ISMA 2014, Le Mans, France

Acoustical Impedance of the Xiao

Y. Lan and C. Waltham
University of British Columbia, Dept of Physics & Astronomy, 6224 Agricultural Road, Vancouver, BC,
Canada V6T 1Z1
[email protected]

517
ISMA 2014, Le Mans, France

The xiao is a Chinese end-blown flute with a history of over a millennium, traditionally made of bamboo, notched
at the blowing end, with six or eight finger holes. The tone range of the xiao is two to three octaves. Tones
starting from the second octave come from over-blowing, and cross fingerings have to be used for the third
octave. Currently most xiaos have difficulties in sounding the higher notes, which also have serious intonation
problems. This paper aims at explaining and solving the xiao’s problems by studying its input impedance. As
an air-reed instrument, the xiao plays at its input impedance minima. We use the transmission-matrix method to
model the instrument, and experimentally measure the input impedance to validate the model. For finger hole
configurations of 24 tones in the two and a half octaves under test, the model has a maximum deviation of 8 cents
from measurements. Then the player’s effects are taken into account, and the model is able to predict the tuning of
a xiao with any tone hole positions, sizes, and arbitrary bore shape along the symmetry axis. Based on this model,
numerical optimizations were applied to find the best configurations. A xiao made of PVC pipe with optimized
tone holes shows good tuning results. Modifying the bore shape shifts the frequencies of the impedance minima
and can be used for controlling the brightness and volume of the instrument. Our optimizations of the bore shape
are ongoing.

1 Introduction cannot correct tuning of every note. In addition, the bore

of a xiao is known to have effects on balancing the tuning
The xiao[1] is traditionally made of bamboo with a node of notes in different octaves and a conical shape tapering
at the blowing end for embouchure. The node is either to smaller diameter at the pipe’s end is preferred, similar
broken through in the southern style (see the top view of to the Baroque flute. A bamboo pipe near the root part
xiao D in Figure 1), or kept with a rectangular or an oval is usually chosen to make the xiao because the tapering
cutting in the northern style (xiao Z, B, K). The styles refer will be naturally formed after removing the bamboo nodes
to regions of China. When playing a xiao with the southern and smoothing the inner wall. A more recent technique
style opening, the top end is blocked by the jaw of the player, for tuning a xiao is by adjusting its bore shape, probably
leaving only the embouchure in the front of the pipe open, learnt from the tuning techniques of the Shakuhachi as an
similar to playing a xiao with the northern style. An air application of the perturbation method of Benade [4, 5].
jet is blown over the edge of the embouchure to excite the Both the hole positions/diameters and bore shape
instrument. The edge usually has a 30◦ to 60◦ angle with the adjustments require the maker to test and modify the
pipe, and has three kinds of shape: “U”, “V” and “Arc”. The instrument again and again. It would be very beneficial if
“U” or “V” shape is made by notching from inside of the tunings can be predicted and the instrument be optimized
bamboo pipe and the “Arc” shape is made by cutting off an in a computer. Plitnik el al. used a transmission-line
oblique plane from outside of the pipe. A xiao with southern model to calculate the input impedance of the oboe in the
style opening and “Arc” shaped embouchure edge is similar 1970s[6], and Ando el al. modelled the input admittance
to a Japanese Shakuhachi[2] in the head part. of the Shakuhachi using the same method in the 1980s[2].
The xiao is played vertically, each hand of the player Later the method was adapted to be the Transmisson-Matrix
controls half of the six or eight finger holes, and the top Method(TMM) and which has been widely used in various
finger hole is located at the back (thumb hole). For the wind instruments like the flute and saxophone[7, 8].
first octave, finger holes are opened in sequence to raise a Similar to the flute and other air-reed instruments,
half tone or a full tone (for all upper finger holes, lower the xiao plays near its input impedance minima. Input
holes are kept open, except for a cross fingering[3] note). impedance of the xiao is measured as described in Section
The second octave comes from over-blowing with the same 2 and modelled by TMM in Section 3. Section 3 also
fingering except for the cross fingering one. For all notes in discusses some irregularities observed in the impedance
the third octave, cross fingerings are used. The note played curves, and their relationships with the woodwind cutoff
by opening all the lower hand finger holes defines the key of frequencies[4, 3]. Numerical optimizations on the xiao’s
the instrument. All the instruments under study in this paper tuning is described in Section 4 .
have a U-shape embouchure edge with eight finger holes and
are in key of G, the lowest note D4 (293.7 Hz) is played with
all finger holes closed. The lengths of our instruments range 2 Acoustical impedance measurements
from 650 to 980 mm, but they all have two to three pairs of
adjustment holes (tassel holes) starting at around 540 mm We followed the method of Dickens et al.[9] and built
from the embouchure end. an impedance tube for the xiao, made out of brass with
A good xiao is expected to play in tune for two octaves inner diameter 7.9 mm, outer diameter 9.5 mm and total
and a half. The finger hole positions and diameters have a length 400 mm. One end of the tube was coupled to an
predominant effects on the tunings. With knowledge from compression driver through a cone and the other end,
experienced xiao makers and practice of making several defined as reference plane, was to be connected to the xiao’s
xiaos, we acknowledge that the xiao makers locates and embouchure or calibration loads. Four 6 mm microphones
drill the finger holes (and the lower adjustment holes) were located at 10 mm, 50 mm, 150 mm and 250 mm from
using parameters from existing instruments or given by the reference plane. Impedance measurements were made
experienced makers, then the instrument is played to test at the UBC anechoic chamber[10] with a Presonus 44VSL
its tunings and finger holes’ positions/diameters are slightly 24 bit sound card to collect the microphone signals and
adjusted accordingly. However each finger hole controls output a log-scaled swept signal to the driver. We used the
pitches of at least two notes and sometimes the adjustments full calibrations for the impedance tube as described by

518
 ISMA 2014, Le Mans, France

an end correction -3.78 mm is applied at the discontinuity.

For measurement on the tube with it far end closed, the end
correction is -3.22 mm.

101 Theory
Measurement

334 504 674

Z/Zc
100 165

673.2
334.1 504.0
164.9
10-1
0.5

Phase (π)
0
-0.5 250 500 1000 2000 4000
Frequency (Hz) (log-scaled)

Figure 2: Theoretical and measured impedance of an open

cylindrical pipe with inner diameter smaller than that of the
impedance tube.

Impedance of the xiao is measured by blocking the top

part of the pipe and attaching the U-shaped opening to the
impedance tube (see Figure 3). Modelling clay (blue) was
used to fill the gaps. The measurement results are shown in
the next Section together with the modelling results.

Figure 3: A PVC xiao under impedance measurement.

Figure 1: Four xiaos Z, B, K, D made of bamboo. Two xiaos

O, P and a xiao head H made of PVC pipes. Holes of xiao P
are labelled: f1 to f8 are eight finger holes and a1 1 to a2 2 3 TMM model of the xiao
are two pairs of adjustment holes (tassel holes).
The xiao’s body part from 100 mm below the
embouchure to the pipe end can be easily modelled by
Dickens et al., except for changing their 97 m resistance the TMM . The special geometry of xiao’s embouchure is
load to a brass tube of 2880 mm (measured by a tape) in not a standard element in the TMM and need to be tested
length and of the same diameters as the impedance tube. experimentally.
At first, the measured impedance curves show ripples of
about 55 Hz interval. The ripples come from inaccurate 3.1 Model of the xiao head
measurement on the calibration tube’s length L, and were
removed by adjusting L in the calculation of the tube’s A xiao head of inner diameter 15.2 mm, wall thickness
theoretical impedance. 2.8 mm and pipe length 100 mm was made of a PVC pipe
After the above calibrations, what’s measured is the (see H in Figure 1). The U-shaped embouchure hole was
impedance at the reference plane with a circular area of formed by cutting from one end of the pipe with a 7.9 mm
diameter 7.9 mm, which is poorly matched to our xiaos’ drill, the pipe was mounted with a 45◦ angle to the drill to
smaller U-shaped openings. To get the real input impedance form an oblique edge. The outside U-shape is 4.5 mm in
at the the embouchure, the discontinuity needs ultimately length and the area is equivalent to a 6.06 mm circle. The U-
to be modelled by the multimodal theory[11]. For now we shape’s length extends to 7.3 mm at the inside of embouchure
used end-corrections instead at the discontinuities. A test and is equivalent to a 8.06mm circle. Then the embouchure
measurement was made on an open tube with inner diameter hole was modelled as a short conical waveguide located at
6.4 mm, outer diameter 7.9mm and length 999.0 mm. The the geometrical center of the inside U-shape.
measurement matched well with the theoretical impedance To model the xiao embouchure part, the equivalent
of the tube, see Figure 2. For the theoretical calculation, T circuit[12] of a side hole was used with the following
modification: the conical waveguide was used in this T

519
ISMA 2014, Le Mans, France

circuit with an end-correction temb . Then a transfer matrix[7] To validate the model, the xiao P was played by one
was used to calculate the impedance at the outside U-shaped of the authors (Y. Lan), each note was played at moderate
opening. The xiao head with its end closed was measured volume and recorded for 5 s. The xiao was played normally
and modelled, and an end-correction of temb,c = 3.52 mm is and intentional pitch adjustment at the embouchure was
used. For the xiao with open end, radiation impedance at the avoided. The playing frequencies were measured right after
end is accurate enough for the imaginary part[13], but the the impedance measurements in the same anechoic chamber
real part depends on the geometry of the end[14]. and can be to avoid temperature change. The SPL of the notes were
written as: averaged to 1Hz resolution and plotted together with the
<(Zrad ) = x ka2 Zc (1) admittance curve in Figure 5.
Here, x is a coefficient depending on the flange condition and
sharpness of the edge, k is the wave number, a is radius of the 293
100 585 102
opening and Zc is the characteristic impedance. The value of 293 878 Z/Zc , TMM

Normalized impedance/admitance
1175 Z/Zc , Measurement
586
x affects the amplitude of impedance curves. For our flute 80 879 1172 1467 1470 Y = Zc /(Z + Zemb)
head with its end open, fitting results was obtained as x = 101
0.212 and temb,o = 3.72 mm. Measurement and modelling 60

SPL (dB)
results of the xiao head are shown in Figure 4. 100
40
10−1
102 (closed) 20 1195
1488
1491
Theory 595 891 895 1195
101 Measurement 297 595 SPL of playing D4 (293.7Hz)
0 296
100 500 1000 1500 2000 2500 3000 3500 4000
Frequency (Hz)
Z/Zc

10-1
120 521 102
10-2 Z/Zc , TMM
1176

Normalized impedance/admitance
10 -3 899 Z/Zc , Measurement
102 100
(open) Theory Y = Zc /(Z + Zemb)
101
1171
101 Measurement 80
SPL (dB)

100
Z/Zc

60 100
10-1 40
10 -2 902 10-1
500 1000 1500 2000 2500 3000 3500 4000 20 911
Frequency (Hz) 1207 1206
538 SPL of playing D6 (1174.7Hz)
0 536 10-2
500 1000 1500 2000 2500 3000 3500 4000
Figure 4: Impedance of the xiao head H. Frequency (Hz)

Figure 5: Impedance, admittance and SPL spectrum of xiao

P. Upper: playing D4 with all finger holes closed, lower:
3.2 Model of the whole instrument playing D6 with cross fingering oxxx-ooxo.
Now the input impedance of the xiao can be modelled by
adding the body part in TMM. Similar to the xiao end, real In the upper figure, the SPL spectrum of tone D4 has
parts were added at the side holes’ radiation impedance and obvious harmonics located at approximately integer
inner length correction as Eq. (1). We obtained x = 0.48 multiples of the playing frequency (293 Hz). The
for the xiao side holes. Results of measured and modelled harmonics are generated by non-linear effects of air-jet
impedance of xiao P for two representative fingerings are at the embouchure[15], and will not necessarily match the
shown in Figure 5 (the red and blue curves). maxima of the admittance curve. However, alignment of the
The measured impedance curve cannot be directly admittance maxima affects the amplitude of the harmonics.
used for predicting the playing frequencies of air-reed To play a brighter tone color and larger volume, a xiao
instruments, because the player’s effects are not considered should have the admittance maxima (impedance minima)
in the measurements. The player’s effects include: covering aligned by integer multiples.
ratio of the embouchure hole, radiation impedance at the In the lower figure, results of a cross fingering oxxx-ooxo
embouchure, velocity of air-jet, temperature and air-content playing D6 is shown (starting from the upper most thumb
changes inside the pipe. To the authors’ knowledge, these hole f8 to the lowest finger hole f1, x means the hole is
effects have not been well modelled yet. Dickens used an closed and o means open). This fingering is also used for
empirical formula for the flute embouchure (Eq. (7.2) in C5 (523.3 Hz, the model predicts 521 Hz for xiao P) in the
[7]). We found the formula also works for the xiao and the first octave. So naturally impedance minima of this fingering
embouchure effects can be expressed by a single formula: will not be harmonically aligned and C5 of our xiaos has
    obviously different tone color with other notes.
Zemb f (m) = 2.937log(m) − 11.6284 Zh . (2) For three PVC xiaos made and finger hole configurations
Here f (m) is the frequency of midi number m and Zh is the of 24 tones in the two and a half octaves under test, the model
radiation impedance of a hole on the pipe with the same area has a maximum deviation of 8 cents from measurements in
as the embouchure. Finally, the minima of Z + Zemb should Z. The accuracy of Zemb of xiao is still under improvement,
predict the playing frequencies of xiao. The xiao plays at the but for now the minima of Z + Zemb predict the playing
first few minima by changing velocity of the air-jet. frequencies of all the notes at least within 20 cents.

520
 ISMA 2014, Le Mans, France

3.3 Irregularities in impedance curves and

Distance to the embouchure (mm)

700
woodwind cutoff frequencies
600
It is known that an open or closed cylindrical or conical 500 101

pipe has harmonic impedance curves (see the typical open 400

Z/Zc
pipe impedance in Figure 2). However impedance of pipes 100
300
with open side holes is not always harmonic. While studying 200
10−1
the impedance and cavity mode of a guqin[16], it was
100 Z at embouchure
recognized that the irregularities in impedance curves come Zc
0
from resonances happening at shorter segments of the pipe. Frequency (Hz) 101
The resonances were validated by as below. 100

Z/Zc
Z
Zc at hole f8 10−1
Sound pressure of the longer PVC xiao O was calculated
500 1000 1500 2000
by TMM and shown in Figure 6. The contour shows the Frequency (Hz)
real part of pressure, red for positive and blue for negative.
The pressure is normalized to the embouchure pressure, Figure 7: Xiao O with fingering oxxx–ooxo, see the caption
and input impedance at the embouchure is superimposed. of Figure 6 for details.
At about 1.2 kHz, a pattern can be seen between the pipe
end and the adjustment hole a1 1. Input impedance at a1 1
was calculated by TMM and plotted in the narrow box, cutoff frequency fc [4, 5]. Above fc , the impedance curve’s
it correctly shows minima at around 1.2 kHz. This may amplitude reduces and its number of minima increases,
explains why the xiao O can be played at the first three and interval of the minima correspond to standing waves of
the fifth impedance minima, but cannot be played at the the whole instrument (as if the holes were closed)[3]. The
fourth. In contrast, xiao P has a shorter segment formed by bottom of Figure 5 shows that at fingering oxxx-ooxo, xiao
the pipe end to a1 1. As the impedance in the top of Figure P has fc ≈ 1.5 kHz. However, above fc , xiao P with this
5 indicates, xiao P with all finger holes closed can be played fingering can still be played at about 1.6 and 2.5 kHz, which
at the first five and the seventh impedance minima, but not at is the third and fifth impedance minima of the embouchure
the sixth one. segment. All other impedance impedance minima cannot be
played at the embouchure, but some of them can be played
at the open holes of their segment, traced by their resonance
Distance to the embouchure (mm)

700 101
locations in the pressure contour (not shown). So above fc
600
the woodwind is not equivalent to a whole pipe with all holes
500 closed, although the standing waves were also observed to
100
400 extend to the whole pipe in the contour.
Z/Zc

300
10−1
200
100
4 Numerical optimizations
Z
Zc at embouchure
0
Frequency (Hz) 101 An optimization algorithm (L-BFGS-B[17]) was used
Z
Zc at hole a1_1 100 to optimize the xiao based on the playing frequencies
Z/Zc

10−1 calculated by the TMM. The optimization works by

500 1000 1500 2000 minimizing the total error, which is the sum of squares of
Frequency (Hz)
the notes’ deviations from the equal temperament[8]. At
Figure 6: Pressure distributions and impedance curves of first, hole positions and diameters were used as the variables
xiao O with all finger holes closed. The horizontal lines for the optimization. Xiao O is a halfway optimization result
indicates hole locations, solid line for closed holes and when Zemb had not been taken into account, and it has bad
dashed lines for open holes. tunings. Xiao P was made using an earlier optimization
result on January 2014.
The playing frequencies of xiao P was measured as
The resonances were further tested by the cross fingering described in Section 3.2 and plotted in Figure 8. Also plotted
oxxx-ooxo of xiao O, (see Figure 7). A pattern shows at are the playing frequencies of xiao Z and B. Z and B are of
about 900 Hz and position of the three closed holes. Input moderate quality and they have no sign of delicate tuning
impedance at f8 is plotted below the contour, it has minima by adjusting the bore shapes. But their tunings are better
at about 500 Hz and 900 Hz. Blowing at f8 can exactly sound than the xiao K and D made by Y. Lan when he was an
the two resonances. inexperienced maker.
Based on the observed relationship of impedance Our latest optimization result on the hole positions/diameters
irregularities and locations of resonances along the pipe, we with cylindrical bore have all 24 notes within deviations of
can state that any irregularities in the impedance curve have ± 15 cents. Then we include the bore shape optimization
a corresponding resonance at a shorter segment S of the by describing the bore by N conical segments, and the
pipe, formed by the pipe end with an open hole, or formed deviations reached ± 10 cents. Some preliminary results on
by two open holes. The resonances of segment S can only bore shape optimization on alignments of the impedance
be excited by blowing at either of its end (for example, at minima are given in Figure 9. We got the typical bore
Figure 7, the resonance at about 900 Hz cannot be played at shape of the Shakuhachi[18], shown as shrinking from the
the emboucure). embouchure and enlarging near the end. This type of bore
The above statement can be related to the woodwind shape has also been adopted by some xiao makers.

521
ISMA 2014, Le Mans, France

[2] Y. Ando et al. Measuring and calculating methods of

60 shakuhachi input admittance. Journal of the Acoustical
xiao Z xiao B xiao P
Society of Japan (E), 6(2):89–101, 1985.
40
Deviation (cents)

[3] J. Wolfe et al. Cutoff frequencies and cross fingerings

20 in baroque, classical, and modern flutes. J. Acoust. Soc.
Am., 114(4):2263–2272, 2003.
0
[4] A. H. Benade. Mathematical theory of woodwind
20 fingerholes. J. Acoust. Soc. Am., 31(11):1564–1564,
1959.
40 D4 E4F4 G4 A4 B4C5 D5 E5F5 G5 A5 B5C6 D6 E6 F6G6 [5] A. H. Benade. Fundamentals of Musical Acoustics.
300 500 1000 1500
Frequency (Hz) Dover Books on Music Series. Dover Publications,
1990.
Figure 8: Playing frequencies of the xiaos versus the
deviations from equal temperament. [6] G. R. Plitnik et al. Numerical method for calculating
input impedances of the oboe. J. Acoust. Soc. Am.,
65(3):816–825, 1979.
9.0
N=1, total dev=29 [7] P. Dickens. Flute acoustics: measurements, modelling
8.5 N=5, total dev=34 and design. PhD thesis, PhD Thesis, University of New
N=10, total dev=15 South Wales, 2007.
8.0 N=60, total dev=16
Radius (mm)

[8] A. Lefebvre. Computational acoustic methods for the

7.5 design of woodwind instruments. PhD thesis, McGill
University, 2010.
7.0 [9] P. Dickens et al. Improved precision in measurements
of acoustic impedance spectra using resonance-free
6.5
calibration loads and controlled error distribution. J.
6.00 100 200 300 400 500
Acoust. Soc. Am., 121(3):1471–1481, 2007.
Distance to the embouchure (mm) [10] C. E. Waltham et al. Acoustic imaging of string
Figure 9: Optimization results of the bore shape, × means instrument soundboxes. In Proc. Meet. Acoust.,
1
the finger hole locations. total dev=( di2 ) 2
P volume 19, page 035004. Acoustical Society of
where di is the deviation of the ith impedance minima to the America, 2013.
harmonic frequency in cents. di includes the first, second [11] V. Pagneux et al. A study of wave propagation
and fourth harmonics of all finger holes, cross fingering in varying cross-section waveguides by modal
notes are not taken into account. decomposition. part i. theory and validation. J. Acoust.
Soc. Am., 100(4):2034–2048, 1996.
5 Conclusions [12] C. J. Nederveen et al. Corrections for woodwind tone-
hole calculations. Acta Acustica united with Acustica,
The xiao’s acoustic impedance has been measured 84(5):957–966, 1998.
and a TMM model has been built considering the xiao’s
special embouchure. The TMM model has also been [13] J. P. Dalmont et al. Radiation impedance of tubes
used to calculate the pressure and resonances of the xiao, with different flanges: numerical and experimental
irregularities in the impedance curves and their relationships investigations. Journal of sound and vibration,
with the cutoff frequencies have been examined. Numerical 244(3):505–534, 2001.
optimizations on hole positions/diameters has produced
[14] J. M. Buick et al. Investigation of non-linear acoustic
a xiao with good tunings. Bore shape optimizations are
losses at the open end of a tube. J. Acoust. Soc. Am.,
ongoing and some preliminary results has been obtained.
129(3):1261–1272, 2011.
[15] N. H. Fletcher et al. Harmonic generation in organ
6 Acknowledgement pipes, recorders, and flutes. J. Acoust. Soc. Am.,
68(3):767–771, 1980.
The authors thank Alan Thrasher and the UBC Chinese
Orchestra for help with playing the xiao, Murray Hodgson [16] C. E. Waltham et al. Vibroacoustics of the guqin. In
of UBC Mechanical Engineering for sharing the anechoic This conference.
chamber and the Natural Sciences and Engineering Research
[17] R. H. Byrd et al. A limited memory algorithm for bound
Council (Canada) for financial support.
constrained optimization. SIAM Journal on Scientific
Computing, 16(5):1190–1208, 1995.
References [18] Y. Ando. Input admittance of shakuhachis and their
resonance characteristics in the playing state. J. Acoust.
[1] A. R. Thrasher. Xiao. Grove Music Online. Oxford U. Soc. Japan (E), 7(2):99–111, 1986.
Press. Accessed 12 May 2014.

522