th

Proceedings of 20 International Congress on Acoustics, ICA 2010

23-27 August 2010, Sydney, Australia

The acoustics of wind instruments

and of the musicians who play them
Joe Wolfe, Jer-Ming Chen and John Smith
School of Physics, The University of New South Wales, Sydney 2052, Australia

PACS: 43.75.EF, 43.75.FG, 43.75PQ

ABSTRACT
In many wind instruments, a non-linear element (the reed or the player's lips) is loaded by a downstream duct the
bore of the instrument and an upstream one the player's vocal tract. Both behave nearly linearly. In a simple
model due to Arthur Benade, the bore and tract are in series and this combination is in parallel with the impedance
associated with vibration of the reed or player's lips. A recent theme for our research team has been measuring the
impedance in the mouth during performance. This is an interesting challenge, because the sound level inside the
mouth is tens of dB larger than the broad band signal used to measure the tract impedance. We have investigated the
regimes where all three impedances have important roles in determining the playing frequency or the sound spectrum.
This talk, illustrated with demonstrations, presents some highlights of that work, looking at several different
instruments. First order models of the bore of flutes, clarinets and oboes the Physics 101 picture are well known
and used as metaphors beyond acoustics. Of course, they are not simple cylinders and cones, so we briefly review
some of the more interesting features of more realistic models before relating performance features and instrument
quality to features of the input impedance spectrum. Acousticians and sometimes musicians have debated whether the
upstream duct, the vocal tract, is important. Setting aside flute-like instruments, the bore resonances near which
instruments usually operate have high impedance (tens of MPa.s.m-3 or more) so the first order model of the tract is a
short circuit that has no effect on the series combination. In this country, that model is quickly discarded: In the
didjeridu, rhythmically varying formants in the output sound, produced by changing geometries in the mouth, are a
dominant musical feature. Here, the impedance peaks in the tract inhibit flow through the lips. Each produces a
minimum in the radiated spectrum, so the formants we hear are the spectral bands falling between the impedance
peaks. Heterodyne tones produced by simultaneous vibration of lips and vocal folds are another interesting feature. In
other wind instruments, vocal tract effects are sometimes musically important: as well as affecting tone quality, the
vocal tract can sometimes dominate the series combination and select the operating frequency, a situation used in
various wind instruments. In brass instruments, it may be important in determining pitch and timbre. Saxophonists
need it to play the altissimo register, and clarinettists use it to achieve the glissandi and pitch bending in, for example,
Rhapsody in Blue or klezmer playing.
Now, how to play the flute. Well you blow in one end and
move your fingers up and down the outside. Monty Python.

that are real musical instruments, including one that explains

the puzzle in Figure 1.

In a simple model of a musical wind instrument, the bore is a

cylindrical or conical pipe, either open-open (flute family) or
closed-open (most others). This acts as a resonator, which
loads a nonlinear element (air jet, reed or players lips) that
converts DC air flow at higher pressure into AC. Together,
the system oscillates at a fundamental frequency near one of
the lower resonances of the bore and its higher harmonics are
impedance matched to the radiation field by the higher
resonances. In this model, the role of the player is to vary the
length of the resonator (using keys, valves or slide), to supply
the air at high pressure and to control some parameters of the
nonlinear elements.
Of course this is not enough, as Figure 1 demonstrates. In this
paper, we shall look at just a few of the reasons why. We
begin with a quick look at the pressure-flow behaviour of one
simple nonlinear generator. Next we review Benades
argument about how the ducts load the generator. Next we
look at some of the subtleties and complications of the ducts
ICA 2010

Source: Dickens et al. (2007a)

Figure 1. A sound spectrum of the same note played
by a flute (open-open pipe) and a clarinet (closedopen). But which is which?

23-27 August 2010, Sydney, Australia

Proceedings of 20th International Congress on Acoustics, ICA 2010

One complication is that the reeds of woodwinds or the lips

of a brass player are acoustically loaded by two ducts: the
instrument bore is located downstream; the players vocal
tract, upstream. We look here at some examples where the
tract resonances have dominant musical roles.

take AC power out of the system. (The radiation by far the

smaller term.) Note that the negative resistances are of the
order of 10 MPa.s.m 3, or 10 M.

Separating the two ducts is the autonomous (self-sustaining)

oscillator, the nonlinear element mentioned above. A clarinet
or oboe reed tends to close when the upstream pressure
increases, whereas the lips of a brass player tend to open into
the mouthpiece, or to move sideways. A model of all three
types is given by Fletcher (1993). As an example, we look
first at a clarinet reed.

The DC flow is asymmetrical: from tract to bore. The

nonlinear generator also: a clarinet or oboe reed tends to
close when the upstream pressure increases, whereas the lips
of a brass player tend to open into the mouthpiece. For
acoustic waves, however, the two are symmetrical. Both
ducts have resonances that fall in the acoustic range: after all,
we use our tract resonances for speech. In both cases, the
resonance frequencies can be varied.

A reed generator
The reed is thin and elastic and has its own natural frequency,
known to clarinetists as a squeak, which is usually higher
than the range of notes played on the instrument. It is
attached to the mouthpiece so that small deflections make big
changes to the aperture through which air enters, as shown in
Figure 2. The graphs on that figure how steady air flow into
the bore as a function of the pressure in the mouth minus that
in the mouthpiece. (Acoustic waves in the mouthpiece were
damped.)

The Benade model for two ducts

Benade (1985) made the following argument for the loading

of the reed or lip generator. Figure 3 shows a schematic in
which a reed or lip separates two ducts. First, continuity of
flow requires that the flow into the mouth and that into the
bore satisfy Umouth = U bore. The force that acts across the
reed or lip is P = P mouth Pbore . From the definition of the
impedance of each duct, looking into the ducts: P =
UmouthZmouth UboreZ bore. Combining these equations gives
P = U mouth (Z mouth + Z bore).
So the tract (measured at the mouth) and the bore act in
series.

Figure 3. A schematic of the reed or lip that lies

between the tract (left) and instrument bore (right).

Figure 2. Air flow U vs the difference between

pressures Pm in the mouth and Pc in the mouthpiece.
Different experimental curves (Dalmont and Frapp,
2007) show different force applied by the lip to the
reed. The inset shows a cross-section of a clarinet
mouthpiece and reed. When the pressure in the mouth
increases, the reed bends (black arrow) and reduces
the opening through which air enters the bore. The red
arrows indicate regions of positive and negative AC
resistance (the reciprocal of the slope).
At low mouth pressure (reed not bent significantly), the flow
increases with increasing pressure. (Assuming that the kinetic
energy of the high-speed air in the narrow aperture is all lost
in turbulence downstream, we should expect the pressure
difference to be proportional to the square of the flow). At
high mouth pressure, however, the reed closes with
increasing pressure, so the flow decreases with increasing
mouth pressure, and of course it closes completely at
sufficiently high pressure, that pressure being reduced if the
force applied by the lips is increased.
If we consider the AC behaviour implied by these curves, we
see that, at low mouth pressure, a small change in flow U
and the associated change in pressure P imply a positive
resistance for AC signals. Over a range of higher mouth
pressures, however, the AC resistance (the reciprocal of the
slope) is negative. This region is of course the region in
which the reed converts DC power to AC power: its negative
resistance will offset the positive resistance due to viscothermal losses in the bore and sound radiation, both of which
2

Consider the flow Ureed, that due directly to the motion of the
reed or lip, here assumed to have the same value on both
sides. Consider also Uair, that passing through the aperture
left by the reed or lip. If the generator operates to produce Uair
and pressure difference P, then the impedance loading that
generator is

Substitution for P from the previous equation gives

(1),
where the last equation indicates that the reed is in parallel
with the series combination of the mouth and the bore. We
shall look at these three impedances in turn.
Impedance measurements
Measuring the impedance spectra of instrument bores
requires precision and dynamic range: musicians are sensitive
to even small changes and sometimes pay large sums for
instruments whose physical properties differ only a little from
those of much cheaper models. Measuring the impedance
spectra in the mouth, during playing, requires measuring a
small probe signal in the presence of the much louder signal
ICA 2010

23-27 August 2010, Sydney, Australia

radiated inside the mouth. We have found two techniques

very helpful: The first is adjusting the spectral envelope of
the probe signal to compensate for the gain of the
measurement system (Smith et al., 1997) and for the noise
spectrum (Dickens et al., 2007b). The second is using only
nonresonant loads for calibration (Smith et al., 1997).
Aspects of this system are described elsewhere in this volume
(Dickens et al., 2010).

Proceedings of 20th International Congress on Acoustics, ICA 2010

have an input impedance: if it were complete, there would be

no aperture for air flow. In practice, the oboe, bassoon and
saxophone approximate truncated cones with, at the input, a
volume approximately equal to that lost by the truncation.

Impedance spectra of the bore

Lets begin with simple geometry. Figure 4 shows the
measured input impedance spectrum of a cylindrical pipe,
15 mm in diameter, 600 mm long and open to the air at the
far end. It shows the expected regularly spaced maxima and
minima whose magnitudes decrease with increasing
frequency because of visco-thermal losses at the wall. The
length and inner diameter of this pipe correspond
approximately to both a flute and a clarinet, both of which are
largely cylindrical. So which is it?

Figure 5. Standing waves in a cylinder. For the openclosed pipe (left) the pressure (red) has an antinode at
the closed end (left) and a node at the open end. The
open-open pipe has pressure nodes at both ends. These
waves correspond (left) to the first two maxima in
Figure 4, and (right) to the first four minima.
For a complete conical bore, solutions to the wave equation
are written in terms of spherical harmonics rather than sine
and cosine functions. The lowest frequency solution has a
wavelength twice the length of the radius of the sphere or the
length of the conical tube. Consequently, the lowest note on
the nearly conical instruments has a wavelength twice the
length of the instrument. So the oboe, which is also
approximately the same length as the flute and clarinet, has a
lowest note similar to that of the flute, while the clarinet
plays nearly an octave lower.
Real instruments
Figure 6 shows five measured impedances: three real
instruments and two simple geometries. One is a cylinder, as
in Fig 4 but for a shorter length. Another is a truncated cone,
where the truncation is replaced by a cylindrical section
having the same volume as the truncated section.

Figure 4. The measured impedance of a cylindrical

pipe, 600 mm long and 15 mm in diameter, and open
at the far end. the frequencies of maxima and minima
are indicated as pitches (Dickens et al., 2007a). A
1 M line is drawn on this and subsequent curves for
reference.

The real instruments in Figure 6 show a number of interesting

features. The flute shows very little variation in Z above
about 3 kHz. This is due to a Helmholtz resonance, at about 4
kHz, in parallel with the bore and that shorts it out at the
frequency of resonance. The mass of this oscillator is located
in the embouchure riser, a very short flaring duct at right
angles to the main bore. The spring is that of the air
enclosed upstream from the mass: a small volume between
the riser and the cork. The purpose of this parallel impedance
is to improve the intonation of the high registers. However,
by shorting out the bore resonances, it also has the effect of
imposing an upper limit to the playing range (about G7).

At the embouchure, a flute is open to the radiation field. It is

driven by a jet that must flow easily into and out of the bore,
so it operates near the minima of impedance. A flute, with all
tone holes closed, will indeed play the first eight or so of the
minima shown in Figure 4, if the player provides an air jet
with appropriate speed.

Z for the clarinet shows the effect of the cut-off frequency.

The tone holes in the lower half of the instrument are open
for this note. At low frequencies, each tone hole acts as a
short circuit to the outside radiation field, so the most
upstream tone hole determines the effective length. At higher
frequencies, the picture is more complicated.

A clarinet, in contrast, has the mouthpiece weve seen above:

it requires a large acoustic pressure to move the reed, so one
might expect it to operate near the maxima of impedance
shown in the figure. This simple, cylindrical model
approximately predicts the first two notes played on a clarinet
with all tone holes closed.

In each tone hole is a mass of air with a finite inertance:

although the tone hole provides an open pathway from the
bore to the outside, to produce flow through this hole requires
a pressure difference. At higher frequencies (larger
accelerations of the mass), larger pressure differences are
required.

This section on idealised bore geometries has omitted the

cone, for the obvious reason that a complete cone does not

At sufficiently high frequencies, therefore, the inertia of the

air in the tone holes effectively seals the bore from the

ICA 2010

23-27 August 2010, Sydney, Australia

outside so, above about 1.5 kHz, we note that the peaks in Z
for the clarinet appear at frequency spacings about half that
of low frequencies. Thus, for low frequencies, the bore is
effectively terminated by the first open tone hole, about half
way along. In contrast, high frequencies dont see the open
tone holes and travel the whole length of the bore before
reflecting to make standing waves. (In the case of the flute,
the cu-off frequency cannot be seen in the frequency range
plotted, because of the Helmholtz shunt discussed above. See
also Wolfe and Smith, 2007.)
The cut-off frequency can be estimated by treating the boretone hole array as a continuous transmission line (Benade,
1960) or as an infinite array of finite elements (Wolfe and
Smith, 2007). For the clarinet, this is about 1.5 kHz, for the

Proceedings of 20th International Congress on Acoustics, ICA 2010

flute, about 2 kHz. (This resolves the riddle posed by Figure

1: the higher harmonics produced by the vibrating reed of the
clarinet fall above the cutt-off frequency, so there is no
systematic difference between odd and even harmonics. As to
which is which: the jet of the flute produces some broad band
sound, which is part of its characteristic timbre. In the concert
music tradition, the clarinet, alone among the woodwinds, is
played without vibrato, and so its spectrum usually has
narrower harmonic peaks.)
For the simple cone-cylinder, the extrema in Z become
weaker more rapidly with increasing frequency than for the
cylinder. This is one effect that deprives the saxophone of
strong, high frequency resonances. Well see later that this
has important consequences for performance technique.

Figure 6. Measured input acoustical impedance spectra (Chen el al., 2009a). From the bottom, they are a cylinder, a flute, a
clarinet, a soprano saxophone and a cone-cylinder combination, where the volume of the cylindrical section equals that of the
truncation of the cone. The flute and saxophone have the fingering that plays C5, the clarinet C4. The length of the cylinder
was chosen so that its first maximum is at C4 and its first minimum at about C5. The length of the cone gives a first
maximum at C5. Thus these pipes could all be said to have the same acoustical length, as indicated by the vertical line at
right. For comparison, the 1 M bar is included on each.

Figure 7. Measured acoustical impedances for brass instruments: a Bb bass trombone (first position, valve not depressed), a
Bb trumpet (no valves depressed) and a horn in the open Bb and F configurations (no finger valves depressed) (Chen, 2009;
Wolfe, 2005). The frequency scale for the trumpet is twice that of all the others, which shows that the trumpet is rather like a
one-half scale model of the trombone: in Italian, tromba means trumpet and trombone means big trumpet.
4

ICA 2010

23-27 August 2010, Sydney, Australia

Proceedings of 20th International Congress on Acoustics, ICA 2010

Impedance spectra of brass instruments

closer to 1000, 2000 Hz etc.

Unlike earlier relatives such as the serpent and the keyed

bugle, modern brass instruments have no tone holes: their
lengths are varied by valves or slides. Figure 7 shows that
they also have a cut-off frequency however: for sufficiently
small wavelengths, the bell radiates efficiently, so the
reflection coefficient is low at high frequencies, so there are
no strong standing waves and no strong impedance peaks. In
Figure 7, we see that the players hand in the bell of the horn
increases reflections. This not only increases the number of
playable resonances, but also affects the tuning.

The tract is not cylindrical, of course: if the player constricts

the tract at any point, e.g. with the tongue, then that lowers
the frequency of modes having displacement antinodes near
that point. Usually, changes due to changes in tongue
position are less important for the first Z maximum, because
its wavelength is rather longer than the tract.

These instruments all have substantial cylindrical sections,

which include the valves and slides. They also have a flare
and a bell at one end, and a cup- or cone-shaped mouthpiece
at the other, linked to the cylindrical section by an
approximately conical section. The net result of this
geometry is that the second and higher impedance maxima
fall close to the harmonic series, i.e. at frequencies 2f0, 3f0,
4f0 etc. The first, however, falls well below f0, and is not
played. Players can, however, play what they call a pedal
note at approximately f0. While there is no resonance at this
frequency, there are resonances at several of its harmonics.
The seventh resonance (here and in Figure 4) does not fall on
a note in common Western scales, hence the half-sharp
symbols. (The resonances are discussed in more detail by
Backus, 1976 and Wolfe, 2005.)
Impedance spectra of the vocal tract
The shape of the vocal tract is complicated, Further, it varies
with time as we move tongue, jaw, lips, palate and glottis.
We shall see some measurements later but, for the moment,
lets ask what features we might expect in Z(f).
There are two simplifications: To play a wind instrument, one
doesnt want a short circuit through the nose, so the palate
seals the nasal tract from the bucal tract. Further, the jaw and
lip positions are almost fixed so as to make an air-tight seal at
the mouth.

During performance on a range of instruments, we have

observed a maximum in the mouth impedance somewhere
around 200 Hz for nearly all players and conditions, and
another whose frequency can be varied with different
articulations between about 400 and 1800 Hz. We shall see
some of these in the figures to come. However, many of our
measurements do not include frequencies below 200 Hz. The
reason is the difficulty of making measurements of
impedance spectra in the mouth during wind instrument
performance: in the presence of a very large signal produced
in the mouth by reed or lips, one must limit the frequency
range of the probe signal so as to concentrate sufficient
power in the frequencies to be measured. Fritz and Wolfe
(2005) give some comparisons between measurements and
models. The measurements, however, were made while
subjects mimed.
Once we could measure impedance spectra during
performance, our first target was the didjeridu. One reason
was that the instrument is iconically Australian. More
importantly, it is clear from listening to the instrument that
vocal tract effects are not only involved, but are of the
greatest musical interest.
The vocal tract and the didjeridu
The didjeridu is made from the trunk or sometimes the
branch of a eucalypt tree that has been hollowed out by
termites, leaving an irregular duct. The ends are trimmed and
a ring of wax is usually applied to the narrower mouth end, to
achieve an airtight seal around the mouth and to improve
player comfort (Fletcher, 1996).

The glottis is the name of the aperture at the larynx. Even

with the glottis open, there is still a considerable constriction
at the glottis so, at high frequencies, one would expect a
strong reflection. Further, Mukai (1992) reports that
experienced wind players tend to keep their glottis aperture
rather small. The respiratory tract below the glottis branches
many times before terminating in the alveoli: its resonances
at acoustic frequencies are thought to be weak.
So, as our first approximation, lets picture the wind players
tract as a cylinder, nearly closed at the glottis. The nearly is
very important. If the glottis were closed (easy enough to do,
but one cant play a wind instrument for very long with no air
supply), Z would be infinite for DC. The small glottal
aperture, however, makes the impedance low at low
frequencies. Where is the first maximum?
For round numbers, lets consider a tract of length 0.17 m
from mouth to glottis. If the glottis were completely closed,
then we should expect maxima in Z at zero frequency, and
also at f = nc/2L = 0, 1000, 2000, 3000 Hz etc. If ideally
open, maxima at c(2n+1)/4L = 500, 1500, 2500 Hz etc.
Of course its neither ideally open, nor even very open. At
sufficiently low frequencies, with >> L, it might operate as
a Helmholtz resonator, with mass of air in the glottis
supported on the spring of the air in the tract. At the
Helmholtz resonance, there would be a maximum in
impedance. For the higher resonances, with higher reactances
at the glottis, the peaks in Z (for a cylindrical tract) would be
ICA 2010

Figure 8. Ben Lange, of the Mara people of Northern

Australia, worked on the didjeridu project while
studying engineering at UNSW.
The instrument is typically 1.4 m long, and roughly
approximates a truncated cone with a small angle, that varies
among instruments. The lowest resonance is typically at
about 60 to 80 Hz (about B1 to E2) and the next about 2.5 to
2.8 times (a tenth or eleventh) higher. The instrument is
blown somewhat like a tuba, with three obvious differences:
First, it usually plays only one note, near the first resonance,
with the second used only occasionally as a brief contrast.
Second, the musical interest is in the rhythmic changes in
timbre rather than in variations in pitch. Third, the player
5

23-27 August 2010, Sydney, Australia

does not stop playing in order to breathe.

The instrument is called the yidaki by the Yolngu people of
Northern Australia, one of the many peoples to whom the
instrument is culturally significant. The English name
didjeridu is thought to be onomatopoeic, the rhythmic
succession of vowels in the name resembling the rhythmic
variation in timbre that is idiomatic to performance.
The long, continuous note is achieved by an unusual
breathing technique, misleadingly called circular breathing.
While blowing into the instrument in a manner similar to that
used for brass instruments, the player fills his cheeks with air.
(Traditionally, the player is a man.) He then seals the mouth
from the respiratory and nasal tracts by lowering the velum or
soft palate, which allows him to expel the air from his cheeks
into the instrument, while simultaneously inhaling air through
the nose to fill the lungs.
While the mouth is thus sealed by the velum, its acoustic
properties are rather different from those it has when
connected to the vocal tract. Consequently, the timbre
changes abruptly.
Like the tuba, which plays in a similar pitch range, the
instrument requires a substantial air flow, so the player
regularly interrupts the normal tone with one or a few short
timbral contrasts that accompany the inhalation. This
provides the regular rhythmic ground. The player can then
introduce a range of different timbres during the more
sustained normal exhalation and thus create extended, varied
musical patterns.
We conducted a series of experiments in which we measured
the impedance spectrum inside the mouth (Figure 9) while
performers played using different articulations.

Proceedings of 20th International Congress on Acoustics, ICA 2010

The sound and impedance spectra in Figure 10 are for an

articulation in which the tongue is raised close to the roof of
the mouth, in a gesture described by some players as the ee
position. This produces what they called a high drone. Two
strong peaks in Z are seen and, at the same frequencies the
harmonics of the radiated sound are suppressed. At these
frequencies, the high impedance in the mouth prevents flow
between mouth and lips.
At about 1.8 kHz, a strong formant is seen in the sound
spectrum and heard in the radiated sound. This formant is
produced by the harmonics lying between the two strong
impedance peaks, which are therefore not suppressed. Similar
comparisons for different players and articulations showed
that formants (peaks in the spectral envelope of the sound)
coincided closely with local minima in the impedance
measured in the mouth and that minima in the sound spectra
coincided with the impedance peaks (Tarnopolsky et al.,
2006).
Interference tones on the didjeridu
Another method of varying the timbre of the didjeridu is
called vocalisation. In traditional performance, this can
simulate animal or bird sounds; in contemporary performance
it has a range of roles. While playing at note at the drone
frequency f, a player sings into the instrument at frequency
g. So, two different periodic vibrations (vocal folds and lips)
modulate the flow of air into the instrument. The radiated
sound has not only the harmonics of f and g, but also the
heterodyne components gf, f+g, etc, as shown in Figure 11.
In this experiment (Wolfe and Smith, 2008), one of us (JW)
played a simple cylindrical pipe in the manner of a didjeridu.
Electroglottograph electrodes were attached both in the
normal position (across the neck at the level of the vocal
folds) and either side of the lips. The MHz electrical
admittance of the two signals is also shown for lips and vocal
folds. (Sound files and more examples are given at
www.phys.unsw.edu.au/jw/yidakididjeridu.html)

Figure 9. An impedance head inside the mouth of a

player. The microphone capillary makes an inertance
divider that reduces the magnitude of the signal to the
microphone (Tarnopolsky et al., 2006).

Figure 10. Sound pressure spectrum radiated by the

instrument (fine lines) and impedance spectrum inside the
mouth
(thick
lines)
measured
simultaneously
(Tarnopolsky et al., 2005).
6

Figure 11. The spectrum of the radiated sound of a

didjeridu vocalisation. The middle trace shows the
MHz electrical admittance across the players lips, the
lower that across the neck at the level of the vocal
folds (Wolfe and Smith, 2008).
ICA 2010

23-27 August 2010, Sydney, Australia

The vocal tract and other lip valve instruments

The brass family (trumpet, horn, trombone, tuba and others)
are modern lip valve instruments. Compared with the
didjeridu, they have a narrow bore near the mouthpiece, and
the mouthpiece itself has a narrow constriction. This gives
these instruments resonances whose impedance peaks are
somewhat greater than those of the didjeridu: The didjeridu
impedance curves shown by Smith et al. (2007) have peaks
of typically about 10 to 30 M at low frequencies whereas
the instruments in Figure 7 have peaks of 30 to 100 M.
However, the magnitudes of peaks in Z for the didjeridu
decrease more rapidly with increasing frequency for the
didjeridu than for the brass.
This has the consequence that the effect on timbre of similar
articulation changes in the mouth are less striking on the
brass than they are on the didjeridu. However, they are still
significant enough for composers to include these effects in
works for trombone (Berio, 1966; Erickson, 1969).
Small changes in articulation have acoustical effects that can
change the pitch in two ways. First, a change in the
contribution of Zmouth to the series impedance Z mouth + Z bore
can change the frequency of the impedance peak at which the
mouth-lip-bore system operates. More dramatically, it can
change which peak in Z bore determines the playing frequency.
Both of these effects are shown in Figure 12. An artificial
trombone playing system used highly simplified models of
lip, vocal tract and glottis (Wolfe et al., 2003). To simulate
the relatively non-resonant lower tract, an acoustically
infinite duct was used (Dickens et al., 2010). The model
vocal tract representing the high tongue configuration played
sharper than the low tongue model when they operated on the
same impedance peak of the bore. As the slide was extended,
there was also a range over which the high tongue model
played on a higher resonance. Experienced players reported
the same effect: when they lowered the tongue while playing
a sustained note, and while holding all else constant,
sometimes the pitch fell slightly, while sometimes it dropped
to the next lower register.

Proceedings of 20th International Congress on Acoustics, ICA 2010

appraratus similar to that shown in Figure 9, though a probe

microphone replacing the inertance divider.

Figure 13. The impedance spectrum measured inside

the mouth of a trumpeter. For the upper graph, he was
playing written A3 (sounding G3), near the bottom of
the instruments range. For the lower, he was playing
written B5 (A5), over two octaves higher. In each
graph, the sharp peaks are (useful) artefacts: they are
the harmonics of the notes being played (nominally
196 and 880 Hz respectively), which are, of course,
strongly radiated inside the mouth. The broad peaks at
about 200 and 800 Hz are due to resonances in the
vocal tract. To improve signal:noise ratio, only part of
the spectrum was measured for a given note, so only
one tract resonance appears in each plot. From (Tusch,
2010).
In this case, the player does not tune the tract resonance to
match the note being played. It is possible, of course, that he
is using it for fine control of tuning (cf Figure 12), but he
evidently has sufficient control of the lip oscillator not to
need the assistance of vocal tract resonances over this range.
Further, this player reports no deliberate changes in the
position and shape of tongue or other articulators playing
over the standard trumpet range.
The vocal tract and single reed instruments

Figure 12. The playing frequency of an artificial

trombone playing system as the slide is extended from
the closed position (0 mm). The open filled circles
refer to geometrically simplified vocal tracts
represented by the sketches for high tongue (top) and
low tongue (bottom). From (Wolfe et al., 2003).
In the simple models of lip valve instruments (e.g. Fletcher,
1993), the playing frequency lies reasonably close to the
natural frequency of vibration of the non-linear oscillator.
Players of brass instruments have considerable control over
this frequency and are adept at choosing one of several
impedance peaks of the bore (see Figure 7).
Can the players select resonances by lip control alone, or do
they need to adjust the resonances of the vocal tract? Figure
13 shows measurements of the impedance spectrum
measured inside the mouth of a trumpet player using an
ICA 2010

In contrast with trumpetters, the players of reed instruments

have relatively modest control of the natural frequency of the
reed, although they can vary the stiffness and vibrating mass
by the position and force with which they bite. How are their
vocal tracts involved? We have been studying vocal tract and
embouchure effects on single reed woodwind instruments
(clarinet and saxophone) using measurements on both live
players (Chen et al., 2007-10) and artificial playing machines
(Almeida et al., 2010).
Concerning these instruments, acousticians and sometimes
even musicians have debated whether the acoustic effects are
important: while Clinch et al. (1982) stated that, for the
clarinet, vocal tract resonance must match the frequency of
the required notes. Backus (1985) wrote that resonances in
the vocal tract are so unpronounced and the impedances so
low that their effects appear to be negligible.
Direct measurements were difficult. Wilson (1996) used a
microphone inside a clarinet mouthpiece and another in the
players mouth. The ratio of the two pressures is
approximately proportional to that of the impedances of bore
and tract. The problem is that data are only obtained at the
playing frequency and harmonics. She deduced, however,
that vocal tract effects are sometimes involved in pitch
bending and in playing the second register without the
7

23-27 August 2010, Sydney, Australia

register key. We made measurements in the mouths of

players while they mimed (Fritz and Wolfe, 2005) but were
unable to relate the frequency of impedance peaks to that of
notes.

Proceedings of 20th International Congress on Acoustics, ICA 2010

the professional players tune the tract impedance peak near to

or slightly above the frequency of the note played.

More recently, we were able to make impedance

measurements in the mouth during playing, using an
impedance head built into the mouthpiece of a tenor
saxophone (Chen et al., 2007, 2008). At the same time,
Scavone and colleagues (2008) made measurements on the
saxophone, using a technique similar to that of Wilson.
The saxophone offers a spectacular example of the use of
vocal tract effects. Partly because of its largely conical bore,
the third and higher modes of standing waves in the bore
usually produce relatively weak impedance peaks, as shown
in Figure 6 and www.phys.unsw.edu.au/music/saxophone/
Without using the vocal tract, players can play the first
register, using the first impedance peak and the second
register, using one of two register keys that weakens and
detunes the first peak in Z. Not only beginners, but also some
players with considerable experience, are therefore limited to
about 2.6 octaves, the standard range of the instrument. The
altissimo register, using the third and higher peaks in Z,
requires tuning the vocal tract.

Figure 15. Frequency and magnitude of peaks in the

vocal tract impedance, measured during playing the
tenor saxophone. The red rectangle encloses the region
of normal playing, and the green line summarises
playing in the altissimo region.
Figure 16 shows measurements made in the mouths of
clarinettists, also made using an impedance head in the
mouthpiece, while they played the second and higher
registers (clarino and altissimo ranges). On the clarinet,
because it is not conical and because it doesnt have a
Helmholtz short circuit like the flute, the impedance peaks
remain relatively large, even at high frequency. Thus the
altissimo range can be played without the extensive practice
needed to tune vocal tract resonances. (Sound files and
examples are at www.phys.unsw.edu.au/music/clarinet/)

Figure 14. The broad, pale grey line shows the

impedance of the bore for the fingerings that play G4
in the standard range of the tenor saxophone and A#5
in the altissimo range. (On this transposing instrument
these notes are written A5 and C7.) The vocal tract
impedance, measured while playing these notes is
shown in red and blue respectively. The sharp peaks
are harmonics of the note played; the broad peaks are
resonances in the tract (Chen et al., 2007).
Figure 14 shows the impedance of the vocal tract of a
professional saxophonist playing notes in the standard and
altissimo register. In the former, there is no relation between
the tract resonance at about 550 Hz and the note played,
whose harmonics are visible as (useful) artefacts superposed
on the Z measurements. In the latter, the strong resonance in
the vocal tract lies close to the fundamental frequency of the
note played, which in turn is very close to the (relatively
weak peak) of the operating resonance of the instrument.
Figure 15 shows the frequency of peaks in the vocal tract
impedance plotted against the sounding pitch over both
ranges. On this plot, the magnitudes of Z are indicated by the
size of the circles used to plot them, and empty and filled
circles distinguish professional and (less experienced)
amateur players. This figure shows that, over the lower part
of the standard range, neither amateurs nor professionals tune
the tract near a note played. Over the altissimo range
however, and also over the upper part of the standard range,

Figure 16. Frequency and magnitude of peaks in the

vocal tract impedance, measured during playing the
clarinet. The red line summarises typical playing in the
clarion register while the green line summarises pitch
bending in the same range.
For the vocal tract of a clarinettist to influence or to
determine the pitch, the tract resonances must have rather
large magnitudes. This is used by advanced players when
bending the pitch over large intervals, or in producing the
glissandi that are used in klezmer playing and also in the
well-known solo that begins Gershwins Rhapsody in Blue.
The clarinet results include both normal playing (grey
circles), and pitch bending (black circles) where again the
size of the circle indicates the magnitude of the peak in Z.
For the pitch bending, the tract resonances have large
magnitudes and the playing frequency is close to that of the
peak in Z. For normal playing in the clarino register, as
expected, the magnitudes are not so large. What is perhaps
ICA 2010

23-27 August 2010, Sydney, Australia

surprising is that they are tuned to frequencies about 100 to

200 Hz above that of the note played.
So far, we have not mentioned the phase of the impedances
in equation (1). Lets return to the Benade model mentioned
above, where the impedance loading the reed generator is
(Zmouth + Zbore) || Zreed. The reed is largely compliant, with its
compliance being larger if the reed is soft. Consider what
happens when this is added in parallel to the series
combination of mouth and bore.
Figure 17 shows the magnitude and phase for all of these
terms in an example where the player was pitch bending on
the clarinet. In normal playing of this fingering (with no
strong peak in Zmouth), the note produced has a frequency
almost exactly at that of the peak in Zbore || Zreed, which is not
far below that of the peak in Zbore. In this example, however,
the note produced fell close to the peak in
(Zmouth + Zbore) || Zreed, which lay nearly 20% or a minor third
lower.

Figure 17. Measurements of the magnitude and phase

of (Zmouth + Zbore) || Zreed and its components. A
clarinettist was pitch bending on the clarinet (Chen et
al., 2009).
Near one of the resonances in the bore, the impedance is
largely inertive (pressure leads flow) at frequencies below
the peak, and compliant at frequencies above. The same is
true for the peaks in Z mouth. Adding the reed compliance in
parallel lowers the frequency and raises the magnitude of the
resultant peak. Because the peak in Zmouth is broader than
that in Zbore, the effect of the reed compliance on the
frequency of the peak is greater when the series spectrum is
dominated by Zmouth. This has the consequence that a peak in
Zmouth located below a peak in Zbore with comparable
magnitude can have a bigger effect on the playing frequency
than can a peak in Zmouth located above that in Zbore. This
explains why it is easier to bend notes down using the vocal
tract than up. More details on the single reed research are
given by Chen et al. (2010) in the proceedings of ISMA, a
satellite meeting of ICA, and published in the same
collection.
Provided that the peaks in Zmouth are of small magnitude, it is
possible to play the clarinet in tune without tuning the tract
resonances, however. Figure 18 shows a clarinet playing
machine built in our lab in collaboration with the ICT
research organisation NICTA. It was built to contest a
competition for automated musicians (Artemis, 2008). The
original version was designed to have no strong impedance
peaks in its mouth, so that only the resonances of the bore
determine the playing frequency. (Sound recordings,
including a duet with a live player, are available at
www.phys.unsw.edu.au/music/clarinet/). This version plays
fairly well in tune in summer. Although originally built for
the competition, it now provides us with an alternative
experimental tool for studying the effects, not only of vocal
ICA 2010

Proceedings of 20th International Congress on Acoustics, ICA 2010

tract geometries, but also some of the other control

parameters used in performance, including the air pressure,
the force and damping on the reed and the coordination of
finger motions. This research is also presented in the ISMA
proceedings (Almeida et al., 2010).
By comparing and contrasting measurements on human
players of woodwind and brass instruments with those on the
artificial systems, we expect to learn more about the
subtleties of control used by expert players. We shall also
learn more about the acoustics of wind instruments and of
the musicians who play them.

Figure 18. The NICTA-UNSW artificial clarinet

player (see Almeida et al., 2010).
Acknowledgments
This work involved a number of colleagues and students,
who are listed as authors on the relevant papers below.
Among these, we thank especially Paul Dickens, Neville
Fletcher and Alex Tarnopolsky. Our teams research is
supported primarily by the ARC. Thanks to our volunteer
subjects. Yamaha, The Didjshop and The Woodwind Group
provided instruments, Lgre and Vintage provided
respectively synthetic and natural reeds.

REFERENCES
A. Almeida, J. LeMare, M. Sheahan, J. Judge, R. Auvray,
K.S. Dang, S. John, J. Geoffroy, J. Katupitiya, P. Santus,
A. Skougarevsky, J. Smith and J. Wolfe, "Clarinet
parameter cartography: automatic mapping of the sound
produced as a function of blowing pressure and reed
force" Proc. Int. Symp. Music Acoustics, Sydney (2010)
Artemis
Orchestra
competition:
https://www.artemisiaassociation.org/photo_gallery_2007-2008

J. Backus, "The effect of the players vocal tract on

woodwind instrument tone" J. Acoust. Soc. Am. 78, 17-20
(1985)
J. Backus "Input impedance curves for the brass
instruments", J. Acoust. Soc. Am., 60, 470-480 (1976).
A.H. Benade, "Chapter 35: Air column, reed, and players
windway interaction in musical instruments" in Vocal
Fold Physiology, Biomechanics, Acoustics, and
Phonatory Control, edited by I. R. Titze and R.C.
Scherer, (Denver Center for the Performing Arts, Denver,
CO, 1985) pp. 425-452
A.H. Benade, "On the mathematical theory of woodwind
finger holes", J. Acoust. Soc. Am., 32, 15911608 (1960)

23-27 August 2010, Sydney, Australia

L. Berio, "Sequenza V; Solo Trombone" (Universal, NY,

1966)
J.M. Chen, Vocal tract interactions in woodwind
performance, Ph.D. thesis, the University of New South
Wales, Australia (2009)
J.M. Chen, J. Smith and J. Wolfe, "Vocal tract interactions in
saxophone performance" Proc. Int. Symp. Music
Acoustics, Barcelona, Agull, J. and Barjau, A., eds.
(2007)
J.M. Chen, J. Smith and J. Wolfe, "Experienced saxophonists
learn to tune their vocal tracts" Science 319, 726 (2008a)
J.M. Chen, J. Smith and J. Wolfe, "Saxophone acoustics:
introducing a compendium of impedance and sound
spectra", Acoustics Australia 37, 18-23 (2009a)
J.M. Chen, J. Smith and J. Wolfe, "Pitch bending and
glissandi on the clarinet: roles of the vocal tract and
partial tone hole closure" J. Acoust. Soc. Am. 126(3),
1511-1520 (2009b)
J.M. Chen, J. Smith and J. Wolfe, "How players use their
vocal tracts in advanced clarinet and saxophone
performance" Proc. Int. Symp. Music Acoustics, Sydney
and Katoomba, Australia (2010)
P. Clinch, G.J. Troup and L. Harris, "The importance of
vocal tract resonance in clarinet and saxophone
performance" Acustica 50, 280-284 (1982)
J.P. Dalmont and C. Frapp, "Oscillation and extinction
thresholds of the clarinet: comparison of analytical results
and experiments" J. Acoust. Soc. Am. 122, 1173-39
(2007)
P. Dickens, R. France, J. Smith and J. Wolfe, "Clarinet
acoustics: introducing a compendium of impedance and
sound spectra" Acoustics Australia, 35, 17-24(2007a)
P. Dickens, J. Smith and J. Wolfe, "Improved precision
measurements of acoustic impedance spectra using
resonance-free calibration loads and controlled error
distribution J. Acoust. Soc. Am. 121, 1471-1481 (2007b)
P. Dickens, J. Smith & J. Wolfe, "Improved method of
measuring reflection or impedance spectra using adapted
signal spectra and resonance-free calibrations" Proc. 20th
Intl. Cong. Acoust., M.Burgess, ed.,Sydney (2010)
N. Dell, R. James, J. Davidson & J. Wolfe, 'The effect of
mouthpiece, hand, and mute on the impedance spectra of
the horn' Proc. Int. Symp. Music Acoustics, Sydney,
(2010)
R. Erickson, "General speech: for trombone solo (with
theatrical effects)" (Smith publications, Baltimore 1969)
N.H. Fletcher, "Autonomous vibration of simple pressurecontrolled valves in gas flows" J. Acoust. Soc. Am. 93:
2172-2180 (1993)
N.H. Fletcher, "The didjeridu didgeridoo" Acoust. Aust.
24, 1115 (1996)

10

Proceedings of 20th International Congress on Acoustics, ICA 2010

N.H. Fletcher and T.D. Rossing, The Physics of Musical

Instruments (Springer-Verlag, New York, 1998) pp. 470480
C. Fritz, and J. Wolfe, "How do clarinet players adjust the
resonances of their vocal tracts for different playing
effects?" J. Acoust. Soc. Am. 118, 3306-3315 (2005)
M.S. Mukai, "Laryngeal movements while playing wind
instrument" Proceedings of the International Symposium
on Musical Acoustics, Tokyo, Japan, 239-242 (1992)
Music Acoustics "Didgeridoo acoustics"
www.phys.unsw.edu.au/jw/didjeridu.html (2004)
Music Acoustics "Clarinet acoustics"
www.phys.unsw.edu.au/music/clarinet/ (2006)
Music Acoustics "Saxophone acoustics"
www.phys.unsw.edu.au/music/saxophone/ (2008)
G.P. Scavone, A. Lefebre and A.R. Da Silva, "Measurement
of vocal-tract influence during saxophone performance"
J. Acoust. Soc. Am. 123, 2391-2400 (2008)
J.R. Smith, N. Henrich and J. Wolfe, "The acoustic
impedance of the Boehm flute: standard and some nonstandard fingerings" Proc. Inst. Acoustics 19, 315-330
(1997)
J. Smith, G. Rey, P. Dickens, N. Fletcher, L. Hollenberg and
J. Wolfe, "Vocal tract resonances and the sound of the
Australian didjeridu (yidaki): III. Determinants to playing
quality", J. Acoust. Soc. Am., 121, 547-558 (2007)
A. Tarnopolsky, N. Fletcher, L. Hollenberg, B. Lange, J.
Smith and J. Wolfe, J. "The vocal tract and the sound of a
didgeridoo". Nature, 436, 39 (2005)
A. Tarnopolsky, N. Fletcher, L. Hollenberg, B. Lange, J.
Smith and J. Wolfe, J. "Vocal tract resonances and the
sound of the Australian didjeridu (yidaki) I: Experiment",
J. Acoust. Soc. Am., 119, 1194-1204 (2006)
S. Tusch, A system for measuring impedance spectra in the
mouths of brass players, Practicum project report, The
University of New South Wales, Sydney (2010)
T.D. Wilson, The measured upstream impedance for clarinet
performance and its role in sound production, Ph.D.
dissertation, University of Washington, Seattle, WA
(1996)
J. Wolfe and J. Smith, "Cut off frequencies and cross
fingering in baroque, classical and modern flutes" J.
Acoust. Soc. America, 114, 2263-2272 (2003)
J. Wolfe, A.Z. Tarnopolsky, N.H. Fletcher, L.C.L.
Hollenberg, and J. Smith, "Some effects of the player's
vocal tract and tongue on wind instrument sound". Proc.
Stockholm Music Acoustics Conference (SMAC 03), (R.
Bresin, ed) Stockholm, Sweden. 307-310 (2003)
J. Wolfe, "Introduction to brass acoustics"
www.phys.unsw.edu.au/jw/brassacoustics.html (2005)
J. Wolfe and J. Smith, "Acoustical coupling between lip
valves and vocal folds", Acoustics Australia, 36, 23-27
(2008)

ICA 2010