a ramp slope of 1.0
V/5 msec, the following calculations demonstrate numerically how the
measurement evolves:
Gate period = 3.792
=
Number
0.005
0.01896 sec
of pulses
=
=
0.01896
200,000
3,792 pulses
As shown above, all that remains is the proper location of the decimal
point and indication of the quantity being measured. If the input voltage were reversed so that 3.792 V were to be measured, the first comRemember, howparison would be supplied by the ground comparator.
from either comparator.
would come from the input comparator,
Other logic has to observe which com-
The second
and
this
would
first in
signal
It is
technique.
8-3
but in
220
Reset
Transfer
become more
The
is
digital-to-analog converter
is
it,
too,
The counter begins accumulating counts from the oscillator and causes
the digital-to-analog converter to generate an output voltage equivalent
to the instantaneous count.
As an example, assume that the instrument
under discussion is a three-digit voltmeter and
This means that as each new count is entered
is
output
ramp continues
DVM
221
amplifier.
V. If the input
voltage of the operational amplifier to depart from
voltage is positive, the direction of this change at the output is in a negative direction.
If the
input voltage
is
moves
at a
amount of charge
The polarity
operational amplifier.
into the
summing junction
of this charge
is
such that
of the
it
tends
V. The
process described above continues to repeat itself, and a signal that looks
very much like a sawtooth is generated at the output of the operational
amplifier.
If the input signal were doubled in value, the number of
to restore the output voltage of the operational amplifier to
Coincident
"teeth" in this output signal per unit time would also double.
with each of these teeth is a pulse which passes through Ti and on to the
input of a control gate.
These pulses are allowed to enter the reversible counter when the gate
is opened, and this opening of the gate is the beginning of the measurement
The
cycle.
this
time
is
gate can remain open for any period of time, but typically
Readout
FIG 8-7
grating
DVM.
222
system
is
pulses to be counted.
The accuracy
of the
system
is
totally
back
dependent
to the
summing
Two
of the
The diagram
BH
set.
input signal
is
is
open.
Figure 8-8 shows the ideal transfer characteristics of a voltage-tofrequency converter. There should be a linear relationship between
voltage and frequency until a certain saturation frequency
The
for
is
reached.
both positive
The
223
added
or subtracted
is
Positive counting
up
Positive counting
down
Negative counting up
Negative counting down
FIG 8-8
An
example
Fig. 8-9.
In this example,
an input voltage of
1.0
it is
assumed that
istics
Transfer character-
of voltage-to-frequency
converter.
The timing of the control gate is shown in Fig. 8-9d. Figure 8-9e shows
In the time
the four conditions of count handling mentioned above.
to t = 0.2 sec, the reversible counter is counting in a
period from t =
From t = 0.3 to t = 0.5 sec, the counter begins
positive up direction.
counting in a positive
down
direction.
1.0
(a)
Vj
-1.0
+100,000
(b)
+100;000
(c)\
-f
Gate open
Gate closed
_yd)
Q
02 0305
l_
o.7 0.8
i.Osec
FIG 8-9
waveform.
+20,000
Count
-20,000
Integration of
complex
S24
count reaches zero, at t = 0.5 sec. At this point, the counter begins
counting in a negative up direction until t = 0.7 sec. The counter now
has a net of 20,000 counts entered.
During the period from t = 0.8
sec to gate closure at
down
direction.
At the conclusion
This
to
=
(t
is
as
0.5 sec)
it
DVM
of the voltage-to-frequency
is
maximum
10 msec.
DVM
of the interpolating-integrating
DVM
its
operating cycle,
it is
DVM
equivalent to 100 counts. At the end of the 16.67 msec period, there may
be some charge remaining on the integrating capacitor C\. At this
point, Si switches to an internal reference whose polarity tends to reduce
the voltage on Ci. At the same time, the gate feeding the Is decade is
opened and passes pulses at a 60-kHz rate. This continues until the zero
comparator detects
V at the output of the operational amplifier. At
this comparison, the gate passing 60 kHz closes and the reading is
completed.
The amplitude of the internal reference is such that it can remove all
the charge from a fully charged Ci in the time it takes to enter 100 counts
into the last two decades at a 60-kHz rate.
Since the internal reference
a fixed value, the rate at which this charge is removed is also constant.
This type of voltmeter has the same basic accuracy limitations as does the
is
noninterpolating voltage-to-frequency
DVM. The
pulse generator
must
225
60 kHz
Reversible counter
FIG 8-10
Interpolating-integrating
DVM.
generate precise amounts of charge per operation, and this amount must
be constant as the rate of operation changes. The determination of the
last two digits of the reading need have an accuracy of only 1 percent to
DVM
simplified block
measurement
During the first sample period, a straightforward voltage-to-frequency
measurement of the unknown is made. At this time in the measurement
V, and only the
cycle, the digital-to-analog converter is generating
is
unknown
signal
is
converter.
The
voltage.
These pulses are fed into the 100s decade of the reversible counter during this period. At the end of the period, the count in the counter
226
unknown
voltage.
At
measurement
converter
voltage-to-frequency
initial
At
but typically
is
is
measurement.
this
point,
The
voltage-to-
frequency converter
(combined voltage-to-current and current-to-frequency converters) now generates a train of pulses, the frequency of
which is proportional to the instantaneous value of the difference between
the input signal and the output of the digital-to-analog converter. These
pulses are now entered into the Is decade of the reversible counter.
Any
carry pulses (every hundredth pulse) is fed to the 100s decade. At the
end of this second sample period, the information in the reversible counter
is transferred to the front-panel readout of the voltmeter.
The accuracy of this system is primarily dependent upon the reference
supply and the digital-to-analog converter. The accuracy is good, and
besides, the integrating capability of the voltage-to-frequency converter
Voltage/current
converter
Current/
frequency
converter
~"
jf
-^
l:-
Control
logic
3
1/10 sec
1/60 sec
s
1/10 sec
Reversible counter
FIG
8-1
Potentiometric-intesrating
DVM.
227
Example
Assume
there
is
to-frequency
voltage-to-frequency error, the
voltage is exactly 1.000000 V. Because of the
1.005 =
number of counts that is entered during the first sample period is 1,000 X
to generate 1.005 V
converter
digital-to-analog
the
causes
This
1,005 counts.
be subtracted.
The net
result
is
1.005
-503
0.99997
As
this
are
first
However,
required.
would have to be used, or a period of 1 sec would be
measurement cycle,
Because counts are fed into different decades during the
changed. This is
to
be
has
converter
sensitivity of the voltage-to-current
the
operational amplifier in
accomplished by the variable resistor shown below the
basic
The use of this means of controlling the sensitivity allows the
remain unchanged during a
sensitivity of the current-to-frequency converter to
Fig. 8-11.
measurement
cycle.
Dual-slope Integration.
This is called
conversion of an analog signal into a digital presentation.
diagram is
block
simplified
a
and
technique,
integration
the dual-slope
shown
in Fig. 8-12.
Basically, a
is
DVM
ing
is
The input
228
Switch
control
Reset
Reset
FIG 8-12
time
Dual-slope integrating
DVM.
current.
pulse
is
carry
is
Some
of the
1-MHz
oscillator
and the
five-
229
shown
is
The
line
waveform
in Fig. 8-13a.
is
both cases as
it is
returned to
is
this integra-
seen to be constant in
V.
it
it
V hence
counter.
counts and continues counting at the same rate until zero comparison is
made at the output of the operational amplifier. As Fig. 8-13c shows,
there are 50,000 counts present
when
zero
is
The
0.5000,
The value
It is
reached for Vi
1.0000 V.
The frequency
it
is
constant.
period generated
by the
oscillator
by
of the oscillator, /,
However,
if
line-
1.00
(a)
FIG 8-13
Dual-slope
waveforms.
230
the values of Ri, C\, and the frequency are not impor-
is
The
following expressions
amp
integrate
op
'
-t
D
Ci /ii
1 integrate ==
^
Hi
10 5
anO.
~T~
Accumulated counts =
Equation
~pj
(8-3-2)
1 discharge
^o-O-lj
* discharge
accumulated counts
7
y.
-
V re
10 5
(8-3-2)
refer-
ence voltage.
safely
though
Even
8-4
it
Normal-mode Rejection
cally,
is
line frequency,
but
in practice
desired.
it
was observed that the reading is achieved only after a number of successful decisions by the logic circuitry.
Any incorrect decision along the way
can cause a completely erroneous reading of the unknown signal.
imposed noise
is
the most
common
Super-
231
to the
amount
of noise present.
digital-ramp techniques.
During
integrating period.
DVM
input of the
truly integrated
is
= V
DVM
can be presented by
(8-4-1)
sin at
it is
V.v
av
sin wt dt
J ti
0}
= _ X
By
r\t=u+T
Vi
cos wt
[cos (*!
T)
(8-4-2)
cos i]
7 av =
-^ [-2
sin
K (2ti + T) sin y
is
2 (o>T)]
(8-4-3)
232
where
ttV
sinK(2cofi
coT)
(8-4-4)
27i
=
V av (max)
sin
uT
^shwr
(8-4-5)
sin irjT
A plot of Eq.
length.
50
40
m
'-30
|o
111 jjj^
w
I
\
jXJ*
<u
20
10
0.1
10
1.0
100
Frequency Hz
,
FIG 8-14
DVM.
Normal-mode
rejection
characteristics
of
inte<jrating
is
50
8-5
Common-Moc/e Reject/on
Another type
of disturbance
S33
in
measurements
measurement
the voltmeter
output.
is
floating
is
to be
is
common-mode
cell in
Fig. 8-16a.
components shown
The
instrument.
measuring
resistance
all
Rt and d.
1
The
C 5 is much larger in
frequencies than
is
/?,
Instrument
*f?s
XRa
High
</? 5
tCi
f
FIG 8-15
j^JJfnd
Noise
signals
and
234
500 A
-1/
High
Ry
Bridge unbalance
voltage
High
Rz
t-V/1
Low
(b)
(a)
FIG 8-16
Load
cell
and equivalent
circuit.
a wire or narrow conductor while the low side is a plane or large metal area
power ground of the measuring system, which is
Because of
this,
and C 5
will
and
be neglected.
The disturbing fact about these paths is that current flowing through
them will tend to generate a voltage that is in series with the signal to be
measured. The most troublesome path of this nature is through R 2 and
the parallel combination of Rx and
common-mode
the
voltages
4.
and
Common-mode
4.
is
due to
rejection
2
is
Given a
bination of R^ and
4 ).
9
In a well-designed floating instrument, # 4 may be as high as 10 Q
may be as large as 2,500 pF. These values lead to the following
and
CMRs:
Direct current:
Alternating current
= 10 9
CMR
20 log
120
dB
10 3
CMR = 20 log
= -20
1/2tt/c
10 3
/=60Hz
= -60 dB
10
log
B
10 3
CMR
mode
In
signal across
many
2.
intolerable in a measurement.
This
is
20-mV
ac signal would be
particularly true
if
low-level
/?4< C4-I-
</?6
235
TH
FIG
ing
8-1 7
Configuration of float-
and guarded
input.
is less
C 6 do not physically exist; they cannot be measured directly, but represent what leakage remains "through"
the guard when it is connected as shown in Fig. 8-17. In other words,
the proper use of this guard considerably lowers the effective leakage
between Lo and power ground; that remaining is indicated by the comThe guard is actually driven by the common-mode
ponents of Z 6
The
signal, as is Lo; hence there is virtually no current through Z4.
bottom side of Z 4 is essentially "bootstrapped" to low because the guard is
The
resistance
driven
The
signal as low.
CMR
is:
10
= -20 log= -160dB
3
10
Alternating current:
CMR =
20
1/(2tt60
log
2.5
10 3
-20 log6
= -120dB
10 9
10 3
10" 12 )
236
Now
of ac
normal-mode
CMR.
signal.
CMR
of
Effective
is
CMR.
CMR
dB
of
DVM
DVM
8-6
Principles of
AC
Voltage Measurements
Most ac measurements are made with ac-to-dc converters, which produce a dc current proportional to the ac input being measured and use
-160 dB
-120dB
(a)
-160dB
-120dB
=sJAM
x
FIG 8-18
tegrating
Log f
(b)
CMR
DVM.
Effective
guarded
of in-
237
frequency-selective circuits.
Most ac voltmeters
are classified into three broad types: rms-respondand average-responding. Those that are average
responding and peak responding are generally calibrated to read the rms
value of a sine wave. Only a small minority of voltmeters available today are true rms-responding instruments, but it is expected that in the
future many more true-rms voltmeters will be introduced.
Voltmeters are ordinarily calibrated in rms volts because that is the
equivalent to a dc voltage which generates the same amount of heat power
in a resistive load that the ac voltage does.
For this reason, rms voltage
is synonymous with effective voltage, and the term is used predominately
in discussions of ac voltage without referring to the term rms.
The rms value of a waveform is defined as the square root of the avering, peak-responding,
Theoretically,
rms value can be found by measuring the voltage at equal intervals along
the waveform for one cycle, squaring the numerical value of the voltage
at each point, finding the average value of all the squared terms, and then
taking the square root of the average value. If there were n discrete
values Vk in a series of measurements, this could be expressed as
V
ei")
V rma =
as
n -^
oo
(8-6-1)
replaced
by
Then
integration.
the
rms value
is
expressed as
Vrms =
{-J v*dt)
-gX'-*)
From
this,
v rms =
(8-6-2)
is
|o
(7 max
sin
ey de\
= (M(y ma*)
wave
is
from
to T.
found.
)*
=
The
is
0.707 7max
(8-6-3)
238
would be described
as
n
1
V av = - J V k
n hi
If
(8-6-4)
is
defined mathe-
matically as
vdt
\jo
8" 6 " 5 )
is
~ ['
^ max
sin
Odd = -
7 max -
wave
is
0.636F max
(8-6-6)
mean the average rectified value or the average without regard to polarity.
The use of average-responding rather than rms-responding voltmeters
is
of sine
an average-responding
1 V can be applied to
the scale
is
adjusted to
proportionality to the
waves
It is justified
measurements. In calibrating
meter, a pure sine wave with an rms amplitude of
the meter and the resulting pointer deflection on
read 1 V. This is done by applying a constant of
meter called the form factor, which is defined as
in electronic
follows
Form
factor
= rms
or effective value
average value
The form
Form
factor of a sine
factor
To determine
wave
V
=
7 av
is
calculated as follows:
" 707 V
0.637 7 max
"""
1.11
(8-6-7)
v
;
the average value of a wave with the use of an averageresponding voltmeter, simply divide the reading on the voltmeter by 1.11.
To determine the peak value of a sine wave, multiply the reading on the
voltmeter by 1.414. Obviously, to get the peak-to-peak value of a sine
wave, one can multiply the reading on an average-responding voltmeter
by
2.828.
V outage
8-7
239
Average-responding Detectors
gain stability for measurement accuracy as well as broadening the frequency range of the instrument. Inclusion of the meter circuit in the
feedback path minimizes the effects of diode nonlinearities and meter
impedance variations on circuit performance.
should be noted that the capacitors in the meter circuit tend to act as
filter capacitors for the rectifier diodes and also coupling capacitors for the feedback signal.
However, both capacitors could be replaced
It
storage or
by
resistors and the circuit would work nearly as well with the inherent
meter inertia for filtering. The diodes act as switches to maintain unidirectional meter current despite changes in the instantaneous polarity of
much
as
it
of
aver-
This
240
shown
as
V =
If piv) is
If
lies
Piv) dv
fj
the noise
where
<r
(8-7-1)
dv
V =
voltage
\v\p(p)
/_"
is
is
between
(8-7-2)
vp(v) dv
-^
and
c-'
'
2 ''
+ dv is
(8-7-3)
dv
gives
w *-^
/
2
ve-W^dv
dV =
v -^5b/.""W
I
a(ftr)Wo
is
l/(2)Mg B"
2/ir m
is
of a sine wave, or
r.-^r.
(8-7-5)
"
'
* ~ - J^
(8"7 "4)
rms value
- i."v. =
where Vo
2<r
(2r)'
* -
2(2)'
Then,
0.886.
(8-7-6)
241
= ^ m [sin
where k
is
(8-7-7)
<*>)]
We can
V
= a
mental.
k sin (nd
2tt
[sin
+2x
*'
r*
[sin
/9
k sin (nd
k sin (n0
by writing
4)]
</>)]
dd
cw}
(8-7-8)
which reduces to
^av
ir
cos 0i
cos
+ n- [cos (w0i
-\-
<f>)
cos (n0 2
<)]
}
J
(8-7-9)
where
0i
and
equation
sin
k sin (n0
<f>)
(8-7-10)
1.11.
is,
1
a
11F
a v max
j.
cos 0i
cos
+ n- [cos
(n0i
4>)
cos (nd 2
</>)]}
(8-7-11)
harmonic distortion. If this is the case, Eqs. (8-7-8), (8-7-9), and (8-7-11)
must be modified to integrate the rectified waveform between zero
crossings.
The
rms reading
error in
242
40
80
240
200
160
120
Phase angle of second harmonic, deg
320
280
360
400
form
factor of
of the
with the fundamental.
Figure 8-22 shows the effect on the waveforms caused by second- and
third-harmonic distortion with the harmonic phase set to obtain the
maximum
40
80
of a
240
160
200
120
Phase angle of third harmonic, deg
280
form
distortion can be
320
360
400
^-^
(a)
243
,,Resultant wave
(c)
Resultant wave
Resultant wave
lb)
FIG 8-22
Distorted waveforms with largest time error in averageresponding detector: (a) in-phase second harmonic, (b) out-of-phase
second harmonic, (c) in-phase third harmonic, and (d) out-of-phase
third harmonic.
calculated by taking the square root of the sum of the squares of the rms
value of the fundamental and the rms values of the harmonics:
^rm S
[(Vims fund) 2
,
(V tms 2nd) 2
,
(Vims 3rd) 2
,
]*
(8-7-12)
TABLE
8-1
Maximum
Harmonic
1
5
10
20
30
error
0.7071
0.7074
0.7080
0.7106
0.7211
0.7382
0.02
0.032
0.112
0.485
1.93
4.2
0.34
1.03
1.77
3.8
8.5
13.8
244
occur at the time the fundamental intercepts the zero axis. This is also
the condition often encountered in practice. A square-law term in a
transfer characteristic, for example, produces this phase relation for the
second harmonic. It can also be seen that for second harmonics of typical magnitudes (less than 10 percent), the average-reading meter will give
For a second harmonic of 10 percent
readings quite close to the rms.
magnitude, for example, the error of the average-reading meter is less
than
percent.
reading than second harmonic, but it also causes greater variations than
any other harmonic. The extremes of error with small amounts of odd
harmonics are given by the percentage of the harmonic divided by the
order of the harmonic. Small amounts of harmonic in this case can be
defined as percentages less than 100 /n, where
is
harmonic.
It should be noted that for typical amounts of this worst harmonic, the
Third
third, the accuracy of an average-reading meter is still good.
harmonics up to 10 percent, for example, can cause errors no greater than
5 percent.
When more than one harmonic is present in the applied wave, the
mathematics of each case becomes more complicated. As a result, no
However, some
analytical studies of these situations have been made.
experimental data have been compiled for combined second and third
harmonics with various amounts of fundamental [2].
Two other cases of interest that can be easily investigated are the
differences between readings on an average-responding voltmeter and a
true-rms voltmeter with a triangular wave or a square wave applied.
A calculation of the average value of a triangular wave, shown in Fig.
8-23a,
shows
_?/"*_!
of the triangular
(8-7-13)
wave
is
/0
245
2ir
(a)
FIG 8-23
Triangular
and square
waveforms.
2w
(b)
Multiplying the average value of the triangle wave by the form factor
1.11 and determining the error between the average-responding and true-
rms voltmeters,
Percent error
1.11/2
l/(3)
100
1/(3)"
= -3.81
percent
(8-7-15)
In a square wave the unique relation exists that the average, rms, and
peak values are all the same. Since an rms-calibrated, average-reading
meter indicates 1.11 times the average value, it will indicate 11 percent
high for the rms value of a square wave. Further, a square wave has
the lowest ratio of rms value to absolute-average value of any wave.
It
follows, then, that an average-responding meter will never read more
than 11 percent too high.
8-8
Peak-responding Detectors
The primary
difference
FIG 8-24
The
is
Peak-responding meters.
value of the applied voltage, and the meter circuit then responds to the
capacitor voltage.
Two
shown
of the
in Fig. 8-24.
246
which
is
discussed later.
The
third
train
is
form. The output of the detector is then filtered to measure only the dc
value of the detected waveform. If a pure sine wave or any other symmetrical waveform were measured, the detected and filtered value would
be proportional to the peak-to-peak value of the waveform. For the
positive pulse train, the filtered value would be approximately equal to
However, for the negative pulse
the dc value of the original waveform.
train, the
filtering
would appear to be a
is pri-
247
dc value
(b)
(a)
FIG 8-25
peak-detector
Alternating-current
train
output
with
pulse
train,-
is
instantaneous output,
across the capacitor,
voltage and
its
consists of a dc value
v,
Vd
and a
series of ac
We
can write
Vd
+ V
(8-8-1)
sin ut
Va
sin
wt r^
FIG 8-26
Alternating-current peak
detector.
rent
cur-
is
= I
= J
a
t
(eol>*T
1)
J a ( e (/n*r)(TVI-V .in0
(I
(g(9/n*r)(V,+V a sinD
1)
(8-8-2)
1)
a constant of proportionality determined by the diode manuIn many recently designed silicon diodes, n has been
experimentally found to be approximately 2. In the equation, q is the
where n
is
facturing process.
charge on an electron, k
degrees Kelvin.
is
is
temperature in
248
= V
V =
( e (<ivjnkT)
(qvjnkT)
sin
_ ^
o>t
average
lost
(8-8-3)
average
(8-8-4)
average
(8-8-5)
I.R
Rearranging,
+ S7>"JnkT
gixrini
jr
(
z)
e (qV a lnkT)sinut
2J 2 (x) cos 20
form
(8-8-6)
or
g* S i
Then
= j (-j x )
for the dc
(8-8-7)
e qVJnkT
qVo
I
_|_
nkT
and
e ( g lnkT)V a sinot
-L- Va
TtK J.
SHI
0)t
+ ^(nkTJ
-^Y^
sin'
w*
Kr
(8-8-9)
we
we can
249
and write
+
^
nkT
or
Q
,r
-^f v = ^f v"
-'
The accuracy
of Eqs. (8-8-8)
where
Vfc
"*
facturing processes in the fabrication of the detector diode, but the equations are useful to
approximation.
8-9
The
Peak-to-peak Detection
addition of a capacitor and a diode can change a peak detector
shown
in Fig. 8-27.
The advantage
is
of
the absence of
Cy
JC
FIG 8-27
"
DZ
M
f
Peak-to-peak detector.
and
must be large enough so that the voltage does not
change appreciably across C% during one period of the input voltage, and
the voltage across C\ does not appreciably change in the process of
of the circuit, C\
recharging
C2
250
8-70
Root-mean-square-responding Detectors
There are many occasions when measurements of the true rms value of
When measurements of electrical or
acoustical noise, low duty-cycle pulse trains, or voltages of undetermined
waveform are made, it is almost imperative that an rms-responding voltmeter be used.
In past years, rms-responding voltmeters have not been used widely
a voltage are highly desirable.
because of the
of
difficulties in
somewhat reasonable
cost.
to take
which exhibit a
fairly accurate
hot junction.
This measurement
rms value
is
251
dc ampl.
^)
FIG 8-28
rms voltmeter
technique.
null-balance
with
may
power
Meter
Crest factor
crest-factor rating.
to the
rms voltage
by the
is
waveform.
of a
The
determined by
waveform will be
spectrum are
frequency
accurately detected.
shown in Fig. 8-29. It should be observed that the width of the frequency spectrum is inversely proportional to the width of the pulse and,
quency spectrum
its
as will be
shown below,
If
*\
*
(a)
FIG 8-29
1*.
ill.
of a pulse
Frequency
train: (a)
frequency spectrum.
~0
(b)
spectrum
252
input waveform would not pass through the amplifiers, which would
result in an error in the voltmeter reading.
Since a pulse train represents an extreme case of a nonsinusoidal periodic
waveform and
some
in
+to
\\
Va
vb
-<
Pulse
Time
'
>
FIG 8-30
is
volts
cases
*-
train.
noise,
it
crest factor
waveform period
That
T.
is,
D =
(8-10-1)
we can assume
va + v = vpp
(8-10-2)
and
Va to = V
(T
(8-10-3)
Then
V = VPP D
Va = V(l - D)
(8-10-4)
of
Va
(8-10-5)
and Vb
_ / TVU ~
vv tms -
is,
D)Ho
after integration,
+- VPP D
2
(T
Y
J
= V pp [D(l - D)]
Since crest factor
y(i
CF =
V[D(1
CF
~
-
is
equal to
D)
D))
Va /VTms
e-y
for values of
(8-10-6)
<
D <
}/2
(8-10-7)
If the duty factor D is small compared with 1, the crest factor is approximately equal to the reciprocal of the square root of the duty factor, or
CF
(8-10-8)
253
Thus, the crest factor of a pulse waveform with low duty factor turns out
somewhat different from that which might be presumed from conA pulse waveform with a duty
sideration of the peak-to-average ratio.
factor of 1 percent has a crest factor of approximately 10, not 100 as might
to be
be assumed.
High crest-factor performance is not easily obtained. An rms voltmeter with a high crest-factor rating must have amplifiers with sufficient
dynamic range to pass signals that have a peak amplitude many times
For example, a pulse train with a crest
larger than full-scale rms value.
factor of 7 must have a minimum dynamic range corresponding to 7.14 V
to read full scale on the 1-V range of an rms voltmeter arid a dynamic
range corresponding to 14 V to read correctly if the pulse train should be
turned over.
8-77
V ima =
where
fci
laV,
and
fc 2
k 2 Va
= fci^
Va
Va
Vp
Peak
detector
fU
FIG 8-31
Average
detector
Va
WW-Wr-i
^T
\K
1 1/_
|
Dividing by
(8-11-2)
/c 2
R/k,
peak and
(8-11-1)
its
u-routc
^g Q^prf
f~
Quasi-rms detector.
Using Eq. (8-11-2), we can solve for fci and kz for any two waveforms,
knowing their form factor and peak-to-average ratio. It should be noted
that the detector will indicate the true rms value for only the waveforms
whose form factor and peak-to-average ratio are used in the design. Any
254
If the waveforms chosen are representative of those to be measured, the detector will
read the rms value with a relatively small amount of error, which can
be calculated by using Eq. (8-11-1).
FIG 8-32
which the average and peak detectors are combined as shown in Fig.
Figure 8-32a shows a simplified average detector and Fig. 8-32b
shows a peak detector. Figure 8-32c shows the combination of the average and peak detectors in which the approximation to the rms value is
determined. The ratio R1/R2 can be adjusted empirically to provide the
rms output for a group of closely related waveforms over a limited
dynamic voltage range and a limited frequency range.
Another possibility is to make Ri nonlinear by using a diode shaping
network or other suitable nonlinear network to more closely approximate
a true rms reading over a large dynamic range.
The shaping network
could be used to "square" the input voltage, which is averaged by the
average detector, and applied to a meter with scale resembling a squarein
8-32.
M
FIG 8-33 Quasi-rms detector with nonRi substituted in previous figure.
linear
root function.
shown
in Fig. 8-33,
is
by empirical means.
8-72
255
Sampling Voltmeters
motion
is
apparently "slowed
by which an
oscillating or repetitive
voltage
is
of the frequency
age.
function
is
is
is
is
assumed
irrelevant,
not.
Incoherent sampling, shown in Fig. 8-34c and d, is especially advantageous in a voltmeter, because it gives the meter the sensitivity, accuracy,
and broad frequency range of a sampling instrument, and yet it is less
costly than coherent techniques and, unlike coherent sampling, it does
not require that the input signal be periodic. The sampling voltmeter
FM
signals.
operates equally well with sinusoidal, pulsed, random, or
it is
all
situations
work
in
to
sampling
incoherent
of
the
technique
For
necessary that there be no correlation between the sampling times and
If the sampling frequency were
the motion or signal under observation.
being measured, the motion
signal
of
the
frequency
a subharmonic of the
would be completely stopped. Thus, all the samples would be exactly
sampling intervals
is
it
would be impossible to determine the peak, averOne way to produce uncorrelated or nonuniform
256
Ah
If
(b)
n'ir"'
(d)
FIG 8-34
random sampling.
Sample
Pulse
generator
'
Input -
\fcrtage -controlled
10 Hz
triangle wave
generator
'
Sampling probe
assembly
FIG 8-35
oscillator
hold drive
<
Attenuator
'
Sample hold
meter
circuit
Sample
hold
output
8-73
257
Synchronous Detection
but essential voltage measurement required for many applithat of measuring low-level signals obscured by noise or other
nonrelated signals. Such conditions are frequently encountered in communications systems, medical research, and control systems. Unfor-
difficult
cations
is
Input
Reference
input
FIG 8-36
r\?
*\
v^
Synchronous
>-
Meter
Filter
detector
circuit
-J
Synchronous
rectifier.
tunately, almost
noise
in the
amplifying circuitry.
designed.
shown
in Fig. 8-37.
The
Inhibit
'
circuit
*
Input
Attenuator
'
Synchronous
detector
t
Meter
circuit
'
Phase-
lock
de tec tor
FIG 8-37
Filter
Voltage
oscillator
internal
258
oscillator
The voltage
is
derived
from a phase detector, in which the input signal and internal oscillator
output are compared in phase.
Ordinarily, a synchronous detector (or demodulator) is simply a phasesensitive detector in which the reference signal is in phase with the
signal to be demodulated.
This relationship gives maximum efficiency of
amplitude demodulation; to get
maximum
= W,
Vvei
sin out
+ V
r,
sin 3o>it
+V
Ti
sin 5o>it
(8-13-1)
Vi
Vi sin
ait
+ V
sin
2w 2 t
+ V
sin 3wrf
(8-13-2)
It can be seen that upon multiplying Eqs. (8-13-1) and (8-13-2) and then
taking the average, several things become apparent:
1. No dc due to even harmonics of the input signal appears in the
detected output.
numbers
age
is
This
is
zero.
to their
harmonic number.
3.
If
This
at this frequency.
by an
is
an undesired
is
259
effect
inhibiting circuit.
is an
measuring signals in the presence
However, care must be taken in the interof noise and spurious signals.
pretation of the meter reading, particularly when the input voltage is not
The synchronous
detection,
or frequency-selective voltmeter,
sinusoidal.
8-14
Direct-current Probes
is
to measure the average magnetic field produced by the current in the vicinity of the conductor, and this is not easy to do accurately.
Unless a magnetic path of low reluctance is provided around the con1
alternative
is
the
ductor, the held measurement varies greatly as the distance between
conductor and the magnetic sensor changes. Of course, a torroidal core
surrounding a conductor will concentrate the flux, but ordinarily one
would have to break the circuit to insert the core or leave the core permanently installed. Some years back, however, a tiny split core was
designed* into a probe (Fig. S-38) in such a manner that the two halves
can be opened like jaws by squeezing the flanges on the probe handle
(Hewlett-Packard model 42SA). A spring return closes the jaws tightly
around a conductor when the flanges are released. Practically no air gap
remains.
of the core contain windings that arc connected as a
magnetic amplifier [3]. That is, an ac excitation is applied to the wind-
More
precisely,
il is
a line integral
which
is
being evaluated,
Hdt
released.
when
flange* are
I.
260
ings,but they are balanced and no output voltage occurs if the square-loop
magnetic materials in the two halves saturate at the same electrical
angles of the excitation voltage.
A flexible cable
ment where
it is
amplified, demodulated,
shown
convenient.
One
10~ 6
261
as
much
as the
8-75
/measured
Output
Ci'p-onheod
imvVroA
Scope
or
VM
l_jf___,
FIG 8*40
Current
2-f+We
5 -ft cobie
tiansiitorixed Amplifier
ac voltmeter.
winding.
FIG 8-41
An
262
wiih
oscilloscope or a voltmeter,
at)
made
of 1
mV/mA
[5].
One commercial
The frequency
The
and 0.05
^H
in series
pF
is less
than 50
10~ s
In addition, there
is
Si
ground.
Broadband noise is less than iiO-jiA rms. Better sensitivity can be
achieved by looping additional turns through the probe head. This
raises the loading effect of the probe by raising the primary inductance.
This increased inductance in combination with loop-to-loop capacitance
can result
in
to
probe
in
mm analyzer.
conjunction with a
is
Some wave
analyzers gen-
Use of this signal to stimulate the device under test ensures that the narrow bandwidth of the analyzer can be used to improve the signal- to -noise
Measurement of currents as low as a few microamperes can be
ratio.
+ 111
10 Hi
""^s^
IkHi
1O0H2
1CMHJ20MB;
(MH2
frequency
res
pome
of
current
probe
when
made
in this
The
way.
full
bandwidth capability
of the
probe can
263
still
be used.
The probe can also be used for current summing and equalizing. Since
the probe is effectively a current transformer, it has the property that it
the instantaneous values of the currents in two or more conThis property makes the probe
it may be clipped around.
in which it is desired
applications
a valuable and easily applied tool in
currents.
alternating
to equalize
will
sum
ductors that
CITED REFERENCES
1.
2.
3.
4.
DC
5.
Milliammeter,
Forge, Charles 0.
MHz
A New
CHAPTER NINE
IMPEDANCE MEASUREMENT
Henry
P. Hall
its reciprocal,
systems.
for electrical
9-1
Definitions
and Formulas
Impedance Measurement
current through
it
by Ohm's law
R =
E/I.
For sinusoidal
ac,
This quantity
is
265
often called
is
called reactance
These quantities
is
The
of
reciprocal of dc resistance
impedance
part called
is
effective,
susceptance B.
is
The
called admittance Y.
or ac,
However,
G and B
and
X respec-
tively, for
for
is
given.
Polar Form.
Z = R
+ jX
\Z\e
|Z|(cos
\Z\
+j
(9-1-2)
sin 6)
= *vR 2
+X
2
,
Likewise,
Y = G
+ jB
\Y\e>*
\Y\(cos
<t>
+ j sin
(9-1-3)
<f>)
where
|F|
= VG+ B
1
and
\Z\
We
Q and
<f>
= -8 =
tan" 1
-7?
G
by
266
of all three
^VW
are additive.
0.05/i
f^
,000 A
FIG
9-1
Series
and
1 2,ooo a
parallel circuits
1,592 Hz.
The words
effective
and
Impedance Measurement
TABLE
9-1
267
Cs Rs
C.-(l+ D*)CP
-
D*
ir^
Rp =
R> =
r^& Rp
D =I=
W R,C,
+ Z>
+D
R =
~d^ r
=
-
(1
+Q2)R>
<oC pxi/p
P
l*
L =
-
R.
Rp
*s
Q*
fTo^*
i
t
+
,
_,
Q2
rT^ L
L* =
RP =
RP
n -
1 -~ ^1
D ~R~,
4-
~W~ l
(1
'
(1
+ D2)L
Q 2 )tf.
- 3l
~ ^L.
j'500 n,
different.
268
components
in Sec. 9-2,
is
a voltage
is
t0 -
Likewise,
is
The
te .
transfer
'.
part of
3.
by impedances Z h Z 2 Z 3
and Zi, by an instrument that could measure the open-circuit transfer
impedance of this network. Such a network nicely "defines" a lowvalued impedance Z because anyone making a proper transfer impedance
accurately measured, without error caused
z2
Zi
t-W
VSAr
Z*
FIG 9-2
four-terminal
Z =
impedance
FIG 9-3
admittance
three-terminal
Y =
I a /Ei.
E./Ji.
Impedance Measurement
z
Zz
Zo
Zc
z1+ z3
+ z + zc
,,
'*
^2
The
FA
Y,
+ Yz + Yz
(b)
(a)
FIG 9-4
+^A
*
Zb
269
transformation.
symmetry.
This transformation
is
designers
make
use of
it
impossible to obtain.
9-2
An
Resistors
[1,
2]
Behavior of Resistors.
shock or humidity.
The
most
resistors,
270
/?dc
Re
FIG 9-5
of a
An
equivalent circuit
resistor.
The
temperature coefficient
Many
degree Celsius.
is
is
(ppm) per
types of resistors increase in value as temperature
in percent or parts per million
maximum
#dc
=
'
a>
# dc
<7p
Impedance Measurement
This effect
is
values of resistance
it is
possible to
271
At medium
L and Cp
is
masked by
others.
is
represented
by a lumped
capaci-
tance Cd which can simulate the first term in a power series in O'o>) 2
This
resulting from capacitance between various parts of the resistor.
If the
capacitance reduces both series and parallel values of resistance.
distributed capacitance were between the resistor
[1].
where R/l
ohms per
is
centimeter.
is
(if
wire-wound) in
tance values
If
of
an iron-core
value of R e
Resistor Phase
calculated
Qof
Q =
The
last
tan
co
(^ - Rc\ -
^|^
(9-2-2)
272
good standard
resistor is
of the causes
is
much
higher
However, at this resistance level, shunt leakage resistance must be kept very high and a guarded measurement is recommended.
The range of resistance standards extends down to 10 /zft at which level
precision four-terminal shunts used for high-current measurement have
an accuracy of 0.04 percent. While wire- wound resistors over 100 Mfi
have been made, film types are usually used in this range and on up to
10 13 ft. Sometimes T networks are used as high-valued three-terminal
resistance standards because their transfer resistances can be very high
resistance value.
(see
YA
9-2-2
Capacitors
[6,
7]
FIG 9-6
An
of a capacitor.
equivalent circuit
Impedance Measurement
TABLE
9-2
273
Approx.
Best
Usual capacitance
Type
range,
Tantalum
TC,
ppm/C
mF
electrolytics.
electrolytics.
D,
1
-120
0.001-1
0.001-10
0.001-10
0.001-100
Paper, impregnated
Aluminum
tolerance,
10
10,000$
1,000
20
10,000t
0.05J
100
Nonlinear
500
100
pF-0.02
0.0001-0.01
0.001-0.5
0.005-100
5-100,000
20
to
- 750
Insulation
resistance,
kHz
0.0005
0.002
0.01
0.005
0.001
0.001
0.01
0.04
voltage
recovery,
25C, fi-F
0.1
0.2
0.5
>10
>10*
>10*
>10
>10
>1,000
5
3
2
>700
>100
>30J
4
2
10
t 10-second discharge,
t Varies widely.
The
small capacitance
Ca which
,
is
tance over
its audio-frequency value resulting from interfacial polarizaThis effect can be considered a gradual redistribution of charge
throughout the dielectric and is most pronounced in composite dielectrics.
tion
[8].
It causes
what
is
been charged and the gradual recovery of voltage after the capacitor has
been momentarily short circuited. The R a C a time constant may be
many seconds, hours, or even days. Table 9-2 gives typical values of
voltage recovery after a 10-sec discharge which is approximately equal to
Ca/Co if the time constant is very long.
Capacitors with dielectrics having dipole moments have a molecular
polarization in the rf range which causes a further change in capacitance.
This could be represented by a second RC combination having an appropriate time constant.
series
Eq. (9-2-3)
C =
rf
0>>LCo
(9-2-3)
{f/frY
274
even though
resistance),
value.
in
some
RC
t
cases
is
humps
make well-defined
humps are often
nonhomogeneity
in
In simpler models,
variable resistor
Rd
could be represented
by a
single
the use of a single fixed series or parallel resistance to simulate all loss
can be very misleading (see Sec. 9-1). The often used equivalent series
resistance
therefore
is
is
much
it
much
and
higher
was determined.
ohm
pF.
Such capacitors are very expensive and are kept by only a few national
laboratories.
The most stable working standards are three-terminal
10-pF capacitors, designed by the Bureau of Standards, that use fused
silica disks as the dielectric.
Even though their temperature coefficient
is relatively high (about 11 ppm/C), their stability and ruggedness make
them
excellent
for
use
as
traveling
standards
for
interlaboratory
Impedance Measurement
at better than 100
ppm/year
stability.
9-2-3
pF and up
Inductors
275
down
to
to 1 F.
[9,
10]
The value of an inductor depends more on the conditions of measurement than does that of the other types of component. This is particularly true for iron-core inductors, for which measurement repeatability
within a few percent is considered good and is usually adequate.
Because
many
is
While
this
is
a more meaningful
Iron-core inductors are nonlinear, and therefore the measured inductance depends on the level of the test signal and the level of any dc bias.
The combination of ac and dc used should simulate that of the actual
application if possible.
Two widely used plots are inductance versus ac
level with no dc applied and the incremental inductance plot of inductance
versus dc bias current with very small ac signal (see Sec. 9-5-4).
series
inductance
is
If
because the core losses, which behave as parallel resistances, vary widely
and change the effective series inductance value (see Table 9-1).
The initial condition of the core will affect the measured value, for it
may have been left in a state of residual magnetization. It should be
demagnetized by applying a large saturating ac signal and slowly reducing
this to zero.
by the proximity
rf coils
where the
is
much
FIG 9-7
higher.
An
of an inductor.
equivalent circuit
276
The
but
is
nearly independent
Rh represents hysteresis
It has a value proportional to frequency and to the flux density
loss.
raised to some fractional power, depending on the Steinmetz exponent [11].
All coils have capacitance that increases the effective inductance
through resonance. This capacitance is actually distributed and can
of level
and frequency.
parallel resistance
Rd
The
capacitance
of air-core coils at
high frequency.
Standard Inductors.
wound
as
on ceramic
cores.
frequency.
9-3
9-3-1
Meter Methods
to
Measure Impedance
Direct-current Meters
Impedance Measurement
current in the voltmeter.
It is
not
difficult to
make
277
and electronic meters, which draw extremely low power from the
measurement circuit, have negligible error from these sources, except
errors,
FIG 9-8
of resistance.
vy&
y&FIG 9-9
--^w\,
practical
A)
ohmmeter
circuit.
While digital ohmmeters can be quite accurate and 0.01 percent is not
uncommon, simple ohmmeters incorporated in multimeter testers are
limited in accuracy to several percent.
The ohmmeter
278
Wo
FIG 9-10
meter
if
circuit
FIG 9-11
ohmmeter
milliohm-
Ri
a sensitive voltmeter
is
The
basic
VTVM
meg-
circuit.
Megohmmeters.
very
common
measuring high-valued
Here the "ammeter" consists
of a high-resistance standard shunted by a vacuum-tube (or FET) voltmeter whose deflection is proportional to R S /(R S + R x ). Such instruments measure resistances up to 10 12 ft or more, but at relatively low
resistors is the
megohmmeter
circuit for
of Fig. 9-11.
accuracy.
them
necessary.
is
The
two
Two measurements
are
more accurate
the measurement
method
tiometer measures
Rx
and
in turn.
Calibrated
voltage divider
FIG
9-1 2
The potentiometric
of comparing two resistors.
method
",
Impedance Measurement
279
9-3-2
Ammeter-Voltmeter-Wattmeter
Method.
If
the
ammeter-voltmeter
method is used for ac, the ratio of the meter readings is the magnitude of
impedance and admittance. To separate resistance and reactance, some
type of phase-measuring or discriminating circuit is necessary. The
>
U&
z,
^-ammeter-voltThe
meter-wattmeter method for mea-
FIG 9-13
suring impedance.
method
shown in Fig.
to
classical
is
as
9-13.
9,
trace
-*
FIG
9-1
[13].
A capacitance meter
circuit.
ient for
E
Ei
Cx
Cx
C,
is
conven-
280
Stabilized
oscillator
tft|
Phase
Ampl.
sensitive
detector
"^
Rz
Reference
signal
FIG 9-15
may
if
is
little effect.
by
An RL
designed in which an
oscillator
Resonant
and
this
circuits
applications.
(Refer to Sec. 9-6-4 for discussion of the
resonant capacitance meter.)
9-3-3
a
Q meter and
current is passed through an unknown impedand quadrature components of the resulting voltage
are measures of the resistance and reactance of the unknown.
Alternatively the magnitude and phase angle of the voltage may be measured to
obtain impedance magnitude and phase. Likewise a known and constant
voltage may be applied and the resulting current measured to get the
corresponding admittance quantities. This simple principle has been
the basis for several impedance meters that characterize the unknown
If
Impedance Measurement
281
This instrument
Vector Impedance Meter (Hewlett-Packard 4800A) [14].
above over
described
manner
the
measures magnitude and phase in
to
Hz
500 kHz).
frequency
and
(5
Mfi)
10
fi
to
(1
wide ranges of impedance
is constant and the output
current
input
kfi,
the
1
to
up
impedance
For
magnitude. Above
voltage magnitude is used to indicate impedance
the output signal
that
so
measured,
1 kfl, voltage is applied and current is
the indiHowever,
magnitude.
admittance
to
is proportional
actually
magnitude on
cating meter has an inverse scale which reads impedance
phase angle
the
is
angle
The impedance phase
these higher ranges.
is output.
which
and
input
is
which
matter
no
between voltage and current
is a power of
that
frequency
a
at
operating
meter
impedance
A.n
1/2t
will
read L(l
+ 1/Q
+D
9-3-4
Resistance
may
*s
4r;
Z +
(bi
(a)
FIG
9-1
circuit as in Fig.
to use a precision resistance divider to form the bridge
The output voltage as measured by a high-impedance voltmeter
9-166.
is
proportional to
R x Rs
(R,
+ Rs)/2
AR
R*v
(9-3-1)
282
R,
is
in the
unknown from
An
often used.
and
systems.
is
small.
of
for
Re
&
Ei
(\Zs\
=
1
[cos (8 X
ds )
\Z S \)/(\Z X
1]/(1
\Z a \)
\Z X \/2\Z S
\Z a \/2\Z x \)
and
sin (0 X
Ei
If
cos (0X
ds )
8,)
\Z X \/2\Z S
is less
\Z S \/2\Z X
and
K
\
|Z S |)/(|Z X
\Z S \)
8a in radians with
9-4
Direct-current Bridges
9-4-1
The
ance
between a meter and a bridge method of measuring impedanalogous to that between a spring scale and an analytical
relation
is
Impedance Measurement
283
balance for measuring mass. The spring scale and meter methods
depend on the calibration, linearity, and stability of a measuring device.
The balance and bridge depend on the calibration and stability of a
passive quantity similar to that being measured and on the sensitivity
Just as a mass standard can
of detecting the difference between them.
movement.
H. Christie,
with
Wheatstone
Charles
Sir
credits
but
referring to a paper
drawing attention to Christie's idea and with the first application to the
Hague
History.
[16]
in 1833,
resistance balance,
The term
bridge
because points
- = R2
Ri
or
R< =
if
(9-4-1)
R\
may
be deduced from
The
is
Eg
FIG 9-17
bridge.
The Wheatstone
284
its
source impedance.
in the expression.
2.
The balance
detector, the
detector.
remain unchanged.
4. A given resistor (R
u for example) is proportional to the resistance of
the adjacent arms (R 2 and R 3 ) and inversely proportional to the
resistance
(or proportional to the conductance) of the opposite arm
(R A ).
The sensitivity of a bridge can be expressed in many ways
Sensitivity.
by using output-to-input ratios of voltage, current, or power or any
combination. The most useful expression for any given application is the
one that uses the quantities that represent the actual limitations of
applied power and detector sensitivity. A general expression for voltage
ratio near balance is
Eo
8Rd
Ei
RX
+ R2 + RS + RI + Rd (2 + R /R + R /R,) + R 2
+ Ri/R, + Rt/R + R dR (l/Ri + 1/R + l/R + 1/R
x
x)
g(
4)
(9-4-2)
by
and a very highimpedance detector, in which case the expression is very simple and gives
maximum output voltage for a 1 1 ratio bridge R x = R s Maximum
output power for a given power in the bridge is achieved when the resistances of the detector and all bridge arms are equal. If the bridge arms
are
unequal, reversing the source and detector connections will give increased
power out if it results in a better impedance match between the detector
and the bridge output resistance, which, near balance, is
:
+ Rt/Ri)
+ Ri/Rz
Ri(l
1
Impedance Measurement
FIG 9-18
sistor
four-terminal
285
re-
measured on a Wheatstone
bridge.
9-4-2
Measurement of Low-valued
The measurement
Resistors
is affected by
the resistance of the connecting leads and terminals, so that low-valued
standards are
Such four-terminal
resistors
Rb
much
percent.
7-
/.Ra)100 percent
286
FIG 9-19
The
Kelvin
double
bridge.
which
zero
is
small
if
Ra and Rb
by adjustment
are large or
if
this expression is
made
equal to
The voltage Ey
rY
r4
r5
+ ty
(R b
Ri
n)R A - (R a
Ra(R(i
r 3 )R B
100 percent
(9-4-3)
Rb)
The second
factor
is
(r Y
oo).
The
tance omitted.
FIG 9-20
Warshawsky's
tance bridge.
rcsis-
Impedance Measurement
9-4-3
287
Megohm
Bridges.
of resistors
up
of such a bridge, one has the choice of either using very high valued
resistance
is
if it is
a rheostat.
proportional to
its
conductance.
unknown
the
Guarding.
by series
by shunt-
leakage resistance.
(duals).
between the terminals of the bridge and its internal parts, and across the
unknown resistor itself. High-valued resistance standards are made
FIG 9-21
megohm
bridge
288
The
simplest
(guard) terminal
connection
is
for
this
third
of .Ri
and
is
The extra guard arms can also be added across the bridge from A to C,
where their ratio should be that of Ra to R x
Putting arms in both
directions, tied in the middle, results in a guard whose balance is much
less critical, but requires two preliminary adjustments [19].
.
9-5
9-5-7
Low-frequency Bridges
Ge
In 1865, Maxwell used a ballistic galvanometer to meaby breaking the battery connection to a Wheatstone bridge when one arm contained an inductor. Later he balanced
this transient by adding a second, variable coil to the bridge.
This bridge
History [16],
Guard
Impedance Measurement
289
required two adjustments: one for the steady-state condition, the resistance balance, and one for the transient, the ballistic or inductive balance.
Ayrton and Perry used ganged commutators to reverse both the battery
and detector connections. This not only provided a continuous series
of transients,
it
The advent
sensitive detector.
Zx = Z
This equation
and
inversion.
son of the
= Z Z2 = Z Z2 Y
^
Z3
Z3
is
(9-5-1)
If
that
is
sarily of the
Rx =
R1R2G3
or
L x = RiRzCz
(9-5-2)
If either of
resistance
R,
+ jX
(fl,
+ jX{) 3
or
flj
f&Z
series
is
+ jX
a series combination of a
2
l
(9-5-3)
fZd%
impedance components
A bridge
if
the
290
impedance
if
the opposite
1/Z,
= Y3 =
wise,
it is
+ jB
arm Z 3
is
jX x = (G, jB 3 )Z 1 Z 2 Like7?,
easy to show that a parallel combination in an arm adjacent to
the unknown or a series combination in the opposite arm will result in a
bridge that indicates parallel components, for the balance equation can
also be written as Yx = G x
jB x = Y t Y 2 /Y 3 = YxY 2 Z 3
G,
that
so
3,
In order to
make
real
Both adjustments in the same arm. In this case the bridge reads the
and imaginary parts separately, such as R and X or L and R.
Example
R x +jwL z =
R^G, +jo>C
3)
= RiR&t
+ jcoR
R^G 3
dials, or
other
readouts.
2.
of one
Example
Rx
=
Lx =
4-j'wL,
A,fi 2 (G,
C^,
C0O3
Rx
To read
+ jo>C
3)
A,fl 2 Ca
of its resistance.
read
and Q.
Example
R X + J0>L x
+R Y
fll#2(G 3
+ jut!,)
R x = RiRiG Ry
3
L x = RiRiC 3
All arms of all actual bridges contain some resistance,
Bridge Residuals.
inductance, and capacitance, and as a result their actual balance condi-
Impedance Measurement
291
9-5-2
While
1.
2.
many
list
if
one
in the basic
4.
listed in
Table
9-3.
the
Wien
a series
RC
no "unknown" impedance.
many RC
The
best
bridge used in
oscillators.
vice versa.
9-5-3
Arms
[22, 23]
worst choice because of their low Q, poor frequency characteristic, nonHowever, if two induclinearity, vulnerability to pickup, and size.
tors are tightly coupled, they become the best choice, for they have two
low frequency).
is the mutual inductance (which is nonlinear in iron-core devices).
R represents eddy-current and hysteresis loss and is a function of frequency and level.
at
292
TABLE
9-3
Series Capacitance
Comparison Bridge
circuit.
it 2
D =
components
Series
Parallel Capacitance
wClft!
Comparison Bridge
Cx =
ft,
C,
Ri
ft*
R2
fti
W
3
cally high.
D =
.Cifti
Parallel
Schering Bridge
Used
Cx
in high-voltage bridges
ft7
in high-
because variable
capacitors can be used for both adjust
ments. Used in dielectric measurements
because C 2 gives high-resolution loss
bridges
measurement. A
Ci and ft2 reads
series
ft2
Ci
with a high
Used
frequency
components
Rx =
fti
Dx =
wft 2 C2
combination of
parallel
capacitance.
Series
components
1
Lx =
Rx =
R1R2
^ftT
coils.
Series
components
Lx = L
fti
t
Basic
inductance
comparison
bridge.
Additional resistance added to inductor
with lower Q to get resistance balance.
Series
circuit
shown;
could
also
'
ft*
= Rs
R2
be
parallel.
Series
components
Impedance Measurement
TABLE
9-3
293
(Continued)
Maxwell-Wien Bridge
L x = RiRtC
r' =
components
L x = R1R2C
Bridge
inductance measure-
Rx =
Q =
R1R2
~rT
1
oRiC
all
Lx
-rT
Q = wCR*
Series
Hay
R1R2
Parallel
Series
components
L x = R1R2C1
Owen Bridge
Gx =
C2
R\Ci
Parallel
and
Gx
Owen
is
formed
if
and
Lx
d connect in
parallel.
Series
mM-tu
^1
FIG 9-24
An
for a transformer.
equivalent circuit
components
294
upper one
Capacitance
is
Although distributed,
not shown.
it
can be approxi-
If
coil still
a voltage
maintains
its
attractive features.
two voltages
is
N +N
is
very nearly
)\R
+ jo>Mj]
(9-5-4)
is
This ratio can easily have an error of less than 1 ppm even
quite large.
To obtain an accurate ratio, one should:
1.
Use a
3.
l/M
if
the ratio
Tape wound,
The material
l/M
is
Use
size for
both
(JVi/AT 2 )(r 2
+ jh) + (Nt/NiH^+juh)
+ N )/N ]Z
[(N 1
If iVi
result
if
2,
this
is
/o
inductance only. Thus, inductively coupled ratio arms are very accommodating: They exhibit a very high value of
and R to provide an
excellent ratio and to present a high impedance to the supply, but they
Such behavior is,
exhibit a very low effective impedance to a shunt load!
of course, due to their coupling and the resulting reflected impedance.
Ratio accuracy and immunity to loading are the reasons that "transformer" ratio arms are used in three-terminal capacitance bridges, such
Here, one stray capacitance shunts the
as the one shown in Fig. 9-25.
Bridges are made where
detector, while the other shunts one winding.
Impedance Measurement
295
tC,
FIG 9-25
,
!q
three-terminal ca-
coupled
ratio arms.
error
would be
less
low errors described above and is preferred for high-accuracy measurements on low impedances because the impedances of the other two bridge
arms do not affect the transformer ratio. Usually a second transformer is
used to drive such a bridge to provide isolation.
Third-winding drive [23] (Fig. 9-266)
This circuit is particularly good
when comparing high impedances because the open-circuit voltage ratio
of the two secondary windings depends only on the ratio of the mutual
inductances between these windings and the primary, which can be very
accurate, and not on winding resistance and leakage inductance.
However, here these residual winding impedances are part of the other bridge
arms and, thus, cause large errors when low impedances are measured.
Lynch connection [24] (Fig. 9-26c). In this autotransformer connection,
full input voltage is applied across one-half of the transformer, and as a
result, loading across this winding has no effect on the balance condition.
Thus, this connection is useful when extreme loading may be present such
as in in situ [25] measurements.
However, this connection has the poorest
.
la)
FIG 9-26
(b)
Methods
(c)
296
FIG 9-27
compound
trans-
former connection.
impedances
An
of the driven
in series
with Zs-
interesting
advantages of
secondary of
primary is nearly equal to the voltage applied to the series connection of
the other windings.
Because only a small current flows, the voltage ratio
is quite independent of winding resistance and leakage inductance.
What is shown in Fig. 9-27 as two transformers can be made as one
unit by using two toroidal cores.
The primary of T\ is wound on one core,
but the secondaries of 7\ and the T 2 windings are formed with windings
that enclose both cores.
Transformers are not very linear devices, and therefore one would not
expect that the source and detector of a bridge using them could be
reversed without changing the balance conditions.
However, as long as
saturation is avoided, the difference caused by this reversal is very small
arm
1.
bridges:
The
leaves one
arm
for the
Impedance Measurement
297
z,
NiT~
H
t IC
\nz
z,
zs
FIG 9-28
former
in
excellent
The
YA
trans-
two
of
Special-purpose Bridges
Hague
poses.
is
[16] discusses
Many
method
of adjust-
ment, and even though some have additional network branches, they
can be reduced to one of the common four-arm bridges by use of the YA
transformation. Another catalog of these bridges would be of little use,
but a mention of the basic problems is worthwhile, along with some of
the more interesting solutions.
High-impedance Measurements. The techniques used for high-resistance dc measurements (Sec. 9-4-3) are applicable to ac measurements,
techniques of making three-terminal measurements.
Measurements on small capacitors are subject to large errors owing to
the shunting of the unknown by stray capacitance unless a guarded three-
particularly the
little
effect.
Likewise,
Wagner
guard circuits for precision capacitance measurements have been available [28].
However, the use of transformer-ratio-arm bridges in the last
298
TABLE
9-4
Arms
An
main
balance.
Nt
is
deviation reading.
Cx =
Cs^
D = wR s Cs
(GRJ
Digital-summing Adjustment
The
1615)
C z = aiCi + a t C 2 + a C 3 +
d = IOC2 = lOOCj etc.
D x = uR(Ci +C2 + C +
z
"\ V
s
many
is
used on
Cx =
- C
c3
Gi(2a
Gx =
)C 3
+c +
Ci
1)
+GiR(a)(l -a)
YA
transformation of
lated inductance.
for balance unless
"Spoiler" Rz
Rx >
L x = RiRtC
R x = R1R2G
Ri
if
is
necessary
R2.
R3 = Ri
C3
fl3
differential
C,
(BEC^f 75C)
Differential-capacitor Adjustment
t*
*z
-f-
^2
X10
Impedance Measurement
FIG 9-29
299
four-terminal bridge
an accurate current
ratio.
1
decade has made this measurement very convenient and has almost completely replaced the use of guard circuits which require additional balance
adjustments.
Low-impedance Measurements. Likewise, the techniques used for lowmeasurements may be adapted for low-impedance ac measurements. Several commercial bridges make four-terminal measurements on large capacitors simply by placing the lead resistances in
the adjacent arms as illustrated in Fig. 9-18. Also, the Kelvin double
bridge has been adapted for ac measurements [26, 29] even though it may
require two adjustments in both the main and auxiliary bridge arms
resistance dc
former type of three-terminal bridges described above. Just as a transformer can produce voltages of precise ratio, it can also produce currents
of precise ratio.
In Fig. 9-29 transformer Ti has no flux if Nih =
2I 2
However,
and, therefore, presents low impedances to these currents.
any unbalance current is presented the full open-circuit inductance and
thus
is
ages which
make
not be precise.
FIG 9-30
four-terminal
ca-
*CS
300
EY
drop
component
There
is
some
disagreement on whether a series or parallel representation of this comThe parallel representation is used in
plex impedance should be used.
Here mutual
Fig. 9-24 because it simulates circuit behavior better.
inductance becomes the effective parallel inductance of the open-circuit
transfer impedance between a pair of coupled coils.
This mutual inductance produces a special kind of four-terminal impedif Fig. 9-24 is compared with Fig. 9-2, one sees that:
Because the coils may be tied together at one end, there are only two
"lead" impedances, and the network forms a T instead of an H.
2. These lead impedances may be very large, larger than the mutual
impedance itself, unless the turns ratio is unity.
3. The mutual impedance may be positive or negative depending on
ance, for
1.
the connection.
inductance.
is less
In spite of the
than l/uC(Ri
many
R2).
mutual impedance is
measurements. From
series of two-terminal
lc
y&
FIG 9-31 Campbell's frequency
bridge used to measure lossless
mutual inductance.
jp-
FIG 9-32
null network for
measuring lossy mutual inductance.
Impedance Measurement
301
if
mutual impedance
is
first
Applying
Bias
The capacitance
The capacitance
Voltage or Current.
is
applied.
of
many
capaci-
of semiconductor
and
resistive devices
Two
feed.
With
tance.
series feed
no error
is
the current
is
limited
There
is
Ri.
feed,
must be made.
but
unknown and
added
circuit,
its
is
is
applied
Their use provides additional flexibilmainly by their ability to provide isolation and phase inverUnity gain amplifiers, both inverting and noninverting, with
302
FIG 9-34
long-term stability of better than 0.01 percent, are possible at low frequencies.
The most critical part of the inverting amplifier can be the
passive components used for current summing.
introduced.
9-5-5
FIG 9-35
An
Impedance Measurement
303
is
required.
ments
of an ac bridge
is
One
interesting commercial
CRL
bridge [36] uses the nonlinearity of biased diodes to make an automatic loss (D or Q) balance and a phase-sensitive detector to facilitate the
balance of the main adjustment, which is manual. As a result, the value
main balance
made
Digitally Balanced
detectors on the bridge output, and their settings provide information for
a digital readout.
9-6
9-6-1
As frequency
increases,
many
components
304
switching
considered.
As a
have variable
Chap.
9-6-2
17).
General Radio Company RF Bridge (GR 1606) [38]. A favorite for use at
radio frequencies is the Schering bridge, because both null conditions may
be met with the use of only variable capacitors. An example of a commercial instrument with this circuit is the GR 1606 of Fig. 9-36. This is
a substitution bridge, for one adjustment C p is in series with the unknown.
Both components of the unknown are read as difference measurements.
Using subscripts 1 and 2 to refer to successive balances made with the
terminals shorted and then with the unknown in place, we have
Rx =
~Z~ (Ca2
Cai)
(9-6-1)
to
\Cp2
Cpl/
The dials associated with both adjustments are calibrated in ohms, but
the reactance dial must be divided by the frequency in megahertz to
FIG 9-36
series substitution
Company 1606).
Impedance Measurement
305
RX meter (Hew-
FIG 9-37
The
lett-Packard
Company 250B).
MHz
RX
Company
Meter (HP 250-B) [39]. This instrument
a parallel substitution device measuring admittance.
It is shown in
Fig. 9-37 as a transformer bridge with opposed T networks.
The four
Hewlett-Packard
is
main arms are similar to those of a Schering bridge while the small capaciC 5 and C 6 are used as a summing connection to the detector. This
instrument includes an internal oscillator and detector and operates from
tors
500
[40].
M M
,
proportional to the currents in the susceptance standard B s the conductance standard Gs and the unknown admittance Gx
jB x Because
the same voltage is applied to all three impedances, the detector is nulled
,
when
M (G + jB
MG Gs + MBjBs
(9-6-2)
The dials associated with MG and M D are calibrated in terms of G and
B and M
used as a multiplier to extend the range. This instrument
X
x)
x,
is
x is
9-6-3
T Networks
1,500-MHz range.
[42]
306
FIG 9-38
An
Company
602).
mon input and output terminal (without the use of transformers) that are
netbased on the simple T configuration and, therefore, classified as T
their
frequencies,
low
used
at
been
works. While these networks have
main application is at radio frequencies where their common input-output
coupling.
is very important to avoid unwanted
setting
null conditions for these networks are easily obtained by
network
all
of
admittances
transfer
short-circuit
at zero the sum of all the
of the
paths from source to detector. If there are only two paths, the sum
terminal
The
short-circuit transfer
must be used
The Bridged
to balance resistors.
T.
If
a single
T is shunted by an impedance Z
4,
a bridged or
Impedance Measurement
Two
FIG 9-39
shunted
Zi
bridged-T null
circuits.
The balance
established.
is
+Z +
3
307
+Z
expression
is
(9-6-3)
The two
Two T
Twin-T Networks.
impedance of one
filters,
networks in parallel
Many
twin-T networks
most important one for impedance measurements is
shown in Fig. 9-40. This is not the same twin T used in many oscillators
and filters. Its main advantage is that the two grounded variable capacitors can be used for the two necessary adjustments.
The balance equais
tions are
d + C + C +
2
+RG
(1
5)
= --
(9-6-4)
and
"fl,CiC,
04:)
=G
Ci
t-zf
Gzi
<CZ
A'
-&
<X>
Gsk
FIG 9-40
&*
>B'
Hf-
The
tance-measuring
*e
twin-T
circuit.
admit-
308
from which
B xA =
Gxa
where
tors
A
that
AC 2a
o)
AC 2a and AC
two capaci-
unknown
the
commercial instrument
(75
CxA = AC 2a
~ Re ACtA
when
or
It
0).
was
[43]
measurements on dielectric samples, over a range from 460 kHz to 40. MHz.
The same basic network has been used for precision measurements [44]
with additional terminals* B and B' across C&. With the unknown connected to these B terminals, the admittance of the unknown is
B xB =
co
AC 5 b
or
CxB = AC5 b
C AC 2B
(9-6-5)
Gx
CidRe
where AC 2 b and AC 5 s are the changes in these settings when the unknown
added to the B terminals. If a given unknown is measured first on the
A terminals and then on the B terminals, Gxa = Gx b = Gx which may be
determined from the expression
is
Gx 1
co
AC 5A AC 2B
(9-6-6)
9-6-4
Resonance Methods
circuit elements
Impedance Measurement
resonance with a known conductance Gs connected and noting
peak voltmeter readings. Now
Gx = Gs
This
is
what
FIG 9-41
ment
V {V - V
- V
Z
VM
is
309
all
three
2)
(9-6-7)
t)
Parallel-resonance
measure-
circuit.
The unknown
is
place
first in
This susceptance variation method has been used for precise dielectric
measurements [45]. In one method [46] a negative conductance,
made from an active circuit, was used to cancel out most of the loss of the
loss
-*-
6
<>
FIG 9-42
Series-resonance
measurement
circuit.
inductor and so to
circuit.
The duals
310
They
With
is
may
be determined
difference.
FIG 9-43
The
The
meter.
method
is
lossless,
Eo
= E
it is
(9-6-8)
fl,+i( w L,-l/C)
^=
Ei
ccR x C
=
R
(9-6-9)
Q,
The meter
9-6-5
The
of this
methods
of Sec. 9-3
may
measures
0.
Impedance Measurement
n
U
Voltoge
sampler
5- kHz
Det.
ampl.
t
Pulse
gen.
zx
Phose
fc
*1
det.
-H
*
Current
sompter
311
5-kHz
4-
Det.
ampl.
rf
mod
rf
osc.
FIG 9-44
The
rf
vector impedance
meter (Hewlett-Packard
9-7
Precision
9-7-1
Measurements
Determination of the
of Standards
(NBS)
Ohm
is
[48, 49].
and dissemination of standards for all units of measure, including the elecimpedance the ohm and the related units of capacitance
and inductance. In 1864, long before NBS existed, a British committee
led by James Clerk Maxwell made the first determination of the ohm.
They gave values to certain wire resistors, based on the mechanical
system of units and Wilhelm E. Weber's proposed electromagnetic system
of units.
While several of these resistors were distributed to various
trical unit of
made a need
Soon
NBS
made
in
after
Germany and
1.
312
realized.
several
The determined
relates inductance (and capacitance) to resistance.
value differed from the international values, and in 1948 the absolute ohm
came into use by international agreement. It was 495 ppm smaller than
the
mean
495
ppm
international
in the
new
units were
larger.
Since then, additional determinations have been made with the use of
more recently, the Thompson-Lampard cal-
culable capacitor
[50],
which
is
more
shown
axis,
easily
made from
It
Then a
came within
of light.
National
Bureau
of
Standards
Calibrations.
Ohm
determinations are
Impedance Measurement
reference set of
is
313
maintained by a
are intercompared
is
revised regularly.
needed.
Traceability.
The word
traceability is
national standard.
necessary to meet this goal, it is not sufficient to ensure high accuracy, for
a standard may change in value between the times it is measured and used,
and good measurement technique is essential, especially when meaIn spite of these and other
suring odd values at odd frequencies.
inherent limitations, the concept of traceability has greatly improved the
general level of measurement accuracy, particularly at the componentinspection level.
Associated with traceability are two additional terms: recall cycle and
ratio.
The first represents the time period between calibrations
accuracy
which is left indefinite by some standards. The United States Navy [54]
however, has established recommended recall cycles for a large variety of
commercial standards and instruments. Accuracy ratio is the ratio of the
accuracy of the device being tested to the accuracy of the standard used
While a high accuracy ratio results in a short tracefor calibration.
ability chain, in many cases it is not feasible, particularly at higher
accuracy
possible
9-7-2
levels.
Many
measurement
Methods of
error,
Precision
Measurement
314
ing accuracy:
1.
2.
3.
Measurement
While the
first is
of
odd values
it is
accuracy required.
Precision Resistance Measurements [19, 55, 56].
One-to-one resistance
comparisons are most often made on a Kelvin direct-reading ratio set,
such as the one shown in Fig. 9-19. Both resistors being compared are
measured on the same terminals with an auxiliary standard or "tare" on
the other side of the bridge.
Because the Thomas resistor has a 1-fi
much less important This limits the adj ustment range to a rather
small deviation, and as a result, such ratio sets can only measure resistors
tance is
of approximately the
Decade
by connecting 10 standards of one value in series or by a 10:1 ratio in a ratio set that
is calibrated by such a series buildup.
The error for such a procedure is
the error in the 10:1 ratio multiplied
is approximately 95 2 /100.
This procedure allows
_
extreme precision and scaling from 1 Q to 10 kO with
PP m precision.
Odd
Impedance Measurement
315
accuracy requirements for inductors are correspondingly low. One-toone comparisons can be made on high-resolution Owen or Maxwell bridges
ppm at medium values, but scaling or odd-value measurements on such bridges are limited by the accuracy of the bridge arms and
other bridge errors. Series connections of standard inductors are sometimes used to provide scaling, but residual lead inductance, shunt capacitance, and low Q values limit the accuracy.
Perhaps the most precise inductance measurements are those made on a
precision capacitance bridge by measuring the change in the effective
capacitance at the unknown terminals as an inductor is added in series or
in parallel with a capacitor [63], but residual parameters and low Q values
to within 1
limit this
method
similarly.
While three- and four-terminal measurements are commonly made on low-valued admittances and impedances
respectively, many instruments contain even more terminals.
Several
Multiterminal Measurements.
316
[65].
REFERENCES
2.
3.
York, 1960.
Gibbings, D. L. H.:
1.
Blackburn,
J.
Series, p. 65,
Ratio, Proc.
IEE
{London),
PL
4.
5.
Ibid., p. 226.
6.
Dummer,
7.
8.
AC/DC
Resistance
John Wiley
&
Sons, Inc.,
New
9.
Blackburn,
J.
Series, p. 115,
10.
Hersh,
J. F.:
14.
15.
16.
17.
18.
p. 948,
of Precision
Impedance Measurement
317
22.
23.
24.
C: A Bridge Network for the Precise Measurement of Direct CapaciIEE {London), Pt. B, vol. 104, p. 363, 1957.
E. C: Impedance Measurements and the In Situ Component Bridge,
Lynch, A.
tance, Proc.
25.
Crawford,
26.
IEE
Ratio
Arms
T.
R.,
Network with
Precise
September, 1963.
31. Hall,
H.
P.:
The Measurement
Radio Experi-
Pitman
&
April, 1965.
38.
Soderman, R. A.:
Notebook,
Summer,
1954.
13, 1950.
Impedance
at Frequencies
up
to 30 Megacycles, Proc.
IRE,
July, 1940.
44.
Huntley, L. E.:
Self -calibrating
NBS
318
Impedance
Co.
Notebook, no.
48.
Winter, 1955.
Silsbee, F. B.: Establishment and Maintenance of the Electrical
Units, NBS
Circ. 475, June, 1949 (available from the U.S. Government
Printing Office
Washington, D.C.).
Silsbee, F. B. Extension and Dissemination of the Electrical and
Magnetic Units
by the National Bureau of Standards, NBS Circ. 531, July, 1952.
Thompson, A. M., and D. G. Lampard: A New Theorem in Electrostatics and Its
Application to Calculable Standards of Capacitance, Nature, vol.
166, p. 888, 1956.
Cutkosky, R. D.: Evaluation of the NBS Unit of Resistance Based on a' Computable Capacitor, J. Res. NBS, A, vol. 65, p. 142, 1961.
Thompson, A. M.: An Absolute Determination of Resistance Based on a Cal4,
'
49.
50.
51.
52.
53.
54.
55.
4,
no.
NASA NTC
1,
p. 107,
January 1968
Pomona,
Thomas,
Calif.
J.
York, 1952.
Hamon, B. V.:
and
Certification,
IEEE
60.
IEEE
61.
62.
McGregor, M. C, et al.:
Measurement, IRE Trans.
New
63.
ber, 1960.
CHAPTER TEN
AUDIO FREQUENCY
SIGNAL SOURCES
Donald
E.
Norgaard
Audio-frequency
signal
generators
as convenient sources of
maintenance, or operation of
Although the term audio frequency
serve
is
responsive,
many
audio-
both higher
and lower in frequency than the usually accepted audio range from 20 to
20,000 Hz. In general this extension in range is accomplished by techniques identical with those employed for the audio range. It should be
noted that the discussed generators provide electrical signals only, not
acoustical signals.
70-7
319
320
3.
4.
Low
1.
2.
spurious output,
i.e.,
hum,
noise,
jitter,
modulation,
etc.
The questions of how low the harmonic and spurious outputs should be
and how stable the frequency and amplitude should be depend, of course,
upon the application for the signal generator. Many applications do not
require extremes in
some
do.
all
The general-purpose
is
audio-signal generators
100
mW. Whenever
electrical energy from some source such as a battery or rectifier and converts a portion of this energy into its output signal.
This signal is a
cyclic repetition of electrical voltage or current of predetermined fre-
waveform
in analysis
and
test of a
The
by which such
signals are
321
produced
and controlled.
Practical sinusoidal signal generators can be classified by the method of
frequency control into the following types: (1) the LC oscillator, (2) the
RC oscillator, (3) the beat-frequency oscillator, and (4) special methods.
It will be noted that the term oscillator is introduced as a means of
The logic of this terminology
describing the first three classes listed.
70-2
The
LC
Oscillator
power
oscillation
the
in
LC
The vacuum tube, with its capability of power gain, was applied
by Armstrong to supply energy lost within a continuously oscillating
LC circuit by a regenerative association of the tube with the circuit.
The low-power signal generator became a practical reality with this
combination.
The impedance
characteristics
of
inductor-capacitor
circuit
voltage.
the
when
it is
J
a
fig 10-1
T~
|
b
Generalized
antiresonant circuit,
322
mechanism
control
useful:
L =
Y =
c
Ya =
we
let
If
<
jo>C
n+
10 - 2 - 2 >
LC =
l/w
-j+.
F = juC
c
+ jR p
(a>
LC -
Rp
uL
1)
4)
- 2-
then
(1 - 2 - 5)
ZLRP
the impedance seen between points
~ 1/r~ "
z =
and 2
HjftWW-i)
\za \/_*
2o)
d<f, t
o>
2R P
d4> z
From Eq.
<t>
uqL
at the frequency
(10- 2 " 6)
form becomes
i2 p (w
/wo
-1)
(10-2-7)
is
oioLR p {u 2 /ui
+ fl p
when
2
(
w/ojq
/a>
-l)
(10-2-8)
1)
(10-2-9)
(10-2-7)
= -
rad
4
o>i
slope
LR p /w -
d(co/uo)
d(o)/o)
in polar
/tan- 1
\z\
resist-
(10-2-3)
" uL
is
Rp =
= \/ 1.
ZL
co
then
at
and
(frequency),
2ir
Ftot
a)/2ir,
The
capacitance in farads,
(10-2-1)
Ftot = Ya +
If
C =
=-j
Yb
co
in a signal generator.
inductance in henries,
ance in ohms,
LC
of antiresonance
when
R P (l -
wi
is
WiVcoo 2 )
determined as follows:
coiL
(10-2-10)
i,
+ VLW/JV +
_ -La>oVfl p
Similarly,
rad at w
(10-2-11)
4coq
(10-2-12)
0)iU>2
Wl
mean
of
toi
and
to 2
is
(10-2-13)
Wo
W2
4to
= w 2 when
+ VLW/flg +
LwoVflp
It follows
7r/4
is
<f> z
U>2
323
o> 2
and
coi is
Wo
(10-2-14)
tip
The
VwiW2
RP
O>oi>
0)1
10-1
(o> 2
on) is
may
RP
when u = w
factor
to the difference
(10-2-15)
Equation (10-2-15)
resistance
mean
Thus
is
defined
Q = R P /o)oL, where Q
to = too-
is
the quality
In terms of
d<t>z
Rp
_2^=
n
is
-2Q
(10-2-16)
tooL
d(to/too)
Thus, when
03
= w and
and
co 2
and
(10-2-12),
^
oj2
The
wi
= Qo = -
_^_
(10-2-17)
2a{a/uo)
relationship between
output impedance
324
FIG 10-2
LC
oscillator.
of the amplifier
If it is
real
G/a X
/J#i|Z|/*,
where |Z[/^,
is
ing frequency.
G/a\Z\/j> t
The voltage
= Ex
is
(10-2-18)
Equation (10-2-18)
may
be expressed as
= -o
(10-2-19)
G/a\Z\/4>.
when
driving
its
load \Z\/^ t
is
(10-2-20)
Equations (10-2-6) and (10-2-7) indicate that -x/2 < <j>, < tt/2.
Therefore, the allowable range of a lies between ir/2 rad if the system is
to be capable of self-sustained oscillation.
Since the frequency_of oscillation departs from the nominal frequency
/
co
/2tt
l/2ir
VLC
frequency stability
(for
in order to
a given
LC
is
a,
greatest at
co/g>
where
<f> z
itself is zero.
Instabilities in
a is zero.
Equation (10-2-19)
calls for
when
will
have
G/a
for
Audio Frequency
any stated values
of \Z\/
and
j8
Signal Sources
325
of oscil-
A value of
when starting, and then \GZ\ must decrease to satisfy the equation at
some predetermined level.
In most LC oscillator designs amplitude limiting occurs because G
eventually becomes a decreasing function of the magnitude of either the
input signal or the output signal of the amplifier. Energy storage in the
reactive circuit elements L and C of Fig. 10-2 tends to keep the output
noise
to be maintained.
Eq. (10-2-19)
is
output-voltage
satisfied
on an average basis
at
some reasonable
amplifier-
level.
is
more consequential
due to change of
shape of the amplifier output current pulse or its time relationship to the
input signal j8Ji. The operating frequency then responds to this phase
shift in order to satisfy Eq. (10-2-20).
in a change of effective phase shift within the amplifier
326
LC
expense
system
is
anti-
from the
The implementation
of
many
specific
oscillator itself.
of the
LC
oscillator
arrangements possible.
LC
antiresonant
circuit for
in all regenerative
LC
frequency control.
The
essential principle
to that delivered to
regenerative
LC
useful output
power
10-1
is
shunted by
resistance, then a
same
result at
some chosen
LC circuit
accombe
may
an
LC
circuit
Fig. 10-2.
only a portion of their possible voltage ranges; outside these ranges the
characteristic resistance becomes positive.
Thus, the average value of
negative resistance so obtained is voltage dependent, with the result
that the level of oscillation produced is a function of the equivalent
327
up
builds
by the negative-
LC
LC oscillator.
oscillator
is
It
Vl/C =
o>
= w
LC
oscillator.
(10-2-21)
chapter,
why
it will
clear
vl0:l and
it is
328
an
70-3
transfer characteristics of
Zt
= R
-^-
(10-3-1)
o)Ci
iRil&C*
Ri 3/^2
/E\
jRz/(j)C2
R1R2
(l/a> 2 CiC 2)
- j(Bi/uC*
+ Rt/uCx + R
/o>C 2 )
it 2/ 03C2
R1/0C2
+ R*/uCi + R
/wC 2
+ j{RxR 2
l/ w 2
dC
2)
(10-3-3)
Audio Frequency
When the
quantity RiR 2
For given values of JBi, R 2
w2
then
0,
this occurs
2 is
in
329
when
(10-3-4)
RiR 2 C 1C2
which defines w
in
terms of
circuit parameters.
=tC,
RC network.
FIG 10-3
Now,
bers,
C h and C 2
O)
l/w 2 CiC 2
Signal Sources
if i?i
k and
E2
(10-3-5)
Ei
+ m + 1 + j Wkm
maximum
(or peak)
(a>/a>
co
/w)
at /&j
may
where
<f>
is
tan
km
fc
plot of
<t>
+m+
wo/w)
(10-3-7)
1
versus
co/co
(co/coo
1,
at which value
<t>
<t>
The
by the
RC
RC
circuit.
330
100
,/ErB&!
Phase
80
1.0
3+/'
(ai
/ay
u)
/ujT~
60
40
/X\
P >v
20
-20
<dXa>
X-
OCT
-60
-80
-100
R*\
\*
fL_
04
-a:
si
-40
*nrF
<
07
-Nj^-
0.3
0.2
0.1
Relative
magnitude
0.01
0.1
I00
oj/oj
FIG
0-4
in Fig.
value of
7r/4 rad.
applied to the
Vkr
\W
Solving for
02
Wo\
C0 2
+m+
of
circuit
/Ei
is
RC
circuit con-
t/A
at
o> 2
(10-3-7)
when
(10-3-8)
co 2 ,
(l/\//cm)(/c
"
COo
+m+
1)
V[(l/fcm)(fe
+
m +
l)] 2
+ 4
(10-3-9)
Similarly,
<*>1
-{l/Vkm){k
wo
+m+
1)
V[(l/km)(k
+m+
l)]
(10-3-10)
The apparent Q
V w wi
Vfcra
w2
Q =
H when
A;
wi
is
+m+
= m =
1.
(10-3-11)
1
Audio Frequency
The normalized
Signal Sources
d tan-1 [V
km
(w/o
a>
/w)/(fc
may
331
be obtained by
+m+
1)]
(10-3-12)
d(co/a>
d(w/wo)
V km
-1
d<f> u
d(co/co
[fcm(w/
u /w)*/(k
+m+
l)
2
]
fc
+ wo/co
+m+1
(1
2
)
(10-3-13)
At
w/ojo
1,
2 vfcra
d<f> a
d(u/o}
+m+
(10-3-14)
1
d0u
d((d/a)
= -2Q
(10-3-15)
10-4
[1]
:*i
%t
TCz
Wien
FIG 10.5
^/?4
brid 3 e.
E* =
k
Ez
2.
From
(<o/o>
(10-4-1)
/w)
XI4
Ei
is
and
+ m + 1 + j Wkm
#u3
XI3
where
-f"
Zt4
(10-4-2)
332
The
bridge output
E%
Ez Ei
is
+m+
+ j \Zkrn
(w/oo
co /w)
COn/w)
(10-4-3)
If
- A
0o
(10-4-4)
where
is
(10-4-5)
+m+
A;
Ev
Ez = Ei
+ m + 1 + j *Vkm
(co/wo
-j-
o/w)
m+
It follows that
\k
co
"
J&2
-C(3
= Ei
_(fc
+m+
l)
km(u/(*)
w /) 2
(10-4-7)
Equation (10-4-7)
may
A
where
<t>B
is
a real
- w /)
Vkm + m + l)
+ m + l) - fcm(l/A - l)(tt/o - wo/co)
2
(1/A)
tan-1
(fc
(Jo
(a>/o>
(10-4-9)
When
w/ojo
1,
may
be
written
y/ki
d<l>B
d(/w
Now, by
A(/c
+ m+l)
(10-4-10)
d<t> B
d(o)/o>
(fc
+m+
(10-4-11)
1)
A(fc
of the bridge
+m+
1)
arrangement Qb
333
is
A
(10-3-15).
balanced),
w/wo
1 is
its
output.
1,
^-^ = Ex
3
(10-4-13)
1,
amplifier
El
= - =
A
E%
(10-4-14)
Ez
can be made very
large, the
respondingly small.
Under conditions
of steady-state self-oscillation,
A precludes
self-oscillation,
while too
small a value will cause the strength of the oscillation to increase to a level
at
the average voltage gain of the amplifier conform exactly with that
stated.
oscillator
W. R. Hewlett
[2]
by providing one
of the
arms of
334
is a function
element an almost perfectly
linear one during the period of the intended oscillation.
Hewlett used a
tungsten-filament lamp which has a positive temperature coefficient of
FIG 10-6
Wien
it,
while
still
making
this
brid 3 e oscillator.
The self-heating of the filament due to flow of signal-frequency current serves to control the average resistance of R t so that /3 of
Eq. (10-4-2) is dependent upon signal level. In this manner, A becomes
dependent upon the operating level of the system, so that Eq. (10-4-14) is
resistance.
The
made very
is
to permit the
high,
of the amplifier.
Audio Frequency
Signal Sources
335
In general, amplifier phase shift is slightly different from zero. Therethe system usually does not operate at exactly w
regardless of
amplifier gain.
Since the departure of operating frequency from the
design value w /2ir is a function of amplifier gain, frequency stability is
linked with gain stability of the amplifier at all operating points where
amplifier phase shift is other than zero.
Amplifier phase shift is generally
fore,
stability.
70-5
The foregoing
aspects of a successful
many
of the practical
Wien
A commonly
that a single set of calibration marks serves for several overlapping decaderelated frequency ranges, which are selected in steps by switching fixed
resistors (Ri and #2) simultaneously.
The output of the oscillator is the
bridge driving voltage El
This output may be taken directly through
it
may
336
is
generator.
ment described
earlier.
these designs a ganged pair of variable resistors permits vernier interpolation between the smallest steps provided by the switching arrange-
ment.
The impedance
amplifier input
Operation of the thermally controlled Wien bridge oscillator at frequencies lower than about 1 Hz is generally impractical because the
thermal element is a source of odd-order harmonic distortion. If the
Zi
is
+z
= S.(
at every value of
Zx
+Z
o>
the
Zi
+Z =
2
Hl)(l-iV^)
m+
where
w/co
(1W1)
1.
If
fc
ra
1,
% V2 Ri/^-
minimum
Hz when k = m =
If
value of
1.
1,200
is
(10-5-2)
20 pF, then
Thus, when Ri
V2 /
Cl
= R 2 = 800
= Ri =
800
12
at 10 7
Q,
(10-5-3)
Audio Frequency
Signal Sources
337
In addition to this reactive load, the amplifier must drive the remainder
# 4 of Fig. 10-6) and any external load as well. The
oscillator operates at w only if the phase shift attributable to the amplifier
of the bridge (R 3
shift at
Since
it is
high frequencies,
its
made
condition of balance.
Amplifier gain and phase shift must be controlled outside the intended
operating frequency range in the interest of satisfying the Nyquist
stability criterion.
Failure to do so can result in spurious oscillation
which is not subject to control by the R 3 R 4 mechanism, and which is not
,
and C 2
The frequency
constants.
Its
amplitude,
and only
uncontrolled,
drives
the
primarily
is
slightly influenced
by bridge
amplifier
into
saturation.
bridge oscillator
quency
Wien
limit of the
bridge oscillator
is
The
high-fre-
most
spectrum.
10-6
[3,4]
The three-section low-pass ladder structure of resistance and capacitance shown in Fig. 10-7 exhibits a voltage transfer characteristic of such a
nature that
FIG 10-7
capacitors
W
F!
may
Three-section
lator.
it
RC
ladder.
and three
29A
10 " 6 - 1 )
338
Ganged
FIG 10-8
oscillator.
at the frequency
(10-6-2)
2ttRC
This network
will
29.
The frequency
The
the
network. Frequency stability, therefore, will be closely related to
in
used
product
of
RC
value
phase stability of the amplifier at any fixed
the ladder.
70-7
Audio Frequency
RC
utable to each
RC
section.
its
shift of 60 in
phase
and
and
section
The combined
RC
is
\/S/2irRC,
system
The
if
2.
loss
is
is
is
controlled automatically
stabilized at a value
where amplifier
acceptably low.
Frequency
capacitance
120,
is
339
section
arrangement
effective
each
Signal Sources
of oscillation is controlled
or the resistance
by ganged variation
of either the
as
by
signals
of
the ring.
The nominal
where amplifiers
is
is
and phase
tively simple.
As
value stated.
system
value of
-3 V3/4.
tion
is
0-9
Output
FIG
The apparent Q
oscillator.
No Q
multiplica-
340
It is practical to provide a
of greater
than 10
automatic-gain-control action.
oscillators are
10 kHz.
70-8
made
to cover a desired low-frequency range serve as the basis of the beatfrequency oscillator. In its elemental form, the beat-frequency oscil-
lator comprises
Fig. 10-10.
The mixer
process
Signal
Signal 2
= E
= E2
1
Mixer output
sin
mt
sin
w 2*
(10-8-1)
(10-8-2)
= EiE 2
sin
a>i
[cos
sin
(coi
o) 2 t
u 2 )t
cos
(a>i
0)2)2]
(10-8-3)
is
Mixer output = k
dc component
[cos (wi
on)t
cos
(o>i
u> 2 )t]
(10-8-4)
where k
signals.
is
The
341
two high-frequency
frequency
than that of either of the
two high-frequency sources. If, for instance, the desired beat frequency
is 1 percent of the frequency of one of the high-frequency oscillators, a
shift of 0.1 percent in the frequency of either oscillator results in a 10
between those
of the
is,
in general, poorer
instability
is
oscillators of the
by the
(Pulling
other.
is
generally
with the
practicable
heuristically
simple
arrangement in
Fig. 10-10.
It is necessary that
(/i
/2 )
<
fz/2,
where f2
<
/i.
Consequently,
Oscillator
(A
f\
)
<
Low-pass
A
/'
\
\
r
(/ ,
0-1
filter
f,
Output
Oscillator 2
FIG
'
Mixer
f.-f9
t
+ / 2 ),(/ 1 -^).
Beat-frequency oscillator.
amplifier
342
unbroken continuum
ratio) is available,
Many
permits.
of
magnitude than the desired signal. Since these spurious signals are all
higher in frequency than the desired signal, the lowest cutoff frequency of
the filter consistent with uniform response over the desired range is
indicated.
70-9
two
will
in Eq. (10-8-4)
Here
an analysis:
is
= kE*
= kEn
cos
(a>i
cos
(o) 2
ui)t
ui)t
(10-9-1)
when
In this case,
An
and
all
signals
may
sin out
and Ei
sin
corf
respectively.
mixer
sin (w 2 t
1) will
a) (displaced
produce
= kE% cos
= kEi cos
[(wi
[(w 2
os^t
a]
oj\)t
a]
(10-9-2)
Audio Frequency
Signal Sources
343
<h
High-frequency
0.-360
oscillator
phase
ISfo)
shifter
Mixer
2
Filter
ifo)
Mixer
Filter
-^9
(f
(fa)
Frequency
bus
:Af
control dial
9^0
Variable-frequency
>Af
fir:
oscillator
(fo-Af)
fo
fo-Af
-\ooAf
-io"
..
1st-decade module
{\Of
u O
-Afl
Band pass
Mixer
-o -^
E?
-10
filter
<j>
L.T3
.J
2nd-decade module
\Of
-Af)
Bandpass
Mixer
filter
-MO
Output
otten
Reference
phase output
FIG 10-11
Zf
/>
polyphase beat-frequency
oscillator.
344
quency
/o
oscillator.
The
filtered
The
quency
signal at 9/
decade module.
2. A bandpass
9/o
is
shown
oscillator in a mixer,
+ /o
A/,
filter
selects
and supplies
of the variable-fre-
the
sum term
this signal,
.
10/
of
A/,
the
mixer output,
to a divide-by-10
frequency divider.
3. The output of the frequency divider, a frequency of (/ A// 10),
is brought to a range-selector switch and to the second decade module,
where it is mixed with the 9/ signal for a repeat of the process of mixing,
filtering,
4.
and
division
by 10
as indicated.
and so
forth,
A// 100),
(/
in the
A// 1,000),
two-channel
The
process
shown
Audio Frequency
Signal Sources
345
frequency ranges.
Any
capabil-
10-10
Sine-wave Synthesis
series of straight-line
[6]
percent.
Symmetry of the sinusoid about its coordinate axes requires
use of only seven different lengths to perform straight-line synthesis
within the error stated.
sine
shown
in Fig. 10-12.
Triangular
wave input
Ro A
\RP
Ru
Ri&
6A
3A
1A
dc-bias
source
Synthesized
wave output
6B
5B
4B
-WSi
^6B
FIG
0-1 2
^5B
3B
-V\ArR-*o
2B
IB
(Reference)
346
1A through
is
manner when
diodes IB, 2B, and so forth, progressively become conducting and serve to
switch resistors
of the
It
tion of the synthesized signal for a specific set of fixed bias potentials
and a
The
process
is
i}
is
supplied to a
Audio Frequency
Signal Sources
347
Bistable
flip- flop
VC 2
VC,
^3
* Reference
potential
FIG
0-1 3
Triangular
wave output
Triangular-wave generator.
reference potential.
is
to change
that which
required.
This mode
is
cycle, interspersed
or discharging of an
348
Reference
Attenuators
AA
-d=
FIG
0-1
<\t-
wave synthesizer
is
wave may be considered the integral of the square wave, while integration
of the triangular wave an even number of times gives a sine wave as an
approximation. Such a signal generator is commonly called a function
generator.
&
is
The
effect of
shunt resistance
loses
The
symmetry and thereby degrades the quality
wave
as well.
triangular
wave thus
Audio Frequency
Signal Sources
349
Fig. 10-13.
control voltage,
CITED REFERENCES
1.
Bauer, B.: Design Notes on the Resistance Capacity Oscillator Circuit, HewlettJ., vol. 1, no. 3, November, 1949, and vol. 1, no. 4, December, 1949.
Terman, F. E., R. R. Russ, W. R. Hewlett, and F. C. Cahill: Some Applications of
Negative Feedback with Particular Reference to Laboratory Equipment, Proc.
IRE, vol. 27, pp. 649-655, October, 1939.
Nichols, H. W.: United States Patent 1,442, 781, Jan. 16, 1923 (filed July 7, 1921).
Ginzton, E. L., and L. M. Hollingsworth Phase-shift Oscillators, Proc. IRE,
Packard
2.
3.
4.
6.
February, 1941.
Crawford, R. A Low Frequency Oscillator with Variable-phase Outputs for Gainphase Evaluations, Hewlett-Packard J., vol. 16, no. 11', July, 1965.
Brunner, R. H.: A New Generator of Frequencies Down to 0.01 CPS, HewlettPackard J., vol. 2, no. 10, June, 1951.
:
CHAPTER ELEVEN
OSCILLOSCOPES
Charles H. House
Hewlett-Packard CompanyColorado Springs, Colorado
upon the
Oscilloscopes
351
on mechanical
ing high-speed variations that could not be demonstrated
"variable cura
constructed
had
apparatus. By 1897, Karl F. Braun
of the
forerunner
first
the
tube,
Crookes
using
the
rent apparatus" by
modern
An
oscilloscope.
oscilloscope
is
an instrument capable
of presenting a
One
luminous
set of electrical
electronic
instrumentation displays,
In recent years, however, the definition of an oscilloscope has been
used
altered in common usage to mean primarily an electronic instrument
with time on
for display of electrical signals on the vertical axis, compared
the horizontal
axis.
oscilloscope
is
a time-
phenomena.
and medical,
as well as
The chapter
many
11-1
The function
displaying,
more
may
be defined as capturing,
All instrument
oscilloscope are refinements in one or
first
[1, 2, 3].
transient
Even
352
Vertical
amplifier
Signal input"
T
I
Power
Gate
supply
CRT
amplifier
h"*"t
-ri
Ext. trig,
'
Trigger
input
Ext
FIG 11-1
horiz. input
Time
Horiz.
base
amplifier
oscilloscopes
still
serves
this function).
2.
Vertical amplifier
(including probe
or transducer to obtain an
electrical signal).
3.
Time
base.
Horizontal amplifier.
5. Trigger or sync circuit (to start each sweep at a desired point in time
for the signal to be displayed).
4.
6.
Gate amplifier
(to
turn
CRT
beam
is
swept
Laboratory oscilloscopes
distinctions are based either
may
be
classified
many
ways.
Usually the
on frequency-response capability or on
CRT
MHz
Thus
loscopes,
CRT
used.
Oscilloscopes
353
Oscilloscope CRTs
77-2
CRT
The
is the most distinctive component of an oscilloscope since it is
In recent years,
the obvious output or display portion of the instrument.
much research has been conducted to obtain a better display means than a
CRT, and several alternatives have been developed, including electroluminescent panels, solid-state light-emitting arrays of gallium arsenide
For the present and foreseeable future,
diodes, and plasma cells.
however, it seems clear that today's refined versions of the original
Crookes tube will continue to dominate the displays found in oscilloscopes,
because of advantages in cost, brightness, and speed of response.
Cathode ray tubes may be classified in several important ways. By the
number of independent electron beams a tube is classed as single-beam,
dual-beam, or multibeam. There are differences in beam-deflection
There
plate
CRTs.
The absence
deflection plates
persistence
CRTs.
and
for
between
internal-graticule
external-graticule
CRT
intensity,
many
applications
and
size,
[4, 5, 6, 7].
coordinates;
the
controls the
beam
beam
velocity;
is
beam-postacceleration
section,
354
Beam
Beam
Beam
Beam post-
focus
deflection
acceleration
generation
Beam
target or
screen
Limiting aperture
Crossover
Fj na
aperture
/Horizontal plates
Fluorescent
screen
Grid-
Cathode
fyar^rj=^s-^p. ^:
"Anode
r*
N- Focus
Vertical plates
+*-
q
Bulb wal
FIG 11-2
volts,
The
intensity of the
by adjusting the
beam
at the target
is
is usually driven by
the z-axis amplifier in a grid-blanking CRT.
It is also possible to make
a double anode aperture, with intermediate deflection plates that may be
used to deflect the beam to miss the second aperture. Because the beam
not stopped at the source but merely deflected enough that it is not
seen at the screen, this method is termed deflection blanking.
One advan-
is
critical as in
life
reduction,
Beam
cost saving.
Focus.
to cause the
The two usual electrodes are the focus and astigmatism electrodes.
Although these two electrodes interact to some extent, in general it is
may
Oscilloscopes
355
z/-axis
More commonly,
deflection patterns.
control
CRT
Electrons emitted from the cathode surface are subject to a lens action
which leads to a minimum beam cross-sectional
area called crossover (Fig. 11-26). The beam then reexpands within the
at the grid aperture
lens,
which
index of the
beam
minimum
so that
it
Many
CRT.
and performance
which accounts for
Magnetic
the great variety of CRTs that differ only in deflection method.
deflection allows a wider beam-deflection angle than does electrostatic
Moreover, when the full-screen beam-deflection bandwidth
deflection.
desired is less than 20 kHz, the electromagnetic deflection system (amplifier
and CRT) has a substantial cost advantage over an electrostatic system.
Thus television sets, many medical monitors, and some oscilloscopes rely
Beam
Deflection.
field
At
high frequencies, inductors with few turns are necessary to obtain fast
current changes, and larger currents are required to obtain the required
Consequently, above repetition rates of 20 kHz, large
field strength.
power dissipations are required to obtain full-scan displays. Electrostatic deflection, involving voltage charging of capacitive plates, is capable
of speeds several orders of
magnitude higher
for a
comparable amplifier
CRT cost).
CRTs
deflection or sensi-
(the
number of
356
(the
is
number
of screen divisions
before
often achieved
the
beam
full-
tion plate
it
may
is
itself).
Figure 11-3 illustrates these three parameters, and
be seen from the first diagram of the figure that the deflection Y
length
may
DF
be written as
DF =
e.
where K\
is
(11-2-1)
ah
a constant determined
by the postaccelerator
field.
and resolution
much
sen-
-4
4 -2
+2 44
Horizonta cm
Distance along axis
I
fdJScan
-2
+2 +4
Vertical, cm
centered oppositely)
linearity
Target
Maximum
Beam
Beam
(c)
FIG 11-3
Cathode-ray-tube parameters.
scan
(d) Bent
plate
scan
Oscilloscopes
357
Neck pins
i(amplifier connection)
Neck glass
of CRT envelope
FIG 11-4
Even
1^
(b) Distributed
or
curved-plate
and
(Fig. ll-3c
designs
Note that
d).
(Fig. 11-4),
may
of
(full
be increased by using
parallel-plate
same tube
length.
structures
Other electrostatic
and
effects for
distributed plates
extremely high-
The
Postacceleration.
is
instead
speed deflection
Beam
'
deflection-plate
region,
plates
bent-plate
same
t_
beam
brightness at fast
Monoaccelerator designs are intended for low-frequency applications; the name suggests that no accelerating field is found
deflection
rates.
is
the potential
first
type,
field
and a spiral, or
with a variable field potential represents the second type. They all
use a large positive bias voltage V p beyond the deflection plates to increase
the beam velocity' and energy to obtain a brighter dot on the target.
helix,
The monoaccelerator
CRT
of Fig.
In such a tube, the beam is given all its acceleration before it passes
through the deflection plates and arrives at the screen with sufficient
velocity to present a bright display for relatively low writing speeds.
In electrostatic CRTs, the voltage Vu from cathode to deflection plate
is usually less than 4 kV in order to keep deflection sensitivity high.
Consequently, at higher writing speeds, the beam must be accelerated
358
(a) Monooccelerator
Helix
FIG 11-5
after
deflection
accelerator
which
CRT
to
produce
bright
display.
shaped-field post-
upon
or, in effect,
To
Oscilloscopes
359
same
of the
length.
Carried one step further, the helix can be eliminated and the mesh
shaped into a configuration that will expand or magnify the display.
The result is a short tube that behaves like a long one, Fig. ll-od. This
CRT, called a high-expansion (or high-magnification) tnesh tube, achieves
mesh
made
The
which
electrodes of the
is
first
glass envelope.
The
target
end).
Typically,
FIG 11-6
tion
CRT.
all
(Hewlett-Packard Company.)
made through
360
The
target;
it
is
When
observed.
the excitation
beam
is
removed, a phosphorescence remains for some time and indicates where the
phosphor had been stimulated into light emission. Standard CRTs use
refreshed phosphor targets, so named because the phosphor must be
restimulated or refreshed before the phosphorescence has decayed in order
Storage CRTs by contrast do
not require continual refresh to maintain a displayed trace.
to avoid a blinking or flickering display.
P39
as radar)
CRT
Oscilloscopes
TABLE
361
11-1
Color
1
Phosphor
Relative
writing
speed
Phospho-
type
Fluorescence
rescence
Persistence
PI
P2
Yellow-green
Blue-green
Yellow-green
White
Green
White
Medium
Medium
Medium to
medium short
45
60
50
35
70
75
Scopes, radar
P4
P7
Blue-white
Yellow-green
Blue-med.
45
95
Radar, medical
PU
Blue-violet
Blue
Medium
25
100
Photographic
P15
Blue-green
Blue-green
Visible-short
15
25
recording
Flying spot
scanners for
P18
Blue- violet
Blue- violet
Very short
P18
White
White
Medium
P19
Orange
Orange
Long
Relative
luminance
Display
application
Scopes
Black & white
television
short
Yellow-long
short
UV-very
TV
short
0.1
25
Flying spot
scanners for
18
35
Low-frame rase
25
TV
television
P22
3-color do t pattern,
red, blue
Medium
Color
green
P26
P28
Orange
Orange
Yellow-green
Yellow-green
P31
P33
P39
Green
Orange
Green
Green
Orange
Green
Radar
television
Very long
Long
Medium
short
Very long
Medium to
medium long
17
50
100
20
50
50
75
7
40
Radar
Radar, medical
Scopes
Radar
Computer
graphics
than does a Pll, P16, or P19. Camera film is often sensitive to ultraviolet or blue; hence for high- writing-speed photography, a Pll or P16 is
well suited.
Monochromatic television using P4 phosphor (black and
white) is certainly more pleasing to the home viewer than a P31 green.
The kinetic energy of the electron beam is converted into both light and
heat energy when it hits the target. The heat gives rise to phosphor
burn on occasion, which is damaging and sometimes destructive. Excessive beam current density for a period of time at one spot can degrade the
light output of a phosphor, and in extreme cases can burn a spot in the
phosphor which amounts to complete phosphor destruction. Phosphors
may be classified according to burn resistance as low (P19, P26, P33),
medium (PI, P2, P4, P7, Pll), and high (P15, P31). Thus, the typical
choice of P31 for oscilloscope CRT phosphor is partially based on color
(maximum eye response), short persistence (to avoid multiple-image
362
when the image moves rapidly), and high burn resistance (to
avoid accidental damage by the operator) as well as on its high luminance
displays
level
Aluminizing a CRT (depositing a thin layer of aluminum on the nonviewed side of the phosphor) accomplishes three goals. Its original
function was to avoid buildup of charges on the phosphor, which tends to
slow down the electrons and limit brightness. Also, aluminizing serves
to reduce light scatter (emission of light in all directions)
With
is
reflected again
shown
when
the
beam
brightness, as
phosphor burning.
considerable
beam
in Fig. 11-7.
acceleration potentials,
although
is
it
CRTs
significantly
is
low at low
higher at higher
potentials.
Light
filters
smoke-gray
filter will
reverse.
may be
filters
Black
Filters are also used to reduce glare from ambient lighting.
wire-mesh filters restrict the viewing angle and thus minimize reflections
from oblique light sources (such as overhead lights) while also enhancing
the contrast ratio.
Polarized
filters
Spot Size and Luminance. Spot size is not only determined by crossover-spot size and image-to-object ratio, as earlier indicated, but also
A uminized
Standard
phosphor
FIG 11-7
/
f
minance.
/
J
10
Accelerator voltage
12
kV
14
Cathode-ray-tube
lu-
Oscilloscopes
363
^ = 3kV
250 l-<25
^, =
17kV
C 200 - t20
3
.o
o
l!
>^-
Spot size
150
15
Beam current/
u
r
100 - 10
/
'
50 -
y*
''Light
output
>
30
20
10
40
50
FIG 11-8
Cathode-ray-tube spot
size.
charge
Above 40 V, space-charge
electrons in the
beam) begins
effect (the
to cause
torily achieved.
the
CRT
in
on
beam
film.
trace (single-shot
It is affected
waveform
not only by
display) that
CRT parameters
364
d*iign*.
(Hewlett-
and
sensitivity.
(such as
beam
is
the eye.
energy,
filters,
more
similar fashion.
Grticulet.
number
CRT
screen
The
by the appropriate switch settings
and time duration between any two points
beam
deflection.
Projected graticules
CRTs
cm
8.0
cm
to 8
10
Larger-
Oscilloscopes
screen electrostatic
many
available for
77-4
CRT
CRTs up to 8 X
10
in. in size
and
365
oscilloscope measurements.
Storage-target Characteristics
storage.
storage
mesh-storage
mesh
CRT
deflection plates
is
This mesh
beam
of the standard
hit.
The
a,
CRT,
charges the
storage target
flood gun,
is
then
charged areas of the storage target allow these electrons to pass through to
the standard phosphor target and thereby reproduce the stored image for
the viewer to observe.
Thus the mesh-storage CRT has both a storage
CRT
is
tion
is
simultaneously
A mesh-storage
contains a dielectric material
deposited on a storage mesh, a collector mesh, a flood gun, and a collimator, in addition to all the elements of a standard CRT (Fig. 11-10).
Mesh
Storage.
CRT
FIG 11-10
structure.
Mesh-storage
CRT
as
as
366
secondary emission.
common
11-11,
is
is
effectively stored.
is
eVc
eVK
Electron volts
Stable--
^^"
s?
sf
\
FIG 11-11
nstable
first crossover
characteristics.
-Stable
1
1
i
^40/
01/
(Flood gun)
{b
Vc
(Collector
Storage-mesh potential
mesh)
Secondary-emission
Oscilloscopes
367
Post-accelerator
electrode
Storage surface
Collimated
(dielectric)
flood electrons
v
Faceplate display
area of CRT
envelope
03
Collector mesh ^
(
based at+i
volts)
-y
FIG 11-12
Phosphor
CRT.
To view the stored trace, a flood gun is used when the write gun is
turned off. The flood gun, biased very near the storage-mesh potential,
emits a great flood of electrons, which migrate toward the collector mesh
since
CRT
approach the
storage target perpendicularly.
When the electrons penetrate beyond
the collector mesh, they encounter either a positively charged region
on the storage surface or the negatively charged regions where no trace
has been stored (Fig. 11-12). The positively charged areas allow the
electrons to pass through to the postaccelerator region and the display
target phosphor.
The negatively charged regions repel the flood elec-
CRT
The stored pattern eventually degrades, primarily because ions genby flood-gun electrons charge other regions of the storage surface
erated
368
Press
erase
Release
erase
+156V
+1 52.7
+142.7V-
<-200 msec
--100->
>
msec
+13.3V
+ 10V
.Storage
/ mesh
+3.3V-
potential
* Storage
i
/
'
surface
potential
Time-
FIG 11-13
quickly charge
The
all
CRT
Oscilloscopes
369
may vary
threshold level
slightly
more
current with a constant time dwell. The useful levels may approach as
many as the 10 required for good television-picture presentation, but this
requires a highly uniform storage surface.
Four levels, roughly representing - 10 (full off), -5, -2.5, and
(full brightness) on different regions
of the surface, are practically achievable.
Phosphor Storage. Although work on storage targets with secondaryemission properties has been conducted since the start of the century, a
storage-target CRT design based on the dielectric mesh target was first
constructed in 1947 by Dr. Andrew Haeff. Some years later, Robert
Anderson developed the bistable phosphor storage target by using similar
principles of secondary emission.
These two approaches comprise nearly
all commercial storage-CRT construction today.
Significant improvements in both types are regularly forthcoming.
is
Electron
beam
FIG 11-14
Target structure of a
phosphor-storage CRT.
Faceplate -
370
electrically continuous.
ing; a typical
Variable Persistence.
CRT
life.
is
capable of using a
on the screen.
This is really a storage mode with continuous electrical control of the
duration of storage. A phosphor-storage tube is capable of erasure
electronically at the end of every sweep, but continuous persistence
control
use), to
show
five or
more sequential
traces (to
new one
replaces
it [16].
Figure 11-15 shows the erase pulse used to obtain variable persistence
As mentioned before,
in the mesh-storage CRT discussed previously.
Oscilloscopes
371
Storage surface
potential
22
Time
FIG 11-15
tion.
surface
storage surface will be about V, while the written areas are near
10 V.
The written areas, still below first crossover, will now attract flood-gun
duty cycle
CRT
of,
trace
is
is
interrogating
in a raster-scan
television)
sequenced according to the xy coordinates of the raster-scan beam locaThis type of signal lends itself well to many emerging industry
needs, such as digital communications [17, 18].
tion.
77-5
General-purpose Oscilloscopes
The general-purpose
deflected
CRT
372
CRT
circuitry.
is
displayed.
test or
frequencies under
MHz.
This
is
for
electrical
ical,
many
Storage CRTs are usually used with real-time display circuitry; their
chief value is in either storing single-occurrence events or providing displays at low-frequency repetition rates without flicker.
All general-purpose oscilloscopes use similar techniques for vertical
amplifiers, time bases, and trigger generators that are fundamentally
The most pertinent
different from the techniques of sampling circuitry.
specifications among different general-purpose oscilloscopes are number of
channels, deflection factor,
common-mode
rejection
(CMR), and
rise time of
the vertical amplifier system; the sweep speeds and number of display
modes of the time base; and the trigger circuit capability [19, 20].
Bandwidth and Rise Time. Just as electronic circuits in general have
of bandwidth even though pulse-transient
so the oscilloscope is ordinarily described
meaningful,
more
analysis is often
Hz. This means that at X Hz, the
by a rated bandwidth: from dc to
vertical amplifier is down no more than 3 dB from its dc or zero frequency
Since an oscilloscope is an instrument often intended for pulse
gain.
analysis, it is more directly specified in capability by its limiting rise
historically
time" or
beam
deflection rate.
Most
Rise time
from 10 to 90 percent
is
rise
time
TT
B=-
exists:
tf-0.35
(11-5-D
Oscilloscopes
where bandwidth
35-MHz
is
oscilloscope
in megahertz,
is
and T r
is
Thus a
in microseconds.
373
may
rise
time.
Since
T Td
where
[(T r .)>
is
ra
(T,
2
)
(11-5-2)
This
may
Tr
r,
{(Trdy
For example,
(T roy\
(11-5-3)
rise
if
time
frequency.
The low-frequency
two
and high
MHz). Although the basic amplifier and time-base circuits are similar,
low-frequency oscilloscopes often feature very low deflection factors (high
high-frequency oscilloscopes have good rise-time
Thus, low-frequency design specifications deal with noise
levels, drift, CMR, and small nonlinearities since the signals to be measured
are quite small.
High-frequency design parameters include CRT
writing speed, fast rise time, good high-frequency pulse response, and fast
sensitivity),
while
performance.
trigger capability.
374
because of the time delay difference between the vertical amplifier and the
time base. A postaccelerator CRT is usually required both to view highspeed waveforms with low repetition rates and to present adequately
bright multichannel displays.
Because of the added expense of post-
compartment
flexibility.
Amplifiers.
all
three
{x, y, z)
deflection axes of
The purpose
and the
CRT
of
an amplifier
is
to provide gain
controls are necessary for calibrating the gain of the amplifier and also for
modifying the gain to view signals of different amplitude. Additionally,
the input impedance should be high enough so that the circuit or signal
under test is not seriously loaded by the oscilloscope input impedance.
Finally, because ac signals
some controls
may
levels,
deflection,
compensated
RC
attenuator
is
is
Without
all
compensation, high-frequency signal measurements would, always have to take the input-circuit
RC time constant into account. This would be very restrictive since the
time constant would change on each range, and errors would be noticeable
at frequencies as low as 3 kHz on many oscilloscopes.
The input impedance for most oscilloscopes is 1 Mfl + 1 percent, shunted by a small
capacitance (typically between 10 and 80 pF), which leads to the term
high-impedance attenuators. Very high frequency oscilloscopes sometimes
use attenuators terminated by 50 Q, which are more desirable at frequencies
where transmission lines are used. Adjustments are usually provided on
this
Oscilloscopes
375
U/w-l
J
Zi(s)
Ri
=
1
~ ~
Zds)
+ sRid
R
Zl
+Z
~
2
R,(l
Choose R1C1 =
(l
aRiCi)
+ SR2C2
+ sRrd)
+ Ri(l +
SR2C2)
!f
et
R2
Ri
R2
FIG 11-16
To ampl input
10 pF
FIG 11-17
is 1
376
o=*o,-*o 2
-100V
% 1*
|fff--500
Bo lance
FIG 11-18
Differential amplifier.
+1
Note that
CRT.
/,
-,-',+',
first
jt3
changes by about
voltage across
i\
Re
Rl2.
Since
mA
is
tEi
signal
Also, through
i\,
CMR
Oscilloscopes
+X V
If a signal of
is
(X
and
gives
ZV
>
Common-mode
By
Y).
contrast,
if
>
volts of differential
+XV
of differential output (Z
Y _
Z
377
is
X).
rejection
is
is
100-dB
level
important when
common-mode
signal of 10 V.
measurements
Such
Common-mode
rejection
is
fields
usually
One
fact affecting
parameters such as
effects
by comparison.
measurements.
In such
tests,
a standard signal
unknown
may
comparator
The
CMR
the
unknown input
signal.
Measurements with
such amplifiers may approach 0.1 percent absolute accuracy, with relative
accuracy limited primarily by the reference standard [25].
Multiple-trace displays are of great advantage in oscilloscopes, for they
enable the operator to make time and amplitude comparisons among
Such measurements, useful in production test and
several waveforms.
The operator
service areas, are indispensable for many laboratory tests.
usually desires the horizontal time scale to be the same for all channels
378
Sweep
x
axis i**
axis
Retrace
<\
.,
iChanne
1
\\
Channel
1
|
1
\
Intensity
bright
!
.
z axis
1
I
I
K- 50u
r*-100/j. -i
sec
Time
FIG 11-19
in
sec
ALTERNATE
operation.
while the attenuators and position controls of the vertical channels are
independently controlled. Both dual-trace and four-trace vertical displays are commonly available.
channel.
The
mode
is
easiest
way to obtain
commonly provided.
is to provide an
between two vertical preamplifiers (containing attenuators and positioning) and one vertical deflection amplifier (Fig. 11-21).
This allows for time sharing the beam between the two preamplifiers as
desired.
Several modes of operation may be derived from such a dualchannel amplifier: channel a only, channel b only, dual-channel
electronic switch
Oscilloscopes
379
Sweep
.innnnn^inpiuinnnnniinii
Channel
Chop transients
Intensity bright
Chop transient
Intensity off
blanking
FIG 11-20
in
CHOP
operation.
B, or the difference A - B.
alternate or chop), the sum of A
In dual-channel displays, the A - B differential mode serves as a
(either
compromises
A - B
obtains the
(Fig. 11-22).
0ff-Q7V
-100 V
+15V
Channel
-100
V-
|200
*!
9k7j-10V
<7
3V
^ 15V|
1
-0.7
FIG
1 1
-21
A dual-channel
10
|i ,6 k
electronic switch.
mA
differential
380
XV
(Part of
function switch)
(a)
A-B
ov
(b)
FIG 11-22
ferential (A
B polarity switch
A+B or A-B
Channel
Joey
A-B
(allows
B)
amplifier operation.
CMR
it
capability of a similar
does not allow an additive
CMR
preamplifiers.
of
CMR
is
available in the
Oscilloscopes
A B mode
381
whereas 60 dB is achievable
with the first type. Also, the first type typically is
able to achieve a wider dynamic range for overdrive common-mode
for
comparable
effort
Deflection Amplifiers.
systems have a number of restrictive design parameters in highfrequency oscilloscopes. The gain-bandwidth product of the output
device, /r, is often a determining factor in the ultimate deflection speed.
Usually the constraints imposed by the CRT capacitance Cp ar>d the
deflection factor
VD V
CRTs
per division are even more important in determinFor very high speed systems, distributed-plate
line of
Fig. 11-23.
An
pin,
current
v<
kI
ie dt
(11-5-5)
is
interval.
of the device,
an equation
for the
minimum
full-scale deflection
Device
Device
Load
*G)
CRT
A,
FIG 11-23
total capacitance C g
C, = CT the more
C<*
required to obtain the same voltage deflection in the same time
Since center screen v e times i c equals the power dissipation Pd
1 4-
time
may
382
side
is fci/2
divisions,
which gives a
maximum power
dissipation of
Pd
(11-5-6)
is
T,
-g-
Thus, the
rise
(11-5-7)
may
be expressed
T
r
(11-5-8)
The sequence
many
intrinsic problems.
an amplifier greatly affects their interaction and their consequent use for modifying signal displays on screen.
For example, if the position control precedes the gain vernier, any
vernier rotation attenuates a dc position offset toward center screen.
If
the controls are reversed, no position shift occurs relative to center screen
when gain
is
of controls within
changed.
signal ground.
sequence
either
is
The
method may
Time
is
Bases.
Time-domain
is
oscilloscopes require a
with display gating functions are termed time bases. A high-quality time
base may feature sweep-time variations from 10 nsec to 5 sec per division,
with time accuracy from range to range of better than 3 percent and
Oscilloscopes
CRT
383
display
Down vern
IOV 8VC
Signal
Position
-5V
20V
Signal
Position
vw
Assume
at a starting voltage
16--
OV
K)-"OV
K)-"
+ 5V
l/
plates will
beam
upper
move
goes from
is
the
left to right,
is
CRT screen.
traced.
As the
signal to
the gate amplifier turns on the z-axis intensity, which allows the sweep
trace to be displayed.
The right-to-left movement, called retrace, or flyis blanked out.
A constant-current source charging a capacitor will
produce the desired sawtooth waveform if a discharge path is provided
(Fig. 11-25).
Relaxation oscillators, with neon tubes or unijunction
transistors, provide an inexpensive sawtooth generator with low accuracy
bootstrap techniques allow much greater linearity with additional expense.
back,
The Miller integrator is the most common time-base generator in laboratory oscilloscopes today (Fig. 11-26).
It basically is an operational
Closed
Sweep
Retrace
Ground
b) Waveforms
circuit
FIG 11-25
384
Gateo-|
input
[X
Block diagram
Vout constant on
all
ranges
JWo-'KRC
Schematic
FIG 11-26
Miller integrator with 24 selectable sweep speeds
(1 00 nsec per division to 5 sec per division in a 1 , 2, 5, 1
sequence).
amplifier with the Miller feedback capacitance to convert a step-function
input into a sawtooth output. Historically, special-purpose vacuum tubes
such as the phantastron, sanatron, and sanaphant were developed to
facilitate time-base generator design based on the Miller integrator.
One
choice of both
Fig. 11-28.
and C
Oscilloscopes
385
Single sweep
Reset
Schmitt
Trigger
Sync, signal
generator
Ho Id -off
Miller
integrator
'
Horiz.
amplifier
Zaxis
FIG
1 1
-27
Other
common sweep
circuits include
an auto
circuit
time
sweep
setting.
free-run mode
which restarts the Miller integrator immeA single-sweep mode is possible also,
which allows the sweep to run once from a trigger pulse after which the
This mode is of
gate generator is not reset until the operator desires.
is
possible,
1.
Sync
signal
2. Trigger pulse
3.
Gate generator
4.
Zaxis signal
output
-A
i$$?
CRT
display
5.
Reset Schmitt
6.
Holdoff
7.
Sweep output
Time
FIG 11-28
relate to points
on
Fig. 11-27).
386
broad variety of measurements formerly' accomplished by dualoscilloscopes is being done today by single-beam oscilloscopes with a
beam
CRT
di<
play
Trigger
^A
Sweep
JLA
g,
b2
C
l
-1/isec/div
A\
^A
A
A Sweep
B Sweep
===="
(c } Mixed mode
Sweep
1
(d j Switched mode
(A,B)
50 n
sec/div
Upper //sec/div
1
A Sweep
//sec/div
Sweep
Time-
=i
Lower 50 n sec/div
axis displays
relations
of four
common
Oscilloscopes
387
'i"^o
tc
FIG 11-30
of
DELAY mode
time-axis displays
[29],
time-base
ramp generated by
pulse at time
it
is
started
by a
trigger
generator.
delayed sweep
time-interval
for the
knob
control.
388
reference
If
,
the
a single pulse
is
to be
measured
is
from a
centered on
and the pulse time delay is then, calculated. If, for example, the
main sweep time is 10 /xsec per division, and the division delay setting
screen,
6.215 divisions, the pulse time delay is 62.15 /xsec ( + 3 percent for
sweep-time accuracy) or 62.2 /xsec 1.9 /xsec.
If a reference time-mark pulse is used in conjunction with the pulse to
be measured (with a dual-channel vertical scope), the accuracy may be
increased substantially. Assume, for example, that a reference pulse
occurs at t +50 /xsec. If the delaying sweep time is set to 2 /tsec per
division, the CRT display should show the two pulses separated by 6.1
divisions, which then gives a difference reading of 12.2 /xsec 3 percent for
sweep-time accuracy and 3 percent for deflection nonlinearities and
resolution.
This results in an overall measurement of 62.2 sec + 0.7 Msec.
Further improvement could result if the division delay knob is used first
is
many
mode is very similar functionally to the delay mode, except that the
delaying sweep generator is displayed first. An external trigger is
generated at time h, whereupon the delayed sweep generator (at a faster
sweep speed) is started. Once it reaches the same ramp height as the
delaying generator, it is displayed instead for the rest of the sweep cycle.
This mode is especially valuable in pulse-train studies, when the time
interval of a specific set of pulses in a complex pulse-code train is of
interest.
dual sweep
sweep
Oscilloscopes
389
a time-mark generator.
The
calculations
shown
The
first
method
defines the
interval.
(and perhaps
is
derivations of
use.
maximum
mark divided by the timeThe second method, which gives a larger error number
in disfavor for this alone), involves measuring the max-
mark
sweep
are
common
mark from a
graticule
The
third
by the
difference
Inconsistent sweep
interval widths.
shown in Fig. 11-326 are not reflected by a specification
based upon the first method; this is probably its major limitation. If a
sweep linearity error is of the form shown in Fig, ll-32a, which is a more
typical form, then any of the three methods give comparable results.
Note also that each method includes linearity errors in the horizontal
deflection amplifier and CRT as well ax the time base.
The circuit block responsible for starting the sweep
Trigger Generators.
Trigger
at the desired point on a waveform is called the trigger generator,
between
nonlincarities as
390
(a)
(b)
Linearity measurement
Maximum
deviation be-
Method
1:
Maximum
LOO
interval
Method
2:
Maximum
between
time-mark interval =
smallest time-mark in1.05, average time-mark terval and largest timeinterval
0.95, largest
1.00,
linearity error
1.05
0. 95
inor
= 10%
LOO
10.00, linearity
1.05
0.95
mark =
0.05
time-mark
c^
=5%
0.95, largest
time-mark interval =
1.10, average time-mark
interval = 100,
linearity error =
L10 -- 95
Method 3: Same as
method 2, except
divided by the total
= 15%
1%
10.00, linearity
1.10
time-mark interval
10.00
FIG 11-32
graticule
1.00
difference
deviation be-
div, average
005
L00
= 5%
terval
Maximum
0.95
10.00
1.5%
Sweep
linearity.
comparator to set the desired trigger level and a trigger pulse generator
which starts the sweep generator (Fig. 11-33).
The three
The internal
polarity selection, one should be able to derive a signal that will give a
frequency signals.
Oscilloscopes
391
is
characteristics [30].
Complex pulse train signals, such as those found in computer logic and
pulse-code-modulation communications, are very difficult to display on a
scope because of the difficulty of developing a satisfactory trigger signal.
Figure 11-34 illustrates a complex pulse train of seven pulses for which a
stable scope display
sweep time
is
desired.
may
be used to
'Trigger output
to gate gen.
FIG
1 1
-33
392
Start of second
pulse train
5 //sec /div
10//sec/div
ffl
5 //sec /div
10 //sec /div
Time 5//sec/div
FIG
1 1
-34
Complex
pulse-train triggering.
allow the calibrated full horizontal display to be retained while a synchronous trigger is obtained. If the range of the hold-off variation is greater
than the total sweep time (including recovery time), stable triggering is
ensured [31].
Delay Lines. High-frequency oscilloscopes nearly always include delay
lines in the vertical amplifiers.
The purpose of such lines is to delay the
vertical signal enough to keep it from reaching the CRT deflection plates
before the horizontal sweep circuits are running. As shown in Fig. 11-35,
the vertical signal triggers the sweep generator and enables the horizontal
amplifier to begin tracing a sweep at full intensity before the vertical
Oscilloscopes
Signol
input at
Delay line
= 160 nsec
Vertico
preampt.
I
time
tn
ta
fa
=IOnsec
-r
Vertical
deflection
amplifier
fa =40 nsec
393
axis
tv-tc
+210
nsec
CRTdisploy
Sync
Gate
amplifier
=20 nsec
amplifier
td
fa
CRT
Wo+
150 nsec
fa
=50 nsec
Horizontal
Trigger
Time
generator
fa = 20 nsec
10nsec/div
/=
deflection
amplifier
base
fa=80 nsec
axis
<rfo+l50
nsec
fa =20 nsec
fa
FIG 11-35
Delay-time relationships
loscope block diagram.
in
time delay
high-frequency oscil-
may
be viewed
The
first
parameter
much
like
any other
if it is
signal.
/TOT*
numbering 50 cascaded
^TW>Cm\
*TL
>*
>*
/'
Core
Ul
Lfl
Shield
Etched
Dielectric
insulation
Shield
s*
circuit
Cross section
(see below
-Dielectric
Core
-Etched
nnL-^ross
^
^mam
^s
-flom
gg
FIG 11-36
circuit
hbo
w*/
O Current out
Current in
394
perturbations:
(a)
pulse
LC delay segments.
wound
line),
con-
helical coil
to
reflections.
are
shown
in
Fig.
11-37.
Anomaly
2,
called preshoot,
is
a dis-
Oscilloscopes
395
fourth anomaly, one delay time later than the rise-time delay, is due to
radiation or coupling from the deflection amplifier back into the preamplifier.
The
large
power supply
disturbance
is
CRT
lines
The
final
a signal travels through the line to point C and meets a disconback through the line (in Ud ). If at the other end,
point B, it encounters a second mismatch, it travels again to point C, arriving with the distortion seen at 2td
A perfect termination at either end
line.
If
tinuity,
it reflects
would stop
this distortion,
difficult to achieve.
less
than
difficult to discern.
Plug-in
Oscilloscopes.
The
plug-in
feature
of
many
oscilloscopes
CMR
396
HG
11-38
x and y axes.
circuits
(Fig.
Oscilloscopes.
and the
11-39).
size of
CRTs, the
Low-power
circuitry allows
make
size
Oscilloscopes
FfG 11-39
997
(Tek-
tronix, Inc.}
Storage Oscilloscopes.
very
difficult to
of
CRT
unless the
itself.
waveforms
CRT indicators.
Lastly, comparison of
two
CUT
contrast, being able to store one signal trace permits comparison with a
FIG 11 -40
variable-persistence
low-frequency
oscilloscope.
(Hewlett-Packard Company.)
398
11-6
Sampling Oscilloscopes
made subnanosecond
was no convenient way to
Real-time scopes with traveling-wave tubes had been
Recent advances in
solid-state
devices have
analyze them.
developed for such speeds, with the vertical signal coupled to the display
tube either directly or through a distributed amplifier [22]. Scopes with
1 GHz or more ?/-axis bandwidth were produced by such techniques, and
while they were unquestionably of value, they were seriously handicapped
by small CRT display area, low sensitivity, and poor brightness on lowduty-cycle displays.
real-time scopes
domain.
motion.
Oscilloscopes
399
(a)
(b)
FIG
-41
or a CRT phosphor to respond to very rapid pulses of informaand to store the image between pulses so that it appears continuous.
The sampling technique allows the design of an oscilloscope with wide
bandwidth, high sensitivity, and a bright, clear display even for relatively
human eye
tion
low-duty-cycle pulses.
allows nearly
all
the
flexibility of
this century.
Since 1950,
when
J.
M.
L. Janssen built a
35-MHz
strobo-
GHz
may
main frame
400
6j
o-AW>
o\
-o ^tomp
Sampling
^o
^_
-p Us
switch
(a)
FIG 11-42
ground.
is
is
sampling gate
As a
result,
CRT.
plates of the
is
also fed
detected.
Example
input capacitor
is
Oscilloscopes
401
is
in
CRT
will
command
The
x-axis
system
The horizontal amplifier builds the time-base signal to sufamplitude to drive the horizontal deflection plates.
A conventional time base produces a linear sawtooth sweep, to continuously move the beam horizontally across the CRT. The sampling
time base also moves the beam across the screen, but usually not as a
continuous movement. It positions the beam horizontally after a sample
is taken and holds the beam at this location until the next sample is taken.
The beam is then repositioned to a point slightly later in time on the CRT,
where it again remains until the next sample. Thus, the time base is
termed a staircase-ramp generator.
switches.
ficient
Sampler Operation.
The
Sampling
switch
Trigger
input
FIG
11
-43
shown
402
/Transmission
line
-v
Bias
Sampling pulses
Diode
barrier
voltage
rop
HF|-A-*-
-t
Normal diode
back bias
(b)
FIG
1 1
-44
e,
Bandwidth
(11-6-1)
which
-q
is
l/\/2
dc or low-frequency value.
The time interval during which the act of sampling takes place must
necessarily be finite.
This sampling time and the charging-time constant
of the sampling circuit limit the resolution and hence the rise time or
bandwidth that can be achieved. These same factors also determine the
Reduction
transfer loss, or sampling efficiency, of the sampling circuit.
times
its
Oscilloscopes
The manner
403
samples.
it
rectangular with
= -e-^
ec
forO
<
(11-6-2)
T
If an inpulse is now considered scanned by the sampling gate, with the
stored value at termination of the gate taken as the sampled amplitude
and time taken as the interval between the trailing edge of the gate and
the impulse, then a similar wave will be reproduced except that it will be
cut off at a time which corresponds to the gate length T.
= - -'/'
e cl
<
for
< T
(11-6-3)
e el
for*
> T
(11-6-4)
MS) =
When
columns
g(t)
^>>
with t
(11-6-5)
or t
and 2
:>
Sampling
g(l+y2x/T)T/r
...
9
1 + j2irfr
T
<5C
in
are
Table
shown
in
column
3.
above expression
easily since
it is
A(0)
e- T 'r
(11-6-6)
404
TABLE
1 1
-2
Waveforms
in a
CRO
Sampling
g(t)
Impulse
fo^o-y^r-
||
(Somple)
Observed siqnol
Sompling
circuit
T> C
r c<&<
<k
ir-
C/<J,lmpulse response
as g (t) if /
of charging CKTif 7"' C/C
L-7"--l
Atf)- ttTt,
Log
T
I
3 4
TTT
G/C
A < f>-G/e+j2TTt
G
Log
Sin2
A(f) "
freq.
(ttiT)2
ZAT
4,/T
\ \Slop-
lAv
ATft
Scope bandwidth 3 dB down
TrScope
rise
time
10-90%
0.8 T
2.2
C/G
0.554 T
The
between the input and output because of the diode resistance, and a
"kickout," or transfer of the sampling pulse into the system under test,
was quite noticeable.
A four-diode bridge overcomes both problems reasonably well, provided
the diode characteristics are well matched. A bridge combination keeps
the total diode resistance in the path the same as for a single-diode
sampler, since there are two series diodes in each parallel
major
difficulty
sions preclude
maximum
is
leg.
The
speed.
Oici lotcapet
405
received,
on the
level
signal line
is
conveyed
sit
to C,
and stored
Very
there.
fast
in order to
to a .sampling gate.
It
may
somewhat
different
approach
by a known delay line section, replacing each diode in the two-diode gate.
This arrangement allows a wide, strobe or pulse gate, with a narrow time
window for sampling defined by the different responses of the two diodes
due to the line. Consequently the sampled signal is defined as the charge
trapped on the delay line section during the short time window when one
diode
is
is
off.
Potentially this
may
be the
Triggered end
technique
is
may work
at
have
This
to be
requires, then,
FIG 11-45
vertical-amplifier
406
Time
Triggering
level
^ window
H*y^>jLlJ s y
la)
FIG 11-46
Random
signals.
clock,
and
by the same
clock.
avoided, the inherent distortions and bandwidth limitation of the line are
eliminated, and display jitter can be reduced further than is typically
The y (or vertical) coordinate is obtained by the same sample-andhold process used in a conventional sampling oscilloscope. The x (or
horizontal) coordinate is derived as shown in Fig. 11-46.
The figure
illustrates five randomly placed samples of the signal.
Each of these
samples was taken on a successive repetition of the signal.
The y component e v of the first sample is held and subsequently used to
Oscilloscopes
407
CRT
position the
The resulting
along the input signal at t
spot horizontally.
held and later used to position the
repeating this process, subsequent samples supply both vertical
started
by
sample
e x is
By
trigger recognition
CRT
77-7
Special-purpose Oscilloscopes
are
made with
special-purpose instru-
Within the
class of time-base
is
displayed on the x
axis,
the unit
may
be
Often, the
called a swept-frequency indicator or a spectrum analyzer.
x and y axes are simply used for a linear plot of any variable, much like
an xy recorder
display.
Phase plotting
is
a very
common
requirement, as
and alphanumeric readout in computer terminal disSuch displays are also used in component test instruments such
operator.
These
levels
408
HiHHilj|j|j|j|j|jB
FIG 11-47
o o
(Tek-
tronix, Inc.)
The biggest use of digital oscilloscopes today is in automated productionsuch as in the semiconductor manufacturing industry, where
tests of a simple waveform-analysis nature must be
repeated on each component. The digital-readout scope, especially when
provided with programmable controls, is uniquely suited to this application (Fig. 11-47).
As analog- to-digital techniques at oscilloscope test
line testing,
numerous go-no-go
prises the
Essentially a closed-
and design
advent
of
Oscilloscopes
The
distance to a discontinuity
is
ct
409
(11-7-1)
200*
where
pulses.
If
dl-2
c is
transmission
and t
line,
is
c(h
is
tj)
2(e)*
it
less
is
achievable.
line.
a system transmission characteristic that sweptfrequency generator and monitor techniques provide for the frequency
domain picture
of
domain.
[ncident
Step
generator
050 A
J*
Feed- through
sampler
System
Zo
test
under
Reflected
To scope
(a) Reflectometer
Incident
Step
generator
50X1
Transmitted
System
under
test
J~
Z
Feed -through
sampler
50
(b) Transmission
FIG 11-48
To scope
410
The
classic
measurement
of
transmission-line
characteristics
by
ratio
is
calculated.
measurements must be made at many frequencies, a rather timeconsuming and tedious task. Swept-frequency reflectometers are available today to facilitate the determination of standing-wave ratio.
Such units give a magnitude reflection coefficient as a function of frequency, but no phase or location information is available.
Time-domain reflectometry is a powerful design and analysis tool
unmatched by any standing-wave-ratio technique, for it truly separates
the location and magnitude of a number of minor discontinuities conPerhaps more important, the type of
tributing to the aggregate effect.
mismatch at each discontinuity (R, L, or C or a combination) is readily
ratio
determined.
series,
may
-z =50
e^^
/?/=
-t-l^so
4_j
i
a
,
<?EE$
shown
*=50
J- ^so
l
Rl ~Zo
RL+Z
^.50^)
fco
Real impedance
fa J
1-100T
I
50 "J"
&EFB
(b)
Series inductance
zt/z c
mnF\-e'
c- 25
FIG
1 1
Oscilloscopes
=
411
50
100 t
Zcf 50
R-C
Shunt
Shunt R-L
_-t^50(lJ,
Pj
Series
R-C
FIG 11-50
istics
C=
T{\-P)
Of prime value in electrical cable and transmission-line testing, timedomain reflectometry has also been applied to many other situations where
For example, in
location of moving materials in conduits is unknown.
time-domain reflectometry can quickly indioil product (kerosene, gasoline, butane, etc.).
Studies of the effect of light-source stimuli on cockroach behavior were
rendered possible by using time-domain reflectometry to watch the
movements of the insects in a closed system. It is used in England to
monitor soup cauldrons for proper soup homogeneity. Checking thermocouple characteristics after implantation in atomic piles is another
difficult job rendered practical by time-domain reflectometry [44].
The xy Plotters. An external horizontal input which omits the time-base
generator is a common provision on oscilloscopes (Fig. 11-1). This
provision allows the user to plot another variable instead of time on the
x axis. For example, spectrum analyzers plot frequency, curve tracers
plot voltage, and vector cardiology instruments plot phase vectors.
oil-refinery cracking towers,
412
Test
x y oscilloscope
oscillator
CRT
display
FIG 11-51
If the x
with
it.
Figure
shown
in
Fig. 11-52.
Network
under test
f-
kHz
CRT
displays
f- 10 kHz
To
scope
600
0027^ F
10
leoo
!!!!!!!!
mH
!:!!!!!!
10
mH
0.027
FIG 11-52
Phase-measurement displays.
f- 100 kHz
Oscilloscopes
41
BH
between the
spectrum information is necessary. In recent years, the oscilloscope display and frequency sweeper have been combined into an extremely versatile
instrument called a spectrum analyzer. See Chap. 16 and Refs. 45 and 46.
Medical Displays. Variations of the standard oscilloscope are used in
many medical applications. Patient monitoring in surgery, coronary care,
is greatly facilitated by several scope
Various physiologic functions such as electrocardiographic waveforms, arterial and venous pressures, and respiration rates
are monitored continuously by time-base scope monitors, which often
intensive care,
types of display.
is
usually
a large-screen multichannel
tions
pressures
systolic, diastolic, or
mean blood
[47].
41 4
FIG 11-53
ments, Inc.)
the body.
Oscilloscopes
415
CRT
CRT
raster-scan
display circuitry.
need
for
and then go relatively slowly to the next square to repeat the fast sequence.
Many more characters may be written this way than with a conventional
magnetic display unit, yet only a small electrostatic deflection amplifier
and a simple magnetic drive amplifier are required. The CRT cost
in
offsets part of the advantage, and the resulting computer constraints
display programming are also disadvantageous.
Electrostatic displays historically have not found much use in computer
Compared with magnetic deflection tubes, electrostatic
terminals.
deflection CRTs typically have very low beam-scan angles, which means
length must be much longer to obtain the same display
Also, since voltage rather than current is the deflection input in
that the
area.
CRT
416
in. long,
certain types of
conditions over a considerable range, enough that low-level Vbe characterand collector saturation resistance (re sat ) can be examined as well as
istics
BV
As with most
[56].
Of el lloicopci
FIG 11-54
CRT
display.
recent
417
phosphor-itorage
{Tektronix, Inc)
improvement
information on the display panel so that the knob settings may be more
While its major advantage is in locating the
readily noted (Fig, 11-55).
scale information in a more convenient place for the user, additional
benefits include the ease of photographing both the analog picture
and
its
range coordinates, plus provision for some simple derived parameters, such
as the beta division number provided on the illustrated instrument [57],
There are numerous other variations of oscilloscope
Miscellaneous.
types of instrumentation.
is
passband;
TV
line
41 e
FIG 11-55
alphanumeric
(Tek-
ironix, Inc.)
11-8
Accessories
number of accessory items is available to increase the usefulness of the oscilloscope for special applications.
Transducers for signal
large
capture and recorders for a retained hard copy of the displayed phenomena
are the two most important accessories.
Other items include CRT
filters or viewing hoods to reduce light reflections or to enhance contrast,
scope carts for ease in moving larger oscilloscopes, carrying cases for
portable oscilloscopes, and special servicing equipment such as storage
cabinets (with power) and plug-jn extenders.
The following discussion
will be restricted to transducers and recorders.
The transducer
is
requirements:
1 JVliJ)
but,
it
power
lines)
under
A coaxial
it
vertical amplifier.
signal to be observed
This technique
suffers
circuit
but
The
test.
OiclMofcopei
input
fi-\0 MCi
* 10 pF
419
mputC
11.1
pF
Shielded
proie Coble
Probe
-G
Ma
'P
>
To scope
amplifier
GcsMbT
;fr
$7
I
J
Cdiltlib
+ C,=10QpF.
the compensated divider probe has been developed (Fig. 11-56). Such
a probe, available usually with a 10 1 or 50 1 voltage attenuation ratio,
:
under
test
capacitor
is
itself.
The probe is capable
when the adjustable compensating
discussed previously.
is
by
far the
most common
Probe
FIG
1 1
probe
-57
compensated 1 0; 1 divider probe at tower right; 1 ;
upper right. Other accessories include camera beiel
at
{Hewlett-Packard Company.)
420
Electronic
compensation is readily accomplished by displaying the calibrator waveform and adjusting for a flat pulse response {Fig. 11-58). Mast compensated-divider probes are capable of probing signals to 600-V dc without damage, and special high-voltage unite with attenuation ratios to
1,000:1 are capable of probing' to 40-kV levels.
Active voltage probes use an internal amplifier to achieve a 1:1 gain
ratio while providing the attributes of higher input impedance and shielding of the compensated attenuator probe. Their chief limitation is a
relatively low dc dynamic range (0.5 to 5 V) since they use an active
device such as an FET at the probe input.
Bandwidth specifications for probes are somewhat different from those
CRT
rPVKtflii
I probe
T
-I-
i*
(11-8-1)
Lfocope
0.35
DW.
(11-8-2)
*tyetem
At higher
signif-
Probe
under compensated
Proper compensation
Overcompensation
CRT
a is pray
Protse
Scope
Scape
JUL
I
input
khz
FIG 11-58
vertical
Calibration of probe
compenMtion,
Oscilloscopes
421
line technique.
and pulse-wave
electrodes.
Hard Copy. There are two common ways to provide a hard-copy print
Film photography is the
of an oscilloscope display for a historical record.
more flexible and popular method for the occasional user, while xy
recorders have several advantages for certain measurements which
justify their use.
and convenient
it is
CRT, and
The use
the
full
CRT
and a hot
display
development
few seconds,
be used to etch the
of the picture in a
may
in. film
422
mounted
on
oscilloscope,
(Hewletl-
black trace, the background in gray, and the trace in white (Fig. 11-6Q&).
Choice of a camera and film system varies with the requirement almost
as
much
as the scope
itself.
If ultrahigh
writing speed
is
desired in order
to record single-shot
f/1.3 lens,
film sensitivity.
For standard
with a 3,000 American National Standards
equivalent, and a 1 L object-to-imagc ratio are more satisfactory for both
cost and usability.
Film types and characteristics, camera capability,
and mating with the proper oscilloscope for a given measurement are
photography, an f/1,9
lens, film
A paper hard copy from an xy recorder has two advantages over a film
photograph hard copy: It uses a cheaper material than film (of interest if
OiCll lotCOp*!
great quantities of records are required), and
it is
423
The disadvantages
The wide
disparity between the electron-beam deflection rate and the recorderstylus deflection rate requires a frequency transducer
copy
may
commercially,
FIG 11-60
Graticule illumination.
424
CITED REFERENCES
1.
2.
3.
4.
Guillemin, E.: "Introductory Circuit Theory," John Wiley & Sons, Inc., New
York, 1953.
Scott, R. E.: "Linear Circuits
Time-domain Analysis, Part 1," Addison-Wesley
Publishing Company, Inc., Reading, Mass., 1960.
Churchill, R. V.: "Modern Operational Mathematics in Engineering," 2d ed.,
McGraw-Hill Book Company, New York, 1958.
Bakish, R.: "Electron Beam Technology," John Wiley & Sons, Inc., New York,
1962.
5.
6.
1948.
Ore., 1968.
Waters,
W. M.:
Electronic Half-tone
17.
vol. 38,
Nov.
Harsh,
M.
29, 1965.
D.: Display and Storage Tubes, Electronic Inds. Tele-Tech, vol. 24,
April, 1966.
ed., transl.
nical Library,
20. Rider, J. F.,
and
Their Uses," 2d
S.
ed.,
Oscilloscopes
425
24.
25.
Rad. Lab.
22.
23.
Company,
27. Pettit, J.
Inc.,
New
1959.
and H. Taub: "Pulse, Digital, and Switching Waveforms," McGrawBook Company, New York, 1965.
Millman,
Hill
Hayden Book
York, 1964.
J.,
Understanding Delaying Sweep, Service Scope, no. 50, Tektronix, Inc., June, 1968.
30. Zimmerman, H. A.: Pseudo-Schmitt Eliminates Uncertainties in Trigger Logic,
Electronic Design, vol. 13, Nov. 22, 1965.
31. Grein, W.: Real-time Oscilloscope Triggering, Elec. Design. News, Nov. 22, 1967.
32. Lewis, F. D., and R. M. Frazier: "Distributed-parameter Variable Delay Lines
Using Skewed Turns for Delay Equalization," Proc. IRE, vol. 45, pp. 196-204,
February, 1957.
29.
33. Siegel,
F. G.:
Packard
MHz
August, 1966.
Krakauer, K. Magleby, R. Monnier, V.
IRE WESCON
3,
pp. 44-51,
1959.
35. Frye, G. J.: Oscilloscope
vol. 24,
June, 1965.
36.
Howard, D.
L.,
WESCON
An Analysis,
37.
Zimmerman,
A.:
"The State
of the
W. M.: "Sampling
X-band,"
IEEE
for Oscilloscopes
and Other
RF
Systems:
Zimmerman,
A.:
DC
through
MTT-14,
pp.
IEEE Trans.
pam-
phlet, 1967.
IEEE
219-224, 1963.
Time-domain Refiectometry
44.
45. Unter,
426
46. Siegfried, L.
Frequency-domain Oscilloscope
Now
&
Febiger, Philadelphia,
1968.
50.
Bryden,
J.
CRT Displays,
Computer
Ore., 1968.
53.
House,
C:
vol. 19,
December, 1967.
Large-screen High-frequency
X-Y-Z
Display, Hewlett-Packard
J.,
54.
Winningstad, C. N.: The Simplified Direct-view Bistable Storage Tube in Computer Output Applications, Proc. Natl. Symp. (8th), Soc. Inform. Display, pp.
55.
A New
May, 1967.
Look in Information Display, Tekscope, vol. 1, June, 1969.
Kurshaw, J., R. D. Lohman, and G. B. Herzog: Cathode-ray Tube Plots Tran129-136,
56.
February, 1953.
A New
Tekscope, vol.
1,
Packard
60. Forge,
61.
Charles
November, 1968.
O.:
A New
Clip-on
Oscilloscope/Voltmeter
Probe,
Hewlett-
August, 1960.
R. W., C. Straub, V. I. Saunders, F. C. Eisen, and T. Gentry Veal: Photography and Photometry of Cathode-ray-tube Displays, Phot. Sci. Eng., vol. 7,
pp. 289-304, September-October, 1963.
63. Bird, G. R., and Allan E. Ames: High-speed Oscillography with Transfer Films:
A System Analysis, Paper 342, Polaroid Research Laboratories, 730 Main, Cambridge, Mass.
62. Tyler,
64. Gilbert, B.
An X-Y
Deans, J. N.: A
Hewlett-Packard J., vol.
17,
September, 1965.
CHAPTER TWELVE
RECORDERS
Arthur Miller
Consulting Engineer
Otto S. Talle,
Jr.
C. D.
Meef
in a
is
form easy
427
428
moving pen
We
the
little
A familiar example
is
Some
from zero
move a
is
If
recorder.
72-7
Galvanometric Recorders
Recorders
429
Fig. 12-1.
The symbols and quantities to be used in the analysis of this
instrument follow:
c
c
load,
g-cm 2
= torque, dyn-cm
= deflection of coil about its axis, rad
S = torsional stiffness of coil suspension, dyn-cm/rad
Rf = mechanical resistance, dyn-cm-sec/rad
t
For uniform flux density in the air gap, the torque due to a given
current and the back emf generated in the coil by its motion are both
independent of coil position. The torque developed by the coil current is
coil
BNlbi
10
= Gi
G = BNlb
Section
(12-1-1)
abvolt-sec/rad
(12-1-2)
A-A
Output shaft
Bearing
r*-Coil
Soft
iron
[|lSoft iron
hi
Torsion
Magnet
rod
Anchor
_=i
Soft
iron
1A
Magnet
FIG 12-1
nometer.
Moving-coil
galva-
430
where
is
is
characteristic
is
stiffness torques, or
For the
electrical circuit,
= (/*.+
R.)i
+ G^ 10+L ^
at
at
(12-1-4)
G X
+R
R,
Rs
is
10- 8 dd
~l~
(12-1-5)
Re
dt
10(R.
(X2-1-6)
l + S
=I ^
+ D dt
dt
(12-1-8)
'
R,)
_, + [>,+ G X ) f + Se
dt'^\R, + R,Jdt
Then
*
+R
10 (R,
e)
GE
+R
10(fl.
where
d
'
E
=
is
(12-1-9)
SB,
c)
1
10S(R.
+R
is
e)
Recorders
If the
solution
damping
becomes
=l
is
~~k
~ Dtm
(t + tan_i H)
sin
(i2- i - ii)
is
*"-*
and the period
431
(12-1-12)
of the actual
damped
oscillatory
motion
is
2tt
"
(12-1-13)
- DV4/*)*
(S/I
A, r it = 2(5/)*
For
(12-1-14)
~\ 1 + VN ) e
Tf
When damping
is less
~ u " TN
than
critical,
by the
peak
negligible.
sine
e-Dtiu
(12-1-16)
Of
Damping
is
simply e~ DTIU
= JT~
T
T =
(l
(12-1-17)
K>)
(12-1-18)
- =
1
e-'*/"-* )*
(12-1-19)
432
The magnitude
K in Fig.
~Kt)i*
is
plotted as a function of
12-2.
r-7i
H sin
-K
\-
TN
(1
tan
(12-1-11),
when
(12-1-20)
There are differences of opinion as to the damping level that will prooptimum fidelity of reproduction of complex waveforms. In commercial recorders, D is usually set between 60 and 100 percent of the
vide
critical value.
cent of
is
is
70.7 per-
critical.
In Fig. 12-3
critical,
is
An
arbitrary definition often given for deflection time (or rise time, or
response time)
is
on the response
to a step.
30
+-
o
o
.c
$ 20
o
o
01
10
n
20
40
Percent of critical
FIG
2-2
80
60
100
damping
critical
damping
Recorders
433
1.2
_A
1.0
Damping
0.8
B
C
0.60
critical
0.6
0.4
0.2
0.4
0.2
FIG
2-3
0.6
0.8
1.2
1.0
1.4
1.6
1.8
we would have
TV = 0.307V
TV = 0.357V
TV = 0.547V
for
for
for
D =
D =
D =
0.60D crit
0.707i)crit
1.00Z) crlt
Another test signal that is useful in evaluating the response of a galvanometer is the ramp. Since the ramp is the integral of the step, the
response to the ramp can be obtained by simply integrating the response
to the step.
If this integration is performed for the last two damping
values illustrated in Fig. 12-3 and for an input ramp that should produce
a deflection of df at Tn then we get, for critical damping,
,
1-(
|_-J
TV
+
^T
2>V
,,
-2rt/TN
(12-1-21)
damping,
e- 2H ''i T cos
-^T
2*V
2*wt
TN
(12-1-22)
434
1.2
/ /h
1.0
Ramp input/
0.8
0.6
0.4
0.2
-7
Q8
0.6
0.4
0.2
K0.225*\
1.0
1.4
1.2
0.319-
FIG 12-4
ramp
ramp by a
fixed time
known
These delay times are obvious in Eqs. (12-1-21) and (12-1-22) and are
TD =
TD =
The
-^
0.3197V
for
D =
Dent
(12-1-23)
= 0.2257V
for
D =
0.707 Dcrit
(12-1-24)
results of the
An
to the calculation of
isosceles triangle
can be
From
we
critically
X,
1.
Recorders
435
critical
accompanying
Type
of step response
Time
Loss of amplitude
in the
critical
damping
of occurrence
of
peak
W + 0.277V
W + 0.237V
0.347V/JF
70.7%
shown
table.
0.182V/W
FIG
pulse
2-5
Synthesis of triansular
of ramps.
by summation
436
100
80
60
40
^
^^-Reproduced
20
Q-
Input pulse
0.5
1.0
2.0
1.5
pulse
3.0
2.5
3.5
TN
FIG
2-6
= 22V
of the
GE
10S(R,
where
+R
c)
/7/at
+ j2irfD/S
(12-1-25)
deflection of coil
e lf
G*
(12-1-26)
+R
Also, expressing D as a fraction of the critical damping value D
2tt/)
\0S(R,
4tt/K
c)
\s)
e>
we
get
(12-1-27)
Recorders
437
and
Bit
/7/at
The magnitude
which
characteristic,
(12-1-28)
+ j2Kf/fs
is
is
1.0, 0.707,
is
and
0.60.
a phase lag
2Kf/fN
_,
7 = tan
1
(12-1-29)
/7V
When
Conversely, as
120
Damping =C .60
100
Damp ng
^-
'
80
0.70
critic
\ ^- Dam ping*
criti cal
J\
60
40
20
0.10
0.20
0.50
1.0
FIG
2-7
2.0
3.0
438
2 5
FIG
2-8
in Fig. 1
bump
2-7.
initial
low-frequency
level.
If
we
to the
define the
3-dB
sponding
loss point,
rise
Bandwidth (BW)
Rise time
0.643//*
1.00//*
0.54//*
0.35//*
0.30//*
1.0
0.707
0.60
15//*
BW
X TB
0.347
0.350
0.345
This table shows that over the practically useful damping range, the
time Tr is given by the close approximation
TR
rise
0.35
BW
(12-1-30)
Recorders
The
is
439
system.
The ideal recording system would exhibit a phase lag of constant slope
and an amplitude response of constant magnitude versus frequency.
From Figs. 12-7 and 12-8 it is seen that the curves for a damping factor
of 0.707 show the best combination of flat amplitude response and
approach to linear phase lag.
A signal that can be reproduced without serious distortion must be
described by a frequency spectrum most of whose energy lies well within
the passband of the recorder. The time delay associated with this portion of the spectrum is given by the initial slope of the curves of Fig.
12-8.
This slope can be calculated from Eq. (12-1-29).
TD
i-
(sec)
Zir
tan
^
af
(12-1-31)
2Kf/fN
_,*
- p/fN
2Kf/fN
i
nit
r/fN*
2Kf
as
77
y approaches zero
(12-1-32)
Thus
TD
2K
2ir
fN
KTN
(12-1-33)
IT
when
1,
7V
TD _
7T
and when
Tn
= V2/2,
V2
7V
which are the identical values previously derived on the basis of the
time-domain analysis.
Equation (12-1-28) can be solved for the value of ///# which will yield
a response of \0/6i f = V2/2 to establish the 3-dB loss points which we
\
440
60
120
^vBond width
100
50
1 80
40 s
Rise time
60
30
40
20
20
10
70
60
50
90
80
100
FIG
1 2-9
damping.
Bandwidth,
rise
The
(-3 dB) =
From
[1
2K*
+ VHK* -
K*)
,M
2]'
result
is
(12-1-34)
Eqs. (12-1-34),
bandwidth,
72-2
rise time,
that drives
E =
(R e
+ JU>L )I +
C
(12-2-1)
-^motional
emf generated
The
term
Emotional is the
motion in the
last
Recorders
magnetic
and
field
The angular
is
= G X
^motional
441
10
-8
angular velocity of
coil
(12-2-2)
Angular velocity
jlw
+R
^-
jb/w
(12-2-3)
F
jQ^motK
FIG
2-1
Equivalent electrical
impedance of a salvanometer.
and
(77/10,
which makes
G X
2
+ j<*L)I + jib)
Z Ic + Z mot iona\Ic
E =
(R c
=
and
moment
is
G X
2
^motional
The impedance
z '
is
Rf
(12-2-4)
jc - i/L +
of inertia)
(12-2-5)
#LC
circuit
10- 9 Cco
G X
10- 9
= Iu
^- = -Leo
a;
and
G X
10- 9
4
R
= Rf
or
G x
G x
2
10--9
10--9
is
(l2 - 2 - 6)
i/ fi
10- 9
of a parallel
G X
jS/o)
(J c is current,
which
10- 9 7 e
if
442
and
G X
R =
2
1Q- 9
Rf
and the friction, which is a series element in the mechanical system, has
become a shunt element of the equivalent electrical circuit.
The voltage developed across the motional impedance represented by
the parallel resonant circuit
= jGwd X
#motioni
From Eq.
C)
we can
is
10- 8
(12-2-7)
and the
(12-2-7)
EG X
(R.
1Q- 1
+ R )S
C
P/fN
(12-2-8)
+ j 27r/V7
f
/n
where
G X
2
Rf
Rs
10- 9
(12-2-9)
Re
E _
~
10(R S
+ R )S
C
(1
If this is
it
+ J2K
(12-2-10)
jo> L c
1.^
J
"
2 xicr
FIG
2-1
galvanometer.
Equivalent circuit of
fre-
Recorders
quencies,
we
HL
443
get
E/e lf
_fL +j2K
lJ
w^
(12-2-11)
fs
*</*
_ . + Jj -lL_ L
Ic/Otf
Away from
Ic/d
2irfN
(12-2-12)
IfN
_ -
(12-2-13)
2
fN
Ic/Slf
fftermin^/0
-^terminal/A*/
_ _ + j 2R
(12-2-14)
/iV
fjf
where K' is the damping factor which would be obtained if the galvanometer were fed from a zero impedance source.
For galvanometers used in direct writing recorders typical values of
K' are around unity.
Equations (12-2-13) and (12-2-14) are important when amplifiers are
being designed that are intended to drive galvanometers. They are
plotted in Fig. 12-12, the voltage plot being based on a value of unity
for K'.
It is seen that if full-scale deflection is to be maintained at or
above the undamped natural frequency of the galvanometer, the ampli-
which drives it must be capable of delivering an output voltage considerably in excess of that which could maintain the same full-scale
fier
deflection at
low frequencies.
The
(R s
in Fig. 12-11).
is
can and does experience a substantial temperature rise owing to its own
power dissipation. This temperature rise can reach values of 40 C, and
as a result, the resistance of the copper winding will increase by 16 percent over its initial value.
444
5.0
MM
4.0
3.0
//
2.0
/Current
1.0
0.10
FIG
2-1
Ratios of currents
In order to make
0.50
0.20
and voltage
(or
this resistance
2.0
1.0
3.0
low-frequency values.
change unimportant,
obvious
it is
damping
factor.
R fb
FIG 12-13
Feedback
for galvanometer.
amplifier
is
gen-
Recorders
erated
445
(*
to
e fb
Jl_
R fb + kGX
_ kG X
Ra
Equation (12-2-17)
(12-2-15)
at
(12-2-16)
Rfb
dt*
10fty
d8
10 8de /dt
J^ _
10- 8
( Rf
is
kG2
10
><
R/>
identical in
AMl +
d
80
(12-2-17)
(12-1-6),
but the
coil
resistance
The curves
of Fig. 12-12 show that if the galvanometer-driving ampliwere equalized so that, beyond the natural frequency, the galvanometer current rose rapidly with frequency, it would be possible to stretch
the range of flat system response beyond the galvanometer natural
fier
frequency.
rises quite rapidly, however, and soon the amplicannot deliver enough of either current or voltage, or the required
current will burn out the galvanometer coil. Thus, if a particular galvanometer could tolerate a sinusoidal input current twice as great as that
which would produce full-scale deflection at low frequencies, and the
amplifier output were limited to that value, then it would be theoretically
possible to equalize the system to provide flat response at full scale out to
1.75 times the natural resonant frequency of the galvanometer alone.
fier
if
is
added to the
446
l-WV
V\A^-S
FIG 12-14
Equalized feedback
amplifier.
72-3
Pen-driving Mechanisms
The conversion
locus described
The
distance
= R
Using the
sin
series
(12-3-1)
= r(b -
= Re(i
6
To a
first
(12.3-2)
If
Pole piece
FIG
2-1 5
is
restricted to
about }4 rad,
this
term
is
Recorders
447
about 1 percent. If the system calibration is carried out at some intermediate point, say at 6 = 0.2 rad, then the error would be substantially
reduced over the entire 0.25 rad assumed as full swing of the pen.
This error can be completely eliminated by special printing of the chart
coordinates in accordance with the sine function, but a linear scale is
more attractive.
The time lines on the chart must be arcs of radius R, and in order to
avoid distortion of the timing of any details drawn out by the recording
pen, the galvanometer shaft must be located exactly at the center of
curvature of a time-line arc. Improper positioning of the galvanometer
or misalignment of the chart paper in the recorder can give the response
to a step input an appearance of having either a negative rise time or an
shown
in Fig. 12-16.
is
The
stylus.
stylus
is
rigidly
The
tact with the paper regardless of the stylus position on the chart.
paper could, of course, be electrically sensitive, in which case the stylus
would serve to carry current into the paper at the point of contact.
left on the paper has a y coordinate,
measured from chart center, of
tip
where
= R
tan
is
(12-3-3)
Fixed clamp
FIG 12-16
rectangular
cording arrangement.
re-
448
and
By
*-*(' + +)
=
^(l+f +
Again, to a
first
from linearity
(12-3-4)
approximation, y
is
proportional to
0.
The departure
is
full scale.
that there
is
construction
Recorders
449
rror.
.Nt.2.
(b)
the torque
is
COS a
10
2
(12-3-5)
produce a
coil
rotation
T
=
(12-3-6)
Gi
105
it
with
its
2
is
10S\
light
beam
(12-3-7)
approximate value
GH*
(a)
Gi
lOV
recording:
200Shits the
(12-3-8)
beam
will
As the mirror
is
reflected
from the
swing through the angle 6. If the disis R, then the recorded deflection
be
y
= R
= R
tan
tan 2a
(12-3-9)
450
enough with
y
GH
(l
LlO>S \
200S 2 /
-^
Comparing
GH* \
6,
it
10 3S 3 J
-**-)+ HGV)
- 2B\ 2-
= 2RGi
is itself
/10 3S 3
(12- 3 - 10 >
U + ^4
(12-3-11)
recorder characterized
the compensatory effect of the torque's falling off in accordance with the
cosine function.
The small galvanometer used in optical recorders differs from the large
units found in direct writers in another important respect, namely, the
method of achieving proper damping. From Eqs. (12-1-7) and (12-1-14),
the condition of critical damping
G X
If
,
VSI
given by
10~ 9
Rf
is
is
(12-3-12)
damping must
When, however, the stiff-
10-)/(/2.
+R
c ).
and
less
on the
circuit resistance.
When
the recording
is
to be
made by
ink, the
Recorders
Pen
451
tip
Sliding
pivot
Galvanometer
shaft
FIG 12-18
Straight-line
pen
mechanism.
pen
along a straight line which passes through the galvanometer shaft. This
straight line establishes the axis of the linkage.
Assuming small values of a so that only the first two terms of the
cosine expansion are significant, it can be proved that x is constant as
a and y vary
c
if
(12-3-13)
a
Actually, the motion of the pen does not stay perpendicular to the
direction of motion of the recording paper because the above assumption
value of c slightly greater than that
becomes worse as a increases.
shown
in
FIG 12-19
Eq.
(12-3-13)
is
Another straight-line
a better compromise.
The
nonlinearity
linkage.
between y and deflection angle in Fig. 12-18 is the same as that in Fig.
12-15 if we let the pen length R = a(l + c/b).
Another straight-line linkage is shown in Fig. 12-19. Here the rear
end of the pen slides through a guide which is free to rotate about a
452
fixed axis.
the galvanometer
may
to drive the
crank
a,
or
Here again
b 2 /a,
(c
(
(c
+ d)
a\
a + d)
a
tan
-i
tan sin -1
4.
(d 2
a sin
a2
(12-3-14)
v
'
2da cos 0)
surfaces,
position.
These
frictional forces, which exist even under conditions of zero velocrepresent a form of hysteresis that can be tested by applying a
deflecting current first in one direction and then in the other, with periods
of zero current between the successive reversals, as illustrated in Fig.
ity,
12-20 in exaggerated form. For this test the applied signal is passed
through a low-pass filter to ensure a gradual application and removal of
the signal, or the galvanometer damping is set for at least critical so that
there will be no overshoot.
The offset 5 should approach zero. In
practical direct writers
it
can be kept
less
full
chart width.
FIG
2-20
pen
motors.
Recorders
453
their amplitude
The major
recorders are the friction of the pen or stylus against the paper, bearing
out of which the torsion elemade, and hysteresis effects at the clamping points where the
torsion element is attached to the moving coil and at the attachment to
the fixed supports.
Torsion bar design is a mechanical engineering
problem beyond the scope of this treatise, and the same can be said of
ment
is
if
any, that
as a lubricant.
process
lost
is
If
the stylus
is
and performance
The
is
itself.
friction.
direct proportionality
Transducer output
where
KT
is
e/&
KT 6
(12-3-15)
454
FIG
2-21
Position feedback in
pen
drives.
A =
where
coil
KT Ae
(12-3-16)
current
KT BA
R
(12-3-17)
c
_Gi._ GKT dA
~ 10 ~ 10Rc
(12-3-18)
back
is
stylus.
SB
The apparent
T
e
stiffness, then, is
_ GKT A
~ ior
c
Recorders
455
mm
also
bound
if it
may
stiffness,
much
of the range.
high stiffness for small deflections, but the fact that large deflection errors
are not accompanied by proportional increases in restoring torques does
not matter.
If the apparent stiffness brought by position feedback is so great, then
part
it is obvious that a conventional torsion element is not a necessary
This means that a given static deflection
of the galvanometer structure.
of the stylus can be maintained at any position of the chart with no
when
When
the stylus
moving.
is
is
will
go through
a minimum
required current remains below that which would need to be supplied to
the same galvanometer without a torsion element.
For a given maximum permissible current, therefore, a recorder with
a mechanically resonant galvanometer will be capable of providing a
The presence of the torsion element, however,
greater bandwidth.
at this resonance.
is
called for to place the stylus near the chart edge that there would not be
much reserve left to make any corrections of errors sensed by the feed-
back transducer.
A compromise can be made which would require substantially less than
maximum permissible current to reach chart edge, but which would still
provide a significant increase in bandwidth. The curve shown in Fig.
12-22 for S = 2 X 10 6 for example, represents a practical limit to this
compromise, and it results in an increase from 64 to 79 Hz in the maxi,
mum
456
20
10
30
50
70
100
Frequency, Hz
FIG
2-22
72-4
Servorecorders
Servorecorder uses fall into either one of two general types. The first
type is the plotting of a variable versus time. The recorder performing
this task
Types of Servos. The servo systems for these recorders can take many
These forms vary from open- and closed-loop galvanoservos
forms.
Recorder
FIG 12-23
Hewlett-Packard
Hewlett-Packard
for
latter, the
Input
scale
facte*
Input
457
vy
Xf>\
1
Amplifier and
gear train
It
Servo
motor
i
1
Feedback
scale
factor
FIG 12-24
Position
transducer
all
the closed-loop
The
of recorders
458
PEN
\
FIG 12-25
of-^WWWWWWWV-|
is
Potentiometric-input
recorder.
is
a closed-loop, null-balance
It derives
Performance Characteristics.
systems are covered in detail elsewhere in the literature and will not be
directly dealt with here.
However, there are some unique design aspects
that must be considered when designing a servo null-balanced graphic
recorder that are learned
by
experience,
here.
and (3)
Accuracy
ity,
resettability.
partially determined
input attenuator
initial
if
graph paper.
Linearity
is
line, offers
This definition, as
the most meaningful and useful speci-
fication.
Recorder*
459
compensated
for.
Linearity can
The
starting voltage
is
influenced greatly
by the
fric-
on the motor.
dynamic
A > p-
(12-4-1)
2w/
where V,
460
acceleration limit
is
is
a V,
A =
2irf(2irfV,
in.
peak-to-peak
(12-4-2)
oo)
method
Since these
of using the
3-dB point
in
null-balanced system depends on its amplifier's having high gain to minimize the error signal. Any input noise that appears on the error signal is
Reliability.
components is
some components with an attitude
even though they are unreliable. A
often neglected.
Many
designers use
filter
example
is
FIG 12-26
Velocity
celeration limits.
100
Frequency of input signal, Hz
and
ac-
Recorders
transducer.
and
The use
stable circuit.
461
of a slidewire
Only
its reliability
tact resistance.
72-5
Magnetic Recording
system.
Typically,
voltage
is
some form
of finely
induced.
It is possible to
Magnetic tape
Magnetic heads
FIG
2-27
of the other
two main
This
way how
axes,
mode
the tape
is
magnetized.
Magnetic wire was used at one time, but tape has superseded the wire.
If a magnetic field is applied to any one of the iron oxide particles
The relationship
in a tape and then removed, a residual flux remains.
between the residual flux and the recording field is determined by the
previous state of magnetization and by the magnetization curves of the
particular magnetic recording medium.
A single magnetic particle on the tape might have the BH relationships shown in Fig. 12-29, where H is the magnetizing force and B is the
Complete degaussing of the material gives
flux density in the particle.
the condition at point 0.
head
Now,
if
Oxide, or
magnetic
coating
Plastic base
FIG 12-28
diagram
process.
Trailing
edge of gap
Tope motion,
of
Simplified
block
magnetic-recording
Recorders
FIG 12-29
Typical
463
magnetiza-
tion curve.
until
1,
2,
is
brought
from
only as far as 1, and H is then brought back to zero, B follows a
minor hysteresis loop back to the point 6. A greater value of coil current would leave a higher residual flux, and a lower current a lower
However, the
residual, and a very simple recording process results.
linearity between residual flux and recording current is extremely poor.
Below, in Sec. 12-6, a method for securing linearity in direct recording
be described. Also, the use of
in recording, to yield independence from amplitude variations and to extend the lower frequency range
to dc, will be discussed.
Finally, various pulse recording schemes for
FM
will
change of
flux.
FM
for
memory.
FM
464
The
FM
bias.
The dynamic range can exceed 50 dB, and recovery from saturation
3.
is
virtually instantaneous.
form.
trical
5.
6.
times.
analysis.
Several or
7.
many
tape.
8. The time base of the signal can be changed by merely playing the
tape back at a speed different from that for recording.
9. Magnetic recording can give exceedingly high density of data points
to simplify storage
The unique
and handling.
it.
Recorders are made in many sizes and are available in portable, semiA few of the applications follow:
portable, and rack-mounted form.
1. Industrial research and production monitoring and control, including stress and vibration recording, fuel-consumption logging, and noise
analysis
2.
and
missiles,
and
satellites
3.
4.
12-6
The
and
analysis on aircraft,
Recorders
a variety of magnetic recording systems.
The
465
and
systems with magnetic tapes, strips, disks, drums, or loops use the same
recording and reproducing transducers and magnetic medium. For these
components, the techniques of recording and reproducing will be described
in this article.
is
is
tances y/g from the recording-head surface, where g is the gap length.
It is an objective of high-resolution recording to produce a very rapid
decrement of the recording field as the tape moves away from the gap
region in the x direction.
Since this condition occurs near the head
surface as shown in Fig. 12-30, only a thin magnetic layer can be effective
The change in direction of the recordfor short-wave length recording.
is illustrated by plotting the longitudinal and perpendicular
components of the field on the right-hand side of the ordinate in Fig.
12-30.
It can be seen that, at the gap center plane, the field is entirely
longitudinal and rotates toward the perpendicular direction as a tape
element moves away from the center plane.
In order to understand how the tape magnetization depends on the
ing field
made
is
FIG 12-30
Field
distribution
and g
gap length.
466
FIG 12-31
60 Hz
7-Fe 2
field
Hysteresis loops of
for oriented
3;
particles
H^ = 1,000 Oe:
of
(a)
to (a).
field.
obtained (Fig. 12-316), but here lower maximum magnetizaThus, taking account of. the tape magnetization
characteristics and the recording-head field contour, it is expected that
the recording will be primarily longitudinal for most of the tape.
However, for the surface layer in contact with the recording head, large
perpendicular fields are encountered which produce some perpendicular
magnetization. That this is the case has been verified by examining the
tape magnetization direction at various depths into the magnetic layer
with the use of a large scale model. It is an unfortunate feature of
conventional recording that perpendicular recording, with relatively poor
acteristic
tion
is
is
achievable.
would be obtained
where the
maximum
resolution
The
head
simplest
in this
manner
is
known
signal.
as non-return-to-zero
Recorders
clock track, corresponding to the time
when a
or
is
recorded.
and
is
467
Self-
reversed at
between these
clock signals.
With
number
(a)
KHs
+ X2
V*"
_
IX (K'Z)
N.
Hx
f~-~h-
/3
/ Ix (K--4)
(b)
1.0
/l\
/
0.5
'
'
FIG 12-32
1
1
Recorded
/7\\
<^i
(a)
tran-
\
V.
468
different amplitudes
of the reproducing
The reproducing-head
field function in Fig. 12-30.
approximately expressed as a convolution integral of the longitudinal field function of the head
x and the longitudinal tape magneti-
but
is
given by the
flux is
zation I x
For a tape
<f>c
K fa
is
of thickness c
given by
a+C
dy
p" HJ
dx
(12-6-1)
From
NRZ recording,
it is
evident
modules or foreign
is
To minimize
recording head and a "dropout" of information occurs.
the effect of dropouts, large recording fields are used and resolution is
Present high-density data recording
sacrificed for increased reliability.
is in the range of 1,500 to 2,000 flux reversals per
coatings with high coercive force, extension
using
metallic
thin
By
up to 10,000 reversals per inch is envisaged for the future. Corresponding improvements in the resolution of the reproducing heads is required,
on oxide-powder tapes
inch.
and
this
may
pulse slimming
filters.
Analog Recording.
to analog recording
Recorders
469
is
saturation level.
Anhysteretic Magnetization Process.
The
shown
is
in the anhysteretic
tape.
For ac fields greater than 500 Oe a limiting condition is
reached and the characteristic curve is obtained. In ac-bias recording,
the anhysteretic magnetization process is modified, since the dc field (the
recording signal) acting on a tape element reduces in amplitude at the
same rate as the ac-bias field, as the element traverses the recording-head
gap zone. This has little effect on the anhysteretic magnetization curves
of Fig. 12-33 until the maximum ac field exceeds the coercive force.
Then, instead of repeating the characteristic curve for large ac fields,
powder
This
is reduced but keeps the same general shape.
understood since the final magnetization of the particles is determined by the amplitude and direction of the dc field when the total field
In anhysteretic maghas reduced to the particle switching field H'c
netization this condition is always the same, providing the maximum
However, in the modified
ac field exceeds the particle switching field.
magnetization process, the dc field occurring when H^ +
dc = H'c will
be reduced as the maximum ac field increases. Thus, for recording conthe slope of the curve
is
easily
has a
maximum
value for a
maximum
field.
amplitude.
It is
O bU
c
/x* 500
t>
>
40
20
y/^250
/x*\Q
>
FIG 12-33
y^x=o
400
200
Applied field,
Hai
Oe
470
curve
is
similar for
Whether or not
ic
randomly directed throughout the material and will have an amplitude distribution which determines the shape of the anhysteretic magnetization curve. Locally, the
magnitude and direction of the internal field will depend on the packing
density of the particles and their magnetization directions.
Consequently, if the magnetization direction of one particle is changed by an
Essentially these local internal fields will be
external
field,
its
and the
effec-
is
Practical Conditions.
The recording
maximum
sensitivity.
Eventually, however, a limit is reached since a
decrease of the magnetization of the surface layers in contact with the
shown
in Fig. 12-34.
As has been
Recorders
explained, ac-bias recording takes place
Hc
#dc
when the
falls to
471
fields.
Due
to
0.25H C
for typical oxide powders, recording will take place over the finite region
which the applied fields fall into this range. Recording regions corresponding to three different amplitudes of applied field are shown as
shaded zones in Fig. 12-34a and b. In Fig. 12-34a the zones correspond
to the resultant applied field, and it is seen that the narrowest longitudinal
recording region, giving highest short-wave length resolution, occurs
for zone 1 where the field is not sufficient to magnetize the whole coating.
in
Owing
12-346 where only the longitudinal component of the constant field conis plotted.
In practice, the short-wavelength recording resolution
reduces so sharply with separation from the head gap that for a recorded
wavelength of 2 jum, the loss caused by a 0.1-/xm surface roughness on the
tours
is about 50 percent.
For the example shown
tape
in Fig. 12-34a
\ \ \ \ \ \ \
and
FIG 12-34
head
b,
Constant
recording-
wide gap,
wide sap,
resultant
field;
(b)
WW^P
\ \ \ \
472
FIG 12-35
Schematic of repro-
length
in Fig.
is
12-34c, correspondingly
becomes the limiting factor. Other losses can occur in the recording
head itself, and thin metallic laminations or dense ferrites are used for the
head core material to minimize the eddy current losses at the bias frequency. This frequency must be high enough to avoid any intermodulation with the signal and also to obtain anhysteretic magnetization
conditions.
A number
is
The
The low-reluctance
Recorders
473
shown in Fig. 12-35, most of the magnetic flux entering the head
core from the tape takes the desirable low-reluctance path C consisting
of the core reluctance R e and the back gap reluctance #,.
However,
ditions
the head gap provides a shunting reluctance R g >, which may be low when
the gap length is made very short to reproduce short recorded wave-
The
lengths.
is
given by
*<
* - Re
o'
Rg
<&**>
Ri
For high efficiency in the reproducing head, the gap reluctance is usually
kept high by use of a very small gap depth, the disadvantage being the
reduction in head life due to wear of the pole pieces. The core reluctance should be as small as possible, and high-permeability nickel-iron
laminations are used for heads operating at relatively low frequencies.
For video and pulse recording, however, the frequency losses in metal
laminations are excessive, and ferrite heads are superior. Other practical reproduction losses associated with the head geometry arise when
the pole-piece length is of the same order as the recorded wavelength
and when the head gap either is not straight or is misaligned with the
recording.
assume that
is
and the field distribution in the gap region may be looked upon as a
measure of the spatial sensitivity of the head to the magnetization in the
gap region. Integrating the contribution of all elements through the
tape thickness and along the tape length yields the total head core flux $e
as shown in Eq. (12-6-1).
For a finite gap length g' a tape thickness c,
and a sinusoidal recorded magnetization Ix of wavelength,
,
*c =
47rc/ x(max)
ABC
2wvt
cos
(12-6-3)
where
v is
Were
it
all
three
474
[1
coating thickness.
elements of tape occurs with regard to their contributions to the head core
This leads to a series of minima in the core flux on further reducing
the wavelength.
Thus, in order to obtain a continuous wavelength
flux.
response,
minimum
it is
first
In addition,
Recorders
infinitesimal coating thickness
ex
is
475
given by
=
dt
x2
Klxbntx.) In
+(a +
+a
a;
(12-6-4)
c)
number
of
unwanted
is
signals arising
from imperfections
port and in the recording and reproducing transducers; if the tape speed
is caused to vary, either due to periodic changes in the transport system
or due to oscillations set up by the friction between the moving
tape and the stationary heads, undesirable changes in signal occur, which
amplifier.
in the tape
itself.
When
effect
known
Finally,
unwanted
as print-through,
which
476
occurs in a
wound recorded
reel of tape.
During storage
of a recorded
tape, the magnetic fields of the recording penetrate the adjacent layers
and can sometimes produce a low-level print or recordIt is found that this undesirable
effect increases with temperature and with the presence of small external
fields.
The single-domain particles of the recording medium have a
distribution of shapes and sizes, and it is known that as the volume of
such particles decreases, the relaxation times for magnetization change
of tape in the reel
ing which
is
detectable on reproduction.
are reduced.
external field
is
form closely
to
also a resolution
by self-demagnetization effects. Because of these requirements there has been a return to considering the advantages of metallic
magnetic media. However, the metallic tapes now under development
losses caused
consist of a thin metallic layer deposited onto the plastic base material;
in this
way
are both
optimum.
remanent magnetization
is
Recorders
477
external fields.
In selecting a suitable material for shape-anisotropy-controlled singledomain particles, attention is paid to the size of range over which singledomain behavior exists. Outside this range, multidomain and superparamagnetic properties are obtained which reduce the magnetization
When the chosen particles are mixed with the plastic binder
stability.
plastic base material, the aim is to produce complete
orientation of the particles so that a recording field can be applied along
their long axes to obtain the same magnetization characteristic in all
It is also desirable to
internal fields
this material.
first
478
FIG 12-36
Horizontally
oriented iron oxide particles,- magnified
sight to offer
20,000 X.
characteristics.
The single-domain
size
Another approach to the manufacture of very thin metallic magneticlayer tapes has been attempted by using electrodeposition or electrodeless
deposition of the material onto the plastic base material. In this method
the inherent advantage of coating dispersions
is less
room
is
479
Recorders
intensity of
trol
magnetization,
CITED REFERENCES
1.
Components for
Davis, Sidney A., and Byron K. Ledgerwood: "Electromechanical
Servomechanisms," McGraw-Hill Book Company, New York, 1961.
2.
3.
4.
Company, Amsterdam,
5.
1964.
New
York,
1964.
6.
OHG,
Berlin, 1960.
CHAPTER THIRTEEN
MEASUREMENTS ON
AUDIO AND VIDEO
AMPLIFIERS
From Notes by
Paul Baird
Fred Hanson
Charles Kingsford-Smith
Terry
E.
Tuttle and
Larry A. Whatley
Hewlett-Packard Company, Loveland, Colorado
study amplifier characteristics further. Enough theory and design considerations will be given in this chapter to clarify the various measurements treated. For a fuller treatment of amplifier calculations the reader
is directed to a good text [1, 2].
Familiarity with Chap. 2 is assumed.
As stated in Chap. 7, amplifiers serve (1) to increase the power avail480
481
source impedance for a load, and (6) to attenuate or reject the commonmode component of voltage on a pair of conductors (the common-mode
is the average of the voltages on the two conductors at each
with
respect to ground or some designated reference potential).
instant
voltage
Transfer Gain
13-1
and
Transfer Function
of
s,
the ratio
is
known
as the transfer
function.
in an amplifier can be conThis leads to four possible gain
ratios: voltage gain, current gain, transconductance, and transimpedRefer to Fig. 13-1 for the basic circuits of an amplifier with
ance.
Thevenin and Norton equivalents for the signal source. Regardless
of how gain is defined for these circuits, it is usually desired that the
sidered
either
voltage
or current.
YL AIZL
YS =VZS
w
Zs
>)>
/Zl
FIG
3-1
/22
Y\\
/IZ
rL
TVs Vs
>2i
i^E
yzz
ratio.
The
482
TABLE
1 3-1
Preferred
Preferred
Convenient
parameters for
Zi n
analysis
\z,\
|Z.|
|Z.|
\z.\
\zL
\ZL
MZl\
|Zi|
Transfer gain
parameter variations.
Consider, for example, the operational (or shunt) type of feedback,
The
gain,
EL /E ^ Zf /Z,.
frequencies
which
is
commonly
is
Here we have
done.
transimpedance rather than a voltage gain. The voltageinverting amplifier, Fig. 13-la, combined with the shunt impedance Z f
as providing a
can be viewed as forming a new amplifier which has very low input and
output impedances. The y parameters of this new amplifier can be determined by adding term by term the parameters of the original amplifier
to the y parameters of the shunt impedance.
For this reason and for
simplicity in discussing the ratio
to load
and source
E L /I
The
variations, while
El/E
is
variations.
One
symbol T, the
Sk T )
is
Sk
The
defined
sensitivity of
by
dT/T
(13-1-1)
dk/k
sensitivity of
to k
is
divided by the
all
fractional
of
EL
-2/21
/,
(?/ii
Y,)(y<n
+ YL )
qel/i.
2/i22/2i/[(2/n
of concern here
+
of K
2/122/21/K2/11
F)(2/22
in terms
and
0,
+ YL
Y,)(V22
(it is
+ YL
(13-1-3)
)]
is
__
- pK
by
(13-1-2)
2/122/21
a El/ I,
483
identifying
K=
(2/11
Y.){y u
+ YL
ohms
= yn
and
j8
^ = K8P =
EA
(2/n
^^
F.)(y
+ YL
|,
(13-1-4)
)
stability
FIG 13-2
Circuit
for
measurement
it is
to be
484
13-2
Steady-state Gain
in
Amplifiers
signal
have
(jw), the
specific values of
phase of the gain ratio. The signal source is adjusted to the desired
frequency and amplitude, and then the phase meter and the two ac voltmeters are read. The voltage gain is calculated as
K
where
= Tjll+i^
<j>
is
(13-2-1)
Chapter 6 discusses digital phase meters, and analog phase meters measure the ratio of the interval between zero crossings of two signals to the
period of one cycle of the test frequency. See p. 25 for general treatment
phase detectors.
of
When
before
it is
used as a reference.
Amplifier
under test
J
~<j(ya))
Signal
source
Phase
meter
$
FIG
3-3
Direct
method
for
485
Amplifier under
test
Mixer
Channel
BP filter
(Aw)
voltmeter
"
C [(/Yw+
Ei(ju)
Signal
_|
source
R?
Signal
Aw/l
source
V7
BP
filter
\E,
,,
Phase
meter
^1-^2
-D>
voltmeter
Splice
(Aw)
FIG
3-4
Channel 2
Ei
is
proportional to
\Ei\.
identical as possible.
it is
If
Operating frequencies in amplifiers are often much higher, and cirmore difficult to determine analytically to the required
degree of precision.
486
2.
frequencies.
3.
shift at
When
trol
operating with large signals, both feedback amplifiers and consystems sometimes exhibit a variation in gain over the signal dynamic
Bode Plots. Network analyzers are available that provide a sweptfrequency source and outputs proportional to the phase and log amplitude
of the
easily
and pole at
and 20 MHz,
respectively.
The gain
487
+30
+20
CD
300
10
Frequency,
FIG
3-5
Bode
MHz
satisfactory.
However, the
of a network
by means
Naturally, such
not evident from the amplitude
is
plot alone.
closed-loop response.
and log
of gain as ordinate.
loci:
loop behavior.
A Nichols plot of the amplifier with the Bode plot of Fig. 13-5 is reproduced in Fig. 13-6. The solid trace represents the compensated loop,
while the dotted trace includes the effect of a 10-nsec transport lag.
Notice that the actual closed-loop performance can be read from the
488
-210
-180
-180
-150
-120
-90
-150
-120
-90
-60
-30
FIG 13-6
The Nichols
chart: 1.6
dB
chart.
peaking at 6
MHz.
dB
at 9
MHz.
Immittances.
one-port device). A common technique is to convert the variablefrequency generator into a constant-current source (for impedance measurement) or a constant-voltage source (for admittance measurements)
489
Network under
.test
BP filter
(Aw)
Mixi
Voltmeter
^rii
E\
(j<*)
Ec
Phase
meter
fl//0D+Aa;)]n
\<r->
Voltmeter
Ez
BP
filter
(Aw)
FIG 13-7
to drive the
network under
test.
is
73-3
1.
phase shifter
is
and
Guard
shield
FIG
3-8
Arrangement to mea-
Phase
shifter
it
can be
490
switched to either side of the null detector, depending upon the sense of
To use this setup, the resistive divider and
the amplifier phase error.
phase shifter are adjusted until a null is obtained to the desired sensiThe resistive divider is then removed, and its ratio is measured
tivity.
with dc. This ratio equals the gain (except for a phase-shifter correcThe frequency range is limited primarily by the frequency
tion).
response of the resistive divider, since it is not pure resistance. The
This
is
capacitor)
is
slightly less
if
than at
its
is
input.
this
To
setup.
TABLE
13-2
by
quired
and
Phase-shifting
13-8 f
Fig.
Angle, deg.
10
9
8
7
6
5
4
3
2
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
t
Correction,
1.5192247
1.2311659
0.973193
0.7453848
0.5478
0.3805
0.2436
0.1370
0.0609
0.0152
0.01234
0.00975
0.00746
0.00548
0.00381
0.00244
0.00137
0.00061
0.000152
Phase-shifter output
is less
than
its
would
491
then be added to drive the guard box separately. Since no input divider
has been shown, the guard box is correctly driven.
Even better accuracy than cited above can be obtained over a narrow
frequency range by substituting a suitable inductive divider for the resisInductive dividers are available that give a direct readout of
tive one.
and accuracies
ratio,
Common-mode
73-4
Rejection
network that one wishes to measure is frequently comoffending signal, and the problem is to meaSometimes the offending sigsure the first in the presence of the other.
When
nal is simply superimposed upon, or added to, the desired signal.
this series condition exists, separation can only be made by filtering, and
filtering requires a priori knowledge of the frequency content and time
Of course, the interfering signal can often
relationship of the two signals.
The
signal in a
bined in some
way with an
(b) Balanced
(a) Unbalanced
FIG
3-9
Signal
differential amplifier.
E =
Kd
where
KjEi
is
c,
the
KE
e
(13-4-1)
Kd =
fj
and
-\-
Ee:=0
common-mode
E t
(13-4-2)
gain, is
(13-4-3)
492
The
signal.
is
defined here as
CMR, = El
(13-4-4)
Kd
mode
generally gives
CMR4
negative number.
CMR,
it is
or
<<C 1.
Some
CMR ratio,
common-
components
of output.
Expressed in decibels,
it
This definition
Kd /K
as
CMR
be no confusion. A
not a large number.
For small signals, one -can use either a or & of Fig. 13-9 to represent the
operating conditions. However, when source impedances and amplifier
input impedances are considered, the equivalent circuit becomes more
complex, as shown in Fig. 13-10. Here, Ec can in general produce differential components of signals at the input and output of the amplifier,
ratio that the whole circuit would
even if CMRa is zero. The
have if CMR^ were zero is called balance. The symbol is CMRs. In
the figure, Z,i, Z S 2, and Z s3 are impedances associated with the signal
source, and Z t Z h and Z% are characteristic of the input to the amplifier.
Referring to Fig. 13-10, observe that E actually consists of four components, which arise from the following causes:
1. The differential gain
dd of the input network and differential gain
CMR
of the amplifier,
Eoi
The
2.
FIG
3-10
NddK Ei
d
and
CMR.
de
of the input
network
493
of the amplifier,
N K Ei
3. The common-mode-to-differential gain N
the differential gain Kd of the amplifier,
=
Eoi
E =
z
dc
ed
and
NcdKaEc
E = N KE
i
CC
#34
Since
cd
E on /Ei
is
common-mode
CMR
^
(K + ^ K
Ndd (Kd +
Eon =
K^j Ei
(13-4-5)
(13-4-6)
gain, the
E o3 i/E
the
is
is
= CMR.
dc
where
CMRX
Ndd
Equation
and CMR B
(13-4-7)
.
shows that
CMR is
CMRA
is,
CMRx -> 0,
CMR
balance
CMRB
(13-4-8)
Ndd
Equation (13-4-7)
CMR
may
CMRS
ri
Ll
be written
+ (*-/*>CM^1
+ (N /Ndd )CMRA ]
Ki6
>
dc
from
Fig. 13-10,
we have
Fig. 13-11.
494
FIG 13-11
13-10.
By
definition,
N
N
+ E )/2]/E
[(Ex
cc
-E
(Ex
ed
+ Zi2 Zx
ZxZ 2
2 )/E e
Ex
+E
2(E 1
-E
2ZxZ 2 (Zs x
-j-
2(Ze2 Zj
-f-
2)
Ec
Z i2
= input
-\-
Zij/Zj
(13-4-10)
ZsiZz)
and
E +E
2(Ex - E
)/Ei
(Ex
Ei = input
2k)ZxZ + Z (Z - Zx) + (1 - k)Z slZ
(1
-kZ Zx + 2ZxZ [(\ - k)Zsl - kZ ]/Zi
N*
[(Ex
+ E )/2]/Ei
t
-E
Ndd
t)
a3
s2
s2
-2[Z 8l Z 2 (l -k)
kZa2 Zx
+Z
iS (Zx
+Z +ZZ
2)
2\
(13-4-11)
The balance
N
Ndd
is
Z*xZi
cd
(1
k)Zsl Z 2
kZti Zi
In a balanced transmission
Z,X
and Zi =
k
3^.
Z, 2
Ro,
Z,3
where
CMRB
+Z
s3
(Zx
+Z +ZZ
2)
(13-4-12)
2
Ro
is
Under these
line
Zi2 Zx
is
N
=
Nm
2(Z 2
S(Zx
Zi)
Zx)
(13-4-13)
4(ZxZ 2 )/R a
we
If
Balance
In
(^2
<$C
2,
Zl)Ro
oy _
(13-4-14)
nAm
CMR
is
CMRA
For
[3Z(Zi
(13-4-15)
K
'
iM
balance /
CMRa
(13-4-16)
CMRa
(13-4-17)
is
CMR* CMR +
and to emphasize
CMR
we
write
CMRA
under the requirements that CMRA <C
The relationship between CMR and
Balance
Z2)Ro/8Z 1 Z 2 ]CMRA
1,
CMR CMRB (l -
different sign,
495
(13-4-18)
and Ri,
R R
2,
<3C
Z Z
1}
2.
CMR
73-5
Common-mode
Rejection
Coupling
Figure 13-12 shows how a shielded transformer is connected between
a signal source and an amplifier to reduce common-mode output. The
impedances to ground, Z\ and Z 2) comprising R\ and C\ and R 2 and C 2
,
and winding capacitance, and the impedance of these elements can approach 10 9 fi at low frequencies. The Zi
of the earlier derivations is Zl reflected to the primary in accordance
with the turns ratio of the transformer. Therefore, in the lower frequency
and balanee is not difficult.
( 80 to 90 dB)
However, at higher frequencies, the stray capacitance of the system
CMR
496
FIG
3-1
CMR.
of Ci
C2
and
to Ci since
it
respectively.
windings,
wound
C3
just adding
couples the
in sections
alent values of Ci
l/j'wCs).
and interconnected in a manner that makes the equivand C 2 almost equal, and this gives good balance over
borhood
13-6
of several
hundred rms
volts.
Obtaining good
CMR
Low
Ratios
come
One
of the better
methods
for ensuring
good
CMR
is
to use a
compen-
is
497
CMR
final
FIG
Simple
3-1 3
common-mode
Kcm
E,
1
Ei
the
ratio.
K C m,
the out-
(13-6-1)
K
E
+ R/2Ri
can make
no current
is
Ei
= E 2 = EC m and
(13-6-2)
Kcm we
EoCM =
if
+E
CMR
R/2Ri
but
low
If
is
applied, EC m
0.
By
adjusting
E cm = E C m,
in either
or
2.
Q2
CMR
fier.
Let
PS
FIG 13-14
have a
Another
differential gain
2d ,
common-mode
A",
effective differential-amplifier
scheme.
gain
K 2C m,
498
and
across the
two inputs.
With
dEo
dE[
+ K)(-K u +
(K/l
1
(K u
Kzcm/2)
K, C m/2)/(1
K)
^f
^p
and
dEo
(K/l
dE'2
K)(K 2d
Kzcm/2)
- K 2C m/2)/(1 +
(K 2d
K)
It follows that
Eq = F\Ei
+FE
2
+ F ^-^ +
2
(Fx
2)
^^
(E[
Eft
commonmode
commonmode
differ-
differ-
ential
gain
signal
gain
ential
signal
(13-6-5)
+
^ Fi
I*/* 1
=
1
(K 2d
- K 2C m/2)/(1 +
Fi-F
2
so that
-[K/(l
=
1
from E[ and
CMR
= -
(K 2d
K)
\K\,
^^
K
(13 - 6 -6)
2d
\K Cm\.
Under
these
same
is
K))K 2d
- K 2CM /2)/(l +
^ _R
if)
(13-6-8)
K 2d
At
KR
resistors are
CMR
shown
in Fig. 13-15.
If
Kld
is
(13-6-9)
Kg
The common-mode gain of the first stage is established by the commonmode amplifier to be very nearly unity since all ground returns in A x are
FIG
3-1 5
The
first
499
stage in Fig.
13-14.
referred to (E x
mode
E*)/2.
Therefore, Eq. (13-6-6) gives the
gain of the whole system, and the system
ratio is
CMR
common-
CMR
K*(1+K)(1
(13-6-10)
2R /R B )
1
usually
procedure
1.
is
used:
Apply a
and adjust
its
amplitude
Short the two input terminals together, apply the signal source
input, and adjust the input level to give the same
output indication as in 1, above.
2.
as a
common- mode
500
10
12
11
13
14
15
FIG
3-1
its
CMR ratio.
Balance
mon
is
the
CMR
CMR
is
CMR
73-7
Distortion
ratio of specified
501
is
Even
dis-
When
by the
maximum signal must actually be present during the measuremaximum signal level may not be close to the
level; that is, at such times the maximum level is limited by a
system, the
ment.
saturation
small nonlinearity that causes spurious signals near the noise level.
There is no single best way to specify and measure dynamic range.
The
usually
system
is
may
be considered.
its
and
will
502
some
similar bandwidth,
may be
The
is
pro-
surement of noise
is
specified to be
made with
a flat-response instrument.
may not
When
Distortion.
be a faithful reproduction of its input. If the frequency response of the
system is not flat, different frequency components of the input signal are
amplified or attenuated by different amounts and the system will cause
a signal
is
line, their
phases
If the value of dd/du is dependent on frequency, different frequency components are delayed different times and
This can be seen by recalling
the signal undergoes phase distortion.
that in the Fourier series representation of a waveform, the phase of
each frequency component must be specified to define the waveform
While the correct phase relationship may be relatively unimuniquely.
portant in some signals, such as an acoustical waveform, other signals
such as a train of data pulses may be completely destroyed by phase
distortion, although the amplitudes of the component frequencies may be
so that dd/du
is
constant.
Measurement
range of interest.
Traditionally dd/du has not been measured directly, the accurate measurement of phase at different frequencies being difficult; instead, a
method developed by Nyquist and Brand has been used [6]. This techsignal which has been fed through the system
nique evaluates an
It can be shown that measurement of the phase of the
to be tested.
modulation envelope yields values of dd/dco equivalent to phase measurements made at the carrier frequencies and is much easier to accomplish
AM
is
the type
most commonly dealt with and is caused by a nonlinear transfer characAmplitude distortion can be evaluated in a few cases, especially
teristic.
503
is dc coupled, by determination of the transfer characterand noting its departure from a straight line. This type of measurement is not generally made, however, since it lacks sensitivity, and results
One frequent
are not directly meaningful in most system applications.
if
the system
istic
exception
is
accurately.
Here a
direct
measurement
of the nonlinearity
is
most
meaningful.
(AS
+ AS +
The
)
fundamental amplitude may be expressed
and is the value of total harmonic distor-
as a percentage or in decibels
Rather than use the wave analyzer to measure each harmonic indisome prefer to use the distortion analyzer, which rejects the
fundamental frequency and measures the average value of all harmonics
simultaneously, the result being a single measurement of the total harmonic distortion. This method is less accurate but much quicker than
the previous one.
In making the measurement with the distortion analyzer it must be realized that components other than distortion products,
such as hum and noise, may be present in the measurement results.
A second method of distortion measurement involves the passage of
two large sinusoids through a system simultaneously and the measurevidually,
ment
wave
analyzer.
The sum
504
Frequency
-I
3 3 3 3
3 3
3 3
tortion
trum of the system output, shown in Fig. 13-17, is seen to consist of the
two driving frequencies a>i and co 2 and distortion products grouped around
The order of a distortion product is the sum of the coefficients of the
W2.
two frequency terms; for example, u 2 + 2an is odd distortion of the third
SMPTE
order.
In the
ratio of the
sum
by envelope
detection.
second technique of intermodulation-distortion measurement introduced by the International Telegraph and Telephone Consultation Com-
mittee
method,
-.
3 3 3
Frequency
3,
dis-
o>i
to the
505
sum
of the
order distortion
is
products to the
sum
CCIT method
is particularly useful for evaluating performance of a communications channel, where the driving signal and odd-order distortion
products can be made to fall within a specified passband. Whether harmonic measurement or one of the intermodulation methods should be
used depends mainly on the application and what kind of instrumentation is available.
One advantage in making intermodulation distortion
measurements is that the driving signals do not have to be spectrally
pure since their harmonics, if moderate, will not cause significant intermodulation distortion and the harmonic frequencies themselves do not
generally
73-8
fall
Slew Limiting
rate, expressed in volts per second at the input terthe time derivative of input voltage at which some
stage in the amplifier is incapable of delivering enough current to charge
minals, or det/dt,
is
de
= c
(13 - 8- 1)
i
t
any stage
rate,
gain
is after
When
first
in
stage
is
an amplifier
is
is
e*
in that stage.
= E
506
maximum
^=
and
ir
rad, or
(13-8-2)
o>E cos ut
dt
The maximum
is
Knowing
until distortion begins to increase very rapidly.
can then calculate de/dt by means of Eq. (13-8-2).
In a servorecorder, the
writing rate.
maximum
ramp waveform
or a triangular
wave
and
w,
one
maximum
of voltage can
be
applied to the input of a recorder amplifier, and the slope can be increased
Similarly, a voltuntil the recorder no longer increases its writing rate.
rate observed
maximum
slew
the
and
applied
be
can
function
age step
directly.
CITED REFERENCES
2.
3.
1.
4.
5.
6.
Linvill,
CHAPTER FOURTEEN
MEASUREMENTS ON
TRANSMITTERS AND RECEIVERS
D. B. Hallock
Collins Radio Company, Cedar Rapids, Iowa
with assistance of
W.
B.
Bruene
Collins Radio
Company
The procedures
507
508
be stressed that these special tests are quite subject to variaby the particular method of performance
specification.
As an example, the crossmodulation performance of a
receiver may be measured with the undesired signal producing an audio
output of 10 dB below reference level or 3 dB above the quieting audio
level.
The basic measurement technique remains the same.
It should
74-7
General-performance Characteristics
video signal.
AM
stringent adjacent-
level, linear
AM
509
transmission.
figure)
since
noise sources.
PM
FM
Entertainment FM broadcast
200-kHz channels with 75-kHz
peak deviation. The transmitter audio distortion is very low (3.5 percent maximum) and background noise is greater than 60 dB below peak
Frequency-modulation transmitters for stereophonic broaddeviation.
casting use a subcarrier system which requires excellent phase linearity
modulation to transmit binary data.
108-MHz band
in
The
teristics similar to
deviation
is
those for
peak deviation. The aural carrier is positioned 4.5 MHz above the
visual carrier, and the receiver recovers the audio with a discriminator
tuned to the 4.5-MHz beat between aural and visual carriers.
The preceding paragraphs have briefly outlined some of the chief oper-
510
ment
commercial and government publications relating to radio equipis given at the end of this chapter.
specifications
74-2
Basic Measurements
It has
also
quency
AM
audio output
is
is often observed that the receivnoise rather than a tone at the undesired-signal
dyned
14-3
511
of
one receiver could easily be tuned to the local-oscillator frequency
problem.
interference
another receiver, and this causes an
interConducted Interference. A test setup for measuring the conducted
For
ference present on power or control leads is shown in Figure 14-1.
reproducible results, the test should be carried out within a shielded room
sheet of
or chamber and the equipment mounted on a large, grounded
to the
bonded
be
should
test
under
equipment
The
brass or copper.
grounded sheet in the same manner as it would be in an actual operating
The current probe is a toroidal transformer through which
installation.
together with
natively, a series of uncalibrated receivers may be used
emission
signal generators to measure the amplitude of each narrow-band
limits of
MIL-STD-461
kHz
to 50
The
test
MHz
are
Current probe
Unit under
test
Control leod
Control
box
"T
'
--::}
Q)
ion]
Isolation
1
Calibrated
Current
probe
receiver
trans
former
X"
lO^F
feedthroughj
capacitor
X
FIG
4-1
indicates a
Power source
Conducted-interference
bond
to a
test setup,
common ground
plane.
512
60
50
40
6 30
20
10
10 kHz
II llll
100 kHz
llll
1MHz
Ml
10 MHz
II
100 MHz
Frequency
FIG
leads.
4-2 Conducted-interference
(From MIL-StD-461.)
simple parallel-tuned
ment frequency
is
LC
well
network in a shielded box when the measureremoved from the transmitter output frequency.
that control the actual emission of desired output. When the rejection
network is used, its frequency response must be known or measured to
allow calculation of the power actually flowing at the antenna terminals.
There are two possible sources of spurious responses in this measure-
unit
under
test
Attenuator
FIG
ence
test setup.
4-3
Rejection
network
Calibrated
receiver
Antenna conducted-interfer-
513
if used.
The EMI meter is a superheterodyne receiver with its own spurious responses, notably the image frequency. Network or high-order receiver spurious products can be identified by increasing the attenuator setting by a small amount (3 dB) and
noting the indicated level decrease on the EMI meter. If it is the same,
is
being Observed.
If
the
EMI-meter decrease
The present
plied to the
mW
as the
maximum
is
is
at least
state of the
available
may
As an example
EMI
meter.
MIL-STD-461.
test specification
but can
m_ 40
o>
o
a.
50
s.
60
.._
-10
10
20
30
FIG 14-4
interference.
limits
for antenna
40
dBW
50
conducted
60
70
514
The emission of unwanted energy from equipform of a nearby magnetic field or an electromagnetic
field escaping through seams in the enclosure or other openings.
The
magnetic-field measurement is carried out with a loop probe, while the
Radiated Interference.
ment can be
in the
electromagnetic
field
emission
is
known
(dBpT).
E =
V2fANB X
by
10- 4
E =
where
is
rms-induced voltage,
/ = frequency, Hz
A =
loop area, cm 2
= number of turns
B = flux density, T
in the loop
The standard loop has 36 turns of wire and has an area of 139 cm 2 The
frequency range is specified from 30 Hz to 30 kHz, and the test limits are
140 and 20 dBpT, respectively.
These limits result in an induced voltage
of 666 mV at 30 Hz and 0.666 /xV at 30 kHz.
Determination of the radiated electromagnetic field is carried out in a
.
shielded
least 1
receiving antenna
is
at
test.
The walls of the room should be covered with carbon-filled
absorption material to suppress reflection of electromagnetic energy.
under
rf
STD-462
The
specification in
MIL-
when
requires measurements
and antennas
Quantitative measurements
are needed of performance degradation as a function of unwanted signals
conducted through power and control wires into the equipment or radiated into the enclosure from external sources.
are simultaneously operating at one site.
515
frequency range,
Conducted Susceptibility. In the 30-Hz to 50-kHz
leads (or control
susceptibility to undesired signals on the input power
signal in series
leads if required) is measured by inserting the undesired
inductance
with the lead and using a transformer with low leakage
(1
mH
-*
Unit under
Power
Sinusoidal
oscillator
amplifier
ll \
test
I
11
Power
source
ac
Null
network
voltmeter
'
susceptibility
measurement (30
The
test setup is
sists
of the
under test and then measuring a critical performance parameter
noise-figure
or
sensitivity
a
be
might
this
equipment. For a receiver,
measurement; for a transmitter, it might be the carrier signal-to-noise
ratio.
To measure conducted
50 kHz to 400 MHz) the
11
frequency of measurement.
measuring
used in the test to ensure that the rf voltmeter is accurately
As in
test.
the voltage at the power input terminals of the unit under
Coupling capacitor
/
rf signal
generator
Power
amplifie
|50
1
rt
Unit under
test
rf
voltmeter
FIG 14-6
lead
(50
Power
source
Conducted-powermeasurement
susceptibility
kHz
to
400 MHz).
516
74-4
of the equipment.
differ
This noise is band limited by the audio response of the receiver, and, in
the case of superheterodyne receivers, is often band limited
by the intermediate-frequency selectivity. The noise power comes from the ampli-
thermal (Johnson) noise of the receiver input-terminal source resistance (the resistance that the receiver input "sees") and from the
fied
noise
sources within the receiver itself.
When the receiver is connected to an
antenna, additional input noise is introduced from galactic, atmospheric,
signal
In
is
is
due
FM
This property of a receiver in the presence of an unmodulated input signal is termed the quieting characteristic.
AM
signal level
signal
is
In
FM receivers,
With no
517
As
the signal
is
but
ping" sound.
amount
of
is
and
hum
dB below
maximum
Amplitude-modulation and single-sideband quieting characteristics commonly exhibit changes in slope as the stages in the receiver automatically
change gain at different rates as a function of input signal.
Measurement
sideband, and
trol
FM
receivers
and
of
is
AM,
single-
AM
FM
AM
FM
receivers
is
that with signal, and then measure the noise power output.
Fixed and mobile voice-communication
Impulsive-noise Rejection.
receivers commonly employ squelch circuits, which mute the audio out-
is
muted
518
Impulse
generator
1
Mixing
pad
Test
receiver
af power
meter
rf signal
generator
FIG
4-7
lated)
is
set to
some
specified level
MHz,
rf
Many
noise ratio
is
open only in the presence of a usable signal-toa given noise power spectral density and
made by applying
then increasing the rf signal until the squelch opens. The noise generator
then turned off and the drop (in decibels) in receiver power output is
This test is repeated for several values of noise level to characternoted.
The drop in receiver power output meaize the squelch performance.
is
sured in this way is actually the ratio of the signal plus noise to the noise
and can be converted to signal-to-noise ratio by noting that
+NN
When
519
connected to 50-ft sources and load, the resistance lookis 50 2 and the attenuation between any two ports is
6dB.
74-5
Sinad Sensitivity
This type of receiver sensitivity test is used primarily with voice-communication equipment since it gives a measurement that is more indicative of the expected intelligibility for a given signal strength than the
usual ratio of the signal plus noise to the noise. Sinad is an acronym
Although this test is commonly
of "signal plus noise and distortion."
specified for
FM
equipment,
it is
equally applicable to
AM
(dual sideconsists in
band) and
applying a signal of known amplitude and modulation to the receiver
antenna input. The audio power output is then measured; this power
consists of the total sum of the recovered modulation, the distortion
products produced by receiver nonlinearity, and the amplified thermal
single-sideband.
than the amount of dB sinad measured for negligible error due to audio
feedthrough. Additionally, the rejection band should be narrow compared with the audio bandwidth of the receiver, since it is desired that
Signal
generator
Test
input
receiver
afj
output
I
Null
of power
meter
network
FIG 14-8
Sinad sensitivity
test setup.
The
flat loss of
the null
520
metering.
The output meter of these instruments is usually not a true
rms (power) meter; the error is negligible with the meter damping that is
provided.
The null depth is typically in excess of 80 dB, allowing sinad
measurement down to at least 60 dB. This is very adequate since the
harmonic distortion of the receiver will not usually permit sinad ratios
as high as 60 dB.
14-6
Modulation-acceptance Bandwidth
FM
In
voice receivers a measure is needed of the maximum deviation
that can be accepted for a certain degradation in sinad sensitivity.
This
The measurement is made by applying a signal that is frequency modulated at 1,000 Hz and has a deviation that is two-thirds of the rated system
deviation.
The signal level is initially set to 1,000 nV, and the receiver
audio gain control is adjusted for rated power output. Next, the signal
reduced until a sensitivity of 12-dB sinad is obtained. Finally,
is increased 6 dB over that just obtained for 12-dB sinad,
and the deviation is increased until a sensitivity of 12-dB sinad is again
level is
and audio
distortion.
It therefore
14-7
The techniques
of measuring
AM
rf,
apply to complete receiving systems. Similarly, noise figure determination of a receiver embodies the principles given in Chap. 4.
Since many
equipment specifications give one form of sensitivity requirement, but
not both, it is necessary to relate the two for design purposes.
Figure 14-9 shows an equivalent input circuit for a receiving system in
521
Rs
'
.
FIG 14-9
Perfec+
Rn (?)
v^
s?'
.receiver
input.
QS/AQin
(S/AQpointl
(
K
(S/N) vointi
(S/AOout
"
where the signal-to-noise ratios S/N are available power ratios. The
available signal power at point 1 is E a 2 /4R while the available noise
power is KTB, where K is Boltzmann's constant, T is the temperature in
degrees Kelvin, and B is the bandwidth of the receiver. Similarly, at
point 2, S is E, 2 /[(R t + R n )], and
With these relais still just KTB.
tions, Eq. (14-7-1) reduces to
F=l+^
R
(14-7-2)
F R, =
The
R,
-\-
Rn
total open-circuit noise voltage at the perfect receiver input (point 2),
is
En = VKTB(R, + R n
(14-7-3)
En = VlKTBFR,
Considering
power
AM system,
is
+N
___
S
where
now an
(14-7-4)
is-
_+ E^
(ME,)*
,,.__.
(14-7-5)
This
522
^-^
+ N/N) -
F =
The
4KTBR,[(S
is
(14-7-6)J
K
1]
just 10 log F.
S +
N =
,
,
10 (10-dB power ratio)
B = 2AFBW (AFBW =
=
E,
nV
F =
^
AFBW
(14-7-7)
and
postdetection bandwidths.
state response,
it is
W
where x
is
of poles, n.
bandwidth analytically by
The squared
[3] is
(14-7-8)
w/umb
for low-pass
5S3
4-1
Attenuation
a,
3
4
5
or
dB
5.4
3.8
3.65
3.46
3.38
QBW/fo
bandpass
for
Gn
response.
Jo
The response
ii*
when
a
By
Let
filters.
oj3ds is
|(? 12
'
dx
|
'
of the
= f"
Jo
x2
=
2n
sin
(14-7-9)
*/2n
is
rrWnormalized to unity.
10 log
(1
+ W 2n
The attenuation
in decibels
is
using the above equations, the values in Table 14-1 are obtained.
The bandwidth
14-8
Automatic-gain-control Characteristics
AM
AM
again measured.
The
increase in audio
5-
524
This
is
oscilloscope
rf
signal
r F
switch
generator
dc power
s upply
FIG 14-10
rf
of
Test
receiver
(j^\
V
T
1
v3
Manual
switel1
Automatic-gain-control
settling-time
test
setup.
done since
it
The
any point
in the
should also be slowly varied over the specified range while any discontinuities or peaks of audio output are noted.
Settling Time.
The manner in which a receiver output varies after
application of a step in signal input is of interest since it characterizes the
envelope distortion of
and single-sideband voice receivers and the
receiver.
rf signal level
AM
AM
accuracy of special
receivers used for navigation.
Measurement
of step response is made with the equipment shown in Fig. 14-10, which
applies a modulated-rf step input to the receiver and simultaneously
is
525
In an
AM receiver,
Countermodulation.
14-9
There are many specialized measurements taken on transmitters, determined by the particular equipment mode of operation and its specifications.
This section will discuss some of the transmitting measurements
applicable to systems and not considered basic device tests.
Single-sideband Distortion.
A basic method of measuring the linearity
of a single-sideband transmitter is to apply a two-tone test signal to the
input and to observe the output spectrum on a spectrum analyzer. This
has been discussed in Chap. 2.
A newer method of measuring the distortion produced by single-sideband transmitter nonlinearity is called the noise-loading test. This test
has been necessitated by the extensive use at high frequency of multichannel single-sideband data transmission. For example, a 3-kHz voice
channel may be used to multiplex by frequency division up to 18 uncorrected frequency- or phase-shift-keyed signals. When the signals within
a channel are uncorrelated, and channels are likewise uncorrelated to each
other, the rf-envelope signal in the rf stages of the transmitter approaches
noise with a Rayleigh amplitude distribution (central-limit theorem).
Since the distortion products of all the other channels appearing in one
channel tend to approach a noiselike signal, the error rate of the data
system is affected by the drop in apparent signal-to-noise ratio. The
noise-loading test provides a quantitative means of predicting digital data
error performance when the results of the test are known.
526
The concept
is
rather simple.
white-noise
is
and the output power spectral density lying in the notch is measured.
The ratio of the power in the same bandwidth with the notch removed
to that with the
notch present
is
power
FIG 14-11
Essential
single-sideband transmitters.
test (or
Band pass
Notch
filter
filter
100Hz-6kHz
test
input
rf
mixer
-? 0-10 KHz
Amplifier
Transmitter
under
af
Calorimetric
rf
coupler
power
meter
rf
output'
rms
Distortion
detector
filter
-*j
Log
dc
volt-
ratio
meter
rWoonvert^r
Crystal
oscillator
527
rms
Signal
filter
detector
FIG
1 4-1 2
Block diagram of automatic noise-loading test set with
continuous display of single-sideband noise power ratio.
3-dB bandwidth
of 50
Hz
is
100 Hz) so that noise power outside the notched band does not limit the
measurement dynamic range to less than 50 dB. A crystal filter, typically with a 100-kHz center frequency, is necessary for the distortion
filter in order to meet these selectivity requirements.
The signal filter is
not strictly necessary since the loss of total power due to the notch can
be calibrated out.
When
power
ratio (single-sideband)
The
which are
signal filter
is
rela-
also
filter
is
528
The
voltmeter.
power output
calorimetric
directly,
is,
The
measured to
is
Voltage,
<r
1
2
3
t
of
0.0796
0.0484
0.0108
0.000898
0.005
0.005
0.003
0.0002
0.2<r,
A representative performance requirement for a single-sideband transmitter can be a 40-dB single-sideband noise power ratio measured at
rated noise power.
529
This concept may be further illustrated with Fig. 14-13 in mind. Suppose a complex waveform is passed through a linear network from A to B.
The sinusoidal components at point A must have a certain time relationship to produce the complex wave; for example, component Ai passes
through zero at time h with positive slope, and
A 2 passes through zero with positive slope at t 2
.
Now,
Network
B must
be exactly
FIG 14-13
tion
of a
Representanetwork with
nent Bx must cross zero at T + h and the component B 2 must cross zero at T + t 2
Thus the value T must be
independent of frequency.
Absolute time delay may be thought of as phase delay. If a sinusoid
with a frequency of 1 Hz enters point A of Fig. 14-13 and suffers a time
delay of 1 sec by passing through the network to point B, it may be seen
that the equivalent phase retardation is 360 or 2tt rad. If the phase
retardation were at 2 Hz and the delay still 1 sec, the phase retardation
would be 720 or 4tt rad, and so forth. Therefore, the absolute time
delay or phase delay is rabs = 0/, where
is the phase shift in radians
and o) is the angular frequency in radians per second (2tt/). This relation
is used in sketching the relationship between constant absolute time
delay and the corresponding phase shift versus frequency in Fig. 14-14.
Note that for the time delay to be constant versus frequency, the phase shift
must be linear with frequency and have a slope equal to the delay. Additionally, each component frequency must be attenuated (or amplified)
by an exactly equal amount if there is to be no distortion of the waveform after passing through the network.
Communication systems are, by nature, band limited; that is, only a
certain range of frequencies is passed by the receiver or transmitter.
For
.
Frequency
(a)
FIG 14-14
phase
shift
The
relationship
of a wave.
530
MHz
to 2.003
MHz.
may
be characterized
time delay
Td
is
(14-9-1)
dc
envelope delay,
This quantity is also referred to in the literature as group or
rf envelope
certain
a
having
frequencies
spaced
since a group of closely
It is evident that
T
amount
the
d
by
delayed
equally
almost
will be
measurement of the quantity A|8/Aw, where A/3 is the change in phase
approximate Td when Aw
for an increment in frequency Aw, will closely
.
is
documented
methods of
phase versus
an envelope
AM
carrier
Suppose that an
of frequencies.
If the
wo is passed through the system.
modulating frequency is w m then the sideband
frequencies wi and W2 are wo w m and wo + wm
Upon passing through the sysrespectively.
tem these sidebands are rotated with respect
,
>
w,
aj
w2
to the carrier
sideban(i)
and
by the
/3 2
'"
region
t"<
"<Z
is
angles
/So
/3
(upper
/3i
(lower
sideband).
and may be
Aw producing
imated by
Td
AjS is
o>
called
.
A/3.
531
is
approx-
A0
03m
where
A/3 is
a>
Corre-
spondingly,
Td =
A<f>
360^
where A<j> is in degrees and fm is the modulation frequency in hertz.
These equations say that the time delay may be measured by finding the
input-to-output phase shift (A/3 or A</>) of the modulation envelope and
dividing by the modulation frequency.
It is to be noted that a real system does not have /3 2 j8 equal to
Therefore, in general, the output envelope does not exactly
/3o /3i.
have the shape of the input envelope. A compromise is necessary in the
choice of a> m such that A/3 is accurately discernible (high co m ) and the
envelope distortion
may
rf signal
generator
i
rf
Test
Envelope
transmitter
detector
'
>
Dummy
Phase
load
meter
i
oscillator
_^
Envelope
detector
direct
532
by obtaining a Lissajous pattern. The accuracy and resolution requirements of the phase comparison are quite high; a 5.4 change in phase A<
equivalent to a 500-/xsec change in delay Td when the modulation frequency fm is 30 Hz. The accuracy of the envelope detectors may be
checked by interchanging them and repeating the measuring run. The
average of the two runs is then due to the delay of the transmitter under
is
test.
is
less
than 360.
The
results of a
Hewlett-Packard Company.
1.0
1.5
Input frequency,
FIG
4-1 7
Time delay of
KHz
a single-sideband transmitter.
swept frequency
533
dis-
FM
rf
is
AM
AM
AM
signal.
AM
AM
AM
by a
resistor equal
in
AM broadcast-
on the rf envelope, the distortion being brought into the instrument from
a sampling loop mounted on the antenna.
noise test
This measurement is similar to the
Residual FM Noise.
except that an FM detector is coupled to the output of the transmitter.
The FM detector may be a deviation meter or an FM receiver whose
quieting characteristic is greater than the FM noise level intended to be
measured. For FM transmitters, the FM detector output ratio for fullrated system deviation and no audio input is the residual FM noise level.
For
equipment, the FM detector output, both with and without AM,
is measured with the reference being the FM detector output with an
FM input signal having some standard amount of deviation, such as
AM
AM
1,000 Hz.
534
international
This follows
AM
e(f)
where
= E[l
m(t)] cos
wc t
(14-9-2)
= instantaneous output-voltage
E = peak carrier amplitude
m(t) = modulating function of time
= angular carrier frequency
oj
function of time
e(t)
When
m(t)
is
the carrier, displaced in frequency from the carrier by integral odd multiples of the square-wave frequency.
On the other hand, if the rise and
fall times are too slow, the code elements tend to merge at high sending
rates,
techniques of Chap. 5.
To obtain a stationary spectrum, the keying
input to the transmitter must be a constant frequency. Since the transmitter normally provides waveshaping, the keying input may therefore
be a square wave of frequency / equal to the highest expected keying
is
in hertz
adjusted over the expected range of keying rate while the leading and
output are observed. Although the
direct connection of a high-frequency oscilloscope is shown in Fig. 14-18, a
trailing edges of the transmitter
Square
Relay or
electronic
keyer
oscillator
rf
coupler
Key
nput
rf
c
Xjtpu *
1
Trigger
;
'
FIG 14-18
Oscilloscope
L*
v_
Dummy
load
at
output/
input
Audio
oscillator
input
i
Key
Audio
output
(XMT)
RCV
rf signal
Power
rf
sampl
Dual beam
osci lloscope
fc
Trigger
4-1
generator
attenuator
Keying
function
generator
FIG
535
scope.
may
diode detector
oscilloscope
is
insufficient.
play should show smoothly rounded leading and trailing edges. The
transmitter should rise to full power, however, at the highest keying rate.
A noteworthy byproduct of the time display is observation of amplitude
in the interval
trailing edges.
Droop
in this interval
certain
another transmission.
A method of measuring the transfer time between reception and transmission and vice versa by using a dual-beam oscilloscope is sketched in
In this case, the transceiver is assumed to be a single-sideFig. 14-19.
band type such that a single rf output is generated by the transmitter
section when a single audio frequency is applied to the audio input
The keying-f unction generator can be simply a relay actuterminals.
ated repetitively by a low-frequency square-wave oscillator. The horizontal sweep of the oscilloscope
is
triggered
by
536
rf
the
rf
signal generator
output attenuator of the rf generator. Alterlevel may be temporarily applied directly to the
rf
With
Communication System Interference. In installations with multiple transand antennas operating within the same frequency band (but
not on the same frequency), severe interference can arise when the isolaceivers
and transceiver
Transceiver
is
537
AM
voice equipment.
the transceivers will be assumed to be
A is modulated to its full capability (80 to 90 percent) at
400 Hz from an audio-frequency oscillator. The rf output is applied
to a power attenuator capable of dissipating the transmitter power and
poses,
Transceiver
Power
Transceiver
of
oscillator
attenuator
Mixing
pad
,
Transceiver
af
VM
Frequency
rf signal
Frequency
control
generator
control
A
FIG 1 4-20
test.
having an insertion loss equal to the actual antenna isolation less 6 dB.
An additional 6-dB loss is incurred in the mixing pad (which is identical
with that discussed in connection with Fig. 14-7), which brings the total
attenuation between transceivers equal to that between antennas in the
installation.
The audio output of transceiver B is monitored with an
(Headphones or a loudspeaker at this point
audio-frequency voltmeter.
are a useful accessory to identify the character of the interference.)
rf
signal generator
is
An
also applied to
Two
can be made with the setup in Fig. 14-20. The first test
and is performed with no signal from the rf
signal generator and with the squelch control of transceiver B adjusted
so that a signal of typically 5 \iS is required to unmute the receiver secFrequency control B is set for some desired channel within the
tion.
is
tests
a squelch-opening test
quency control
is
With
exercised through
transceiver
all
transmitting, fre-
channels noted which cause the squelch of transceiver B to open as indicated by the audio-frequency voltmeter or other audio output indicator.
The test may be repeated for as many receiver channel settings of fre-
quency control
The causes
com-
predominant and
They
in a superheterodyne receiver.
The
arise chiefly
Hf
+ LfL
(14-9-3)
and L are integers of either sign associated with fH the higherfrequency input, and fL the lower-frequency input, and / is the mixer
where
538
receiver.
an/
of 20
to
60
to
80 dB
for orders
above
the tenth.
The second test performed with the equipment in Fig: 14-20 is closely
akin to a standard crossmodulation or blocking test. Transceiver B is
channeled to a desired frequency in its range, and the rf signal generator
used to apply perhaps a 20-mV signal. For very high frequency airapproximates the minimum signal received while
in flight; the test can be repeated for other desired signal levels if required.
With transceiver A transmitting and modulated, frequency control A is
used to channel the transceiver over its frequency range. At each
is
ceiver
B by
it is
The
in
1-MHz
steps.
larly in
14-10
is
The successful design, construction, and measurement of radio receiving and transmitting equipment require detailed knowledge of the per-
539
These specifications generally cover the minimum performance requirements for an equipment. Certain documents, notably those of the
Federal Communications Commission (FCC) and the Radio Technical
Commission
(RTCA), define the legal minimum specifiand operation of radio equipment in the
for Aeronautics
The actual performance of radio receivers or transmitmust exceed these minimum values to be competitive in the
United States.
ters usually
electronics market.
listed,
purposes.
N.W.
two-way
The
fire,
FM
taxicabs, forestry,
minimum
is
usually
much
standards specified.
Publication
Title
RS-152-A
RS-204
mc
Minimum
470
PM
RS-237
FM
or
PM
Transmitters (25 to
FM
or
Receivers
Minimum Standard
Using
FM
PM
for
in the
RS-240
Electrical
RS-250A
Electrical
RS-316
Minimum Standard
TR-107
Electrical
Transmitters
Performance Standards
FM
Facilities
PM
FM
Minimum
mc-108 mc)
Standards
Equipment
mc = MHz.
Relay
munications
or
Equipment 25-470 mc
Performance Standards for
Broadcast Transmitters (88
TR-120
for Television
for
Land-mobile
Selective
Signaling
540
2551 Riva
Road
(ARINC)
an association supported by
is
Specifications of
difficult
to obtain in production
equipment.
Publication,
characteristic
No.
Title
546
547
551
552
566
568
570
572
533A
VHF
Satcom 4 System
HF = high frequency.
SSB = single sideband.
3
VHF = very high frequency.
4
Satcom = satellite communications.
6
ADF = automatic direction finder.
1
Whitman Boulevard
New York 11746
Melville, L.I.,
The
is
It is
Rules, regulations, and specificaby agencies of the United States government are obtainfrom the U.S. Government Printing Office, Washington, D.C.
able either
listed
of radio.
Each branch
541
(some
specifications for their specific requirements; these may be obtained
with restrictions) from the U.S. Department of Defense, Washington,
D.C. 20360.
Federal Communications Commission
Office
following volumes and parts thereof define the FCC regulations which
must be adhered to in the performance and use of radio equipment in the
United States. These regulations chiefly cover any equipment which emits
The
rf
(noise figure)
radiation limits.
lume
Title
Part
Allocations
and
Radio
Treaty
Matters;
Frequency
15
General Rules
Experimental Radio Services
Radio Frequency Devices
73
74
81
Stations on
83
85
87
89
Aviation Services
Public Safety Radio Services
91
93
95
97
99
21
23
25
Services in Alaska
6
6
6
7
7
7
Services
Communications
of
Aeronautics
The
RTCA
is
Copies
may
542
Title
DO-48A
Minimum Performance
cation Transmitting
Standards
Airborne
Radio Communi-
DO-49A
Minimum Performance
DO-57A
DO-86
DO-92
1,800-2,000 Kilocycles
DO-93
Minimum Performance
Standards
Equipment
DO-94
DO-95
DO-109
DO-110
DO-111
DO-112
Equipment
Minimum Performance
DO-114
Standards Airborne
VOR
Receiving
of
108-118 Megacycles
Minimum Performance Standards
DO-123
Altimeters
DO-124
ing
DO-132
Minimum Performance
Standards
Airborne
Receiving Equipment
1
2
3
ATC =
VOR =
Military Specifications.
MIL
are available
D.C. 20360.
in that
The
third specification
information in
DCA-CIR-175-2A
543
on
Publication
MIL-STD-461
ments
for
Require-
Equipment
MIL-STD-462
MIL-STD-188C
ment
of
high frequency.
It
is
The
to very
the U.S.
Government Printing
Office.
Title
Publication
DCA-CIR-175-2A
1
DCS =
DCS
Engineering-installation Standards
Manual
CITED REFERENCES
1.
2.
3.
4.
Gannaway,
AM
vol. 9, 1930.
5.
6.
Brown, T. T.: Mixer Harmonic Chart, Electronics, April, 1951; or Olsen, W. R.,
and R. V. Salcedo: Mixer Frequency Charts, Frequency, March-April, 1966.
Myers, R. T., and T. A. McKee: Receiver Spurious Responses Computer
Improves Receiver Design, IEEE Trans. Vehicular Commun., March, 1966.
CHAPTER FIFTEEN
MICROWAVE SIGNAL
SOURCES
From
notes by
Stephen
John
J.
F.
Adam
Dupre
Douglas Gray
Steve Hamilton
Wallace Rasmussen
Hewlett-Packard Company
Palo Alto, California
and
L.
Besser
The importance of having adequate sources of test signals for measurement procedures must be clear in the reader's mind by now. The development of excellent audio sources and rf sources for laboratory use has
kept pace with user requirements rather easily through the years, but
544
difficult.
545
As the
conventional
behavior.
parameters
Distributed-impedance
For years the most effective strategy in the face of these difficulties
has been to make both the circuit components and the active devices
distributed
by
very nature.
their
of termination.
finite elec-
were born.
The theory
of operation
is
They
will
that is
ator
is
75-7
The
known
character.
standard-signal gener-
Microwave
special
Transistor Oscillators
valuable in laboratory signal sources, but they have the following disadvantages: limited operating life of a few thousand hours or less, large
size, and operating voltages in the kilovolt range.
Some must
be operated in a constant and accurately specified magnetic field. Furthermore, accurate
and
are difficult to achieve in some oscillator
physical
AM
FM
tubes.
Transistor Characterization.
lator transistor
is its
The most
maximum
frequency of
oscillation,
/max
an
oscil-
Mason
[1]
546
and many others have shown that this frequency occurs where the unipower gain of the device is equal to unity. In terms of the scattering matrix S of the device, the unilateral gain [2] is given by
lateral
\su
U(S)
det
where S
is
U may
(15-1-1)
At microwave
frequencies the
scattering matrix
and
Sn\
- S&)
(1
is
is
straightforward.
It is
feedback oscillator and the negative-resistance oscillator, even though feedback oscillators exhibit negative resistance across
certain terminals.
The feedback oscillator consists of an amplifier and
a frequency-selective feedback network connected, for instance, as in
Fig. 15-1.
A z connection (or series connection) of the amplifier and
feedback networks is shown and is readily analyzed by using the z matrix
The condition for oscillation is that the determinant
of each network.
oscillator types, the
be equal to zero
+ Z/ll)<Zo22 + 2/22)
(Zo21
+ Z/2l)(z
[3],
12
or
Z/li)
(15-1-2)
where zaij and zfii are the z parameters of the amplifier and feedback networks respectively. An example of this type of oscillator is the Colpitts
circuit
shown
in Fig. 15-2.
Amplifier
Feedback
network
FIG 15-1
feedback
The
z-connected
oscillator.
FIG 15-2
lator as
nected
The
Colpitts
oscil-
547
By
letting
reflection coefficients at
frequency,
r9
IA
Resonant
circuit
Active
network
(a)
T
(b)
FIG 15-3
a specific circuit,
if
at
some
548
|r|>
^
t
and
arg T R
= -
15- 1 " 3 )
arg T A
This implies that the Smith chart trace of Tr versus frequency must
encircle (r^)- 1 plotted on the same chart at the frequency of interest.
As an example of this, consider the circuit in Fig. 15-36. A negative
resistance
is
common
collector cir-
to 4.3
GHz
is
apparent.
Tuning.
power
Piecewise linear
networks are required to linearize the voltage- versus-frequency relationThe ratio of maximum to miniship to better than within 1 percent.
mum capacitance of a varactor is limited by the breakdown voltage and
the parasitic capacitances, which makes the tuning of bandwidths exceeding one octave difficult to obtain.
field.
If
is
supplied
by
is
The equivalent
0.035
549
Ls
in.
1.2
Ga YIG
nH
400 G
2400 A I
0.16
T ^ 17.5 pF
nH I
(b)
(a)
FIG
5-4
shown.
Although yig tuning
is
its
equivalent circuit
as
disadvantage.
Example
1 .8-
to
4.2-GHz yig-tuned
Oscillator
Amplifier
Oscillator
TL\ to
FIG 15-5
1 .8-
to
4.2-GHz
oscillator
and
amplifier.
TL
microstrip
5 are
lines
transmission
550
FIG 15-6
The
its
plete
,8- lo
(a)
The
holder.
cuit in
thin-film cir-
(6) The
4.2-GHi
com-
oscillator.
mW
15-2
Solid-st&te
its
Microwave Amplifiers
in
microwave
amplifiers.
551
amplifiers.
very accurately,
is
of
brief
shown
is
in Table 15-1.
culator
broadband
'*
)'"'
*l
?o
'
FIG
5-7
One-port
device
signal source.
shown
552
TABLE
Type
5-1
of
microwave
Advantages
amplifier
Traveling-wave tube
Disadvantages
Broad bandwidth
High-overload capability
Usable to 100 GHz
Wide dynamic range
High gain and output
Limited
life
Warm-up
Large
required
size
power
Traveling-wave maser .... Ultralow noise
High gain stability
Low intermodulation
distortion
Large
size
High
price
Pump
required with
intrinsic
frequency
Intrinsic operating
temperature
Parametric
Tunnel diode
Pump
Ultralow noise
High
Low
Low
noise
Octave bandwidth
available
Transistor
Broad bandwidth
Low
price
required
price
output power
Burnout protection required
Difficult to stabilize
Low
/ < 8 GHz
Burnout protection required
tunnel diode
is
reliability
Ultimate
Smallest size
Currently limited to
/ < 200 MHz
Lowest cost
High noise
reliability
figure
considerably greater than unity. If a three-port circulator is connected to the tunnel diode as shown in Fig. 15-8, then a signal entering at
port 1 will be amplified and will leave the circulator at port 3. A vector
cient
is
63
Sll
S12
13
ai
S21
S22
S23
a%
S31
SS2
33
a3
parameters of
(15-2-1)
553
43
1*1-
03
by
u
Tunnel
diode
circuit
FIG
5-8
But, by definition, a 2
IY&2 and a 3
= T Lb
3,
or
ai
6i
62
\S\
bs
T T bi
T Lb 3
(15-2-2)
The
is
is
shown
in Fig. 15-9.
when
Note
FIG
5-9
554
+ S i(l S22IY)
+ szzTl + r rr Ai
+ r riA + tt tl a + r a TT T L &
S21S32IY
G
(iir,
A2 =
A3 =
A4 =
S21S12
S31S13
S32S23
S11S22S33
(15-2-3)
where A x
s 22 ri>
4)
SuS 22
S11S33
S22S33
S31S23S12
S13S21S32
S11S23S32
S22S31S13
S33S21S12
A typical tunnel-diode
shown in
The
Amplifiers.
amplifier circuit
is
Fig. 15-10.
for true
it is usually necessary
cascade several stages. A typical microwave-amplifier design is
described below.
The design is for a broadband, two-stage, 10-dB
module used in a multistage 0.1- to 2.0-GHz amplifier [13].
To achieve flat gain through such a wide frequency range, the designer
to
and
carefully
match imped-
Output
1
Stabilizing
network
Input
FIG 15-10
Tunnel-diode amplifier ac
circuit.
555
TL
FIG
Diagram of a 0.1
5-1 1
to
2.0-GHz
amplifier module.
and low reverse gain. The last two requirements are extremely important when modules are cascaded.
The circuit configuration of the twostage module is shown in Fig. 15-11. Although no numerical values are
given, the diagram is still instructive.
Rather than transistor equivalent circuits, it is suggested that the
measured two-port parameters be used directly in the design. All other
circuit elements are described by some two-port parameters {y, z, or
transmission parameters), and the two-port blocks are interconnected in
the computerized design program in the following manner:
1. All shunt connections are made by using ^/-parameter matrices,
Y =
Z
Fx
+F
2.
2.
'=
Zi
3.
+z
T = TiT 2
made by
2.
made by
The matrices can be continuously converted as the circuit is prostep by step [14].
The quickest way to do these conversions
to have a subroutine when necessary.
Once the complete circuit is
grammed
is
circuit
is
to be optimized for
minimum
556
dc bias
ferfc
^gen
Output
Input
>
FIG 15-12
filter
filte
filter
pass fj
stop fo
pass
poss
only
s/?
S^100 "
multiplier.
deviation from the desired gain and for best input and output impedance.
Since no approximations are used, the calculated and measured
results
75-3
The use
microwave signals
been discussed above. Microwave signals can
be generated by frequency multiplication with the use of varactors
at low
also
power
levels has
or step-recovery diodes
and by new
acteristics will
be compared.
Frequency Multipliers.
plier is
on
shown
many
in Fig. 15-12.
specific circuit
power
used depend
levels at input
and
is
made
to
application.
as
1/n 2
efficiency,
term n
is
is
zero [16,
by
17].
The
its
input
frequency.
step-recovery diode.
Mi crowa ve
igna
V =
where k
q
is
557
proportional to the
(15-3-1)
kq>
=
=
S<
proportionality constant
charge in coulombs
Usually 7
is
denned
tance) as follows:
= ( *
where S
=
Vb =
=
v
<t>
for v
<
<f>
elastance
i-pack
pock
Metal contact
tni
Diode construction used for most frequency-multiplier
diodes whether pure varactors, step-recovery diodes, or other type.
FIG 15-13
558
<r, where
When
all
/,-
is
input frequency
(15-3-2)
removed by negative
current, the
If this transition
time
much
is
That
is,
if
the
the charge.
circuit.
"
Front
contact
559
Back contact
20:
f,o
5*
O
o
\
:
J
-
SRD
\
:"A^\
\
(a)
FIG 15-14
acteristics: (a)
doping
profiles, (b)
recovery
frequency range of
1 /t
is
The equivalent
</ <
1 ft T .
= e*p(-)
Po
where
fc
Cmin
R,
V~b,
0o>iCmin (<t>
=
=
=
=
=
(15-3-3)
- vBr y
(15-3-4)
output frequency
cutoff frequency = (2C m n Rs)~ 1
i
breakdown voltage
560
Equation (15-3-4) shows that for the power, output increases directly
with the diode capacitance. As the capacitance is increased, however,
the input and output impedances decrease, since they are given by
= R.A
(15-3-5)
R = R.B
(15-3-6)
Ri
Values for
When
finite Q u
The varactor
1.
2.
3.
Its
is
reached.
of:
Being calculable
Being broadbandable as a doubler
that:
1.
2.
3.
circuit
power levels.
becomes extremely complicated
circuits.
The broadbanding
of doublers of this type is feasible, but the broadbanding of idler-dependent multipliers is quite difficult because of the
idler interaction with circuits that determine /, and /.Figure 15-15 shows a step-recovery multiplier circuit. Both lumped
circuits and distributed circuits can be used with the step-recovery diode
The operation of this type of
to form frequency multipliers [24, 25].
4.
multiplier
is
briefly as follows:
The diode
pulse)
561
Experimentally and theoretically it has been found [24, 25] that the
should be one-half to one cycle wide at /. This width is contp
trolled by Li through
pulse,
U = ttVl~C v
(15-3-7)
1<x<2
"%
The capacitance Ct resonates with L, at fi and should itself be nonresonant up to at least the frequency of the highest Fourier components
of the diode voltage (harmonics of /,- appear in the diode voltage up to
/ = 3/2t p ). The input resistance across the Ct terminals is known to be
approximately equal to the inductive reactance uuLi, which value can be
used to design the input matching network. The output inductor is
approximately equal to L, for optimum energy transfer from input to
output circuits. The output filter's first element can be either a series
or shunt capacitor which resonates with L at /.
A popular output resonator is shown in Fig. 15-156. Ideally, an impulse traveling down the
line reflects from C and returns to the diode just as it is closing again.
It then resonates in this shorted quarter-wave link until the energy in
the pulse has been dissipated in the load to the right of C.
t
resonates L/ at f
Li
f,
R6 |
Matching
^L
J_
_i_
output
1*0)
SRD
Ri^WiLj
Matching
network
M
T^Ta"
-prjTK
~~
"
If
iPT
Rest of
output
filter
pass f
J'
562
experimental
filter
been designed.
Spectrum analyzer
A'
'
563
inc
'ir+c
Hl^0
^dc)j I
Sweep
oscillator
VSWR
10--
2--
it
FIG 15-18
amount of
occur
driven
in
The
small
does
varacter
develop
very
rectification that
can be used to
self-bias.
that:
1.
The
circuitry,
ratios, is simpler
than for
Once set up and in mode, the self-biased step-recovery diode multihas a linear characteristic of power input to power output (an important feature in a large system that has realistic power-level tolerances
over temperature and frequency).
2.
plier
564
3.
The design
pass
filter,
fect filter
2.
overall multiplier)
is
BW
/,.
to less than
about 20
GHz
definite relation
power output.
= 2 GHz and 1
able
W at 10 GHz.
problem
in the practical
L m --U.8nH.
Cr =
Diode
'(SRD)
hpa
0300
0.1
pF
0.52I
1 5-1 9
Cross-sectional drawing of a X 1 (200 to 2,000 MHz)
The tuning bandwidth of
step-recovery diode frequency multiplier.
this multiplier is 100
(to -1-dB point) and P = 2.5 W.
FIG
MHz
pF
If it is larger
nation of
than the
565
Elimi-
all
stable even
put that is
put of 2.5
flat
within
dB
75-4
Solid-state
Microwave
MHz, and
a power out-
Oscillators [26]
modes
of operation for
devices are
somewhat
Gunn
or bulk-effect
broadly tunable, whereas the impatts provide higher average power output and are
useful to higher frequencies.
Gunn
is
mobility, and
of
carriers
in
most commonly used, although some other GaAs is a function of the applied E
intermetallic (III-V) compounds are someJj 16 * ? 9 on of
negative differential mobility, A,
,.
mi
times used.
These compounds have been corresponds to a region of negafound to have a region of negative differen- tive differential resistance.
tial mobility above a certain threshold electric field (3,000 V/cm) but below avalanche breakdown, as shown in Fig.
5-20.
Current flow is proportional to the mobility, J = en nE, and so a
is
dj
dE=
t
W.
dfx
en
dE
<a
when dJ/dE
is
negative
<
18 -*-
Actually, the basis for the diode that exhibits negative resistance was made by
T. Read, Jr., in his 1958 paper [32]. J. B. Gunn reported microwave oscillations
in bulk
III-V semiconductors in his paper entitled "Microwave Oscillations of Current in III-V Semiconductors" (Solid State Communications, vol. 1, p. 88, September,
1963).
566
I
-
200
40
Distance,
200
40
jjl
1 5-21
Sequence showing development of domain in 200-/um
sample of GaAs. The domain was initiated by a small doping
change near the cathode end of the sample.
FIG
where n
is
in region
Negative resistance
When biased
is
Bias
When
Gunn
\L
Gunn
FIG 15-22
device in
voltage
across the tank circuit can force
the terminal voltage below the
value required to sustain the
during
its
transit
domain
an
rf
circuit.
The
rf
quenched-domain mode. At
higher frequencies the domain
may
not
even
form
(LSA
in
Fig.
up higher than
device
shown
# threshold,
an accumulation
form and
cathode across the
is
con-
trolled
When
is
567
the terminal voltage drops below that required to sustain the space charge
layer for the complete transit.
The limits on the LSA mode are often thermal. Gallium arsenide has
high thermal resistance that leads to overheating, which can impair the
negative mobility mechanisms of the device.
Therefore, LSA oscillators
are often pulsed, rather than continuous-wave, devices.
Broadly tunable
oscillators
several laboratories.
in
Power outputs
for
hundred milliwatts
in the 4- to
to be reasonable,
100 GHz.
AM
FM
Impatt
down
When
568
-90i
Read
-100'
oscillator
A-
-110
1-120
Gunn oscillator
-130
-140
100 kHz
10 KHz
kHz
This pulse of charge enters the drift space (width W) where the E field is
7
high enough to maintain saturated velocity of the carriers (10 cm/sec).
trancharge
the
circuit
while
external
The particle current flowing in the
space.
drift
the
current
in
average
to
the
is
equal
space
drift
sits the
Since the velocity is constant, the current (at the contacts) is constant
from the time the charge enters until it leaves the drift space. If the
= %f the generated square pulse of current will
drift-space width is
be exactly 180 out of phase with the ac voltage that generated it.
Idealized current, voltage, and charge waveforms are shown in Fig. 15-25
M ici
igna
569
-'displacement
FIG 15-24
Read
structure
and
Avalanche zone
somewhat
less like
process technology
is
further advanced.
Gallium arsenide
is
Iporticle
a
-^displacement
-G
wcr
FIG 15-25
An
ac cycle
in a
its
Its chief
Diode
Diode
current
and
primarily
570
drawbacks are
its
arsenide technology.
considerations.
device.
1.5
10~ 4
rent of 150
mA
to produce
of rf output.
This corresponds to a
2
The
a power density of 10 5 W/cm
state-of-the-art data of Fig. 15-26 were obtained at Bell Telephone
Laboratories by using diamond heat sinks, which have 10-times better
A/cm 2 and
GHz
with 200- to
300-mW
output.
100
50
10
100
fQ (gHz)
FIG 15-26
solid-state sources.
1000
3X
fa
II
T(WM-
571
n L
Impatt diode
Impatt turnable
(a)
oscillator cavity
X.
HirI
dc bias
circuit
Impatt diode
(b)
Double slug
tuner
Two
FIG 15-27
impatt-oscillator
double slug
circuits:
(a)
impatt-tunable
tuner.
The
FM
High-Q
A/L = tt2
FM
FM
noise, as
can
noise [22]
kToBM
p
(15-4-2)
^T
(15-4-3)
is
Sufficiently far
on the bias
circuit [23].
When
the diode
is
is strongly dependent
terminated improperly, para-
572
AM
designed, the
4kT
PxMnoi..
2
At frequencies
BM/P
AM
75-5
/ 2<?et/
further
FM
/1K A A
(15-4-4)
away from
AM
dB
per octave.
noise data for free-running impatt oscillators are
Typical
and
given in Fig. 15-23.
of the oscillator
Oscillators
ability, long-
have several important parameters: power output, tunand short-term stability, AM and FM noise, and capability
for modulation.
wave Gunn-effect
devices.
stages to achieve the desired output frequency, and at each stage a very
high power device (for that frequency) is required. The power output
of a
is
less complicated
with the step-recovery diode multiplier.
Mechanical or electrical tunability, or both, of oscillators is an imporEven a small amount of tunability may
tant feature in many systems.
make the difference in whether a parametric amplifier performs or not,
system is achievable. Ceror whether a low-noise, phase-locked
tainly the most tunable of the group is the low-power Gunn-effect oscillator.
Devices of this type that are yig-tuned have been built to tune
FM
with 40
13
GHz
[30].
Impatt
oscillators
573
mW
mW
chain.
To
fundamental
FM
oscillator cavity, the presence of this signal tends to "lock" the differ-
made
as in Fig. 15-29.
two
An
identical.
onto a holding capacitor once per cycle of the low-frequency (50- to 100MHz) reference oscillator. Any drift from sample to sample in the sample voltage
is
amplified
and used
to correct the
microwave
oscillator fre-
AM
FM
function of fm the distance of the measuring slot away from the carrier.
example
The significance of local oscillator noise is illustrated in an
A single-ended mixer and an intermediate frequency close
in Fig. 15-30.
to the carrier were chosen deliberately to illustrate the point that local
,
AM
Circulator
G>
Microwave
reference
oscillator
dB
in detectable signal
pi
CF
Pw
degradation of 14
Load
lTu
oscillator
An
FIG
Microwave
574
0-
Sampler
dHocked
",i
Voltage-
rf output
controlled
Pulse=forming
oscillator
FIG
circuit
sampler as
phase detector.
Low-frequency
O/
reference oscillator
>"ref.
Quiet LO
LO;
AM
N/C
bandwidth,
LO
noise in bandwidth,
Wb
FkTB
PL0
For
mw
A7S--1CT
/I//C--10
^=
FIG 15-30
noise.
14
B
-
-12
10
FkTB
kHz
(-144 dBm)
(-120 dB)
4 dB degradation
dB
Degradation
in
575
level.
FM
systems balanced mixing does not cancel out local oscillator noise.
AM
Gunn
oscillator
AM
The
and
behaved
AM noise.
The
The
FM
noise deviation
A/drive
=
=
FM
FM
is
linear
is
Af = n Afdtiv
where A/
intermediate
FM
complicated.
60-MHz
(15-5-1)
deviation at output
deviation at input
is
Pn _ ( n
~ \
Po
A/drive
2/m
(15-5-2)
As a
D ,*200-400mW
^,*1mW
X15
J
MHz * SRD
\-srfMHZ
100
osc.
X7,
/?,*10-50mW
SRD
multiplier
~7ilc
L-band
trans,
osc
FIG
5-31
576
smaller
AM
the
AM
must have a
its linear power
be frequency modulated by
FM
75-6
may
also be
This section treats generators of standard microwave signals with frequencies above 2 MHz. The major characteristics of these signals are
specified and calibrated; the specifications can cover tuning ranges,
amplitude, type and degree of available modulation, output impedance,
and
the accuracy and resolution of these quantities. Also specified are spurious outputs such as distortion, hum, and noise.
It is important for the
engineer to understand the various configurations of standard-signal generators, so that he can compare and use them satisfactorily.
Block Diagram.
starts
fre-
We
Figure 15-32
is
heterodyne generator.
a block diagram of the direct generator, and Fig. 15-33
AM
2Frequ
Frequency
dc voltage
control
fm mod
in
Detector
tunin.g
Tunable
oscillator
AM/pulse
modulator
Step
attenuator
input
Video amplifier
(optional
Range
switching
FIG
5-32
Output
FIG 15-33
to 4.5
577
GHi.
shows a photograph of such an instrument. The desired output frequency is formed In a single tunable oscillator usually covering a range
of at least 2:1 in frequency continuously, with provision for switching the
range-determining elements to increase the total frequency range covered
mine the output characteristics of the generator and also serve to isolate
the basic oscillator from the effects of variations in load.
Figure 15-34 is a diagram of the heterodyne generator, usually associated with instruments that automatically sweep through a range of freHowever, sweepers usually have a continuous-wave mode of
In Fig. 15-34, the output frequency is the difference frequency between a fixed oscillator and a tunable oscillator. Obviously,
wide range on a single band is possible, and tin- user may consider this
quencies,
operation.
Step
Uttenuqtor
FIG 15-34
generator.
Simplified
block
diagram
of
heterodyne
output
signal
578
MHz
now determined by
how well
sta-
now determined by
largely
emphasized here.
cally about 0.05
FM
quency effects can be ignored, then they are random and can be treated
by means of the principles in Chap. 4. The power spectral density of the
phase variations (symmetrical about the carrier) is a valuable indication
and the curve is easily measured. If Wm is the phase
of oscillator quality,
deviation,
and
co
is
dB
FIG 15-35
6 dB /octave
ZQ
sity of
aim
log
Power
an oscillator.
spectral
den-
the oscil-
and Q
is
device.
Sf ( Um )
579
FM.
is
= f* = *.*()
(15-6-D
a curve similar to that in Fig. 15-34 can be plotted with the frequency
spectral density as ordinate [35].
The following paragraphs, although
Signal Generator Specifications.
far
material in Chap. 10
Output Frequency
Range.
is
also germane.
Characteristics
Standard-signal
generators
usually have
limit
multioctave
is less
than
fre-
GHz.
readout.
Many
in the
form
oscil-
lator frequency.
Resettability
is
is
a useful
580
and desirable
accuracy.
ual
FM,
drift,
and microphonism.
FM
is less
than
10
and the tests are accomplished by substituting a signal generator for the antenna signal.
A signal generator with spurious outputs
responses,
by
division
Output-level Characteristics.
In addition to the information on standard-signal generators in Chap. 10, several things need to be said specifically about generators in the microwave range.
For instance, electronic
leveling of output as the tuning dial is moved is used more frequently in
microwave generators than in those operating at lower frequency. Automatic leveling ranges from about 0.25 to about 1 dB across the dial.
Meter indication of output level into a matched load is required in
standard-signal generators,
Output
levels ranging
dBm
more
often,
power.
127 dBm
581
applications.
This
specified into a
the standard source impedance for a signal generator, but in the lower
frequency regions ( < 500 MHz) both 60- and 75-0 source impedances are
available and others on special order.
In any case a source vswr of less
than 1.5:1 is to be expected.
Radio-frequency Interference or Leakage.
Radio-frequency interference
on a signal-generator
data sheet, and yet they are among the major specifications defining a
standard-signal generator.
Radio-frequency leakage from a signal generator can create practical difficulties in measuring receiver sensitivities,
particularly if the receiver itself is not well shielded.
Additionally,
inaccuracies in attenuation at the microvolt level may result if strong
Most rf engineers become acquainted with a simiproblem very early in their careers when they attempt to put 150 db
The usual
result
is
output of the cable shows no decrease in level for the last 30 dB or more
of attenuation added.
This result is obviously not due to the attenuator
errors but to leakage from input to output connectors and from cables to
the attenuator.
coaxial connectors,
and
must be compatible
in leakage
582
nal generator
of modulation,
and
must be
it
calibrated.
Four forms
of
AM,
AM
AM
AM
AM
AM
this
mode
is
useful for
is
greatly increased
measurements involving
TV
and
video
modulation.
Pulse Modulation (PM). Pulse modulation is to the microwave freHistorically
is to the lower frequency ranges.
quency range what
microwave signal generators have included provisions for pulse modulation, but this has normally been accomplished by gating the oscillator on
and off in some fashion with resulting problems of "moding" and high
AM
FM. In recent years these limitations have been circumvented by the use of PIN modulators which absorb power on the transmission path. This technique allows the oscillator to operate continuously as in the lower frequency ranges and reduces reflections below levels
For reasons of cost and versatility, highthat cause frequency pulling.
quality pulse modulation for modern signal generators is usually provided
incidental
by an accessory instrument.
Frequency Modulation (FM). Applications of FM within the range
from 2 to 1,200 MHz are required of fixed and mobile communications,
broadcasting,
links,
583
requirements for all these applications in one signal generator, but many
signal generators are available covering substantial portions of the
applications.
Because of this broad range of generators optimized for
particular applications, only general specifications and characteristics will
FM
be discussed.
settable and accurately monitored over the
minimum, from
kHz to 0.2 percent of the
carrier.
The deviation, once set, should not require resetting with band
change and should be monitored by a peak-reading meter to within at
full
range provided
as
lation
is
in addition,
if
stereo
modu-
AM
FM
FM
situation,
It
is
stabilizer) is
some harmonic
dc signal.
In a simple form the operation of a phase-lock frequency stabilizer is
indicated in Fig. 15-36. A sample of the signal-generator rf is mixed
with a comb of harmonics of the reference oscillator. The difference frequency which falls within the intermediate-frequency amplifier bandpass
carries the frequency instabilities of the signal generator to be stabilized
and the relatively smaller instabilities of the reference oscillator harmonic.
The
difference frequency
is
rf sample
amplifier
from
generator
signal
Harmonic
Phase
mixer
detector
5L
To frequency control
input of signal
generator
Tuneable
i.f. reference
oscillator
or fixed
reference
oscillator
FIG
5-36
Ext. ref.
Ext.
i.f
ref
in.
(tunable)
584
is
which
is
is
Two
ator.
special requirements, a
For exam-
of 50
kHz
many
different
to 500
since the dc frequency control has usually pulled the. oscillator off cali-
For
monly used as
bration.
75-7
this reason
Amplitude Modulators
is
very com-
when a synchronizer
is
used.
programming that output level. The signal level of any oscilby varying the voltage on some electrode of the active
oscillating device, but this approach to AM usually has at least two disand
for
advantages.
voltage,
First,
M icrowave
Signal Sources
585
dQ+Q_ =
could also
affect
:P P- doped loyer:
intrinsic silicon loyer
\N N- doped
loyer
sec-
tion of
oscillator,
Let us
dc
first
itself.
Diode conductance
is
g+!at
by
(15-7-1)
where
I
lie
and where r
driven by ac
+
is
COS
(t)t
offset
=
=
Irt
by an arbitrary
[1
g dc [l
+ /
layer.
When
the diode
^TT^
rf/7dc
cos (
008
*)1
is
is
M+ *>J
(15-7-2)
586
=
=
I =
t =
nn =
Up =
d =
<7do =
where
arctan 1/cor
dc component of drive current
rf component of drive current I t t cos ut
<t>
Zdc
l{
carrier lifetime in
layer, sec
electron mobility in
hole mobility in
i-layer thickness,
cm 2 /V-sec
cm 2 /V-sec
layer,
layer,
cm
dc conductance term
The time-varying term in Eq. (15-7-2) represents a nonlinear resisThus long
tance to rf, but this term approaches zero for large cor.
charge lifetime is desirable.
The lowest frequency for which the diode is usable is determined by
the harmonic generation one is willing to accept. A typical diode might
have a lifetime t of about 100 nsec. With the diode biased so that its
100 ft, the second harmonic would be about 30 dB down at
There are PIN diodes now available with lifetimes in excess
of 1 msec, which makes them controllable microwave resistors at frequencies as low as 10 MHz.
An attenuating array can be formed by placing a number of PIN diodes
The diodes are spaced at quarter-wavelength
across a transmission line.
resistance
500
is
MHz.
band of interest [36]. In a simplican be assumed that the diodes are pure resistances. At
is
As
bias cur-
For an attenuating
array they are never biased to less than about 2 Q in order to maintain
an adequate match. Input match can be further improved by tapering
In Fig. 15-38 a tapered
the diode impedances at the ends of the array.
array is shown along with the input standing wave ratio as a function of
frequency.
The analysis used to derive these curves assumes that the
array is infinitely long. The theory has been applied to arrays as short
as seven diodes, however, with little difference between theory and actual
rent
is
results [37].
As was
tances; this
is
resis-
due to the
mounting structure must be taken into account.
Another class of modulators which has been extensively used in recent
years makes use of the PIN diode as a reflective element. A schematic
The PIN switches are
of this type of modulator is shown in Fig. 15-39.
simply two or four diodes mounted in shunt. The center conductor of
the transmission line which interconnects the diodes is a very thin wire.
This line looks inductive and at zero bias, in conjunction with the capaci-
wave
diode and
its
filter.
The
30
10
FIG 15-38
filter
70
50
Attenuator array
90
8
130
110
up
to 18
170
in transmission line.
structure
150
587
GHz.
When
this permits a
well-matched
At
PIN
through the other switch and the signals recombine in phase at the output port. When the PIN switches are biased to a reflection coefficient
of unity, the power reflects from the two switches and combines in phase
at port 3 and is absorbed in the 50-ii termination.
Bandwidths of up to
4:1 and operation to 18 GHz have been achieved with this type of
modulator.
3-dB
PIN switch
coupler
<50.a
> termination
>
3-dB
so n,
(4)
coupler
PIN switch
termination
FIG 15-39
PIN
diodes.
588
fref
LP
filter
Flat
detector
Unieveled
source
HP
PIN
HP
filter
line
filter
Directional
coupler
Leveled
output
tJ
Unieveled output
(a)
Freq.
Input
LP
"hr
Driver
filter
HP
Source
HP
PIN
#-
Pulsed output
filter
filter
(b)
LP
Shaper
\h Driver
filter
Source
HP
PIN
HP
filter
line
filter
(O
Unottenuated output
Si-Kl
HP
Source
LP
DC
filter
supply
PIN
Attenuated output
Freq.
HP
filter
filter
(d)
FIG 15-40
AM,
PIN
modulators (PIN
"lines").
75-8
Swept
signal generators, or
sweepers, are therefore important, and the swept range should be at least
an octave or the range over which a particular waveguide operates.
tuned
Sweep
out
Sweep
p nn rnn)
Osc
IN
mod
ckts
589
\LC amplifier
Marker
control
Morker
Ckts
Ext.
ALC
input
*
Referen
*
"
Modulation signals
FIG 15-44
oscillator.
to
CRO
no
o
Vertical
Sweeper
Device under
test
1
1
1
1
>
,
broaaoano
detector
Sweep out
FIG 15-42
Typical
setup
for
measurement
of
transmission
590
frequency limits, and output to be adjusted. Also, after the frequency has swept through certain increments, sharp marker pulses are
injected into the PIN modulator to produce a visual frequency calibrarate,
Another
circuitry.
function
microwave sweep
oscillator
is
the leveling
is
facilitate
serves
it
is
to
would seem
scheme
is
is
It is better
is
desired.
The schematic for a complete leveling loop is shown in Fig. 15-43. This
The detector is generally operating at a level where
circuit is not linear.
can be considered to have a square-law response so that V det = FkP
where F is the coupling factor of the directional coupler and k is the
The PIN modulator is approximately linear in dB
detector sensitivity.
it
P = P e-<
(15-8-1)
where a denotes the sensitivity of the PIN line in nepers per milliampere.
For small signals this characteristic can be approximated by
P =
P,(l
(15-8-2)
al)
l/REF
K)ET
Oscillator
FIG 15-43
PIN
modulator
Directional
coupler
P = P
From
iel
(15 " 8 - 3)
if
where a
yields
+ ccKV
+ cKFkP.
this equation it
P.
=
F
k =
1
-
PIN
591
is
high,
(15-8-4)
line sensitivity,
nepers/mA
coupling factor
detector sensitivity
electrically tunable.
15-44.
592
Modulation
input
2.2
GHz
Low-pass
PIN
fixed
Mixer
filter
modulator
oscillator
0.1
i,
2.3-4.2
GHz yig-tuned
oscillator
2.2
GHz
fs
for a
.r\
Hewlett-Packard
hetero-
LP
fo
'V
8699B
filter
\fu>
2.3-4.2
GHz
tnfuj -
m fs
4.5 _
4.0 -
2.5 -
FIG 15-45
m =
Mode
low-pass
filter.
to
2D
GHz
amplifier
>
Output
593
cent at 2
GHz
CITED REFERENCES
1.
Mason,
2.
Power Gain
S. J.:
CT-l, no.
in
Feedback Amplifiers,
IRE
Ku, W.
2,
Parameters, Proc.
of Scattering
IEEE
Terms
11,
pp.
4.
1617-1618, 1966.
Cote, A. J., Jr.: Matrix Analysis of Oscillators and Transistor Applications, IEEE
Trans. CT-b, no. 3, pp. 181-188, September, 1958.
Voltage Variable Capacitor Tuning: A Review, Proc. IEEE, vol. 56, no. 5, May,
5.
3.
1968.
:
IRE, February,
6.
Gordon,
and
New
1947.
J. P.,
H.
NH
Bloom,
S.,
linear Reactions,
8.
Esaki, Leo:
RCA
of Parametric Amplification
New Phenomenon
in
Using Non-
December, 1957.
Rev.,
IEEE
Trans.,
March,
1965.
10.
Microwave
11.
12.
13.
I.
C. Amplifiers, Intern. Solid State Circuit Conf. Dig. Tech. Papers, February,
1969.
14.
15.
16.
Besser, L.: Combine <S Parameters with Time Sharing, Electron. Design, vol. 16,
Aug. 1, 1968.
Mason, S. J.: Feedback Theory--Some Properties of Signal Flow Graphs, Proc.
IRE, vol. 41, September, 1953.
Page, C. H.: Frequency Conversion with Positive Nonlinear Resistors, Res.
Paper 2664, /. Res. Natl. Bur. Std., vol. 56, no. 4, April, 1956.
594
17.
Manley, J. M., and H. E. Rowe: Some General Properties of Nonlinear Elements Part 1, General Energy Relations, Proc. IRE, vol. 44, pp. 904-913, July,
1956.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
Gray, D. A.:
How
November, 1964.
to Design
CHAPTER SIXTEEN
MICROWAVE SIGNAL
ANALYSIS
From
notes by
Stephen
F.
Adam
Roderick Carlson
In the previous chapter, the generation of microwave signals for the purpose of measurement was studied. It follows naturally that the next
step is to study the other measurement instruments and the techniques
used for characterization of signals. Magnitude, frequency, harmonics,
and modulation are some of the characteristics of microwave signals that
one may need to analyze and describe quantitatively.
76-7
Power Measurement
596
Power
Sensors.
basis of heating.
rise in
a thermocouple,
of the device.
The films are controlled in thickness to yield a resistance
termination equal to the characteristic impedance of the transmission line
The sensor output is a dc voltage very nearly
to which it is connected.
proportional to the microwave power dissipated. In the simplest embodiment, the thermocouple is connected to a dc amplifier with the right gain
means
for
597
rf input
the two required contacts to the auxiliary power meter at rf ground, and
the coaxial center conductor is loaded only by the desired sensor elements.
Figure 16-2 is a drawing of a thermoelectric subassembly containing
two sensors bonded to a microcircuit on a sapphire substrate. The dissipation of microwave energy in the resistance of the bismuth and antimony films causes a temperature rise proportional to the power input and
inversely proportional to the effective dissipation constant of the device.
Vd c = K(Ssb
where
SsijiTj
$Bi
Ti
T am
\>
power
to bulk-metal
of the
+36 juV/C
74 fiV/C
Transmission
center
conductor
FIG 16-2
supporting structure.
thermo-
line
allowed
(16-1-1)
& mb)
~ 0.8
power,
bulk thermoelectric coefficient of Sb,
bulk thermoelectric coefficient of Bi,
temperature of Bi-Sb junction
=
=
=
= temperature
is
is
$Sb
in
the microwave
is
also
598
can be seen from Eq. (16-1-1) that the dc voltage output of the
is proportional to the applied microwave power.
The sensitivity
of the sensor depends upon the difference in the thermoelectric coefficients
of the pair of metals forming the hot junction of the sensor.
When the
It
device
Sb
FIG 16-3
microwave
thermo-
couple.
metallic films are evaporated onto a substrate of poor thermal conducmake the sensor sensitive, then the effective dissipation con-
tance to
stant
is
An
recorder or
The dc
DVM.
For one
range,
Currently, the dc amplifier with the lowest drift for use with source
12 is one that uses a mechanical chopper, or modulator.
The input circuit in Fig. 16-4 shows such a chopper feeding a transformer.
impedances of 200
Current from the sensor flows first down through the top primary windand then up through the bottom winding, which causes almost a triangular wave of flux in the core and a square wave of output voltage.
A high step-up ratio is used to keep the noise in the first stage of ac
amplification from being more significant than the noise in the source
resistance, and the transformer is very well shielded.
The FET used in
the first amplifier stage provides high input impedance and is chosen for
low noise.
ing
The shielding requirement is reduced, however, by operating the chopping switch at 110 Hz, which, together with synchronous demodulation,
avoids intermodulation with harmonics of either 50- or 60-Hz line frequencies.
The arrangement
in Fig. 16-4
is
presently capable of a
full-
599
600
scale setting of
for
high-impedance sources.
ambient temperature. Calibration is easy, however, because the thermocouples respond to audio frequencies as well as microwave frequencies.
The power meter can be obtained with a self-contained audio generator
of good amplitude stability.
A "calibrate" switch is used to apply a
known audio power to the two couples in series while microwave input is
zero, without interfering with the performance of the amplifying system.
The gain of some part of the system is adjusted to give the correct reading on the output meter while the calibration signal is applied.
Thermistor Power Meters.
Thin-film techniques had to be developed to
construct the tiny Bi-Sb couples described above.
Previously, thermis-
"H
0.018 in.
K-
*)
0. 014
in.
Side view
FIG 16-5
bead
End view
Microwave bead
for use in a
thermistor.
mount
measurement
for the
it difficult
of
601
Microwave input
JC>
Vv
N.
FIG 16-6
^Thermistor
mount
Two
mount.
temperature, and one such circuit will be described later, but an elemenObserve that the thermistor bridge
tary scheme is shown in Fig. 16-6.
Figure
is excited by an amplifier that is fed by the bridge output El
2 in a qualitative manner.
16-7 shows Ei versus the excitation voltage
16-6 is high enough to
Fig.
in
amplifier
dc
of the
Assume that the gain
fi
FIG 16-7
voltage
in a thermistor
bridge.
versus input
602
drive
to the vicinity of
E = -0(E - E
x
-0
where
E =
2
Then
(Xe.i.2)
is
= -!KB,
0K
ft
when
+ fE.
The use
1.
But
(16.,.3)
An
E
methods have
been devised to measure AE 2
The main trouble with the simple circuit of Fig. 16-6 is that it is sensitive to changes in ambient temperature.
To a good approximation, the
power required to bias the thermistor to the proper operating temperature
.
is
P = Kd (T b
where
T.)
(16-1-4)
T = operating temperature, C
T a = ambient temperature, C
Kd = dissipation constant of the
mW/C
in
practice
The change
of
dPb _
~ ~ K*
dTa
with respect to T*
is
(16-1-5)
mW
kind of
603
Bridge circuits
rf
lTLTL
rf
bridge
f l^comp-^rf
5 kHz
I'rf
multivibrator
fcomp
^rf
\5kHz
Voltoge
to time
converter
^comp+^f
FIG 16-8
\rep.
Switch
(multiplier)
'TT^rv
compensation will be discussed in connection with Fig. 16-8 (HewlettPackard model 432 A).
Two matched sets of thermistor beads are located in the mount or
sensing head, and two complete dc self-balancing bridge circuits similar
to those in Fig. 16-6 are used, though the circuit details are not shown.
Both the sum and difference of the two bridge output voltages are made
by a resistance network.
Both bridges produce self-balancing output voltages proportional to
the square root of the power introduced into their respective thermistors.
The power into one bridge (compensating bridge) is P c or compensating
power, and the power into the other bridge (rf bridge) is P rf plus P M the
power to be measured. Therefore, the difference between the outputs of
the two bridges can be used to measure the true rf power input if the
two bridges are identical, or
available
Pc = Pn + Pm
P M = PC~ Prf = K(Vc ~ Frf
= K(V C + VrdiVc - V
2
lt )
(16-1-6)
2
(16-1-7)
where the subscripts on the voltages correspond to the bridges that produce the voltages and K is a constant.
604
sum
Vc
+ V
is
r(
pulses
Vc - V
t{ .
circuit is
power was
across a
way to
to use a solid-state rectifying diode to
measure microwave
measure the voltage
known
resistance.
At low signal level, diodes are nonlinear
the actual voltage-versus-current curves depending upon the
choice of diode: point-contact, ordinary junction, or hot-carrier junction.
In any case, the proper choice of operating level and load produces a dc
output current proportional to the square of ac input voltage from a lowrectifiers,
mechanical damage.
The barretter, a tiny metal resistive element with a positive temperature coefficient, has been used extensively in the past to measure power.
most common physical form is a short platinum wire with a diameter
and it is used in a power meter with some sort of
resistance bridge.
At least one such power meter is a self-balancing
Its
of only 30 to 60 nin.,
76-2
burn
it
is
necessary
power
level
out.
mW
often achieved
605
(to
be
studied later) at both the input and output ports of the attenuating
device.
The
The coupling
measurements.
factor
For the direct measurement of power levels above 100 mW, calorimmost often used. That is, the power being measured is dissipated in a thermal device, and the resulting temperature rise is observed.
Classical calorimetry consists of observing the temperature rise in a
well-insulated thermal mass of known thermal capacity after power is
absorbed in that mass for a known time. This is more directly a measurement of energy, and a different approach is commonly used in the
eters are
measurement
of electrical power.
Usually, the
This
may
be a temperature
rise of
unknown power
some point
a fluid of
is
made
in a thermal network.
known
physical character-
and flow rate, a fluid carrying the heat away from the dissipative
element.
Or the power may establish a temperature gradient in a physical structure of known, fixed characteristics.
istics
16-3
Pulsed-power Measurements
In the material above, it has been assumed that either the high-frequency power being measured was unmodulated or the average power in
a modulated signal was desired. Pulsed rf power introduces some special
problems in measurement, and the measurement of peak pulse power has
been very important since the early development of pulse radar.
Peak Power Calibrators. Probably the most direct approach is to
attempt to use a power sensor with a response so fast that its output
faithfully follows the pulse modulation.
This approach rules out thermistors, barretters, and thermocouple sensors at present because of their
long thermal time constants. Semiconductor crystal diodes can be made
to follow rf pulses with widths as narrow as 0.25 fisec with repetition frequencies as high as 1.5
characteristic
and
MHz.
If their variation in
curvature of response
much
higher levels.
606
FIG 16-9
crystal detector to
is
That
is,
607
the output
To
a straight line with a slope proportional to the peak pulse power.
reconstruct the pulse from its integrated form requires an active or
is
With adequate
The other way to use slow sensors for peak pulse measurement is to
measure average power in the repetitive pulses and use some auxiliary
method to measure the duty cycle of the pulses, which is the product of
pulse width and repetition frequency.
Attenuating directional couplers
The peak
are used to reduce the measured power to a convenient level.
power out of the coupler into the instrument setup is
rP _
average power
,,.,
^io-o-i;
duty cycle
is
most
easily
it
pulse.
The detector output is fed to a calibrated oscilloscope, where
width and repetition frequency are observed.
Detailed information on the instruments and systems described tutorially here can be obtained from instrument manufacturers prominent in
the microwave field.
76-4
Mismatch Considerations
equation
_ q -
|r c |*)(i
|i
where To =
Tl =
rGrL
|r,n
2
|
608
The
polar form
is
If
is
achieved.
Zo Match.
is
ThusEq.
(16-4-1)
becomes P D =
(1
\T a
2
\
=P
)PA
to a nonreflecting load.
If neither of the foregoing conditions is satisfied and
T Q ? 0, then Eq. (16-4-1) is not readily solved, because
The T L
usually arg r G is not known and this may also be true of |r G
can be measured with a network analyzer. When arg TgI^ is not known,
the limits of the quantity |1 ToT L 2 should be noted:
Mismatch
Errors.
in addition
|.
(i
If \T G
<
|i
0.2 (swr
1.5),
limit
on
+ 5.3
\t l \)*
|r G
<
t g t l \*
\T L
|1
(i
|r G
|ri|)
(16-4-2)
0.13 (swr
r G r.z,| 2
is (1
percent error.
76-5
if
TG =
It is
and
will
Application Considerations
Let us now consider the problem of measuring power under the two
most frequent matching situations. A system would be designed for
conjugate match when it is desired to get maximum power delivered to
a load from a source of limited output. A good example would be maximizing the power delivered from a radar transmitter to the antenna.
It is likely that both the transmitter and antenna have tuning elements
to match them to the line so that excessive standing-wave ratios will not
exist on the transmission line.
Most other equipment is designed to be matched to the characteristic
impedance of the line. This facilitates standardization, measuring techFor instance, it is very desirable to design a
niques, and compatibility.
609
The
meter.
ve
There
fore the
effective efficiency
is
P<?
(16-5-1)
will be some error in the electronics of the power meter; therepower indicated Pid will not be a perfect representation of Pd
(1
When making
source,
it is
e.
e)P s
Therefore,
(16-5-2)
a measurement of the
maximum power
mount and
source
available
from a
by using a
tuner.
KT ~
0.97,
which
is difficult
to
In a good
measure and varies with probe
penetration.
(1
is
e)KTV eP A
measured
610
jugate match,
= KtvPa
"load
(16-5-4)
Krii]e(l
-qe)
p
Pind
(1
[r G
e)r,e
|1
However, we want
Also
P ind
in
)(l
1 <?i
terms of
|IV! 2 )
....
P A
(16-5-5)
M\
P =
(1
|r G
2
|
)PA.
is
the mount.
It differs
from
effective efficiency
by the mismatch
loss at
(1
)ve
(16-5-6)
- e)KJ>.
- T.TM'
\r
Rewritten,
PM
(1
|1
(l
~1FT\
b {l
jr |')|i
\il
e)\l
TgT m 2
Pind
2
^
t^ 12
TqT
l
is
con-
(16-5-8)
Mounts.
specified separately
NBS.
The uncertainty
of
NBS
611
from 0.5 to
a wavecompared with an NBS-tracecalibration
is
it is
efficiency
at 0.1
is
a function of frequency.
coaxial
mount
is
typically 0.99
To
which
known, it
measure the
reflection coefficient |r m accurately at that frequency, and determine the
calibration factor from Eq. (16-5-6).
Dual-element error is found in mounts that have two sensor elements in
parallel to microwave power, but are in series for the substitution power.
Errors of a few percent have been observed in thermistor mounts. However, the technique of thermal matching used by some manufacturers
in temperature-compensated mounts limits the error to about 0.1 percent
in these mounts.
The dual-element error is significant only on the highest
range or two of a power-measuring system.
factor at
is
at
it is
effective efficiency,
16-6
The measurement
612
can be easily made to, say, 1 part in 10 10 but there is a large market for
simple inexpensive wavemeters that operate on the basis of electrical
resonance in mechanical structures. These structures are tunable.
While these tunable resonant wavemeters (or frequency meters) are
,
basically simple, they are fabricated with a precision that keeps errors as
is
made
by the
its
power,
resonance.
it
of
As
of the structure.
mission line
Zi
where
j8
(16-6-1)
jZo tan 0l
Z =
I
is
characteristic line
impedance
= length of line
= phase constant = 2w/\
= wavelength of signal
an approximation for lines that are long in comThe impedance Z approaches zero
for even numbers of quarter wavelengths, and it approaches > for odd
numbers. Half-wave shorted lines are frequently used, and Fig. 16-10
shows a coaxial example with a short circuit adjustable by a plunger.
The resonant frequency of such a half-wave cavity is
is
(ie-6-2)
/;//;;////
\<
FIG 16-10
cavity.
Coaxial half-wavelength
/////;;;/;
/ y
613
/-vv
I
1YV ;////;; ;/;//
'
where
n = an integer
X e = cutoff wavelength of the line
which approaches for the principal transverse electromagnetic mode
(TEM).
If one end of the cavity is left open as in Fig. 16-11, a quarter-wave
cavity results, one that is resonant at odd multiples of quarter-waveIn this construction, it seems more convenient to make the
lengths.
center conductor of the coaxial system movable than to slide the shorting
on a quarter-wavelength cavity at
= j(2n-l)
where X
is
\ \ \ v
(16-6-3)
\^^^-
a 5S
^
:\
^ \ \ \ \ \ m
The bandwidth free of spurious resonances can be increased in a quarterwavelength cavity by loading the open end with a lumped capacitance.
Physically this is accomplished somewhat as shown in Fig. 16-12. When
the center conductor, which acts as the tuning plunger, is far from the
61 4
=
=
I =
n =
A =
B =
where /
d
The
+ Bn
frequency,
(-)
(16-6-4)
MHz
cylinder diameter,
cylinder length,
in.
in.
index of
TEm
depends
615
Power
Frequency
filter
specifications.
Type
of cavity frequency
meter
TEin
TEon
Bandwidth
2:1
3:1
Accuracy
2,000-4,000
2,000-4,000
4-10:1
1.5:1
1.1:1
500-4,000
8,000-12,000
50,000-100,000
Input
Output
FIG 16-14
coaxial arrangement for magnetic field coupling
for reaction or absorption.
FIG 16-15
coupling
in a
Transmission type of
coaxial structure.
0.1
0.1
0.1
0.02-0.05
0.005
616
76-7
Spectrum Analysis
of observing a signal
it on an
Chap. 11. This is
a view of the signal in the time domain. It is also very useful to display
signals in the frequency domain.
This measurement method, often providing unique information unavailable, or practically unavailable, in the
time-domain view, is called spectrum analysis and is the subject of this
section.
The instrument providing this frequency-domain view is the
spectrum analyzer. On its CRT, the spectrum analyzer provides a
is
to display
Amplitude -frequency
(O
(b)
Aft)
-*
(f)
FIG 16-16
The time and frequency domains: (a) Three-dimenshowing time, frequency, and amplitude. The
addition of a fundamental and its second harmonic is shown as an
sional coordinates
617
(a)
(b)
FIG 16-17
incoherent signals.
Displayed as vertical
and
composed.
From
is
its
horizontal location,
Figure 16-16 shows how a simple signal A(t) looks as viewed in the
time domain on an oscilloscope (6) and in the frequency domain A(J)
618
each other. Each has its own area of almost exclusive application, and
is a gray area where both types of measurement yield important
information.
The digital computer operates almost exclusively in the
time domain. Also in the time domain is the time-multiplexed pulsecode-modulation telephone communication system, where each telephone
channel is sampled sequentially and each sample level digitized into a
digital word.
The transmitted signal is then a series of digital words
each representing the level of a particular channel sampled at a particular
there
time.
Figure 16-186 shows how the spectrum analyzer can measure extremely
small amounts of distortion, far beyond the capability of an oscilloscope,
by measuring very small harmonic components, generated from a sinusoidal test signal by the device under test.
However, Fig. 16-18a shows
how at higher levels the oscilloscope can show where in the test-signal
is occurring, which can lead to uncovering its cause.
Spectrum-analyzer applications include the measurement of signal level
and the measurement of frequency and its response, harmonic and intermodulation distortion, frequency stability, spectral purity, modulation
index,
and attenuation.
Scanning and Real-time Spectrum Analyzers. There are two types of spectrum analyzers scanning types, which scan in frequency, andnonscanning
:
Microwave Signal
A na
61
(h)
FIG 16-18
in 6 severely distorted
Overdriving an
waveform eaiily observed, with the oscilloscope,- however, quantitaThe
tive measurements of distortion levels are difficult to obtain.
scope calibration is 0.2
per division vertically, 0.05 jjiec per
When the input power level if reduced (>),
division horizon tally.
amplifier results
the distortion
If
hardly observable.
Trie
two signals.
The frequency scale is 10 MH* per division centered at 50 MHz,
and the reference level is -J- SO dBm. (The response at the far left is
The 10-MHz signal input Is at 30
the lero frequency indicator.)
dBm in A and at 40 dBm in B. Since the normal amplifier gain is
40 dB, gain compression is about 7 dB in A. Second harmonic distortion is reduced from 14 dB down to 38 dB down by the 1 0-dB
reduction in signal input. The effect of input signal level on the
gives quantitative information about the distortion of the
other harmonics
is
620
16-8
identification.
division,
1
is
CL
Q
621
622
Mixer
Input
to
1GHz
/v \
2GHz
.ow-pass
__fc
filter
^\cy
if.
V
1
Detector
ampl.
2 GHz
GHz
Sweep
>
Voltage
generator
1X1/
tuned
Motor
2-3 GHz
local osci
FIG 16-20
to
GHz.
frequencies that are the sum and difference between the frequencies of the
input signal and the local oscillator signal.
The intermediate-frequency amplifier is tuned to a narrow band around
GHz.
2
As the local oscillator is tuned over the range from 2 to 3 GHz,
only input signals that are separated from the local-oscillator frequency
FIG 16-21
tuning chart.
Superheterodyne
623
by
From
sawtooth signal sweeps the local oscillator linearly from 2 to 3 GHz, the
tuning of the spectrum analyzer as a receiver is swept linearly from to 1
GHz. The sawtooth scanning signal is also applied horizontally to the
CRT to form the frequency axis of the display, the left edge representing
Any horizontal driving wavezero frequency and the right edge 1 GHz.
form could be used for the scanning function, but a sawtooth provides a
linear time axis useful in certain applications.
Notice in Fig. 16-21 that
the spectrum analyzer will also be sensitive to signals from 4 to 5 GHz,
referred to as the image frequency of the superheterodyne.
A low-pass
filter with a cutoff a little above 1 GHz at the input suppresses these
spurious signals.
Figure 16-22 is a more detailed block diagram of the spectrum analyzer
The frequency of the first local oscillator is controlled
just described.
the
8^
CX
/^
a
g8
a.
0-3.-
.o
a.
IQ.Q-
fc
a.
^
ex
O
"a
'
ii
c
624
ex,
T?5>
FIG 16-23
teristic
for
Selectivity
i
625
charac-
3-db bandwidth of
10 kHz.
20 kHz
/division
window
separately.
Two continuous-wave signals with a separation
than the intermediate-frequency bandwidth would both be in the
passband at the same time and could not be distinguished. Multiple
synchronously tuned intermediate-frequency filters are used. These
approach a gaussian response, which will not ring and produce outputs
that could be misinterpreted when a signal is swept rapidly through them
or in an impulse-input situation.
the
of less
filter.
10
dB
(b)
FIG 16-24
The dip
60-dB
in
(a)
the center
Two
difference in
is
626
Residual
any
and
and phase noise will add noise skirts to the filter skirts, both reducing
shown in Fig. 16-25.
Sweep desensitization is an effect, caused by
Sweep Desensitization.
scanning a spectrum analyzer too fast, which results in loss of amplitude
It is easily detected and corcalibration, sensitivity, and resolution.
rected if understood. During scan the signal must remain in the bandpass of the intermediate-frequency filter long enough to allow the amplitude of the signal in the filter to build up to the proper value. A simple
rule of thumb for avoiding sweep desensitization is that the scan velocity
in hertz per second must not exceed the square of the 3-dB bandwidth of
resolution as
the intermediate-frequency
filter
in hertz
FIG 16-25
Effects
of (a) residual
FM
time.
627
=,1000
FIG 16-26
versus normalized
B =
TS B 2
sweep
loss
rate,-
^ =
bandwidth.
Ph = kTB
where
B =
=
T =
k
The
(16-8-1)
bandwidth of system
Boltzmann's constant, 1.38
absolute temperature,
10- 23 W-sec/K
the passband of
its
628
-150
-140
-100
|0
Hz
FIG 16-27
1kHz
100 Hz
10 KHz
Bandwidth
100 kHz
10
MHz
can be calculated.
is
124 dBm +
25
dB = 99 dBm.
above the noise "grass" level on the display, a continuous-wave signal should be 10 dB above the spectrum analyzer's own noise
To be
easily seen
level, or
-99 dB
10
dB = -89 dBm,
or 8
mV
across 50.
Figure
iV A
+G
is
N
NA
NB
where
decibels, that
is,
gain of 10)
100, not 20
dB
629
Transistor, tunnel-diode,
is
ia
Dynamic Range.
linear
a selection of
rithmic
modes
The
of amplitude display.
frequently used.
fotja-
its
the most
is
final
300
of the
itB
ability to
itself
.small
of
Wilh
he instrument, spurious
.sig-
by the analyzer
and could either mask the small signals or erroneously appear J
The dynamic range h:it is
signals, will not appear on the display.
which
nals,
dynamic range
result
free of spurious signals can be defined as the ratio of the signal level to
tin?
just
630
Mixer
Input
2 GHz
low-poss
0-2GHz\
filter
GHz
Detector
if
YIG
band pass
filter
lijiiyi
^r
Voltage -tuned
,
FIG 16-29
Harmonic mixing
analyzer from
local oscillator
2 to 4 GHz
to
in
GHz.
Sweep
generator
^1 ^
\^^
1
begin to become visible above the noise level of the display. Present
spectrum analyzers have 60- to 70-d.B dynamic ranges, free of spurious
signals, for continuous-wave signals except at very high frequency.
^^Dnve
/
/
from
\
j_~~
/
<
/
/
Alt)
signal
^^ur
r~
\
/
Bios
\_
/
\
Time
FIG 16-30
A(t)
(3w
A(t)
dj 3 )
= A
Ai
sin (o>
Ai
sin (2a>
<t>i)
+A
sin
631
and on by the first local-oscillator drive signal. This transfer funcis shown expressed as a Fourier series function of frequency.
Bias is applied to the drive waveform to unbalance the duty cycle so that
even as well as odd harmonics are present. An input signal is multiplied
by this transfer function, which produces sum and difference frequency
output signals with each term of the Fourier series. In other words, the
off
tion of time
etc.
Bandpass
filters can be used to eliminate the confusion.
broadband filters can help with higher-order harmonic mixing modes, but
the most effective solution is a tracking narrow-band filter which can be
adjusted to track a desired harmonic mixing mode as the spectrum anaSuch a filter is the YIG filter, consisting of
lyzer is tuned and scanned.
one or more coupled yittrium-iron-garnet resonators whose resonant frequency is proportional to the strength of the field from an electromagnet
This filter can be biased and electrically swept
in which they are placed.
to allow only the signal matching a desired mixing mode to enter the
input mixer. This is, of course, like the tuned front end on an ordinary
but preselection
broadcast receiver.
sion
local oscillator.
3
Local oscillator frequency, GHz
632
16-9
Wave
Analyzers
wave analyzer
It is a
FIG 16-32
Selectivity
characof the spectrum analyzer
and wave analyzer compared.
teristics
633
Broadband
detector
YIG
V_/
input
+H
filter
FIG 16-33
yig-tuned
Low
noise
Log/
pre-
in
amplifier
amplifier
Scan
Scan
controls
generator
trf
spectrum analyzer.
The advantages over the direct use of a counter are that the frequency of
very small signals can be measured and a signal can be selected out of
many even larger signals that might be present. Wave analyzers usually
have automatic frequency control, in which the tuning automatically
locks to a signal
and so makes
it
16-10
The heart
CRT
vertical
and poor
Its
sensitivity
disadvantages
compared with
the microwave
filter,
so
it
compared with
tion
is
100 dBm
Present applica-
634
16-11
all
its
Bandpass
filter
45-55 Hz
Bandpass
filter
56-70
Hz
Rms
detector
Rms
detector
Bandpass
Input
O
50 Hz
filter
Rms
71-89 Hz
detector
Electronic
switch
Bandpass
Preamplifier
filter
-@h
output
scanner
90-II2 Hz
to
10 kHz
etc
(24 channels)
Scan
generator
Bandpass
filter
9.0-
FIG 16-34
1.
kHz
Rms
detector
FIG 16-35
0.1
0.3
0.5 0.7
Multifilter
635
real-time
spectrum-analyzer display.
these
awkward.
analyzer techniques
is
flicker will
If
16-12
r^?^
^^Z>
!^8
<>
c
CQ
j_
o>
a
CO
<2>
<Z>
"
O
636
637
Spectrum analyzer
Spectrum analyzer
@@
o@o
o@o
r
Local oscillator
outputs scan
output
Device
under
test
Local oscillator
outputs scan
output
__,
O O
O O
With test
device
(b)
(a)
FIG
Intense marker
crystal-controlled oscillator signal at the same frequency as the spectrum-analyzer final crystal filter is mixed with local oscillators as shown,
which ultimately produces a signal that will track the spectrum-analyzer
passband even on its narrowest bandwidth.
Figure 16-37 shows the frequency-response measurement setup. The
tracking generator signal is connected to the input of the test network and
The amplitude frethe spectrum analyzer connected to the output.
quency response of the network is plotted on the spectrum-analyzer CRT.
The frequency response can be determined over very wide range in
This is due to the high sensitivity of the
amplitude, over 100 dB.
analyzer and the advantage that this tracking, tuned system is not
sensitive to harmonics of the test signal that
16-13
From
may
be present or generated.
ponents of a
of using
of the
638
(b)
(a)
(c)
(f)
(e)
FIG 16-38
Accurate determination of
signal
frequency on a
spectrum analyzer.
639
read as 1,847.35 MHz (Fig. 16-38/). To obtain these displays, the output
of the comb generator is superposed on the signal through a coaxial tee.
Modulation Measurement. With its frequency scan set to zero and the
representing time rather than frequency, the spectrum analyzer
axis
x
operates as a fixed tuned receiver to measure amplitude versus time.
This has been called the synchroscope mode. When the analyzer is
tuned to the carrier frequency with bandwidth at least twice that of
the modulation frequency and with a linear display, the envelope of an
a,
AM
/'
/' n\
1
/
/
\\
v/
/
\\
vV- >
"
1
AM
VP +VT
640
T
Vr
Distortion
Vs
*-2fm
AM
Simple
AM with distortion
producing 2nd hormonic
modulation sideband at 2fm
(a)
(b)
FIG 16-40
Spectrum of
modulation frequency, and
as small as
m =
a
0.0006.
C.
As seen
in Fig. 16-40,
it is
also easy to
measure
knowledge
culate the
FM
FM
However, when Ad
FM
AM
The amplitudes
AM
or pure
FM
of
exists.
Therefore,
FM
when
first
AM
exercised.
FM.
Modulation
Ad =
(16-13-1)
Jm
deviation.
as
WW
Jl
WMOCT
(a)
Frequency-modulation
FIG 16-41
FM signal
at
MHz.
displays
on
spectrum
FM,
This
is
641
642
FM
approximately
Usually the fre-
is
lation frequencies, subaudio, but with large relative deviation, the spectral display will be the signal sweeping slowly back and forth across the
display by an
amount exactly equal to twice the deviation. As the moduis now increased, this same display holds approximately
lation frequency
When
643
frequency deviation.
High-frequency wideband noise-modulation components produce what
sidebands
is sometimes called phase noise in a signal and result in noise
in Figs.
seen
carrier,
as
the
of
side
each
on
distributed
continuously
Since these are distributed noise signals, they cannot be
16-25 and 16-43.
Their voltage amplitude, but not that
resolved as individual sidebands.
used.
The amplitude
must be
of these sidebands
given.
since distortion
is
treated in so
many
texts.
The
measurement by means
of
an analyzer
is fairly
straightforward.
The
noise
in Fig. 16-44,
selectivity characteristic.
is
is
mode
of vertical display
ordinate.
Video filtering is
spectrum analyzer.
Video
superheterodyne
filtering of a
axis,
directly
if
calibrated,
or
by continuous-wave
noise
bandwidth
to
644
^f
^r
tlH
"^"^"^
t~
wen
The
645
f,
NB =
A
area of
height,
77
vs.
MHz
X SW)
f
selectivity characteristic in
cm 2
cm
FIG 16-44
Noise bandwidth.
mW/MHz
over a range of 100 MHz is a total power of 100 mW, which could overload the instrument.
The widest bandwidth consistent with resolution
The duration of these pulses is usually short compared with the rise
time of the systems with which, as noise, they interfere. Also, the pulses
are spaced by time intervals that are long compared with the decay times
The response in these systems is thus an impulse
of these systems.
646
FIG 16-45
(a)
PRF
Time-domain rep-
(a)
Pulse
/Ts
III,
,|
1/7-
PRF
3(PRP)
(t>)
i/r.
2(PRF)
It is characterized
spectral intensity.
by
its
The voltage
is
while with white noise the voltage increases as the square root of the
IB =
where SI
(16-13-2)
01
spectral intensity
Vr = response
IB =
A =
VA =
Measurement
is
of impulse noise
is
similar to
bandwidth IB
is
measurement
is
desired.
of white
filter is
not
was shown
for white-noise
It
can be
bandwidth
in Fig.
way is to apply
known
647
better
spectral intensity,
in
measuring white noise, since peak voltages can be especially high and
overload the analyzer input even though the distributed spectral intensity
might be low. The widest bandwidth consistent with spectral-intensity
envelope resolution requirements should be used. Overload is checked as
always by seeing that changes in the display follow changes in input
attenuation.
AM
SI--/ZA
(a)
SI-
IB
^L
(PRF)
mi T
-(PH?)
(b)
FIG 16-46
648
~+\Tr
N
Z^f\
2 TIT-??
Ep=Ep Tw
[PRF)
IrrnumLi
^rms/HID:
w
--
(b)
FIG
6-47
AF
*-Af-
the spectrum,
its
envelope,
is
given by
Sin (2tt/7V2)
2wfTw /2
2irfTr /2
with nulls spaced at each side of the carrier frequency by multiples of the
reciprocal of the pulse midwidth.
It is informative to examine, one by one, the effects that changes in the
time-domain parameters of the signal have on the frequency-domain
spectrum. If the carrier frequency is changed, the whole spectrum is
completely unchanged in amplitude and frequency dimensions, but Shifts
649
The responses that a modern swept superheterodyne spectrum analyzer can have to this periodically pulsed rf signal can be of two kinds
and give two different but similar displays. One response is called a line
because it is
described in
responses to
terminology
a periodically pulsed
must stay below the pulse repetition frequency to stay in the line
spectrum mode.) The displayed height may change if the analyzer gain
changes with B, but the measured signal amplitude will not.
A pulse spectrum occurs when B of the spectrum analyzer is greater
than the pulse repetition frequency. The spectrum analyzer in this case
cannot resolve actual individual Fourier frequency-domain components
since several lines occur within its bandwidth, as with impulse noise.
However, if B is narrow compared with the spectrum envelope, then the
envelope can be resolved. The display is not a true frequency-domain
display, but a combination of time and frequency display.
It is a
time-domain display for the display pulse lines since each pulse line
occurs as each rf pulse occurs.
The display lines occur at the actual
pulse repetition frequency.
It is a frequency-domain display of the
spectrum envelope. The display has three distinguishing characteristics:
1. The spacing between the pulse lines on the display increases linearly
650
The
3.
PRF
is
dB
by continuous-wave
per octave as
persists until
substitution,
B is increased.
This
width of the main lobe of the spectrum envelope, and then no further
increase occurs.
At this point, B has collected almost all the spectral
components. The display is now in the time domain. If the frequency
scan were stopped by setting the spectrum width control to zero and one
could sweep the display fast enough, one could begin to distinguish the
time-domain shape of the pulse envelope. If B were increased further
and became much wider than the main-lobe width, the detailed shape, in
time domain, of the pulse envelope could be observed.
In the pulse spectrum just described, the response of the spectrum
analyzer to each rf input pulse is the impulse response of the analyzer
The height of these impulse responses
intermediate-frequency amplifier.
traces out the shape of the input spectrum as the front-end tuning is
swept across the input spectrum. This impulse type of response is the
explanation for the characteristics of this display.
Why use a pulse spectrum response? The spectrum envelope, its
shape, amplitude, and spectral extent, is usually of interest instead of
individual spectral
component
lines.
The use
of the pulse
spectrum
dis-
play does. The display amplitude at the center of the spectrum envelope
with a line spectrum is, from Fig. 16-47, Tw (prf)E p and with a pulse spectrum it is the product of the spectral intensity at the center of the spectrum and the impulse bandwidth IB of the analyzer, or IB X Tw X E p
,
making
Pulse spectral response
;
IB
-
^lO-JLo-Oj
prf
The
own
B, so the
Operating with B
greater than the pulse repetition frequency increases sensitivity to pulsed
The input level should be
rf signals and increases the dynamic range.
analyzer's
651
held to a value such that the peak pulse signal barely does not overload
the input mixer.
Some
16-14
Sensitivity
Rules of
Control Settings
versus
Envelope Resolution.
The
should be as wide as
possible for the highest sensitivity but not wider than 0.1 /Tw
spectrum envelope
the pulse
be seriously
impaired. Another way of saying this is that B should be less than
5 percent of the width of the spectrum-envelope main lobe.
If higher
resolution is required to look at faster falling spectrum envelope skirts
or to look more deeply into nulls, narrower B is required.
To resolve
20 to 30 dB into nulls, a good rule is
of the
will
<
0.03
(16-14-1)
Avoidance
of Base-line Lifting.
when
stant
five
B >
1.7 prf
(16-14-2)
dB below
the spectrum
envelope.
FIG
6-48
Base-line
lifting.
if
something
is
cutting on and
off.
652
The
1/prf.
rule
df
j<
0.2prf
df
0.2prf
is
AF
(16-14-3)
df
where
AF =
= 2AF
Tw = rf pulse width, sec
This rule assures that the peak will be displayed on each sweep and
avoids having to watch several sweeps to catch the peak value of the
displayed spectrum.
Loss in Sensitivity Due to Sweeping Too Fast.
If one sweeps a spectrum
analyzer past a continuous-wave signal too fast, a loss in response occurs.
The rule for avoiding this sweep desensitization, already explained, is
f < B.
(16-14-5)
Loss in sensitivity can also occur from sweeping too fast through the
The rule for less than
dB
of loss in sensitivity
is
that
2.5
df
(16 " 14 - 6)
* < T?
Frequency Domain. Most of the preceding has dealt with the pulsespectrum mode of response. The area between it and the line-spectrum
mode of response is a gray area that does not have quite the characteristics of either type of display.
In this area, interpretation of the display
is difficult
The
and
is
best avoided.
B^ <
is
that
P rf
(16-14-7)
pulses,
The
there
less
is
mode
is
that
B >
prf so
,
653
16-15
measurements
K.
signal delivered
TABLE
16-1
V/m
by antenna to instrument,
Antennas Specified
in
MIL-STD-461
rf
current transformer.
for
Frequency
14kHz-25MHz
Description
4i-in. rod (electrical length \i
m)
Biconical antenna
CHAPTER SEVENTEEN
MICROWAVE NETWORK
ANALYSIS
From
notes by
Adam
Stephen
F.
Douglas
K. Rytting
William Heinz
Hewlett-Packard Company
Palo Alto, California
by impedances and
and networks are characterized at
lower frequencies. However, both the methods of measurement and the
parameters measured are profoundly affected by the fact that a wavelength at microwave frequency is short in comparison with the dimensions
of the system.
It becomes imperative to consider the existence of waves
traveling at finite velocities and being partially transmitted, partially
reflected, at impedance discontinuities in a system.
Microwave impedance and transmission measurements are also conditioned by the universal practice of transmitting microwave signals
are characterized
through virtually
654
lossless,
uniform transmission
lines
having
real, positive
655
impedances Z or R
For years, instruments have been
designed to measure the reflection coefficient T at a reflection plane or discontinuity in such a transmission system, where T = E r /Ei. The incident
characteristic
many
years.
measurement and
design.
77-7
Reflection
viewed.
The measurement
ratio,
that transmission
easiest to
vswr.
One way
line.
Coaxial slotted lines, using true coaxial cross section, are severely
limited in their performance because of the finite width of the slot to
enable the probe to penetrate into the line. This causes slight deviations
of characteristic impedance and excess leakage through the slot.
Wholey
and Eldred [2] suggested a slab line arrangement, now widely used, a
cross section of which is shown in Fig. 17-1.
With proper design,
be maintained even when a sliding probe is used.
can
656
Cioh
FIG 17-1
FIG 17-2
lection of a
Slotted-Iine
Packard
Company 447A.
Modem
as
shown
in Fig. 17-2.
Oscilloscope with
log display
Probe/
detector
ALC
Swept signal
FIG 17-3
Slotted
Sweep
adopter
source
ilotted-line technique.
line
SWR
with swept-frequency
Unknown
load
657
minimum
A P = A + B P + C P
where
A =
B =
slope
and loss
[5]
(17-1-1)
of slotted line
and residual
C =
probe reflection
discontinuity in a line
Ratiometer
Detectors
[jri^r
>
Signal
source
\y
couplers
FIG
7-4
Reflectometer arrangement.
Device
under
test
Remember
658
x-y
recorder
m
Calibration
attenuator
Swept signol
Short or
open circuit
source
FIG
7-5
Calibrated reflectometer.
1+P
1 ~P
o-
<7
a,
and
reflection coefficient p
is
(17-1-2)
With the advent of leveled signal sources the use of ratiometers has
been largely eliminated. Leveled reflectometers do not need to make the
Tuned
x -x
recorder
indicator
Calibration
ALC
Swept signal
source
attenuator
|c<
Short
Measurement port
Device
tested
FIG
7-6
Leveled reflectometer.
7-1
Error factor
Measurements
No
Calibration grid
Directivity ; attenuator
659
calibration grid
Directivity
accuracy
Directivity attenuator
;
accuracy
Directivity; attenuator
accuracy
Coupler /detector
tracking; detector
square-law error
Reflectometer source
Reflectometer source
match
match
AP = A
Bp
Cp 2
(17-1-3)
17-2
Attenuation Measurements
Measurement
classes: rf substitution,
Signal
source
~"
r~
*|
Standard rf
*
attenuator
Attenuator
undf;r test
FIG
7-7
rf
substitution.
Indicator
known microwave
660
Signal
source
Standard
af or dc
Detector
v---.
Indicator
attenuator
L
1
FIG 17-8
Attenuator
under test
J
1
Audio frequency
or dc substitution.
Audio-frequency or dc substitution uses low-frequency or dc attenucompare microwave attenuation (Fig. 17-8). This technique
Of course,
takes advantage of more accurately calibrated attenuators.
in this method the error in the detector law does come into consideration,
and great care has to be taken to keep detectors in their square-law region,
while in the rf substitution method the detector always operates at the
same level, which eliminates the need to worry about detector law. The
combination of the two techniques is often advantageous.
ators to
40
to
60 dBm.
dynamic range,
their
100 dBm
Errors in
as scalar, vectorial, or
ified
random
(uncertainty).
tors.
If their
values
known
most
FIG
17-9
single crystal
Possible
domain
configuration showing
in
of the
such measurements.
Uncertainty terms or random errors are those that
cannot be predetermined. Such errors are caused
by variations connector repeatability, instrument
Many repeated meareadability, resettability, etc.
661
urements and the averaging of the results can cut down the size of these
errors.
There are statistical methods taking these into account.
77-3
Ferrite Devices
Directional couplers, discussed above, yield output signals with amplitudes depending upon the direction of propagation in the main channels,
but these couplers are passive, static devices in which reciprocity holds.
and phase
oxides of iron,
to achieve any
from oxides of
They
electrical resistance.
just as of
property of the spinning electron. The relationships between the individual atoms determine the macroscopic behavior of the material.
In ferromagnetic materials, the individual magnetic moments align
themselves
in
parallel
within
the
crystal
(see
662
FIG 17-10
Saturation
magne-
*7"
the
sample.
17-96).
moments
aligned
net magnetization
oriented
when no
applied, the
is
produced
if the oppositely
Just as in the ferromagnetic
external field
domains
applied.
is
As an increasing
field is
When
Ferrimagnetic Resonance.
ferrite,
moments
the individual
a dc magnetic
field is
applied to a
In doing
however, they precess about the field as a spinning top precesses in the
earth's gravitational field when disturbed (see Fig. 17-11).
The freso,
quency
of precession
o>
is
magnitude
Path of moment
vector after
disturbing force
is
removed
Angle of
precession
of the
the ferrite, Hi or
= yH
coo
663
(17-3-D
and to the
lattice
absorption occurs
when
frequency of
absorbed from an rf field circularly polarized in the opposite sense.
Since the rf permeability is a function of the direction of the rf magnetic
If
field relative to the applied dc field, it is in general a tensor quantity.
we restrict our attention to the case of CP rf magnetic fields in a plane
perpendicular to the dc
field,
= n h
where m = M
(17-3-2)
The
ju- versus
664
lossy
AH
in absorption
ferrites
having applied
fields
below
wlf
= y[H a
+ 4*M
when
the frequency of
(17-3-3)
t]
An
isolator
is
power passing
It
Reflections from
665
Power
tion.
is
absorbed by the
as the precession,
"dc
i
;
"dc
I
"EJ-
Ferrite
L!
E5i
(a)
(b)
H<ic
Dielectric
vtsm
(c)
FIG
7-1 3
Resonance
isolator configurations.
666
Ferrite
FIG 17-14
displacement
showing
electric field patterns for forward
and reverse waves.
isolator
Field
configuration
Forward wave
Reverse wave
TEM
of
CP
[13, 14].
of this
applied
(b)
n
>
n-port
-2->
circulation;
circulator
shown above.
shown
in Fig. 17-16.
Ferrite
667
Ground
substrate
plane
(cl Microstrip
H-pIane wave-
(a)
involves splitting of
This
is
resonator
is
shown.
The
field,
field configuration
mode in
a cylindrical
is
668
field,
represented
by
lines, is entirely in
rotated 30 as shown in
case,
Power
into port 2
is
pattern,
(b)
diverted to port
Signal
669
Load
(a)
(O
FIG 17-18
is
phase
shift of a
wave
The gyrator
circuit
symbol
is
wave traveling
shown in Fig.
670
IT
FIG
7-1
symbol.
(a)
A gyrator circuit
(6) Four-port
circulator
\)
and
gyrator.
is
17-19a,
realized
of
[10, 21],
but these
will
not be
discussed here.
.If the magnetic field applied to the waveguide device in Fig. 17-13a is
below the resonance value, the different dispersive (real) parts of n+ and
ju_ will yield two different propagation constants. fi + and /3_ for the two
directions of propagation.
Thus the differential phase shift A</> = (/3+
/3_)/, where I is the active length of the device.
The device shown in Fig. 17-20 is a digital phase shifter capable of
being switched rapidly between two states. When a single current pulse
is applied to the wire, the ferrite is magnetized in one direction and
remains in its remanent state. Since the magnetization will be as shown
by the solid arrow, the conditions for differential phase shift are satisfied
It is
[23].
tion
is
FIG 17-20
digital
phase
Waveguide
shifter.
latching
FIG 17-21
Reciprocal
671
phase
shifter.
77-4
microwave networks.
the network.
Although a network may have any number of ports, network paramecan be explained most easily by considering a network with only two
ports, an input port and an output port, like the network shown in Fig.
17-22.
To define the performance of such a network, any of several
parameter sets can be used, each of which has certain advantages.
Each parameter set contains four variables associated with the termiters
I\
Two
-I
-<
"
f
Port
FIG
V\
7-22
Two -port
network
t
Pc
Two-port network.
672
network
of Fig. 17-22
nal currents 7i
=
=
I*
VxxVr
2/iVi
where Vi and
and 7 2
+
+
V2
driven by voltage sources Vi and 7 2 the termibe determined by the following equations:
is
will
y 12 V 2
(17-4-1)
*/22F 2
(17-4-2)
2/22
circuited.
shown
1/21
is
short
ratio of
in Eq. (17-4-3),
Vi
Ii
V =
2
(output short-circuited)
(17-4-3)
be defined.
The s parameters are being used in microwave design because they are
easier to measure and apply at high frequencies than other parameters.
They are conceptually simple, analytically convenient, and capable of
providing insight into a measurement or design problem.
Scattering parameters are well suited for describing transistors and
Measuring most other parameters calls for the
other active devices.
input and output of the device to be successively opened and shortcircuited.
This is difficult to do at rf frequencies where lead inductance
may
673
77-5
Theory of
Parameters
now be
discussed.
The
next section.
He
Kurokawa
defines
the network:
[6*]
(17-5-1)
[*][a k ]
where
ak
Vk
bk=
Vk =
=
Zk =
Zklk
..,_
_,.
(17-5-2)
V|ReZ*|
Vk
2
and
+
.
;JlIl
V\ReZ
k
(17-5-3)
one
as the transformation
674
Pk
Vk =
{Z*
k ak
V\ReZ k
/"
V |Re Z
(ak
k
+ Zkb
(17-5-4)
k)
(17-5-5)
bk)
where
when Re Z k >
when Re Z k <
Pk
Thus any
result in
The voltage
Vk
I k and
,
Zk
are
given by
Vk = V - Zk Ik
If
we
of the
magnitude, we have
a
which
=
is
|2
(17-5-6)
internal impedance.
Re Z k >
and negative
Next, consider
\a k
if
Re Z k <
\b k
0.
2
.
\
Re (V k I*) = P k (\ak
(17-5-7)
2
)
Ik
-
IT
-,
7,
.
b+
'
<
/rth
port
675
For a moment, consider the case where the real part of the internal
impedance of the source is positive, that is, P k is equal to 1. Then Eqs.
(17-5-6) and (17-5-7) can be interpreted as follows: The source is sending
Howthe power |a 2 toward the load, regardless of the load impedance.
of
Z*,
a
part
equal
does
not
Z
is,
matched,
that
is
not
L
ever, when the load
This reflected power
the incident power is reflected back to the source.
2
so that the net power absorbed in the load is equal to
is given by |6*|
_
Associated with these incident and reflected powers, there
|5 |2
0fc |2
fc
fc
are
_ V + Rkh _
k
2VR
bk
Vk
VR
\/fi
fcth
voltage incident on
port
^7-5-8)
v Rk
V Rk
Vk
Rklk
Vik
fcth
port
(17-5-9)
-VRk
fc
V k = Vik + V
Ik
Rk
lik
Irk
and
If all o's
ik
Vrk
lik
Irk
and
b's in
Two -port
network
FIG
7-24
source and
Zh
load.
impedance R then we can divide both sides of the equation by Vi? and
obtain an expression relating incident and reflected voltages directly,
,
[Vrk]
[s][V ik ]
(17-5-10)
676
=
=
Sndi
^i
$2101
+
+
Si 2 a 2
(17-5-11)
S 22 d 2
(17-5-12)
where
ai
+Z
I1
a2
V|Re Zi\
~ ZUi
h =
V|Re Zil
V +ZI
V Re Z
V 2 - ZUt
V|Re Z
2
Vi
and
11
6i
a2
ai
S22
ai
S21
a2
ai
ai
S12
&i
a2
V rl = SuVn +
Vr2 = s iVa +
2
s 12
s 22
V
V
i2
(17-5-13)
i2
(17-5-14)
where
input
reflection
coefficient
*n =
III
22
III
S21
= Zl2
transmission
gain
with the output terminated
in a matched load
(17-5-17)
Yn
v
transmission
(reverse
gain
with the input port terminated in a matched load.
(17-5-18)
Va
v i2
Vi
Va =
(17-5-15)
(17-5-16)
(forward
S12
Va
i2
677
Equations (17-5-13) through (17-5-18) are very important and practiparameters can be simply determined by measuring the
incident and reflected voltages from the network.
cal since the s
Notice that
Sll
Vrl
Vi
- Roh
V x /Ii - Ro
Vn
+ Roh
Vi/U
Ro
Zi
Rq
Zi
Rq
i2
=
(17-5-19)
and
Zi
n
Ho
(1
u)
(1
Sn)
(17-5-20)
From
and Eqs.
Fig. 17-23
coefficients (normalized to
air
fl 2
62
_ (power
Ro ~ 1 power
Fii|
|F, 2
12
W
^L =
1
|5 2 |2
and
(17-5-9),
have physical
we
significance.
That
available
from a source
of
impedance
power
*21
load.
loss,
quan-
s 22
'
Sl2 |2
=
=
Sill
and
is,
(17-5-8)
678
TABLE
Su
,,.
17-2
Parameter Interrelationships!
(Zn
1)(2 "
(2ii
1)(*
(0.1
l)(zn
+
+
1)
~
-
Zli z 21
1)
012^2.
1)
212221
Zu
??i*
2"
12
5fi
(211
22
(211
Sl2
,
*21
1)
1)
~2
2/22)
Sn =
Sj2
l)(h t ,
_
~
n )(l
_
"
S22 )(l
(1
Sll )(l
(1
y 2l
(1
(1
1)
1)
hiihtl
^ 21
and
0'
and
0iiZ o
<7
,
0i 2 o
_
-
02i/5 o
_
-
022^0
'
2 21
2 22
77-6
22 )
si 2 s 21
S 12 S 21
Sl 2 S21
22
s 22 )
s, 2 s 2 i
-2s2i
(1
_
~
sn)(l
(1
(1
s n )(l
s 22 )
+s
S 22 )(l
Sn)
S12S2!
s u )(l
s 22 )
s 12 s 21
12 s 2 i
normalized to Z
If h',
is obtained as follows:
.
y',
2U
012
parameters
S12S21
Sl 2S21
21
h, y,
2si 2
(1
The
Si 2 s 2 i
(1
+ 1)(A + 1) - hiAti
+ hu)(l - A22) + ^12^21
(hn + 1)(A + 1) - hnh
22 )
(1
(1
huhii
~ 2/t "
2s 2 i
12 s 2 i
+S22)
(1
U =
Sl 2 S 2 l
-2s 2i
2/212/12
ftft
S22 )
~2si2
1)
5l2 S 2l
Sl2S21
22 )
+s u )(l
2/22)
"
S 22 )
Sn )
(1
2/122/21
2/122/21
_
_
-
(1
12 s 21
+ *2i)(i ) +si
+s a )(l +s - s
2/122/21
-s
2s 21
V"
(An
2s, 2
222
2 ^i2
(hu
-*n)(l -)
2/22)
2/ll)
(1
22 )
2/2i
+ 2/n)(l +
(1 ~
+
(1 + 2/n)(l +
(^11 ~ !)(& +
(flu + l)(h M +
(!
+ n)(l - S + SnSn
- n )(l -*) - SliSn
(1
(1
12
(1
(1
~ 2 V"
2/ii)(l
_ ~
2i
8tl
0i 2
2/22)
j/ 22 )
212221
(1
1)(Z 22
(1
S22
(1
(1
1)(2 22
1)(Z M
+ 1) 012021
~ y"^ 1 + ) + yiy
- ^122/21
+ 2/n) (1 +
(2n
Sll
+
+
+
=
Zo
u =
y'
'
y 12
'
2/21
J/ 22
A'n
2/12
^ 21
~7f
^0
= AnZo
12
Ai 2
l'
h
n.
"21
/i
2/22
/l
22
A22
This
insertion gain,
s 22
are a measure
679
loop rule."
The complete flow graph of the network in Fig. 17-24, including source
and load impedances, is shown in Fig. 17-25. The load and source are
described by their reflection coefficients T L and T s respectively, referenced
Notice that 6o represents the
to the real characteristic impedance Ro.
normalized voltage delivered from a source with a Z, internal impedance
2
to an Ro termination and that |fro| equals the power dissipated in an R
load.
We
The
now
shall
Two branches, whose common node has only one incoming and
Rule 1
one outgoing branch (branches in series), may be combined to form a
single branch whose coefficient is the product of the coefficients of the
.
Vs
Wo
Zs +
Ro
bz
bs
-*
>
Szx
'
TL - Zl-Rq
ZL + Ra
wSu
ZS + R
b\
az
th.
Two - port/network
FIG
7-25
source and
An
Z&
load.
680
a
az
S 21
o\
a-i
S 3Z
$31
SZ1
5 32
fa)
az
S Z\
'
SA
+ Sg
fb)
r
Szz
*/
o-i
az
S
1
a-i
*32
-522
(c)
original
branches.
is
eliminated
(see
Fig.
17-26o).
parallel)
whose
sum
coefficient is the
Fig.
(see
pointing from a
may
17-266).
must
Mason's nontouching loop rule will now be discussed. The nontouching-loop rule provides a method for writing the solution of any
flow graph by inspection.
The solution T (the ratio of the output variable to the input variable)
T =
?* T* A *
is
(17-6-1)
681
(a)
04
01
S4I
S54N.
al
02
"
*32*43
(b)
04
*32
*21
S43
$42
04
<7 2
42
1-S 22
(C)
FIG 17-27
branch, (b)
Node
node
where
Tk =
A =
path gain of
1
(sum
A;th
forward path
(sum
(sum
of all second-
Ak =
pliers
first-order loop is a
682
terminates on the same node, no node being encountered more than once.
second-order loop is the product of two nontouching first-order loops
(that is, two first-order loops that have no branches or nodes in common).
A third-order loop is the product of three nontouching first-order loops,
etc.
Loop gain is the product of the branch multipliers around the loop.
As examples
touching-loop rule,
we
The same
coeff icient=
"1
FIG 17-28
*i
(d)
reduc-
1-5 22
(e)
Flow-graph
PL
we
First
rule.
683
61,
using
Fig. 17-28a:
Forward path
Forward path
Next we find
Sum
all
= Su
= S2irz,Si2
The
therefore,
Ak
with k
2,
We
sn
(with k
Ai
s 22 T L
forward path
first
only one)
s 22 r L
A =
is
it,
and
1) is
s 22 T L
= A =
Ai
1.
_ a u (l -
Cll
s 21 Sx 2 T L (l)
S 22 Tl
SilSl2TL
8ll
s 22 T L )
ai
(17-6-2)
s 2 tT L
In this example,
-^
K ~
Kv
V,
Vx =
V*
a2
61
and
(a x
= (at +
= Tl&2
= s'n ax
VRo = V + V
V + V
61)
62)
V^ =
iX
rX
i2
r2
so
62 1
Oi 1
+
+
TL
*n
rule,
TL)
S22IVKI
(17 . 6 . 3)
4i)
Power Available from the Source. The flow graph for the source with
an arbitrary load T L is shown in Fig. 17-29. The power absorbed by the
2
The solution of the flow graph for a 2 and 6 2 in terms
load is |a 2 2 |
684
b
o
bs
)
aL
>
'
bL
1*0
I-
ps
of 6 o is as follows
o-l
aL
bL
T L T s aL
TL T S
VLb a
= TLaL =
\a L \>
T L Ts
2
|6o|
|i
(i
r^r,
|rx|
\a 2
|6o
= P avs =
max
- \U
(17-6-4)
PL
Pav8
r _
Pl = P
=
* avs
(i
6o|
(i
(reflected
from load)
|r^|*>
|r s
b2
Gt =
(incident on load)
(1
2
|
|r,|)(i
\t l \*)
60
From
60
--
snTs
s 22 T L
s 2 iS 12 T L Ts
SuTsS22Tl
S21
(1
s 22 T L ) SuSuTlTs
- |r a |)(i - \r L
- s n r s )(l - s 22 T L - s 21 Si T L T s
iil\s)(l
\s2i\
|(1
(i
\*)
2
\
(17-6-5)
lists
many
685
often-used network
TABLE
17-3
Two-port
network
ai
V,
6i
Parameters
a2
b2
=
=
bi
62
Input reflection
S2lOl
S21S12TL
= sn
Output
reflection coefficient
s 22 rz,
with arbitrary
si2S2ir,
S22
Si 2 a 2
S220(2
with arbitrary Zl
coefficient
sn
+
+
Snai
- snr
Vi
power input
|s2i|
(1
TL)
-*ri)(i +s'n )
(1
Power gain
G =
S2l(l
*.=!-'_
|s|)
|r L
2
|
to
network
(i
(|s 2 2|
|rz,| )
|>| )
Re (TlN)
M Q -2
G
(1
|r s
(|s
|r s
2
|
|>|
2
)
Re (T.M)
|(i
where D =
M=
N=
Gt =
Sus 22
Sn Ds M
s 22
Ds n
|rz.|
2
)
i2s 2 irz,r s
2
|
686
77-7
Measurement of
Parameters
earlier.
If we want the complex reflection
Sn and s 22 the measurement system used for determining
the driving-point impedance of a network or a slotted line can be used.
Because |s 2 i| is the insertion loss (or gain) of the two-port network, it
can be measured by the techniques described in the section on measurement of attenuation. The above techniques have been discussed in
It is important to realize
detail earlier and will not be reviewed here.
that the above classical methods are valid techniques for measuring
coefficients
parameters and
many
of our
measurement needs.
How-
device characterization.
Both the magnitude and the phase of all four s parameters (in the twoThis section will describe
port network) will sometimes be required.
one measurement system used for complete network characterization.
The Network-analyzer Concept. The network-analyzer system shown in
Fig. 17-30 contains three main functions: signal source, test set, and
receiver.
The signal source can be either continuous-wave signal generator or a sweeper.
The sweeper can be used for broadband network
characterization.
The amplitude of the swept rf output may or may not
Network
under
test
Port
Port 2
<
'
Reference
Signal source
rf
s- parameter
signal
test set
channel
Receiver
Test
channel
FIG
7-30
Network-analyzer system.
687
Gam
vernier
Reference
channel
Frequency
converter
AGC
Detector
Test
if.
Display
attenuator
channel
i
1
Phase
vernier
Local
oscillator
FIG
7-31
688
77-8
Measurement Techniques
4.
689
We shall now develop the error models needed to characterize the error
terms in our measurement system. For any two-port s-parameter
measurement system, Table 17-4 lists the possible error parameters and
their physical interpretation.
Figure 17-32a through d shows the four
basic error models with the use of these parameters.
The general error-analysis procedure is to write the flow graph of the
entire system and then solve for the appropriate transfer function by
using the flow-graph reduction technique or the nontouching-loop rule.
This can become difficult for a measurement system like that shown in
Fig. 17-30.
However, these error models can be simplified in specific
measurement situations after reducing the models to the basic forms of
Fig. 17-32.
The error models and error parameters will be described in
detail when we discuss specific measurement examples.
Assume that the error terms and ambiguities are now accounted for
and minimized. We are ready to initialize the measurement system.
During the initialization we may equalize the electrical length in the
and test channels and adjust the phase and amplitude vernier
reference
balance the electrical length in the reference and test channels. The
error model, the equation describing the reflection measured by the
Table
Fig.
7-4
Common
Models
in
17-32
Error parameters t
Physical counterpart
7*00
r 33
*30
t03
Transmission crosstalk
"01
7"32
^32
tai
ril"22
r ii r 22
twtii
< 22
Reflection port
mismatch
"
s r
PortO
^>
"'bo
bo
-521
$22 M
"5||
1i
Ms
r22
7,
Measured 5 M
*0
PortO
^1
'"Ol
4-
-5||
Port
/"||
Port 2
Port 3
- ^^
>
tf
-Sll
r00
Oo
521
^-a
|
Measured s 2
_ b 3
= CO
^22
"
12
A,
"
5 22
'5||
/So
+"
S||
/||
T 7~1 7
1
-5 2 2 Tz2-Sz[S\ztT~Z
n t22 + Su T\\S 22' 22
(b)
Port
Port 3
*rr-K
*2I
1
'
5 22
5||
i-^e
"fZZ
I2
<
Measured s 22
--
r 33 +
1_ S22
03
<
^32^1-51,
5 22
"
<
<}C
^33
^ -5
r,|
;+5| 2 5 2l
r,j
^ 32
'
/-,',-
l2
5 2 r2 2 r,|+s 22 /-2'2Sn i.
,
/W
PortO
Port
Port 3
Port 2
r"
*2I
A,'"
^22'
S|2
roi
5 22
'S
-<-
Meosured s l2
*o
^03
+-
5 12
1-S 22
A 22 -S||
/-|'|
^o
-5, 2 5 2 / 22
|
A||
S22 tzzS[\
(d)
error models: (a) for measuring
measuring S22, and (c?) for measuring
Ports 1 and 2 are the actual ports of the network analyzer; port
812.
and port 3 are fictitious ports used only for notational purposes.
FIG 17-32
su,
(.b)
Measurement-system
03
t\\
691
Port
1
1 r/
r \\
oo
ro\
rQ0
5 n rQ\
_5
ii
rn
r 00
as
receiver
f\
i
P r f
mismatch
(b)
FIG 17-33
and
system, and a description of the various error sources are given in Fig.
17-336.
This error model is simplified from the model in Fig. 17-32a
since Si 2 = S21 =.
for the one-port network.
We shall now describe the various error terms and the sources from
which they arise. If we place a perfect load on port 1 (s n = 0), we still
measure an apparent reflection. This may be caused by a signal leaking
either from the reference channel to the test channel in the receiver, or
via the directivity path through the test channel coupler.
The leakage
term will typically be down 80 dB or more below a full reflection, but the
directivity will typically be no better than 40 dB below a full reflection.
Hence, the directivity of the test channel coupler is the greatest contributor to the crosstalk term r 0Q
If the amplitude and phase response of
the coupler arms do not track or the two channels of the receiver do not
track, we measure an apparent ripple as we vary frequency, which cannot
be attributed to the network being tested. This ripple term is known as
.
692
Load
reflection
moving 36Q
Directivity 0.031 L
135
FIG 17-34
phasing
a sliding
quency.
From the center of the display to the center of the small circle
obtained when the load is phased through 360. Fullscale
8n = 0.05.
match
rii
many
Finally,
we know
times
is
typically
when measuring
becomes significant
measurements. The tracking error r Dl causes a percentage error in the
measured value of Su and is sometimes neglected when su 5s small. Also
the mismatch error r n is not as important when &n is small. The
mismatch and tracking terms, rn and r i, affect the measurement, of Su
most severely when Su is large.
In the continuous- wave mode of operation the errors can be calibrated.
First the source mismatch term r-a can be reduced by using a slide screw
tuner, etc.
Then a sliding load can be used to isolate the directivity.
Attach the sliding load to the output port and slide the load back and
forth (phasing the load).
The resultant output on a polar display is
small, rii
shown
in Fig. 17-34.
The
is
not
but when the load vector is rotated through 360 its center can be
located, which isolates the directivity vector r m
To determine the
aero,
The mea-
on the display.
For swept measurements, we
first
electrical length
693
Is
Ja
(a)
w
W
t?
Ijt
o\
8
?
k
FIG 17-35
.,
in
reflection
mea-
adjusting phase
in the reference
and
test channels.
The
polar display,
we obtain
measurement system.
Using a
is
adjusted until the line becomes a small cluster, and then the phase
vernier control is adjusted until the display reads an angle of 180.
694
termination placed on
Figure
17-.i7/j
one-port system.
The above
FIG
7-36
measurement*.
Test
set
r.f.
Source
Two-port
*o
network
termination
Portl
Reference
695
Test
channel
channel
Receiver
PortO
Two-port network
(b)
Measured
S\\
feo
Snr
+
(1
ri
r2 2
i(l
siir u )(l
s 2 2r 2 2)
+ S2iSi
S22?"22)
2 r 2 2r"oi
S2lS 1 2r l l"22
termination mismatch
so forth.
Most
696
FIG
tive
1 7-38
Compressed Smith chart used for direct readout of negaimpedances. Full-scale \s n = 3.16.
\
of the
crosstalk in the system ^oLet us now discuss the importance of these various error terms.
crosstalk term tso is usually a very small number.
Typically it is
The
down
80
dB
or
still
of attenuation.
Many
and are many times neglected in measuring a network with low input
and output reflection coefficients. It is difficult to measure the attenuation of a bilateral network with high input and output reflections.
zero
in
697
However,
This isolates, the crosstalk term t 30
$12 = S2i = 0.
sometimes hard to recover from the noise. Next we connect ports
1 and 2 and adjust the gain and pulse vernier controls until the magnitude
This calibrates the tracking term t 3 2is unity and the phase is 0.
For swept-transmission measurements the electrical length in the test
and reference channels must first be equalized. This can be accomplished
by the same technique used in the reflection measurement. For the
transmission measurement we connect the two ports and adjust the line
stretcher and phase vernier until the small cluster is centered at 0 on the
A phase-versus-frequency display could equally well be
polar display.
so that
t
30 is
used for adjusting the electrical length. With this display we adjust the
out the phase slope and then offset the phase
display with the phase vernier to obtain an approximately 0 reading for
rf
Source
Two -port
network
1
Port 2
Reference channel
Test channel
Receiver
(a)
PortO
Port 3
Port 2
Portl
"1
^r
S 22
\'5||
S|2
Measured
521
d
bz
00
1"S|| ^ll
^30
2*
crosstalk
~ SZZ ^ZZ~ 5 2l 5 I2
f$2
An
ca
port
/ 22
c.
^ll
^2Z
splitter
+S
&
\\
^n s Z^ZZ
receiver
mismatch
port 2 mismatch
1
(b)
698
FIG 17-40
in
transmission
mea-
till
If desired,
much
the
same way
as
ill
us rated
I
in
Fig.
Port
Two-port
network
rfo-
r\
**-
Port 2
Line
stretcher
699
V-^
Switch
FIG
and
reflection
and transmission
test set.
merits.
1
Combined
7-41
Test channel
Reference channel
to port
2.
far
from the known network, the tracking and mismatch errors are
comparatively small.
Many times we should like to measure
Multiparameter Measurements.
more than one s parameter with the same test set. Figure 17-41 shows a
^30
Port
Port
*r~
*2I
bo
ro\
iTwo-port/network
Error model
measurement system.
for
Hz
5 22
-<%
FIG 17-42
Port 3
Port 2
combined
reflection
and transmission
bz
700
rf
4w
<_
s,
Reference
channel
J~
Test
pad
A,
P^ v
channel
\Q
FIG
7-43
test
set.
switched from the reflection port to the transmission port. The analysis
and equations for this test set and the operation of it are the same as
described in the previous reflection and transmission measurement sec-
we had a three-channel
If
respectively.
They can be broadly classified as fixed-frequency or sweptfrequency techniques. The power of the fixed-frequency technique is
that the mismatch, tracking, and directivity errors of the measurement
system can be minimized by "tuning out" the residual errors and high
accuracy can thereby be achieved. Fixed-frequency techniques, howworks.
701
Network
under test
Reference
Signol
source
r.f.
signal
5 -parameter
test set
channel
Receiver
Test
channel
,
s-parameter
Freque>ncy
contro
Receiver
control
control
Output
data
Computer
FIG 17-44
Computer-controlled
automatic
network
analyzer
system.
mally, it is difficult to account for all errors when using the swept-frequency
technique, since they cannot be completely tuned out in the broad bandwidths.
We should like the microwave measurement system to provide
results.
702
used are such devices as shorts, opens, and loads, which are
relatively
easy to characterize and manufacture to accurate tolerances.
(3)
Speed.
The computer can easily control all the instrument functions
normally operated by the user. Then, too, the calculating power of the
computer greatly shortens the time required for complete network
characterization.
at
now
available,
CITED REFERENCES
1.
2.
3.
4.
5.
6.
7.
Adam, S. F.: "Microwave Theory and Applications," Prentice-Hall, Inc., Englewood Cliffs, N.J., 1969.
Wholey, W. B., and W. N. Eldred: A New Type of Slotted Lice Section, Proc.
IRE, vol. 38, no. 3, March, 1950.
and B. O. Weinschel: Swept Frequency Height Resolution VSWR
Measuring System, Weinschel Engineering Co., Internal Rept. 90-117
p 723
March, 1966.
Adam, S. F.: Swept VSWR Measurement in Coax, Hewlett-Packard J., vol. 18,
no. 4, December, 1966.
Ely, P. C: Swept Frequency Techniques, Proc. IEEE, vol.
55, no. 6, pp. 991Sorger, G. U.,
IEEE, vol.
IEEE, vol.
Fresh, D. L.
IRE,
Methods
of Preparation
703
of Ferrites, Proc.
12.
Heller, G. S.,
IRE
13.
6, p. 97,
Mc,
1958.
Duncan, B. J., L. Swern, K. Komiyasu, and J. Hannwacker: Design ConsideraBroadband Coaxial Line Isolators, Proc. IRE, vol. 45, pp. 483-490,
tions for
April, 1957.
14. Fieri, D.,
mission Lines,
IRE
in Dielectric
Loaded
9,
TEM
Mode
Trans-
1959.
15.
Weisbaum,
S.,
and H.
Seidel:
The
16.
Weisbaum,
24 KMC,
Treuhaft,
S.,
Isolators at 4,
6,
11,
and
IRE
UHF
Circulator,
G., and V. Dunn: Broadband, Lumped Element
Trans. Microwave Theory Tech., vol. 15, pp 198-199, March, 1967.
22. Ince, W. J., and E. Stern: Computer Analysis of Ferrite Digital Phase Shifters,
IEEE Intern. Conv. Record, 5, 1966.
23. Reggia, F., and E. G. Spencer: A New Technique in Ferrite Phase Shifting for
21.
McChesney,
IEEE
Beam Scanning
24.
CHAPTER EIGHTEEN
AUTOMATED MEASUREMENT
SYSTEMS
From
notes by
M. D. Ewy
Hewlett-Packard Company
Palo Alto, California
and
useful form.
An
1.
automated system
Many
repetitive
is
economically advantageous
measurements
must be
if
made
performed; or
2. A complicated operation requiring
a high degree of
performed.
will
or
operations
skill
must be
be described, which
illustrate the
1.
2.
3.
Data acquisition
Data analysis
Programming and
At the time
704
[1],
Low- level
Multiple analog
scanner
signal sources
Digital
705
Printer
voltmeter
t
I
l_.
Tinning
and
control
FIG
8-1
is
78-7
the need
is
for
and presented
measurement
a few microvolts.
At
this level,
introduced by
measurement capability
is
often limited
by the
noise signals.
errors
Signals
706
introduced identically on both the high and low input lines are referred
common-mode signals (see Fig. 18-2). Figure 18-3 shows how both
dc and ac common-mode noise signals are typically introduced into a data
to as
acquisition system.
The
is
High
Common-mode
Output
noise (E,
source
FIG 18-2
Common-mode
voltage.
common-mode
reject these
If
will
CMR
No
and 13
amplifiers themselves.
Output
ac interference
^->. source
--09
Earth ground
FIG
8-3
_|_
common-mode
voltage.
707
Normal- mode
noise source
High
Signal
fe
source v-s
FIG
Normal-mode
8-4
voltage.
Common-mode Noise
78-2
Rejection
Floating
measured.
Xd
= 1/wC
is
large
the
compared with
extent
that
CMR
interference
RL
the
to
approximately
ratio is
(18-2-1)
~ -48 dB
For long input lines, the use of shielding as shown does not improve the
An insignificant part of
c
noise rejection because of the low value of
flows through R H
normally
the current generated by the noise generator
For
voltmeter.
of
the
required
impedance
because of the high input
High
Floated
digital
voltmeter
=<Ti(0.01/iF)
Chassis
Earth
ground
FIG
Earth
ground (2)
(
8-5
floated
measurement system.
708
Earth
ground
some
8-6
floated
to 10
FIG
V rms)
special applications,
further
charging C 3 through
Since the potential on the guard shield nearly
s
follows the potential on the Lo input line, very little noise current
flows
through L and C 2
Also, the guard shield can be made quite effective,
so that Ci is only a few picofarads.
Therefore, very little noise
current
RL
and CV
Xci
in Fig. 18-5), 3
pF
of 0.01 M F for
C 2 and C 3
and 60 Hz,
for C,
RL
Xa ~5> Rl
Xcz ^> Rs
The
En
_j
(dB)
= -20
log
20 log
RsRl
Rl
XczXcz
Xci
approximately
Xci/
'
is
Rl\^
RsR L
\Xc2Xc3
CMR
RL
/ RsRl
Rl\
\Xc2Xc3
Xci/
(the
same value
709
and
CMR
Note that
-20
this
value of Xci
CMR
If
is
log
Xci
formula
is
log
decibels
(18-2-2)
-119 dB.
Rh
leakage resistances
tors Ci,
= -20
C 2 and C
,
Ri,
and
of Fig. 18-6, it
CMR
chassis or earth
78-3
ground
[2, 3].
filter of
some kind.
(see next section), is also significant at the 1-juV level. Therefore, a desirable filter not only would have
high attenuation at 60 Hz, but also good attenuation at all higher frequencies.
The filter should have minimum attenuation in the passband.
is preceded by a scanner, a filter at the input terminals
Because the
Otherwise, accurate
of the voltmeter must have fast transient response.
rates, to allow the
scanning
slow
measurements could be made only at
filter output to settle, or a preamplifier and filter would have to be used
DVM
of the scanner.
technique often used, which meets
all
710
of the interfering
is
approxi-
78-4
As previously shown,
it is
An
impedance
of 10 3
1 fxV,
in order of magnitude.
DVM
with an integration period of 3^o sec nas an equivbandwidth of approximately 30 Hz. Thermal noise in a pure
resistance is given by
integrating
alent noise
En = (KTR
where
En =
A/)^
(18-4-1)
noise voltage,
rms
resistance of 1,000
ideal source
/zV.
18-5
Crosstalk
E
E sn
Ri, R n
R P) R
si
Cp
C,
Ri
=
=
=
=
=
=
where
measured
situation,
signal being
interference signal
source resistances
leakage resistances
leakage capacitances
vw
711
Rn
I
(Esn
I
I
>Rp
-jCp
parameters
Crosstalk
(only Hi-line switches are shown).
FIG 18-7
(*1
Ri
The
total leakage
z =
t
impedance
R,Rp
R RMC, + C v +
)
At
is
(18-5-1)
R,
+ RP
dc,
Zt
= Rt =
R
R,
Rj
-\-
(18-5-2)
Rp
Z =
t
Crosstalk
X
is
1
t
(C.
typical applications, Ri
Then
x.
(18-5-3)
C P)
CT = R lU (C. + C P
CT is
(18-5-4)
or
CT
(dB)
= 20 log
RMC + Cp)
(18-5-5)
Sources
-To digitizer
FIG
8-8
712
In most applications,
guarding, but C,
system.
is
78-6
A
emf,
Thermal Voltages
commonly
magnitude
of the voltage
is
Reference
cold junction
<
Copper
S
\
To measurement
system
Hot
junction
FIG 18-9
Thermocouple.
Constantan
<
A5*
Copper
High
'f
SCoppei
-o
Connections'
Scanner
FIG
8-1
Thermocouple input
circuit.
713
very low.
and
exist
drift
the input.
And
if
two pieces
drift.
78-7
Scanners
DVM.
few milliseconds.
High-speed Scanners
called a multiplexer.
or Multiplexers.
high-speed scanner
is
usually
ft.
If
is
desired, Ri
must be
714
Source cable
Rn
&-L
Source cable
Signal
cable
FIG
8-1 1
(only
shown).
fect
This time
period of
to
is
make measurements
fore, a
device
is
usually a
DVM.
The
to
methods
78-8
Chap.
8.
is
in Figure 18-12.
sweep
715
[4]
of the device
being characterized.
channels
test
shown
[5,
6].
Any
of the device
in Fig. 18-12, the port-1 reflection characteristics
in the
converters
analog-to-digital
Two
measured.
test would be
under
for
network analyzer digitize the magnitude and phase information
computer input.
The computer is a small instrumentation computer [7]. All important
comfunctions of the microwave instruments can be controlled by the
freAlso, the computer can be used to store the system's own
puter.
perform
well
as
phase)
as
and
quency-dependent errors (both amplitude
Control of the system is as folnecessary mathematical calculations.
generator and network analyzer
sweep
the
sets
computer
the
First
lows:
signal through the multito the required frequency and then routes the rf
be measured is selected,
to
parameter
Next, the
plexer to the test unit.
no.
Port
no.
"*. Device
is
under test
Test
unit
Network
Sweep
analyzer
Generator
Frequency
Frequency
control
control
Digital
computer
Display
Teletype
FIG
1 8-1
Scope display
data
digitized
and
trans-
716
Error Correction.
The error correction capability allows the system to
make microwave measurements to the accuracy of a standards laboratory
and other
calculations
made
calibration constants.
This calibration procedure is automatically performed at each frequency of interest. Then, after the s parameters on the
unknown device are measured, they are corrected by an iterative computing technique that uses the stored calibration data, and this yields
in only a
few minutes.
The data transformation capability of the complay after error correction. The corrected s-parameter
Data Transformation.
The
digital
78-9
made
inaccurately or not at
all
because of
We
Fig. 18-13.
717
Unit under
test
Power
<
Power
^L
'
'
'
II-
Output
signal
Measurement
Switching
Input
signol
(-
Progrom
Program
Stimuli
Progrom
Control
and
Test
analysis
Progrom
digi tol
(
FIG 18-13
Automatic
:ompi
Display
or
record
test system.
programmed
connected to the system, the switching unit would be
would be
power
the
then
and
interconnections,
system
to make the proper
programmed
programmed on to the proper level. The stimuli are then
measurement might be comon, and after an appropriate time interval, a
is satisfactory, the
comparison
the
if
and
pared with a desired result,
At the commade.
be
to
test
next
the
for
programmed
system would be
on an
recorded
are
the unit, the significant results
test
is
all tests on
automatically,
output device. The entire sequence of tests can be made
that can be
measurement
any
Almost
without operator intervention.
with such a
automatically
made
be
now
can
device
a
on
made manually
and
repeatability
system, with a great saving of time and with increased
pletion of
accuracy.
and
calibration references
[8].
718
as these, the system could be used automatimeasure the gain and phase versus frequency, the time response
to a step input, and time delays and logical operation of
digital circuitry,
for example.
The measurements made are usually the same measurements that could be made manually, but the computerized system can
dramatically speed up the process.
The computer can also be used to
analyze the results more thoroughly and even to perform diagnostic
troubleshooting on test units that exhibit improper performance.
The instruments used in such systems are in most respects the same
as the instruments used in manual test setups.
The primary difference
is that the instruments must be programmable.
That is, the rangechanging, level-setting, switching, etc., that would otherwise be
done
manually must be programmed by the computer. Also, data generated
by the measurement instruments must be in a form suitable for reading
by a computer.
cally to
Programming.
modern automatic
is
in the
on (that
form
is,
test
systems
[7].
and analysis in
All information in such a computer
word
programming
data.
trolled
in
to an oscillator, for example, could control the switching of eight frequency ranges. Each bit is represented by an electrical signal, and only
Sometimes program data take the form of numbers, for example, the
channel number to which a scanner should be connected. This information is transferred as a binary-coded octal or binary-coded decimal
number. Some instrument controls, such as signal output-level controls,
are analog in nature.
Here a digital-to-analog converter is needed to
transform the digital program data before it is used to program the
instrument. The digital-to-analog converter could be a part of the pro-
grammed
instrument, or
it
719
hardware.
All digital program data transferred out of a
computer are transferred from a common data buss in the computer.
The program data must somehow be directed to the appropriate instrument. This is done by addressing. Each instrument has a unique
When an instrument is addressed, the comaddress, an octal number.
digital
computer could successively read program data from its memory and
transmit the data to instruments to connect system switches, program
power supplies, program an ac source, program an ac voltmeter, and
examine the voltmeter for measurement data, all in a time interval of less
than 100 /xsec. The instruments obviously cannot respond this fast.
Typically, the switches would require a few milliseconds to close and stop
bouncing, a fast power supply would require 100 jusec to settle, and an ac
source and voltmeter could each take a good part of a second to settle
and obtain an accurate measurement. The point is, system timing
information must be incorporated into the computer program. There are
basically two methods.
The most obvious approach is to program the computer to wait, or
"count time," for an appropriate period after each time that it programs
This is a completely workable
or communicates with an instrument.
approach, but it requires that the person programming the system (writing
the computer program) have a good knowledge of the timing requirements of all instruments in the system. Another approach is to provide
a line between the computer and the instrument, over which timing
information is transferred. With such a line, an instrument can "tell"
the computer when it has "timed out" or settled, by changing the digital
The computer then simply monitors this
level on this line at that time.
line until the appropriate level change occurs and then proceeds with the
next system operation. The system then operates at maximum permissible speed, determined by the instruments, without the need for putting
explicit instrument timing information in the computer program.
More
720
the computer program goes back to the operation with the instruments.'
This technique of course makes more efficient use of computer time.
CITED REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
5,
May,
1968.
INDEX
INDEX
Ac
(see
Alternating current)
Admittance, 265
AGC (automatic gain control), 523
Alternating-current comparator, 282
Alternating-current probe, 261
Alternating-current resistance, 265
Amplifier:
audio, 480
chopper, 204
common-mode
rejection,
491
deflection, 381
differential, 201, 376,
496
direct-coupled, 188
direct-current, 186
distortion,
500
Amplifier (Cont.):
video, 480
Amplitude modulator, 584
Analog-to-digital conversion, 86
Analyzer:
spectrum, 616
wave, 632
Analyzer system, automatic, 714
Anhysteretic magnetization, 469
Aristarchus of Samos, 2
Astronomical observations, 2
Astronomy, 4
Atomic frequency standards, 170
feedback, 485
gain measurement, 484
impedance, 482
input balance, 491
loop gain, 483
low-noise, 710
measurements, 480
microwave, 552
phase measurement, 48
slew limiting, 505
solid-state, 550
transistor, 554
tunnel-diode, 551, 553
Bode
plot,
486
Brahe, Tycho, 2
Bridge:
active, 301
automatic, 302
basic circuits, 292
frequency, 291
723
724
Indej
Bridge (Cont.):
inductively coupled ratio arm, 298
Kelvin double, 285
loading error, 294
low-frequency, 288
megohm, 287
radio frequency, 304
special purpose, 297
Warshawsky, 286
Wheatstone, 282
Curve
Data
Dc
tracer,
416
(see
Capacitor:
frequency
272
Detection
standard, 274
Goherent, 99
Detector:
average-responding, 239
peak, 245
peak-to-peak, 249
phase, 25
quasi-root-mean-square, 253
root-mean-square, 250
Differential amplifier, 201
Chopper:
amplifier, 204
Digital voltmeter
counting, 219
transistor, 207
Circulator, 666
(DVM), 211
Coherent detection, 99
Comite" Consultatif International
ramp, 218
potentiometric, 213
Conducted
Conducted
Mesures, 12
Common-mode
707
Communication system
interferences,
536
interference, 511
susceptibility, 515
measures, 41
725
Inde:
113
Disturbance, background, 102
Gunn
oscillator,
565
Gyrator, 669
Emu,
RX
meter, 305
266
117
circuit,
filter,
Ergodic systems, 68
Esu, 7
Impedance, 265
precision measurement, 311
transfer, 268
Impedance measurements, 264, 655
common
pairs, 109
measurement, 176
VLF
broadcasts, 160
564
Gain margin, 36
Gain measurement, 19, 489
Galvanometric recorder, 428
Group delay, 31, 34, 530
by
crosscorrelation, 101
meter, 227
Interference, 536
conducted, 511
radiated, 514
Interference measurement, 653
Poids et Mesures), 12
International system of units, 12
Isolator,
664
LC
oscillator, 321
Line shape, 126, 127
Linear systems, analysis
Loop gain, 35
of,
17
726
lnd<
Loss measurement, 19
amplifier, 710
National
Low -noise
159,
station
Matched
filter, digital,
141
Mathematics, 66
Mean
WWVB,
160
traceability, 313
Network,
305
electrical,
bridged T, 306
Network analyzer, 29
s-parameters, 686
background, 90
common-mode
Mean-square values, 69
Measurement:
automated systems, 704
computer systems method, 701
electromagnetic interference, 653
guarding, 708
effect
spectrum, 138
ensemble average, 66
mean-square values, 69
on measurement of power
of,
475
spectral measurements, 87
in spectrum analyzers, 628
as a test signal, 87
modulation, 639
multiparameter, 699
mutual-inductance, 300
presence Of noise, 97
radio-frequency, 303
Nonlinearity, classes
reflection, 655,
Normal-mode
time average, 67
white, 71
689
transmission, 695
Microwave
Microwave
amplifier,
46
709
552
attenuation
measurement
Ohmmeter, 277
Optical recorder, 448
659
Modulator:
absorption, 585
amplitude, 584
of,
rejection,
558
frequency, 556
step-recovery diode, 562
varactor-diode, 556
Oscillator:
avalanche, 569
beat-frequency, 340
Gunn, 565
impatt diode, 567
LC, 321
microwave
transistor,
phase-shift, 337
polyphase, 342
RC, 328
ring,
339
565
578
tuning, 548
Wien bridge, 331
solid-state,
stability,
Oscilloscope:
accessories for,
418
545
Index
meter, 310
Quartz frequency standards, 167
Oscilloscope (Cont.):
amplifier,
374
classification of,
delay
lines,
352
392
digital readout,
407
413
plug-in, 395
Radio receiver,
508
sampling, 398
Random
storage, 397
PAM,
647
Period measurement, 175
Phase detector, 25
Phase margin, 36
Phase measurement, 23
Phase-shift oscillator, 337
Physical constants, table, 14
oscillator,
342
Power meter:
thermistor, 600
specifications for,
Recorder:
amplifier for, 440
galvanometric, 428
475
448
pen mechanisms, 446
phase response, 438
position feedback, 454
servo, 456
sine response, 436
straight-line writing, 450
transient response, 430, 459
noise,
optical,
Reflectometer, 657
Resistor:
frequency
power
effects,
coefficient,
270
270
standard, 272
Gaussian, 72
sampling
and military
540
modulation, 521
noise figure, 520
quieting, 516
Recording:
digital, 466
magnetic, 461
thermocouple, 598
Power sensor, 596
of
amplitude-modulation,
federal
Polyphase
727
115
Probe:
alternating-current, 261
direct-current, 259
Programming, 718
Pulse-amplitude modulation (PAM), 647
Pulsed-power measurement, 605
temperature
coefficient,
270
728
Ind.
Signal generator:
Standard
atomic, 12
prototype, 11
time, 157, 162
examples
of,
50
Susceptibility:
conducted, 515
electromagnetic, 514
Sweep generator, microwave, 588
319
Sinad sensitivity, 519
Sine-wave synthesis, 345
Test signal:
509
Slotted line, 655
pseudorandom, 93
random, 91
SMPTE
ephemeris, 157
solar, 158
sidereal, 158
standard, 157, 162
method, 504
Society of Motion Picture and Television
Engineers (SMPTE) method,
504
Solid-state oscillator, 565
mean
universal, 158
microwave, 556
Spectral analysis:
Time-domain reflectometry,
by
digital
methods, 126
power, 134
628
real-time, 635
superheterodyne, 620
with tracking generator, 635
trf, 633
46, 61,
408
430
and military
specifications,
540
Index
Vector voltmeter, 29
Voltage measurement, alternating-current, 236
Voltage-to-frequency conversion, 221
Transmitter (Cont.):
keying waveshape, 534
noise-loading, 525
residual noise, 533
(cathode-ray oscillo-
Trigger generator
729
scope), 389
Tunnel-diode amplifier, 551, 553
Voltmeter:
averaging, 239
effects of distortion,
242
peak-reading, 245
root-mean-square, 250
sampling, 255
Units
electrical, 7
Warshawsky
electromagnetic (emu), 7
Wave
electrostatic (esu), 7
of, 6
International System
definitions of 12
history
of,
12
Wavemeter, 611
Wayne Kerr radio-frequency bridge, 305
Wheatstone bridge, 282
White noise, 71
Whitney, Eli, 5
Wien
Variable persistence in cathode-ray tube,
370
bridge, 286
analyzer, 632
333
Writing speed (cathode-ray tube), 363
of,
Variance, 84
in digitizing, 123
Vector impedance meter, 281, 310
X-y
nm
ca-o
Alllf/j
**
OTHER McGRAW-HILL
INTERNATIONAL STUDENT EDITIONS
tMIi^
Uvarez:
Auslander:
Bartee:
Belove:
go
AND DEVICES
ANALYTICAL AND COMPUTER METHODS IN
FOUNDATION ENGINEERING
Davis: INTRODUCTION TO ELECTRONIC COMPUTERS 2/e
Davis: COMPUTER DATA PROCESSING, 2/e
Director: INTRODUCTION TO SYSTEM THEORY
Dorfman: LINEAR PROGRAMMING AND ECONOMIC ANALYSIS
Gass: LINEAR PROGRAMMING, 4/e
Ralston: INTRODUCTION TO PROGRAMMING AND
Bowles:
COMPUTER SCIENCE
07-085544-7
YnVi