The physics of the didgeridoo (didjeridu/ yidaki/ yiraki),
presented for non-specialist readers.c
The didgeridoo or didjeridu is the yidaki or yiraki
in the language of the Yolngu,
one of the peoples of Northern Australia, where the instrument
originated. This page explains how the instrument works and gives a non-technical introduction to some research from our lab. (for which the formal publications are below.)
The instrument is deceptively simple: it is just a wooden tube,
about 1.2 to 1.5 m long, hollowed out by termites in the thin
trunk or branch of a eucalypt tree, and with a ring of beeswax around the
mouthpiece for sealing and player's comfort. It is a very unusual
instrument: although it usually plays only one note, it is capable
of a spectacular range of different sounds, and the rhythmic variation
of these sounds is its chief musical interest. The painting on
the instrument can have considerable cultural significance.
The instrument is closed at one end by the player's lips and face. The
difference between closed and open pipes is explained in Open
vs closed pipes , which compares them using wave diagrams, air motion
animations and frequency analysis. Go there for an explanation of the
animation below. (The animation is for a simple, cylindrical pipe. However, most didjeridus are somewhat flared and have complicated surface geometries.)
How does the instrument work?
The player's lips are alternately opened and closed
by the higher pressure air flow from the lungs and the tension in the lips. This part of the mechanism is rather like that of a tuba
or trombone, and is discussed in Introduction
to brass acoustics. Like the brass instruments, the didjeridu
or yidaki is a closed
What makes its sound so unusual and varied is the strong interaction
- the sound waves in the instrument,
- the sound waves in the player's vocal tract,
- the motions of the player's lips and
- the flow of air between the player's lips.
In the brass instruments, the vocal tract has only a minor effect,
but in the didjeridu
A schematic diagram of the acoustic system (not to scale).
The player's lips produce a sound wave that travels into the instrument,
but it also travels in the other direction, into the vocal tract. The
vocal tract is a resonator that, in normal speech, can assist the radiation of some frequency bands, but not others. In fact its resonances are what allow us to produce different speech sounds: see voice acoustics for an introduction.
For a didjeridu player, the vocal tract is working backwards: it still has resonances, but the vibration is (usually) coming from the lips, rather than from the vocal folds. In either speech or in didjeridu playing, the frequencies at which the vocal tract resonates are determined by the shape of the tract, especially by the position and shape of the tongue, and the state of closure of the vocal folds.
In the spectrogram at right, Lloyd Hollenberg illustrates several
of the effects discussed above. The horizontal yellow line at the bottom of the spectrogram is the drone (at constant frequency.) The other horizontal lines are harmonics. The larger patches of light colour – the ones that vary – are formants. These are changing with time as Lloyd changes the shape of his tract.
. . . Download didj sound file in .wav format (470 k) or
In a few places, he vocalises: that is, his vocal folds vibrate
while he is playing. In this case, the vocal tract is being driven
by vibrations at both ends, by the lip vibration and the vocal
fold vibration. The frequencies are different, and the result
is not just two pitches, but also notes with pitches corresponding
to the sum of the two frequencies, and others as well. (Details
in the experimental
What are the key findings of our recent research?
Some background: Depending on how you position your mouth and tongue,
vibrations in the vocal tract can be enhanced or inhibited at different
frequencies. A band of enhanced frequencies is called a formant.
They are important in speech, where different formants identify different
speech sounds — so our ears are good at detecting them, especially
when they change. Producing strong formants, and varying them, is important
in didjeridu performance.
The frequency response of the vocal tract is quantified by its
spectrum, a measure of how difficult it is to produce vibration
at a given frequency. Our key technical advance has been to measure the
impedance spectrum, inside the mouth, during playing. We do this
by injecting a signal having hundreds of different frequencies,
and measuring how the tract responds to those frequencies. It
is a little bit like sonar, because we can use this information
to work out the shape, but that analogy is limited. In the photo,
Lloyd Hollenberg, physicist and didjeridu player, plays while
having the acoustic properties of his vocal tract measured.
This process is difficult because it is so noisy inside a
wind player's mouth — over 100 dB. That's because the vibrating
lips transmit a sound in both directions: into the instrument,
and also into the vocal tract. The waves that go into the vocal
tract interact with its resonances, and then some frequencies
pass into the instrument to emerge in the output sound.
To produce the strong formants characteristic of didjeridu performance, we found that the player produces two
or more strong peaks in the impedance spectrum. These inhibit
sound at their frequencies, and the uninhibited frequencies
in between are heard as a formant in the sound. For instance,
with the tongue very near the roof of the mouth, one can strongly
inhibit the frequencies around 1.5 and 2.0 kHz. This results
in a strong isolated band around 1.8 kHz that gives the insrument
a sound a bit like "ee". (See the technical
page for spectra, sound files and film clips.) Our hearing
is very good at identifying formants in the speech range, so
we are very conscious of their presence, especially when they
change over time as the player moves his tongue.
How do experienced and novice players differ?
It's easy to make a basic sound. Then you have to learn circular breathing.
Learning to make these strong formants takes a while. To be able to
produce the strong, clear formants requires learning how to produce strong resonances in the vocal tract. Other
techniques involve vocalising and playing at the same time: one gets
interactions between the vibrations from the lips and from the vocal
A virtuoso like William Barton
(recent soloist with the London Philharmonic and Sydney Symphony) can
produce an amazing range of sounds.
Implications for other instruments
The vocal tract effects are much less spectacular in brass instruments,
but still important enough that some composers, such as Luciano Berio,
have used them. In a previous paper, we used an artificial system that
'played' the trombone to show the effect of the 'vocal tract' geometry on
register, timbre and pitch. The vocal tract in this case was mechanical
and therefore of known geometry and acoustic properties. We have since measured those effects on real, human players, during playing.
Have there been previous studies on the acoustics of the yidaki?
Yes. Several papers have been published, including a few by members
of this team. The sound of the instrument has been analysed, and the
motion of the player's lips has been photographed.
In one of our previous papers, we measured the acoustic response
of players' vocal tracts while they mimed playing, and that certainly
gave clues about the mechanism. But miming does not come naturally
to players of this instrument, and we were especially uncertain about
the position of the players' vocal folds when miming. Hence the importance
of the measurements during performance.
The yidaki or didjeridu is an iconic Australian instrument and,
as an Australian research lab specialising in music acoustics, we
thought that we should.
Further, this instrument demonstrates the most spectacular case
of the vocal tract influencing the sound. By understanding it, we
hope to understand vocal tract effects in other
instruments, where the effects are much smaller, but still musically
Then there is curiosity. It is an unusual system with some subtle
features and some interesting physics. Science and technology often
benefit in unexpected ways from researchers looking into questions
just from curiosity.
Circular breathing and vocalisation
Circular breathing is idiomatic for the instrument (and also gives it
its onomatopoeic Western name). For most of the time, the player exhales
through the mouth in the ordinary way, but, when new air is needed,
the cheeks are filled with air, the soft palate is used to isolate the
mouth from the respiratory tract, and the air in the mouth is used to
continue playing by contracting the cheeks. Meanwhile, the player inhales
quickly through the nose. This requires coordination and a lot of practice.
The result is an uninterrupted sound from the instrument. The timbre
changes, however: during the normal part of the cycle, the whole vocal tract
is connected via the lips to the instrument. During the inhalation,
only the mouth itself is involved. There are great differences in the
resonances in the mouth for these two configurations, so the timbre
of the sound produced changes dramatically. Far from being a disadvantage,
players use this timbre difference, along with differences in loudness,
to establish the rhythm of the performance.
Vocalisations are much more common in modern styles than in traditional
performance. When the vocal folds vibrate during play, the vocal tract
is driven by two vibrations, at different frequencies, from the lip
vibration and the vocal fold vibration. The result is not just two
pitches, but a complicated set of frequencies corresponding to the
sum and difference of the two frequencies, and others as well.
We discuss the physics of circular breathing and vocalisation in this paper and the effects of vocalisation in this paper.
Does a skilled player use the glottis subconsciously when playing?
Apparently. None of the players to whom we have spoken are aware
of it, except during vocalisations.
Did your method of sound measurement impair the players' ability
Sometimes. Many players play with the instrument a bit to one side
of the mouth. It was easier for such players to deal with our system.
Why are some didjeridus better than others?
This is a subtle question and there are several subtle effects. However, there is one over-riding effect: players prefer instruments that do not have very strong resonances in the range 1-2 kHz. This is easy to understand in terms of the story that we tell above: this frequency range is the one in which players can manipulate the formants using the resonances of their own vocal tracts. They prefer instruments whose resonances that fall in this range are weak compared to the resonances of the vocal tract. This is explained in more detail in this paper.
Sounds and pictures
Download didj sound file of Lloyd Hollenberg playing
Because of the publicity aroused by our paper in Nature, various
journalists have been asking for pictures. Benjamin Lange, one
of the authors of this paper, is a member of the Marra people of
Northern Australia, where he learned to play yidaki in the traditional
style. He recently graduated in Electrical Engineering at the
University of New South Wales, where he worked on the didjeridu/
yidaki project on a vacation scholarship. On this
link, we provide photos of Ben taken by Kate Callas. These
photos may be used, with acknowledgment of the photographer and
of this URL, in stories about our work.
Links for more information
Some papers are reproduced with kind permission of the editors.
Wolfe, J., Tarnopolsky, A.Z., Fletcher, N.H., Hollenberg, L.C.L.
and Smith, J. (2003) "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.
Fletcher, N., Hollenberg, L., Smith, J. and Wolfe, J. (2001) "The
didjeridu and the vocal tract" Proc. International Symposium
on Musical Acoustics, Perugia. D.Bonsi, D.Gonzalez, D.Stanzial,
eds, pp 87-90.
Hollenberg, L. (2000) "The didjeridu: Lip motion and low frequency
harmonic generation" Aust. J. Phys.53, pp 835-850.