Sound Transmission Through A Double-Leaf Wall: Received 6 July 1971 2.9

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Received 6 July 1971

2.9

Sound Transmission through a Double-Leaf Wall

R. J. DotqAXO
Building Physics Section, Division of Building Research, National Research Cormoil of Canada, Ottawa, Canada

The classical approach to the calculation of sound transmission loss through double-leaf wallsis reformulatedand various devices employed to reduce the resulting integrals to a tractable formfor analytical

solution. An approximation is made at lowfrequencies whereby, using a transform technique, thefinite size ofthewallmaybeincluded. Comparisons withexperimental measurements show good agreement with
theoretical predictions.

and Ss is the energyspectral density.White* assumes that R,'=R/, which allows R/S'--R,S, to be There are essentially two methods for calculating written as R/(S,'--S/). On the other hand Eichler the soundtransmission lossthrougha double-leaf wall. writes the equationof power flow betweenthe two The classical approach chosen by London t andBeranek systems as and Work2 usesthe easily derivedformulasfor transmission at a fixedangleof incidence andthensums over all possible incident angles.These authors found the whereV is velocity.If we take systems of equalmodal resulting integrals to be too intractable for analytical densit)', then becanse,7,--nv/, where r, no are solution and computed the transmission loss numeri- the modal densities(in this caseequal), there is no cally. The other method is the statistical mechanical power flow when S=S, ' or M',V?=M,V, , whereas approach adoptedby White and Powell, s who use the White and Powell say there is no power flow when ideasof Lyon and Maidanik 4 in considering energy V?=V, . Further, if /=,', then flows between two or more coupledmultiresonant Clearly, then, the relationship between systems as an analogof a heat transmission problem; must include the modal masses as well as modal the analog of temperature is the modalenergy. This densities asn,,'=nl/, where n is themodal density2 .a techniquehas the advantageof considering finite This possibly accounts for the absence of any term radiators andvibrators, whereas theclassical approach representingthe panel spacingin the derivation of White and Powell. a deals with infiniteareacomponents. From a practicalpoint of view, the requirement is Even without this objectionone still has to make for an analysisthat will showthe contributions of the some guesses as to the value of radiation resistanceto various parameters to the transmission loss. For use. For the externally radiated power one may use instance, the effectof wall spacing is not readilyseen Maidanik's expression, 8 but there is still doubt about from the analysis of White and Powell.This is partly the correct value to use for energy radiated into the dueto a doubtfulfeaturein the analysis. Theseauthors cavity. White and Powell use a value equal to twice take the power flow betweensystems r and s to be the externalradiationimpedanc% but this is probably R/(S,--S,u), whereR, is the nondissipative (gyro- only true below coincidence. scopic) coupling resistance between the systems, and A later paper by Price and Crocker applies these Sv is the velocityspectral density.This is at variance techniques to the double-wall problem? Althoughthe with Lyon and Scharton ' and Eichler, who take the resultsare impressive, there still remaina few peculiarpower flow to be proportional to ities. While the resultsagreewell with our own experimental ones for a double wall having an absorbing (n,'S, -,'S,), layer,theresults do not agree with those for an absorpwhere tion-freespace,the discrepancies beingaround 15-20 n,'= R.'/M,, dB. Further, theseauthors showthe effectof varying
The Journalof the Acoustical ociety of America 807

INTRODUCTION

aded 23 Dec 2010 to 200.23.51.1. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/te

R.

J.

DONATO

Uo/2p---[-Zq-Zx-3-Z2(Z-3-Z)/(Z,-kZ2q-Z,)'] -1, (1) u/2p = (uo/2p )[-Z2/(Z2+ Z+ Zp),


(2)

2p

whereu0and u represent the velocityof the two leaves, and p is the incidentsoundpressure.

Fro. 1. Model usedin the analysis:Z=jpc tan0; Z=jpc/sin20; O= &L/2c) cos,.

The impedances Z, Z2 givenin Fig. 1 should strictly be multiplied by (cos) - to conform to the subsequent use of impedances. However, this createsproblems. It is well known that the range of incident angles presentin the incidentfield must not spreadover a full quadrant, for thenthe transmission loss is calculated
to be lower than the measured values. The effect is worsened with a double wall because there is an extra

the wall space from 1 to 40 cm to be negligible--aresult certainlyat variancewith usualexperience. There still
remains the doubt as to the amount of the contribution from the nonresonant modes. Price and Crocker take

it to be basically that given by applying the masslaw


to the two leaves.It is difficult to seehow this can be so,

expecially at frequencies just belowcoincidence. For a

very largewall, the resonant modeat sucha frequency will be very sharply peaked in the k domain, the radiation impedancevery low, and the radiation given predominantlyby the nonresonant modes.Here the modeswill also be narrow but more densely packed. Then, on the above argument,we shouldexpectthe transmission to be given by the masslaw. However, classical theory shows that at this frequency there Upon simplification we find will be the well-knowndeparturesfrom the masslaw given by a 1--(f/f)2 siniqv term in the denominator u joc .... (Z cosO+ joc sin0)-(Z, sinO-joc cos0) -. (3) of the transmission lossequation? 2p 2 The following methodrevertsto classical procedures, but simplifies the integralsobtainedso that they may The plate impedance term Zv=jcoM-l-oc/cos% where be evaluated analytically. We have attempted to M is the massof the wall per unit area and pc is the compensate for the finite size of the wall at low frecharacteristic impedance of the ambientmedium. quencies by usinga wavenumber approach. This is at Substituting for Zp in Eq. 3 gives best an approximation, for at the low frequencies the mode spacing is too wide for either the statistical
method or the infinite-area method to be used. One

sourceof angularlimitation. Putting in the (cos)- factor givesresultshighly dependent on the limiting angle chosen, which of coursewill in turn dependon on the wall size and spacing.Any modificationis to a certainextentarbitrary, but in the present analysis we have chosento take the free-fieldimpedance of the cavity to bemultipliedby cos. The modification affects the subsequentintegrand in such a way that the resulting integration carried out between the limits 0 and *r/2 gives the same result as the more rigorous expression integratedbetween0 and an anglelessthan

should sum the individual modes, which would make both methodsrather unwieldy.
I. THEORETICAL THROUGH ANALYSIS OF TRANSMISSION VrALL

2p

u =--JjcoM cos0+-+joc sin0 pc sin0


cos

tomsin0-3

A DOUBLE

oc cos 0
cos p

jc cos0

(4)

It is interestingto note that equatingthe imaginary part of the secondterm in the denominatorto zero gives both the zero-orderresonance when the masses area wall are discussed in Sec. IV. In the analysisthe and the air cavity resonateas a lumped system,and angle of incidence , of the incomingsoundis usedas the higher-order modeswhen path differences in the the variable although the use of k, the wavenumber air spaceare multiplesof a wavelength. along the wall, would be equally acceptable. If the spacingL betweenthe two leavesof the wall
Let us assume first of all that the wall is infinite in area. The modifications needed to correct for a finite

in Fig. 1. Zx is the seriesarm of the network and Z2 the parallel. The impedances of the two leavesof the wall are assumed to be equaland denotedby Z, where
this term for the moment is taken to include the

The air space is represented by a .Tnetwork asshown is zero,Eq. 4 reduces to


(u,/2p)c:o: cos,.(jwM cosv+c) -, (5)

which is the well-known form for a wall of surface

radiation impedance of each leaf as well as a reactive density 2M. component. It may be shownthat the second term in the denomFrom the equivalentcircuit of Fig. 1, inator is the predominant factor in determining the
808 Volume51 Number3 (Part I) 1972

aded 23 Dec 2010 to 200.23.51.1. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/te

SOUND

TRANSMISSION

THROUGH

DOUBLE-LEAF

WALL

maximum amplitude of u/2p. Approximating 0 for sint,

then

2p

r=2D 2 ua --jac[(joM-{c q jpccoLcosl


cos 2c /
2c

xdx(A2x+l)-l[-(Ax-D)+l] -
2D

.t O

(D+4)

/Lx -1

;i [ x_2/AD x--2(D'q-1)/ADqd x
When the integration of Eq. 11 is performed we obtain

-jocl}

X2c/

cosq-jac+c}l
LX COS/X wL /J

. (6)

r=(D+ 4) 2A
2A

Thus, we my write for Eq. 6

2D { 1 AD 2 1ln[(A-Fi_[D 2(D+l)
_ lnA_-- tan-tA

L D+I

X[tan-(A--D)+ta (12)
and for large A this approaches
D

The radiationimpedance associated with the second leaf of the doublewall is pc/cosyper unit zre, and so for n incidentdiffuse sound field the radiatedpowerW
from this unit rez becomes

+2(D+)(;+tan-' (12')
As previously stated,the lowerlimit of coseusedin the determination of Eq. 12 hasbeentaken to be zero. The integrandhas a maximum amplitudeat = qm fixedby the second term in the denominator. Typically, for the lower frequencies this term peaks at around
60 to 70 from the vertical, and the lower limit of the

[cos

integration is of no great concern. However,whenthe In the extreme caseL=0, M=0, frequency is raised, v.... approaches 90 and the value of this limit assumes greaterimportance if the rigorous W0=cospsindq =, (9) formulais used.By modifyingthe free-fieldimpedance pC a 0 pC in the cavity by the cosq factor, it is found that this dependence is removedand the resultsobtainedare whichis the powertransmitted in the bsence o[ the closeto thoseobserved experimentally. wMl. Thus,ff is the transmission fctor o[ the The solutiongivenby Eq. 12 is valid only as longas 0 approximates sin&Thisin turndepends uponthevalue of D. WhenD becomes equalto unity we assume that the second term in the denominator of Eq. 6 no longer exhibitsits peak. This is more or lessjustifiedby the previous discussion, where we showed cosq ..... to approachzero progressively. In the regionDI, we

(8)

2p f.l

/L -f'/

F/cos -
+U '
i

;;

calculate

the transmission

loss to be

As check,we seethat for L 0 Eq. 10 reduces to

3)

f'/

F/2M cos

r=2J0 cossincdd{ LX 2pc t / +1/ J (10') chosen here is when D = 1. Around this point, it may be
To simplify the integrationinvolved in Eq. 10, we
substitute

The point wherewe change overfrom oneformulato the other is arbitrary to a certainextent; the criterion
shownthat the two formulasagree.

For frequencies below and approachingthe coin-

A=wM/pc,

D=2c/wL,

x=cos;

cidence frequency f, M is replaced by M[1--(f/f)g Xsinq.TM Normally this correction is only appliedat
The Journalof the AcousticalSocietyof America 809

aded 23 Dec 2010 to 200.23.51.1. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/te

R.

J.

DONATO

The peakof the integrandnowoccurs near oneof the limits of integration, and so we must modify the II. FREQUENCIES ABOVE COINCIDENCE procedure of Sec.II. In fact, we restore theZ, Z2 terms For thefrequencies above coincidence (fc), weneglect removed from the derivation of Eq. 16 and obtain Z, Z2 in the denominator of Eqs. 1 and 2. We replace tom by coM{-1--(f/fc) 2sin4)] and pc by X (f/fc)2 sin4cosq.These modifications arisefromthe lower impedanceowing to waves traveling along the plate where is the dampingfactor. Thus Eq. 3 becomes

the two third-octavesnearest to coincidence, and for this rangewe take sinq'l.

III. COINCIDENCE FREQUENCY

2p/

u/2p = --jZo(2 sinO cosOZ) -x,


with

(14)

[./2coM col XlA 2c cos,+(l__ / x pc coso


XLJ[, pc
coL/ \ pc

Z:

+M sin cos Xfd

(A

cos . (19)

Over the rangeof integration wherethe integrand peaks, (coL/2c) cos 2<< (2coM/tc) cos a
and so will be neglected. Thus the transmission factor r is given by

and with Zo=pC.


is modified to

+jM[1--(5 )sin

When the necessary substitutions are made, Eq. 10

r=8[-)

w 2Jocosa sind(sin cos) -a

k 6 /

d0

tLX

p6

p6

A fairly good approximation may be made to Eq. 20 by writing the integrandin the form ax exp(--x), which is immediatelyintegrableif is large enoughfor f us to write the limitsas0 and m. We fit the approximate The integral in Eq. 15 clearly has a maximumwhen curve at the maximum of the integrand. If we write sin4q0 = (f/f)L If we expand the denominator in the A =2wM/pc, B=M/9c, and if we assumethat the

1M (fsin,cos]

F/2M 2c5 / wM 5q}-.

+ cosy 1-sin

. (15)

integralaroundthe angleq0,Eq. 15 reduces to


r= cosaq0 sinqod'(cos0 sin0) -:

maximum

is not far distant

from

the maximum

of

x/(l+Bx) 4, we find the maximumxo to occurat x0 (3B)-L The maximumof x exp(--x ) occurs when 4x04=1, and so= (3B)4/4. Now

LX

pc

a-'e=L(+(1.33):JL[ a) +(1.33):] (21)


2sin40cos0,

, F A

F/2A

2c

For the particular caseshownin Fig. 2, a = 1/2.9 and + x/ cs4sin%[--} (')l' (16)
If we make the substitutions

A = 1 (M/pc)(f/f)

r=8[-)

/wLx - 2exp(--2.1XlO4x4)dx x
r=
\5.8/\ c / \4.2/

n=
Eq. 16 reduces to

sbo(fVf?),

( =k/k/ exp(--2'lXlO
(17)
Thus

r = 4, cos*o sino(AaB) -,

where we have replacedcos sin20by its mean value

Substituting backagainfor A, B, 0,we finally

X 10 -2.

(22)

obtain

f/kpc
810 Volume51

f/

pc k
1972

With our particular value of coL/c=2.8 (coincidence frequencyis 1600 Hz, L=3 in., M=4.6 lb/ft , and =0.01), this reduces to - 10 log0r = 28.2dB.

Number3 (Part I)

aded 23 Dec 2010 to 200.23.51.1. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/te

SOU'ND

TRANSMISSION
0.02

THROUGH

DOUBLE-LEAF

WALL

Fro. 2. Curve fit around coincidence.

00.01
o

(------): theoretical curve, Eq. 20; (- - -): x exp(--/x ) approximation.


z

10

//// /
TO

I
0.05 0.1

I
O. 5

IV.

CORRECTION

TO

BE APPLIED

assumethat the correlationfunction W is dependent

LOW-FREQUENCY CALCULATIONS
wall surface could be considered to be infinite in extent.

We assumed at the beginning of the analysis that the simplerform,

solelyupon (r--r); in otherwordswe have circular symmetry.Equation23 can now be expressed in a

In practical situationsthis conditionis of course violated, especiallyfor the low frequencies. To see what kind of correctionis necessary to allow for a Xh(r,r)dradr, (24) finite wall, we shall restate the problem using the spatialFourier representation. It is well known that, whereh(r,r) is the impulse response of the wall, and using a statistical representation for the incident R(co;Ir-r[) is the incident sound-field correlation soundfield, TM the mean amplitudepower spectrumof function. Then the vibratingwall surface is givenby

<(,0; r)> =fIP01 g(o; I--rl )h(r,r)

s fsiw(; r,r2)
X ,,(r),n,(r2)drdr2, (23)

r)) =[pal 2fF(.,; k)[H(co; k) lkdk2r, (25)


where F is the Hankel transform in the k domain

of R, and H is that of h. We can replace H(co;k) by k) -l, thewall displacement impedance. where//is the vibrationamplitude at radianfrequency [-Z(o; Up to nowwe havesaidnothing aboutthe formof co and point r. r-r:). For an infinitearea this term will be the Wv(co; r,r2) is the pressure correlation at obetween R(co; pointsr, r on the wall surface. The quantity q,, is autocorrelation of the incident sound field. In a rearea,however, a betterapproximation wouldbe an orthogonalcharacteristic function satisfyingthe stricted boundaryconditions at the edgesof the wall and given by R(w;ri--r).(1--r2/4R), where circular symmetry is maintained and the area of the wall is fulfillingS{,,,n 2)= 1. S is the areaof the wall and X, is the modalimpedance of the wall. Two subscripts m, given by rR=S, and r= Iri--rg[. To simplifythe we find the Hankel transformof the n are usedto satisfy the generalcaseof a two-dimen- calculations, sional vibrator. modulating function(1-r/4R ) and then use this If we now consider the practicalcaseof an 80-ft2 transform to convalve the infinite area solution. Thus, k-domain spectrum. Finally, wall, we find that there are around0.32 modes per unit weendup with a modified this modified spectrum between 0 and bandwidth, and in a third-octaveband centeredon we integrate of sound in air) to 125 Hz there will be about 5 modes.The number per k (k=co/c,wherec is velocity
third-octave band will increase proportionally to the find the fraction of the incident energy being radiated.
The Hankel transform

frequency so that at 1250Hz there will be 50 modes. We shall,however, assume that the numberof modes is alwayssufficiently numerous that we may replacethe summation by an integrationprocess. We shallfurther

I(k) = ff( --,'2/4R)Jo '


The Journalof the AcousticalSocietyof America 811

aded 23 Dec 2010 to 200.23.51.1. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/te

R.

J.

DONATO

--

--

--

(oJR/2.3c)
Fro. 3. Low4requency corrections to transmission loss.

canbe shown to be givenby


and

the actualwall area or--whichis possibly a better the radius of the largest circlewe can I(k)=J(kR)[(kaR)-+(R/k)]--(2k2)-Jo(kR ) (26) procedurefind
fit into the rectangular area.
I(0) = 3R2/8.

We mustnow convolve I(k) with the k spectrum derived for the infinite areacase. The valueof F(o;k) for the infiniteareacase is givenby F (o ; k) = [-k(k--k2)i] -, three-dimensionally diffuse with theresulting
R(o;r)= (sinkr)/kr.

As a further simplification we shall find the effective

bandwidth of I(k) and use a rectangular wave I(0) for the convolution process. Now

(28)

where we have taken the incident sound field to be

L k3R Jo(kR) 2k 3RJffkR)lkd -_1k foi(k)kdk=fj[J(kR)


----IJffkR)/kgl+JJo(kg)j3
=1. (27)

For the frequencies we shall be considering ko<k, andsowe cansplitthe convolution into threeparts
k

Thenthe effective bandwidth k0of I(k) is givenby

f0 for k<ko,

fo k I(O)kdk = 1
or

koR= (16/3) = 2.3.

(27')

To find the effective R we may eitherequate rR to


812 Volume51 Number3 (Part 1) 1972

for ko<k<k, for k<k<k+ko. (29

aded 23 Dec 2010 to 200.23.51.1. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/te

SOUND

TRANSMISSION
I''t''1' ' I ''1'

THROUGH
'l

A DOUBLE-LEAF

WALL

6O

5O

o
40

30

20

10

(b)

6O

5O

O o

0 30
o

20

!0

,,
125 250 500 1K IhJ 2K HERTZ 4K

I l,I
250 500

,,

I I , I II
1K
IN

25

2K
HERTZ

4 K

FREOUENCY

FREQUENCY

(c)

(a)

Fr.4.Theoretical and experimental transmission losses forstudless double-leaf walls. (----): experimental curve; (O):theoretical n= 0.015; (): theoretical n- 0.010; ( and IZl): as above but with coso correction athigher frequencies. (a)Each leaf 2.7 lb/ft *, spacing 3- in.;(b)each leaf 2.1 lb/ft ,sparing 2in.;(c)each leaf 2.1 lb/ft ,spacing 3] in.;and (d)each leaf 4.6 lb/ft ,spacing 3 in.

The Journal of the AcousticalSocietyof America

oaded 23 Dec 2010 to 200.23.51.1. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/te

R.

J.

DONATO

If we denote F(k)*I(k)=G(k), we find


k,,(8/3R2)S(k) =k sin-I{k/k)+(k--k:)--k,

k<ko; =k sin-(k/k)--ksin-[(k-k0)/k3+ (k2--k) ko<k<k; =k sin-'(k,/k,)--k sin-'[(k--ko)/k,[] --[k:-- (k--ko):], =0, k>ko+k. (30)

the curveis strongly peaked in the coincident region, one wouldexpectthe calculated lossto be too low. A computerstudy was made of this regionand it was
concluded that a fair allowance for the third-octave

By carrying out theintegrals outlined in Eq. 29with


the values of G(k) calculatedfrom Eq. 30, we may derive values of the total normalizedk spectrumEv

andthe part of thek spectrum E, whichtakes part in soundradiation.If k=fiko is written 8kEr/3R=ko a )< (r/6+2), then
8k,,E. / 1--a\ -- 1 ra
2

3R =kat--)sin-' 6
We also note that

k3

91

(I)
Substituting into Eq. 32 from Eq. 28,

+ I].

(32)

containing the coincidence frequency wasto add 4 dB to the coincidence loss figure.With thisadjustment the agreement is good over most of the range.On the heavier wall (4.6 lb/ft * on each side) two calculated valuesare shownfor the 1000-Hz point (D-I). The lowerone,whichis closer to the experimental point, is obtained by usingEq. 13 at thisfrequency. The worst fit is in the high-frequency region. It appears that Eq. 18 is not a very good description for frequencies close to coincidence and that in this region a modification of Eq. 19 wouldbe moresuitable. For frequencies abovecoincidence, it seems that =0.01 is the best value to use. At coincidence, on the other hand,7=0.015 givesbetter results. It might be that in the high-frequency regionwhere the transmission losschanges rapidlywith slopea correction should be appliedfor the third-octave bandwidth. This would tend to bring the theoreticalresultsdown. It could alsobe that theoriginal assumption of a cos0 multiplier in the cavity no longerapplies at thehighfrequencies. Removing thismultiplier brings these calculated figures
closer to those observed.

It is heartening that the fit is as goodas it is in the midfrequency range.One would expecta goodfit at Er = (,/3) (ko/k) + 1. (33) the lowestfrequencies, but at the middlefrequencies onemightexpect to runinto trouble with theintegration For an initely largepanel,ko 0 and limits in Eq. 12. The goodfit goessomeway towards E. = L (3) justifying the cavitymodel we havechosen. This resultcan,of course, be obtained directlyfrom
VI. COMPARISON WITH PRACTICAL CASE

fo (k/k)(k2--k')-tdk = 1.
As an illustration of the above procedurelet us
consider the case of a 10X8-ft wall. The radius of the

largestcirclefitting into this area is 4 ft and from Eq. 27' k becomes 0.57 ft-L From Eqs. 31 and 33 we may preparethe correction factors of Fig. 3 whichare
to be added to the results of the transmission loss

calculation from Eq. 12.


V. COMPARISON WITH
RESULTS

EXPERIMENTAL

Figures 4(a)-4(d) show the comparison between


the calculated and observed results for four studless

gypsum boardwalls.Throughout, the damping factor


has been chosento be 0.010, which agreeswith labora-

tory measurements made on the vibration of strips and panelsof the wall material.
An extra correction has been added at the coincidence

The analytical representation, while introducing some simplification, doesrepresentthe transmission lossfairly accurately. It must be emphasized, though, that the problemsolvedis somewhatartificial--as are indeedmost solutionsto this problem. In a practical problemwe would have to includethe effectof the supporting studs.This effectmay be twofoldin that (1) the angular distribution of waves through the cavity may be distortedso that, in general,the effect wouldbe to increase the transmission loss, and (2) the studsthemselves might be actingas a coupling medium betweenthe two pncis. The effect of any absorbing medium placedin the cavity would also have to be includedin any comprehensive theory. This should not be too much of a problem,however,as we could replacethe real wavenumber for the waves in the cavity by a complex one. The resultingequations then become more complexbut remain of the samegeneral
form.

An empirical attack on these problems usingexperifrequency itself. The calculated resultsapply for a mental data is givenin AppendixA. One very importantparameter is the internalloss in particularfrequency, whereas the experimental points are thoseobtained usinga third-octave filter. Thus, if the panel. In the present analysis we have taken
814 Volume51 Number 3 (Part 1) 1972

oaded 23 Dec 2010 to 200.23.51.1. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/ter

SOUND

TRANSMISSION

THROUGH

DOUBLE-LEAF

WALL

7=0.010 to givea good fit, and thisagrees with experi- with the 16-in.studarrangement. Then the transmission mental evidence basedon vibration studies,where r/ factor of the wall will be ?q-[ctro. For the studless was found to be 0.012. absorption-free case,the corresponding factor will be r0. The ratio is thus/tq-T/r0, and --10 log0(a
--?/r0) is taken to be the difference of the STC values for the two cases. At first sightit might seemthat we This paper is a contribution from the Division of have sevenequations with sevenunknowns.This is Building Research, National Research Council of not so, however, as thereare four equations with three Canada,and is published with the approvalof the unknowns a, r/ro, /2 and only two equationsto Director of the Division. determine , a, */r0. A least-squares fit is first made to the four equations. It is found that 2 is not too far APPENDIX A. EMPIRICAL APPROACH TO from /t0, and since tt>/l>0, we take tit=tt. The THE EFFECTS OF STUDS AND resultsare shown in Table A-I. As we should expect, CAVITY ABSORPTION rt>r (the stud contributionis greater for 16-in. than for 24-in. separation). Similarlytt2</. Onemay carryout a very crudeanalysis of experi- separation simple mental resultsto get someidea of therelativecontribu- The puzzlingfeature is that a>a, whereas tions made by studtype,studspacing, andan absorbing reasoningwould lead one to expect the reverse. To cavity. We shall take as our measures of the transmis- makea<ax onewouldneedr/ro to be largerandr/ro sionfactor the soundtransmission class(STC) of the smaller,which is not very likely. Another possibility wall. We have results on various walls for the STC of is that there is somecoupling throughthe resilient This wouldmakea2smaller, whichis a change the wallswith (a) no studsupports, (b) wooden studs channels. direction. Yet another possibility is that 16 in. apart, (c) wooden studs24 in. apart, and (d) in the wrong case there may still be some metal channelstuds24 in. apart. These resultsare even for the studless available both with and without 2 in. of mineral wool contribution from the perimeterframework.The data absorptive nmterialin the cavities. We assume that the show thisto besmallenough to be neglected. The most metalchannels remove theeffect of studcoupling. If the we can say is that both a and a2 are lessthan unity, the dependence uponstudspacing effect of thestuds modifies theair transmission pathby but at the moment a for the 16-in.spacing anda2 for the 24-in.spacing, seemsto be oppositeto what one would expect. It is too blunt a tool if the effectof studcoupling for the wooden studs is , might alsobe that the STC measure for the 16-in. spacing and3'2 for the24-in. spacing, and, to separatethe variouseffects. finally, if/0, /, tt2 are the effectsof the absorbing
materials for the studless case and for the two different of ReverberantSound through studseparations, wemaywritethefollowing equations t A. London, "Transmission Walls," J. Acoust.Sac. Amer. 22, 270-279 (1950). for the STC valuescompared to the studless no- Double L. L. Bernsink and G. A. Work, "SoundTransmission through
ACKNOWLEDGMENT

absorption caser0 as

a.,q-?2/ro=C, dB a2=D, -- 42 dB ',typically --1] dB [haq-'/ to=A r o= B, dB


fi2a2-'V2/ro=C t, 2a2= D' ,
[

Multiple StructuresContaining Flexible Blankets," J. Acoust. Sac. Amer. 21, 419-428 (1949). a p. H. White and A. Powell,"Transmission of SoundVibration througha RectangularDoubleWall," J. Acoust.Sac. Amer.
821-832 (1965).

dB dB.

(35)

As an example we take the case of absorption present


TABLE A-I. Air, stud, and absorbenttransmission factors.

R. H. Lyon and G. Maidanik, "Power Flow BetweenLineaxly Coupled Oscillators," J. Acoust.Sac. Amer. 34, 623-639 (1962). s R. H. Lyon and T. D. Scharton,"Vibrational Energy Transmission in a Three Element Structure," J. Acoust. Sac. Amer. 38, 253-261 (1965). E. Eichler, "Thermal Circuit Approachto Vibrations in CoupledSystemsand the Noise Reductionof a Rectangular Box," J. Acoust.Sac. Amer. 37, 995-1007 (1965). P. H. White, "Sound Transmission of Double Flexible Walls Excited by Random PressureFields," PhD thesis, Univ. California, Los Angeles(1965), 157pp. a G. Maidanik, "Response of Ribbed Panelsto a Reverberant SoundField," J. Acoust.Sac. Amer. 34, 809-826 (1962).

A. J. Priceand M. J. Crocker, "Sound Transmission through


0.06 0.12
0.03

0.07 0.11

0.0 0.11
0.15

0.54 0.67

0.42 0.33
0.20

0.46 0.33

0.19 0.29 0.09

Double Panels Using Statistical Energy Analysis," J. Acoust. Sac. Amer. 47, 683-693 (1970).
York, 1960), p. 298.

L. L. Beranek,Ed., Noise Reduction (McGraw-Hill, New


tiE. Skudrzyk, Simple and Complex Vibrating Systems

0.08

0.16

0.32

(McGraw-Hill, New York, 1968), p. 321.

The Journalof the Acoustical Saclelyof America

815

aded 23 Dec 2010 to 200.23.51.1. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/te

You might also like