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

THE JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA VOLUME 34, NUMBER 4 APRIL, 1962

Sound-Insulation Ratings and the New ASTM Sound-Transmission Class


T. D. NO•TrtWOOD
Division of Building Research,National ResearchCouncil,Ottawa,Ontario, Canada
(ReceivedDecember 18, 1961)

A survey is made of past and presentsystemsfor rating the sound-insulationvalue of buildingwalls and
floors.It is observedthat in mostcountriesratingsbasedon "average"transmission losseshavebeenreplaced
by systemsthat comparetransmission-loss characteristicswith a standard contour. An exampleof this
approachis the new sound-transmission classcontainedin ASTM E90-61T. This is discussed from various
theoreticaland practicalviewpoints,and it is concludedthat the sound-transmission classis a usefulrating
systemfor commonarchitecturalproblems.

HE sound-insulation for a partition constitute


requirement a recommended standard but in most other
dependson the occupanciesit separates;more casesthey are mandatoryprovisionsof buildingcodes.
explicitly, it dependson the magnitudeand frequency Actual minimumrequirementsare beyondthe scopeof
distribution of the noise producedon one side of the the presentpaperand of ASTM E90-61T. Nevertheless,
partition and the amount of intruding noisethat will it is hoped that this discussionof the theoretical and
be tolerated on the other. In rare instances, the char- practical basesof sound-insulationrequirementswill
acter of the noisemay be predictableand constant,and lead first to the general use of the sound transmission
the tolerancelevel of transmitted noise may be accu- classas a rating systemand ultimately to its use as the
rately specified.Then it may be possibleto specifyin a basis of minimum standards.
straightforward way the detailed transmission-loss
EMPIRICAL APPROACHES
requirementsas a function of frequency. But more
typical is the problem of designingan apartment or For many years, both here and in Europe, the
office building, where the required sound insulation commonlyusedrating wasthe arithmeticaverageof the
varies widely with the individual occupantsand their transmissionlossesin decibelsmeasuredat a specified
activities of the moment. Here the architect and his
seriesof test frequencies.On this continentthe standard
client need a simplefigure of merit that will help them method of test, described in ASTM E90-55 and its
provide enoughsoundinsulation to satisfymost of the counterpartAmerican Standard Z24.19-1957,used the
building occupantsmost of the time. arithmetic averageof the transmissionlossesmeasured
It is the purposeof this paper to examinethe problem at the nine frequencies125, 175, 250, 350, 500, 700,
of providingsucha figureof merit, and in particular to 1000, 2000, and 4000 cps. Two objections have been
descr•e the recently developed Sound Transmission raisedto the useof suchan averageas a figureof merit'
Class, which appears in the revised standard ASTM (1) it givesequalweightto all testfrequencies regardless
E90-61T. •
of their importance in sound-insulationproblems
The soundtransmissionclasswas designedprimarily (althoughby includingthe half-octaves below1000cps
for assessingthe sound-insulationvalue of walls and the U.S. nine-frequencyaveragegivesextra weight to
floors for use between dwelling units in apartment the low-frequency range); (2) it givesequalweightto
buildingsand similarstructures.It alsohas application both high and low transmissionlosses,as if super-
in hotels,hospitals,schools,and other situationswhere latively high values at some frequenciescould com-
the requirementsare similar in character,though not pensatefor deficienciesat other frequencies.
necessarilyin degree, to those encounteredbetween Attempts to meet the first objectionhave led to the
dwellings.For partitions betweenofficesit is usual to introductionof severalother "averages,"obtainedby
require only that transmitted speechbe unintelligible. altering the selectionof frequenciesincluded in the
This suggestsan approachslightly different from the average.There is sometimesgoodreasonfor concentrat-
dwelling problem, but the soundtransmissionclassis ing on a specialfrequencyrangebut unfortunatelyit is
again found to be a usefulrating system. rarely made clear, especiallyin trade literature, when
The new classificationsystem is an intermediate or why a nonstandardaverageis being employed.As
stage between the physical measurementof sound- one step in producingan orderly presentationof infor-
transmissionloss and the specificationof minimum mation ASTM E90-61T now requiresthat only the
requirementsfor separatingvarious occupancies.It "average"reportedbe the nine-frequencyaverage,and
might be noted that similar criteria are usedin Britain that it be so labeled.
and severalEuropeancountriesfor specifyingminimum The secondobjectionis basedon the premisethat a
requirementsfor dwelling separations.In Britain they partition is no better than its lowest transmissionloss.
• ASTM E90-61T, "Tentative recommended practice for It led on this continent to the development of the
laboratory measurementof airborne sound transmissionloss of "energy average,'"'obtained by averagingthe trans-
building walls and floors" (American Society for Testing and
Materials, 1916Race Street. Philadelphia,1961). • R. V. Waterhouse,J. Acoust. Soc.Am. 29, 544 (1957).
493
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.88.90.110 On: Sat, 20 Dec 2014
22:10:27
494 T. D. NORTHWOOD

missioncoefficientscorrespondingto the transmission were disturbedby traffic noise,the less they were
lossesand then taking the decibel equivalent of the disturbedby noisefrom their neighbors.
averagetransmissioncoefficient.Such an averageis Correlation of tenant disturbance and airborne-sound
dominatedby the lowestvaluesof transmission loss. insulationwasin generalvery complex.For example,
Unfortunately,the lowestvaluesare almostalwaysat it is difficultto distinguishbetweenreactionsto airborne
the lowesttest frequencies, which are usuallythe least andimpactsounds; tenantsdissatisfied with oneaspect
importantin the evaluationof partition performance. of their dwellingstend to expressdissatisfaction with
Moreover,evenin the laboratoryit is difficultto make other aspects;a tenant'spast experience with very
precisemeasurements of sound-transmissionlossat the inferiorhousing may causehim to commentfavorably
lowest test frequencies;thus, the energy average is on a slightly less inadequate environment; and of
usuallybasedon the leastreliablemeasurements. coursethereis the complicating effectof trafficnoise,
Clearly,nosimpleaverage,or evena complicated one discussedin the foregoing.Nevertheless,both the
such as the energy average, can properly rate the British and the Swedishsurveys showeda definite
performance of a partitionunlessit takesinto account reduction in complaintsfor increasedairborne-sound
the variation of transmissionlosswith frequency.The insulation.The Netherlandsstudy, which involved
usefulnessof simple averagesin the past may be sevendifferenttypesof structure,yieldedlittle corre-
attributed to the fact that the massive structures lation between tenant reactions and sound insulation.
that predominated had rather simpletransmission-loss The Britishsurveyincludedtwo main typesof floor
curves,all risingin a similarway with frequency.But structure for which the transmission-loss characteristics
the presentuse of lightweightstructureshas resulted were similar, with averagesof 49 and 44 db.ø In the
in both lower and more irregular transmission-loss 49-dbstructures, 22% of the tenantscomplained about
curves,and necessitates a criterionthat is moreclosely noiseproblems,andnoisewasaboutequivalentto other
related to actual performancerequirements. sourcesof complaint.In the 44-db structures,36% of
the tenantscomplainedabout noise,and noisewas the
Subjective Reactions and Standard Contours major complaint.Thesetwo transmission-loss
character-
istics,somewhatidealized,formthebasesof thegradeI
In at least three countries, Holland, Sweden, and and gradeII curvesshownin Fig. 1. They constitute
Britain, the dwellingseparationrequirementwas in- recommended minimum values for party walls and
vestigateddirectlyby canvassing the tenantsof apart- floorsin certainclassesof apartments.7
ment buildings and row dwellings. 3-• Each of the The sound-insulation datafromthe threesurveysare
surveyswasconducted withoutknowledge of theothers, shown in Fig. 2, in which the abscissais the ASTM
and it is of interest to examinethe points of agreement sound-transmission class,whichis describedlater; for
and disagreement amongtheir conclusions. the moment,it can be taken as roughlyequivalentto
The British and Dutch surveys showed that noise averagetransmission loss.The resultsservemainlyto
from the floor above constitutedthe greatest disturb- indicatethe order of soundinsulationrequiredfor
ance,apparentlybecauseof the specialimportanceof acceptable dwelling separation. The one anomalous
impactsoundssuchas footstepsand childrenplaying. pointin theBritishresultswasfor a newblockof apart-
The Swedishsurvey also showedthe importance of mentswhosetenantsweremostlyrefugeesfrom a con-
transmissionthroughfloors,but did not indicate that demned-housing area; they are thereforenot a repre-
impactnoisewasasserious a problemasairbornesound. sentativesample.Someinformationwasgleanedfrom
Nevertheless,current British and Europeanstandards, the Dutch resultsby discarding two structuraltypes
includingthe Swedishone, deal with both impact and for whichimpactinsulationwasexceptionally low,and
airborne sound. On this continent there is as yet no by separatingthe restinto two groupswith highand
standardtest methodfor impact (althoughone is now low insulation. Despitethesecautionaryremarks,the
being considered by ASTM CommitteeE-6). Hence
this paperdealsonly with the airborne-sound problem,
despite
Thetheevident
Swedish importance
sample ofimpact
included transmission.
similar ø
structures Fro. 1. British recom-
located
onnoisy
thoroughfares
andonquiet
residential mended minimum require-
ments for airborne-sound
streets.
Asmightbeexpected
these
showed
aninverse ] 40 insulation between apart-
relationship between disturbance from traffic and ment dwellings. (A sound
disturbancefrom adjacent dwellings;the more tenants transmission class contour
i 20 .... ' is shownfor comparison.)
aC. BitterandP. VanWeeren.
"Sound
nuisance
andsound •00 400 •000 '40'0•
insulation in blocks of dwellings I," Rept. No. 24, Research
Institute for Public Health EngineeringT.N.O. (1955).
40. Brandt and I. Daldn. Byggm•staren 31, 145 (1952). 6The standardEuropeanaverageis basedon measurements at
5p. G. Gray, A. Cartwright,and P. H. Parkin, "Noisein three third-octaveintervalsfrom 100to 3200cps.
groups of flats with different floor constructions,"National 7BritishStandardCodeofPractice,Chap.III: "Soundinsulation
BuildingStudiesResearchPaperNo. 27, H.M.S.O. (1958). and noisereduction,"British StandardsInstitution (1960).

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.88.90.110 On: Sat, 20 Dec 2014
22:10:27
SOUND-INSULATION RATINGS 495

4O i i i l
transmissioncharacteristicselectrically and had 20
subjectsdetermine the subjectivereduction in loudness
BRIT
• 3o provided by various transmissioncharacteristics.For
Fro. 2. Results of • noise sourceshe used samplesof music and speech
three surveysto deter-P= (garbledto eliminatemeaning)and severalspectraof
minethe acceptability
• 20 .9P '•.,,,. '\
ofexisting
sound
insula-• /'•" •t-...•sw•o•. filtered white noise.His principalnoisesamples,includ-
tion betweenapartment•z NETHERLANDS• x\
10 ing speech, music, and one sample of white noise,
dwellings. • >50 eachhad a broadmaximumin the frequencyrange200
0
to 800 cps.
SOUNDTRANSMISSION
CLASSOF PARTITIONS
Startingwith the GermanDIN curve,he investigated
the effectof decreasingthe attenuation by 10 db in one
octave band, as compared to a uniform attenuation
results suggestthat a sound transmissionclass 45 is over five octaves. He found that the 10-db decrease in
borderlinefor reasonablysatisfactorysoundinsulation, oneoctavewasequivalentto 1 to 4 db overfive octaves,
and that a soundtransmission class55 representsabout the greatest effect being, as might be expected,in the
the maximum value that is worth attempting for rangeof highestsourcepower.In anotherseriesof tests,
ordinaryhousing. he comparedfour attenuationcurvesof differentshapes,
A similar set of rating curves (Fig. 3) was incor- each correspondingto an average transmissionlossof
porated in the German standard DIN 52211.8The same 48 db. One characteristiccorrespondedto the DIN
shapewassubsequently adoptedalsoby theScandinavian curve, the secondincreasedat 3 db per octave, the third
countriesas the basisof sound-insulation requirements increasedat 6 db per octave, and the fourth had a
for row-dwellingand apartmentbuildings. 9.•øThe DIN plateau characteristicextendingto 800 cyclesand then
curves,alsobasedon the characteristicsof actual walls, a sharp increasein attenuation. The first three were
differ from the British curves in requiring slightly rated subjectively at about 48 db (when tested with
greatertransmission lossin the middlefrequencies(300 speech or music),but thefourthwasratedat only37 db.
to 1000 cps) and slightly lessat lower frequencies.It This result might be anticipated sinceonly the fourth
is to benotedthat the standardrequiresa slightlyhigher wall differed substantially in the range of maximum
value (2 db) for a laboratorytest than for a field test. source power. Rademacher dismissed this fourth
The differenceis an allowancemainly for flanking characteristicas physically unreal and concludedthat
transmission.
a simplearithmeticaverageis adequatefor rating most
The use of a standard curve rather than a simple structures.Actually it is similar to many lightweight
arithmetic averageis basedon the hypothesisthat the partitions, and it is to protect against transmission
transmission-loss requirementvaries in a certain way curves of this type that the standard curves were
with frequencyand must thereforebe specifiedfor each introduced.
frequencyband. The direct evidenceconcerningthe Although the British and European standards em-
contribution of each band is rather sketchy although phasizethe importanceof the shapeof the transmission-
there are theoreticalreasonsfor supportingthis view.
loss curve, they compromisesomewhatin their inter-
In a survey of occupantsof row dwellings,the British
pretation of actual transmissioncharacteristics.Gener-
were able to compare a 9-in. solid-brick wall with an
l l-in. cavity wall; they found little superiorityfor the ally, an averagedeficiencyof 1 db (in somecases2 db)
latter despitethe fact that its averagetransmission loss is allowedrelativeto the prescribedcurve.In both cases,
was 5 db better. It was concluded that the lower and the deficiencymight conceivablytake the form of a
middle frequencies,for which there was little difference large deviation in a limited frequency band. It is
between the walls, were more critical than the high commented 8that this averagingarrangementis adopted
frequencies for which the superiortransmission lossof in the absence of sufficient information about the
the l l-in. wall is most evident. The transmission loss is contributionof individual narrow bands.In any case,
presumablyhigh enoughat high frequenciesin either
case.Both walls just meet the British gradeI require- 70 i , . ,, ] , , ,,
ment.
•m LAB
TESTS-
FLOORS•2.
More extensive information is provided by the u-r60 LAB
TESTS-WALLS• / _
experiments of Rademacher,• who simulated wall-
FtG.
3.German
minimum
$ FIELD
TESTS-FLOO • LLS

a "Bauakustiches Prufungen Schalldammzahl and Normtritt-


schallpegel,"DeutschenNormenausschusses DIN 52211 (1953).
0 F. Ingerslev and J. Kristensen, "Lydisolation I Boligbyggeri"
(Sound insulation of dwellings) with English summary,
Statensbyggeforskningsinstitut,
Rapport39, K•benhavn (1960).
requirements
for
airborne-
g50
•A -i
sound insulation between •
apartmentdwellings.(A • 40
sound-transmission

parison.)
class
contouris shownfor com- g
•_

= 3o
•0"Anvisningar till byggnadsstadgan," BABS Stockholm i , , , , ! , ,

(1960). •ø ,oo ,oo ,ooo ¾oo


n H. J. Rademacher,Acustica5, 19 (1955).

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.88.90.110 On: Sat, 20 Dec 2014
22:10:27
496 T. D . N OR T H W OOD

80 I I I I I I I I I I I
they are intendedto illustrate spectrumshapes,rather
Fro. 4. Half-octave-band
- • 7o .• -"'-'•'•.'"y• levels of typical household
than absolutelevels. Severalimportant noises,partic-
noises. Curve A--room air- ularly airbornenoiseresultingfrom footstepsand doors
.• • <' ..z' 's xx \.
conditioner, Curve B-- slamming,have been omitted for lack of information.
• • 60 / 'X k vacuum cleaner, Curve C--
•• / "X '.,, 'xc It is assumedthat the optimum shapeof a rating
• 6 50 x." '........
'"'B normal speech (levels ex-
ceeded by 1% of speech curve will not vary much with level in the range of
•• "X. peaks), Curve D--radio, applicationof the rating system,and that a family of
• 40 '•. television (peak levels,
speechand music), Curve parallel curves can be adopted. This is analogousto
•0 I I I I I I I I I I
E--assumed "standard- assumingthat equal loudnesscontours are approxi-
125 250 500 I000 2000 4000 household noise."
MID-B•ND
FREQUENCY,
cYCLES•SEc mately parallel in the range 20 to 60 phons.
On the listeningside, the most obviousprocedureis
to consider the loudness of the transmitted noise. The
since excessesabove the curve are disregarded, the work of Stevens,•' Quietzsch,• and BeraneU4 provides
allowabledeficiencyis actually rather limited. a goodbasisfor such a criterion, although their results
Summarizingfield experienceand experimentalevi- are more directly applicable to office and industrial
denceto date, the consensus
is that a simplearithmetic problemsthan to dwelling separation.Stevensdevised
averageisnotadequate,especially for ratinglightweight a methodof calculatingthe loudnessof complexsounds
structureswhichfrequentlyhave seriousdeficiencies in that, particularly in its latest form, agreeswell with the
the mostcriticalfrequencyrange.Severalcountriesare subjective ratings obtained by Quietzschfor a wide
now using standard transmission-loss contours to variety of complex noises.Beranek used these data,
representminimumrequirements for dwellingsepara- supplementedby office-noisesurveys, to develop the
tions.Ideally, the transmissionlossof a givenpartition well-known noise criterion (NC) curves, which take
shouldat no frequencyfall belowthe standardcontour; into account both loudness and speech-interference
in practice,a limited amountof tradingbetweenfre- levelsapproximatelyas they affect the acceptabilityof
quencybandsis permitted,althoughthe consequencesoffice noise. These criteria are used in later sections.
of this procedureare not yet fully understood. Assumingthat the subjective reaction to noise is
THEORETICAL CONSIDERATIONS
related to loudness,it is appropriate to determine the
transmission-loss characteristic that reduces each band
The British and Europeanstandardswere developed of "standardhouseholdnoise" (Fig. 4) to a particular
largelyby informedinterpretation of the comments of equal-loudness contour.For purposesof this paper the
occupants ofactualbuildings. It isofinterestto examine 0.5 sone contour, correspondingto a loudnesslevel of
also various theoreticalapproaches,utilizing informa- 46 phons,is used. (Judgingby tenant surveys,this is
tion on typical noisesand varioussubjectivecriteria probably of the right order.) The transmission-loss
relatingto loudness, noise,or annoyance. Theseprovide characteristicthat accomplishesthis is shownas curve
someinsightinto the importanceof variousfrequency (a) in Fig. 5 (solid circles).This is the most efficient
bands and assist in selectingan optimum shape of partition for achievingthis transmittedloudnesssince
transmission-loss characteristic. all bandsof transmitted soundare reducedto the point
Figure4 showshalf-octavebandspectrafor a few where they are equally important in determining
typicaldomestic
noises.The trendin domestic
appli- over-all loudness.
ancesis toward control of the high-frequencycompo- Kryter•5recentlydeviseda slightlydifferentseriesof
nents of noise, so that low- and medium-frequency "equal noisiness"curves,leadingto "perceivednoise
components predominate in the residualnoise.Speech, levels" instead of loudness levels. The calculation is
radio,and televisionnoisesare broadlypeakedin the essentiallythe same as Stevens' except that Kryter
middle-frequency range.Speechintelligibility,as dis- providesgreaterweightingfor high-pitchedcomponents
tinct from power,involvesa slightlyhigherfrequency of noise.Kryter claims for his curvesonly that they
rangeextendingwell beyond4000 cycles,but this is correlate better than loudnesswith subjective judg-
irrelevantfor dwellingseparationsincethe transmission ments of acceptabilityfor certain high-pitchednoises
lossshouldbe substantiallygreater than the amount suchasare producedby jet aircraft. It appearsthat the
requiredmerelyto reduceintelligibility.
Musicalinstru- high-frequencycomponentsmay be especiallyim-
mentsand high-fidelityrecordplayerswill extendthe portant when a noise is intrinsically obnoxiousor
range,especiallytowardthe lowerfrequencies. Noting alarming.Reviewingthe most troublesomesourcesof
from the surveysthe specialimportanceof radio, dwellingnoises,it seemslikely that someof them fall
television,
andspeech noises,it appears that onemight into the intrinsicallyobnoxiouscategory,and it might
considera "standard householdnoise" spectrum flat therefore be more plausible to use Kryter's criterion.
from 250 to 1000cpsand diminishing by 4 to 6 db per
octave below and above this frequency(Fig. 4). It •' S.S. Stevens,J Acoust.Soc.Am. 33, 1577 (1961).
• G. Quietzsch.Acustica5, 49 (1955).
shouldbe noted that the curvesof Fig. 4 are derived •4L. L. Beranek,J. Acoust.Soc.Am. 28, 833 (1956).
from severalsourcesand are not strictly comparable; •5K. D. Kryter, J. Acoust.Soc.Am. 31, 1415 (1959).
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.88.90.110 On: Sat, 20 Dec 2014
22:10:27
SOUND-INSULATION RATINGS 497

weighted five times as much as the other bands. On

,•60
t •• •••
•, 50
o

g 40
A ß
o
either basis,it appearsreasonablypreciseto require
that a given transmission-loss requirementbe met in
eachband,rather than just on the average.
For dwellingseparation,however,wherenoisesources

-[
• .30

125 250 500 I000 2000 4000


MID-BANDFREQUENCY,CYCLES/SEC
are highly variable, anotherapproachis to considerthe
probabilitythat noisewill occurin a givenfrequency
band.Both noisyand quiet conditionsmay fluctuatein
the mannerindicatedin Fig. 4. A reviewof the spectra
FIc. 5. Theoretically derived transmission-loss curves.Trans- of the moretroublesome noisessuggeststhat high noise
missionlossrequiredto reducestandard-household noiseto: (a)
0.5 soneequal-loudness contour(solidcircles),(b) 0.5 Noy equal- levels are more probablein the mid-frequencyrange
noisiness contour (open circles), (c) NC-25 contour (triangles). thanat highor lowfrequencies; hencethisregionshould
The solid line is the contour of sound transmission class 48.
be givenspecialattention.For this reason,in applying
the proposednew rating, the followingprocedureis
Proceedingas before,one arrivesat the requiredtrans- used: there shall be no deficiencies below the middle
missionloss defined by curve (b) of Fig. 5 (open segmentof the STC curve,but deficiencies averaging
circles). 1 db are allowedbelowthe outersegments of the curve.
The transmitted loudnessor perceivednoiselevel,
per se, is not necessarilya good index of satisfactory OFFICE PARTITIONS AND SPEECH PRIVACY
soundinsulation.More importantis the degreeto which
the transmitted sound is masked by ambient noise In the foregoingsection,the soundtransmissionclass
existingon the listeningside of a partition. For com- wasdiscussed primarilyfrom theviewpointof dwelling
pletemasking,a goodapproximationfor complexnoises separation.It is of interestalsoto considerits applica-
is to requirethat the transmittednoisebe no greater bility to the problem of officeseparation.This is the
than this ambient level in any half-octave band.•6 secondlarge-scaleproblemconfrontingthe designer,
Curves(a) and (b) could,on thisbasis,be regardedas anddiffersin severalwaysfromthe dwellingseparation
possiblespectraof ambient noise.SinceBeranek'snoise problem.
criteriaare frequentlyusedfor similarpurposes,a third The primary requirement for sound insulation be-
transmission-loss characteristic,whichreducesstandard tween offices
is to preventthetransmission of intelligible
householdnoiseto NC-25 contour,hasalsobeenshown speech.This is soboth for the speaker,whomay wish
as curve (c) in Fig. 5 (triangles). to speakprivately, and for the listener, sincethe dis-
Comparingthese curves with the DIN curve, it is tracting quality of speechnoise is intensifiedwhen it
seen that the latter is reasonablysimilar in shape. beginsto be intelligible.The mostcomprehensiverecent
Following these considerations,and with an eye to studyof factorsaffectingspeechintelligibilitywasthat
internationalstandardization,it was decidedto adopt of Frenchand Steinberg,•7 who developeda straight-
the shapeof the DIN curvesas the basisfor the ASTM forwardmethodof calculating speech
intelligibilityfrom
soundtransmission classcontours(Fig. 3). Plotted on thepropertiesof eachlink in a communication system.
conventionalsemilogpaper the STC contoursconsist They first showedthat speechintelligibilitycan be
of a horizontalsegmentfrom 1400 to 4000 cps, at a expressedrather preciselyin terms of the available
level corresponding to the soundtransmissionclass;a dynamicrangein eachof 20 equallyimportant"critical"
middle segmentthat decreases6 db from 1400 to 350 frequencybands.A maximum range of about 30 db
cps; and a low-frequencysegmentthat decreases14 db (aboveeitherthreshold of audibilityor maskinglevel
from 350 to 125 cps. setby noise)is requiredin eachcriticalbandto getthe
Equating the over-all loudnesses
of transmitted noise full contributionof the band.To a goodapproximation
andambientnoisewouldnot resultin completemasking the contributions of the individual bands are inde-
if the loudness level of the transmitted sound were pendentof eachother,althoughthereare secondary
particularlyhigh in a narrow band. But the distinction effectsdueto maskingby adjacentbands.The contribu-
betweenthe equal-loudness and maskingcriteria is not tion of eachcritical band is expressed
in terms of an
as great as might be anticipated, since the over-all "articulationindex,"and the averagearticulationindex
loudness levelof a complexsounddependsmarkedlyon for the twentycriticalbandsprovidesa quantitythat
the level in the loudestband. For examplein Stevens' is relatedin a knownway to the othercommonmeasures
calculation for half-octave bands the loudest band is of speechintelligibility.
Beranek,in his earlierapproachto the problemof
•6H. Fletcher,M. R. French and J. C. Steinberg,E. Zwicker, specifying minimumnoiserequirements within spaces
andothershavedemonsrated that calculationsof loudness,mask-
ing,articulationindex,andsimilarquantitiesinvolvingperception such as officesand conference rooms, concluded that
of complexsoundsare mostprecisewhen"critical bands"are used. easeof speech communication provideda goodcriterion.
Nevertheless, it is commonpracticeto use data in octave,half-
octave,or one-thirdoctave bands.The approximationis valid if •7M. R. FrenchandJ. C. Steinberg,
J. Acoust.Soc.Am. 19,90
the spectraare reasonablysmoothand continuous. (1947).

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.88.90.110 On: Sat, 20 Dec 2014
22:10:27
498 T. D. NORTHWOOD

Hence, he used the work of French and Steinbergto and female voices. The idealized speechspectrum of
develop the criteria known as speech-interferenceFrenchand Steinberg(Fig. 4) is assumedto existin a
levels,TMobtainedby consideringspeechlevelsrelative source room separated from a listening room by a
to noisein the threeoctavesmostimportant to intelligi- partition havinga transmission lossdefinedby an STC
bility (300 to 2400cps).Later, however,he foundthat contour.The transmitted-speech level in the listening
acceptabilityof officenoiseis more closelyrelated to room will depend on the room absorptionsand the
loudness. This led to the NC curves mentioned earlier. •4 partition area; roomabsorptionsof 100 sabinseachand
The speech-privacy problemis the inverseof that a partition area of 100 sq ft will be assumed.It will
considered by French and Steinbergand by Beranek, further be assumed that the ambient level in the
beingconcerned with the marginalconditionin which listening room is defined by an NC curve. Then the
speech is not quiteintelligible.In thisregion,the simple signal-to-noise ratio in eachband is dependenton the
linearapproximationthat workssowell whenintelligi- sumof the STC and NC values(assumingthat the NC
bility is highis no longerstrictlyapplicable.Neverthe- curvesof interestare all parallelto NC-30). The articu-
less,a linear approximationhasbeenusedsuccessfullylation index in eachband is determinedby the speech
in office-separation problemsby Hardy•ø and more level (exceededby 1% of peaks)relativeto noiselevel;
completelyby Cavanaughetal.2øThe latter is probably for level differenceslessthan 12 db, the band-articula-
the most useful method available when the problem is tion index is shownin Table I, and for level differences
well defined:forexample,whenthe roomconfiguration greaterthan 12 db, the valueis W-(E-6)/30, whereE
is known and the ambient noiselevel known or specified. is the level difference.This procedureapplieswhen the
For low-intelligibilityconditions,the effect of their ambient level is moderately above the threshold of
linear approximationis to obtain articulationindex audibility, but not sohigh that nonlineareffectsbecome
valuesthat are somewhatlargerthan thoseobtainedby significant.
the Frenchand Steinbergmethod.Sincethey deduce The relationshipbetweenarticulationindex and the
fromofficesurveysa maximumacceptable valueof this sum of STC and NC numbersis shownin Fig. 6. It is
index,calculatedby the samemethod,the discrepancy seen that for an articulation index of 0.03 and for
is perhapssafelycanceledout again;but it shouldbe S/(AxA2)=O.01, STC plus NC should equal 68. For
noted that their resultsare not comparablewith those example,assumingan ambient level corresponding to
presentedbelow. NC-35, the transmission lossshould then be equal to
For the usualoffice-building designproblem,a rating or greater than STC-33.
that ensures a reasonable probabilityof generalsatisfac- Now, considerthe variables affecting this result.
tion is againnecessary. The usesandlimitationsof the Differencesbetweenindividual speakersmay affect the
sound-transmission classfor thi• purposeare considered speechlevelby up to q-10 db; differences in voiceusage
below,with the helpof the Frenchand Steinbergstudy betweena smallofficeand a largeconference roommight
and their nonlinear treatment for conditions of low introducea similarchangeof q- 10 db, althoughthe two
intelligibility. Following Beranek's example, TM the effectsareprobablynot cumulative(i.e., to someextent
critical bandsof French and Steinbergare replacedby a loud talker is one who habitually declaims).Differ-
bandsof equal-frequency ratio, in this caseby half- encesin room absorptionand partition size from the
octave bands. The conversion is accomplishedby assumedvalues will affect the transmitted level by
weightingeachhalf-octaveby a factorproportional
to changingthe factor10log[S/(A 1A,)•, but, in any case,
•øThe weighting this factor, derivedfrom reverberant-room
the numberof criticalbandsit contains. theory,is a
factorsare derivedfrom composite studiesof both male crudeapproximationin modernofficeswith absorbent
ceilings.
TAB,.r.I. Band-articulation index for small differences between
the levelexceeded
by 1% oPspeech
peaksand the ambientlevel. .O8
(Adaptedfrom reference17.)

Level Level
diff 30W a diff 30W
1 0.2 7 2.4
2 0.4 8 3.0
3 O.7 9 3.6
4 1.1 10 4.4
5 1.5 11 5.1
6 1.9 12 6.0
60 65 70 75

• 30W rather than W is tabulated for convenience in calculations. (SOUND-TRANSMISSION


CLASS)+
(NCMASKING
CONTOUR)
Fro. 6. Combinedeffectof idealizedpartition (conformingto a
28L. L. Beranek,Proc.Inst. RadioEngrs.35, 881 (1947). givensoundtransmission classcontour)anda givenambient-noise
29H. C. HardyandJ. E. Ancell,NoiseControl4, 9 (1958). level (conformingto modifiedNC contour)on articulationindex
W. R. Farrell,P. W. Hirtle, B. G. Watters, of transmitted speech. (S--Area of partition; A2 and A• are
•.0W. J. Cavanaugh,
"Speech
privacy
in buildings," Soc.Am.34,475(1962). absorptions
J. Acoust. of sourceand receivingrooms.)

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.88.90.110 On: Sat, 20 Dec 2014
22:10:27
SOUND-INSULATION RATINGS 499

TABLE II. Analysis of four walls.

Equivalent sound transmissionclass "Averages"


Test 1 Test 2 Test 3
Sound Equal masking Equal
transmission Equal by ambient artic. 9-Freq. 11-Freq.
Wall class loudness noise index arithmetic arithmetic Energy
A 47 51 49 50 46 48 40
B 36 37 35 39 32 34 26
C 30 36 31 32 33 36 27
D 20 27 23 25 21 22 21

Deviations of the ambient-noisespectrumfrom the =0.01. More-precise values can be determined for
assumed NC-curve will also affect the result. In the othervaluesof S/(A •A2)by referringto the appropriate
calculation leading to Fig. 6 it was found that the curvesof Fig. 6.
articulation index dependson the bands from 350 to
1000 cycles,with a lessercontributionfrom the 250- RATINGS OF ACTUAL WALLS
and 1400-cyclebands; hencethe NC-curve shouldbe
matched to the noise in this range unlessthe noiseis To illustratethe useof the ratings,a detailedanalysis
concentrated in other bands. is made of the performanceof the partitions whose
Finally, thereis the error introducedby the method transmission characteristicsare shownin Fig. 7. Wall A
of matching a transmission-loss curve to the STC is a solid concrete wall 3 in. thick. The coincidence
contours.Typically, an actual curve will fall on the frequencyfor this wall is below the test range and it
STC curveat oneor two frequenciesin the middlerange therefore has a smooth, steeply rising characteristic.
and be above at the other 3 or 4 frequencies.Conse- Wall B is a 2- by 4-in. stud and plasterwall. Suchwalls
quently,a partitionwith a sharpmid-frequency dip will have a characteristichigh-frequencydip and a low-
providea lowerarticulationindexthan its STC rating frequency dip that varies in detail from sample to
would indicate. In a few cases calculated for actual sample; this particular curve was taken from reference
partitions the standard matching procedurewas too 20, Fig. 16. Walls C and D are two officepartitions that
conservativeby 0 to 6 db. This seemssmall enoughfor have pronounceddeficiencies in the middle range.
a generalrating of this type in view of the other vari- The sound-transmissionclass, determined as pre-
ablesin the calculation.Moreover, it is suspectedthat scribed in ASTM E90-61T, is given in the second
althoughspeech intelligibilityis a primaryconsideration columnof Table II. The reliability of the classrating
in officeseparation,loudnessis probablya secondary was tested in three ways describedin the following.
but significantone.This is illustrated,for example,by Since the standard test frequenciesform a half-octave
experimentsof Cavanaughet al.,2øwho usednarrow- series,the analysiswas made on a half-octave basis
band transmission in a study similar to those of throughout.A sampleset of calculationsis given in
Rademacher,but with speechprivacy as a criterion.It AppendixA.
was found that a narrow-transmission band in the
Test 1
regionof maximumspeechpower had an annoying
effectout of proportionto the intelligibility it carried. Loudnesslevels were calculated using the method
The averageerror due to matching curvescan be (Mark VI) given by Stevens n for the fraction of
minimizedby relaxing the requirementslightly; it is standard-household noise(Fig. 3) transmittedby each
thereforesuggestedthat STC plus NC valuesshould wall. Similar calculations were made for a range of
total 66 for roomconditionscorresponding to S/(A •A2) sound-transmission class contours in order to obtain the
curveof Fig. 8. From this it waspossibleto determine
7O i i i i i i i i i i i

B
FIG. 7. Transmission
losses
of four typicalwalls. /,-',/ /
Curve A--3 in. concrete FIG. 8. Loudness•.
levels of standard- •
wall, Curve B--2X4 in. i I/ '.,j _ household noise trans- •40
stud wall (Reference 20, •.." / O
Fig.16),Curves
C andD-- / ./ / mitted by partitions•
Officepartitions.
• .. /
\o- -.t--*'
..'
/
/ conforming to STC õ
contours. .• 20

o I I I I
I I; •, I I I I •
125 2 0 500 I000 2000 4000 20 30 40 50 60

FREOUENCY,
CYCLES/SEC SOUND TRANSMISSION CLASS

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.88.90.110 On: Sat, 20 Dec 2014
22:10:27
500 T. D. NORTHWOOD

the STC contour that would transmit the same loudness the energyaverage,undulyinfluencedby low-frequency
as each wall. This "equivalent sound-transmissiontransmissionlosses,grosslyunder-rateswalls A and B.
class"is given in the third columnof Table II. This is
SUMMARY
a test of the systemon the assumptionthat the absolute
loudness of transmitted sound is an index of wall Evidencehas beenpresentedshowingthat a simple
performance.(A similar calculation,using perceived- averageis an unreliableindex of the sound-insulation
noiselevelsinsteadof loudnesslevels,producedalmost value of a partition. It is noted that in many other
identical results.) countriesthe simple averagehas been replacedby ß
standard contour that defines transmission loss as ß
Test 2
functionof frequency.The significance of suchcontours
A better criterionis to requirethat transmittednoise has been examinedtheoreticallyfrom the viewpoint of
be low enoughto be maskedby ambient noise. This dwellingand officeseparation,and it is shownthat the
condition would be approximately attained if peak sound-transmission classnow incorporatedin ASTM
levels of transmitted noise were not greater than the E90-61T providesa simpleand accuraterating system.
ambient level in any half-octaveband.•ø Assumingan
ACKNOWLEDGMENTS
NC contourof ambientnoise(modifiedfor half-octaves)
the NC curve that just-maskedtransmitted noisewas The author is gratefulfor considerable help from the
determined for each wall. Then the STC contour just other membersof the Writing Group responsiblefor
maskedby the same NC curve was also determined. ASTM E90-61T, and especially for comments and
This is given in the fourth columnof Table II. criticismsreceivedfrom Ralph Huntley, William Jack,
ThomasMariner, T. J. Schultz,R. W. Young, and Bill
Test 3 Watters. This paperis a contributionfrom the Division
of Building Researchof the National ResearchCouncil,
Articulation indiceswere calculatedfor each wall, on Canada, and is publishedwith the approval of the
the assumptionthat ambient noisecorrespondingto an Director of the Division.
appropriateNC curvewaspresent.(The NC curvewas
chosenso that the STC rating of the wall plus the NC APPENDIX A. SAMPLE CALCULATIONS
value totalled 66, correspondingto the speech-privacy OF WALL PERFORMANCE

requirementdiscussedearlier, but the exact criterion The calculationsmade in preparing Table II were
used is not important in the calculation.)The actual quite straightforward,but it might stimulateadditional
articulation index obtained was used to deduce from studies if the details are set down here. Actual calcula-
Fig. 6 the equivalent sound-transmission class for tionsare given in Table III.
speech,i.e., an actual STC contourthat would combine
with the assumedNC curve to give the observed Transmitted-Noise Levels
articulation index.
The standardhousehold noise(Fig. 4) is taken to be
The three equivalent sound-transmission classes,
the level in a sourceroom containing100 sabinsof
basedon the three criteria describedabove, are shown
absorption.The receiving-roomcorrectionis zero for
in Table II. Since the masking criterion (test 2) de-
receiving-roomabsorptionof 100 sabinsand a trans-
pendson the highestband of noise,relative to the NC
mitting-wall area of 100sq ft. Hencethe receiving-room
curves, it agreescloselywith the actual sound-trans-
level is obtainedsimply by subtractingthe wall trans-
missionclass,whichis determinedin a similarway. This mission losses from the band levels of the standard
is the safestcriterion,sinceit is a measureof the proba-
noise.This is the first step in tests 1 and 2. A similar
bility that the transmitted noise will be masked by
procedureis usedin test 3, after first adapting French
ambient noise and thus unnoticed. As might be ex-
and Steinberg'sidealized-speech curve (level exceeded
pected the other two, since they tend to average out
by 1% of speechpeaks)for half-octaveband levelsin
peaks,fall slightlyabovethe sound-transmission classes the source room.
found by the standard procedure.But, apart from a
slight shift in scale, the sound-transmissionclass Loudness Calculations
accuratelyrates partitions in comparisonwith any of
the three tests considered here. Stevens'procedurewas followed exactly, using the
For comparison,three "averages,"the 9-frequency formula St-S•-}-O.2(•:&-Sm), where S is total loud-
average,an l 1-frequencyaverage (includingdata for ness in sones, Sm is the loudnessin the loudest half-
1400and 2800 cps),and the energyaverageare shown octaveband and (•:S•-S•) is the sumof theloudnesses
in the last three columns. These serve to illustrate the in the other bands.
inconsistencies
that can arisewith simpleaverages.All
three show considerable scatter relative to the sound- Masking by Ambient Noise
transmissionclassor to any of the three tests; none of An NC curvewasfoundthat just equaledor exceeded
the three givesthe proper ranking for walls B and C; the transmitted standard noise in each half-octave

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.88.90.110 On: Sat, 20 Dec 2014
22:10:27
SOUND-INSULATION RATINGS 50I

TABLEIII. Samplecalculationsfor wall A.

Half-octave band (centerfrequency)


125 175 250 350 500 700 1000 1400 2000 2800 4000

1. Standard household noise


(L•=L2+TL) 70 72 74 74 74 74 74 72 70 68 66
2. TL, wall A 33 35 39 41 44 48 52 55 58 61 64
3. Transmitted-noise level 37 37 35 33 30 26 22 17 12 7 2
Test 1--Loudness
4. Loudness index 0.23 0.40 0.45 0.50 0.50 0.37 0.26 0.12 0.05 0.02 0.00
5. Total loudness, St=0.50+0.2(2.40)--0.98. Loudnesslevel=40 phons.
Equivalent sound-transmission
class=51 (from Fig. 8).
Test 2--Masking
6. Modified NC-28 a 44 40 37 33 31 28 27 25 24 23 22
(This just exceedstransmitted noisein all bands.)
Equivalent sound-transmission class= 77-28=49.
Test 3--Speech intelligibility
7. Speechnoise'L•--I,2+TL 68 71 74 75 75 75 71 67 61 58 55
(Level exceededby 1% of peaks)
8. Receiving-roomlevel 35 36 35 34 31 27 19 12
(subtract line 2)
9. Subtract ambient level
NC-19a (- 66- STC 47) ßßß 5 8 10 10 8 1 ............
10. 30W (per critical band) ß ßß 1.5 3.0 4.4 4.4 3.0 0.2 ............
11. Band-weightingfactor (10
times no. of critical bands
per half-octave) 0 0 4 10 13 16 20 26 32 33 26
12. 6000 A= 165. A=0.028 ...... 12 44 57 48 4 ß ßß
From Fig. 6, STC q-NC -- 69; hence,equivalent STC -- 50.

• NC curves are modified for « octaves and made parallel to NC-30.

band. This was taken to be the ambient level in the Then the maskingconditionis expressed simplyby the
receiving-roomthat would just mask the transmitted formula STC+NC=77; for example, in the sample
noise.The equivalent sound transmissionclassis the calculation of Table III the standard household noise
STC contour that attenuates the standard noise suffi- attenuated by STC-49 is just maskedby NC-28.
ciently for it to be just maskedby the sameNC con-
Articulation Index
tour. This calculationis facilitatedby the fact that both
STC and modifiedNC contoursare familiesof parallel The method outlined in an earlier section is illustrated
curves (the latter are taken to be parallel to NC-30). in Table III, lines 7 to 12.

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.88.90.110 On: Sat, 20 Dec 2014
22:10:27

You might also like