why is it so hard to design a concert hall with excellent acoustics? · 2016-11-03 · receiver...

ProceedingsofACOUSTICS2016 9-11November2016,Brisbane,Australia

ACOUSTICS2016 Page1of15

Why is it so hard to design a concert hall with excellent acoustics?

Tapio Lokki

DepartmentofComputerScience,SchoolofScience,AaltoUniversity,Helsinki,Finland

ABSTRACT Concert hall acoustics is a multidisciplinary research topic, although often people think that it is onlymeasuring andanalysingimpulseresponses.Thispapergoesbeyondtheroomimpulseresponsesandexplainsinparticularthefeaturesofsoundsourcesandlistenersinaconcerthall. Inaddition,thepapercollectstherecentresearchresultsonsubjectiveevaluationsanddiscusseshowhallsaffectthemusicandtheimpressionsthatlivelisteningproducestolisteners.Theaimofthepaperistoexplainwhytheacousticsofaconcerthallarehardtodesignwithconventionalmethodsandwhataretheissuesthatwouldneedmoreattentioninfutureresearch.

1. INTRODUCTION Theacousticsofconcerthallshavebeenstudiedscientificallyover100years.Still,whenanewconcerthall is

openeditsacousticalqualityismoreorlessamystery.Whyisitsohardtodesignexcellentacousticsforacousticmusic? Could we just measure existing halls to understand their acoustical deficiencies and improve the nextdesign?Thispapersuggestssomeofthereasonswhytheresearchinthisfieldhasnotbeenabletodisentangleallproblemsanddiscussessomeissuesthatneedmoreresearch.

One of the major challenges is a common disagreement on what constitutes excellent acoustics. The widevariation among preferences poses the issue of whether a new room should aim for a strong, enveloping, andreverberant sound, or are clarity and definition better design criteria. Very often acousticians refer to objectiveparameters,defined intheISO3382-1:2009standard,asdesignguidelines.Thestandard listsparametersthatcanbe computed from measured impulse responses. Although, the standard is quite recent the basis for theparameterswerepresentedalmost30yearsagoinanexcellentpaperbyBradley(1990).Thestandardrecommendsmeasuring impulse responses with an omnidirectional sound source and an omnidirectional or a figure-of-eightmicrophone to capture sound in the audience area. In addition, the standard recommends computing theparametersintheoctavebandsfrom125Hzto4kHz.Moreover,peopleoftenwanttodescribetheacousticsofaconcert hallwith only a fewnumbers and therefore the standard recommends using the averages of all source-receiver combinationsaswell asaveragesatmidor lowandmid frequencies (seeTableA.1 “Acousticquantitiesgroupedaccordingtolisteneraspects”in(ISO3382-12009)).

Figure 1 illustrates the information that can be extractedwhen following the ISO3382-1:2009 standard. Thefigurevisualizesprincipalcomponentanalysis results for14concerthallsmeasuredwith24sourceand5 listenerpositions. Based on this analysis it is obvious that shoebox type concert halls aremore reverberantwith strongsound and good envelopment. In contrast, vineyard and fan-shaped halls have a clearer sound but with lessreverberation, strength, and envelopment. It should also be noted that over 90% of the variance in the data isexplained by two first principal component dimensions. If the same analysis is done without figure-of-eightmicrophone measurements (i.e. only EDT, C80, and G, the most commonly used parameters) over 97% of thevariance in the data is explained with two dimensions. On the one hand, this means that these five (or three)parametersaresufficienttoexplainthemaindifferencesbetweenthesehallsobjectively.Ontheotherhand, it isobvious that these halls sound quite different, even among shoebox halls, and the parameters fall short indescribingindetailtheacousticsofthesehalls.

9-11November2016,Brisbane,Australia ProceedingsofACOUSTICS2016

Page2of15 ACOUSTICS2016

Figure1:Principalcomponentanalysisof14Europeanconcerthallswithobjectivedataaccordingto“TableA.1”in

ISO3382-1:2009standard.Allparametersareaveragesof24sourceand5listeningposition,i.e.,120measurements.Left:PCaxes1and2explainingintotalof91.8%ofthetotalvariance.Right:PCaxes1and3.

2. THE PURPOSE OF A CONCERT HALL Aconcerthall isaspecialvenuededicatedtoperformingand listeningtomusic.Aconcerthall isolatesother

soundstoenablethesilencethatmusicrequires.Inaddition,thehalladdsreverberationtothesoundemittedbymusical instruments to make the musical performance more loud, rich and full-bodied. The reverberation alsocolors the sound by emphasizing some frequenciesmore than some others, i.e., the hall changes the perceivedtimbreofinstrumentsounds,andmanyfamoushallshavetheircharacteristic"sound"or"timbre".

Engineers often use a source-medium-receivermodel to explain a process. In case of a concert hall such amodelispresentedinFigure2.Themajorityofroomacousticsresearchisconcentratedonlyonthemediumpartofthiscommunicationprocess.Forexample, thepreviouslymentioned ISO3382-1:2009standardacknowledgesthatallmeasuresarefrequencydependentandthereforeoctaveoronethird-octavebandanalysesarerecommended.Moreover, the standardeliminatesvariability inmeasurementsby requiring thatmeasurement loudspeakersandmicrophonesareomnidirectional, i.e.,emitting/capturingthesameamountofsoundenergyfromalldirectionsatall frequencies. This is reasonable for the repeatability and for the comparison ofmeasurements performed bydifferent researchers,but it doesnot represent the real situationwithmusical instrumentsandhuman listeners.Theirfrequencydependentdirectivitiesaswellasleveldependentspectraareextremelyimportant,butveryhardtomeasurerepeatedly.

Oneimportantthingtonoteisthatpeoplealsogatheratconcerthallstomeetfriends.Formany,aconcertisprimarilyasocialeventandtheacousticsofthehall isonlyonepartofthewholeexperience.Thearchitectsandacousticians need to pay attention in particular to flows of people and design of lobbies to enable socializingwithout problems. An example of unsuccessful design is the newHelsinkiMusic Centre (Pätynen& Lokki 2015),wherethelocationsofthecloakroomandrestroomsareimpractical.Moreover,flowofpeopleisreallyslowtakingalmost five minutes for the last spectator to reach the lobby at intermission. Such logistic issues are reallyimportant,althoughthispapernaturallyconcentratesindetailonlytotheacousticsofconcerthalls.

-3 -2 -1 0 1 2 3

-2-1

01

23

14 European Concert halls

Dim 1 (73.13%)

Dim

2 (1

8.64

%)

AC

BB

BKBPCP

FT

LS

MG

MHMT

SB VMWS

ISO_EDT

ISO_GISO_C80ISO_ASWISO_LEV

Shoebox hallsVineyard hallsFan-shaped hallsOval-shaped hall

-3 -2 -1 0 1 2 3-3

-2-1

01

23

14 European Concert halls

Dim 1 (73.13%)

Dim

3 (5

.19%

)

AC

BB

BK

BPCPFT

LSMG

MHMTSB

VM

WSISO_EDT

ISO_G

ISO_C80

ISO_ASW

ISO_LEV

Shoebox hallsVineyard hallsFan-shaped hallsOval-shaped hall



Figure2:Source-Medium-Receivercommunicationprocessforaconcerthall.

3. SOURCES AND RECEIVER IN A CONCERT HALL

3.1 Musical instruments as sources The sound sources on the stage of a concert hall aremusical instruments andhuman voices. The size of an

ensemblevariesfromasoloinstrument,e.g.,agrandpiano,tolargeorchestrasandchoirsconsistingofhundredsofmusicians.Regardlessoftheensemblesize,everysinglesoundsourcesharestwoimportantfeaturesfromaroomacousticspointofview.Theyhavea frequencydependentradiationpattern(Meyer2009,Pätynen&Lokki2010)andaleveldependentspectrum(Luce1975,Meyer2009,Pätynenetal.2014).

3.1.1 Directivity of instruments Musicalinstrumentsradiatesoundtoalldirections,butthespectrumvarieswithdirection.Thelowfrequencies

aremoreorlessomnidirectionallyradiatedwhiletheradiationpatternbecomeslessuniformathigherfrequencies.Theradiationpattern isdefinedbythegeometryandthepropertiesof theradiatingpartsof the instrument.Forexample, the main radiators in string instruments are the f-holes and the top plate under strings. In brassinstruments, all sound is emitted from the bell and the size of the bell defines the frequency belowwhich theinstrumentisomnidirectional.Woodwindshavevaryingradiationpatternsassoundisemittedfromtheopentoneholesandfromthebell.Theruleofthumbisthatat lowfrequencieswoodwindsareomnidirectionalandathighfrequenciesmostof the sound isdirected in thedirectionof thebell.Agrandpianoandpercussion instrumentshaveprobablythemostcomplexradiationproperties.

Tosumup,thedirectivityofeachinstrumentisdefinedbytheradiatingmechanismofeachinstrumentandtheradiationataparticularfrequencyisalwaysthesame.Naturally,theradiationpatternvariesbetweenplayednotes,as different frequencies are excited, but if all frequencieswouldbe excited simultaneously the radiationpatternwouldalwaysbethesame.Moreover,theradiationpatternsdonotdependonplayingintensity(Pätynen&Lokki2010).Thisconceptisimportanttounderstandandtoseparatefromtheleveldependentcharacteristicsofmusicalinstruments.

3.1.2 Level depend features of music instruments Inmusicalacousticsresearch,theleveldependencyofinstrumentspectraiswellunderstood(Luce1975,Meyer

2009).Forexample,abrassinstrumentexcitesmanymoreharmonicswhenplayedinfortissimothaninpianissimo,resulting in a different timbre, and a trumpet soundsmuchmore bright when played loudly than when playedsoftly. In room acoustics research, such spectral changes are traditionally ignored and only recently has it beenfound to be important (Pätynen et al. 2014, Lokki&Pätynen2015, Lokki 2016). As roomacoustics researchhaslargelyfocusedonmeasurementandmodelingofimpulseresponses,itisevidentthatsuchleveldependentissueshavenotbeengivenattention.



Figure3illustratesthespectraldifferencesofpianoandfortissimoofafullorchestra,analyzedfromcommercialrecordings.Thedataaretakenfrom29recordings1ofBruckner’s4thSymphony(bars19-26,IIImovement)andtheplot shows the spectraof anorchestraplayinga long crescendowithout significant change in thenotes thatareplayed.Theplotintherightshowsthedynamicrangeandhighlightsthelargedifferencesatdifferentfrequencies.At low frequencies the larger dynamic range is due to timpani playing really loudly, and at high frequencies thedifferenceisrelatedtotheincreaseinhigherharmonics.Inmanycasesthespectraldifferencescanbeevenlargerasloudinstruments(grancassa,timpani,tuba,trombone,trumpet,piatti)aresilentinpianopassages,butjointhefortissimos,whichmakesthedifferencereallypronounced.

Figure3:Left:Thespectraofafullorchestrasoundindifferentdynamics.Right:thespectralchangebetween

fortissimoandpiano.

Figure4:Left:Binaurallevelforreflectionsfromlateralandfrontaldirections.Theresponsesaremeansofallcoloreddirectionsillustratedontheright.

3.2 Human listeners as receivers Similarlytothesoundsources,humanlistenershavetwoimportantfeatures;frequencydependentdirectivity

(Møller1992,Mølleretal.1995),andleveldependentsensitivity,thelatterofwhichisgenerallyknownintheformoftheequalloudnesscontours(ISO2262003).Theequalloudnesscontoursalsochangetheirshapesaccordingtothelevel,asdiscussedhere.

1Commercialrecordingsareusuallycompressedandthereforetheoverallleveldifferenceisquitemodest.In-

situinaconcerthallthedynamicrangecouldbeeven60dB(Meyer2009).

Frequency [Hz]102 103 104

Re

lativ

e m

ag

nitu

de

[d

B]

0

10

20

30

40

50

60

70

80Spectra of full orchestra sound

FortissimoPiano


Rela

tive m

agnitu

de [dB

]

0

2

4

6

8

10

12

14

16

18

20

22Fortissimo - Piano


Bin

au

ral l

eve

l [d

B]

-8

-6

-4

-2

0

2

4

6

8

Comparison of lateral and frontal reflection

Mean of frontal reflection (azi = -20...20, ele = -10...80)Mean of lateral reflection (azi = 30...80, ele = -10...45)Difference



3.2.1 Directivity of the human head The frequencydependentdirectivityof thehumanheadhasbeenunderactive research for severaldecades.

Head-related transfer functions (HRTF) (Møller 1992,Møller et al. 1995)describe theeffectof thebodyand theouter ear to the frequency response of sound and these direction-dependent functions have been measured,modeled,andapplied instudiesrelatedtobinaural technology(Blauert1997).FromtheroomacousticalpointofviewtheHRTFsshowhowthehumanheadmodifiesthefrequencyresponsesofreflectionsarrivingfromdifferentdirections.Figure4illustratesthemeandifferencebetweenareflectionfromthefrontaldirection(e.g.,fromceilingorfromcloudsabovetheorchestra)andthelateraldirection(e.g.,areflectionfromasidewallorfromabalconyoverhang).Notethattheplotshowsbinaurallevel,whichisapowersumofbothears(Sivonen&Ellermeier2006).The spectra illustrate how a human head emphasizes the level of lateral reflections relative to median planereflections (Lokki&Pätynen2011).Analogously to thedirectivityofmusical instruments, theHRTF functionsareindependentofthelevelofexcitationsignal.Inaddition,HRTFsarenotdependentonthedistanceifthedistancefromtheearismorethanonemeter(Brungart&Rabinowitz1999),i.e.,notdistancedependentinthecontextofaconcerthall.

3.2.2 Level dependent sensitivity of human hearing Correspondinglytotheleveldependentspectraofmusicalinstruments,thesensitivityofhumanhearingisalso

level dependent. Equal loudness contours ISO226 (2003), plotted in Figure 5, describe the level of differentfrequencies producing loudness perceived as equal to that of a 1 kHz tone at a given level. These contours aredefinedonlyforsingletonesanditisnotknownexactlytowhatextenttheyapplyforcomplexwidebandsounds.Itisevidentthisleveldependentsensitivityparticularlyaffectsourperceptionoflowfrequencies.

Figure5:Left:EqualloudnesscontoursaccordingtotheISO226standard.Right:Thesamecontoursshiftedsothat

thelevelsarealignedat1kHz.

3.3 Combination of level dependent features Figure6combines the two leveldependenteffects together for the indicativemusicalpassageassessed.The

equalloudnesslevelfor40phonweightedthespectrumofanorchestralpianoandtheequalloudnesslevelsfor60to100phonweightedthespectrumofanorchestral fortissimo.Thefigure illustratesthat inadditiontothe levelincrease the overall frequency response flattens markedly. The relative change in loudness-weighted spectrabetweenthepianoandfortissimopassagesishighlightedintherightplot,whichshowsallresponseslevelalignedat1 kHz. The largest change is at low frequenciesbelow200Hz, but also the change for frequenciesover3 kHz ispronounced.Thesefrequencyregionsarethusimportantandtheyshouldbeconsideredmorecarefullyinconcerthallacousticsresearchandinparticularwhendesigningnewhalls.

Figure6 (left)also reveals that, fromtheengineeringpointofview, thedefinedobjectiveparametersatmidfrequencies are good, as the level dependency (see Figure2) isminimal.However, from theperceptionpoint ofview, current mid frequency parameters do not describe how a hall renders broadband music. Mid frequencyresultstotallyignoretheleveldependentcharacteristicsofthecombinationofthemusicandtheroomacoustics.


So

un

d P

ress

ure

Le

vel [

dB

]

30

40

50

60

70

80

90

100

110

120

130

[phons]

Equal loudness contours for different loudness levels

40 50 60 70 80 90100

Frequency [Hz]

102

103

104

"Rela

tive S

PL"

[dB

]

-10

0

10

20

30

40

50Equal Loudness Contours shifted to match at 1 kHz

40 dB

50 dB

60 dB

70 dB

80 dB

90 dB

100 dB



Figure6:Left:Loudnessweightedorchestraspectraatvariousplayingintensities.Right:Relativechangesin

loudnessweightedorchestralspectra,betweenpianoandfortissimo,contourslevelalignedat1kHz.

4. MEASUREMENTS AND AURALIZATION OF CONCERT HALLS A concert hall acts as a frequency dependent transmission channel in the source-medium-receiver

communicationprocess(Figure2).Ahallconveysthesoundfromthestagetothelistenersdirectlyandthroughthereflections that form the reverberation. The shape of the hall defines the reflection patterns and the frequencycontents of reflections are further modified by the absorption properties of the surface materials. In addition,propagationthroughairattenuatesthehighfrequencies.Assaidintheintroduction,theISO3382-1:2009standarddefines method to measure this transmission channel. However, the standard ignores the frequency and leveldependentfeaturesofsourcesandreceivers.Therefore,itisnotasurprisethatifthedesignofanewconcerthallisbased on these standard parameters (often atmid frequencies), the resulting sound of an orchestra cannot bepredictedverywell.Itisalsoobviousthatauralizationofamonoresponsefromonesingleomnidirectionalsourceisinadequateforperceptualstudiestounderstandhowtheacousticsofaconcerthallaffectsmusic.

4.1 Concert hall acoustics research at Aalto University Our research team at Aalto University startedworking on concert halls in 2008.Wewanted to havemore

detailedinformationfromthemeasurementsbyapplyingmorearealisticsoundsourceandamicrophonearraytocapturespatialimpulseresponses.Inaddition,wewantedtodosubjectivestudieswithrealsignals(i.e.music,notnoiseorsyntheticsignals)andtheacousticsofrealhalls.Veryoftenroomacousticsresearchisbasedonmodeledand simplified impulse responses, whichmight give biased results although they can be verywell controlled. Inshort,ourgoalwastohaveinourlisteningroomasanauthenticauralizationofexistingconcerthallsaspossible,toenable real-time comparison of different existing acoustics. The research led to an invention that we call a“loudspeakerorchestra”,whichisasymphonyorchestrasimulatormadeoutof34loudspeakersonstage(Pätynen2011).Thefundamentalideabehindthisorchestra,showninFigure7,isthatithasdozensofsourcesasforarealorchestra, it can be calibrated, and it is accurately repeatable in every hall. Naturally, the music played byloudspeakershastoberecordedinananechoicroomandsuchrecordingswereanotherlargeproject(Pätynenetal.2008,2011).TheloudspeakerorchestradoesnotfulfilltherequirementsoftheISO3382-1:2009standard,butitimplementsboth frequencydependentdirectivitiesof sourcesand leveldependent spectra (seeFigure2).Whenthemeasurementsandrecordingsaremadeatequaldistanceseatsineachhall(Tervo,Pätynen&Lokki2013),thecomparisonofhallsrevealsverydetailedinformationonthedifferencesinacousticsofstudiedhalls.

First perceptual study (Lokki, Pätynen, Kuusinen, Vertanen & Tervo 2011) was made using a spatial soundrecordingandreproductiontechnique(DirectionalAudioCoding;Pulkki2007)totransfersoundsinaconcerthalltoourmultichannellisteningroom.Although,thespatialsoundqualitywiththatmethodwasalreadyhigh,itwasnotauthentic.Therefore,thesecondstudywasbasedonthesamerecordingtechnique,butimplementedwithimpulseresponses(SpatialImpulseResponseRendering;Merimaa&Pulkki2005,Pulkki&Merimaa2006).InsuchamethodtheimpulseresponsesfromeachloudspeakerchannelonstagearemeasuredwithaB-formatmicrophoneandthe


Magnitu

de [dB

]

-40

-20

0

20

40

60

80

100

120"Auditory" frequency response

fortissimo [60 dB]fortissimo [70 dB]fortissimo [80 dB]fortissimo [90 dB]fortissimo [100 dB]piano [40 dB]


Ma

gn

itud

e [

dB

]

-5

0

5

10

15

20

25

30Relative change in spectra

fortissimo [60 dB] - piano [40 dB]fortissimo [70 dB] - piano [40 dB]fortissimo [80 dB] - piano [40 dB]fortissimo [90 dB] - piano [40 dB]fortissimo [100 dB] - piano [40 dB]



capturedspatialimpulseresponsesareprocessedforamultichannellisteningsetupbeforethemusicisconvolvedwith reproduction loudspeaker responses. The appliedmethod, basedonB-format impulse responses, increasedthesoundquality,butinsomecasesthehallsstilldidnotsoundsufficientlyauthentic.However,thesecondstudywassuccessful(Lokkietal.2012,Kuusinenetal.2014)confirmingtheperceptualaspectsofacousticsfoundearlier,and highlighted the importance of proximate and intimate sound for preference. Finally, the research towardsauthentic auralization resulted in our novel method, which is called the Spatial Decomposition Method (SDM)(Tervo, Pätynen, Kuusinen & Lokki 2013). The SDM is based on the simultaneous measurement of impulseresponses using several closely spaced microphones and analyzing time-difference-of-arrivals of sound energybetweenmicrophonepairswithveryshort timewindows.Asa result, theSDMaugmentsonemeasured impulseresponsewithmetadatathatcontainsazimuthandelevationanglesforeachsample.Therefore,intheauralizationprocess, an impulse response can be decomposed into several reproduction loudspeakers according to themetadata.Naturally,therecanbealargenumberofsoundsourcesandtheauralizationprocessisundertakenforeachofthemindividually.TheSDMhassofarbeenusedinstudiesonconcerthalls(Kuusinen&Lokki2015,Lokkietal.2016),studiocontrolrooms(Tervoetal.2014),andcarcabins(Tervoetal.2015).TheSDMiscurrentlythestate-of-the-art auralization method for room acoustics studies, but it is not a direct recording technique for spatialsound.

Figure7:TheloudspeakerorchestraonthestageoftheConcertgebouw,Amsterdam,theNetherlands.Toprightimageshowsthespatialsoundmicrophoneconsistingofsixomnidirectionalmeasurementmicrophones.Figure

adaptedfromKuusinen(2016).

Beforediscussingindetailtherecentresearchtopicsinconcerthallacoustics,itshouldbeemphasizedthatthespatial impulse responses, analyzed with SDM, enable intuitive visualization of sound energy distribution in thetime-frequencyandthespatio-temporaldomains(Pätynenetal.2013).Figure8illustratesanexampleanalysis.Inthisvisualizationtheideaistoanalyzeimpulseresponseswithalengtheningtimewindowandthentoillustratethefrequencyresponsesforspatialdistributionofsoundenergyinspaceincumulativetimewindows.CurvesinFigure8showtheresponsesat0-30ms(blackcurves),thencumulativelylongertimeframesat10msintervalsupto200ms(greycurves)andfinallytheresponseswhenthewholeimpulseresponseisappliedinanalysis(redcurve).



Figure8:Anexamplevisualizationofsoundenergydistributionintime-frequencyandspatio-temporaldomains.

5. WHAT MAKES A PREFERRED CONCERT HALL? Architectsandacousticianswanttomakethebestpossibleacousticsforeverynewhall.Asadesignproblem,

this is almost impossible, as the “best possible acoustics” is a matter of taste (Lokki 2014) and different musicrequires different acoustics.Many research teams have tried to find the optimal acoustics. Our research team’slatestcontributionwasalargelisteningtestinwhich28assessorsevaluatedsixconcerthallsatthreedifferentseatswithtwoshortmusicsignals,BrucknerandBeethoven(Lokkietal.2016).Threeofthehallswereclassicalshoeboxshapedhallsandthreeothersofmodernvineyardorarenadesigns.TheresultsofthepreferencetestsaredepictedinFigure9,whichclearlyshowsthatassessorscanbedividedintotwopreferencegroups(i.e.classes).Moreover,music signal affected the results in each group and changed the preference order of halls, although the overallpreferencedidnotchangeremarkably.

Ourlisteningtestalsoincludedanindividualvocabularyprofiling,inwhicheachassessorverbalizeshis/herownperceptions of the differences he/she hears between halls. The process resulted around 100 attributes withdefinitionsforbothmusicexcerpts.Inaddition,theassessorscomparedthehallsinpairsusingtheirownattributes.The results reveal the perceptual differences between halls and results can be associatedwith the preferences.Statisticalanalysiscategorizedtheattributesintothreedifferentclasses.Thelargestclassofattributesconsistedofindividualtermsrelatedtoreverberance,loudness,andwidth.BasswasalsointhisfirstclasswithBruckner,butnotwithBeethoven.Differencesintheseperceptualaspectsarequiteuniversalandeverylistenercanheardifferencesinreverberanceand loudness, forexample.When lookingat theorderofhallswiththeseattributes theshoeboxhalls (VM,AC,andBK)havemoreof theseaspectsand thepreferenceclass1 (seeFigure9) clearlyvalueswide,reverberantandstrongsound.Here,itshouldbenotedthateventhoughtheseperceptualaspectswereclusteredinthesameclasswithinthestudiedhalls,theycouldbeseparatedperceptuallywhena listenerpaysattentiontoonlyoneof themata time. Inourstudythehalls thatweremorereverberantwerealsoperceivedaswider,butthere could, e.g., beahall that soundednarrowand still very reverberant. The smallestof theattribute clusterswerenamedasclaritywithBruckneranddefinitionwithBeethoven.Itcouldbeconcludedthatassessorsbelongingto class 2 value clarity and definition over reverberance and loudness. Finally, the last group of attributeswererelatedtotimbreandconsistedofattributessuchasbass,brightness,brilliance,andproximity.



Figure9:Preferencesofconcerthalls.VM,AC,andBKareshoeboxhallsandBP,CP,andHMarevineyardorarena-

typeconcerthalls.Seemoredetails(Lokkietal.2016).

Theaboveresultswerenotsurprisingasseveralearlierstudieshavefoundoutthatsubjectscanbedividedintoatleasttwopreferencegroups,seee.g.(Hawkes&Douglas1971,Schroederetal.1974,Barron1988,Soulodre&Bradley 1995, Sotiropoulou & Fleming 1995). In addition, the attributes that described perceptual differencesbetweenhallshavealsobeenlistedinseveralearlierstudies.IfwereflectonresultsinFigure1,wecanseethatthefirstprincipaldimensionisthe“preferenceaxis”,asindividualpreferencesoflistenersaredistributedalongthisaxisfromnegativeend to thepositiveend.Thus, if theobjectiveparametersof the ISO3382-1:2009standardexplainthisaxis,shouldwejustadmitthatpreferencesofpeoplevaryalotandindividualsweighttheseaspectsdifferently?

It is clear that the total picture ismuchmore complex and there are certainly aspects in acoustics that theISO3382-1:2009standardismissing.Onemajorshortcomingisthatthestandardignoresalltimbrerelatedaspects,suchasthestrengthofbass,brillianceandbrightness.Interestingly,Barron(2005,Table1)alsolistedbrillianceandwarmth as subjective qualities of concert halls, but he did not find any objective measure for them. Anotherimportantissueisproximityorintimacy,whichhasbeendiscussedinmanypapersandwhichhasbeenfoundtobeadriverforoverallpreference(Kuusinenetal.2014).Outofthese,thetimbrerelatedfeaturescouldbeexplainedobjectivelywith strengthG at certain octave bands, but no consensus has been found. Proximity ismuchmorecomplexanditisnotevidentwhatfeaturesinacousticsmakeaproximatesound,althoughtheloudnessiscertainlyoneofthekeyfactors(Kuusinen&Lokki2015).

STALLS FRONT PARTERRE BALCONY

0.0

0.2

0.4

0.6

0.0

0.2

0.4

0.6

Class 1: 43 %Class 2: 57 %

VM AC BK BP CP HM VM AC BK BP CP HM VM AC BK BP CP HMHall

Pi

MusicBrucknerBeethoven



6. FROM PERCEPTION POINT OF VIEW ACOUSTICS OF CONCERT HALLS IS LEVEL DEPENDENT Section3andFigure2explainthatallperceptualaspectscannotbeexplainedbystudyingmeasured impulse

responses. The level dependent features of sources andof humanhearing suggest that level dependent aspectsmayplayamajorroleforpreference.Fromthecomparativelisteningtestsofdifferentconcerthalls,excitedwiththesameanechoicorchestrarecordings, ithasbecomeobviousthatdifferenthallsrendercrescendosandsuddendynamicchangesdifferently.Thisobservationledtothestudyofdynamicandleveldependentaspectsofconcerthallacousticsandourfirstpublicationsconcentratedoninvestigatingobjectivelymeasurablereasonsforperceiveddynamic range (Pätynen et al. 2014, Lokki & Pätynen 2015). Concurrently, when we did the above-mentionedpreference tests and individual vocabulary profiling,we studied the perception ofmusic dynamicswith listeningtests (Pätynen & Lokki 2016b) and with psychophysiological measurements (Pätynen & Lokki 2016a). The latterstudy revealed interesting results as it showed that the most renowned concert hall, Musikverein in Vienna,increasedskinconductivitythemostduringthepassivelisteningofacrescendo(Pätynen&Lokki2016a).Thisalsohappenedwiththeassessorswhopreferredlessreverberanthallswithgreatclarity.OurinterpretationofthisresultwasthattherenownedMusikverein,andsomesimilarshoeboxhalls,rendermusicmoreexpressivethanothertypeof halls (Lokki et al. 2015). In addition, the vineyard hallsmight reduce themusical dynamics (Pätynen & Lokki2015),andthereforesomepeopledonotlikesuchhallsatall,whileotherspraisethesehallsduetoextremeclarityanddefinition.

Our recent studies suggest that acoustics of concert halls is “perceptually level dependent”, as outlined inSection3.Intheliterature,Beranek(1996)didnottotallyignoreleveldependentaspectsandhewroteinhisbook:

Thethrillofhearing[orchestralmusic...]isenhancedimmeasurablybythedynamicresponseoftheconcerthall.Dynamicresponsemeansbothquietsupportforthepianissimopartsandmajesticlevelsatthefortissimos.[...]Unfortunately, Beranek did not elaborate further on dynamic responsiveness and he associated such

phenomenontolinearlybehavingobjectivestrength(G)parameter.Anotherexample,evenearlierthanBeranek,isfromMarshall(1967):

Narrowhallswithhighceilinghavespatialresponsiveness,whereasthemoremodernbroad,withlowceilinglacksspatialresponsiveness.Both of these quotes are clear evidence that such dynamic aspects are apparent in live concerts, but past

researchhasnotbeenable toexplain thereasonsbehindthem.Themajorityof researchhasbeenconcentratedaroundimpulseresponsesandthewholecommunicationprocess(Figure2)hasnotbeenunderstoodindetail. Inthe following, we discuss some of the features of concert hall acoustics that lead to situations in which theperception of the room acoustics could be level and signal dependent. Figure 6 indicates that level dependentfeaturesplaythe largestroleat frequenciesbelow200Hzandabove3kHz, frequencyregionsthatdeviatemorethan5dBfromzeroinFigure6.

6.1 Low frequencies below 200 Hz and the seat dip effect Atlowfrequenciesbothsourcesandreceiversaremoreorlessomnidirectional,thusthebassresponseisnot

affectedatallbythedirectivitiesofinstrumentsandlisteners.Nevertheless,thereareremarkabledifferencesinthelowfrequencyresponsesbetweendifferentconcerthalls,asseeninFigure10.Themainreasonsforlowfrequencyattenuationarewallstructures,materialabsorption,stageconstruction,andtheseatdipeffect(Schultz&Watters1964,Sessler&West1964).Aportionofthesoundpropagatingatagrazingangleovertheseatingareadiffractsdownbetweenthechairsandreflectsoffthefloor,usuallyformingdestructiveinterferencewiththeundiffractedsound at low frequencies. Themain dip in the frequency response, due to this destructive interference, usuallyoccursbetween100and200Hz,dependingontheseattypeandinclinationofthefloor(Tahvanainenetal.2015).When the seat dip frequency is high, close to 200 Hz, the interference seems to be constructive below 100 Hzresultinginstrongboostatreallylowfrequencies,asseeninFigure10.Inouropinion,thefrequenciesbelow100Hz,allthewaydownto20Hz,arereallyimportantforstrongbassandwarmandintimatesound.Whentheseatdipfrequency iscloser to100Hz (typical for rakedfloorandseatswithoutunderpass, (Tahvanainenetal.2015)) thefrequenciesbelow100Hzarealsoattenuatedandthelowestoctavesareweak,eveninaudible.Tovalidatealltheseassumptions,more research isneeded tounderstand the roleof frequenciesbelow100Hzandwhat featuresofconcert halls attenuate or strengthen these frequencies. Unfortunately, there are few measurements at 63 Hzoctaveband,asthenormalpracticehasbeentomeasurehallsstartingfrom125Hzoctaveband.



Futureresearchshouldalso lookmoredeeplytothecombinationofmusicandacoustics. Inorchestralmusicwhen an orchestra is playing in piano only a few instrument sections are in voice, but during crescendos theremaining instrumentsections (e.g., trombones, tuba,grancassaandtimpani) join intothepassage.At thesametime,usually,thepitchoftheleadingvoiceincreases.Furthermore,thecounterpointinclassicalmusicdictatesthatthebasslinemovesinoppositiontoarisingmelody--towardlowerfrequencies.Insuchcases,notonlythenominallevelchanges,butalsotheexcitedfrequenciesspanamuchwiderbandandtheimportanceof lowfrequenciesispronounced.

Figure10:Spatiotemporalvisualizationsofcumulativesoundenergy11meterstothestage.Thecolorsindicatethelevelofcumulativeenergyat5,30,120,3000msafterthedirectsoundandthedataistheaverageof24sourcechannelsonstage.Fromtop:frequencyresponses,soundenergyinplane(2ndrow),insection(3rdrow),in

transverseplane(4throw).Inspatiotemporalplots-6,0,and6dBcurvesindicatethelevelre.levelinfreefieldat10m(i.e.thesameleveladjustmentthanwithstrengthG).(Lokkietal.2015).

6.2 High frequencies above 3 kHz and early reflections High frequencies attenuate quickly in a concert hall due tomaterial and air absorption. Therefore, it is very

importanttopayattentiontothesurfacesthatreflectearlysound.Inmanymodernvineyardhallshardlyanyearlyreflectionsarepresent(Pätynen&Lokki2015)orthemajorityofearlyenergyforlistenersreflectfromtheceiling.Ifwe look at Figure 4, we can see that ceiling reflections do not emphasize high frequencies as much as lateral



reflections due to the directivity of the humanhead. Therefore, lateral reflections are "more efficient" and theyconveybettertheharmonicsoftheinstruments(Lokki&Pätynen2011)resultinginlargerdynamicrangeandbetterspatialresponsiveness(Meyer2009,Pätynenetal.2014).Inaddition,thereflectorsonthesidearemoreefficientthanabovetheaudienceareabecausetheyspreadthesoundenergyoveragreaterlisteningareaasexplainedbyJurkiewicz, Wulfrank, and Kahle (2012). For these reasons, early lateral reflections from sidewalls, and fromunderneathofsidebalconies,aremandatoryinaconcerthall.

Manyacousticdesignerstendto likediffusivestructures, includingonsurfacesthatproduceearlyreflections.Obviously,diffusorsonthewallsareusedtodistributethesoundtoawiderarea.However,theperceptualaspectofearlydiffuse reflectionsshouldbestudiedmorecarefully,asdiffusorsmightalsocolor thehigh frequencies inreflections(Lokki,Pätynen,Tervo,Siltanen&Savioja2011).Recentresearchonstudiocontrolroomsalsosuggeststhat reflections fromhard flat surfaces indeedarebeneficial, notdetrimental (Kingetal. 2012).Moreover, ifwelook at the photographs of the renowned concert halls (e.g., Boston SymphonyHall, AmsterdamConcertgebow,ViennaMusikverein), sidewalls are not covered with diffusors; in contrast the walls are flat, covered with hardplaster.

6.3 Temporal aspects in acoustics -- room and source presence Afewpeople,e.g.,Kahle(2013)hassuggestedthattheauditoryperceptionofasymphonyorchestraplayingin

aconcerthallcanbeunderstoodwithrespecttotwomainpercepts:thesourcepresenceandtheroompresence.Thesourcepresenceisthecontinuousperceptionofthesoundsourcesinthehallwhiletheroompresenceistheperceptionofthespacethemusicislistenedto.Thesetwoareseparateentitiesintheperceptualdomain.Ifahallcancreate these two"auditorystreams", i.e., theyaredistinctandseparate, then it isproposed thismaypermitbothgoodclarityandplentiful,envelopingreverberationatthesametime.Theformationoftheauditorystreamsispossible through stream segregation (Griesinger, 1997) and is subject to the perceptual grouping laws therein(Moore, 2012). The early reflections are perceptually groupedwith the source streams through the precedenceeffect(Litovsky,Colburn,Yost,&Guzman,1999),andaffectthewidth,loudness,andtimbreoftheauditoryevents(Blauert,1997).Inthisway,thedirectsoundoftheorchestraandtheearlyreflectionsofthehallcombinetomakeupthesourcepresence.Thelatereflections,i.e.reverberation,formthecontextandspaceforthemusic,andlendthemusicsupport,embellishment,andasenseofdepth,providingthelistenerwithasenseofenvelopment;thatis,roompresence. At themoment, there isnoclearconsensushowthese twostreamsare formed,ordoweevenneedthem.Naturally,moreresearchisneeded,includingthespatialaspectsofearlyandlatereflections(Lokkietal.2015).

7. CONCLUSIONS Thispapersummarizesissuesthatareoftennotconsideredinconcerthallacousticsresearch.Inparticular,

the paper highlights the perceptual consequences of the level and frequency dependent phenomena ofmusicalinstruments and human spatial hearing. In addition, recent research results on the preferred concert halls arediscussed.

Infutureresearchweneedtounderstandmuchbetterhowthecombinationoforchestralmusic,concerthall,andhumanperceptionworkstogether.Thetraditionalpathofexplainingdifferencesbetweenhallsexclusivelywithmeasured and modeled monaural impulse responses is not sufficient. It is hoped that this paper providesmotivationforfurtherresearchintotheperceptionofmusicinperformancespaces.

ACKNOWLEDGEMENTS IthanktheVirtualAcousticsresearchteamatAaltoUniversityfortheirhardworkandgreatworkingspirit.In

addition,JamesHeddleisthankedforgreatcommentsontheearlierversionofthispaper.



REFERENCES

Barron,M.(1988),‘SubjectivestudyofBritishsymphonyconcerthalls’,Acustica66,1–14.

Barron, M. (2005), ‘Using the standard on objective measures for concert auditoria, iso 3382, to give reliableresults’,AcousticalScienceandTechnology26(2),162–169.

Beranek,L.(1996),ConcertandOperaHalls—HowTheySound,AcousticalSocietyofAmerica,NewYork,NY,USA.

Blauert,J.(1997),SpatialHearing.Thepsychophysicsofhumansoundlocalization,2ndedn,MITPress,Cambridge,MA.

Bradley,J.S.(1990),Contemporaryapproachestoevaluatingauditoriumacoustics,in‘Proc.AES8thInter-nationalConference’,Washington,D.C,USA,pp.59–69.

Brungart, D. & Rabinowitz,W. (1999), ‘Auditory localization of nearby sources. head-related transfer functions’,JournaloftheAcousticalSocietyofAmerica106(3,Pt.1),1465–1479.

Griesinger, D. (1997), ‘The psychoacoustics of apparent source width, spaciousness and envelopment inperformancespaces’,ActaAcusticaUnitedwithAcustica,83,721–731.

Hawkes,R.&Douglas,H. (1971), ‘Subjectiveacousticsexperience inconcertauditoria’,Acustica24,235–250. ISO3382-1(2009),‘Acoustics–measurementofroomacousticparameters–part1:Performancespaces’,

InternationalStandardsOrganization.ISO226(2003),‘Normalequal-loudness-levelcontours’.

Jurkiewicz, Y.,Wulfrank, T.&Kahle, E. (2012), ‘Architectural shapeandearly acoustic efficiency in concerthalls’,JournaloftheAcousticalSocietyofAmerica132(3),1253–1256.

Kahle, E. (2013). Room acoustical quality of concert halls: Perceptual factors and acoustic criteria – return fromexperience.In‘InternationalSymposiumonRoomAcoustics(ISRA2013)’,Toronto,Canada.

King,R.,Leonard,B.&Sikora,G. (2012), ‘Thepracticaleffectsof lateralenergy incritical listeningenvironments’,JournaloftheAudioEngineeringSociety60(12),997–1003.

Kuusinen,A.(2016),Multidimensionalperceptionofconcerthallacoustics-Studieswiththeloudspeakerorchestra,PhDthesis,AaltoUniversitySchoolofScience.Onlineathttp://urn.fi/URN:ISBN:978-952-60-6904-3.

Kuusinen,A.&Lokki,T. (2015), ‘Investigationofauditorydistanceperceptionandpreferences inconcerthallsbyusingvirtualacoustics’,JournaloftheAcousticalSocietyofAmerica138(5),3148–3159.

Kuusinen,A.,Pätynen,J.,Tervo,S.&Lokki,T. (2014), ‘Relationshipsbetweenpreferenceratings,sensoryprofiles,andacousticalmeasurementsinconcerthalls’,JournaloftheAcousticalSocietyofAmerica135(1),239–250.

Litovsky,R.Y.,Colburn,H.S.,Yost,W.A.,&Guzman,S.J.(1999),‘Theprecedenceeffect’,JournaloftheAcousticalSocietyofAmerica,106,1633–1654.

Lokki,T.(2014),‘Tastingmusiclikewine:Sensoryevaluationofconcerthalls’,PhysicsToday67(1),27–32.

Lokki,T. (2016),Concerthalls–conveyorsofmusicalexpressions, in ‘Proceedingsofthe ICA2016’,BuenosAires,Argentina.

Lokki,T.&Pätynen,J.(2011),‘Lateralreflectionsarefavorableinconcerthallsduetobinauralloudness’,JournaloftheAcousticalSocietyofAmerica130(5),EL345–EL351.

Lokki,T.&Pätynen,J.(2015),‘Theacousticsofaconcerthallasalinearproblem’,EurophysicsNews46(1),13–17.

Lokki, T., Pätynen, J., Kuusinen, A.& Tervo, S. (2012), ‘Disentangling preference ratings of concert hall acousticsusingsubjectivesensoryprofiles’,JournaloftheAcousticalSocietyofAmerica132(5).3148-3161.



Lokki, T., Pätynen, J., Kuusinen, A. & Tervo, S. (2016), ‘Concert hall acoustics: Repertoire, listening position andindividual tasteof the listeners influence thequalitativeattributesandpreferences’, Journalof theAcousticalSocietyofAmerica140(1),551–562.

Lokki, T., Pätynen, J., Kuusinen, A., Vertanen, H. & Tervo, S. (2011), ‘Concert hall acoustics assessment withindividuallyelicitedattributes’,JournaloftheAcousticalSocietyofAmerica130(2),835–849.

Lokki, T., Pätynen, J., Tervo, S., Kuusinen, A., Tahvanainen, H. & Haapaniemi, A. (2015), The secret of themusikverein and other shoebox concert halls, in ‘The 9th International Conference on AuditoriumAcoustics’,Paris,France.

Lokki, T., Pätynen, J., Tervo, S., Siltanen, S. & Savioja, L. (2011), ‘Engaging concert hall acoustics is made up oftemporalenvelopepreservingreflections’,JournaloftheAcousticalSocietyofAmerica129(6),EL223–EL228.

Luce,D.A.(1975),‘Dynamicspectrumchangesoforchestralinstruments’,JournaloftheAudioEngineeringSociety23(7),565–568.

Marshall, A. H. (1967), ‘A note on the importance of room cross-section in concert halls’, Journal of Sound andVibration5(1),100–112.

Merimaa,J.&Pulkki,V.(2005),‘SpatialimpulseresponserenderingI:Analysisandsynthesis’,JournaloftheAudioEngineeringSociety53(12),1115–1127.

Meyer,J.(2009),AcousticsandthePerformanceofMusic,Springer(NewYork).

Moore,B.C.J.(2012),Anintroductiontothepsychologyofhearing(6thed.),Bingley,UK:Emerald,283–312.

Møller,H.(1992),‘Fundamentalsofbinauraltechnology’,36(3-4),171–218.

Møller,H.,Hammershøi,M.S.D.& Jensen,C. (1995), ‘Head-relatedtransfer functionsofhumansubjects’,43(5),300–321.

Pätynen,J.(2011),AVirtualSymphonyOrchestraforStudiesonConcertHallAcoustics,PhDthesis,AaltoUniversitySchoolofScience.Onlineathttp://urn.fi/URN:ISBN:978-952-60-4291-6

Pätynen,J.&Lokki,T.(2010),‘Directivitiesofsymphonyorchestrainstruments’,ActaAcusticaunitedwithAcustica96(1),138–167.

Pätynen,J.&Lokki,T.(2015),‘Theacousticsofvineyardhalls,isitsogreatafterall?’,AcousticsAustralia43(1),33–39.

Pätynen, J. & Lokki, T. (2016a), ‘Concert halls with strong and lateral sound increase the emotional impact oforchestramusic’,JournaloftheAcousticalSocietyofAmerica139(3),1214–1224.

Pätynen,J.&Lokki,T.(2016b),‘Perceptionofmusicdynamicsinconcerthalls’,JournaloftheAcousticalSocietyofAmerica.Conditionallyaccepted.

Pätynen,J.,Pulkki,V.&Lokki,T.(2008),‘Anechoicrecordingsystemforsymphonyorchestra’,ActaAcusticaunitedwithAcustica94(6),856–865.

Pätynen, J.,Tervo,S.&Lokki,T. (2011),Simulationof theviolinsectionsoundbasedontheanalysisoforchestraperformance, in ‘IEEEWorkshoponApplicationsof SignalProcessing toAudioandAcoustics (WASPAA)’,NewPaltz,NewYork,USA,pp.173–176.

Pätynen,J.,Tervo,S.&Lokki,T.(2013), ‘Analysisofconcerthallacousticsviavisualizationsoftime-frequencyandspatiotemporalresponses’,JournaloftheAcousticalSocietyofAmerica133(2),842–857.

Pätynen, J., Tervo, S., Robinson, P.W. & Lokki, T. (2014), ‘Concert halls with strong lateral reflections enhancemusical dynamics’, Proceedings of theNational Academy of Sciences of theUnited States of America (PNAS)111(12),4409–4414.



Pulkki, V. (2007), ‘Spatial sound reproduction with directional audio coding’, Journal of the Audio EngineeringSociety55(6),503–516.

Pulkki,V.&Merimaa,J.(2006),‘SpatialimpulseresponserenderingII:Reproductionofdiffusesoundandlisteningtests’,JournaloftheAudioEngineeringSociety54(1),3–20.

Schroeder,M., Gottlob, G. & Siebrasse, K. (1974), ‘Comparative study of European concert halls: Correlation ofsubjective preferencewith geometric and acoustics parameters’, Journal of theAcoustical Society ofAmerica56(4),1195–1201.

Schultz,T.&Watters,B.(1964),‘Propagationofsoundacrossaudienceseating’,JournaloftheAcousticalSocietyofAmerica36(5),885–896.

Sessler,G.&West,J.(1964),‘Soundtransmissionovertheatreseats’,JournaloftheAcousticalSocietyofAmerica36(9),1725–1732.

Sivonen, V. P. & Ellermeier, W. (2006), ‘Directional loudness in an anechoic sound field, head-related transferfunctions,andbinauralsummation’,JournaloftheAcousticalSocietyofAmerica119(5),2965–2980.

Sotiropoulou,A.&Fleming,D.(1995),‘Concerthallacousticevaluationsbyordinaryconcert-goers:II,Physicalroomacousticcriteriasubjectivelysignificant’,Acustica81(1),10–19.

Soulodre,G.&Bradley,J.(1995),‘Subjectiveevaluationofnewroomacousticmeasures’,JournaloftheAcousticalSocietyofAmerica98(1),294–301.

Tahvanainen,H.,Pätynen, J.&Lokki, T. (2015), ‘Analysisof the seat-dipeffect in twelveEuropeanconcerthalls’,ActaAcusticaunitedwithAcustica101(4),731–742.

Tervo,S.,Laukkanen,P.,Pätynen, J.&Lokki,T. (2014), ‘Preferenceofcritical listeningenvironmentamongsoundengineers’,JournaloftheAudioEngineeringSociety62(5),300–314.

Tervo,S.,Pätynen,J.,Kaplanis,N.,Lydolf,M.,Bech,S.&Lokki,T.(2015),‘Spatialanalysisandsynthesisofcaraudiosystem and car cabin acoustics with a compactmicrophone array’, Journal of the Audio Engineering Society63(11),914–925.

Tervo,S.,Pätynen,J.,Kuusinen,A.&Lokki,T.(2013),‘Spatialdecompositionmethodforroomimpulseresponses’,JournaloftheAudioEngineeringSociety61(1/2),16–27.

Tervo,S.,Pätynen, J.&Lokki,T. (2013), Spatio-temporalenergymeasurements in renownedconcerthallswithaloudspeakerorchestra, in ‘the21st InternationalCongressonAcoustics (ICA’2013)’,Montreal,Canada. Invitedpaper.

why is it so hard to design a concert hall with excellent acoustics? · 2016-11-03 · receiver...

Documents