Professional Documents
Culture Documents
Presented AT THE 51st Convention 13-16
Presented AT THE 51st Convention 13-16
1031 a
BY
DON DAVIS
SYNERGETIC AUDIO CONCEPTS
TUSTIN, CALIFORNIA
PRESENTED AT THE
51st CONVENTION
r
INTREPRETING FIELD MEASUREMENTS OF DIRECTIVITY FACTOR AND THEIR
RELATION TO THE PROPOSED STANDARD METHOD OF MEASURING THE DIRECT-
TIVITY FACTOR OF LOUDSPEAKERS USED IN COMMERCIAL SOUND WORK.
by
Don Davis
Synergetic Audio Concepts
Box 1134
Tustin, CA. 92680
ABSTRACT
Increasingly accurate results attend the field measurements of critical distance.
Multipliers and divisors of critical distance have been sufficiently identified
to allow usable correlation between the proposed anechoic chamber method and the
semi-reverberant field measurements. Various anomolies regarding geometric Q,
maximum Q, relative Q, and a will be discussed. A more usable critical distance
equation will be presented. A product goal evolved from these techniques will be
suggested.
THE ROLE OF THE DIRECTIVITY FACTOR IN ESTABLISHING THE SOUND FIELD
Hopkins and Stryker, in their seminal paper in 1947,ldeveloped a basic equation
that has found wide usage in the noise control field as well as in architectural
acoustics. 2
Experience with the use of these equations reveals that at a distance from the source,
the level ceases to decrease whereas close to the source the change in level with
distance follows the inverse square law relationship. (See Figure 1)
DEFINING THE CRITICAL DISTANCE (Dc)_
From Figure 1 we can see that where the extension of the two rates of level change
intersect is Dc , critical distance. This is the distc_noeat which, according to the
Hopkins-Stryker equation, the direct sound level is equal to the reverberant sound
level or the ratio of direct-to-reverberantsound is 1/1. Therefore, we can derive
a further equation from the original one as follows: Assume that at Dc
4nDx2 R
Then:
3!
Dx2 : 16_
and
Since we defined for this special case Dx as equal to Dc, then we can substitute
D c = 0.141_ 4
and its other forms are:
(Dc)
Q: 0.01_
(Dc)2
R = 0.019881Q
Q io[D___
]._ Where: PI : the directivity index in dB
-2-
expected >2D c.
The critical distance equation then becomes:
Dc = Ref.dist. Ireference
x 10
_ distance SPL
20- reverb, point SPL]
Th_s method allows two frequency responses to be taken on a 1/3-octave real time
spectrum analyzer (one at the reference distance and one at the reverberant field
measuring point >2Dc) and plotting either Dc , Q, or R at each frequency by the
difference in level between the two frequencies. Therefore, we can write, for
single souroe devices:
~3=
direct-to-reverberant sound. If, in an overhead distributed sound system, five
loudspeakers provide the listener with both direct and reverberant sound energy
and 35 provide the listener only with reverberant sound, then the first five
loudspeakerscan be thought of as one group and 35/5 can be thought as the re-
maining groups of equal acoustic power contributing to the reverberant sound field.
Therefore, if the first group of five = 1 and the remainder of the groups = n,
then n : 35/5 = 7.7 The total number of groups contributing to the reverberant
sound field, then equals n+l, or 7+1=8. This number used as a divisor in the
critical distance equation correctly modifies the Dc that will be measured. We
can now write a more useful form of the critical distance formula:
Dc : 0.141W'_?_
r r- D l ]
Q = 4HDx2llo[lO J - 4(_1) and m ....
RLlo[_e]
_4(n+l)[4__Dx]
]
In large arenas with large arrays covering very wide angles, the effect of (n+l)
is crucial and any chance to manipulate a modifier must not be overlooked.
Ability to switch off sources not needed at a given time is a common method of
reducing the detrimental effects of (n+l).
DEFINING Q, Re , Df, etc.
In the Appendix of the 7th Edition of Handbook of Noise Measurement by Arnold P. G.
Peterson and Ervin E. Gross, Jr., directivity factor is defined as follows:
"The directivity factor of a transducer used for sound emission is the ratio of
the sound pressure squared, at some fixed distance and specified direction, to the
mean square sound pressure at the same distance averaged over all directions from
the transducer. The distance must be great enough so that the sound appears to
diverge spherically from the effective acoustic center of the sources. Unless
otherwise specified, the referenoe direotion is understood to be that of maximum
response." (Italics by the author.)
In a note to the above definition it is stated that:
"This definition may be extended to cover the case of finite frequency bands
whose spectrum may be specified. The average free field response may be obtained,
-4-
for example, by integration of one or two directional patterns whenever the
pattern of the transducer is known to possess adequate symmetry."
Several very important concepts emerge from this definition:
1. Q is always at some point and does not cover an area. A large area may have
the same Q all over that area but that is because any point within that area
satisfies the above definition.
-5-
The power level (PWL) is found by
PWL : lO log Pwatts - 10 log lO-12
Where: PWL : The power level to a reference 10-12 watts
Pwatts = The acoustic power
Therefore, 10 log 1 watt - 10 log 10-12 = 120 dB-SPL. And at four feet, we can
write:
SPL = PWL_12 + l0 log - 20 log Dx - 0.5 dB
Therefore,
SPL = 120 + 0 - 12 - 0.5 = 107.5 dB-SPL
DI : 10 log Q
Then, of course, another method of expressing Q would be:
We can see in Figure 5 that the coverage angle of our theoretical loudspeaker with
a Q: 2 would be, in the ideal case, 180°x180 °. Wall-mounted or ceiling-mounted
cone loudspeakers approximate this case at low frequencies but as we shall see,
-6-
the situation becomes more complex when dealing wit_ real-life cone loudspeakers
at higher frequencies. 8 9 l0
THE Q AND DI FOR SPHERICAL SEGMENTS
We can now construct a simple table of Q and DI for spherical segments from the
total sphere to 1/8 to a sphere. See Figures Figure 6 and 7.
Q: 10[dB-SPL directional
lO -
dB---sWq.
Manufacturers of loudspeakers have not used this method but rather have concentrated
over the years on gathering polar response data in only the horizontal and vertical
planes. Various methods have been sought out to economically utilize the wealth
of data at hand to calculate Q. While recognizing that the first attempts were
crude, it should be recognized that at the time these cruder methods were used, the
alternative was no Q data at all.
MEASURING THE Q OF REAL LOUDSPEAKERS
Wolff and Malter, is in a paper in 1930, were the first to describe the directional
characteristics of loudspeakers, so far as the author is aware. A second giant
step was_ken by Hopkins and Stryker in 1947. (This work, plus the papers by C. T.
Molloy, 16 are heavily quoted by L. L. Beranek in his works.) It remained, however,
for Bob Beaversl? of Altec to take the big jump to measure and assign values by
frequency for a line of commercial sound loudspeakers. His method was an adaptation
of the one used by Ben Bauer to measure the Q of microhpones. 18 This same method
was extended to other transducers 7 years later. 19 The author, by measuring Q
indirectly in the field in reverberant spaces, helped in the construction and
identification of the method used today.
Because real-life loudspeakers exhibit lobing (in some cases severe), it became
necessary in a majority Ofothe cases to average readings on a horizontal and verti-
cal polar pattern every 10 in order to arrive at a resolution that correlated
with field data taken to confirm the effect of Q and of Dr modifiers. Fortunately,
it was also found, at least for the higher quality commerCial-type loudspeakers
involved, that symmetry betweeen the horizontal and vertical axes was smooth
enough to require only two polar plots per frequency of interest. Again, fortunately,
though 1/3-octave bands were first employed, the 1-octave band intervals proved
sufficient for planning the effect of Q by frequency on acoustic gain, articulation
loss of consonants in speech, etc.
A manufacturer who measures a horizontal polar plot and a vertical polar plot at
each of the following center frequencies has sufficient raw material to adequately
state coverage angles and Q (See Figures 11-16): 63 Hz, 125 Hz, 250 Hz, 500 Hz,
1000 Hz, 2000 Hz, 4000 Hz, and 8000 Hz. Because of the %ALnnNs formulas, the
1000 and 2000 Hz bands are felt to be the most important oE_, and if only two
frequencies can be measured, it should be these. Since both a vertical and horizontal
polar plot are made at each frequency, the manufacturer must then process 16 polar
plots in order to obtain the desired data. e°
PROCESSING THE POLAR PLOTS
Start at the on-axis 0° point on the horizontal and vertical plot. The SPL is
recorded and weighted according to the area surrounding the center point compared
to the area surrounding the next center point. (This weighting method was first
suggested to the author by ArnoldP. G. Peterson). See Figures 17-19.
These are then summed:
fdB-S___Z_PL] fdB-SZP_L_LPL]
l0_ lO J x weighting + lO[ 10 J x weighting + ......
The same is then done for the vertical pattern.
-8-
This summed power level for both horizontal and vertical is then re-converted to
dB-SPL by l0 log Summed Power Level. 10 log SPL summed gives the average dB_.
Choosing any given directional dB-SPL, for instance, the on-axis point, allows the
calculation of Q for that direction:
Q: [dB-SPLD
,
lO,' TO- dB--2_-
The same polar charts can be used to assign the effective coverage angle for the
loudspeaker at each of the same octave bands. The -6 dB point below the on-axis
level is a normally chosen parameter for such angles. Thus, for each octave band,
it would be possible for the conscientious manufacturer to assign his loudspeaker:
1. An on-axis sensitivity figure such as x dB-SPL at y feet from z watts
2. An effective vertical and horizontal coverage angle
3. The Q by octave bands.
The generation of such data would immeasurably benefit the commercial sound field,
especially in having an accurate knowledge of Q. By comparison with loudspeakers
of known Q, various Dc modifiers could be properly investigated.
INFLUENCE OF THE SOURCE POSITION ON Q
The position of the source within the environment can have substantial influence
on the lowest Q encountered. The wall or ceiling mounting yields a Q = 2. The
wall and floor or ceiling intersection provides a Q = 4, and the intersection of
two walls, and a floor or ceiling (a corner) equals a Q TM 8. This is another form
of geometric Q and has had advocates for over 40 years (Knowles, Klipsch, etc.)
See Figure 20.
CALCULATION OF R AND OTHER PARAMETERS USED IN THE FIELD TO OBTAIN Q
One interesting problem we have encountered in the field is the frequent misuse
of the standard reverberation equations. These equations are employed to obtain
the a for use in the critical distance equation, the Hopkins-Stryker equation, etc.
Professor Sabine's original equation was
0.049V
RT6° : S_
= Slal+s2a'2....sna n
S
-9-
Where: si, s2, etc. = Individual surface areas
Snan : The number of Sabins assigned to objects,
people, pews, etc.
Note: Metric dimensions require a replacement of the constant 0.049 by the constant
0.161.
In the limiting case for a completely absorptive room, the RT60 failed to drop
to zero. R.F. Norris 21 pointed out an even more grievous difficulty in that absorp-
tion coefficients obtained by the use of the Sabine Equation in reverberation
chambers greatly exceeded 1.0. Norris and Eyring then developed an equation that
if used in the same circumstances would yield the true a:
[0.049V t
= 1 - e-I S_°J
This will yield "a" values from 0 t2 1.0 for materials that range from total
reflectivity to total absorption, a values obtained by this equation are converted
back into Sabine equivalents by -In(I-a).
Therefore, the two equations are identical provided the correctly derived a is
inserted:
We are left with the possibility of needing still another RT6o equation
RT6o : 0.049V
R
Since a is a dimensionless quantity, proper use of any of these equations in a
real room gives numbers that are useful. Only the Norris-Eyring Equation properly
scales a for 0 to 1.0, however.
At the present time the most useful of all these equations appears to be the
Fitzroy based on the Norris-Eyring Equation and the averaging of the apparent
RT6o of the three major axes of a given space.
The equations in Figure 21 illustrate the rationalizing of these equations and
provide the necessary conversion equations to change any a value to any other form.
SOME OF THE BENEFITS THAT CAN ACCRUE
Just a few of the current benefits being realized suggest a multitude of others yet
to come:
-10-
7. A better understanding of the Q of arrays and their interaction with R,
(n+l), and m.
8. Study ways to improve the correlation between the coverage angle, Cz ,
and Q in real-life cases.22 23
9. Define new and unique polar responses more suitable to actual audience and
installation requirements thereby stimulating product design in these
areas. (See Figure 22a and 22b.)
WHY THE AES
The author has chosen the AES as the proper forum for this discussion because it is
felt that the type of standard required must, of'necessity, be closely followed
and supported by the concerned transducer manufacturers and the majority of these
concerned manufacturers are directly involved as participants in and supporters of
the AES.
It is also felt that all too often, in the past, standards that should have logically
issued from the AES did so instead from other sister societies because they were set
up to handle such matters internationally and the proposals fell into the hands of
workers with membership and involvement in both societies. In the case of a Q
standard for commercial sound transducers, we feel that the need of participants
involved in such work directly is necessary in order to produce a meaningful stand-
ard. Those in noise control, architectural acoustics, etc., while vitally interested
in Q, do not always appreciate the requirements of the commercial user of transducers.
Therefore, it is hoped that the AES will actively further the development of these
needed standards within our own very well equipped membership for presentation as
an international standard.
-11-
REFERENCES
1 H.F. Hopkins and N. R. Stryker,A Proposed Loudness-EfficiencyRating for Loudspeakers
and the Determination of System Power Requirements for Enclosures, Proceedings of the
I.R.E., vol. 36, no. 3,(March 1948) pages 315-335
2 Arnold P.G. Peterson and Ervin E. Gross, Jr., Handbook of Noise Measurement, 7th Edition,
· General Radio. Especially Chapter 8, pages 189-205
3 Don Davis, Engineering Loudspeaker Locations, Altec Technical Letter # 182. (1968)
4 Melvin Sprinkle, Private communication between author and Melvin Sprinkle, consultant,
Washington, D.C.
5 Richard C. Heyser, Acoustical Measurements by Time Delay Spectrometry, J. Aud. Eng.
Soc., vol. 15, no. 4, pages 373-382 (October 1967)
6 Cecil R. Cable, Acoustics and the Active Enclosure, J. Aud. Eng. Soc., vol. 20, no. 10,
pages 823-826 (December 1972) (Dc modifiers first identified)
7 Don Davis, New Insights into the Critical Distance Formula - Visualizationof N+I,
Syn-Aud-Con Newsletter , vol. 2, no. 1, pages 14-15 (September 1974)
8 Cecil R. Cable, Loudspeaker Q - What Does it Mean?, Preprint No. 956 iJS) Audio
Engineering Society, May 1974
9 Don Davis, Oomments on Directivity o_ Louc_speakersand Microphones,Altec Technical
Letter no. 183 (1968) (Includes a reprint of the Hopkins-Stryker paper referenced above.)
IO. Don Davis, Notes on Loudspeaker Direotivity, Altec Technical Letter no. 211 (1971)
ll Don Davis, On Standardizing the Measurement of Q, J. Aud. Eng. Soc., (Forum) vol. 21,
no. 9, pages 730-731 (November,lC73)
]2 Don Davis, Further Comments on Direotivity Factor, J. Aud. Eng. Soc., (Forum) vol. 21,
no. 10, pages 827-828. (December 1973
13 Geoffrey L. Wilson, Direotivity Factor: Q or R_? Standard Terminology and Measurement
Methods, J. Aud. Eng. Soc., (Forum) vol. 21, no_ 10, pages 828-833 (December 1973)
14 Geoffrey L. Wilson, More on the Measurement of the Directivity Factor, J. Aud. Eng. Soc.,
(Forum) vol. 22, no. 3, pages 180-181 (April 1974)
15 Irving Wolff and Louis Malter, Directional Radiation of Sound, J. Acoustical Soc. America,
pages 201-241 (October 1930)
16 C.T. Molloy, Oalculation of the Direotivity Index for Various _ypes of Radiators,
J. Acoustical Soc. America, vol 20, pages 387-405 (July 1948)
17 Robert Beavers, Useful Direotivity Data for Reverberant Sound Fields, Altec Technical
Letter, no. 20lA (1969) (Includes actual measurements of typical commercial sound
loudspeakers'directivity factors at 1/3-octave intervals)
18 B. B. Baumzeiger (Bauer), Graphical Determination of the Random Efficiency of Microphones,
J. Acoustical Soc. America, no. 11, pages 477-479 (April lC40)
19 Paul M. Kendig and Roland E. Mueser, Simplified Method for Determining Transducer
Directivity Index, J. Acoustical Soc. America, vol. 19, no. 4, pages 691-694 (July 1947)
20 Don Davis, A Proposed Standard Method of Meaeuring the Direotivity Factor (Q) of
LoudspeakersUsed in Commercial Sound Work, J. Aud. Eng. Soc. vol 21, no. 7, pages
571-578 (September 1973)
21 R.F. Norris, A Discussion of the True Coefficient of Sound Absorption - A Derivation
of the Reverberation Formula, Apprendix II, pages 603-605 in V.O. Knudsen's
Architectural Acoustics, John Wiley & Sons Inc (1932)
22 Benjamin B. Bauer, A Diopole Speaker System, Audio Magazine, pages 50-56, (September
1974 / (Illustrates why the required Q of loudspeakers used for quadraphonic reproduction
must >>7)
23 Winston E. Kock, Sound Waves and Light Waves, Anchor Science Study Series (1965)
(II]ustrateshew to photograph loudspeaker directional effects.)
B \
I, \ f
%
Fig. l. RELATIVE
ATTENUATIONFORANY GIVENr
AL.i. EIV_'le_y ""-
_7 _ ._... T_T,_
L R_ _._=o, 16
'.....
1-0.16 = 1.24
M: 1_-_.32
D_=
o_41/_x_
.24
=111
ft
Fig. 2. Dc Modifiers (M)
N+l = 1
N+I
=1
N+l
=2
180
F6,.'I_,l
2 ARC SIN _.._
,v o J
Fs'"_(_'
1 ;*
0:2 v o 2A.cs,.)_--F-_)
Fig. 5. A hemispherical radiating LSmi J
surface,
IV FOR ANGLES:> IS0%THENONEANGLE = lBO°
AND THE OTHER ANGLE" SOMEVALUE BETWEEN
A 180_ AND 380°
ectivityfactorand directivity
index.
tl0t iw 3o __ 2 9
O:4 _.-
Fig.6, A quarter-shperical
radiating -'- "--_ _.
surface. "'
_.. at -- 'J
-'
A a0'
ms --11
_" .o_ --U _.
.s_ --1314
_3s(--22 '_'
, t' 180"
(-0._1_8,O,B??,O.NB) (0,1'4(I,0,8';"7,,0_C._3) OFF AXtS
+
(0,127,
- 0,934,0.353)
o.'
Fig. 9. Plane view of lO points
distributed on a hemisphere of .'
units radius.
:- ii
'; .;-
'%
, o_a3 ,, ol4_ ,_ m \
surface. '"________'___.
_.___'__i,_"'
_Y//_'
_;"; . __
,I?_?_,.o,,,
Fig.'11. Typical equipment set-up for '_'-_"_--
ii
....
measuring Q and sensitivity of a loud- Fig, 14, An actual horizontal polar
speaker in an anechoic chamber, response of a horn loudspeaker·
°[_]_LILLUIUt I I_/UILLULI_uMll
II_
,_1I lIIfl
[[llll
I_
I P liJlllll[[I
I Ifhq_r_t_[_lf
I I IIII I Lllll -I'1 IIII I L_"_,-
·I IIIIIIIII111 I IIIIIlilTl)--I IqlTTTI.j]
'"1 I lill]llJIU I I[llfi[llU I JIIlllffl
.J 1 IIIil11111[ IIIIIIIIIIII IIIIIIILU / '
°] l llllllllllP I Ill[Jr[Nfl J IJJl[IJll
.I I IIIII_IIJW__J_JJJJ]__I]WI] II]111]
'_1 I IIIlllUW] 1 Illllllltll_l_L.J_.l_lJ_lJ_
q f lllllJ Pl[fllfifllJ J IIIIllill
°°1 Illlllllllll I
,,_! I IIIIItlllll I lllllllll_l o£111111111
Fig. 15. Angular distribution of a horn
(plottedat the-6 dB pointson a series
of polar responses).
<_D_ · 1 . ,,',..,,_.,. [ i I,_l]l:llTl_
I I [ irlrlllJlll [ i IlllfllJ [1]1
I ] I I]llllllllll I I IIJIIIflllll ,8o
j ]_ Ij jjjjjjj;jjjj
I ] i ii i iff ij ]j j_,,,_
i ril.fllli[ll ................[ ......?._,,,4-1
' _1 .......,]
,
I J, I, [,,,,,t111,,,
rJlil Ill J, I, ], ]lflllf
_,,,,,,r,, Ill[ ............ _]
Lb! !! ll_l1t",..........
......... I!!! ! ! !!_"_'?!!!!
I)I',I] Fig.18.AreaformulasforA-l,A-2,etc.
L L I I IIlllllllll [ I ImFl[]lll Jill
I I I I I111111 Ill I I I11111111111
irlr LJii I J I [llllrlllEI
I 11/111T]111 / I l
J--'----J---_l [llJIIIJ. lli I I I_
J I I Illlllll_ll I I IIllllll]lll
/ J I I IIIIlllqlll I I I ¥J_LLJJJJ_L_
/,mst \ '
_,,_,,o,,_-
is : 0,049V si
S_ an: 1 - e '(as)
iR=s;r-Y+r
fo.o49v]
as = -L_(1-2n) in = 1 - e (S'-7_6_6oJ iR = S('Zn(1-_L)
_ _2R , (S(-Z.(_-_
n))+s
as = -iR n
(i-R=T-I) a : 1 - e[2_ ) 2R = 1
Sa_.Case Values
When: Then:
80*
2gO °
70 °
290°
60*
300*
50°
310°
40"
320:
Fig.22b. Idealized
polarresponseforoverheaddistribution.