Scavone Vocal Tract

8/12/2019 Scavone Vocal Tract

1/10

Measurement of vocal-tract inuence during saxophoneperformance

Gary P. Scavone, a Antoine Lefebvre, and Andrey R. da SilvaComputational Acoustic Modeling Laboratory, Centre for Interdisciplinary Research in Music Mediaand Technology, Music Technology, Schulich School of Music, McGill University, Montreal, Qubec, Canada H3A 1E3

Received 5 July 2007; revised 11 January 2008; accepted 13 January 2008

This paper presents experimental results that quantify the range of inuence of vocal tractmanipulations used in saxophone performance. The experiments utilized a measurement system thatprovides a relative comparison of the upstream windway and downstream air column impedancesunder normal playing conditions, allowing researchers and players to investigate the effect of vocal-tract manipulations in real time. Playing experiments explored vocal-tract inuence over thefull range of the saxophone, as well as when performing special effects such as pitch bending,multiphonics, and bugling. The results show that, under certain conditions, players can create anupstream windway resonance that is strong enough to override the downstream system incontrolling reed vibrations. This can occur when the downstream air column provides only weak support of a given note or effect, especially for notes with fundamental frequencies an octave belowthe air column cutoff frequency and higher. Vocal-tract inuence is clearly demonstrated when pitchbending notes high in the traditional range of the alto saxophone and when playing in thesaxophones extended register. Subtle timbre variations via tongue position changes are possible formost notes in the saxophones traditional range and can affect spectral content from at least8002000 Hz. 2008 Acoustical Society of America. DOI: 10.1121/1.2839900

PACS number s : 43.75.Ef, 43.75.Pq, 43.75.St, 43.75.Yy NFH Pages: 23912400

I. INTRODUCTION

Since the late 1970s, there has been signicant interestin understanding the role and inuence of a players vocaltract in wind instrument performance. Acousticians generallyagree that, in order for such inuence to exist in reed-valveinstruments, the players upstream windway must exhibit in-put impedance maxima of similar or greater magnitude thanthose of the downstream air column. 1 Most musicians concurthat they can inuence sound via vocal-tract manipulations,though there is less consensus in terms of the extent of suchinuence or the specics of how this is done. 2 A number of previous studies have been reported but attempts to demon-strate and/or quantify vocal-tract inuence have not beenconclusive. These analyses have been complicated by thefact that measurement of the upstream windway congura-tion and impedance are difcult under performance condi-tions.

It is the aim of this study to provide experimental resultsthat substantiate the discussion on the role of vocal tractmanipulations in wind instrument performance. The resultsare obtained using a measurement system that allows theanalysis of vocal tract inuence in real time during perfor-mance.

Most previous acoustical studies of vocal-tract inuencehave focused on the measurement of the input impedance of players upstream windways while they simulat e or mimicoral cavity shapes used in playing conditions. 1,36 Thesemeasurements have then been compared with the input im-

pedance of the downstream instrument air column to showinstances where the vocal tract might be able to inuence thereed vibrations. The majority of these investigations are inagreement with regard to the existence of an adjustable up-stream wind-way resonance in the range of 5001500 Hz,which corresponds to the second vocal-tract resonance. 36

On the other hand, musicians do not appear to manipulate the

rst vo cal-tract resonance, which is typically below about300 Hz. 6 In Ref. 6, players reported using a fairly stablevocal-tract shape for most normal playing conditions, withupstream manipulations taking place mainly in the altissimoregister and for special effects. A brief review of previousstudies and a d iscussion of their limitations is provided byFritz and Wolfe. 6

Numerical investigations have also been reported thatcouple an upstream windway model to a complete wind in-strument system. 710 Simplied, single-resonance models of a players windway have been shown to reproduce bugling,pitchbend, multiphonics, and glissando characteristics inclarinet and saxophone m odels, effects generally associatedwith vocal-tract inuence. 8,9

Many performers have used x-ray uoroscopic or endo-scopic approaches to analyze the vocal tract sha pes an d ma-nipulations used while playing their instruments. 2,1115 Thesestudies have reported some general trends in vocal-tract con-gurations for different registers, produced mainly by varia-tions of the tongue position, but have not well addressed theunderlying acoustic principles involved.

A fundamental limitation of previous acoustical investi-gations is related to the fact that the measurements were notconducted in real time. Subjects were required to mime anda Electronic mail: [email protected].

J. Acoust. Soc. Am. 123 4 , April 2008 2008 Acoustical Society of America 23910001-4966/2008/123 4 /2391/10/$23.00


2/10

hold a particular vocal-tract setting for up to 10 s, a task thatis likely difcult and uncomfortable for the player. Despitethe results of Fritz and Wolfe 6 that indicate players haveconsistent muscle or procedural memory, it is not clearthat subjects can accurately reproduce the exact vocal-tractcongurations of interest in this study without playing theinstrument because auditory, and perhaps vibrotactile, feed-back is important in ne-tuning and maintaining an oral cav-ity conguration. Further, these methods only provide datafor the held setting, without indicating the time-varying char-acteristics of vocal-tract manipulations. This last issue re-mains for a recently reported approach that allows the up-stream impedance to be measured while the instrument isplayed. 16

This paper addresses the previously mentioned limita-tions by providing a running analysis of vocal-tract inuenceover time not solely at discrete moments in time . This isachieved using a measurement system embedded in an Ealto saxophone that allows a relative comparison of the up-stream windway and downstream air column impedances un-der normal playing conditions. The same approach couldalso be applied to clarinets, though we chose to concentrate

on saxophones given our own playing experience with them.This paper is organized as follows: Section II describesthe measurement approach, including the system and proce-dures used to evaluate vocal-tract inuence. Section III pre-sents the measured data for playing tasks involving tradi-tional and extended registers, pitch bend, bugling,multiphonics, and timbre variation. Finally, Sec. IV con-cludes with an analysis of the results in the context of saxo-phone performance practice.

II. MEASUREMENT APPROACH

Measurements made for this study were based on conti-

nuity of volume ow at the reed junction,

U u = U d =P u Z u

=P d Z d

Z u Z d

=P uP d

, 1

where u and d subscripts represent upstream and downstreamquantities, respecti vely. This expression is attributed rst toElliot and Bowsher 17 in the context of brass instrument mod-eling. This implies that a relative measure of the upstreamand downstream impedances can be obtained from the en-trance pressures in the players mouth and the instrumentmouthpiece, which is sufcient to indicate when vocal-tractinuence is being exerted. That is, it is assumed that theupstream system can be inuential when values of Z u areclose to or greater than those of Z d and this can be deter-mined from their ratio without knowing the distinct value of each. Because measurements of the mouth and mouthpiecepressures can be acquired while the instrument is beingplayed, this approach allows the player to interact normallywith the instrument. Further, as modern computers can cal-culate and display the pressure spectra in real time, playersand researchers can instantly see how vocal tract changesdirectly affect the upstream impedance.

In this context, the reed is assumed to be primarily con-trolled by the pressure difference across it and thus, a mea-

surement of these two pressures can provide a good indica-tion of how the reed oscillations are inuenced by the twosystems on either side of it. While it may not be possible touse this approach to reconstruct or calculate an upstreamimpedance characteristic fr om measurements of Z d , Pd , andP u as attempted by Wilson 5 , it is valid to indicate instancesof upstream inuence. This argument ignores the possibleeffect of hydrodyna mic f orces on the behavior of the reeddue to unsteady ow. 18 20

A. Measurement system

The measurement approach used in this study requiresthe simultaneous acquisition of the input pressures on theupstream and downstream sides of the instrument reed undernormal playing conditions. Therefore, it was necessary tond a nonintrusive method for measuring pressures in theinstrument mouthpiece and the players mouth near the reedtip. Such measurements are complicated by the fact that thesound pressure levels SPL in a saxophone mou thp iece un-der playing conditions can reach 160 dB or more 21 and thatthe dimensions of the alto saxophone mouthpiece requirerelatively small microphones to allow normal playing condi-tions and to avoid interference with the normal reed/ mouthpiece interaction.

After tests with a prototype system, 22 a special saxo-phone mouthpiece was developed as shown in Fig. 1. AnEndevco 8510B-1 pressure transducer was threaded throughthe top of the mouthpiece 55 mm from the mouthpiece tip toobtain internal pressure values. For pressure measurementsinside the players mouth, a groove was carved in the top of the saxophone mouthpiece for an Endevco 8507C-1 minia-ture pressure transducer of 2.34 mm diameter such that thetransducer tip was positioned 3 mm behind the front edge of

FIG. 1. Saxophone mouthpiece system.

2392 J. Acoust. Soc. Am., Vol. 123, No. 4, April 2008 Scavone et al.

: Vocal-tract inuence in saxophone performance


3/10

the mouthpiece. This position is well within the mouth giventhat the players teeth typically rest anywhere from about 13to 20 mm from the mouthpiece tip. Both Endevco transduc-ers are rated for maximum SPLs between 170 and 180 dBand were found to be unaffected by moisture.

Ideally, both microphones would be located at the frontedge of the mouthpiece inside and outside , which repre-sents the input to both the downstream and upstream aircolumns. Wilson 5 experimented with two mouthpiece trans-ducer locations on a clarinet and selected the more distantposition 39 mm because of noise due to unsteady ownearer the reed tip. The results from a digital waveguidesimulation using a cylinder-cone model 23 suggest that thesignal recorded by the 8510B-1 transducer at a distance of 55 mm from the tip differs from the downstream input im-pedance peak values by less than 2 dB for frequencies up to1500 Hz and less than 3.7 dB for frequencies up to 2000 Hz.Thus, the microphone positioning does not represent a limi-tation, considering that the frequency range analyzed in this

study does not exceed 2000 Hz and that we are mainly in-terested in the use of a vocal-tract resonance that runs fromabout 500 to 1500 Hz.

The pressure transducers were connected to an Endevco136 differential voltage amplier and the signals from therewere routed to a National Instruments NI PCI-4472 dy-namic signal acquisition board. The acquisition card samplerate was set to 12 000 Hz. An NI LABVIEW interface wasdesigned to allow real-time display of the spectra of the twopressure signals. The Endevco transducers were calibratedrelative to one another prior to the experiment, as describedin the Appendix.

To help distinguish between vocal tract and embouchurechanges, a small circular 12.7 mm diameter force sensingresistor FSR made by Interlink Electronics was placed un-der the cushion on the top of the saxophone mouthpiece toobtain a relative measure of the upper teeth force see Fig.1 . The time-varying sensor voltages were input to the signalacquisition board for storage and to provide a running dis-play of embouchure movement in the LABVIEW interface.Although this setup provided no data on movements of thelower lip, normal embouchure adjustments involve a simul-

FIG. 2. Color online Spectrograms of the SPL ratio between the mouth-piece and the mouth pressures when playing a scale Subjects A and D, fromtop to bottom .

5 10 15 2040

30

20

10

0

10

20

Scale Note Index

F i r s t P

a r t i a l S P L R a t i o s ( d B )

Subject ASubject BSubject CSubject D

0 5 10 15 2040

30

20

10

0

10

20

Scale Note Index

F i r s t P a r t i a l S P L R a t i o s ( d B )

FIG. 3. Average SPL ratios for rst partials of scale: individual by subjecttop and for all subjects bottom .

J. Acoust. Soc. Am., Vol. 123, No. 4, April 2008 Scavone et al.

: Vocal-tract inuence in saxophone performance 2393


4/10

taneous variation of both the upper teeth and lower lip. Theprimary purpose of this sensor was to provide players withvisual feedback to help them focus on maintaining a xedembouchure setting, rather than acquiring an absolute mea-sure of lip or teeth force.

Playing tests were conducted in an IAC IndustrialAcoustics Company double-walled sound isolation booth tominimize external sound interference. In addition to the

mouthpiece described previously, a single Vandoren #3 me-dium hardness reed and Selmer Super Action Series II altosaxophone serial number 438024 were used for the entireexperiment.

B. Player tests

Upstream inuence was evaluated using the measure-ment system through a series of playing tests with a group of four professional saxophonists. Subject C was the rst authorof this paper. The subjects lled out a questionnaire abouttheir saxophone background and experience and were given

about 5 min to become accustomed to the mouthpiece andsaxophone setup. They were allowed to practice the re-

quested tasks before the recording began. After the subjectswere comfortable with all the tasks, data storage was initi-ated and each task was performed in sequential order. Whensubjects had difculty with a given task, they were allowedto repeat it. A large real-time display of the FSR reading wasprovided and the subjects were told to avoid making embou-chure changes while performing the tasks.

III. RESULTS

To examine variations of upstream inuence over time,many of the results in the subsequent sections are displayedvia spectrograms of the ratio of the SPL in decibels or

power spectral densities of the mouth and mouthpiece pres-sures. To reduce artifacts arising from the ratio calculation,spectral bins containing no signicant energy in both theupstream and downstream power spectra were masked. Themeasured upper teeth force is also plotted to indicate therelative steadiness of a players embouchure setting.

A. Traditional and extended registers

The traditional range of a saxophone is from writtenB 3F 6, which on an alto saxophone corresponds to thefrequency range 138.6 880 Hz. Advanced players can play afurther octave or more using cross-ngerings, a range re-ferred to as the altissimo or extended register. In order toevaluate general player trends over the full range of the in-strument, each subject was asked to play an ascending legatowritten B D concert pitch scale from the lowest note onthe instrument to the highest note that could comfortably beheld which was typically near a written C7 in the extendedregister , playing each note for about 0.5 s.

Representative spectrograms of the SPL ratios of up-stream and downstream pressures for this task are shown inFig. 2, as played by Subjects A and D. Breaks in the mea-sured upper teeth force indicate instances where subjectsstopped to breath. SPL averages computed at the fundamen-

FIG. 4. Color online Spectrograms of the SPL ratio between the mouth-piece and the mouth pressures when playing a written D6 without vibrato,with a slow lip vibrato, and bending via vocal tract manipulations SubjectsB and D, from top to bottom .

500 550 600 650 700 750 800 850 90060

70

80

90

100

110

120

130

140

150

Frequency (Hz)

S P L d B ( p

r e f

= 2 0

P a

)

Mouth PressureMouthpiece Pressure

FIG. 5. Overlaid snapshots of the pitch bend spectra on an alto saxophone atthe starting frequency 700 Hz and at the lowest frequency trajectory580 Hz for Subject D.




5/10

tal frequencies for each scale note are shown in Fig. 3 foreach subject and averaged across all subjects. For most notes

in the traditional range, the SPL ratios at the fundamentalfrequencies are below 20 dB. In other words, pressures inthe mouthpiece for these notes are typically 10 times greaterthan those in the players mouths. Based on the assumptionsoutlined in Sec. II, this indicates a similar ratio of inputimpedance peak levels at the fundamental playing frequen-cies on either side of the reed and thus minimal upstreaminuence for notes in this range. These ratios display a localminima centered at the thirteenth note of the scale 466 Hz ,which may be related to the fact that notes in this range arerelatively easy to play. The standard deviation of the SPLratios are shown by the error bars in the lower plot of Fig. 3.

A fairly abrupt change in SPL ratios is evident whensubjects prepare to enter the extended register, a result thatwas also reported by Fritz and Wolfe. 6 Scale note index 19 inFig. 3 is the highest note written F6 in the scale that fallswithin the traditional register. An alternate ngering existsfor F6 on the saxophone that has a playing behavior morelike extended register notes it is based on the use of a thirdair column partial , though the ngering used by subjects inthis study was not specied or recorded. Averaged SPL ratiosacross all subjects for the extended register notes are be-tween 3 and 5 dB, though the standard deviation is signi-cant. In particular, Subjects A and B show large variations inSPL ratios from note to note in this range, whereas the re-sults for Subjects C and D are more consistent. It should benoted that the task called for a legato scale, or slurring fromnote to note. Future studies could investigate potential varia-tions of SPL ratios in the extended register when the notesare attacked individually, with or without breaks betweeneach.

It has been suggested by Wilson 5 that performers mighttune upstream resonances with higher harmonics of a playednote. There are some instances in Fig. 2 and the data for theother subjects where the ratios for upper partials of notes arenear 0 dB, typically in the range 6001600 Hz, though thisvaries signicantly among the subjects. However, no system-

atic note-to-note tuning with an upper harmonic was found.It is unlikely that upstream tuning would help stabilize a noteunless the fundamental is weak and only one or two higherpartials exist below the cutoff frequency of the instrument.That said, variations of upper partial ratios could affect thetimbre of the instrument and this is investigated further inSec. III E.

B. Pitch pendingSaxophonists make frequent use of pitch bends in their

playing, especially in jazz contexts. There are several wayssuch frequency modications can be achieved, including lippressure and tonehole key height changes, as well as vocaltract manipulations. Lip pressure variations, which are usedto produce vibrato, affect the average reed tip opening andcan yield maximum frequency modulations of about half asemitone. The use of vocal tract manipulations for pitch bendcan achieve downward frequency shifts of a musical third ormore. Signicant upward frequency shifts using lip pressure

0 500 1000 1500 20000

1

2

3

4

5

6

7x 10

7

Frequency (Hz)

I m p e d a n c e

M a g n i t u d e ( P a s m

3 )

Register Key ClosedRegister Key Open

FIG. 6. Input impedance of the alto saxophone for the D6 ngering.

FIG. 7. Color online Spectrograms of the SPL ratio between the mouthand mouth piece pressures when performing the bugling task Subjects Band C, from top to bottom .




6/10

or the vocal tract are not possible. On an alto saxophone,bends produced via vocal-tract manipulations are easiestwhen starting on notes above concert E 5. Bends that startabove a concert C5 tend to have a lower limit around the C5frequency 523 Hz . In other words, the ngered note con-trols the starting and highest frequency of the bend rangebut the minimum frequency is generally always between 500and 600 Hz.

Pitch bending has previ ous ly been investigated with re-spect to vocal tract inuence. 5,6 In the present study, subjectswere asked to nger a written high D6 698 Hz with the rstpalm key and to play the note normally, without vibrato, forabout 3 s. They were then asked to play the note with a slowvibrato about 1 Hz , modulating the pitch up and down asmuch as possible using lip pressure only . Finally, subjectswere instructed to perform, without varying their lip pres-sure , a slow downward pitch bend to the lowest note theycould comfortably maintain, hold that note for about 1 s, andthen bend the note back to the starting D pitch.

Figure 4 shows representative spectrograms of the SPLratios of upstream and downstream pressures for Subjects Band D when performing the requested tasks. The average

SPL ratio at the fundamental frequency, across all subjects,when played normally without vibrato was 25 dB, with astandard deviation of 9.3. Subjects B, C, and D demonstrateda consistent vibrato frequency range, using lip pressurevariations only, of about 677700 Hz or 54 cents down andabout 5 cents up . Subject As range for the same task was655690 Hz, but his SPL fundamental ratios were as high as3.4 dB at the vibrato frequency dips, indicating that he usedsome vocal-tract manipulation as well. For the pitch bendperformed without lip pressure variation, the lower fre-quency limits of the four subjects were 583, 597, 558, and570 Hz, respectively, with corresponding fundamental fre-

quency SPL ratios of 29.1, 22.7, 30.7, and 25.1 dB. The subjects averaged a frequency drop of 330 cents with a SPL ratioof 26.9 and a standard deviation of 3.7. Snapshots of theupstream and downstream spectra at the starting frequencyand lowest frequency trajectory of the pitch bend task forSubject D are overlaid in Fig. 5. At the starting frequency of about 700 Hz, the upstream and downstream SPLs differ byabout 35 dB, while at the bottom of the bend, they differ byabout 25 dB.

The input impedance of the alto saxophone used for thisstudy with a D6 ngering is shown in Fig. 6, from which it isclear that the instrument has no strong resonance between400 and 650 Hz. This measurement was obtained using atwo-microphone transfer function technique. 24,25 Given thexed ngering and relatively constant lip pressure used bythe subjects, as well as the fact that lip pressure variationsalone can only produce bends of about 54 cents, the pitchbends of 300 cents and more can only be the result of vocal-tract manipulation. The high SPL ratio during the bend indi-cates that a signicantly stronger resonance exists in theplayers mouths during the pitch bend than in the down-stream air column. The implication is that performers cancreate a resonance in the upstream windway in the range700550 Hz for this task that is strong enough to overridethe downstream air column and assume control of the reedvibrations.

The upstream resonance frequency is mainly controlledvia tongue position variations. The pitch bend is only pos-sible for notes higher in the traditional range, where the aircolumn resonance structure is relatively weak. We speculatethat the lower frequency limit on the pitch bend is related tothe vocal-tract physiology and players control of the secondupstream resonance within an approximate range of abou t5201500 Hz, which is close to that reported by Benade. 3

This is corroborated by the fact that a similar lower fre-quency limit is found when pitch bending on both sopranoand tenor saxophones. Informal tests on a B clarinet indi-

0 2 4 6 830

20

10

0

10

20

30

Overtone Series Note Index

S P L R a t i o s ( d B )

Subject ASubject BSubject CSubject D

0 1 2 3 4 5 6 7 8 930

20

10

0

10

20

30

Overtone Series Note Index

S P L R a t i o s ( d

B )

FIG. 8. Average SPL ratios at each overtone series frequency: individual bysubject top and for all subjects bottom .

FIG. 9. Color online Fingerings and approximate sounding notes writtenand frequencies for four requested multiphonics.




7/10

cate a greater variance in the lower pitch bend range withdifferent ngerings, though the overall range still falls withinthe limits mentioned earlier.

C. Bugling

Bugling involves the articulation of the notes of an over-tone series while maintaining a xed low note ngering. Be-ginning students can normally produce the rst and second

overtones, sometimes inadvertently while trying to play thefundamental. It typically takes many years of practice to de-velop the exibility that allows one to cleanly attack thehigher overtones. In this task, subjects were asked to nger awritten B 3 all holes closed and to play an overtone se-ries, attacking each note individually.

Representative spectrograms of the SPL ratios for thebugling exercise are shown in Fig. 7, as played by Subjects Band C. Individual subject averages, as well as averagesacross all subjects, computed at the fundamental frequenciesof each overtone are shown in Fig. 8. Subject D did not playovertones 89 and Subject A was unable to play overtones24 and 9. Vocal tract inuence is not clearly evident untilthe third overtone, where downstream instrument resonancesbegin to weaken. In general, it is difcult to play the thirdovertone without some vocal tract manipulation and it ap-pears that a ratio of about 7 dB is sufcient to allow this tohappen. It also may be possible that, when playing the thirdovertone, players reinforce the 1130 Hz component with anupstream resonance rather than the component at 563 Hz,which falls at the lower end of the adjustable upstream reso-nance range. Averaging across Subjects B, C, and D, the SPLratio of the 1130 Hz component when playing the fundamen-tal of the overtone series is 8.7 dB. The ratio at this samefrequency when playing the third overtone is 4.8 dB.

D. Multiphonics

Multiphonics involve oscillations of the reed based ontwo inharmonic downstream air column re sonance frequen-cies and their intermodulation components. 26 Two relativelyeasy and two more difcult multiphonics were chosen foranalysis with respect to potential vocal-tract inuence. Thefour multiphonic ngerings and their approximate soundingnotes are illustrated in Fig. 9.

Figure 10 shows the SPL ratios for four different multi-phonics, as played by Subjects B and D. The sounds pro-duced by the two subjects differed signicantly. From theplots, there are more spectral components evident in thesounds of Subject D. We expect that a player can inuencethe sound quality of a multiphonic by aligning an upstreamresonance with one or more components. For example, bothsubjects appear to reinforce a group of partials in the range8751075 Hz for the last multiphonic. SPL ratio levels forSubject D were, across all tasks, generally higher than thoseof Subject B. This suggests that Subject D makes use of stronger upstream resonances, which likely explains both thehigher ratios for this task, as well as the greater spectraldensity of the multiphonic sounds.

E. Timbre variations

In a questionnaire, all subjects indicated that theythought they could inuence the sound of the saxophone viavocal-tract variations. They reported using upstream inu-ence to control and adjust tone color timbre , pitch, andextended register notes. To investigate timbre variations viavocal-tract manipulation, subjects were asked to play asteady written F4 207.7 Hz and to move their tongue peri-odically toward and away from the reed while maintaining aconstant embouchure setting.

Figure 11 shows spectrograms of the SPL ratios result-ing from this task, as played by Subjects A and C. Partialsfour and higher show clear variations in SPL ratios withtongue motion. Snapshots of the mouth and mouthpiecespectra for Subject C are shown in Fig. 12 at times of 4 and5 seconds. Although both the upstream and downstreamspectra varied with tongue position, the upstream changeswere most signicant. For example, the downstream SPL of the fourth partial only changes by +3 dB between the two

FIG. 10. Color online Spectrograms of the SPL ratio between the mouthand mouthpiece pressures when playing four multiphonics Subjects B andD, from top to bottom .




8/10

times, whereas the upstream SPL increases by +13 dB. Forthis partial and subject, the SPL ratio goes from an averageof 0.7 to 9.9 dB, or a difference of 10.6 dB between thetwo tongue positions. The change of downstream and up-stream SPL for two tongue position extremes is collated forpartials 19 and Subjects A-C in Fig. 13 . Subject Ds resultsshowed no clear variation of SPL ratios over time and it islikely this person did not properly understand the requestedtask. In general, tongue movements toward the reed tendedto boost upstream frequency components from about 800 Hzto at least 2000 Hz. Note that these spectral changes occursimultaneously over a wide frequency range and thus arelikely the result of a more wide bandwidth upstream reso-nance. Below 800 Hz, the effect of the tongue positionmovement varied signicantly among the subjects. It is notpossible to say whether these variations were the result of differences in physiology or tongue positions.

IV. DISCUSSION

Results of the tongue movement task indicate that vocal-tract manipulations can be used throughout the playing range

of the saxophone to produce subtle timbre variations involv-ing frequency components from at least 800 2000 Hz.These timbre variations simultaneously affect partials over awide frequency range, which implies the use of a relativelywide bandwidth upstream resonance. The pitch bend, ex-tended register, and bugling tasks, however, indicate thatvocal-tract inuence signicant enough to override down-stream air column control of reed vibrations is only possiblewhen the downstream system provides weak support of agiven note. This is normally the case for notes with funda-mental frequencies an octave below the downstream air col-umn cuto ff frequency around 1500 Hz for the altosaxophone 27 and higher. Thus, signicant vocal-tract inu-ence can be exerted for notes near the top of the alto saxo-phones conventional range and on into the extended or al-tissimo register. This type of inuence makes use of anarrow upstream resonance that the player manipulates tocontrol the fundamental vibrating frequency of the reed.

Various special ngerings are used when playing notesin the saxophones extended register but these provide onlyweak downstream support of a note. When students unaccus-tomed with altissimo register playing try these ngerings,they produce a weak tone lower in the traditional range of

FIG. 11. Color online Spectrograms of the SPL ratio between the mouthand mouthypiece pressures when varying tongue position while holding anF3 Subjects A and C, from top to bottom .

0 500 1000 1500 200060

70

80

90

100

110

120

130

140

150

Frequency (Hz)

S P L d B ( p

r e f

= 2 0

P a

)

Time = 4 secondsTime = 5 seconds

0 500 1000 1500 200060

70

80

90

100

110

120

130

140

150

Frequency (Hz)

S P L d B ( p

r e f =

2 0

P a

)

Time = 4 secondsTime = 5 seconds

FIG. 12. Snapshots of the mouthpiece top and mouth bottom spectra forthe tongue variation task at times of 4 and 5 s as played by Subject C.




9/10

the instrument. From the discussion above, this implies theneed for vocal-tract inuence when playing in the extendedregister. There were a few instances when the SPL ratios forSubjects A and B fell below 0 dB while playing in this rangesee Figs. 3 and 8 . This might happen because a given n-

gering produced a relatively strong downstream resonance atthe fundamental playing frequency or it is possible thesesubjects made embouchure adjustments that were not de-tected by the FSR on the top of the mouthpiece. The subjectswere instructed to avoid making embouchure changes for alltasks in this study. However, it is common for performers tomove their lower lip forward, away from the reed tip, inpreparation for playing in the saxophones altissimo register.The authors assume this reduces the reed damping, with acorresponding increase in the reed resonance frequency, andthat this may help to maintain normal operation of the reed inits stiffness controlled region. As well, extended registerplaying is generally easier with stiffer reeds. Notes near thetop of the extended register, with fundamentals in the range14002000 Hz, may be beyond the range of a high- Q sec-ond vocal-tract resonance. It is not clear whether playersstabilize these extreme high notes, which are difcult to pro-

duce consistently even for professionals, with a vocal-tractresonance and/or by manipulating the reed resonance fre-quency.

One of the clearest examples of the use of vocal-tractinuence involves pitch bending, with downward frequencymodulations 56 times greater than that possible via lip pres-sure variations alone. Pitch bends are not possible for noteslower than about a concert C5 523 Hz , at the lower ex-treme of the main adjustable upstream resonance and whereseveral downstream air column resonances can help stabilizea note.

The results of this study are less conclusive with respectto the production of multiphonics. Performers who are un-able to play in the extended register of the saxophone nor-mally also have trouble producing the more difcult multi-phonics. This would imply the use of an upstream resonanceto support one or more intermodulation components. A futurestudy should compare successful and unsuccessful multi-phonic attempts for a xed ngerings in an effort to verifythis behavior. There were considerable differences in thesound and quality of the multiphonics produced by the subjects of this study and this is l ikely related to variations in the

upstream system.Finally, there is a common misconception among somescientists and players that vocal-tract inuence is exerted ona nearly continuous basis w hile playing a single-reed instru-ment in its tradi tional range .5,13 For example, an early studyby Clinch et al. 13 concluded that vocal tract resonant fre-quencies must match the frequency of the required notes i nclarinet and saxophone performance. Likewise, Wilson 5

claims that for most tones in analyzed melodic phrases, theperformers airways were tuned to the rst harmonic or tothe second harmonic, or there was a resonance aligned withboth the rst and second harmonics. A related conclusion ismade by Thompson 28 with respect to variations of the reed

resonance. These suggestions have no basis in performancepractice. Although it is true that effects such as pitch slidescommon to jazz playing may involve some level of vocal-tract inuence, most professional musicians use and advo-cate a relatively xed vocal-tract shape during normal play-ing. This behavior is corroborated by Fritz and Wolfe. 6 Itseems likely that players vary their embouchure and perhapstheir vocal-tract shape gradually when moving from low tohigh register notes, but this should be understood to varywith register and not on a note-by-note basis until one getsto the altissimo register . The results of this study indicatethat vocal-tract inuence for normal playing within thetraditional range of the alto saxophone is primarily limited totimbre modication because most of these notes are wellsupported by the downstream air column.

A website and video demonstrating the me asurementsystem described in this paper is available online. 29

ACKNOWLEDGMENTS

The authors thank the anonymous reviewers for theirinsightful comments and suggestions. They also acknowl-edge the support of the Natural Sciences and EngineeringResearch Council of Canada, the Canadian Foundation for

1 2 3 4 5 6 7 8 920

15

10

5

0

5

10

15

20

25

30

Partial Number

S P L D i f f e r e n c e s ( d B )

Subject ASubject BSubject C

1 2 3 4 5 6 7 8 920

15

10

5

0

510

15

20

25

30

Partial Number

S P L D i f f e r e n c e

s ( d B )

Subject ASubject BSubject C

FIG. 13. SPL changes in mouthpiece top and mouth bottom between twotongue position extremes for partials 19 of held F3 Subjects A, B, and C .




10/10

Innovation, and the Centre for Interdisciplinary Research inMusic Media and Technology at McGill University. The doc-toral research of the second and third authors is supported bythe Fonds Qubcois de la Recherche sur la Nature et lesTechnologies and CAPES Funding Council of the BrazilianMinistry of Education , respectively. Finally, we would liketo thank Bertrand Scherrer for his help in developing the rstprototype of the National Instruments LABVIEW measurementsystem.

APPENDIX: TRANSDUCER CALIBRATION

A JBL 2426H compression driver was connected to asteel cylindrical pipe of 150 mm length and 25.4 mm diam-eter. The calibration is valid up to the cutoff frequency of therst higher order mode, which is f c =1.84 c / 2 a 8 k Hzfor a radius a =12.7 mm and a speed of sound c =347 m / s.30

At the far end of this pipe, the two transducers were tthrough a cap such that they extended a few millimeters be-yond the cap surface. A 60-s noise sequence was playedthrough the system and power spectral densities were deter-mined using a modied averaged periodogram of 1024 datapoints and 50% overlap Hanning windows at a samplingrate of 48 kHz. In order to match fast Fourier transform binsused in the LABVIEW interface and the subsequent data analy-sis, the calibration data was downsampled to 12 kHz. Thepower spectral densities were computed with the pwelchand cpsd functions in MATLAB . The transfer function relat-ing the gain and phase differences between the microphoneswas obtained as: 31

H 12 =S 22 S 11 + S 11 S 22 2 + 4 S 12 2

2S 21,

where

S 11 = PWELCH P1 ,HANNING N ,N / 2 , N,FS ,

S 22 = PWELCH P2 ,HANNING N ,N / 2 , N,FS ,

S 12 = CPSD P1 , P2 ,HANNING N , N/ 2 ,N, FS ,

and

S 21 = CPSD P2 , P1 ,HANNING N , N/ 2 ,N, FS .

A similar calibra tion technique was previously reportedby Seybert and Ross. 24

1J. Backus, The effect of the players vocal tract on woodwind instrument

tone, J. Acoust. Soc. Am. 78 , 1720 1985 .2M. Watkins, The saxophonists vocal tract. 1, Saxophone Symp. 27 ,5178 2002 .

3A. H. Benade, Air column, reed, and players windway interaction inmusical instruments, in Vocal Fold Physiology, Biomechanics, Acoustics,and Phonatory Control , edited by I. R. Titze and R. C. Scherer DenverCenter for the Performing Arts, Denver, CO, 1985 , Chap. 35, pp. 425452.

4P. L. Hoekje, Intercomponent energy exchange and upstream/downstreamsymmetry in nonlinear self-sustained oscillations of reed instruments,Ph.D. thesis, Case Western Reserve University, Cleveland, OH, 1986.

5T. D. Wilson, The measured upstream impedance for clarinet perfor-mance and its role in sound production, Ph.D. thesis, University of Wash-

ington, Seattle, WA, 1996.6C. Fritz and J. Wolfe, How do clarinet players adjust the resonances of their vocal tracts for different playing effects?, J. Acoust. Soc. Am. 118 ,33063315 2005 .

7S. D. Sommerfeldt and W. J. Strong, Simulation of a player-clarinetsystem, J. Acoust. Soc. Am. 83 , 19081918 1988 .

8R. Johnston, P. G. Clinch, and G. J. Troup, The role of the vocal tractresonance in clarinet playing, Acoust. Aust. 14 , 6769 1986 .

9G. P. Scavone, Modeling vocal-tract inuence in reed wind instruments,in Proceedings of the 2003 Stockholm Musical Acoustics Conference ,Stockholm, Sweden, pp. 291294.

10P. Guillemain, Some roles of the vocal tract in clarinet breath attacks:

Natural sounds analysis and model-based synthesis, J. Acoust. Soc. Am.121 , 23962406 2007 .11R. L. Wheeler, Tongue registration and articulation for single and double

reed instruments, Natl. Assoc. College Wind Percussion Instruct. J. 22 ,312 1973 .

12W. E. J. Carr, A videouorographic investigation of tongue and throatpositions in playing ute, oboe, clarinet, bassoon, and saxophone, Ph.D.thesis, University of Southern California, Los Angeles, CA, 1978.

13P. G. Clinch, G. J. Troup, and L. Harris, The importance of vocal tractresonance in clarinet and saxophone performance: A preliminary account,Acustica 50 , 280284 1982 .

14J. T. Peters, An exploratory study of laryngeal movements during perfor-mance on alto saxophone, Masters thesis, North Texas State University,Denton, TX, 1984.

15M. Patnode, A ber-optic study comparing perceived and actual tonguepositions of saxophonists successfully producing tones in the altissimo

register, Ph.D. thesis, Arizona State University, 1999.16J.-M. Chen, J. Smith, and J. Wolfe, Vocal tract interactions in saxophoneperformance, in Proceedings of the International Symposium on Musical Acoustics , Barcelona, Spain, 2007.

17S. Elliott and J. Bowsher, Regeneration in brass wind instruments, J.Sound Vib. 83 , 181217 1982 .

18A. R. da Silva, G. P. Scavone, and M. van Walstijn, Numerical simula-tions of uid-structure interactions in single-reed mouthpieces, J. Acoust.Soc. Am. 122 , 17981809 2007 .

19J. van Zon, A. Hirschberg, J. Gilbert, and A. Wijnands, Flow through thereed channel of a single reed instrument, J. Phys. Paris , Colloq. 54 , C2821824 1990 .

20A. Hirschberg, R. W. A. van de Laar, J. P. Marrou-Maurires, A. P. J.Wijnands, H. J. Dane, S. G. Kruijswijk, and A. J. M. Houtsma, A quasi-stationary model of air ow in the reed channel of single-reed woodwindinstruments, Acustica 70 , 146154 1990 .

21X. Boutillon and V. Gibiat, Evaluation of the acoustical stiffness of saxo-phone reeds under playing conditions by using the reactive power ap-proach, J. Acoust. Soc. Am. 100 , 11781889 1996 .

22G. P. Scavone, Real-time measurement/viewing of vocal-tract inuenceduring wind instrument performance A , J. Acoust. Soc. Am. 119 , 33822006 .

23G. P. Scavone, Time-domain synthesis of conical bore instrumentsounds, in Proceedings of the 2002 International Computer Music Con- ference , Gteborg, Sweden, pp. 915.

24A. Seybert and D. Ross, Experimental determination of acoustic proper-ties using a two-microphone random excitation technique, J. Acoust. Soc.Am. 61 , 13621370 1977 .

25A. Lefebvre, G. P. Scavone, J. Abel, and A. Buckiewicz-Smith, A com-parison of impedance measurements using one and two microphones, inProceedings of the International Symposium on Musical Acoustics , Barce-lona, Spain, 2007.

26J. Backus, Multiphonic tones in the woodwind instrument, J. Acoust.Soc. Am. 63 , 591599 1978 .

27A. H. Benade and S. J. Lutgen, The saxophone spectrum, J. Acoust. Soc.Am. 83 , 19001907 1988 .

28S. C. Thompson, The effect of the reed resonance on woodwind toneproduction, J. Acoust. Soc. Am. 66 , 12991307 1979 .

29G. P. Scavone, http://www.music.mcgill.ca/~gary/vti/, 2007 last viewed11 January 2008 .

30N. H. Fletcher and T. D. Rossing, The Physics of Musical InstrumentsSpringer, New York, 1991 .

31J. H. P. R. White and M. H. Tan, Analysis of the maximum likelihood,total least squares and principal component approaches for frequency re-sponse function estimation, J. Sound Vib. 290 , 676689 2006 .



Scavone Vocal Tract

Documents

Transcript of Scavone Vocal Tract