Methods of Calculating The Average Coefficient of Sound Absorption

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

METHODS OF CALCULATING THE AVERAGE COEFFICIENT OF

SOUND ABSORPTION
Carl F. Eyring

Citation: The Journal of the Acoustical Society of America 4, 178 (1933); doi:
10.1121/1.1915599
View online: https://doi.org/10.1121/1.1915599
View Table of Contents: http://asa.scitation.org/toc/jas/4/3
Published by the Acoustical Society of America

Articles you may be interested in


Measurement of the absorption coefficient of sound absorbing materials under a
synthesized diffuse acoustic field
The Journal of the Acoustical Society of America 136, EL13 (2014);
10.1121/1.4881321

A NEW REVERBERATION TIME FORMULA


The Journal of the Acoustical Society of America 4, 193 (1933); 10.1121/1.1915600

Analysis of Sabine and Eyring equations and their application to concert hall audience
and chair absorption
The Journal of the Acoustical Society of America 120, 1399 (2006);
10.1121/1.2221392

Sabine Reverberation Equation and Sound Power Calculations


The Journal of the Acoustical Society of America 31, 912 (1959); 10.1121/1.1907816

Overview of geometrical room acoustic modeling techniques


The Journal of the Acoustical Society of America 138, 708 (2015); 10.1121/1.4926438

A MODIFIED FORMULA FOR REVERBERATION


The Journal of the Acoustical Society of America 4, 69 (1932); 10.1121/1.1915588
METHODS OF CALCULATING THE AVERAGE
COEFFICIENT OF SOUND ABSORPTION

By CAR•.F. EYRINO
Brigham YoungUniversity

INTRODUCTION

In a paper,A ModifiedFormulafor Reverberation, G. Millington• pro-


posesa formulawhichshouldbegivenconsideration because of itsvalue
in solvingcertainvery specialacousticalproblems.It will be shownthat
he introducesthe method of usingthe geometric meanrather than the
arithmeticmeanin calculatingthe averagecoefficientof absorption,but
that he doesnot changethe uniqueform of the reverberationtime for-
mula. His paper is valuablesinceit bringsout oncemore the needto
heedthe wordsof Dr. Paul E. Sabine,"In giving the absorptioncoeffi-
cientof any material,onemustspecifycarefullythe test conditions."
This may alsobe said of the paper,A New Reverberation
Formulaby
W. J. Sette•'whichhasappeared
sofar onlyasan abstract.*
When oneplacesabsorbingmaterialin a reverberationchamberand
then on the walls of an auditorium and assumesit to have in both places
the sameeffectivecoefficientof absorptionas worked out by somere-
verberationformula, he tacitly assumesthat the absorbingmaterial
playsan identicalralein bothsituations.
The methodsof averagingthe
coefficientsof absorptionsuggestedby Millington, Sette and Andreea
can be testedout experimentallybest when sometechniqueof unques-
tionable validity is establishedto measuredirectly the exact percentage
of soundenergyabsorbed at eachreflectionfromthe adsorbing material.
It shouldbe clearlykept in mind that the soundchambermethodof
measuringthis percentageis not a direct method.
THEORETICAL CONSIDERATIONS

In the summaryof the paperReverberation


Time in "Dead"Rooms,

the authormadeclearthat any generalreverberationtime formulamust
permit variousmethodsof averagingthe coefficients
of absorption.
Quotingfrom this paper, "We have shownthat a correctpicture of
• G. Millington, J. Acous.Soc.Am. IV, 69-82 (1932).
2 W. J. Sette,J. Acous.Soc.Am. IV, 8 (1932).
* Editorial Note. Mr. Sette'sarticle appearsin this issueof the Journal,p. 193
3 C. A. Andtee, J. Acous.Soc.Am. III, 535, (1932).
• C. F. Eyring, J. Acous.Soc.Am. I, 217-241 (1930).

178
1933] C^•n F. EYRING 179

soundabsorptionin an enclosedroom leads to a generalreverberation


formula of the form
r = s -
The manner of the absorption,which has been describedin detail in this
paper, is responsiblefor the logarithm factor and therefore the new
form; the mean free path between reflectionswhich is independentof
the nature of the absorptioncontrolsand magnitude of k; and a• is ob-
tained by assigningproper weightsto the various elementsof surface.
For a diffuse condition of soundenergy, probably the condition most
frequentlyencountered,k=0.05 and a• is obtainedfrom Eq. (3) (the
weightedarithmetic mean); but for eachorderedconditionof thewavesa
particularvalueof k mustbeusedand a• mustbeobtainedbyproperaverag-
ing." Quotingfrom page220, "We shall,followingSabine'sexperiment-
tal results,assumethat a• definedby Eq. (3) (the weightedarithmetic
mean) may be considered as the uniform absorptioncoefficientof the
walls; yet we realize that for an ordered condition of the soundwaves
this proceduremay need to be modified."
G. Millington concludesthat a formula of the form
r = - o.05v/•s• log,(1 - a•) (2)
shouldbe usedfor certainvery specialcaseswhen an orderedcondition
of the soundwaves is maintained during decay. It will be shownthat
his analysisreally leadsto a specialmethodof averagingthe coefficients
of absorption,but doesnot modify the form of the reverberationtime
equation.
To arrive at Millington's formula one of two idealized assumptions
must be made. In one casethe soundenergywill needto be considered
confinedto a sort of "ray-packet"which travels from one kind of sur-
face to another, the number of times a certain surfaceis struck being
porportional to the ratio of its area to the total area of the enclosure.
This ray-packet will need to remain intact--there is to be no spreading
of the beam, and the soundintensity will be measurednot at somepar-
ticular positionin the enclosuresuchas the location of a microphone,
but in the ray-packet itself. If-we representthe total surfaceby $, its
m parts by S•, S•, ß ß ßS•, and the coefficientof absorptionrespectively
of theseparts as, a•, a• ß ß ßa•, then the chancethat the beam will be
reflectedfrom S•, for example,will be &IS. If E0 representsthe initial
energydensityin the ray-packet(not necessarilythe averagevalue in
the enclosure)then the value after n reflectionswill be
E,• = Eo(R•R•Ra. . . R,•R•R•. . . to n factors) (3)
180 JousNx•.or •ar. Acous•cx•. SOCIETY [January,

where R•= 1-a•, etc. At onceit is clear that the value of E• is a dis-
continuousfunction of the time, its value droppingat eachreflection.
This stepdike decay curve which representsthe true decay might be
ironedout into a smoothcurvewhichwould representthe approximate
nature of the decay,a sort of averageway in which the soundintensity
might be thought to fall off continuously.The smoothedcurve would
be written as

(4)

wheren is a continuousfunctionof the time and R. is a properaverage.


SinceEqs. (3) and (4) mustdescribethe samedecaywe haveat once,
R,, = (R•R•... R•R•R•... to • factors)•/•. (5)
This meansthat R, is a geometric
mean,and from this the averagevalue
a, may be calculated.
If m < <n this value of R, may, with limitations pointed out later,
be usedno matter what the sequence
of the reflections.HenceEq. (3)
may be written as follows
E,, = Eo(R•(S,/S),•R,_S(,/s),•.
. . R•(S•/S),•). (6)

This is essentiallythe sameas the followingequationgivenby Milling-


ton

Et• = E0(1 -- a•)s,,,t/s•,(1-- aa)s,,,t/s•,. . . (1 -- a,•)s,,,,,t/s•,. (7)

In this equationhe tacitly assumesthat t is a continuousfunction.His


equation doesnot give strictly the true decay equation, but rather a
smoothedout curve which may be consideredto fall off continuously.
We are,therefore,justifiedin writingEq. (7) in the form
& = (8)
in which

R, = (Rx(S•/S)"Ra(S,/S)(SdS9
". . . R•(S,•/s),,)•/,•. (9)

If m__>nthe average value of the coefficientof absorptionwhich


shouldbe usedin Eq. (8) to bestsmoothout the discontinuous nature
of the decay would certainly dependupon the sequenceof the reflec-
tions. Even with m < <n limitations still exist as pointed out later.
This meansthat Mr. Millington'sequationleadssimplyto Eq. (1)•
the differencebeing that a• is obtainedfrom a geometricaverage,but
never from an arithmeticaverage.Even for the idealized situation de-
5 Eyring, reference4, Eq. (24), p. 234.
1933] C^a•. F. EYRING 181

scribedthere is, therefore, no reasonto assumethe need of the distinct


form he offersin Eq. (2).
The other picture which may be usedto developMillington's equa-
tion is onein whichthe soundenergyis not confinedto a ray-packetbut
uniformly
distributed
intheroom.
Oneofthem•ypes
ofabsorbing
sur-
face is made to cover completelythe surfaceot•the enclosurefor the
periodof time requiredfor soundto travel the meanfree path between
reflections.Then anothersampleis introducedfor the next period,and
so on duringthe decayof soundsuchthat eachtype of absorbingma-
terial shall have been introducedas many times as its size warrants
under the assumptionthat during n reflectionsa certain material will
be usedq timeswhereq is the ratio of the sizeof the absorbingsurface
in questionto the total areaof the enclosure.The intensityof soundis
arrived at by the samemathematicalprocedureusedabove.The aver-
age value of the coefficientof absorptionis found to be the geometric
meanas before,the only differencebetweenthe two idealizationsbeing
that in the first casethe soundenergydensityis measuredin the ray-
packet,while in the secondcaseit is measuredat any positionin the
room.

Ifodesiredthe methodof images


4may be usedin this secondcase.As
an approximateprocedure,the sourceof soundand the walls of the en-
closuremay be replacedby the sourceof soundand imagesourceslo-
catedin evenlyspacedsphericalzones--surfaces of concentricspheres
separatedby the distancep, the mean free path betweenreflections.
The emissivepower of each sphericalimage sourcewill be constant
over-all, but because of the assumed shift of the nature of the total
surfaceof the roomwith eachreflection,thismagnitudewill changefrom
zoneto zonein a mannerdependentuponthe orderin whichthe various
typesof surfacesare introduced.The energyflowwill be the samefrom
all directions,
sowe sumfor all directionsandobtainthe energydensity
in the enclosurewhen equilibriumis establishedas follows,
Soo= E0(1 q- R• q- RiR• -4-R•R,.R•q- R•R•RaR4'4- ''' ). (10)
As the numberof factorsin the termsincreases, a cyclicrecurrenceof
the valuesof R will be noted.This cyclemay be longor shortdepending
on the numberof typesof materialand the areacoveredby each.Dur-
ing eachcyclethe decaywill dependuponthe orderof appearance of
the typesof materialinto the room,but at the closeof the cyclea prop-
erly weightedgeometricmeanmay replaceeachof the valuesof R to
obtainthe correctenergydensity.During the next cyclethis average
182 JOURNAL
OFTI-IEACOUSTICAL
SOCIETY [January,

value may not be used to give the detailed decay, but at the closeof
the cycle it will again give the correct value. This processis repeated
during the whole decay.At best then the weightedgeometricmean rep-
resentsthe true averageonly at certain points on the decay curve. It is
not certain that Millington recognizedthis limitation in his equation.
He did recognizethe limitations of the idealizedpicture which he pro-
posed.
C. A. Andree• in a very interestingpaper has clearly brought out the
fact that various methodsof averagirlgcoefficientsof reflection must
be resortedto under different arrangementsof materials and different
assumptionsas to the nature of the soundwaves. His method suggests
a meansof obtainingan averagevalue to representthe decayduringthe
cyclesdiscussedabove. Such a value might be usedinstead of the ge-
ometricmeanwhich holdsonly at the closeof eachcycle.But when the
rate of decayof the soundintensitylevelin sucha roomis m•asuredon
a reverberationmeter, which if either of theseaveragevalueswill the
datayield?Theproblem
is discussed'in
a general
wayin anotherpart
of this paper.
In the developmentof the generalreverberationformula,the author4
made useof the method of images.Throughout this discussionan aver-
age coefficientof reflectionwas usedto representthe coefficients
of the
materialsactuallyin the room.It was tacitly assumedthat a certain
type of absorbingmaterial coveredthe whole surfaceof the enclosure.
It was recognizedthen that careshouldbe taken in obtainingthe proper
averageto representthe various types of material. Since that time
Andree• has pointedout the exactingconditionswhich need to be im-
posedif the weightedarithmetic mean of the coefficientsof absorption
is to be usedwithout error. He • concludesthat behindthe usingof this
average in reverberationformulae, "lies the assumptionthat sound
which is reflectedfrom material having a given reflectioncoefficient
hasa probabilityof encountering,
onits nextreflection,
materialof (any)
like coefficientwhichis equalto the ratio of the areaof that material to
the total area present."The author has alsocometo this sameconclu-
sionby an analysisbasedon the methodof images.
Supposethat m types of absorbingmaterialswith areasS•, S•, S,,
etc., cover the surfaceof the enclosure.As before, the walls of the room
may be replacedby a seriesof image sourcesdiscretelylocated on con-
centricspheresspacedat a distancep apart. Becauseof dispersionthe
Reference3, p. 546.
1933] C^m, F. EYRING 183

imageswill be blurredand henceit seemslegitimateto assumespherical


image sourcesinsteadof discreteimageslocatedon concentricspheres.
Even for high frequencieswhere dispersionis least, this assumptionis
legitimate to a first approximationprovidedthe soundis emitted from
a point source.This discussion doesnot take into accountthe resonant
effect of very small rooms.
It should be pointed out now as previously, that the actual energy
densityat a point in an enclosureproducedby a point sourcemay be
dividedinto threeparts; (1) that whichis producedby the sourceonly
and this varies inversely as the squareof the distancefrom the source;
(2) that whichis producedby the first few reflectionimagesources,and
the point to point variation of intensity may be calculatedfor eachpar-
ticular enclosure;(3) that which is producedby the remainingsources
which showsno point to point variation. When the rate of decayof the
soundintensity level is determinedby a meter, it is usually the decay
of the third portionwhichis measured.It is to be understood,then, that
the idealizedpictureproposedis to be appliedonly to this third portion,
the first few image zonesbeing strictly fictitious.
The emissivepowerof eachimagesourcewill not be uniformoverall,
but will be made up of patcheswith different emissivepower. The va-
riety of patchesbecomes very greatasthe imagesourcegetsfarther and
farther from the source.The energyflowinginto the enclosurefrom one
direction,in a smallconeAco,will in generalhave a differentcomposition
than that flowing from another direction. Taking this into account we
write for the total energydensityin the enclosure at equilibriumthe fol-
lowing equation,
Soo= (EoAco/4•r)(1
-3-R• -3-R•R• -3-R•R•R• -3- ''' )
+ (EoAco/4•r)(1
+ R• q- R•Ra q- R•RaR• q- ... )

+ (EoAco/4r)(1+ R,. + R,.R• + R•RaR4+ ... )


(11)
+ (Eo/Xco/4r)(1+ R,• + R,Ra + R,.RaR•+ ... )

+ (EoAco/4r)(1+ R• + R•R,.Ra+ R•RaR4 + ... )


184 Jous•x•.o• •x•r Acous•cxx.Socm• [January,

The problemof averagingnow presentsitself. The secondterm in


eachseriesis a coefficientof reflection,and this coefficientis a common
factorin all the remaining
termsof the givenseries.If the weighted
arithmetic mean,
R,, = (S•R•q- S,R, q- SaRaq- ... S,,R•)/S, (12)
whereS isthetotalsurface of theenclosure,
isusedto replacethesecond
termsof all the series,thenthe pointsourcemustbe considered placed
at thecenterof a sphericalroom.Underthisidealized situation,a given
coefficient,
Raforexample, will appearasa common factorin 4•-Sa/AooS
series,andat oncethe properaveragewill be the weightedarithmetic
mean. A still further idealization must be resorted to if the first factor
in all the othertermsof the seriesis to be replaced by the valueR•.
Froman averagepointof view,the emissive powerof the firstimage
source maybeconsidered uniformover-all.If theenergyfromthisimage
source befocussedto a regionin theenclosure andteemitredequallyin
alldirections,thusplayingtheidentical r61eoftheoriginalsource except
for the reductionin magnitude by the factorR•, thenR• mayreplace
thefirstfactorsin all thetermsof theseries. At once,aftera rearrange-
mentof the series,Eq. (11) takesthe form,
s• = (•0a•/4•){• + •(• + • + •, +... )}
+ (E0•/4r) {1 + R•(1+ R•+ R•Ra+... )} (13)

+ (•0a•/4•){• + •(• + • + •• +... )}


The factorsmultipliedby R• haveexactlythe formof the original
series
ofEq. (11),andtheassumptionandargument usedto obtain(13)
from(11)mayagainbeused.Asa result,aftera rearrangement of the
serieswe get,
s• = (••/4•) {•+ • [• + n• (-)]} + • + n•n,+... )]}

ß (14)

+ (z0a•/4=){• + •[• + •(• + •. + •.• + ... )]}


The process
maynowbe repeatedasmanytimesasneededuntil the
averagevalueR. only appearsin the seriesandall the 4•r//X•0seriesbe-
comeidentical.Summing,we get
S• -- E0(1 q- R, q- R, •-q- ... R, • q- ... ), (15)
1933] C^•. F. EYP.ING 185

the equationusedin the formerpaper in the developmentof the rever-


beration formula.
It is important to clearly recognizethe physical requirementsback
of this equation. The simple picture used in the above development
tacitly assumesa point sourceat the centerof a sphericalroom and the
focussingof the energyremainingafter each reflectioninto the equiva-
lent of a point source.It is further assumedthat the new sourceemits
energyidenticallylike theoriginalsource,that is, equallyin all directions.
This means that some sort of mixing mechanismmust be postulated
because,due to the nonuniformabsorptionof the wall surfaces,the en-
ergyflowingin certaindirectionsinto the regionof focuswill needto be
augmented,and that comingfrom other directionswill need to be de-
creased.But if by the useof the mixing mechanism,the energyfocussed
from a given elementof area be reemitted equally in all directions,and
if this be true of the energyfrom all elementsof surface,then certainly,
exceptin magnitude,the new sourceand the original sourcewill be-
comeidentical.Thus the requirementsbackof (15) will have beenmet,
and the value R, may be usedwithout error. At once I may conclude
with Andree that behind the rigoroususe of the arithmetic averagein
reverberationformulae "lies the assumptionthat sound which is re-
flected from material having a given reflectioncoefficienthas a proba-
bility of encountering,on its next reflection,material of like coe.l•cient
which is equal to the ratio of the area of that material to the total area
present."
By analogy one would expect,therefore,that if a diffusestate were
originally achieved,a state which would permit the accurateuseof R,
for the first reflect{on•thentheexactnatureof sucha statewouldneed
to bemaintainedor reestablishedaftereachreflectionin order to rigorously
developEq. (15).
The art and techniqueof developinga diffusestate and the main-
taining of it is at oncebeforethe experimenter.The error involved in
the assumptionof a diffusestate when this state is only approximated
is anotherproblemworthy of investigation.Experimentaldata point to
the conclusionthat with a carefuldistribution of absorbingmaterial and
the use of diffusingpanelsand paddles,a strictly diffusestate may be
at least approximatelyestablished.
We now come to the problem of ascertainingthe physical require-
mentsimposedif Eq. (11) is to be averagedby the useof the weighted
geometricmean.If a cyclicrecurrenceof the valuesof R is foundin the
large terms of Eq. (11), then eachof theseterms will take the form of
186 JournAl. Or THEACOUSTICAL
SOCIETY [January,

Eq. (10), and the geometricmeanmay replacethe variousvaluesof R


underthe limitationspointedout above.But if the large termsof Eq.
(11) are to be addedin orderthat an equationof the form of (15) may
be obtained, then not only must there be a cyclic recurrenceof the val-
ues of R in each large term, but, if the recurrencecomeswith the m'th
term in one of the large terms, it must comeat the m'th term in all the
other large terms. Further, these m'th terms, on the rearrangementof
the factorswhich in generalwill be necessary,must be identical. As dis-
cussedabove,unlessthis cyclicrecurrencecomesa very largenumberof
times during the processof decay, the geometricmean can not be used
accuratelyeven under the exactingconditionsjust pointed out. This is
especiallytrue if it is to be usedas rigorouslyas the arithmetic mean is
used under the limitations discussed above.
I am unable even to suggesthow sucha generalstrictly orderedcon-
dition of the soundwavesrequiredfor the useof the weightedgeometric
mean of all the absorbingmaterial in an enclosurecould be produced
by a practical mechanism.With high frequenciesand very directional
sources,the sound energy might be approximately controlled so as to
remain in a few strictly orderedstates.For example,the soundcouldbe
made to move back and forth betweentwo parallel walls, or diagonally
about the room, each wall beingstruckin proper sequenceand the cor-
ners of the room being carefully avoided. However, if the room is to be
filled with energy and if all the walls are to be used, three distinct
orderedconditions,a back and forth motion betweenall oppositewall,
wouldneedto be established.Thus, aspointedout by Andree,a combi-
nation of a geometricand arithmetic mean would need to be used.If an
attempt were made to fill the enclosurewith energyby sendinga beam
of sound diagonally about the room, the ordered condition would be
vitiated at the corners.
All this meansthat in generalEq. (11) may not be averagedby a
simplemethod.In a roughgeneralmannerthe large termsof this equa-
tion may be arrangedinto three groups'terms which very nearly fulfill
the conditionsfor the strictly diffusestate; terms which very nearly ful-
fill the conditionsfor an orderedstate, and this groupmay have a num-
ber of subclasses;terms which cannot be classifiedwith either of the
other groups.In averagingthe terms of the first group the weighted
arithmetic mean may be usedwith little error. In the secondgroupeach
subclasswill have its geometricmean and the weightedarithmetic mean
of thesewill be taken. For the third group,representingthe statewhich
is neither diffusenor ordered,I am suggestingthe useof an empirically
1933] Cx•. F. EYRING 187

weightedarithmeticmean.(SeeEq. (17).) Of coursesomeotheraverage


couldbe experimentallytestedand empiricallyestablished.The experi-
mental aspectof this problemis discussed
in the next part of the paper.
EXPERIMENTAL CONSIDERATIONS

In certain experimentsthe author has found situations7 in which it


seemedlogical to assumepart of the soundenergyin a diffusestate and
part in an orderedstate. For high-frequencytonesemitted by an ex-
ponentialhorn sucha situationgavedecaycurvesof the soundintensity
level with two regionsof distinct averageslope.The first averageslope,
which representsin the main the decayof the original approximately
diffusesound,after a drop of 18 db changerather quickly to the second
averageslopewhich represents,in part at least, the decayof the sound
which is reflectedback and forth between two hard surfacedparallel
walls. But a "beam of sound"travelling back and forth betweenthe
parallel walls without lossexcept at the reflectingsurfaceswould have
had a reverberation time of eight seconds. Yet, at 1000 cycles the
longestreverberationtime measuredwas three seconds.This meansthat
considerableenergy left the beam of soundand was absorbedby the
highly absorbingceilingand floor. This indicates,I think, that the beam
of soundis essentiallya slowlydivergingcone,and roughlyspeaking,a
certainportionof the energyis passinginto an approximatediffusestate
from the more or less ordered condition of the sound waves.
Another interestingaspectof these experimentsis that the average
effective coefficientof absorptionas calculatedby the usual formula
changedfrom 0.28 to 0.08 during the processof decay.If the ear method
had been used a constant average value of 0.11 would have been re-
ported. Placing diffusingpanelson the hard surfacewalls eliminated the
secondand slow rate of decay of the soundintensity level. Under this
diffuseconditionthe measuredaveragecoefficientof absorption,0.28,
checkedcloselywith that calculatedusing reverberation chamberdata
and the assumptionthat the averagevalueshouldbe the weightedarith-
metic mean of the various coefficients. Had the ear method been used the
changingrate of decay would not have been detectedand someother
type of averagingwould probably have been resortedto in order to ob-
tain as low an averagevalue as 0.11.
In anothersituationaI foundthat an openwindowapproximatingthe
sizeof a wall surfaceand large as comparedto the wave-lengthcould
? C. F. Eyring, J. Sac. Motion Picture Eng. XV• 528-548 (1930).
8 C. F. Eyring, J. Acaus.Sac. Am. III• 188 (1931).
188 JOUI•NA•,
OFTar ACOUSTICAL
SocmT¾ [January,

not be consideredto have a coefficientof absorptionof unity in the


applicationof the reverberationtime formulafor a diffusestate. "Since
the reverberationtime equation
T = O.05V/- S loge (1 - aa) (16)

is built up on the assumptionthat a diffusestate is establishedin the


enclosure,a secondfactor, the lack of a diffusestate, which getsworse
as the openingbecomeslarger can accountfor the calculateddecreaseof
the coefficientas the openingarea increases.The method of calculation
fails to give a coefficientof absorptionof unity to the openwindowbe-
causethe mean free path between reflectionsin this casediffers from
that established in an enclosure of the same size in which a diffuse state
is reached. It fails also because the lack of a diffuse state makes it im-
possiblefor all elementsof surfaceto have equal chancesof registering
on the soundwaves their absorbingability. The mode of wave travel
betweenthe parallel wallshas a greater chanceto survivethan a mode
whichquickly takesthe wave out through an openwindow.This means
that a unit of surfaceon one of the parallel walls has a greater chance
to registerits absorbingability on the soundwavesthan a unit of sur-
faceof the openwindow.This indicatesat oncethat if Eq. (16) is to be
usedto calculatethe reverberationtime for this type of room and if a
coefficientof absorptionof unity is assignedto the open window then
equal weightscannot be assignedto all elementsof surfacein the cal-
culationof the averagecoefficientof absorption."
The more or less ordered or semi-diffuse condition of the sound waves
in the two situationsjust describedis producedin the first casebecause
the soundsourceis directionalfor high frequencies,in the secondcase
becausethe small room is not a true enclosure,and in both casesbecause
of a concentrationof the absorbingmaterial. If a large area of very
highlyabsorbing materialapproximatingthe sizeof a wall surface,were
placed in a reverberationchamber, the room could hardly be called a
true enclosure,and a conditionof the soundwaveswould be produced
which would be neither strictly diffusenor strictly ordered.A concen-
tration of absorbingmaterial of any kind on a singlewall wouldvitiate
the accuracyof resultsobtainedusingthe weightedarithmetic mean of
the coefficientsof absorption.Andrees has clearly pointedout this fact
and hasrightly given warningthat a diffusestate shouldnot be assumed
as existing ameng the soundwaves without good evidence.It is the
opinionof the author, however,that the averagecoefficientof absorp-
tion can be obtained accurately from a reverberationchamberwith
1933] CAm, F. E•m• 189

certainty only if the various formulae proposedfor averaging can be


tested by a techniquewhich permits thedirectmeasurement of the per-
centageof energyreflectedat each encounterwith a given type of ab-
sorbing material.
From all considerationsit is clear that in order to include all possible
casesone might in a generalway divide the soundenergyof an enclosure
roughlyinto three parts' first, the part which is strictly diffuse;second,
the part whichis not strictly diffusenor strictly ordered;third, the part
which is strictly ordered.Let A, B, and C representthefractional part
of the total energyexistingin thesethree statesrespectively,then we
may write,

1--ao=R,= q-•q- ß (17)

The valuesof k will usuallycorrespondto the variousareasof absorbing


material, but becauseof "edgeeffects"this strict correspondence.may
not hold. The valuesof q will dependuponthe type of soundsource,the
room configurationand the concentrationof absorbingmaterials. Only
after a longseriesof experimentswith all typesof soundsourcesand en-
closuresand all possiblelocationsof absorbingmaterial can the tech-
nique of determiningeven approximatevaluesof q be established.If
one can be sure that a portion of the soundenergy movesabout the
room as ray-packets,then the third term must be included.At oncewe
have for this term,
r = sq-tq-pq-.--v. (18)
Someof theseexponentsmay be zero if it is found that certain of the
surfacesare not used; and the weight d of each ray packet and the
weight of each reflectingsurfacefor each ray packet must be deter-
mined empirically.
In order to explain the experimentalfacts reported above for the
small auditoriumit must be possibleundercertainsituationsfor A, B
and C to be variables, functions of the time. In the caseof the small
auditorium,energywas continuouslyleavingthe roughlyorderedstate
and going into at least a semi-diffusedstate. By introducingthe dif-
fusingpanelsthe coefficients B and C weremadeto vanish.In the small
room with the openwindowreferredto above,the coefficientsA and C
were probably very small. Millington thinks that in a small room of
volume1933 cu. ft. and a total surfaceof 932 sq. ft. and absorbingma-
terial with an areaof 27.8 sq. ft. he wasable to set up a conditionsuch
190 JOUSNAI,
OFTHEACOUSTICAL
SOCIETY [January,

that no strictly diffuseor semi-diffusesoundexisted,and thereforethat


A and B were zero. Of course he could have obtained the smaller de-
sired coefficientof absorptionby taking into account"edge effects,"
thus usingonly the first term of the equation,or he couldhave usedthe
secondterm only, adjusting the valuesof q to give the desiredcoeffi-
cient. But doeseither of these procedures,or even the one he used,
representthe reality of the situation? How is one to determine?
In the discussion so far it has been assumed that the values of R are
accuratelyknown. Usually thesevaluesare obtainedin a reverberation
chamber.But when a personplacesabsorbingmaterial in suchan enclo-
sure and then on the walls of an auditorium and assumes it to have in
both placesthe samereflectionor absorptioncoefficientas workedout
by somereverberationformula, he tacitly assumesthat the absorbing
material plays an identical r61ein both situations.If a personcan be
surethat the soundin the chamberis perfectlydiffuseand remainsso
duringthe decay,he may usethe reverberationtime Eq. (1) to calculate
the absorptioncoefficientprovidedthis formulausesthe meanfreepath
calculatedfor a diffusestate, and an averagecoefficientof absorption
obtainedfrom an arithmeticmean.If there are no edgeeffectsthen the
calculatedvalue of the coefficientis probablythe true absorptioncoeffi-
cient of the material, yet there is still somelittle uncertaintywhether
the reverberationtime equation will give the exact coefficientof ab-
sorption,that is, the exactpercentageof soundenergyabsorbedat each
reflectionfrom the absorbingmaterial. If a techniqueof unquestionable
validity couldbe establishedfor the measurementof this percentageof
absorption under the situation found in the reverberation chamber,
then this last uncertainty would be eliminated.
Any kind of reverberationformula used to calculate an absorption
coefficientfrom a reverberationtime, may be usedunder the identical
circumstancesto calculatethe reverberationtime from the absorption
data. In a word, it is simply the processof workingforward througha
mathematicalprocedureand then retracingthe exact stepsto the orig-
inal data. Therefore, if the identical situation found in a chamber is also
foundin an auditoriumthen any kind of formulacouldbe usedwith
success even thoughit describeda fictitiousphysicalstate. It is easyto
see,therefore,if oneshouldusethe geometric meansuggested by Milling-
ton in determiningthe coefficientof absorptiontor a diffusestate in a
reverberationchamber,and then use the samemethodof averagingin
calculatingthe reverberationtime in a rather live auditorium in which
the soundis alsodiffuse,that the final resultswould not differ greatly
1933] C^RL F. EYRING 191

from thoseobtained using the true averagevalue for a diffusestate, the


arithmeticmean. Again, supposeabsorptioncoefficientsare obtained in
a reverberation chamber where a diffuse condition of the sound waves
is maintained, and supposethat a diffusestate is also set up in a live
auditorium, a reverberationformula which very nearly but not exactly
describesthe true physicalstate couldgive excellentresultsin spiteof its
inaccuraciesbecauseof the almostidentical r61esplayed by the absorb-
ing material in the two situations. The need for the new reverberation
time Eq. (1) becameapparentonly whenthe situationin the reverbera-
tion chamber and that in the soundstagewere far from being identical
--.one room was very live, the other was very dead. This brings out
clearly that the checkingof measurementsmade in a reverberation
chamber with those made in an auditorium cannot test with precision
the validity of a reverberationtime equation.It would be best to check
it experimentally against coefficientsof absorption which have been
measuredby somedirect process.
In practice most reverberation chamber and auditorium situations
are not identical. For this reason a reverberation time formula which
truly describesthe physicalreality is muchto be desired.In otherwords,
a true not a fictitiousrelationshipbetweencoefficients of absorptionand
reverberation time should be determined for as many different situa-
tions as possible.We have a formula which describesvery well, I think,
a diffusestate, and if precautionsare taken in a reverberationchamber
to obtain a diffusestate and if this formula is used, then the absorption
coefficientsmeasuredwill be approximatelycorrect,that is, they will
quite truly representthe percentageof the energyabsorbedat eachre-
flectionfrom the surfacein question.
If "the durationof audibilityof the residualsou'udis nearlythe same
in all parts of an auditorium,"if it is "nearly independentof the position
of the source,"and if "the efficiencyof an absorbentin reducingthe
residualsoundis, underordinary circumstances, nearly independentof
its position,"ø the soundwill be at least approximatelydiffuseand we
shallhave goodacousticconditions.Thus the regularreverberationtime
formula for a diffusestate may be usedwith little error. However,in
certaincases,this approximate"ideal" conditionmay not be obtained.
If so,onemustuseempiricallysomesuchequaffonas (16) to determine
the averagecoefficientof absorption,and determineempiricalIf ø the

WallaceC. Sabine,Collected
Papers,pp. 18, 104.
V. O. Knudsen,J. Acous.Soc.Am. III, 314 (1932).
192 JOVSNA•,
O• •r, ACOVS•XCA•,
Socxr,•¾ [January,

valueof k (a valuewhichdependsuponthe meanfreepath betweenre-


flections)which shouldbe usedin Eq. (1).
Millington doesnot give data to provethat edgeeffectsor that a semi-
diffusestate could not explain his results. In other words, he doesnot
showthat he would not have beenjustifiedin weightingthe surfacesin
sucha manneras to get an arithmeticmeanof the smallermagnitudehe
desired.I am not willing to assumewith him that in my experimentsthe
unique state of the soundwavesrequiredby his geometricmeanexisted
in fact. It may be that the soundwasnot in a perfect diffusestate, but
I am convincedthat it wasapproximatelyin that state. Millington does
do this, he setsthe minimum for my values,but the true value I am sure
is much nearer the maximum which I give than the minimum which he
calculates.My explanationof obtainingcoefficients greaterthan unity,
still holds. The difficulty of obtainingcorrectcoefficientswith high
precisionin a highly reverberentroom is recognizedby all who have
made suchmeasurements.Certainly I may claim observationalerror in
measuringcoefficientsof absorptionin a very deadreverberationcham-
ber having a reverberationtime of 0.34 second!
In conclusion I wishto emphasizethat improvementin reverberation
chambertechniqueshouldbe in the directionof givinggreatercareto
the establishment
of a diffusestateamongthe soundwaves,thusgiving
rise to the use of the weightedarithmetic mean of the coefficientsof
absorptionand a valueof the meanfreepath betweenreflections which
is backedby well establishedtheory.If it is foundimpossible
undercer-
tain circumstances to establish this ideal state--this situation would be
encountered more often in auditoriums than in sound chambers--then
the methodsof averagingsuggested by Millington, Setteeand Andree,
when testedexperimentally,will be of great value.
Acknowledgment is extendedto Mr. ClarenceA. Andreewhosesug-
gestions
havestimulatedthe development
of certainportionsof this
paper.

You might also like