You are on page 1of 15

Applied Acoustics 66 (2005) 751–765

www.elsevier.com/locate/apacoust

Measurements of sound transmission


through panels of locally resonant
materials between impedance tubes
Kin Ming Ho, Z. Yang *, X.X. Zhang, Ping Sheng
Department of Physics, The Hong Kong University of Science and Technology, Clear Water Bay,
Kowloon, Hong Kong, China

Received 18 October 2004; received in revised form 10 November 2004; accepted 11 November 2004
Available online 21 January 2005

Abstract

We present a method for the measurements of complex transmission coefficient through


high transmission loss samples using three-sensors, two impedance tubes, monotonic wave
excitation, and phase sensitive detection. Having demonstrated the effectiveness of the method
on perforated plates measurements have been made on locally resonant sonic materials
(LRSMs). The transmission losses of perforated plates are found to decrease with decreasing
frequency down to 120 Hz, following the mass law. For the LRSMÕs panels with the same area
mass density but different compositions, the local resonance frequency (at which the transmis-
sion loss is maximum) is found to vary according to the predesigned value. Transmission losses
as high as 96.5 dB at 630 Hz and 87 dB at 250 Hz can be measured with good accuracy, with
corresponding phase spectra that match the theoretical prediction of LRSMs, confirming the
reliability of the transmission data.
 2005 Elsevier Ltd. All rights reserved.

Keywords: High transmission loss; Impedance tube; Local resonant sonic materials

*
Corresponding author. Fax: +852 2358 1652.
E-mail address: phyang@ust.hk (Z. Yang).

0003-682X/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.apacoust.2004.11.005
752 K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765

1. Introduction

The Impedance Tube Method (ITM) has been widely used in the last few decades
to measure the acoustic properties of materials [1–5]. In an ITM measurement, the
acoustic waves are confined within the impedance tube, which is typically a few cen-
timeters in diameter, and the size of the materials sample need only be large enough
to fill the cross-section of the tube. This avoids the need to fabricate test samples with
lateral dimensions several times the acoustic wavelength, e.g., as required in free field
tests, thus greatly facilitates measurements in the low frequency regime where the
wavelength is on the order of metres. The current impedance tube method employs
either a movable microphone [1] or two microphones at fixed positions [2] to mea-
sure the acoustic field between the sample and the sound source. The readings of
the microphone(s) directly yield the complex reflection coefficient of the sample, r.
The portion of the wave that is not reflected back by the sample, i.e., due to trans-
mission and dissipation, is then the sound power absorption coefficient a = 1  jr2j
To measure more accurately the transmission coefficient (or transmission loss) a
third microphone [6] or even a second pair of microphones [7–9] were placed behind
the test sample, either in open room [6] or in a second impedance tube [7–9]. The
highest transmission loss reported in these early works was about 50 dB [9] at fre-
quencies below 500 Hz.
We have developed a novel type of ‘‘locally resonant sonic materials’’ (LRSMs)
[10] that exhibits negative (dynamic) elastic constants over certain pre-designed
and tunable (local resonance) frequency ranges. This characteristic enables LRSM
to have transmission significantly lower than that dictated by the mass density law
[11], and acquire a large phase shift over a relatively short distance near the reso-
nance frequency. A typical transmission coefficient for even one unit layer of LRSM
is lower than 1%, which implies transmission loss above 40 dB, so the current imped-
ance tube method has to be carefully tested for the accurate measurement of the
complex transmission coefficient of high transmission loss composites. In this work,
we test the limit of minimum transmission coefficient that can be measured with
good accuracy with impedance tubes and small size (<wavelength) samples, and de-
monstrate that our method can accurately measure both the amplitude and phase of
the sound waves transmitted through samples with transmission loss 100 dB, such
as the LRSM.
One obvious potential problem in measuring high transmission loss samples, be-
sides the sensitivity and noise, is the leakage of sound wave around the sample being
tested due to inadequate sound isolation of the apparatus. Another potential prob-
lem is the finite sample sizes which could affect the acoustic behavior of the samples.
A perforated plateÕs transmission is expected to follow the mass law. This provides a
way to test both the presence/absence of leakage of the apparatus and finite size ef-
fects. A more stringent test is provided by LRSMs. An LRSM has distinct transmis-
sion features at its local resonance frequency. A transmission minimum occurs at the
resonant frequency which is usually over 10 times lower than that at the nearby non-
resonant frequencies, followed by a transmission maximum at 50–200 Hz above the
frequency. The phase spectrum of the transmission in the meantime first drops stee-
K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765 753

ply by 180 at the resonance frequency, then remains flat, and finally rises again by
180 at the frequency when the transmission is maximum [10]. Such distinct feature
of LRSM is well suited for the test of apparatus for high transmission loss samples.
If at the transmission minimum the signal detected at the back side of the sample is
due to leakage of sound waves instead of the ones transmitted through the sample,
the phase spectrum of the signal will not display the 180 drop, but will rather in-
crease with frequency. The 180 drop in the phase spectrum therefore serves as a
clear indicator that ensures the reliability of the obtained transmission minimum lim-
it of the apparatus.
The tests are divided into several stages. In the first stage no sample is placed
between the impedance tubes. Under such condition the results should be 100%
transmission and 0 reflection. In the second stage two porous samples are tested.
Their transmission spectra should display a gradual increase in transmission with
the decrease of frequency. In the third stage three LRSMÕs with the same average
area mass density but different compositions are tested. Their local resonance fre-
quencies, and therefore the transmission minima, are expected to follow the prede-
signed compositions of the samples. Finally, having obtained satisfactory results
from the tests in all the stages, a high density LRSM sample is measured to test
the minimum transmission limit of the apparatus, and the minimum transmission
is found to be 0.0015% at 630 Hz, which is equivalent to 96.5 dB in transmission
loss.

2. Apparatus

Our approach is based on modifying the standard method [1,2]. Impedance


tubes are used to generate plane sound waves inside the tube while screening out
room noise. Fig. 1 shows the schematics of our approach. The sample panel being
measured is held firmly on an optic table with clamps between two Brüel & Kjær
(B&K) Type-4026 impedance tubes of 10 cm in diameter, which are the standard
instrument in the standard method [1,2]. The joints between tubes and sample
are well sealed with rubber sheets to minimize leakage. The impedance tubes are
pressed to the sample, and then held firmly onto the optical table as well. The front
tube (27.5 cm in total length, to the left of the sample) contains a B&K loud-
speaker at the far end, and two Type-4187 acoustic sensors at two fixtures,
10 cm apart, of the tube. The distance between sensor-1 to the front surface of
the sample is 22 cm, and that between sensor-2 and sample is 12 cm. The wiring
of the sensors is modified so that the B&K type-2148 Signal Analyzer serves only
as a power supply to the sensors, and the electric signals from the sensors are mea-
sured with two lock-in amplifiers (Stanford Research SR-850). A third acoustic
sensor with an electronic gain about 100 times of that of the front sensors is placed
at the sensor fixture of the back tube, 10 cm away from the rear of the sample. The
rest of the back tube after the sensor is filled with 26 cm of sound absorbing mate-
rials up to the end which is terminated by a well sealed hard aluminum disk. The
back impedance tube effectively shields the room noise from the third sensor, so
754 K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765

Fig. 1. The schematics of the two-tube three-microphone method.

that the measurements can be carried out in a normal laboratory instead of a spe-
cially equipped quiet room [6]. A sinusoidal signal is sent from a third lock-in
amplifier (SR-830) to drive the loudspeaker through a power amplifier, which also
measures the signal from sensor-3. The frequency of the wave is scanned in a range
from 200 to 1400 Hz at 2 Hz intervals, while the electric signals, both in-phase and
out-phase, are measured by the three (two-phase) lock-in amplifiers. It takes about
2.5 s to acquire the data at each frequency for a lock-in amplifier time constant of
0.3 s. Single frequency excitation and phase sensitive detection significantly im-
prove the signal to noise ratio over the more widely employed broad band source
and autocorrelation multi-channel frequency analysis, which is more susceptible to
noise interference at low frequencies [8]. Furthermore, even at a constant drive to
the loudspeaker (sound source) the signals of the sensors can vary over several or-
ders of magnitude at different frequencies, either due to the resonance of the
impedance tube, the coincidence of nodal points with the location of the sensors,
or the local resonance of the sample. A broad band source could cause frequency
cross-over due to slight non-linearity of the sensors because of the large difference
in signal strength at different frequencies, one at tube resonance and one at nodal
point, for example. The use of monotonic source avoids such potential problems,
especially when the transmission loss is high.
To ensure that the acoustic properties of the samples are not altered due to fi-
nite size and limited by the way they are held, measurements are done several times
with samples held in different ways, such as held by the edge of the sample so the
actual size of the sample is preserved, or held at positions closer to the tube so the
effective size of the sample is reduced. For the samples reported in the article, no
noticeable variations in transmission due to different ways of holding the samples
are observed.
K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765 755

3. Theory

Although the derivation of the relevant formulae was already given in [8] for a
broad sound source, for completeness we briefly present it here as well. We first de-
fine the terms to be used in the derivation.

hn ¼ 2pfdn=c
c = speed of sound in air
f = frequency
k = 2pf/c
d1,2,3 = the distance from sample to the positions of sensor-1, sensor-2, and sensor-3,
respectively. d1 = 22 cm. d2 = 12 cm. d3 = 10.5 cm
df = length of the front impedance tube, 27.5 cm
rs = reflection coefficient of the loudspeaker
r = reflection coefficient of the sample
t = transmission coefficient of the sample
Xn = signal at sensor-n
X(z) = wave at position z along the tube length
z = coordinate along the tube length
A = amplitude of the wave emitted by the loudspeaker

Assuming the sound wave in the tube is plane, and taking the Z-axis direction to
the right, we obtain the total waves propagating towards the sample in the front tube
as
Aeikz
X f ðzÞ ¼ Aeikz ð1 þ rrs e2ihf þ ðrrs Þ2 e4ihf þ   Þ ¼ : ð1Þ
1  rs re2ihf
Here the first term is the wave (first wave) just emitted by the loudspeaker (source),
the second term (second wave) is the first wave reflected by the sample and the source
while traveling the entire length of the front tube twice, the third term (third wave) is
the second wave reflected by the sample and the source while traveling the entire
length of the front tube twice. . ., and the final result is the sum of infinite number
of these multiply reflected waves. Note that we ignore any wave that has entered
the back tube and then entered the front tube again. These waves would pass through
the sample at least twice. At 20% transmission and anechoic terminal reflection of
<50%, the amplitude of these waves would be less than 4% of the incidence waves.
The total wave in the front tube traveling away from the sample can be obtained
in a similar way,
Areikz
X b ðzÞ ¼ Areikz ð1 þ rrs e2ihf þ ðrrs Þ2 e4ihf þ   Þ ¼ : ð2Þ
1  rs re2ihf
The total wave in the front tube is than

eikz þ reikz
X ðzÞ ¼ X f ðzÞ þ X b ðzÞ ¼ A : ð3Þ
1  rs re2ihf
756 K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765

Choosing z = 0 at the sample surface, the waves at sensor-1 and sensor-2 are

eih1;2 þ reih1;2
X 1;2 ¼ A : ð4Þ
1  rs re2ihf
The sound wave propagating towards the sample at the sample front surface is,
according to Eq. (3), given by ðA=ð1  rs re2ihf ÞÞ. The sound wave at the rear of the
sample is then ðA=ð1  rs re2ihf ÞÞt. Treating this term as the sound source in the back
tube, we obtain the wave in this tube as
  ikz 
A e þ rb eikz
X ðzÞ ¼ t: ð5Þ
1  rs re2ihf 1  rrb e2ihb

Here rb is the reflection coefficient of the end of the back tube, which has been
measured beforehand using the standard two-sensor method [2–5]. The factor
1=ð1  rrb e2ihb Þ is due to the multiple reflections in the back tube.
Taking z = 0 at the rear of the sample for the waves in the back tube, the signal at
the sensor-3 is then
  ih3 
A e þ rb eih3
X3 ¼ t: ð6Þ
1  rs re2ihf 1  rrb e2ihb
From Eq. (4) we obtain the reflection coefficient r of the sample through

eih2 þ reih2
H 1;2 ¼ X 2 =X 1 ¼ ; ð7Þ
eih1 þ reih1
which is the so called transfer function between the two sensors [2–5]. Note that Eq.
(7) no longer contains the reflection of the loudspeaker rs. Solving the equation, we
have the reflection of the sample as

eih2  H 1;2 eih1


r¼ : ð8Þ
H 1;2 eih1  eih2
Eq. (8) is the same as used in the standard two-microphone method to determine
the reflection r using the measured transfer function H1,2 [2–5].
The transmission coefficient t can then be obtained by X3/X2 once the value of r is
obtained. Using Eqs. (4) and (6), we have
 ih3 
ei#3 e þ rb eih3
X 3 =X 2 ¼ ih t; ð9Þ
e 2 þ reih2 1  rrb e2ihb

or
 
  1  rrb e2ihb
t ¼ ei#3 eih2 þ reih2 ðX 3 =X 2 Þ: ð10Þ
eih3 þ rb eih3

The transmission loss (TL) is then obtained directly from t


TLðdBÞ ¼ 20 logðtÞ: ð11Þ
K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765 757

The lower limit of transmission t obtained by measuring H2,3 = X3/X2 and using
Eq. (10) is mainly determined by how small the value of H2,3 one can still measure
with good accuracy, in which the high sensitivity of sensor-3 together with phase sen-
sitive signal detection plays an important role, in addition to good sound isolation.
The relative error of t is the same as the relative error of the measured H2,3. Using
Eq. (10) it is straightforward to show that for r close to 1 with absolute error of 0.01,
the relative error carried over to transmission is a few percent except when X2 and the
term ðe2ih þ reih
2 Þ is very small. An example of the dependence of the error in t on the
error in r is given in the Section 5 below.

4. Samples

Two perforated samples are fabricated as reference samples. One sample is a


30 cm by 30 cm 1 mm thick aluminum plate with 0.5 mm diameter clear holes at
10 holes/cm2. The other one is a 4 cm thick porous brick, also 30 cm by 30 cm in
area. For LRSM samples, rigid plastic square grids, which are almost totally trans-
parent to sound wave, are used as the frame. A grid is made of a square array of
1.5 cm · 1.5 cm · 2 cm cells. Each cell is surrounded by four 1.5 cm · 2 cm · 0.1 cm
hard plastic walls. A schematic sketch is shown in Fig. 2. Resonant units are made of
metal balls, which serve as the mass in a mass-spring oscillation system, surrounded
by silicone rubber serving as the spring. Each ball is placed in a cell of the grid with-
out direct contact with the grid, and the rest of the cell is filled with rubber. The ball
and the elastic modulus of the rubber determine the local resonance frequency. Typ-
ical dimensions of the samples are 30 cm · 30 cm · 2 cm. Four samples were fabri-
cated. Sample-A is a reference sample with soft rubber filled grid, but without

Grid

Sphere
Rubber

Top View of Panel

Side View of a Cell

Fig. 2. Schematics of the locally resonant sonic material samples.


758 K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765

metal balls. The transmission of Sample-A is expected to decrease smoothly with


increasing frequency, similar to the porous samples. Sample-B contains soft rubber
and a 5.7 g steel ball in each grid cell, and a total of 15 · 15 cells. The soft rubber and
steel ball form the local resonance unit, so the transmission amplitude of Sample-B is
expected to dip at its local resonance frequency. The mass of a steel ball is about 1/4
of the mass of the lead ball used in [10], but the elastic modulus of the rubber in Sam-
ple-B is lower. As a result, the local resonance frequency of Sample-B, which is pro-
portional to the square root of modulus over mass, should fall somewhere near
400 Hz reported in [10]. Sample-C contains the same type of rubber as Sample-B
and a 23.5 g lead ball, the same type used in [10], in each grid cell with dimension
of 3 cm · 3 cm · 2 cm, which is four times the area size of the ones in Sample-B.
There are a total of 8 · 8 cells in Sample-C. The transmission dip of Sample-C
should occur at lower frequency than that of Sample-B, since the mass is larger.
In fact, if the elastic modulus is nearly constant over the frequency range, then the
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dip frequency should be proportional to 1=mass. The last sample, Sample-D, is
the same as Sample-B except that another type of rubber with higher elastic modulus
is used. Its transmission dip should occur at higher frequency than that of Sample-B.
It should be noted that the average area mass densities of Sample-B, Sample-C, and
Sample-D are almost the same. According to mass law they would have about the
same transmission.
Finally, to test the detection limit of the method, a sample with double layers of
23.5 g lead balls coated with hard rubber and embedded in epoxy was fabricated.
The dimensions of the sample are 30 cm · 30 cm · 5 cm. This sample is similar to
the first LARM sample reported earlier [10], which has eight layers of balls in a sim-
ple cubic lattice structure, except that it is thinner in thickness and broader in planar
area.

5. Results

The relative response curves of the three sensors were first measured by using the
conventional switching position method [3–5]. The measurements were done with the
sensors in the front tube. The two B&K front sensors were found to have identical
response within the noise level, while the relative response between the B&K sensors
and the high-gain Sensor-3 was found to be flat in amplitude (with a factor of 110),
with a frequency dependent phase shift.
The frequency dependent complex rb was measured after the relative responses of
the sensors were obtained. Its amplitude is below 50% without sharp features in the
entire frequency range of interest. The same rb is used for all the samples in the sub-
sequent measurements.

5.1. Porous samples

The transmission without sample was tested first and the results (100% transmis-
sion and 0 reflection) are as expected.
K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765 759

100
8
6
4 Metal plate with holes

Transmission (%)
2

10
8
6
Porous Brick
4

1 2 3 4 5 6 7 8 9
1000
Frequency (Hz)

Fig. 3. The transmission amplitude spectra of two porous samples.

The transmission spectra of two porous samples are shown in Fig. 3. The trans-
mission amplitude of the holed Al plate drops gradually from 93% at 120 Hz to 30%
at 1500 Hz. The transmission amplitude of the brick is 10.5% at 120 Hz and drops
gradually to about 1% around 1200 Hz, following quite well the mass law. The trans-
mission spectra of both samples as measured from the apparatus resemble typical
ones as expected from these materials.

5.2. Error due to sensors mismatch

Sample-B was chosen to test how the mismatch of the two B&K front sensors
would affect the measurements, because being a LRSM sample, it is expected to
display unique features in the transmission spectrum that provides a suitable test
ground. The reflection and transmission of Sample-B were measured twice, the
second measurement being done when the positions of the two B&K sensors,
originally at position-1 and position-2, respectively, were switched, while sensor-
3 remained at position-3. The difference spectra between the two measurements
are shown in Fig. 4. The transmission dip at 675 Hz is the local resonance effect
of the ball-rubber units as expected according to the LRSM theory [10]. The
resulting difference of reflection amplitude is less than 0.002 over the 400–
1200 Hz frequency range, while the different of phase is less than 1, as seen in
Fig. 4(a). Below 400 Hz or above 1200 Hz, the error in amplitude increases line-
arly to 0.01 at 200 Hz and 1400 Hz. The relative error in transmission amplitude
Djtj/jtj is 4% at 200 Hz, and decreases to <1% at 400 Hz. The relative error in-
creases to above 1% again and reaches maximum around 760 Hz, the nodal point
of sensor-2 where the wave strength is minimum. The transmission phase differ-
ence is 5 at 200 Hz, and less than ±1 for frequencies >300 Hz. It is evident that
the errors in r and t caused by the mismatch of the two front sensors are well
within the practical tolerance.
760 K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765

(a) 2 4

Absolute Error (10 )


-2

Phase Error (˚)


1
2

0 1

0
-1
Amplitude Error -1
Phase Error
-2 2 3 4 5 6 7 8 9
-2
1000
Frequency (Hz)

(b) 5 3
Relative Error
4 2
Relative Error (%)

Phase Error (˚)


Phase Error

3 1

2 0

1 -1

02 3 4 5 6 7 8 9
-2
1000
Frequency (Hz)

Fig. 4. (a) The absolute error in the amplitude (solid curve, left axis) and phase (dashed curve, right axis)
of reflection coefficient of Sample-B, consisting of 5.7 g steel balls in soft rubber, due to the mismatch of
the two front sensors. (b) The relative error in the amplitude (solid curve, left axis) and absolute error of
phase (dashed curve, right axis) of transmission coefficient of Sample-B due to the mismatch of the two
front sensors.

5.3. Error in transmission due to possible error in reflection

It is a known fact that there is typically a loss of 2–3% in amplitude for acoustic
waves propagating in impedance tubes. As a result, the measured reflection of a rigid
sample is 97–98% [4], instead of the ideal value of 100%. Several rigid flat metal sam-
ples were tested on our setup, and the loss was found to vary from 1% in the low
frequency to 3% in the high frequency.
As has been discussed in the Section 3, according to Eq. (10) an error in reflection
r will cause an error in transmission t. The maximum error due to front sensors mis-
match is 0.01 for r. To test the effect, in calculating the transmission t of Sample-B
using Eq. (10) with experimentally measured X3/X2, an artificial random noise of
0.01 is added to the experimentally obtained reflection r. The resulting transmission
with and without the added noise are plotted in Fig. 5. For clarity, the curves with
noise have been upward shifted. It is seen in Fig. 5(a) that both the amplitude and the
phase of the two transmissions match well. The only region where the noise is signif-
K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765 761

(a) 10 250
8
6 |t|, no error in r
Phase, no error in r
4 200

Transmission (%)
|t|, error in r
phase, error in r

Phase (˚)
2
150
1
8
6 100
4

50
2

0.1 2 3 4 5 6 7 8 9
0
1000
Frequency (Hz)

(b) 0.2 20

0.0 15

Phase Error (˚)


Relative Error

-0.2 10

-0.4 5

-0.6 0
Amplitude
-0.8 -5
Phase
-1.0 2 3 4 5 6 7 8 9
-10
1000
Frequency (Hz)

Fig. 5. (a) The amplitude and phase of transmission coefficients of Sample-B with and without the
artificially added 0.01 random noise in reflection coefficient. For clarity, the curves with noise have been
upward shifted. (b) Relative error of amplitude and absolute error of phase of the transmission coefficient
caused by the 0.01 random noise in reflection coefficient of Sample-B.

icant is around 760 Hz, where both X2 and the term ðe2ih þ reih
2 Þ reach minimum.
There the added noise introduces a relative error nearly 30% in amplitude, and 10
degree in phase, as shown in Fig. 5(b). However, the overall shape of the transmis-
sion with the added noise remains unchanged, and the local resonance feature at
675 Hz is not affected.
We now compare the test results on the five samples using the method.

5.4. LRSM samples

Fig. 6(a) depicts the test results of Sample-A with rubber filled grid, but without
metal balls. The transmission amplitude jtj varies smoothly over the 300–1400 Hz
frequency range, decreasing from 10% to 0.5%, and the transmission phase varies
almost linearly. Critical points in frequency, at which the signal of either sensor-1 or
sensor-2 peaks very high or dips to very low level, are the nodal points of sensor-1
around 400 and 1200 Hz, the nodal point of sensor-2 around 760 Hz, and the front
762 K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765

(a) 100 50
4

Transmission (%)
2 0
10

Phase (˚)
4
|t| -50
2 Phase
1
-100
4
Soft Rubber
2

0.1 2 3 4 5 6 7 8 9
-150
1000
Frequency (Hz)
(b) 100 250
Large Pb balls in soft rubber
200
Transmission (%)

10

Phase (˚)
150
1
100

0.1
|t| 50
Phase
0.01 2 3 4 5 6 7 8 9
0
1000
Frequency (Hz)
(c) 100 100
11mm steel balls in hard rubber
50
Transmission (%)

10
0
Phase (˚)

1 -50

-100
0.1 |t|
Phase -150

0.01 2 3 4 5 6 7 8 9
-200
1000
Frequency (Hz)

(d) 1 Coated Balls in Epoxy 150


4
Transmission (%)

2 100
0.1
Phase (˚)

4
50
2

0.01
0
4
|t|
2 Phase
0.001 2 3 4 5 6 7 8 9
-50
1000
Frequency (Hz)

Fig. 6. The amplitude and phase of the transmission coefficient of (a) Sample-A, soft rubber only, (b)
Sample-C, 23.5 g lead balls in soft rubber, (c) Sample-D, 5.7 g steel balls in hard rubber, and (d) the sample
with rubber coated balls embedded in an epoxy/gypsum mixture.
K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765 763

tube resonance around 630 and 1260 Hz. No anomalous features near these frequen-
cies are present in the spectrum. The wide range of dynamic response of the sensors
together with the high precision phase sensitive signal detection technique plays a vi-
tal role in the precise measurement of signals near these critical points. The reflection
of the sample is almost a constant of 97% at frequencies >300 Hz. As is mentioned
above, this is close to the transmission loss limit of the standard method [2–5] be-
cause of the propagation loss of 3%.
Fig. 6(b) depicts the test results of Sample-C with heavier lead balls
(mass = 23.5 g) in the cells. The dip in transmission due to local resonance shifts
to 340 Hz, reaching a value of 0.08% (62 dB in TL) as compared to 0.8% of Sam-
ple-B. Both samples have the same area mass density, as the balls in Sample-C are
four times the mass of that in Sample-B while the cell area of Sample-C is also four
times of those in Sample-B. The effect of sound blocking efficiency of LRSM at low
frequency is clearly demonstrated here. The local resonance frequency is expected to
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
vary as 1=mass if the elastic modulus of the rubber remains the same over the fre-
quency range. This gives a ratio factor of 2 for the dip frequency of Sample-B and
Sample-C, which is close to the observed value (675 vs. 330 Hz).
Fig. 6(c) depicts the spectra of Sample-D with 5.7 g ball in hard rubber. Due to
higher elastic modulus and smaller loss modulus, the dip in jtj is deeper (0.015%
as compared to 0.2% for the same type of balls in soft rubber) and shifts to higher
frequency (970 Hz). The phase of transmission drops more than 135 within the fre-
quency range of the dip. The dip in jtj is followed by a peak at higher frequency of
1180 Hz with a peak value of 1.2%, with a corresponding rise of phase over 135. The
dip-peak feature in amplitude with corresponding down-and-up phase shifts is the
predicted characteristics of LRSMs [10]. The observed phase shift (135) is less
than the predicted value of 180 under ideal no-loss condition, because of the sizable
loss modulus of the rubbers. The local resonance features of Samples B and C are
similar to that of Sample-D. Due to even higher loss modulus, however, these fea-
tures are not as strong (dip not as deep, peak not as high, phase change smaller)
as the one in hard rubber.
That the LRSM samples can break the mass law is well exemplified by these
samples, which have almost the same area mass density. Since they have different
local resonance frequencies, a frequency range in the transmission spectra of the
samples not in resonance could serve as mass law reference for the one in resonance.
For example, around the local resonance frequency of 340 Hz the transmission of
Sample-C (Fig. 6(b)) reaches minimum of 0.1%, while for Samples B and D, which
are off resonance, the transmissions are around 1%. Similarly, at 980 Hz where
Sample-D is in local resonance its transmission is 0.013%, while that of Sample-
D is 0.3%. In both cases transmissions are at least 10 times smaller than dictated
by the mass law.

5.5. Limit of minimum transmission

Having demonstrated that the setup is able to measure correctly the unique
transmission of LRSM samples, we now test the low frequency sensitivity limit
764 K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765

of the setup with a LRSM panel made of rubber coated lead balls imbedded in an
epoxy/gypsum mixture. The panel contains two layers of balls in a body-cubic cen-
tered lattice structure with 2 cm distance between ball centers. Because of the high
density of the local resonators (the rubber coated balls) the resonance dip is ex-
pected to be deeper. Fig. 6(d) shows the transmission spectra jtj where two pairs
of transmission dip and peak are clearly displayed. As reported earlier [10], the
low frequency resonance is due to the local resonant vibration of the balls within
the rubber coating, while that at the higher frequency is due to the standing wave
in the rubber coating. The transmission dip at 632 Hz is 0.0015% (96.5 dB in TL)
while the other dip at 250 Hz is 0.005% (86 dB in TL). The phase spectrum first
displays a downward shift of nearly 180 at these two resonance frequencies, then
remains nearly flat, until the frequency reaches the value where the transmission
peaks and the phase jumps upward by 180. Such phase change in response to
the transmission dip and peak is exactly what is predicted by the LRSM theory
[10]. The fact that such phase change is clearly observed even when the transmis-
sion loss is as high as 96.5 dB is strong support to the confidence of the transmis-
sion spectrum. Had the transmission spectra been corrupted by imperfection of the
system such as leakage, then the phase would not have undergone the observed
180 downward shift.
We now discuss the limitation of the impedance tube method. It is well known
that finite size of the samples affects the measurement results because different sizes
bring different boundary conditions to the vibrating sample. Therefore, in principle
the method is only applicable to the samples with weak size dependence. Although
one can verify the results by using different clamping positions (near the edge of
the sample, or close to the impedance tubes) to mimic the difference sizes, the
structures and compositions of the composite materials under testing are also
important. One class of materials expected to have little size dependence and there-
fore suitable for the impedance tube method are the composites containing at least
two materials with drastically different elasticity coefficients, i.e., one hard and the
other soft, and the sound waves can pass through the sample by propagating
through the soft material only. In the perforated samples the air in the empty
channels is the soft component and the solid aluminum/brick skeleton is the hard
component. Sound waves can pass through the samples through the air channels,
while the skeleton remains mostly undisturbed. The acoustic property of the sam-
ple is then determined mainly by the air channels where viscosity causes dissipation
of the acoustic energy, while there is little or no lateral propagation of sound
waves. The lateral size of the samples or the way they are clamped is therefore
unimportant. In the LRSM samples the soft component is the rubber filled cells
and the hard component is the rigid plastic grids. Most of the sound waves pass
through the cells and leave the rigid grids mostly undisturbed. Again, little lateral
propagation of sound waves is expected. The rigid grids effectively block the vibra-
tion of one cell to the others, so each cell reacts to the sound waves as if it was
isolated from the others. As our results presented above show, for these locally
reacting materials the overall size of the sample becomes unimportant, so they
are well suited for the impedance tube method.
K.M. Ho et al. / Applied Acoustics 66 (2005) 751–765 765

6. Conclusion

In summary, we have developed, on the basis of the standard ITM, a two-imped-


ance tube three-microphone method to measure the transmission coefficient of high
transmission loss acoustic samples. Both the amplitude and the phase of the trans-
mission coefficient can be obtained with high accuracy, even when the transmission
amplitude is as low as 105, or as high as 1. The effectiveness of the method has been
successfully demonstrated on a series of conventional and LRSM samples. The con-
fidence of the results is strengthened by the distinct phase change at the resonance
frequency in the corresponding phase spectra of the LRSM samples being tested.
The scanning frequency approach allows one to concentrate on a particular fre-
quency range where the transmission is low, thus getting high quality spectra over
the entire frequency range of interest.

Acknowledgements

This work was supported by the Hong Kong Special Administrative Region Gov-
ernment Research Grant Council Grant HKUST6143/00P. We thank C.T. Chan for
valuable discussions.

References

[1] ASTM C384-98, Standard test method for impedance and absorption of acoustical materials by the
impedance tube method.
[2] ASTM E1050-90, Standard Test Method for Impedance and Absorption of Acoustical Materials
Using a Tube, Two Microphones, and a Digital Frequency Analysis System.
[3] Rue Y. Test procedure and user interface for determination of the acoustic properties of materials
using the two-microphone transfer function method. Build Acoust 2000;9(1):73–9.
[4] Katz BFG. Method to resolve microphone and sample location errors in the two-microphone duct
measurement method. J Acoust Soc Am 2000;108(5):2231–7.
[5] Chu WT. Transfer function technique for impedance and absorption measurements in an impedance
tube using a single microphone. J Acoust Soc Am 1986;80:555–60.
[6] Amelin Celse K, van Champoux Y, Berry Alain. Acoustical characterization of absorbing porous
materials through transmission measurements in a free field. J Acoust Soc Am 1997;102:1982–94.
[7] Selamet A, Radavich PM. The effect of length on the acoustic attenuation performance of concentric
expansion chambers: an analytical, computational and experimental investigation. J Sound Vibrat
1997;201:407–26.
[8] Chung JY, Blaser DA. Transfer function method of measuring in-duct acoustic properties. I. Theory.
J Acoust Soc Am 1980;68:907.
[9] Selamet A, Ji ZL. Acoustic attenuation performance of circular expansion chamber with extended
inlet/outlet. Noise Control Eng J 2000;48(2):60.
[10] Liu ZY, Zhang XX, Mao Y, Zhu YY, Yang Z, Chan CT et al. Locally resonant sonic materials.
Science 2000;289:1734.
[11] See, e.g. Brekhovskikh LM. Waves in layered media. 2nd ed. New York: Academic Press; 1980.

You might also like