Вы находитесь на странице: 1из 10

Measuring auditorium seat absorption

W. J. Davies,
a)Fl.J. Orlowski,
b)andY. W. tam
Departmentof AppliedAcoustics,
SalfordUniversity,SalfordM5 4Brl],England

(Received29 September
1993;revised3 March1994;accepted
26 April 1994)
There is a needfor a more accuratelaboratorymeasurement methodto predictauditoriumseat
absorption. The traditionalmethodtendsto overpredict
theabsorption
of theexposedfrontandsides
of seatingblocks.An alternativereverberationchambermethodwasstudiedthatinvolvesthe useof
barriersto obtainrealisticmeasurementsof frontandsideabsorption.
This methodwasvalidatedby
comparingmeasurements of seatsmadein a reverberation
chamberwith in situabsorption
datafor
the sameseats,calculatedfrom reverberationtime measurements
in ten auditoriawith and without
the seatspresent.The accuracyof the alternativemethodwas satisfactoryin all cases,althougha
severelack of diffusionin two of the hallshinderedthe validationprocess.It was foundthat using
a frequency-constant edgecorrectionstrip to accountfor side and front absorptioncould lead to
significanterrorsin auditoriumabsorption prediction.
PACS numbers: 43.55.Ev, 43.55.Gx, 43.55.Fw, 43.55.Nd

INTRODUCTION tween reverberation chamber measurements and accurate in


situ seatingabsorptionmeasurements
in ten auditoria.
Of all the parametersin commonusein auditoriumde-
signtoday,reverberation time (RT) was the first to bc estab-
I. THE AVAILABLE METHODS
lishedand it is one of the mostsubjectivelyimportant.Be-
causethe RT in a hall is dominatedby the absorptionof the A. The required accuracy
scaringand audience,it is essentialthat thesecan be mea-
Several
authors
•-'3haveremarked
uponthe needfor
suredor predictedaccuratelyin the early stagesof design.
greateraccuracyin the predictionof auditoriumabsorption
Howeverthereis currentlyno wholly acceptedstandardtest coefficients.
The questionariseshereof how muchaccuracy
methodfor measuringseatingabsorption andthedataquoted the designerneeds.An estimatedanswercan be given from
in the literaturevary widely. the differencelimen for reverberationtime T obtainedby
The traditionalmethodof measuringseatingabsorption Seraphim(Ref. 4, pp. 505-507). Seraphimmeasuredthe
involvesplacinga small array of seatsin the centreof a smallestpercentagechange8T/T that could be correctly
reverberation chamber.The mainproblemwith thisarrange- identifiedby 75% of his subjectsfor reverberated
bandpass
ment is that it exaggeratesthe absorptionof the exposed noisewith variousvaluesof T andcenterfrequency.For the
front and side of the seatingarray,comparedto the larger midfrequencyoctavebetween800 and 1600 Hz, oøT/Tis
seatingblockscommonlyfoundin auditoria.This resultsin between 3% and 4% for values of T between 0.6 and 4 s. Of
errorsin the predictedauditoriumRT. course,$T/T is likely to be differentfor different source
A modification of the traditional reverberation chamber signals,so it is unfortunateit has not been measuredwith
testmethodfor seatingabsorption
wasproposed
by Kathand music.The differencelimen is likely to be largerfor music,
Kuhlaslongagoas19647Though
thisseemed
toofferthe sowe mightassumea figureof 5%. Sincethe RT in a hall is
possibilityof greateraccuracyby correctlyallowingfor the largelygovernedby theaudienceandseatingabsorption, one
absorption of theexposedfrontandsidesof seatingblocksin shouldaim to measureseatingabsorption to an accuracyof
auditoria,it has not beenwidely takenup. This may have 5%, at leastat midfrequencies.
beendueto the absenceof any large-scale validationof the
method,and a lack of understanding of the effectsof the
B. Averaged data
variousmeasurement parameters. The work reportedhere
aimsto clarifythissituation. Bcranek
• andKosten
6 havebothproduced
dataforthe
This papercoversthe investigationand validationof a averageabsorptioncoefficientsof occupiedand unoccupied
reverberation roommethodof measuring seatingabsorption. seating.The datawereaveragedfrommeasurements in many
First,the availablemeasurement and predictionmethodsare halls and are usefulfor estimatingRT in the early design
briefly reviewed.Then, data on the effectsof parametersof stages.Betarickshowedthat greateraccuracyis obtainedby
the measurement methodare presented.The choiceof the calculatingseatabsorptioncoefficientsbasedon absorption
optimizedparametervalues is justified by comparisons be per unit floor area rather than by the previouslyaccepted
absorptionper seat. Most subsequentseat absorptionwork,
includingthat validatedhere, hasusedthis sortof coefficient.
aJCurrently
atLeeds
School
of theEnvironment,
Brunswick
Building.
Leeds
MetropolitanUniversity.LeedsLS2 8BU, England.
However,the use of averagedata is not reliablefor finished
•'•Currenlly
atAmpAcoustics,
St.GilesHall.Pound
Hill,Cambridge
CB3 designsunlessone can be confidentthat the seatsto be used
0AE, England. in a particularhall will havean absorption
close(within5%)

879 J. Acoust.Soc.Am. 96 (2), Pt. 1, August1994 0001-4966/94/96(2)/879/10/$6.00 ¸ 1994Acoustical


Societyof America 879
to Beranek'sor Kosten'saverage.Sincemodernseatdesigns
can vary greatly,this will typicallybe doubtful.

C. Standard measurement methods

If we cannotvery accuratelypredictthe absorptionof a

-- __
Sp
'•
particulararea of chairsusingthe abovemethods,then we
mustmeasurethechairs.Currently,thisis doneusinga small
sampleof chairsin a reverberation chamber,thoughan at-
tempthasbeenmadeto devisea methodof predictingthe
absorption of chairsfrom the measuredabsorption of their
component
parts.
7
The aim of measuringthe randomincidenceabsorption
/
of a small sampleof seatsin a reverberationchamberis to
predictthe absorption
thata largeareaof the sameseatswill
exhibit when installed in an auditorium. The usual method is FIG. l. Schematicdiagramof an array of seatsin the cornerof the rever-
to placea rectangulararrayof the seatsnearthe centerof the berationchamber,showingthe stripsneededto correctfor pressuredou-
reverberation chamber,using the same row spacingas is bling.The subscript"/" indicatesthe edgeof the front row of seats,sub-
script"s" the edgeof the sideaisleseats,andsubscript"p" theplan areaof
foundin therealtheatre.
Standard
sample
areasvaryfroms the seatswithoutedgecorrection.
6.69m2 too 10-12m2.Hence thelargest
typical
sampleis
likely to be about24 chairs;in the currentwork, the standard
samplewasfour rowsof six chairs.Whenthisis scaledup to still be presentand so the measuredabsorptioncoefficient
a large block of seatsevery fourth row is in effect a front may still vary with samplesize. These effects have been
one, and every sixth seat is on the edge of an aisle. This discussed in Reft 12.

overemphasis of the absorptionof the front row and side This arrangementallowsus to measureabsorptioncoef-
aislesleadsto a predictedabsorption higherthanthatwhich ficientsfor threeconditions
of theseatingarrayshownsche-
will be exhibited in the auditorium. matically
in Fig.1: % withbarriers
covering
boththefront
andsideof the array,oqwith barrierscoveringthe sideonly
and ot2 with barrierscoveringthe front only. Theseare all
O. Bradley's method
related
to theplanareaSv shownin Fig.1. Twoabsorption
Recently,
Bradley
m published
details
of a seating
ab- coefficients
canthenbefoundforthefrontandside:oqand
sorptionmeasurement methodwhich attemptsto take ac- oq;theserelatetoareasSI andSs, respectively,
in Fig.1. If
countof the variationof seatingabsorptioncoefficientwith theareasof thefrontrow(Sya),sideaisles(Ss,,)andplan
samplesize--the failing of the traditionalmethod.This in- area(Sv,,)of a particular
largeseating
blockin thetheatre
volvesmakingmeasurements on five or six differentlysized are knownthen its absorption coefficient%, expressed as
arraysof a seattype.The variationof absorption coefficient the total absorptionthat would be measuredin situ divided
with the ratioof arrayperimeterlengthto area,E, is assumed by the plan area of the largeblock, can now be predicted
to be linear,so thata straightline may be fittedto the data. with the reverberationchamberabsorption coefficients
to be
This is extrapolatedback to the smallervaluesof E that
characterize largeseatingblocksin auditoria.Bradleyfound ot,•= at,+ otfSl./St,
a+ oqSs./St,
.. (1)
thatthismethodcouldgive accurateresultswhencompared Thusamshouldincorporate thecorrectamountof absorption
with measurements of the same seats in situ in auditoria.
dueto the exposedfrontrow andsideaisles.
Thoughit seemsthat this methodcan offer superiorac- The cornerplacingof the seatsis advantageous because
curacyover the methodsdiscussed above,it doesrequire it increasesthe effectivesize of the array.However,thereis
many testsfor eachtype of seatmeasured. a disadvantage: The SPL in a reverberant field is increasedat
theboundaries,
]3 so the absorption
coefficients
measured
E. Kath and Kuhl's method will be higherthan thosefound when the sampleis in the
center of the chamber.To compensatefor this, Kath and
Kath and Kuhl also thought that the over-valuingof Kuhlproposed
TMthattheabsorber
areas
(St,,S[, andSs)
front and side absorptionin the traditionalmethod was a usedin thecalculationshouldbe increasedby stripsof width
major reasonfor poor predictionof auditoriumabsorption M8 as shownin Fig. 1, where ;• is the wavelengthcorre-
coefficients.
Theyproposed
TManalternative
method
thatre- spondingto the centerfrequencyof the measurement.This
quiresfewer measurements thanBradley'sandyet maybe at extra absorbingarea accountsfor the increasein measured
leastasaccurate.In thismethodtheseatingarrayis placedin total absorptiondue to the increaseof up to 3 dB in SPL
the corner of the reverberationchamber,and the exposed closeto the wall. In a corner,there is an increaseof up to 6
edgesobscuredwith barriers,as in Fig. 1. Thoughit seems dB(seetheoverlap
in Fig.1),anda correction
of (M8)2 is
that the array is mirroredin the adjacentwalls of the cham- needed.
t Hence,theeffectivetestareasbecome
ber, thuseffectivelyincreasingits size, it is not effectively
infiniteas Kath and Kuhl thought.Diffractioneffectswill st,=(L+ X/8)(W+X/8), (2)
880 J. Acoust.Soc.Am., Vol.96, No. 2, Pt. 1, August1994 Davieset al.: Measuringauditorium
seat absorption 880
TABLE 1. Descriptivedatafor the ten hallsand their seats.

a Tmi
Hall Use d Nt
(s) N,, V-•)
(m •n• Ss•
2) (S•
•) (m •) Nb Back
rest Rear
• back
b Squab Under
b'c Arm
b'c squab rest
• Otmi
a
B1 c/rap 1.65 1811 1019 10929 441 68 40 1 CF metal CF metal ... 0.59
B2 c 2.13 702 468 6627 208 26 25 0 CF wood CF wood CF 0.68
C mp 1.98 •1000 616 14543 318 56 17 1 CF plastic CF metal plastic 0.56
D1 t 0.86 514 514 2488 242 62 12 1 CF wood CF wood CF 0.67
D2 t 0.88 900 734 3007 301 45 5 2 CF wood CF wood CF 0.69
G c 1.97 2500 2500 28750 1386 170 14 1 CF CW CF CW CF 0.66
H c/mp 2.02 1150 498 9571 177 29 25 2 CF metal CF metal ... 0.55
L t 2.21 700 700 12290 340 50 15 0 CF wood CF CW CF 0.67
M mp 0.67 624 241 1538 202 15 14 0 VF metal VF metal ... 0.37
O mp 1.46 669 669 8271 256 44 14 0 CF metal CF wood ... 0.68

ac=concert
hall,mp=multipurpose
hall,andt=theater.
•'CF=cloth
(woven)onfoam,CW=cloth(woven)onwood,andVF=imperviousvinyl on foam.
'The squabis thepaddedhorizontal
partof theseatwhichis satOB.

Ss=LH+Hk/8, (3) that could be removed from a hall or the number that were
installedif the hall was beingbuilt or refurbished.Addition-
Sf= WH+ Hk/8, (4)
ally,V is thevolume
of theauditorium,
Spais thetotalplan
whereall the symbolsare definedin Fig. 1. areaof the Nm seats,andSsais the total areaof the sidesof
all theblocksof seatscomprisingNm thatareexposedto the
II. VALIDATION METHOD sound
field.Similarly,
Sfais thetotalareaof thefrontrows
of all the blocksof seatscomprisingNm that are exposedto
It was decidedto attemptto validateKath and Kuhl's the soundfield. If the front row of a balconyblock was
methodby carryingout absorption measurements on seating
obscured
bya balcony
front,thenit wasnotcounted
forS/•.
in situ in auditoriaand on samplesof the sameseatsin a Here, N o is the numberof balconies.Table I also lists the
reverberationchamber.This was done for ten auditoria, so materialsforming the major componentsof the different
that RT measurements were made in all ten with the seats
seats. amid is the mid-frequency(averageof 500- and
presentandwith as manyas possibleremoved.
The reverberationchamberusedhasa volumeof 224 m3
1000-Hz
octaves)
valueof % measured
in thereverberation
chamber.
anda surface
areaof 226m2.It hasoneslanting
walltoaid
diffusion;the ceilingis horizontal.Elevenfixed, curveddif-
III. SOME REVERBERATION CHAMBER FINDINGS
fusers
witha totaltwo-sided
areaof 67.1m2weresuspended
in the room.Two loudspeaker positionsandten microphone During the validation process, many reverberation
positionswereusedto obtainaverageRTs.Sabine'sformula chambermeasurements weremadeto investigate the effects
was usedto calculatethe absorptioncoefficients.
The mea- of varyingparametersof the method.The effectsof sample
surementsystemusedin both the laboratoryand the audito- position,sampleconfiguration,
occupancyandcarpetare re-
ria centeredon a NorwegianElectronics real-timeanalyzer ported
in detailelsewhere?
Someof themoreinteresting
(type830). findingsare summarizedhere.Becauseof time limitations,it
The auditom were of varioustypes:Somewere multi- was not possibleto examinethe effectof everycombination
purposehallsin whicha significantportionof the seating of parameters on eachseattype.However,mostparameters
couldbe removed;otherswere newly built or refurbished. wereinvestigated for mostseats,usingan arrayof four rows
All datareportedhereare for unoccupiedauditoria.Consid- of six in almost all cases.Occupiedmeasurements were
erable efforts were made to ensure that each auditorium madeon two seattypes.
changedas little as possiblebetweenthe two RT measure- Sinceonlytheeffectsof thevariousparameters aredis-
ments,thoughit wasusuallynecessary to makesomeallow- cussedin this section,mostof the followinggraphsuse a
ancefor areasof carpetor curtainsbeingaddedor removed. standardmeasurement as a baseline:The seatswere placed
In someof the multipurpose halls,it was possibleto make in the cornerof the chamber,at a 900-ram row spacing,
the "full" and "empty"measurements on the sameday,so surrounded by barriers900 mm high,correctedfor pressure
thatnothingsignificant changedapartfrom the seats.Any doublingat the walls, but with no correctionsfor front and
necessary corrections for air absorption weremadebasedon sideabsorption.The barrierswere constructed from sheetsof
relativehumidityand temperature measurements madedur- 18-mm chipboard.
ing the RT test.
A. The spread of absorption coefficients
TableI listssomedescriptivedatafor the seatsand halls.
Threecategories are givenfor the hall usage:Concert,mul- Figure2 showstherangeof absorption
coefficients
from
tipurpose,or theater.Tmid is the averageof the RTs in the the ten seat types that were measuredin the standardcon-
500- and 1000-Hz octave bands in a hall with all the unoc- figuration.
Mostof theseattypeshavean absorption profile
cupiedseatspresent.
Here,N t is theseatingcapacity,
butN m not unlikethatof a homogeneous porousabsorber.
The most
is the numberof seatsthat were measured;i.e., the number notableexceptions are seattypesD2 and M. D2 was a start-

881 d.Acoust.Soc.Am.,Vol.96, No.2, Pt. 1, August1994 Davieset al.: Measuring


auditorium
seatabsorption 881
015
0.8 OlO

005
0.6

o
0.4
-0.05
0.2
-0.10

0
125 250 500 1000 2000 4000
-0.15
125 250 500 1000 2000 4000

Frequency, Hz Frequency, Hz

FIG. 2. Reverberation
chamberabsorptioncoefficientsof ten differentseat FIG. 4. Difference
caused
by barriers
of different
heights
to theat, of an
types,measuredusingthe "standard"versionof the new method. arrayof 24 seatsof type H, measured
at 800-ramrow spacing.One line is
for the seats in the center of the chamber,the rest are for the seats in the
corner.

dardwell-upholstered model,exceptthatits squabwashol-


low.Underthecushion,thissquabconsisted of 18-mmchip-
to predictabsorptionat onerow spacing froma measurement
board,thena 45-mm air gap,andthen5-mm plywood.This
of absorption at another.No attempthasbeenmadeto pro-
seemsto haveproducedan effectivelow-frequency resonant
ducea straightline regression for prediction,however.The
panelabsorber. Seattype M was coveredwith impervious
magnitude of theeffectof row spacingdepends, notsurpris-
vinyl, limiting its high frequencyabsorptionseverely.Seat
ingly,on how absorbent the seatsare. Hence,if figuresare
type B2 is picked out as representative of the well-
availablefor the absorptionof particularseatsat two differ-
upholsteredcloth-coveredseatsoften installedin concert
entrow spacings anda hall designerrequiresdatafor a third
halls. It was a standardmodelfrom a large manufacturer.
spacing,thenlinearextrapolation shouldbe a goodfirst ap-
proximation.
B. Row spacing
Figure3 showsthe effecton the absorption coefficient C. Barrier height
of seatB2 of varyingthe row spacingover a small range One aspectof Kath andKuhl's methodthat hascaused
commonlyfoundin auditoria.The effectis significantcom- someconfusionin the literatureis the questionof specifica-
paredto the magnitudeof one standarderror.It shouldbe tionof thebarrierheight.To investigatethe influenceof bar-
notedthatincreasingtherow spacingincreases the totalab- rier height, measurements were made on some relatively
sorptionof theseatarray,but the planareaincreases faster, lightly upholstered
seats,from hall H. Theseseatsare 840
and so the combinedeffect is to decreasethe absorptionco- mm high.Modularbarrierswere used,in the form of sheets
efficient.
of chipboard300 mm high and 18 mm thick. The barrier
Thoughthe lines in Fig. 3 are different,they are all surfacewas left untreated.The resultsare shownin Fig. 4,
highlycorrelated
witheachother(thelowestcorrelation co- presented
asthedifference
in a, caused
bythebarriers.
Con-
efficientis 0.9938).This indicatesthatit shouldbe possible siderthe mid andhighfrequenciesfor the seatsin thecomer
of the chamberfirst. When the barrierheightis increased
from0 to 300mm,thereis littledifference
in ap because
a
300-ram barrier obscuresonly the nonabsorbent
chair legs
820 mm-..• (theseseatsdid not havea tippablesquab).As the barrier
0.8
height is increasedfrom 300 through600 to 900 mm,
though,
a, decreases
firstat highandthenat midfrequen-
0.6
1000
cies. Now the barriersare progressivelyobscuringthe ab-
sorbingsurfacesof the squabandbackof the seats.
0.4 When the barrierheightis increasedfrom 900 to 1200
mmthereis littlechange
in apatmidandhighfrequencies.
0.2 This seemsreasonable,since there are no more absorbing
surfaceson the front and side of the array to be covered.
0 With an increaseto 1500 mm, however,thereis a significant
125 250 500 1000 2000 4000
jumpin % atmidandhighfrequencies.
Thebarriers
arenow
Frequency, Hz someway abovethe seattops,so it is possiblethatthe ab-
sorbingarrayis no longerin a diffusefield.As the barriers
are extended further and further above the absorber, any
FIG.3. Effectof rowspacing
onthe% of anarrayof 24 seats
of typeB2,
measuredin the cornerand surroundedby 0.9-m barriers. soundraysenteringthis enclosure
are lesslikely to leave,

882 J. Acoust.Soc.Am.,Vol.96, No.2, Pt. 1, August1994 Davieset al.: Measuring


auditorium
seatabsorption 882
and so the apparentabsorbingpower of the array will in- heightof 900 mm, two setsof barrierswerecommonlyused:
crease.This situationis analogousto thatof seatingundera 900 mm high for unoccupiedmeasurements and 1200 mm
deepbalconyoverhang. CremerandMfiller (Ref. 4, p. 263) high for occupiedwork.
saythatsuchseatingis no longerin thediffusehall field,and
recommend ascribingan absorption coefficientof 1.0 to the
D. Edge corrections
openingunderthe overhang,as one might to an openwin-
dow. The high frequencyabsorptionat 1500 mm is also An awkwardproblemin measuringthe absorptioncoef-
probablydue in part to the surfaceabsorptionof the un- ficient a of a three-dimensionalobject in a reverberation
treatedchipboard. chamberis that the measuredcoefficientis found to vary
There is also evidenceof an increasein low-frequency with the size of the sample.If the absorbingedgesof a
absorptiondue to the barriersin Fig. 4. The most obvious sampleare exposed,thenthis variationis due to two compo-
possiblecausesare that the barriersact as panelabsorbersor nents:(i) The absorptionof the front row and side of the
that the enclosurethey form aroundthe seatscontributesex- seatingarray,causinga to increaseas sampleareaincreases;
tra absorption.If the measurement
is conductedwith the and(ii) diffractionof soundwavesat the edgesof the array,
seats in the center of the chamber, the enclosure is similar, againcausinga to increasewith area.
but the requirednumberof barriersis doubled.In Fig. 4, the In thepast,acousticians haveusuallymadean allowance
low-frequencyabsorptionof barriersused in the centeris for the absorptionof the sidesandfrontof a seatingblockby
roughlydoublethatof barriersusedin the corner.Hence,the increasingthe seatingarea usedin calculationsfrom the ac-
anomalousabsorptionseemsdue mainly to the barriersact- tual plan area.Thesecorrectionsusuallytake the form of a
ing as a panelabsorber. stripof constantwidth intowhichtheplanareaof theseating
blockis supposed to extendat all its exposedsides.Because
the width usedis constantwith frequency,the assumption is
1. Minimizingbarrierabsorption made that the exposedsideshave an absorptioncoefficient
proportionalto that of the plan area of the block.Also, the
The unwantedlow-frequencybarrierabsorptionis less-
samestrip width is usuallyusedfor the exposedfront row
enedby usinga positionin the cornerratherthanthe center
andsides,so theabsorption coefficientsof thefrontandsides
of the reverberationchamberfor seat absorptionmeasure-
are assumed to be the same. However, even once these as-
ments.Reducingit furtherseemsmoredifficult. Becauseof
sumptionshavebeen made,it is not clear what the width of
the panel absorption,it would be advantageous to suppress
the strip shouldbe. Most designersuse the value advocated
the mostprominentmodesin the barrieror movethemout of
byBeranek,
whichhaschanged
from161 to5 0.5m.
the frequencyrange of interest.Unfortunately,during this
When the absorptioncoefficientsof the front and sideof
work, suitablematerialswere too expensiveto buy in an area
seatingblockswas measuredusingKath and Kuhl's method,
largeenoughto form seatingbarriers.
it was found that the resultswere quite differentfrom the
If the problemcannotbe tackledat source,thena crude
absorption coefficientof theplanareaof thesameblock.For
correctioncan be madeto seatingabsorption measurements
clarity,theseresultsare expressed as edgecorrectionstrip
by subtracting the absorptioncoefficientof the barriersmea-
suredseparately,from the absorptioncoefficientof the seats widthskf andk.,for thefrontandsideof theblock,respec-
tively:
with barriers.This is not entirely satisfactoryfor two rea-
sons.First, the "barriersonly" absorptionmeasurement will kœ=L(a•/at,- 1), (5)
not be very accuratedue to the low absorptionbeing mea-
sured,and so the correctedabsorptionwill also be inaccu- kv= W(ce2/a
p- 1), (6)
rate;andsecond,it takesno accountof any possibleinterac- where the symbolsrelate to Fig. 1.
tion between the barriers and the seats. Edgecorrectionstripwidthsweremeasured for samples
It should be rememberedthat the barrier absorption of six differentseattypes(BI, B2, DI, G, H, and L). The
problemis not as bad as it might be, since it occursat the rangeof thesedata is shownin Fig. 5: The spreadis large
lowerendof thefrequencyspectrum. Low frequencyabsorp- and the meansare far from constantwith frequency.If a
tion measurements are always less accurate,especiallyin frequency-constant figureis insistedupon,then0.5 m seems
roomswith less then perfectdiffusionlike auditoria.Also, a betterchoicethan I m. Even so, we might expectthis
fortunately,the humanear is alsolessdiscriminating in this approximation to introduce significant
errorsintothepredic-
region:Cremerand M/filler(Ref. 4, pp. 507-509) quotere- tion of auditorium RTs. It must be concluded that it is far
sultsfrom Plengeshowingthatthe subjectivelimen for rela- betterto measure
thesideandfrontabsorption of a sampleof
tive changein RT increases
with decreasing
frequencybelow seatsratherthan rely on a frequency-constant
edge correc-
1 kHz. tion.
The unwanted high-frequencybarrierabsorption seems The question of howto treataisles,asopposed to com-
easierto dealwith:The barriersshouldbe at leastashighas pletelyexposedsides,remains.If an aisleis I m wide, then
theseatingplusanyauditors, butexcessiveextraheight(say, it seemscorrectto say that the seatson eitherside are not
more than 100 mm abovethe top of the absorbers)shouldbe fully exposed,due to shadingfrom the seatson the opposite
avoided.
Notethatthelowest
values
of % at midandhigh side. If the aisle is carpeted,as it usuallyis, then sound
frequenciesin Fig. 4 are for 900-mm barriers.Sincemostof energy reflectedfrom the floor onto the seatsshouldbe less
the seat typesmeasuredin this work were a little below a than that encountered in the reverberation chamber. Both

883 J. Acoust.Soc.Am.,Vol.96, No. 2, Pt. 1, August1994 Davieset aL: Measuring


auditorium
seatabsorption 883
(a)
1.0

1.5
0.8

0,6
1.0

0,4

0.5 02

(b)
125 250 500 1000 2000 4000
0,8

Frequency, Hz
0.6

FIG. 5. Minimnm,meanand maximumedgecorrectionstripwidthsat each 0.4


frequency
fortheexposed
frontrow(kf) andtheexposed
sidearea(k•) of
six differentseattypes. /' infinite
0.2 small
I auditorium
theseeffectswouldtendtoreduce
theabsorbing
powerof the 0
aisleseats,but bothwill probablybe frequencydependent. •
125 •
250 i
500 i
1000 2010040•00
Frequency, Hz

IV. COMPARISON OF REVERBERATION CHAMBER


AND AUDITORIUM MEASUREMENTS FIG. 6. Absorption
coefficient
of (a) seating
B2 and(b) seating
G measured
in situ in the auditoriumandby threedifferentreverberation
chambermea-
For each hall, many reverberationchambermeasure- surementmethods. The errorbarsrepresent_* one standarderror.
mentsproduced by combinationsof differentmeasurement
parameters were comparedwith one auditoriummeasure-
Figure6(b) showsthe comparison for hall G, a large
ment.The auditoriumabsorption coefficientwas expressed
modernconcerthall.Again,the auditoriumabsorption coef-
as the total measuredabsorptionattributedto the seatsdi- ficient lies between the "infinite" and "small" measure-
videdby thetotalplanareaof theseating.Acrossall tenhalls
ments.Overall,the "large finite" line is the bestmatch.At
therewas one reverberationchambermeasurement configu-
high frequencies,the traditionalmeasurement has signifi-
rationthatconsistently
produceda betteragreementthanany
other:
cantlyoverestimatedthe absorption in the hall. The measure-
mentwas performedwith a 24 seatsample;with smaller
A rectangular arrayof seatswasplacedin thecornerof
samplesthe excesswouldbe evengreater.At low frequen-
the chamberat the auditoriumrow spacingand surrounded
cies,particularly
250 Hz, the "largefinite"dataaresignifi-
by unabsorbent barriers
0.9 m highfor unoccupied seatsand
cantlylower than the auditoriumdata.This was probably
1.2 m high for occupiedseats.The absorption of the plan
causedby wood panelingbeingaddedto the hall during
areawas measuredandcorrectedfor pressuredoubling.Two
construction.Thoughit was difficult to policethe construc-
more measurements were made, with the barrierscovering
the sideandfront of the arrayonly.A separatemeasurement
tionschedule completely,approximately 300m2ofpaneling
over an air gap was addedbetweenthe two RT measure-
of low-frequency barrierabsorptionwassubtracted from all
ments.No absorption data are availablefor the particular
thedata.a,,,wasthencalculated from Eq. (1). In calculating
panelsused,butcoefficients of 0.50 at 250 Hz (Ref. 17) or
S.•, in Eq. (1), thebestresultswerefoundif aislesbounded
0.42at 125Hz (Ref.5) suggest thatabout150m2 of total
by seatingon bothsidesweretreatedas oneexposedside
absorptionwasadded.Thiswouldaccount for an extra0.11
area.The valuesof S.,, in TableI reflectthis.
in the auditoriumabsorption coefficientat 250 Hz in Fig.
The followingsections discussthecomparison for each
6(b).
hall in detail.In eachgraph,"largefinite" refersto amcal-
culatedas described
above,"infinite"refersto ap and
"small" refersto the traditionalmethodwith the array in the
center of the chamber with no barriers.
B. The concert/multipurpose halls B1 and H
A. The concert halls B2 and G
The agreement
betweenthe "largefinite" reverberation
Figure6(a) showsthe comparison
for hall B2. As ex- chamberdata and the auditoriumdata is very good for hall
pected,the auditorium
absorption lies between H, as Fig. 7(a) shows.Again, the auditoriumline lies be-
coefficient
the reverberationchamber data for the "infinite" configura- tweenthe "infinite"data(whichincludeno sideabsorption),
tion(whichincludeno sideareaabsorption)
andthe "small" and the "small" data (which includetoo much side absorp-
configuration
(whichincludetoo much).It is quitewell tion).The accuracyof this measurement was compromised
matchedby the "large finite" curve. by the fact that only 43% of the seatsin the hall were re-

884 J. Acoust.
Soc.Am.,Vol.96, No.2, Pt. 1, August1994 Davieset al.:Measuring
auditorium
seatabsorption 884
1.4 1.2

1.2 ...... large finite (eX.m)


1.0
infinite(c%)
small
1,0
auditorium 0.8
0,8
0.6
0,6
0.4
0,4
0.2
0.2 (a)
' ' (a)
0
0

1.0
1,0

0.8
0.8
/./ 0.6
0.6
0.4

/"•',• (b)
0,4
0.2
0,2
0 I I I I
0
125 250 500 1000 2000 4000 large finite (O•m)
infinite(o•p)
Frequency, Hz 0.8 small
auditorium
0.6
FIG. 7. Absorptioncoefficienlof (a) seatingH and(b) scatingB1 measured
in situ andby threedifferentreverberatkm chambermeasurement methods. 0.4

0.2
movable. Nevertheless, the reverberation chamber measure-
0
menthasaccuratelypredictedthe averageabsorptioncoeffi- 125 250 500 1000 2000 4000
cient of the seats in the hall.
For theotherlargeconcert/multipurpose auditorium,hall Frequency, Hz
BI, the agreementis good at low and midfrequencies, as
shownin Fig. 7(b).At thesefrequencies, theauditorium line FIG. 8. Absorption
coefficient
of (a) seating
C, (b) seatingO, and(c) seating
lies betweenthe extremereverberation chamberlines,sug- M measuredit! situ and by threedifferentreverberation
chambermeasurc-
meut methods.
gestingthat it is a reasonableresult from a diffuse sound
field. At 2 kHz and particularlyat 4 kHz, however,the seat
absorptionin the auditoriumis higherthan the laboratory
measurements. This was one of the measurements made in largeblocksin the auditorium,the increasein absorption
wherestripsof carpetin the aisleswere removedalongwith coefficientdue to exposedsidesis probablysmall. Hencea
the seats.No sampleof this carpetwas availableto measure "large finite" line for this hall would not be far above the
"infinite" one.
separatelyin the reverberationchamber,so a correctionwas
made using a measurement of a typical carpetsampleto In hall C, the auditoriummeasurement was complicated
hand.It may be that this substitutedcarpetdid not absorb by the fact that the folding seatswere fixed to retractable
high frequencysoundas effectivelyas the materialin hall bleachers.Since neither could be removed from the hall, the
B1. "empty" RT measurementwas performedwith the seats
foldeddownandthebleachersfully retracted,andthe "full"
C. The modern multipurpose halls C, O, and M measurement with the bleachers extended and the seats erect.

The ratherhighauditoriumabsorptioncoefficientat 4 kHz is


Thesehallshavein commona multipurpose function. As
perhapsdueto the unfinishedsurfaceof the bleachersthem-
well as,andperhapspartlybecause of this,theyalsohavein
commonlightweightchairs and a poor state of diffusion. selvesabsorbingenergywhen extended.
This hasled to problemsin predictingthe in situabsorption In Fig. 8(b) the agreementbetweenauditoriumand re-
coefficientof theseatsin two halls.Nevertheless,in the larg- verberationchamberabsorptioncoefficientsfor hall O is rea-
est of the three, hall C, the agreementbetweenthe audito- sonablygoodin all frequencybandsexcept1 and2 kHz. The
rium and reverberationchamberabsorptioncoefficientsis sameis true of hall M in Fig. 8(c): in both hallsthe seats
quitegoodup to 2 kHz, asshownin Fig. 8(a). No reverbera- seemto absorblittle energyat 2 kHz. It is thoughtthesedips
tion chamberdata for the side and front absorptioncoeffi- may be due to the soundfields in thesehallsbeing so badly
cientsof theseseatswere measured,so a "large finite" ab- diffusedat 2 kHz that little soundenergyactuallystrikesthe
sorptioncoefficientcannotbe calculated.Becausethe seats seats.This explanationis supportedby the observationthat
are quitelightly upholstered, andbecausethey are arranged decaycurvesin bothhallssaggedbadlyat thisfrequency.

885 J. Acoust.Sec.Am.,Vol.96, No. 2, Pt. 1, August1994 Davieset al.: Measuring


auditorium
seatabsorption 885
1,0
TABLE fl. Percentagedifferencebetweenreverberationchamberand audi-
toriumabsorptioncoefficientsat 1 kHz, comparedwith the statisticaluncer-
0,8 taintyin the auditoriumdataß

Standard error Difference Difference


0.6
in auditorium for "large for "small"
d
Hall measurement(%) finite" measurement(%) measurement(%)
0.4
B2 -+7 +4 +27
G -+5 +6 +30
0.2
H +15 -6 +6
B1 +-7 1 + 10
0
O +10 +32 +52
M _+37 +3 +26
0.8 L +7 +6 +26

0.6
" 1 ' D1
D2
+_6
+-7
-I
7
+13
+ 14
d
0.4
..
0.2
tion chamberlines at low frequencies.These discrepancies
are similar to the one found for hall G, which was also under
0
construction.
The reasonsseemthe same:In hall I approxi-
(c) •
1.0 mately160m2of panelling
wasprobably
addedduringseat
installation.
Thiswouldprovide
about80 m2 of low fre-
0.8
quencytotal absorptionand would accountfor an extra 0.24
d O.6 in the auditoriumabsorptioncoefficientat 125 Hz in Fig.
9(a).Similarly,
inhallDI approximately
110m2ofplywood
0.4 infinite probablyintroducedan extra0.23 in the auditoriumabsorp-
" small
0.2 tion coefficientat 125 Hz in Fig. 9(b).
auditorium
The final hall comparison is for theaterD2 (DI after
0
125 250 500 1000 2000 4000 refurbishment) and it appearsin Fig. 9(c). This time, the
agreementbetween the "large finite" data and the in situ
Frequency. Hz measurementis not quite so good.The matchis best at mid
frequencies;at low and high frequencies,the reverberation
FIG.9. Absorption
coefficient
of (a) seating
L, (b) seating
D1,and(c) chambercoefficientis too high.The mostlikely problemat
seatingD2 measuredin situ and by three different reverberationchamber
measurement methods.
lowfrequencies
isagain
panelling.
Thistime,about
70m2of
wood was probablyremovedat the time the seatswere in-
stalled.This would producea shortfallof about0.12 in the
D. The theaters D1, D2, and L auditoriumabsorption coefficientat 125 Hz in Fig. 9(c). At
high frequencies,the auditoriumfield may be less diffuse
In the last groupof halls to be investigated,the audito- than that in the reverberation chamber. Table I shows that
rium calculationis slightly complicatedby the problemof after refurbishment,
the volumeper seatof D2 stoodat only
the volumeof the stagehouse or flytower.Being theatres,all 3.3m3.Thisisrather
lessthanthe8-10 m'•commonly
used
three of thesehalls have a prosceniumarch couplingthe asa rule of thumbfor concerthalls.It is probablethattheater
volume of the flytower over the stageto the volume of the D2 was just too full of highly absorbingseatsto achievea
auditoriumwhere all the seatsare installed.The problemwas diffusefield at high frequencies.
minimized in hall D both before (D1) and after (D2) refur-
bishment,since the area under the prosceniumarch was
fairly smalland it was coveredby a heavyfire curtainduring
all measurements. It was therefore assumed that the volume
E. Midfrequency accuracy
of the stagehouse
did not play a part in the soundfield in the
auditoriumof hall D. In hall L, conversely,the RT measure- Table II shows the difference between the reverberation
mentswere made during construction.In this theatre,a very chamber and auditorium absorption coefficients at 1 kHz,
large flytower was coupledto the auditoriumby a large expressedas a percentageerror relative to the auditorium
opening.Sincetherewas no evidenceof a dualdecayratein data. In the nine halls for which both "large finite" and
the recordeddecaycurves,it was assumedthat the coupling "small" data exist, the "large finite" data are almostalways
betweenthe two spaceswas perfect,and the volumeof the muchmore accurate.Only hall O is a long way off the ten-
flytowerwas includedin the absorption calculation. tative 5% accuracyrequirement.The discrepancyin the
For boththeatersL and D1, in Fig. 9, the auditoriu m "large finite" data for the other halls could perhapshave
absorptioncoefficientis predictedwell by the "largefinite" beenreducedby improvingthe testmethod,but it is in most
reverberationchamberdata for mid and high frequencies.In caseslessthan the statisticaluncertaintyin the auditorium
both cases,the auditoriumline is well above the reverbera- measurement.

886 d. Acoust.$oc. Am., Vol. 96, No. 2, Pt. 1, August1994 Davieset al.: Measuringauditoriumseat absorption 886
1,2 V. CONCLUSION
(a)
1.0 After validatingthe barriermethodfor measuringseat-
ing absorption in ten auditoria,it can be concluded thatit
0,8
gaveclosepredictions in eightof them.In all tenhalls,rang-
0,6 ingfroma largemodern concert hallto a smalltheater, any
deviationsfroma goodprediction notattributable to random
0,4 errorcanbe explainedby problems in thevalidationprocess
or hall measurements themselves.These mostly take the
0.2
form of uncertaintiesdueto the presence of extraabsorbing
0
materialin the auditoriumduringthevalidation.In two halls,
a largedip in the auditoriumseatabsorptioncoefficientwas
0.8 foundthatwasnot predictedby any reverberation chamber
measurement.
There is evidencethat very poor diffusionin
0.8
the halls was the cause.
In the course of the reverberation chamber measure-
0.4 ments,it wasfoundthattheuseof a frequency-constant edge
Beranek
correctionstripcouldleadto significant
errorsin auditorium
G,
0.2
...... B2, RT prediction.Measurement of the absorptionof the front
and sides of seats should be used to obtain more accurate
0 • estimates. It was also found that the barriers used in the
125 i
250 i
500 i
1000 2010040100
presentmethodcouldform a resonantlow-frequency ab-
Frequency. Hz sorber.The effectsof thisproblemwerereducedby makinga
separatemeasurement of barrierabsorption.
Furtherinvesti-
FIG. 10. (a) Minimum, meanand maximumauditoriumabsorptioncoeffi- gationof the effectsof the barrierconstruction
may allow
cients for nine unoccupiedupholsteredseatscomparedwith Beranek's additional reduction of the effect.
average.
s(b)Values
ofapfortwotypes
ofoccupied
upholstered
seats
com- In all ten halls, the traditional reverberationchamber
paredwith Beranek'saverage. measurementmethod overestimatedthe in situ absorption
coefficient. This means that a reverberation time calculation
fora newhallbased
onsucha measurement
isverylikelyto
F. Comparison with Beranek give too low a value.Becausethe overprediction of the tra-
Figure10(a) showsthe rangeof unoccupied auditorium ditional measurement is .quite large, and the seatingis the
absorption
coefficients
takenfromFigs.6-9, theirmean,and majorabsorber in a hall, the deviationfrom the designvalue
of RT wouldprobablybe greaterthenthe subjectivediffer-
Beranek's
data5forunoccupied
seats.
Theunusual
datafrom
encelimen.The presentmethodwill alsogive moreaccurate
hall M have been excluded.Consideringthe range of the
resultsthan the use of either Beranek'sor Kosten'saverage
currentdata,the agreement betweenthe meanandBeranek's
absorptiondata will allow in almostall cases.Finally,the
valuesis quite good up to 1 kHz. At higherfrequencies,as presentmethodachievesan accuracyat leastequalto the
Bradley
explains,
mBeranek's
absorption
dataarequitepos- morelengthyone proposedby Bradley.It is thereforepro-
sibly affectedby differencesin air absorptionbetweenthe posedthattheoptimizedbarriermethodof measuring seating
many hall measurements he used.Beranek'sdata are below absorptionshouldbe adoptedfor all designswhereaccurate
the meanof the presentdataat all frequencies.
This may be reverberation
time predictionis desired.
becausemoderntheatreseatinghas slightly more padding
thanthe onesformingthe bulk of Beranek'sdata.The spread ACKNOWLEDGMENTS
of datain Fig. 10(a) emphasizes thatthe useof an average
Thanksare due to the managers,acousticconsultants,
absorption coefficientshouldbe for roughearly designfig-
architectsand seatingmanufacturers
from the halls mea-
uresonly, at leastfor unoccupiedRT predictionin a hall.
sured.This work was financedby the British Scienceand
The comparisonof the two setsof occupieddata from
EngineeringResearchCouncil, under contractno. GR/F/
thereverberation
chamber
withBeranek's
average
5 is very 05251.
interesting.
Figure10(b) showsthatBeranek'scoefficientis
very closeto being the mean of the two measureddata sets
U. KathandW.Kuhl,"Messungen
zurSchallabsorption
yonPersonen
auf
for seat typesB2 and G. Althoughthe occupieddata pre- Ungepolsterten
Stuhlen,"Acustica14, 49-55 (1964).
sentedhereare limited,theypoint to the possibilitythat Be- 2H. Kuttruff,RoomAcoustics
(Elsevier,
London,1991),3rded.
ranek's average absorption coefficient may be accurate 3M. Nagata,"Whatwe havelearnedfromthe listening
experiences
in
concerthalls--physicalproperties
andsubjective
impressions
of five con-
enoughto give goodpredictionsof occupiedhall RT, at least certhallsin Tokyo,"Appl.Acoust.31, 29-45 (1990).
for sometypesof seat.This supportsthe ideathatthe absorp- L. CremerandH. A. Muller,Principles
andApplication•
ofRoomAcous-
tion of occupiedupholsteredseatsis dominatedby the ab- tics (AppliedSciencePublishers,
Barking,England,1982).
SL. L. Beranek,
"Audience
andchairabsorption
in largehalls.11,"J.
sorptionof the occupantsand so shouldnot vary muchover Acoust. Soc. Am. 45, 13-19 (1969).
differentseattypes. aC. W. Kosten,"New methodfor thecalculation
of thereverberation
time

887 J. Acoust.Soc. Am., Vol. 96, No. 2, Pt. 1, August 1994 Davies et al.: Measuringauditoriumseat absorption 887
of hallsfor publicassembly,"
Acustica16, 325-330 (1965). Acoust.Soc. Am. 91, 1514-1524 (1992}]," J. Acoust.Soc. Am. 93,
7T.Fukuchi
andK. Fujiwara,
"Sound
absorption
areaperseatof uphol- 2238-2240 (1993).
steredchairsin a hall," J. Acoust.Soc.Jpn.(E) 6, 271-279 (1985). •3R.V. Waterhouse,
"Interference
patterns
in reverberant
sound
fields,"J.
8ASTMC 423-84a--"Measurement
of sound
absorption
andsound
ab- Acoust.Soc.Am. 27, 247-258 (1955).
sorptioncoefficients
by the reverberation
roommethod"(AmericanSoci- 14U.Kath andW. Kuhl, "Einflussyon Streuflache
und Hallraumdimen-
ety for theTestingof Materials,Philadelphia,
1984).
sionenauf den Gemessenen
Schallabsorptionsgrad,"
Acustica11, 50-64
ISO354•"Measurement
of sound
absorption
in a reverberation
room," (1961).
(International
Organization
for Standardization,
Geneva,1985).
mj. S. Bradley,
"Predicting fromreverberation •5W.J. Davies,"TheEffects
theaterchairabsorption of Seating
on theAcoustics
of Auditoria,"
chamber measurements,"J. Acoust. Soc. Am. 91, 1514-1524 (1992). Ph.D. thesis,SalfordUniversity,England(1992).
U. KathandW. Kuhl,"Messungen yonPolsterstu- 16L.L. Beranek,
zurSchallabsorption Music,Acoustics
andArchitecture
(Wiley,NewYork,
hlen mit und ohnc Personen,"Acustica 15, 127-131 (1965). 1962).
•2W.J. Davies,
Y. W. Lam,andR. J. Orlowski, on'Predicting •7A. Lawrence,Acoustics
"Comment and the Built Environment
(Elsevier,London,
theaterchairabsorption
from reverberation
chambermeasurements'
[J. 1989), p. 234.

888 J. Acoust.Soc.Am.,Vol.96, No. 2, Pt. 1, August1994 Davieset aL: Measuringauditorium


seatabsorption 888

Вам также может понравиться