THEPHYSICSOF MUSICALINSTRUMENTS RhysPhillips May2008

THEPHYSICSOFMUSICALINSTRUMENTS RhysPhillips

CONTENTS
1 2 3 4 5 6 7 8 AimsandObjectives AnIntroductiontotheTrumpet InvestigationintoGeneralHarmonicContent InvestigationintotheTransientPart ModellingtheLipReedGenerator SummaryofPossibleFurtherInvestigations ReferencesandAcknowledgements Appendix:RiskAssessment .....................................3 .....................................4 .....................................10 .....................................16 .....................................22 .....................................28 .....................................29 .....................................30

Myprojecthasthreemainaimsandobjectives. Firstly,togenerallyinvestigatethespectralcontentandtimehistoryofsoundpressuresignalsthat aregeneratedbythelipreedonthetrumpetandtoinvestigatehowtherelativestrengthsvarywith pitchand'loudness'.Secondly,tolookindetailathowtheharmoniccontentoftheimportant transientpartofthenotevaries.Finally,toconstructaverysimplecomputersimulationofthelip reedtohelpunderstandthemotionofthelipwhenplaying.

AIMSANDOBJECTIVES

Themoderndaytrumpetismadefrombrasstubing,withthreevalveswhichwhenpressed,change thelengthsoftheeffectivetubingoftheinstrumentallowingdifferentnotestobeproduced.The playerplaystheinstrumentby'buzzing'hislipsintothemouthpiece.Thisproducesastandingwave vibrationintheaircolumnofthetrumpet.Pitchisselectedusingtheaforementionedvalves combinedwiththeadjustmentofthetensionintheplayer'slips. Butthereisalonghistoryofthedevelopmentoflipdrivenwindinstrumentsstartingwithhollow plantstems,seashellsandanimalhorns[1],[5].Veryearlytrumpetsexistedaslongagoas EgyptiantimesanditwasRomantimesthatsawthefirstmetaltrumpetsemerge.Itisbecauseof theirrootsintheserudimentaryinstrumentsthattheearlierinstrumentboreswereconical.Early trumpetswereonlyaroundhalfameterinlength(bywhichwemeanthetotallengthofthe instrumentifitwasunravelledintoasinglelengthtube).Thislimitmeantthatthefundamental harmonicwasaround300Hzandonlytwoorthreeothernotes(correspondingtothemodesofthe fundamental)couldbeplayed.Thisiswhy'buglecalls'suchasTheLastPostplayedon RemembranceDayusealimitedrangeofnotes.InfactalthoughTheLastPostisoftennow playedonamoderndaytrumpet,itisplayedwithouttheuseofanyofthevalves,harkingbackto thebuglewhichdidn'thaveanyvalves. Someoftheseprimitiveinstrumentsweredevelopedbydrillingholesintothesidewitharesultnot dissimilartoamoderndaywoodwindinstrument.Theresthadtheirlengthsincreasedandtheir boresnarrowedmeaningthatthecloselyspaceduppermodescouldbeusedmelodically.Early militarytrumpetsusedthe3rdand6thmodesforsignalssoundedatcamp,ceremonialtrumpetsused evenmoremodesfortheirregalfanfaresandcoilednaturalhornswereabletoplayawholeoctave, albeitinadiatonicscale,notachromaticone. Atthispoint,Iwillbrieflyexplainwhatthewordsoctave,diatonicandchromaticmean.Fig.2.1 showswhatthenotesonapianolooklike:

ANINTRODUCTIONTOTHETRUMPET

Fig.2.1APianoKeyboard

www.purplehell.com

Everyeighthwhitenoteisthesamenoteasbefore,butanoctavehigher.TheintervalbetweenCand Gsay,wouldbeafifth.Onatrumpet,asageneralrule,thesamefingeringsareusedfornotesan octaveapart.Soallofthedifferent'C'sareplayedopen,ie.withoutcompressinganyofthethree valves.Thechromaticscaleconsistsoftwelvenotes,allone'semitone'apart.Thesetwelvenotesare thetwelvenotesonapianoforexample,betweentwo'C'sonapiano,thiswouldbeallofthe whiteandblacknotesfromCtotheBaboveit.Thediatonicscaleisbasicallythescaleonlyusing thewhitenotes.Thismeansthatsomeofthestepsaretonesapart(CtoDforexample)andsome arestillsemitones(EtoFforexample). Thelowermodesoftheinstrumentswerealsoavailabletotheplayerbutwerenotusedsooftenas thenotesthatwereproducedhadalargeseparationinpitch.Oneearlywayofovercomingthis problemwastointroduceaslidetoincreasethelengthofthenarrowcylindricalpartofthebore continuouslywhatwaseventuallydevelopedintothemoderndaytrombone.Howeveramore commonsolutionwastoaddvalves,usuallythreebutsometimesasmanyasfive,whichwhen compressed,increaseordecreasetheoveralllengthofthetubingbyswitchingextralengthsof tubingintotheinstrumentbore. Mostmoderninstrumentshoweverarenotpurelyconicalinfacttheyaredesignedsothatthe majorityofthelengthoftubingisvirtuallycylindrical.Aslongastheflareattheendofthe instrumentisaroundathirdofthetotallength,thenthefrequenciesavailabletotheplayerarestill wideenoughformelodicusage.Onebigadvantageofnearlycylindricaltubingistheabilitytoadd slidesbothintheobviouscaseofatrombone,andfortuningadjustmentinthetrumpet.Italso meansthatonsometrumpets,partsofthetubecanberemovedandextratubinginsertedtoadjust theoverallpitchoftheinstrument.Somoderndayinstrumentsarebasicallyalongcylindricaltube, albeitcurvedtomaketheinstrumentpracticalinshape,connectedtoaroughlyconicallyexpanding sectionandendingwithashortrapidflaredsection.Boththeanglesandproportionsoftubing involvedvaryfrominstrumenttoinstrument. ThemostcommonmoderndaytrumpetispitchedinBb(Bflat).ThismeansthataConthetrumpet (playedwithnovalvescompressed,knownas'openvalves')soundsasaBbonthepiano.Soevery noteplayedonthetrumpetsoundsatone(ortwonotesonthepiano)lowerthanwritten.Trumpets arenottheonlyinstrumentthatthisoccursfor.Inanorchestraforexample,atrumpetplayercould playthemusicwrittenforclarinetasthisisalsopitchedinBb.Howeverif(s)heplayedthemusic writtenforflute,itwouldbeatonelowerthroughoutandsotheplayerwouldhavetoplayeachnote atonehigherthanwrittenonthemusic.Thisiscalledtransposing.Inreality,thiswouldn'tcausea problemformosttrumpetersasorchestralmusicoftenhastrumpetpartswrittenfortrumpetsinA, C,D,Eb,E&F.InfactsomeofthosekeysaremorecommonthantrumpetinBbforcertain periodsinmusicalhistory.

Thereasonthatmusiciswrittenforthesedifferenttypesoftrumpetsisn'tbecausethecomposers wishedtomaketheaverageBbtrumpetplayerstrugglewhenplayingthemusic,butbecausethose differentkeyedtrumpetsexisted,andinfactdostillexist,thoughalotlesscommonlynowthan whenthemusicwaswritten.Manyplayersstillprefertoplaythismusiconthetrumpetitiswritten for,notbecauseitmakestheirlifeeasiertoavoidtransposing,butbecausethedifferentkey instrumentsdoactuallysounddifferent.TheEbtrumpethasamuchbrightersoundthantheBb instrument. Thisisrelatedtotheproportionsofcylindrical,conicalandflaredtubingused.Forexample,more gentlytonedinstrumentsmayhaveaconicalsectionthatisaroundtwothirdsofthetheentirelength whichgivesthemasmallconeangle,whereasbrightersoundinginstrumentswouldhaveashorter conicalsectionandamorepronouncedflareatthebellofthehorn.Itmustbenotedherethatthese arenottheonlyfactorsthataffectthetoneoftheinstrument[2],[14]. AnapproximationtotheshapeofatrumpetistheBesselhorn.Therelationshipbetweentheradius ofthebore,aandthedistancefromthehornmouth,xisgivenbythisequation: (2.1) wherex0isthepositionofthemouthofthehorn,bischosentogivethecorrectradiiatthesmall andlargeendsanddefinestherateofflareasshowninFig.2.2.

Ithasbeenshown[3]thatagoodapproximationofthefrequencies Fig2.2AfamilyofBesselhorns f[8] nofthemaximumimpedanceat thethroatofthehorncanbeexpressedthus: (2.2) wherecisthespeedofthesound,listhelengthofthehornandisaparameterwhichvaries accordingto.Infact~0.6for<0.8and~0.7for>0.8.

Soforahornwith=1,theresonancesaregivenbythefollowingharmonicrelation: (2.3) Notethatifbothx0andarezeroinvalue,wehavetheharmonicresultforacylinderwhereasfor anyothervaluesof,themodefrequencieshavenosimpleharmonicrelations. ABesselhorncaneasilybefittedontoacylindricaltubeaslongastheradiiarematchedcorrectly andthatthepositionatwhichitisjoinedisfarenoughawayfromx0thattheflareangleisvery small.Theflareanglecanbecalculatedusingthisequation: (2.4) Formostbrassinstruments,awouldbearound5mmandx0<10cm,meaningthattheanglewould belessthanhalfadegreeat50cmfromthemouth. Trumpetstendtohaveavalueofclosetoabout0.7[18],meaningthattheflareismuchmore abrupt(asshowninFig2.2).Althoughthesefitevenmoresmoothlyontocylindricaltubes,theydo requiresomeadjustmenttotheirshapetotunetheresonancesoftheinstrument[4],[10],[15],[18]. Thisisnormallydonewhendesigningtheinstrumentandthemodesthatarenormallyproduced giveamodeseriesresembling0.7f0,2f0,3f0,4f0etc.Asyoucansee,thefirstresonanceisclearlynot inlinewiththeothersandproducesaveryweaksound,notusedwhenplayingthetrumpet. Howeveritispossibleforplayerstousethenonlineareffectsoftheinstrumentaswellas manipulatingtheirlipstoproduceanoteatfrequencyf0,knownasa'pedal'note.Forthistheyrely ontheharmonicallyrelatedhigherresonancesofthenote. Anotherimportantfeatureofthetrumpetareit'sslidesandvalves.Theseareusedtofillthespaces betweenthelowerresonancesandthereforetoproduceacompleterangeofnotesatthelowerendof therange.Baroquetrumpetsdidn'thavethisfeature,apartfromthepossibilitytoswaplengthsof tubingtochangetheentirekeyoftheinstrument.Thisexplainswhymostbaroquemusichasthe trumpetsonlyplayingoneortwodifferentnotesthroughoutapiece,withmorevariationinthe upperregister.Forlatercomposers,thiswastoobasicandsovalveswereintroducedtoovercome thisproblem. Thelargestresonancegaptobefilledisbetweenthesecondandthirdnote.Thisisamusicalfifth andthereweresixextrachromaticnotesbetweenthemthatwereunobtainable.Thevalvesareof courselimitedtohavingonlytwopositionseachopenandclosed.Thismeansthatatleastthree valveswouldberequiredtoproducethe6notesinvariouscombinations.Ifwelabelthevalves1,2 and3,thenthereareinfactsevencombinationsofclosedvalves,namely1,2,3,12,13,23and 123.

Eachvalvelowersthepitchby1,2or3semitonesrespectivelyandsoincombinationcanbeusedto alsolowerthepitchby4,5or6semitones.ThevalvesonatrumpetareshowninFig.2.3.

Fig.2.3ATrumpet Thereisaslightmathematicalproblemtothisdesignhowever.Anequaltemperedsemitone involvesafrequencychangeofapproximately6%,sotolowerthefrequencyofanoteby1 semitone,thetubelengthmustbeincreasedto1.06timesit'soriginallength.Formulatingthisas1+ x,itisclearthatloweringthefrequencybynsemitoneswouldrequirethatthetubelengthbe increasedby(1+x)nwhichismuchgreaterthan1+nx.Soifthevalvesareadjustedsothatwhen loweringthepitchby1,2and3semitonesthenoteispreciselyintune,thenwhenplaying4,5or6 semitoneslower,thenotewillbesharp.Thesolutiontothisproblemistomakethethirdvalve slightlyflattocompensatefortheshiftsto4,5and6semitonesbelow.Forplaying3semitones belowpitch,acombinationofvalves1and2canbeusedinstead[2]. Belowiswhatatypicalwaveformofanoteplayedonthetrumpetlookslike.Theonegivenbelow wasrecordedbymyselfatthesoundstudiosattheCardiffUniversitySchoolofPhysicsand Astronomy.

Fig.2.4Waveformgeneratedbyatrumpet

Thisisbynomeansacompletelookatthetrumpethowever.Themouthpieceisanotherimportant featureoftheinstrumentandmustbematchedcarefullywiththeinstrument.Thelipreedisa pressurecontrolledsystemwhichfunctionsbestwhenworkingintoahighimpedance.Thereforeit isimportanttomakesurethatthemouthpieceresonanceemphasizesimpedancepeaksinthemajor playingrange.Theradiationefficiencyisanothercharacteristicthatisimportantinunderstanding thetrumpetasawholenonlinearitiessupplementtheinternalspectrumofeachmodeandthe effectiveradiationoftheseharmonicsmayhaveaneffectonthebrightnessofthetoneofthe trumpet. Therearealsobothsmallamplitudeandlargeamplitudenonlinearities,effectsofbothinputand reservoirimpedance,acousticspectra,theeffectsofmutesandthevariationinplayingtechnique thatallhavetheirowneffectsontheworkingsofandthesoundproducedbythetrumpet.Although theseareallveryimportantcharacteristicsthathavemajoreffectsonthetrumpet,Iwillnotgointo themherebutinstead,havingalreadyunderlinedthebasics,moveontomyinvestigation.

ThefirstthingIlookedatwasthegeneralharmoniccontentofnotesplayedonthetrumpet.Imade severalrecordingsofmyselfplayingvariouscombinationsofnotesatdifferentvolumes.Thiswas doneatthesoundstudioatthedepartmentusingabasicvocalmicrophoneandAdobeAudtionto recordthesoundsto.wavfiles.IthenranthesethroughtheMathCadworksheetprovidedwhich calculateandplottedtheFouriertransformsofmyrecordings.Thisallowedmetolookatthe spectrumoffrequenciesthateachnoteconsistedof.

INVESTIGATIONINTOGENERALHARMONICCONTENT

Fig.3.1Flowdiagramofrecordingprocess Thespectrumofthewholenotewasbeinganalysedatthispoint(detailedinvestigationintothe transientpartofthewavewillbelookedintolater).Icomparedthedifferentspectralpatternsfirstly fordifferentnotesofthescale,thenfornotesthatwereoctavesapartandfinallyforthesamenoteat differentintensities. Itwasimmediatelyobviousthatdifferentnoteshaddifferentharmoniccontents.Thiswastobe expected,soIstartedtolookatthedifferencesincontentbetweendifferentoctavesofthesame note.ImaderecordingsofconcertGatthreedifferentoctavesatroughlythesamevolume. NB: Thegraphsonthenextpagearenottothesameverticalscaleasitdidnotallowthedatato be examinedeasilyinthatformthequieterharmonicsforG3weretoosmalltoseeifthe maximumamplitudewassetto4000

10

Fig.3.2HarmonicSpectrumforConcertG3

Fig.3.3HarmonicSpectrumforConcertG4

Fig.3.4HarmonicSpectrumforConcertG5 Asexpected,eachtimethenoteisplayedanoctavehigher,halfthefrequenciesarelost.Wecansee fromthediagramsthattheremainingfrequencieseachtimecorrespondstotheevennumbered frequenciesoftheoctavebelow.Thisisduetotheratiooffrequenciesfornotesanoctaveapart.For example,theAabovemiddleChasafrequencyof440Hz.TheAthatisoneoctaveabovethathasa frequencyof880Hz,exactlydoublethefrequency.Thedifferentharmonicfrequenciesdisplayedfor eachnoteareintegermultiplesapart,sodoublingthefundamentalfrequencywillhalftheamountof lines.

11

Thisalsoexplainswhythespectralcontentvariedforeachnoteofthescalewherethereisaratio of2:1fornotesanoctaveapart,therearedifferentratiosforalloftheotherintervals.Forexample, twonotesthatareafifthapart(forexampleCandtheGaboveit)wouldhaveamathematical relationof3:2..Theotherthingtobenotedhereisthatasthenotesgothigher,thefrequencieshave ahigheramplitude.HoweverthisisprimarilydowntomyplayingtechniqueeventhoughIwas tryingtomaintainaconstantvolume,itisnaturalwhenplayingthetrumpetforhighernotestobe playedlouderinfactitcanbeveryhardtoplaynotesinthetopregisteroftheinstrumentatalow volume. Ithenlookedathowthespectralcontentvariedwithvolume.Sincethevolumeiscontrolledbythe amountofairpassingfromthelipstotheinstrument,thiscorrespondswithvaryingthestaticair pressurethroughtheinstrument.Interestingresultswerefoundatthispoint.Belowaregraphsof ConcertF4,playedat4differentvolumes.Thereisnodefinedscaletotheintensity,butitincreases graduallyfromVolume1toVolume4.

Fig.3.5HarmonicSpectrumforConcertF4atVolume1

Fig.3.6HarmonicSpectrumforConcertF4atVolume2

Fig.3.7HarmonicSpectrumforConcertF4atVolume3

12

Fig.3.8HarmonicSpectrumforConcertF4atVolume4 Wecanseethatatdifferentvolumes,wehavedifferentfrequenciesthatdominatethesound.It appearsthatwhenanoteisplayedlouder,higherfrequenciesstarttodominatethesound.Itiseasy tocomparethisonasinglegraphinthefollowinggraph,Ihavenormalisedallfourvolumestothe amplitudeofthefundamentalfrequencyforthehighestintensityandplottedthefrequenciesasline plotsonasinglesheet.

Fig.3.9ConcertF4atFourVolumes,NormalisedtotheHighestFundamental Ithenwentbacktotherecordingstudioandrecordedadifferentnoteatroughlythesamefour volumestoseeifsimilareffectscouldbeseen.TheconcertFisplayedasaGonatrumpetonan openpipe,i.e.withnovalvesbeingcompressed.Sotomakethingsfairandtomaketheresults meaningful,itseemedsensibletotryanothernoteplayedwithouttheuseofvalves,awrittenCfor thetrumpet,whichsoundsasaconcertBb3.

13

Fig.3.10ConcertBb3atFourVolumes,NormalisedtotheHighestFundamental Thereisclearlyaverysimilarpatternoccurringwiththisnote.SoIdecidedtohavealookathow theharmoniccontentvariedovertimethroughacrescendo.Irecordedmyselfdoingafairlylong crescendo,againontheconcertF4noteandplottedthefollowing3DwaterfallplotindBscale.

Fig.3.11WaterfallplotofConcertF4Crescendo Wecanseethatastimeprogresses,notonlydoestheoverallamplitudegetlouder,butdifferent frequenciespeakatdifferentpoints,sotheoverallsoundisconstantlyvaryingthroughoutthe crescendo.Eachofthepeaksonthegraphisanovertoneofthefundamentalfrequency.

14

Beforemovingontodetailedanalysisofthetransientperiodofthenote,Iwantedtocheckifthe acousticsoftheroomwerehavingasubstantialeffectonmyresults.Obviously,theroomacoustics willhavesomesmalleffect,soIrecordedmyselfclappinginthesameroomtwentytimes.Ithen lookedattheendofeachclapandcomparedthiswiththeendsofnotesthatIhadrecordedtoseeif anynoticeablesimilaritiesoccurred.Belowisanexampleofoneofthesecomparisons.

Fig.3.12Exampleofendofrecordedwaveform(concertF4)

Fig.3.13Waveformofarecordedhandclap Asyoucansee,thereisnothingthatreallystandsouthere,soitissafetoassumethattheroom acousticsarenotgoingtoaffectmyoverallpatternsandtrendsthathavebeenobservedsofar.If somethinghadbeennoticed,thenextstepwouldtohavebeentorepeattherecordingsmadesofar inananechoicchamberandthenanalysetheresults.Althoughtherecordingswouldhavebeenless effectedbyroomacoustics,itisextremelyunlikelythattheoveralltrendswouldhavechangedgiven thattheyaresostronglycorrelated.

15

Thestarting(orattack)transientisanimportantfeatureofanoteproducedbyanymusical instrument.Bytransient,wemeantheverystartofthewave,beforeitsettlesdownintoregular oscillatoryorperiodicmotion.Itisknowntobeveryimportantinidentifyingtheinstrumentwhich isplayingforexample,ifyoulistentoarecordingofanoteproducedbyatrumpetbutcuttingoff thetransient,itsoundsratherartificial.Infact,youcouldeasilymistakeitforanoboeoranorgan sound.Instrumentsthatusealipreed,suchasthetrumpetdifferfromallotherinstrumentsdueto thelipgeneratorbeingabletoactindependently,withoutusingresonancesofthetube.Luceand Clark[13]measuredmanyofthepropertiesoftheinitialtransientforbrassinstruments.They foundthatpartialswithfrequenciesbelowthecutofffrequencyoftheinstrumenthadrelative amplitudeswhichallbuiltuptogetherandsoreachedtheirsteadystatevaluesatapproximatelythe sametime.Howeverthoseabovethecutofffrequencybehaveddifferently,withamplitudesthat builtupslowerandsopartialswithhigherfrequenciesreachedtheirsteadystateslaterthanthoseof lowerfrequencies. RissetandMathews[16]analysedinstrumentsoundsandsynthesisedsoundsonacomputer. Althoughthebuildupofthehigherharmonicsdidn'tsplitintotwodifferentbehavioursforthetwo regionsthatLuceandClarkfound,theydidfindthathigherharmonicsreachedtheirmaximum valuesafteralongerperiodoftimethanlowerharmonics.Itisthisformoftransient(ie.thetime beforeamaximumvalueisreachedasopposedtothetimebeforeperiodicmotionisobserved)that isknowntobeimportantinsubjectivelyidentifyinganinstrumentfromit'ssound. Lookingattransientsinthetimedomain,thevibratinglipseesthecharacteristicimpedanceofthe tubeofthetrumpetatthestartofthenote.Thehornitselfhasnoeffectonthenoteuntilthefirst reflectionfromtheopenbellreturns.Thetimetakenforthistooccurisroughlytheperiodofthe firstmodeandthisisindependentofthenotebeingplayed.Thismeansthatforhighernotes,the lipsgothroughmoreoscillations,sothetransientlengthislonger,makingitmoredifficultto correctlypitchthenote. Duringthetransientperiod,theamplitudesofthecyclesareincreasingi.e.averyshortcrescendo. SoIattemptedtosimulateatransientfrommyrecordingofacrescendo.Firstly,bycopyingthe sinusoidalcyclesofthewave,Iextendedthelengthoftheloudpartofthenote.IthenusedAdobe Auditiontoplaythenoteat8timesthespeed(meaningthatitonlyplayedoneeighthofthedata) andwhatIheard,althoughdidn'tsoundexactlyasIhadbeenaimingfor,wasnotabadresult.The resultantsoundresembledabadtrumpeterplayinganote,withamuffledsortofattacktothenote. Howeveritwascloseenoughtoidentifythenoteasthatofatrumpet.Byattack,Imeanthewayin whichthetrumpeteruseshis/hertonguetostartthenote.Beginnersoftenjustblowthenoteout resultinginaphwasound,notdissimilartothesoundproducedfrommanipulatingtherecording ofthecrescendo.However,moreexperiencedplayerslearnto'tongue'thenotebyensuringthat thetipofthetonguetouchesthetopofthemouthatthestartofthenote,acleanertasoundis formed.

INVESTIGATIONINTOTHETRANSIENTPART

16

MyinitialinvestigationsintothetransientsinvolvedtakingFouriertransformsofthestartofthe transientperiodofthenotes.However,ascanbeseenhere,thisapproachdidnotreallyproduceany usefulinformation.BelowistheFouriertransformofthetransientofaconcertFnote.Wecansee thesameharmonicsthatwesawforthesamenotewhentheFouriertransformwastakenoverthe wholerecording.Soinstead,Imeasuredthetransientlengthsforvariousnotesandlookedfor patterns.Idecidedtocomparethetransientlengthagainstrelativeintensity,pitchandlengthof tubing.

Fig.4.1FFToftransientperiodofconcertF4 BeforeIanalysedtheseresults,tocalculatetheerrorsinmyreadings,Ifirstmeasuredthetransient lengthof20recordingsoftwodifferentnotesconcertBb3andconcertB4.Plottingtheselengths showsthatthereisaroughlyGaussiandistributionaboutthemascanbeseenhere:

0.07

0.06

Transient Length (s)

0.05

0.04

0.03

0.02

0.01

Fig.4.2ConcertBb3transientlengthssortedintoascendingorder

17

0.05 0.04 0.04

Transient Length (s)

0.03 0.03 0.02 0.02 0.01 0.01 0

Fig.4.3ConcertB4transientlengthssortedintoascendingorder Thismeansthatcalculatingthestandarddeviationwouldgiveameaningfulapproximationofthe errorsintheresults.Sotheerrorbarsoneachofthefollowinggraphsrepresentthestandard deviationofeachnotefoundfromrepeatedreadingstheplottedpointisthemeanaverageofthe readings. Firstly,Ilookedathowtransientlengthvariedwithrelativeintensity.Asbefore,Ihavelabelledmy intensitiesas'1','2'&'3',with'1'beingthequietestand'3'theloudest,butagainitmustbestressed thattheseintensitieshavenotbeenmeasuredinanywaytheyarejustincreasinginintensityas judgedbymyownear.Beforeweanalysetheresults,herearethewaveformsforthedifferent volumesofconcertF5whichclearlyshowthattheintensityhasincreased.

Fig.4.4ConcertF5Waveform(Volume1)

18

Fig.4.5ConcertF5Waveform(Volume2)

Fig.4.6ConcertF5Waveform(Volume3) ThegraphsplottedbelowshowstherelationshipfortwonotesaconcertBbandaconcertF(C andGbothplayedonopenvalvesonthetrumpet).

Fig.4.7Intensityvs.TransientLength(ConcertBb4)

19

Fig.4.8Intensityvs.TransientLength(ConcertF5) Wecanseethatasintensityincreases,thetransientlengthdecreases.Thiscanbeexplainedfrommy earlierresultswhenIfoundthatasweincreasedtheintensityofanote,thehigherharmonicsstarted todominatetheharmonicspectrum.Sohere,byincreasingtheintensity,lessoscillationsneedtobe madebeforethemaximumvalueisreachedandsothetransientlengthisshorter. Secondally,Iinvestigatedtherelationshipbetweentransientlengthandpitch.HereIhaveplayedthe notesBb3,F4andBb4,allatapproximatelythesameintensity.

Fig.4.9Pitchvs.TransientLength

20

Herewecanseethatasweincreasethepitch,weagaindecreasethelengthofthetransientpartof thewave.Thiscanbeexplainedbythefactthatasatrumpeterplayshighernotes,itisnormalto playthesenoteswithagreaterintensity.Infact,itcanbequitehardtoplayhighpitchednotes quietlyduetothelongtransientlengththatshouldbeexpected.Sothenotesarebeingplayed slightlylouder,eventhoughIwasattemptingtokeepthevolumeroughlythesameforallofthese recordings,thusthepatterncanbeexplainedfromtherelationshipwithintensity. Finally,Ilookedatthevariationintransientlengthwithlengthoftubing.Asalreadyexplained, depressingdifferentcombinationsofthevalvesonthetrumpetchangestheeffectivelengthofthe tubingofthetrumpet.Forthispartoftheinvestigation,Iplayedthreenotesofsimilarpitchon differentlengthsoftubing.Rememberthatthethreevalvescorrespondtorelativelengthsoftubing 0.5,1and1.5respectively.Obviouslyitwasnotpossibletoaccuratelyplaytheexactsamepitchon threedifferentvalvecombinationsbutitwasdesirabletokeepthepitchesasclosetogetheras possiblesothateffectsofchangingpitchdidn'tinfringeupontheresults.IplayedaconcertF4on openvalves(relativetubelength0),concertE4onvalve1(relativetubelength0.5)andconcert Db4onvalves2&3(relativetubelength2.5).

Fig.4.10RelativeTubeLengthvs.Intensity Asexpected,asthelengthoftubingisincreased,thetransientlengthalsoincreasesasitwillnow takelongerforthewavestotraveldownthetubemanyFrenchHornplayersusetheexcuseof havingsuchalongtubedinstrumentforsplittingnotes!Theresultsmaypartlybebiasedhereas thepitchhasdecreasedslightlyasthetubelengthhasbeenincreased,andsowewouldexpectthe transientlengthtoincreasefromourearlierresults.

21

Therehavebeenseveraldifferentmodel'softrumpeter'slips,butthemostpopularisasimpleone massmodel.Cullenetal.[6]hasshownthatsuchasimplemodelcaneasilypredictordersof magnitudeforboththepressureandthethresholdfrequencywhicharerealisticwhencomparedto thoseseenwhenusingamechanicalmodelofthelips.Althoughsomemorespecificresultscannot bereproducedusingthismodel(itlacksvitalmechanicalboundaryconditions[17]),fornowitwill besufficientforourcomputersimulation. Thelipreedonthetrumpetcanbethoughtofasapressurecontrolledvalveheldclosedbyaspring. Theblowingpressure,p0tendstoblowthevalveopen,theinternalpressurepisthecontrolvariable, thecoordinatexgivesthepositionofthecaprelativetoitsseatandx0isitsequilibriumposition(so thatx0>0).

MODELLINGTHELIPREEDGENERATOR

Forasimplespring,theequationof motioncanthenbewrittenthus:

Fig.5.1Pressure ControlledValve [8] (5.1)

whereisaconstantthatdependsonboththeareaofthevalvecapandthespringcompliance. Assumefornow,thattheareaoftheopeningbetweentheseatandthevalvecapitselfis proportionaltox.Also,wecanassumethatBernoulli'sequationappliesherebecausethepressures arehighenough.Thismeansthatwecanwritethesteadyvolumeflowthroughthevalve,U,as: (5.2) where1isdifferentpositiveconstanttothatinequation5.1. Wecanthencombineequations5.1and.52toget: (5.3) where2isafurtherpositiveconstant.

22

Rememberingthatxo>0andassumingthatp<<p0,thenwehavethefollowingplotUinterms of(p0p):

[8] Fig.5.2Steadyflowcharacteristicsofthe valve Ifpisanacousticpressurethatvariesinasinusoidalmanneratsomecharacteristicfrequency,say <<r(theresonancefrequencyofthevalvesystem),thenthedisplacementofthevalvewill followthepressurechange(pp0),andequation3abovewillstillbetheequationoftheflow.

[8] Fig.5.3LipReedGenerator ItisclearfromFig.5.3thattheopenareaofthevalveisn'tproportionaltoxasassumedaboveit canbeassumedthatitisinsteadproportionalto|x|.TheBernoulliflow(equation5.2)willtherefore alsoneedtobeadjustedslightlytogive: (5.4) where1~1and2~0.5. Someofthepressuredifferenceisusedtoactagainsttheacousticinterferencefromtheairinthe reedchanneltoaccelerateit.Thepressurebeingusedinthiswayisthenclearlynotavailableto maintaintheflow,andsoequation5.4canbeadjustedtothisform: (5.5) where4|x|1(dU/dt)describestheinertanceoftheair,3isaconstantrelatedto1and4isa constantrelatedtothegeometryofthevalvechannel.

23

Sofarwehaveonlyconsideredthestaticformandsoweneedtoadaptthistoincludetheeffectsof reedresonanceandtheBernoulliforce.Althoughthereedmayinfacthavemorethanone mechanicalresonance,itisacceptabletojustconsiderthefundamentalwithfrequencyasthis tendstobetheonethatismosteffectivetothevalveaction.Takingthisintoconsiderationgivesus: (5.6) wheremristhemassoftheplayer'slips,>0isrelatedtothedampingofthereed,>0isrelated tothegeometryoftheexposedreedfacesand<0isrelatedtothegeometryoftheinternalflowin thenarrowpartofthegapvalveandgivestheBernoulliforce. Equation5.6isnonlinearsoitwasdifficulttofindacompletesolution.Infactnumericalmethods lookedliketheonlyoptionandsoIsolveditusingafinitedifferencemethod.Todothis,Iused thesethreerelationships: (5.7) (5.8) (5.9) Substitutingtheseexpressionsintoequation5.6,givesthefollowingequation: (5.10) Forthevariousparametersintheequation,Ihaveputknownvaluesinforasmanyparametersas possibleandforthosewherethevaluesareunknown,Ihavemaderealisticestimates: U=2 mr=0.02 r=2 (p0p)=0.5 h=0.01 =0.5 =0.2 =0.2 x0=1 x1=1.2

24

Plottingthegraphofxiagainstiwiththesevaluesgivesusanincreasingoscillatingmotionwhichis tobeexpectedthebuzzingmotionofthelipswhileplayingthetrumpetinvolvesthemmoving upanddowninasinusoidalmanner.Theexponentialtypeincreaseisalsotobeexpectedasthelips startoffatrestandincreaseinfrequencythroughthetransientperioduntiltheyreachtheir maximumvalue.

Fig.5.4ComputerSimulationoftheLipReedGenerator Ofcourse,varyingthedifferentparameterschangesthegraphslightlybutthesamesinusoidal motionisstillseen. Thevalveconductanceofthelipreedispositive,andthereforedissipativeoverthewholefrequency rangeofthereed(exceptforaverysmallregionjustaboverandwhen0<<0.5).Thiswillnot functionoveralargerangeoffrequenciesandsoitisbesttoadjustittobeslightlyunderthe oscillatorthatitisgoingtodrive'soperatingfrequency. Tocompletethediscussionofthelipreedgenerator,itisnecessarytoconsidertheacousticflow associatedwiththephysicalmotionofthesurfaceofthereed,U',whichalsocontributestothe acousticadmittanceofthereed.

25

Assumingthatthisflowispositive,wecanwritethisas: (5.11) where5isproportionaltotheareaofthevibratingfaceofthereedasseenfromtheinstrument bore. Theacousticadmittancecannowbeexpressedas: (5.12) p0leadsU'bynearly90andsoY'rhasanegativeimaginarypartandisthereforeinertive.Asa generalrule,wecanignoretheeffectsofY'rbecauseitisnegligiblecomparedtoYr. Thereedgeneratoradmittancecanbeplottedbyusingthecomplexplane.AccordingtoFletcher[7], theparametercanbeeithertheblowingpressureorfrequency.Fig.5.5showsthebehaviourasa functionoftheblowingpressure.

Fig.5.5Complexadmittance asafunctionofblowing pressure [8] ThereactivepartofYrisinductiveabovethethresholdblowingpressurep*0andtheincreaseofYris unlimited,basedontheassumptionswehavemadesofarinthisdiscussion.Howeverinreality, theseassumptionsdonoteholdtrueforverylargevaluesofp0.

26

Fig.5.6belowshowsthecomplexadmittanceYrplottedasafunctionoffrequencyforvarious valuesoftheblowingpressurep0.

Fig.5.6Complexadmittance asafunctionoffrequency [8] Finally,itshouldbenotedthatexperimentsandmeasurementsmadeonpracticalblownreed generatorsbyFletcheretal.[9]confirmtheabovecalculatedbehaviourasexpected.

27

Thereisstillfurtherresearchtobecarriedoutinthisarea.Iftherehadbeenmoretimeavailable,I wouldhaverepeatedmyinvestigationsbutusingrecordingsmadeintheanechoicchamber. AlthoughIhavestatedthatthiswouldbeunlikelytomakeanydifferencetomyoverallpatternsand trends,itwouldstillbeinterestingtoseewhatthedifferencesifanywere. Anotherpossibilitywouldbetorepeattheinvestigationusingmutes.Mutesareplacedintothehorn ofthetrumpetwhichiswhereallofthesoundisradiatedfrom.Asthenamesuggests,theirprimary roleistoreducethevolumeofthenotesbeingplayed,butdifferentmutesalsochangetheoverall soundindifferentwaysbecausethereductioninvolumeisdependentonthefrequency.Some commonmutesforthetrumpetincludethestraightmute(athinmute,normallymadeofmetal, givingasharp,tinnysortofsound),cupmute(similartothestraightmutebutwithacupshapeover theend,generatingasoftersound)andplungermute(effectivelytherubbersuctionpartoffatoilet plungerwhichgivesthe'wahwah'soundoftenusedinGlennMillerstylejazzmusic).Furtherwork couldbecarriedouttolookatthechangeinharmoniccontentwithdifferentmutes,andindeedhow thetransientlengthvarieswithdifferenttypesofmutes. Finally,thelipmodellingpartoftheinvestigationcouldhavebeenextendedtosimulatevarious modelsthathavebeensuggested[11],includingthesinglemechanicaldegreeoffreedommodel proposedbyCullenetal.[6]andthe2massmodelofLousetal.[12].

SUMMARYOFPOSSIBLEFURTHERINVESTIGATIONS

28

Allreferencesfortheprojectarelistedbelow.Itshouldbenotedthatthebulkofthetheory equationsusedinthisprojectarebasedonideasfrom[8],whichwouldmakeitcumbersometo referenceeachindividualpoint.

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] Baines,A.(1966).EuropeanandAmericanMusicalInstruments,Viking,NewYork Bate,P.(1966).TheTrumpetandtheTrombone,Norton,NewYork Benade,A.H.(1976).FundamentalsofMusicalAcoustics,pp391429OxfordUniversityPress Cardwell,W.T.(1970).Cupmouthpiecewindinstruments,U.S.PatentNo.3507181 Carse,A.(1939).MusicalWindInstruments,Macmillan,London Cullen,J.S.,Gilbert,J.andCampbell,D.M.(2000).Brassinstruments:Linearstabilityanalysisand experimentswithanartificialmouth,ActaAcustica86707724 Fletcher,N.H.(1979).Excitationmechanismsinwoodwindandbrassinstruments.Acustica43,6372 Fletcher,N.H.andRossing,T.D.(1991).ThePhysicsofMusicalInstrumentspp347394,SpringerVerlag Fletcher,N.H.,Silk,R.K.andDouglas,L.M.(1982).Acousticadmittanceofairdrivenreedgenerators. Acustica50,155159 Kent,E.L.(1961).Windinstrumentofthecupmouthpiecetype,U.S.PatentNo.2987950 Lopez,I.,Hirschberg,A.,VanHirtum,A.,Ruty,N.andPelorson,X.(2005).PhysicalModelingofBuzzing ArtificialLips:TheEffectofAcousticalFeedback,ActaAcusticaunitedwithAcustica92,10471059 Lous,N.J.C.,Hofmans,G.C.J.,Veldhuis,R.N.J.andHirschberg,A.(1998).Asymmetricaltwomass vocalfoldmodelcoupledtovocaltractandtrachea,withapplicationtoprosthesisdesign.ActaAcustica unitedwithAcustica84,11351150 Luce,D.,andClark,M.(1967).Physicalcorrelatesofbrassinstrumenttones.J.Acoust.Soc.Am.42, 12321243 MorleyPegge,E.(1960).TheFrenchHorn,ErnestBenn,London Pyle,R.W.(1975).Effectivelengthofhorns,J.Acoust.Soc.Am.57,13091317 Risset,J.C.andMatthews,M.V.(1969).Analysisofmusicalinstrumenttones.PhysicsToday22,2330 Vilain,C.E.,Pelorson,X.,Hirschberg,A.,LeMarrec,L.,Op'tRoot,W.andWillems,J.(2002).Contribution tothePhysicalModelingoftheLips.InfluenceoftheMechanicalBoundaryConditions,ActaAcusticaunited withAcustica89,882887 Young,F.J.(1960).Thenaturalfrequenciesofmusicalhorns,Acustica10,9197

REFERENCESANDACKNOWLDGEMENTS

[13] [14] [15] [16] [17]

[18]

IwouldliketothankDr.BernardRichardsonattheCardiffUniversitySchoolofPhysicsand Astronomyforhishelpthroughoutthisproject,andalsoacknowledgethefollowingpaperswhich althoughhavenotbeenreferencedinthisproject,haveprovidedessentialbackgroundreading:

[19] Youshikawa,S.(1994).Acousticalbehaviourofbrassplayer'slips,J.Acoust.Soc.Am.97,19291939 [20] Adachi,S.andSato,M.(1995).Trumpetsoundsimulationusingatwodimensionallipvibrationmodel, J.Acoust.Soc.Am.99,12001209 [21] Chen,F.andWeinreich,G.(1995).Natureofthelipreed,J.Acoust.Soc.Am.99,12271233 [22] Elliott,S.J.andBowsher,J.M.(1982).RegenerationinBrassWindInstruments,SoundandVibration83, 181217 [23] Moore,T.R.,Shirley,E.T.,Codrey,I.E.W.andDaniels,A.E.(2003).TheEffectsofBellVibrationsonthe SoundoftheModernTrumpet,ActaAcusticaunitedwithAcustica91,578589 [24] Vlimki,V.(2004),PhysicsBasedModelingofMusicalInstruments,ActaAcusticaunitedwithAcustica90, 611617 [25] Poirson,E.,Petiot,J.F.andGilbert,J.(2005).Studyofthebrightnessoftrumpettones,J.Acoust.Soc.Am. 118,26562666

29

BeforeIembarkeduponthisproject,aRiskAssessmentwasdrawnupasfollows: Therearetwomainrisksassociatedwiththisproject. 1. Usingthenonventilatedanechoicchamberforrecordings 2. Usingmainspoweredelectricalequipment Toensurethattheworkiscarriedoutsafely,Iwillabidebylocalrulesandensurethatallequipment usedhasapassedPATtestasrequired. ItshouldbenotedherethatthiswasdrawnupbeforeIdecidednottousetheanechoicchamberfor myinvestigation.

APPENDIX:RISKASSESSMENT

30