Transient behaviour in the motion o the brass player s lips John Chick, Seona Bromage, Murray Campbell The University o Edinburgh, The King s Buildings, Mayield Road, Edinburgh EH9 3JZ, UK, john.chick@ed.ac.uk In judging the quality o a brass wind musical instrument, a player evaluates a number o dierent properties o the instrument. One important indicator o quality is the ease with which dierent notes on the instrument can be made to sound. At the start o each note there is a transient stage during which the sel-sustaining oscillation o the coupled system o lips and air column builds towards a ully developed steady-state regime. Use o a high speed digital video camera with a transparent mouthpiece allows visual inspection o the detailed motion o the player s lips. Transient behaviour in the physical motion o the lips has been studied, and related to the psychoacoustically important transients in synchronised recordings o the sound. 1 Introduction The starting transient o a musical instrument is widely acknowledged as being one o the main parameters used by players and listeners to characterise the timbre o a particular instrument or player. In some cases it can be diicult or a listener to identiy the instrument being played without irst hearing the start o the note [1, ]. In the work presented here, we study the start o a note produced on a brass instrument as it is generated at the lips o the brass player. We propose an experimental procedure which enables us to measure the motion o the lips accurately at the start o the note in order to study the transient behaviour. The experimental set up involves high speed photography o the lips as the note is started, together with sound pressure levels measured in the mouthpiece. This is described more ully in Section 3 together with details o instrumentation and data acquisition. Analysis o the data or both movement o the lips and the acoustic emissions produced provide a better understanding o the starting transient and the mechanics o producing a note. The results o the experiments are presented in Section 4. Some related work is being carried out to investigate the sound emission created using artiicial lips. These studies have generally ocused on steady state behaviour [3, 4]. When applied to starting transients this has the potential to provide much greater consistency and repeatability o experimental results than experiments conducted on actual players. However, in the irst instance it is important to investigate some o the characteristics o transients produced by actual players and to identiy the key parameters which aect this transient. Section concludes and describes possible directions or uture work inspired by these experiments. Background Despite the acknowledgement o the importance o the starting transient, to the authors knowledge there has been relatively little work speciically aimed at investigating the beginning o a note as played by a brass player. Luce and Clark [8] presented a study o dierent transient attacks rom a range o dierent orchestral brass instruments. A time dependent ourier transorm was used to analyse amplitudes and relative phase o the irst 11 partials. They ound the duration o the attack transient, which they described as the period in which requency and amplitude changes take place, lasted or about ms, with an average deviation o about ms depending on requency and instrument. However, they also ound that the requency modulation at the start o the note lasted or considerably longer than this, typically 1ms. For producing a clean attack on a brass instrument, the importance o establishing a cooperative regime between lips and relected sound rom the bell is discussed by Benade [, 6]. To initiate a tone, the lips buzz at or near the desired requency and need to sustain this requency or the time it takes or the sound to travel rom the player s lips to the bell o the instrument, where some o the sound is relected back. I the requency and phase are correct, constructive reinorcement is initiated between the resonating air column and the lips. The time taken or the sound to become ully developed depends predominantly on the length o the horn, and the number o oscillations is dependent on the requency o the note played. In their investigation into the psychoacoustic parameters aecting the design o mouthpieces or rench horns, Plitnik and Lawson [7] cite attack clarity as one o their seven key psychoacoustic variables used to assess mouthpiece quality. 73
3 Experimental Procedure In order to analyse the motion o the brass player s lips, an experiment was designed to synchronise audio recordings with high speed camera ootage. A schematic diagram and photograph o the experimental process is shown in Figure 3. Light Source Transparent Mouthpiece enable video recording o lip movement. Although the mechanism or sound production is the same, the embouchure o a rench horn player is normally a little dierent rom that o other orchestral brass instruments with only a little o the lower lip covered by the rim o the mouthpiece. The window at the ront o the mouthpiece is angled allow or better visualisation o the lip opening area. A schematic diagram and photograph o the mouthpiece are shown in Figure 3.1. Ø 8mm High Speed Camera Brass Instrument Ø 18mm Probe Microphone PC 16mm Audio Trigger rom Timing Box Figure : Schematic and photograph o the experimental mouthpiece Figure 1: Schematic and photograph o the experimental equipment used or measuring the motion o a horn player s lips 3.1 The instrument The horn used or these experiments was a medium bore Paxman model 4, B -alto double descant horn. This instrument allowed or tests to be carried out using tube lengths o approximately 1.8m and 3.6m. The mouthpiece design is similar to that used by [4]. The cup o the mouthpiece is transparent and machined rom perspex, with a thin optical glass window at the ront to The mouthpiece is a key component in determining the playing characteristics o any brass instrument[, 7]. The main design parameters are the cup volume, the throat and back-bore, and the rim dimensions. For the experimental mouthpiece, the shank (including throat and back bore) has been taken rom a Paxman 4C mouthpiece, and the cup volume and rim dimensions are also based on this mouthpiece. The experiments have been carried out using two dierent lengths o tubes on the same horn. Both tube lengths were used to record notes F, F, F, and C. The experiments were repeated with the same equipment with two dierent players in order to identiy common patterns in the motion o the lips. 3. Data acquisition To measure the mouthpiece pressure a Brüel and Kjær 1/ microphone was used with a 3mm long mm diameter probe. The entrance section o the probe was packed with metallic damping wool to reduce the pres- 74
sure levels to within the range o the diaphragm sensitivity, but there is still some clipping evident at high amplitudes. The images were captured using a Vision Research, Inc. Phantom v4.1 camera. With the image resolution chosen or this study (6 by 64 pixels) and the necessity or a reasonably long capture time (3.6 secs so as to capture the transient) the rame rate available was rames per second. Filming requires the use o a strong light source; a Schott KL1 LCD swan-neck lamp was used. This had the beneit o being a cool source so was suitable or use within close proximity to the musicians. The maximum exposure time o 18 secs was used; however it was still necessary to use a relatively large aperture on the camera lens. This had to be balanced with the need or a reasonable depth o ocus to optimise image quality, and to avoid lips moving in and out o ocus as they move in the direction perpendicular to the plane o the image. Even with the angled mouthpiece viewing window there was diiculty in visualising part or all o the opening between the lips due to overhang o the top lip. This was particularly a problem with high notes, where the amplitude o motion is small. This diiculty could partly be overcome by adjusting the camera angle, but in practice this proved diicult. As this study was interested in the changing motion o the lips and not the exact dimensions o the open area, the problem has not aected the results except in limiting the range o the notes which could be studied, with C being the highest that it was practical to ilm. The camera was triggered using an external Berkeley Nucleonics Corp Model pulse generator, the signal rom which was recorded simultaneously with the audio signal rom the probe microphone using the Brüel and Kjær PULSE system. This then allowed the identiication o the point in the audio time axis which corresponded to the irst captured image, and hence the synchronisation o the two data sets. shited along the y-axis in order to allow comparison o the two synchronised data sets (open area and mouthpiece pressure). The units o pressure are arbitrary. Figure 3: One image o lip opening and the corresponding thresholded image. 4 Results The high speed digital video records showed signiicant variation between nominally similar measurements. Nonetheless, some interesting patterns can be identiied in the transient behaviour o the start o the note. Results rom only one player are presented, though similar behaviour was observed or both. 4.1 Input impulse response measurements One obvious timescale over which to consider the transient behaviour is the time or the initial pressure disturbance to travel along the instrument, be relected at the open end and return to the lips. It is only ater this time that the lips receive inormation about the instrument resonance to which their behaviour must be coupled [, 6]. 3.3 Analysis procedure The high speed digital ilms were edited to produce a series o images which show the transient, the rame numbers corresponding to part o the time axis o the audio data. The individual rames were also cropped to show just the lip opening, in order to reduce the data size or analysis. A threshold grey level was then chosen to pick out the open area, as demonstrated in Figure 3. These pixels were then counted or each image and the open area plotted against time. We have chosen to look at just the area o opening, though the same analysis process can also provide plots o variations in width and height. The synchronised mouthpiece pressure data have been added to the plots o open area, scaled and in some cases Figure 4: Input impulse response curve or the 3.6m horn showing a round trip time o approximately ms This round trip time can be obtained rom input impulse response measurements o the instrument. Figure 4 was 7
obtained using apparatus in the School o Physics developed by Kemp [1]. The round trip time given or the low F horn (3.6m) is approximately ms. A similar curve or the F-alto horn (1.8m) gives a round trip time o approximately 11ms. It should be noted that the round trip time is determined by the group velocity and so is not necessarily directly related to the periodicity o any o the modes o vibration o the air column. 4. Amplitude Figure 6 shows a series o plots o open area and synchronised mouthpiece pressure or the attack transients o our notes played on the 3.6m horn. It can be seen that the amplitudes o the mouthpiece pressure oscillations remain almost constant, whilst there is a gradual reduction in the mean pressure over approximately the irst ms o the note. Ater this time there is an obvious increase in both the mean pressure and the amplitude o oscillation. Evidence is also present o a change in the rate o increase o the open area at the same time; this can be observed qualitatively when viewing the corresponding series o images (see Figure ). 1 1.16 1.17 1.18 1.19 Figure : A series o lips images (rotated by 9 ) showing the maximum opening or each o the irst 14 cycles o the note F played on the 3.6m horn, and the corresponding graph o open area and mouthpiece pressure. Figure 7 shows the equivalent series o plots o open area and synchronised mouthpiece pressure or the attack transients o the same our notes played on the 1.8m horn. Similar behaviour is observed over a shorter timescale, with changes evident at approximately 1ms. In both cases (3.6m and 1.8m horns) the observed changes in behaviour occur at a time which corresponds well with the measured round trip time or each instrument. 4.3 Frequency A peak detection program was used to calculate the time between consecutive pressure maxima or the mouthpiece pressure data. This was then used to calculate the equivalent requency o the note at that time. Figure 8 shows plots o calculated requency against time or the irst.18 seconds o each o the same our notes, played on both the 3.6m and 1.8m horns, given in igures 6 and 7. The data or the long horn is given in blue and that or the short horn is given in red. It can be seen that the requency settles more quickly or the notes played on the short horn (approximately 4ms) than those played on the long horn (approximately 8ms). Conclusions and uture work In this work, the transient behaviour in the motion o the brass player s lips has been studied. We have concentrated on the eect o length o horn on the nature o the transient; there are o course several other important actors, including the degree to which the mode requencies approach a harmonic relationship. This study is related to the goal o reining brass instrument physical models. The techniques developed provide additional inormation about real behaviour, both in the physical motion o the brass player s lips and the sound produced, with which to compare results o time domain simulations. It is intended that uture work will involve urther experiments carried out on a wider range o brass instruments and players in order to conirm the initial conclusions drawn rom the analysis. The development o a realistic tongue mechanism or the artiicial mouth [4] would allow similar investigations without the variability inherent in human playing. This could allow consistent testing o a range o instruments looking at or example, the ease o starting a note. Future work will also include relating initial playing requency to the amplitude changes, looking or evidence o a cleaner attack i the player is able to initiate lip vibration at a requency close to the instrument resonance. The observed changes in the envelope o the mouthpiece pressure occur on a timescale which is signiicant when compared with the overall transient length. We would thereore expect these dierences to be perceptible to the human ear. To this end listening tests on recorded radiated sounds, and/or synthesised versions o sounds incorporating the measured eects, orm part o ongoing work. 76
4 3 1 1 1 1 note F..3.4..6.7.8.9.6.61.6 1 note F.7.73.74.7.76.77.78.79.8.81.8 1 1 1 1 1 note F 3.66.67.68.69.7.71.7.73.74.7.76 1 note F 3.4.46.47.48.49..1..3.4. 1 8 6 6 4 4 6 8 note F 4 1.13 1.14 1.1 1.16 1.17 1.18 1.19 1. 1.1 1. 1.3 4 4 note F 4.91.9.93.94.9.96.97.98.99 1 1.1 3. 3 1. 1.. 1 1. note C.69.7.71.7.73.74.7.76.77.78.79 1 1 note C.6.63.64.6.66.67.68.69.7.71 Figure 6: and mouthpiece pressure transients or the 3.6m horn Figure 7: and mouthpiece pressure transients or the 1.8m horn 77
requency(hz) 1 1 9 9 8 F 1st harmonic 1.8m F nd harmonic 3.6m Reerences [1] K.W. Berger, Some actors in recognition o timbre, The Journal o the Acoustical Society o America, Vol. 36, pp. 1888 1891 (1964) [] E.L. Saldanha and J.F Corso, Timbre cues and the identiication o musical instruments, The Journal o the Acoustical Society o America, Vol. 36, pp. 1 6 (1964) requency(hz) requency(hz) requency(hz) 8..1.1 1 19 18 17 16 38 37 36 3 34 33 F 3 nd harmonic 1.8m F 3 4th harmonic 3.6m..1.1 F 4 8th harmonic 3.6m F 4 4th harmonic 1.8m 3..1.1 6 C 1th harmonic 3.6m C 6th harmonic 1.8m [3] J.S. Cullen, J. Gilbert, D.M. Campbell, Brass Instruments: Linear Stability Analysis and Experiments with an Artiicial Mouth, Acustica, 86(4): 74 74, () [4] O.F. Richards, Investigation o the Lip Reed Using Computational Modelling and Experimental Studies with an Artiicial Mouth, PhD thesis, University o Edinburgh, (3) [] A.H. Benade, Fundamentals o Musical Acoustics, New York: Oxord University Press, (1976) [6] A.H. Benade, Eect o dispersion and scattering on the startup o brass instrument tones, The Journal o the Acoustical Society o America, Vol. 4, pp. 96 97 (1969) [7] G.R. Plitnik and B.A. Lawson, An Investigation o Correlations between Geometry, Acoustic Variables, and Psychoacoustic Parameters or French Horn Mouthpieces, The Journal o the Acoustical Society o America, Vol. 16 (), pp. 1111 11 (1999) [8] D. Luce and M. Clark, Physical Correlates o Brass- Instrument Tones, The Journal o the Acoustical Society o America, Vol. 4 (6), pp. 13 143 (1967) [9] S.J. Elliot, J.M. Bowsher, Regeneration in brass wind instruments. J. Sound Vib. 198, Vol. 83. pp. 181 17 (198) [1] J.A. Kemp, Theoretical and experimental study o wave propagation in brass musical instruments, PhD thesis, University o Edinburgh, () 4..1.1 Figure 8: Plots o requency settling rom mouthpiece pressures shown in Figures 6 and 7 78