EFFECT%OF%WOOD%ON%THE%SOUND%OF%OBOE%AS%SIMULATED%BY%THE%CHANTER%OF%A%166INCH% FRENCH%BAGPIPE% % MathieuPaquier 1,EtienneHendrickx 1,RaphaëlJeannin 2 1UniversityofBrest,LabDSTICCCNRSUMR6285,6avenueLeGorgeuD29238Brest,France. Mathieu.Paquier@univDbrest.fr 263410Vitrac,FranceDjeanninraphael@yahoo.fr % ABSTRACT% % Many objective and subjective experiments on brass instruments, organs, flutes and clarinets haveshownthattheinfluenceofmaterialwasweak.yet,theinfluenceofwoodonthesoundofoboesis still to be determined. In this study, short musical recordings of ten French 16 bagpipes made of 5 differentwoods(africanebony,santosrosewood,boxwood,africanblackwoodandservicetree)were presentedtosubjects(specialistandnaïve),whohadtogivetheirfeedbackonseveralcriteria(global quality, warmth, aggressiveness, brightness, volume and attack precision). The choice of a bagpipe rather than a simple oboe enables to minimize the influence of the musician, as he is not directly in contactwiththereed.aninfluenceofthereedmaterialwasfound,butnoinfluenceofthewood.ina second experiment, a discrimination task allowed to confirm that the differences between chanters were not principally due to the wood. Several physical parameters calculated from recorded signals couldalsonotrevealanylargedifferencesbetweenwoods. Keywords):)wood,)oboe,)woodwind)instrument) 1. INTRODUCTION% % Very different opinions can be found among musicians, acousticians and musical instrument makersregardingtheinfluenceofmaterialsonthesoundofinstruments.whenthesoundisgenerated bythebodyoftheinstrument(aviolinforexample),thechoiceofmaterialscanbeessential(see[1]for stringinstruments,[2]fordrums).ontheotherhand,thesoundofwindinstrumentsisgeneratedby theaircolumninsidetheinstrumentanddependsonthemodeofaircolumnexcitation,theshapeofthe aircolumn(cylindricalorconical),andtheaircolumn slength,controlledbyopeningandclosingthe fingerholesontheinstrument.material,asitisnotdirectlyinvolvedinsoundgeneration,istherefore lesslikelytohaveasignificantimpactonsoundqualities. 1

1.1.%Brass%instruments% Theeffectofwallvibrationhasbeenstudiedwithbrassinstruments[3].Whitehouseetal.[4] haveshownthatmechanicalwallresonanceswereexcitedwhenasimplewindinstrument,consistingof a mouthpiece and section of metal piping, was artificially blown. The strength of those induced wall vibrationswasdependentonhowcloseinfrequencytheartificiallyblownresonancesandthestructural resonanceswere.thematerialofthepipeaffectedthepositionofthestructuralmodesandhenceits responsetoaparticularnote.in[5],onedthirdoctavesoundlevelmeasurementswererecordedforfour yellow brass and three nickel silver French horn bell flares of varying hardness. The sound level associatedwiththeunannealedbrassflareswashigherinthe1 3kHzrangethanwiththeannealed brassbellflares,whereastheoppositerelationshipwasobservedfornickel silverbellflares. Organisnotpartofthebrassfamily,butsomeoftheorganpipesaremadeofthinmetalwall.In [6],theresonatingaircolumninathinDwalledmetalorganpipewasobservedtointeractwithawall resonance.effectsbecameaudiblewhenawallresonancefrequencywasnearlythesameasthatofthe aircolumn.levelchangesof6dbandfrequencyshiftsof20centswerefound.in[7],theinfluenceofthe wall vibrations on the timbre of flue organ pipes have been studied by measuring wall velocity and sound spectra of wooden and metallic pipes. While large differences have been found in vibration spectra,onlyslightchangeshavebeenobservedinthesoundsignal. 1.2.%Woodwind%instruments:%wall%vibration% Thewallvibration(andthepossibleinfluenceofmaterialuponthisvibration)wasalsostudied for woodwind instruments [8,9,10]. The main physical process at the origin of sounds produced by woodwindsistheradiationoftheopenend(s)ofthewaveguide[11].themechanicalvibrationsofthe instrument wall may contribute to sound production by: i) structure/internal fluid interaction, ii) structure/externalfluidinteractionandiii)interdmodalcouplingduetotheradiationoftheopenendof the waveguide. In [12], a model for the vibroacoustic behaviour of an ersatz clarinet was presented, includingtheabovedmentionedkindsofcoupling.theradiatedsoundpowerfromthelateralwallwas foundtobemuchlowerthanthesoundpowerradiatedfromtheopenend.backus[13]alsoshowed thatthewallvibrationsofawoodwindinstrumentdonotaffect(ortoaverylowextent)itssteadytone either by radiating sound themselves or by affecting the harmonic structure of the internal standing wave. So it seems that the contribution of the wall vibration was quite negligible in pipes with no circularitydefault.however,thevibrationoftheaircolumncouldbealteredbyovalshapingofthewall andthestateoftheinternalsurface[14].moreover,ananalysisofrecordingsofatransversalflutemade fromthedriedstemoftheheracleumlaciniatum(withanirregularcircularity)waspresentedin[15] (Hanssen). While the lower octave exhibited conventional harmonic spectra, the upper octave surprisinglyincludedsubharmoniccomponents.authorsbelievedthatthesubharmoniccontributions wereduetononlinearoscillationsoftheflutematerial. 2

1.3.%Woodwind%instruments:%%wall%losses,%state%of%the%internal%surface% Beyonditsimpactonwallvibrations,materialscanhaveaninfluenceonthestateoftheinternal surface. Some studies [16,17] indicated that wall losses (frictions and thermal energy transfer to the instrumentwalls)haveagreateffectontheeigenvalues.benade[18]andfletcher[19]alsoindicated thattheviscouslossofenergytothepipewalls,alongwiththelossduetoconductionofheatintoand out of the air column from the walls during each cycle of the sound wave, both contribute to the dominantenergyexpenditureofmostinstruments.moreprecisely,yinandhoroshenkov[20]indicated thatporositymodifiedthehighordermodes.wegst[21]indicatedthatthetubematerialinfluencesthe soundoftheinstrumentanditsplayabilitybyvibrationaldampingduetoairfrictionatthetubewalls (lowerintubeswithasmoothfinish)andbyturbulencesinthevibratingairattheedges(whichare reducedwhentheextremityedges,asthoseofthefingerholes,arecutpreciselyandfinishedslightly rounded). It explains why the woods from which the wind instruments of the Western symphony orchestra are made traditionally are dense, have a fine structure and a high dimensional stability, especiallywhenexposedtohighlevelsofmoisture.theycanbeturnedanddrilledwithgreataccuracy, andtheyaresufficientlydimensionallystableundertheinfluenceofmoisture[22,23]. % 1.4.%Woodwind%instruments:%perceptive%effect%of%wood% The abovedmentioned studies have sometimes revealed an objective influence of materials (becauseofwallvibrationsortheinternalsurfacestate).nevertheless,thisinfluencewasnotalways audible: in [24], three keyless flutes of identical internal dimensions and made of thin silver, heavy copper,andwood,respectivelywereplayedoutofsighttomusicallyexperiencedobservers,whohadto indicatewhethertoneswerealikeornot.nosignificantcorrelationbetweenthelisteners'answersand thematerialoftheinstrumentwasfound.in[25],7fluteswithdifferentmaterialswereevaluatedby 110 persons. Although the sound analysis pointed out objective differences, statistical analysis on perceptualresultsshowedsubjectscouldnotdifferentiatebetweenmaterials. However,in[26],anickelsilver/copperalloyBundyandasilverMuramatsuwereused,andthe Bundywasfoundtobemore reverberant, whilemlsmeasurementsrevealedthatthemuramatsuhad morehighfrequencycomponents.theauthorsindicatedalargedifferencebetweenthetwoflutesin tonequality.in[27],twoflutesofthesamemakerandmodel,butwithonebeingmadeofgoldanthe other one of silver, were played and slight differences in the radiated sound of the two flutes were found.yetauthorsraisedsomequestions:wouldthosedifferencesalsobefoundintwo identical flutes (ofthesamematerial),sincenoinstrumentcanbeexactlyidentical?couldtwoflutesofthesamemaker, modelandmaterialsounddifferent,duetoslightdifferencesinmanufacturing? % % 3

1.5.%Wood%of%oboes% Studiesonoboequalityasafunctionofthewoodarerare.Pfeiester[28]usedanoboemadeof Grenadilla wood and another one made of a plastic resin, and found that there were noticeable differences overall such as larger amplitudes of the higher harmonics present in the wooden oboe. Moreover the wooden oboe impedances had higher impedance levels at high frequencies but often lowerlevelsforthefundamentalfrequencyofeachnote.higherimpedancelevelscanindicate(i)that, at that frequency, more pressure waves are bouncing back to the top of the instrument (so at the mouthpiece),makingtheinstrumenteasiertoplay,and(ii)thatthehigherharmonicshaveagreater impactonthesoundoutputresultinginamorecomplexsound(thistrendinthewoodenoboebecame evenmoreapparentintheuppernotes). Moreover, oboes have conical bore, with a pipe radius that gets very low close to the double reed. At this point, as the thickness of the boundary layer (in which turbulences are important and gradients of particle velocity and temperature are high) is large compared to the pipe radius, wood couldpossiblyhaveasignificantinfluence. 1.6.%The%french%16 %bagpipe% Thesoundofareedinstrumentisstronglydependentupontheplayer slipsposition.ifthepipe materialhasaneffect,theplayershouldbeabletocompensateforitwithhislips.bagpipesareworth beingusedforexperimentsonpipeperceptionbecausetheplayerhasnodirectinfluenceonthereed, sincethereedsofchanteranddronesareenclosedinstocks.the16 bagpipeisatraditionalinstrument fromthecentreoffrance.itconsistsofabag,usuallyablowpipeusedtoblowthebag,twodroneswith cylindricalboresandsinglereeds,andaquasidchromaticchanter;thesmalldroneplaysag3,whereas thebigoneplaysag2.thechanter,unlikethedrones,isequippedwithadoublereedandhasaconical bore.theseinstrumentsareexclusivelyhomemade,andthemostcommonwoodspeciesareboxwood, AfricanBlackwood,andServiceTree.Traditionally,thechanterdoublereedshavebeenmadeofcane, butnowadaysmoreandmoreplayersusesyntheticones(plasticisinterestingtomakereedsbecause theyarelessdependentonmoisturelevels,hightemperaturesandageing). Accordingtotheuniquestudyavailableontheperceptionofthematerialsofbagpipechanters (whichusedabagpipeclosetothe16 Frenchbagpipeofthepresentstudy)[29],soundsfromchanters made of various wood species seem to be different. This perceptive observation was confirmed by severalobjectivedifferencesinmeasuredspectra.however,differencesbetweenchanterscouldnotbe relatedtoanyphysicalpropertyofthewood,suchasdensity.moreover,thisstudywaslimitedbecause itreliedoni)theassessmentofchantersoundsbyonlyonelistenerandii)theuseofonlyonereedand onechanterperwoodspecies. In this study, short musical sequences played on a 16 Dbagpipe with chanters made of 5 different woods were recorded and presented through two tests to piperdlisteners and non piper 4

listeners. They were asked to assess the quality of sound during a first session, then to report quantitative feedbacks on brightness, aggressiveness, warmth, volume and attack precision during a second session. In an additional experiment, subjects capacity of discrimination between chantersmadeofdifferentwoodwereinvestigated.% % 5

2.%EXPERIMENT%A% 2.1.%Material%and%methods% % 2.1.1.Chantersandrecordings Thechantersundertestweremadeinduplicatefromdifferentspeciesofwood:AfricanEbony, SantosRosewood,Boxwood,AfricanGrenadillaandServiceTree.Thechanterswere44.5cmlong(the aircolumnwasactually4.8cmlongerbecauseoftheadditionalductofthedoublereed).astheinternal borewasconical,theinternaldiameterofthechanterwas3.8mmclosetothereedand19mmatthe endofthepipe.thedimensionsofthetenchanterswereidentical.whenmakingthechanters,arough cone was firstly dug inside a piece of wood using several drill bits of different diameters. Then, the internal shape was completed with conical reamers, which provide a very smoothed surface of the internalbore.finallythetubesofthetenchantersweretreatedwithboredoil(acommonpracticewith thistypeofinstrument)severalweeksbeforetherecording.thediameterofthe9fingerholesranged from4mmto6mm. Thechanterreedswereeithersyntheticormadefromcane.SincetheywerebrandDnew,they hadtobeusedforafewhoursbeforestartingrecordings.itisworthunderliningthattheaimofthe experimentwasnottoobservetheeffectsattributabletothereeds,butrathertoextendtheconditions ofplayingtomaketheexperimentsmorerealistic(indeed,somestudieshaveshownthat,withsome bagpipes,theroleofthereedcouldnotalwaysbenegligiblecomparedtotheinputimpedanceofthe pipe[30,31]). Thechantersweresuccessivelymountedonauniquebagpipe,sothattherecordingswouldbe madewiththesamedronesandthesamebag.insomebagpipes,themusicianhastoblowinthebag.as theairfromthelungsismoist,theworkingofreeds(especiallycanereeds)islikelytobeaffectedbythe progressiveincreaseofhumidity.inordertofreefromthisproblem,a16 «Bechonnet»bagpipewas used,asitallowstheplayertosendsomedryairinthebagbymovingaswell.thetwodroneswere madeofafricanblackwoodandwereequippedwithcommonsyntheticsinglereeds.allchantersand bagpipecomponentsweremadebyaprofessionalmaker. Foreachchanter,atraditionaltunefromFranceplayedonthechanterwiththetwodroneswas recordedinarecordingstudiowithasingledpa4006microphone,placed1.20mfromthepiperand 1.60mabovethefloor,andconnectedtoaPresonusFirestudiosoundcard(thesamplingfrequencyand quantizationwere48khzand16bitsrespectively).thetuningpitchwascontrolledwithanelectronic tuner.with5woods,2duplicatesperwoodand2reeds,atotalof20sequenceswasrecorded.each sequencewas20secondslong. 6

2.1.2.Testprotocol Subjectswereaskedtoassessglobalqualityinafirsttest,thenthefivecriteriainasecondtest. They were placed in front of a computer screen and equipped with Sony CDR2000 headphones, that theywereinstructednottomoveduringtheentiretestperiod[32]. Aftereachpresentation,thewords globalqualityofsound weredisplayedonthepcscreen with5boxes,from 1 (low)to 5 (high).thelistenerwasrequestedtoticktheboxthatmatchedat besthisfeeling(thetestinterfacewasimplementedinmatlab).forthesecondtest,theprotocolwasthe same,exceptthattherewere5criteriatoevaluate: brightness, aggressiveness, warmth, volume (refers to the volume of the sound by the chanter with respect to the sound by the drones) and detachedprecision.thesetermsweredeterminedonthebasisofapredstudy,duringwhichpipersand nondpipers had been asked to express at best how they qualified and differentiated sounds from bagpipes. Once a subject was satisfied with his answers, he had to press a button to go to the next stimulus. Eachlistenerwassuccessivelygiventhetwotests.Thefirstonelastedabout15minutes,andthe second one 25 minutes. Each test was preceded by a predtest of about 5 minutes to familiarize the listenerwiththeproposedrangeofsoundsandthedifferentcriteria.theaimoftheexperimentwasto assess the sound produced by the chanter played under normal conditions, that is with drones. Yet subjectswereremindedthatdroneswerenotthesubjectoftheassessment,andthattheyshouldfocus onchantersound.thesoundvolumeofthesequencesplayedintheheadphoneswasabout85dbsplto correspondtothetruevolumeofa16 bagpipe(at1meter). Amongthe18listenersinvolvedinthestudy,9werenonDpipermusicians.Theotheroneswere alltrainedpiperswithahighpracticelevel.thisdiversityinthepopulation undertestwasmadeon purpose to determine whether both populations of listeners had similar quality criteria to assess chantersounds. 2.2.%Results% Asthescalecouldnotbeconsideredcontinuous,datawereanalyzedwithnonDparametricprocedures, andrankswereratherusedthanmeans[33]. 2.2.1.Globalqualityofsound Wood)effect:accordingtothelisteners,woodhadnodirecteffectonthesoundproducedbychanters (p=0.103accordingtothefriedmantest[33]). Reed)effect:thelistenersgavesignificantlyhighermarkstochantersequippedwithsyntheticreedthan tothosewithcanereed(p<0.0001accordingtowilcoxontest[33]). Listener s)background)effect:globally,thesetofstimulireceivedhigherratingswithpiperlistenersthan with nonpiper listeners(p<0.0001accordingtomannwhitneytest[33]). 7

Effect)by)chanter)items:despitethelackofadirecteffectofwoodspeciesontheassessmentofsound quality, listeners showed some significant preferences for certain chanters, independently of their wood. The figure 1 indicates the ranks for each chanter. A chanter with a rank n means that the concerned chanter was, on average, sorted at the n th rank with respect to other chanters (the first chanter being the least preferred instrument). The figure indicates that the ranks for chanters made fromthesamewoodcanbemoredistantthanranksobtainedforchantersfromdifferentwoods.for example, the two chanters in African Ebony have very different ranks, whereas the second item of chantermadefromafricanebonyobtainedarankclosetothoseofthetwochantersmadefromservice Tree.Thisexamplesupportstheabsenceofasignificantglobalinfluenceofwood. 7 6.5 6 5.5 Rank 5 4.5 4 3.5 3 Ebo1 Ebo2 San1 San2 Gre1 Gre2 Box1 Box2 Ser1 Ser2 Chanter Figure 1. Global quality: ranks for the two chanters from the five different woods (namely African Ebony,Santosrosewood,AfricanGrenadilla,Boxwood,andServiceTree) 2.2.2.Othercriteria Concerning the criteria brightness, aggressiveness, warmth, volume, and detached precision,thefriedmantestindicatednoeffectofwood. Brightness)criterion:thereedandthelistenersbackgroundwerefoundtohavesomesignificanteffects: indeed, the chanters with the cane reed were judged as brighter than those with the synthetic reed (p<0.0001,wilcoxontest),andthe nonpiper listenersgavehigherbrightnessmarksthanthe piper listenerstothewholesetofsounds(p<0.0001,mannwhitneytest). Aggressiveness) criterion: the reed and the listeners background were found to have some significant effects:indeed,thechanterswiththecanereedwereconsideredasmoreaggressivethanthosewiththe syntheticreed (p<0.0001,wilcoxontest),and for all sounds the aggressiveness marks given bythe 8

nonpiper listenerswerealwayshigherthanthosebythe piper listeners(p<0.0001,mannwhitney test). Warmth)criterion:thereedwasfoundtohavesomesignificanteffects:thechanterswiththecanereed werejudgedaswarmerthanthosewiththesyntheticreed(p<0.0001,wilcoxontest). Volume)criterion:thereedandthelistenersbackgroundwerefoundtohavesomesignificanteffects:the chanterswiththecanereedwereconsideredaslouderthanthosewiththesyntheticreed(p<0.0001, Wilcoxontest);moreover,withrespecttothe piper population,the nonpiper oneassessedallofthe chantersaslouder(p<0.0001,mannwhitneytest). Detached)precision)criterion:thereedandthelistenersbackgroundwerefoundtohavesomesignificant effects: the chanters with the cane reed were considered as providing a better degree of detached precisionthanthosewiththesyntheticreed(p=0.002,wilcoxontest);moreover,thedegreeofdetached precision found by the non piper population was higher than by the piper population (p<0.0001, MannWhitneytest). 2.3.%Discussion% Themainresultofthisexperimentisthatwoodseemstohavenosignificantinfluenceonglobal soundquality. Thequalityofsoundsseemstobestronglydependentonthereedmaterial:inthisstudy,the syntheticreedwaspreferredbymostofthesubjects.thispreferencecanberelatedtotheresultsofthe second test where the sounds produced by the cane reed were felt to be brighter, warmer, more aggressive and louder than those by the synthetic one, and providing a better degree of detached precision.moreover,thispreference(atleastfor piper listeners)maybeduetothefactthatmostof pipersplayswithsyntheticreedsnowadays,andmaybemorefamiliarwiththeirsound. The listener background had a significant effect on the test results: the ratings of the sound quality by the non piper listeners were globally worse than those by the piper population; the former also considered all of the sounds as brighter, more aggressive and louder. Moreover, the detached precision on the whole set of chanter sounds was assessed by the non piper listeners as moreprecisethanbythe piper listeners.itisworthnotingthat nonpiper listeners reportedthat theyhadtroubleassessingthiscriterion. Thecorrelation between globalquality and the other criteria wasvery low, and, surprisingly, lowerforpiperlisteners.themaximumcorrelationwasreachedwithglobalqualityandwarmth,yet thecoefficientwaslow(0.33fornaivelisteners,0.29forexpertlisteners,withp<0.0001accordingto Spearmantest).Thoseverylowcorrelationsaresurprising,especiallyfromexpertlisteners,whohad determinedduringthepredstudythechoiceofcriteria. In this first experiment, subjects did not perceive significant differences of sound quality betweenwoodspecies.yetthey reportedthatthetaskwasdifficult.itisthereforeimpossible atthis point to determine whether wood was found to have a negligible impact on sound quality because 9

subjects could not hear any differences between the woods, or because the difficulty of the task hid potential differences, or simply because subjects could hear differences between woods, but did not haveanypreferences.moreover,thefactthatthedifferencesbetweentwochantersfromthesamewood are sometimes larger than the differences between chanters from different woods is surprising, and suggests that the variability in instrument manufacturing is more important than the choice of the wood. A second experiment was therefore carried out to verify whether subjects could truly differentiatebetweenthedifferentwoodspecies. % % 10

3.%EXPERIMENT%B% 3.1.%Material%and%methods% Inthisdiscriminationexperiment,a3Interval3AlternativeForcedChoice(3I3AFC)response paradigm was chosen. During a trial, three intervals were successively presented. Each of the three intervalswasa5dsextractfromthemusicalsequencesusedinexperimenta.thethreesequenceswere distinct recordings: two with a same chanter, and one with a different chanter (from identical or differentwood).theorderofthethreesequenceswasrandomized,andsubjectshadtoidentifywhich oneofthethreewastheoddballstimulus(thatisthechanterthatwasonlypresentedonce). As experiment A had shown that the influence of reed was very pronounced, chanters with plasticreedswerenevercomparedtochanterswithcanereeds.thenumberofpairstobecompared wastherefore[10*(10d1)]/2=45forthechanterswithsyntheticreedsandalso45forthechanterswith cane reeds. With a total of 90 pairs, the test was about 30 minutes, with a 5Dminute predtest to familiarizesubjectswiththetask. Amongthe22listenersinvolvedinthestudy,11werenonDpipermusicians.Theotheroneswere alltrainedpiperswithahighpracticelevel.thetestconditions(room,hardware )werethesameas thoseusedinexperimenta. 3.2.%Results% Firstly,resultsofexperimentBweresimilarfornaiveandexpertlisteners(p=0.84accordingto themannwhitneytest). Then, the influence of chanters and woods on the detection rate was quite close between syntheticandcanereeds(itdidnotaffecttheorderofrankssignificantly).yetthedetectionratewas globally higher with synthetic reeds (55,25% on average) than with cane reeds (44,8% on average). ThisdifferencewassignificantaccordingtoaWilcoxontest(p=0.001). Figure2indicatesthedetectionrateoftheoddballstimulus(thechanterplayedonetimeonly), foreachofthechanterpairs(pooledacrossallreedsandlisteners).thestarsabovebarsindicatepairs ofchanterscomingfromthesamewood,andrevealthatchantersfromthesamewoodaresometimes distinguishedmoreeasilythanchantersfromdifferentwoods.forexample,thetwochantersinsantos Rosewood were distinguished at 72% (it is even the most distinguished couple). On the other hand, many couples of chanters from different woods were distinguished with a detection rate inferior to 50%. Figure3presentsthesameresultsasfigure2,butitenablestocompareratesobtainedwithtwo chantersmadefromthesamewoodmoreeasily.alightercasecorrespondstoahigherdetectionrate. Thediagonalisblackbecauselistenerswereneverproposedthesamechanterinthethreeintervals.In most cases, the two items of a same wood (consecutive odd/even columns on the figure) gave quite differentresults.thisisparticularlyclearbetweenebonyitem1andebonyitem2:thosetwochanters werenotdifferentiatedfromtheotheronesinthesameway. 11

Figure 2. Detection rate of the oddball stimulus (the chanter played one time only), for each of the chanterpairs.starsabovebarsindicatepairsofchantersthatcomefromthesamewood. Figure 3. Detection rate of the oddball stimulus (the chanter played one time only), for each of the chanterpairs,onagrayscale.forexample,thegraycolorofthesquareattheintersectionofrowebo1 andcolumngre1indicatesthatthedifferentiationratebetweenthefirstchanteriteminafricanebony andthefirstchanteriteminafricangrenadillawas61%. 12

Figure4.Detectionrateofthechanterplayedonetimeonly,foreachwood(pooledacrossitems),ona gray scale. For example, the gray color of the square at the intersection of row San and column Box indicatesthatthedifferentiationratebetweenchanters(pooledacrossitems)insantosrosewoodand chantersinboxwoodwas53%. Figure4indicatesthedetectionrateforeachwood,pooledacrossitems.Thediagonalindicates thediscriminationratesfortwochantersfromthesamewood.itcanbeseenthatthedetectionratewas thehigherwhenthetwoduplicatesofsantosrosewoodwerecomparedbetweeneachother,andwhen the two duplicates of Grenadilla were compared between each other. It confirms that differences betweenchantersfromthesamewoodcanbelargerthandifferencesbetweenchantersfromdifferent woods.ontheotherhand,thetwoboxwoodduplicatesseemtobeveryclose.thoughresultscannotbe legitimately generalized (as there were only two duplicates perwood), they suggest that somewood species (boxwood for example) may have a more «constant» structure, and/or may provide a more constantmanufacturethanotherwoodspecies.constancydoesnotseemtoberelatedtodensity:for example, the two Grenadilla duplicates were better differentiated than the two Boxwood duplicates, whengrenadilladensityisfarsuperiortoboxwooddensity(1270kg/m 3 and975kg/m 3 respectively [34]). It is worth noticing that the reputation of the Service Tree (a particularly different sound compared to other woods, with a warmer and sweeter tone) was not verified in experiment A. Actually,thiswoodwasrarelydifferentiatedfromotheronesinexperimentB(thecolumnassociatedto ServiceTreeisthedarkestinFig.4). 13

Two groups can be distinguished: (i) African Ebony, Santos Rosewood, and Grenadilla, which seemquitedifferentbetweenthemandwithforchantersfromthesamewood,(ii)boxwoodandservice Tree, which seem closer between them and with for chanters from the same wood. In experiment A (figure1),thetwochantersinafricanebonywereverydifferentlyassessed.thesameobservationwas done, to a lesser degree, for Santos rosewood and Grenadilla. On the contrary, the two chanters in BoxwoodandthetwochantersofServiceTreewereclose.Sothetwoexperimentsareinagreement:the chanterswhichwasfoundtohavesimilarqualityratingsinexperimentawerealsothechanterswhich werelessdiscriminatedinexperimentb(andtheabsenceoflargequalitydifferencesinexperimenta seemstobenotduetothedifficultyofthetask). 4.%SIGNAL%ANALYSIS% % Experiments A and B suggest that the perceptual effect of wood is negligible. Objective measurements were also carried out on recordings of isolated notes (from G3 to C4), with the same chantersasinexperimentsaandb:soundpressurelevel(inpascalandindb),oecue(thelogarithm of the ratio between the sum of amplitudes of odd harmonics and the sum of amplitudes of the fundamentalfrequencyandtheevenharmonics),spectralcentroid(averageandtemporalevolution), irregularitycue(whichindicatestowhatextentenergyisconstantthroughconsecutivespectralbands), skewness(whichmeasureshowfaradistributionisasymmetric),kurtosis(measureswhetherthepeak ishigherorlowerthanthatofanormaldistribution),ratioai/a1(betweentheenergyoftheharmonici andtheenergyofthefundamentalfrequency),ratioai/σai(betweentheenergyoftheharmoniciand the total energy of the n harmonics), tristimulus 1, 2 and 3 cues, which respectively indicates the relative importance of the fundamental frequency, of the low harmonics (from i = 2 to 5) and of the harmonicsofrangesuperiorto5[35,36]. Only two physical parameters were significantly different between synthetic and cane reeds: TheOEcue(D10,8forsyntheticreeds,D14.4forcanereeds,p=0.002accordingtotheMANOVA),andthe spectralcentroid(6710forsyntheticreeds,6405forcanereeds,p=0.005accordingtothemanova). Only one physical parameter was significantly different between the woods: the spectral centroid(figure5,p=0.04accordingtothemanova[33]),whichwaslowerforchantersfromsantos Rosewood,thanforchantersfromAfricanEbony(significantlydifferentaccordingtoBonferronipostD hoc test [33]: p=0.025). This result is not really related to the perceptive results of previous experiments.however,theglobalabsenceoflargeobjectivedifferencesbetweensignalsfromdifferent woods is in agreement with experiences A and B, which had not highlighted perceptive differences betweenwoods. It is worth noting that no physical parameter was significantly different between chanters, independently of their wood. Perceptive results have shown differences between chanters, independently of their wood. However, differences in quality assessments (in experiment A) were 14

globallylow,andalltheseresultsindicatethatthedifferencesbetweenchanters,independentlyoftheir woodornot,areweak(perceptivelyandobjectively).thisresultisinagreementwithseveralstudies aboutotherwoodwindinstruments[12,13,24,25]. 7200 7000 Spectral centroid (Hz) 6800 6600 6400 6200 6000 Ebo San Gre Box Ser Wood Figure 5. Spectral centroid for the five woods (namely African Ebony, Santos rosewood, African Grenadilla,Boxwood,andServiceTree) 5.%CONCLUSION% Experiment A has not revealed any influence of wood on the sound quality assessment of chantersfrombagpipes.yetsyntheticreedsweremoreappreciatedthancanereeds,andratingswere globally higher with experts than with naïve subjects. Warmth was the most correlated criteria to global quality, yet the coefficient remains low. Independently of their wood, some chanters were preferredtoothers. Experiment B showed that chanters from the same wood could sometimes be distinguished moreeasilythanchantersfromdifferentwoods. The analysis of signals revealed that there was also little objective difference between wood species. Only the spectral centroid was significantly lower with the Santos Rosewood than with the AfricanEbony. Theinfluenceofwoodonthesoundofchantersfromfrench16 bagpipesisthereforelimited, andappearstobelessimportantthanmicroddifferencesinmanufacturing. ACKNOWLEDGMENTS% Weacknowledgeallthelistenerswho participatedtothisstudy, NicolasGrimault andsamuelgarcia fromthecrnl(cnrs/universitylyoni),andjeandpierredalmontfromthelaum(cnrs/universityle Mans). 15

REFERENCES% [1]Yoshikawa,S.(2007).Acousticalclassificationofwoodsforstringinstruments.TheJournalofthe AcousticalSocietyofAmerica,122(1),568D573. [2]Aramaki,M.,Baillères,H.,Brancheriau,L.,KronlandDMartinet,R.,&Ystad,S.(2007).Soundquality assessmentofwoodforxylophonebars.thejournaloftheacousticalsocietyofamerica,121(4),2407d 2420. [3]PyleJr,R.W.(1998).Theeffectofwallmaterialsonthetimbreofbrassinstruments.TheJournalof theacousticalsocietyofamerica,103(5),2834. [4] Whitehouse, J. W., Sharp, D. B., & Harrop, N. D. (2002, December). An investigation into wall vibrations induced in wind instruments constructed from different metals. In Proceedings of the internationalsymposiumonmusicalacoustics,mexicocity. [5]Lawson,B.,&Lawson,W.(1985).AcousticalcharacteristicsofannealedFrenchhornbellflares.The JournaloftheAcousticalSocietyofAmerica,77(5),1913D1916. [6] Nederveen, C. J., & Dalmont, J. P. (2004). Pitch and level changes in organ pipes due to wall resonances.journalofsoundandvibration,271(1),227d239. [7]Angster,J.,Paal,G.,Garen,W.,Miklos,A., TheEffectofWallVibrationsontheTimbreofOrganPipes, Proceedingsofthe16th.Int.CongressonAcousticsand135thJASAMeeting.Seattle,Vol.3,pp753D754, 1998. [8]RossingTD,FletcherNH.2004.PrinciplesofVibrationandSound.NewYork:SpringerDVerlag.2nd ed. [9]BenadeAH.1990.FundamentalsofMusicalAcoustics.NewYork:Dover.2ndrev. [10]OlsonHF.1967.Music,PhysicsandEngineering.NewYork:Dover.Rev.2nded. [11]Gautier,F.,&Tahani,N.(1998).Vibroacousticbehaviourofasimplifiedmusicalwindinstrument. JournalofSoundandVibration,213(1),107D125. [12]Gautier,F.,&Tahani,N.(1998).Vibroacousticbehaviourofasimplifiedmusicalwindinstrument. JournalofSoundandVibration,213(1),107D125. [13] Backus, J. (1964). Effect of Wall Material on the Steady State Tone Quality of Woodwind Instruments.TheJournaloftheAcousticalSocietyofAmerica,36(10),1881D1887. [14]R.Pico,andF.Gautier, Thevibroacousticsofslightlydistortedcylindricalshells:Amodelofthe acousticinputimpedance,journalofsoundandvibration302,18 38(2007). [15] Hanssen, A., Hindberg, H., OEigard, T. A., Birkelund, Y., & Hanssen, O. (2006, June). Analysis of HarmonicandSubharmonicEffectsinaTransversalFluteMadefromHeracleumLaciniatum.InSignal ProcessingSymposium,2006.NORSIG2006.Proceedingsofthe7thNordic(pp.194D197).IEEE. 16

[16] Benade, A. H. (1959). On woodwind instrument bores. The Journal of the Acoustical Society of America,31(2),137D146. [17] Howle, V. E., & Trefethen, L. N. (2001). Eigenvalues and musical instruments. Journal of computationalandappliedmathematics,135(1),23d40. [18]Benade,A.H.,&Gans,D.J.(1968).Soundproductioninwindinstruments.AnnalsoftheNewYork AcademyofSciences,155(1),247D263. [19]N.H.Fletcher(2000).Thewindofmusic artandscience,westpravii(seventhwesternpacific regionalacousticsconference),3d5/10/2000,kumamoto,japan [20]Yin,Y.,&Horoshenkov,K.V.(2005).TheattenuationofthehigherDordercrossDsectionmodesina ductwithathinporouslayer.thejournaloftheacousticalsocietyofamerica,117(2),528d535. [21]Wegst,U.G.(2008).Bambooandwoodinmusicalinstruments.MaterialsResearch,38(1),323. [22]Fletcher,N.H.,T.D.Rossing.1991.Thephysicsofmusicalinstruments,SpringerDVerlag,NewYork, NewYork,USA,p491. [23]Wegst,U.G.(2006).Woodforsound.AmericanJournalofBotany,93(10),1439D1448. [24]Coltman,J.W.(1971).Effectofmaterialonflutetonequality.TheJournaloftheAcousticalSociety ofamerica,49(2b),520d523. [25]Widholm,G.,Linortner,R.,Kausel,W.,&Bertsch,M.(2001).Silver,gold,platinumDandthesoundof theflute.inproc.int.symposiumonmusicalacoustics,perugia(vol.1,pp.p277d280). [26]Cocchi,A.,&Tronchin,L.(1998).Materialandobsolescenceonflutequality.ProcAcoustSocAm, 103,763D764. [27]Carral,S.(2010).GoldvsSilver:Doesmaterialinfluencethesoundofflutes?.InProceedingsofthe DAGA2010:36.(pp.779D780). [28]NisolePfiester(2008),NFS2008report,PerdueUniversity [29] Bernard Blanc (2004) Recherches sur les bois de cabrette, in Bois de musique, J.M. Ballu, Ed. Gerfaut,100D101. [30] Paquier, M., & Moign, C. (2005). Quality of Bagpipe drone reeds: plastic versus cane?. In Forum Acusticum,Budapest(Vol.29,pp.08D02). [31]Jeltsch,J.&Gibiat,V.(2000).Lefonctionnementacoustiquedescornemusesàtraverslesexemples de la Boha landaise et de la Chabrette limousine ; ou comment se retrouver très loin du modèle classiquedel instrumentàvent.proc.5èmecongrèsfrançaisd Acoustique,1D4/09,Lausanne,318D321 (2000) [32]Paquier,M.,&Koehl,V.(2015).DiscriminabilityoftheplacementofsupraDauralandcircumaural headphones.appliedacoustics,93,130d139. [33]Field,A.(2013).DiscoveringstatisticsusingIBMSPSSstatistics.Sage. [34]http://www.woodDdatabase.com/ 17

[35]Peeters,G.(2004).Alargesetofaudiofeaturesforsounddescription(similarityandclassification) CUIDADO project, IRCAM report, http://recherche.ircam.fr/equipes/analysed synthese/peeters/articles/peeters_2003_cuidadoaudiofeatures.pdf [36]Pollard,H.F.,&Jansson,E.V.(1982).Atristimulusmethodforthespecificationofmusicaltimbre. ActaAcusticaunitedwithAcustica,51(3),162D171. 18