Professional Documents
Culture Documents
Laboratory of Ocean Acoustics and Sensing, School of Marine Science and Technology, Northwestern
Jianhua Yang
Abstract
This study presents a method for measuring the normal incident sound absorption coefficients of
acoustic materials by extracting the first reflected wave in a short pulse tube. Instead of generating a
short pulse in a tube with an open end, the inverse filter method is used when a sample is added at the
end of the tube, which makes it possible to extract the first reflected wave directly from the superposed
sound field in the time domain. The length of the tube used in this proposed pulse separation method is
half as long as that of the traditional pulse separation method, which makes it can be used in a general
commercial impedance tube. Two different materials are investigated. When the proposed method is
compared to the traditional transfer function method and the traditional pulse separation method, good
Keywords: sound absorption measurement; superposed sound field; first reflected wave; pulse
separation
1 1. Introduction
2 The normal incident sound absorption coefficient is an important indicator to characterize acoustic
3 materials. A multitude of studies have involved the measurement of the normal incident sound absorption
4 in an impedance tube, and these methods can be divided into two types: the transfer function methods
6 Chun and Blaser1,2 used two flush-mount microphones on a tube to decompose the wave field in
7 terms of the two directions of propagation and obtain acoustic parameters from their ratio, which is the
8 widely used transfer function method (which can be used to validate a model or algorithm or a designed
9 acoustic structure3-5, and also used as a foundation of the measurement of different acoustic parameters6,7).
10 However, there are still limitations to this traditional method8-11: first, the phase and magnitude
11 mismatches that occur between the two sensors employed in this method can influence the accuracy of
12 the test results, and second, the test frequency range of this method is related to the microphone spacing,
14 In fact, if the duration of the test signal is short enough, the incident and first reflected wave in the
15 tube (i.e., the first two pulses in the time domain) can be separated in the time domain directly and can
16 be used to calculate the absorption coefficient, which is the so called pulse separation method14-19.
17 Experimental measurements of acoustic impulse responses have long been published12,13, which can be
18 used for the pulse separation method. In 2003, R. D. Stevens et al.14 used a maximum-length sequence
19 to yield an accurate estimate of the system impulse response of a pulse tube system and then used the
20 separated incident pulse and the first reflected pulse from the surface of the sample to calculate the
21 complex reflection coefficient of a sample. However, the duration of the pulse response of a loudspeaker
22 is very long (approximately 10 ms), and the tube therefore must be long enough (7.5 m) for the separation
23 of the incident and first reflected waves, which is not convenient for measurement. R. Ramakrishnan et
24 al.15 used the inverse filter method to generate a short pulse in a tube by eliminating the pulse response
25 of the generator system. However, a very long tube (approximately 2.18 m) is still needed to edit out the
26 pipe end reflection. A similar method is described in Ref. 16. In 2006, Jing et al.17 considered the
27 influence of the pipe end reflection and generated a pure short pulse in a tube with an open end. Sun et
28 al.18,19 employed this pulse generation technique to measure the normal incident sound absorption
29 coefficient by separating the first two pulses in the time domain. Thus, the step of editing out the pipe
31 The transfer function method decomposes the sound field in the tube into two terms: the positive
32 wave and the negative wave. Meanwhile, the pulse separation method just uses the first two pulses in the
33 time domain to calculate the reflection coefficient, and both limitations of the transfer function method
34 can be avoided. However, to capture the first two pulses in the time domain, the tube used for the pulse
35 separation method is usually not too short, and at least one microphone needs to be mounted at the middle
36 of the tube14 (some methods15,16,18,19 require another microphone close to the loudspeaker to eliminate
37 the response of the generator system), which means that a special tube still needs to be made for
38 measurement using the pulse separation method (the general commercial impedance tubes are always
40 The work presented here basically lies within the system established by Sun et al. 18,19. In this paper,
41 the length of the tube used for pulse separation method18,19 is further shortened and the method for
42 extracting the first reflected wave from the superposed sound field is proposed. First, the theoretical
43 derivation of the pulse generation method is given, the pulse separation method is discussed for different
44 situations (different tube lengths and sensor spacings). Then, the shortest tube length for the measurement
45 of the incident absorption coefficient using the pulse method is presented, the first reflected wave is
46 extracted, and the normal incident sound absorption coefficient can be calculated by the ratio of the
47 extracted reflected wave and the incident wave. Several corresponding experiments are performed to
48 verify the proposed method. Experimental results for two different materials in this paper agree well with
51 The scheme of the measurement system used in this study is shown in Fig. 1, and this system
52 includes an impedance tube, a B&K2716C power amplifier, a B&K3560C data acquisition system
53 controlled by a computer, and a 1/4 in MPA416 condenser microphone used to record the data at two
54 positions. The input (AI) and output (AO) channels of the B&K3560C board are programmed to be
56
57 The principle of pulse generation in a tube for the pulse separation method, which is introduced in
58 Refs. 18-19, usually consists of two steps: first, a designed broadband signal (i.e., hi in Fig. 1(a)) is fed
59 into the system without a sample added at the tube end, and the acoustic response signal (i.e., hr(xa) in
60 Fig. 1(a)) is recorded by a microphone located close to the sound source (position a in Fig. 1(a)). The
61 impulse response of the sound generation system H(xa, ω) can thus be obtained by H(xa,ω) = Hr(xa,ω)/Hi
62 (ω), where Hi(ω) and Hr(xa, ω) are the Fourier transforms of the input excitation (also the designed
63 broadband pulse) hi and the output signal hr(xa), respectively. Second, the spectrum of the driving signal
What is the difference between input exitation and driving signal? Explain!
64 Hd(ω) can be deduced by the ratio of the spectrum of the desired pulse signal Hi(ω) to the resolved
65 impulse response H(xa, ω) as Hd (ω) = Hi(ω)/H(xa, ω). Applying an inverse Fourier transform to Hd(ω),
67 The impulse response H(xa, ω) corresponds to the frequency response of the generator as well as
68 the microphone position in the tube. In fact, the wave field of the tube can be decomposed in terms of
69 the forward and backward propagation waves7; thus, Hr(xa, ω) can be written as follows:
H i H g
70 H r xa , e jkxa +rt e jk 2l xa (1)
1 rt rs e jk 2l
72 the distance between the surface of the loudspeaker to the left surface of the sample (as shown in Fig.
73 1(a)), and rs and rt are the complex reflection coefficients at the surface of the sound source and at the
74 surface of the end, respectively. It should be noted that the frequency response at position xa (caused by
75 the wave propagation process in the tube) can also be viewed as a linear system, and its frequency
e jkxa +rt e
jk 2 l xa
77 H p xa , (2)
1 rt rs e jk 2l
78 Consequently, Hd(ω) can also be written as Hd(ω)= Hi(ω)/(Hg(ω)∙ Hp(xa, ω)). It is clear that a single
79 desired pulse is generated at position a when hd is fed into the system (i.e., substituting Hd(ω) in Eq. (1)
80 for Hi(ω)).
83 The source is excited using the calculated driving signal hd, the single microphone is moved to another
84 position to test the sound pressure (i.e., hr’(xb) in Fig. 1(a)), and the frequency response can be expressed
85 as
H r xb , H d H g H p xb ,
H p xb ,
H i
H p xa ,
86 jk 2 l xb (3)
e jkxb rm e 1 rs rt e jk 2l
=H i jk 2 l xa
e jkxa rt e 1 rs rm e jk 2l
H i e
jk xb xa jk 2 l xb xa jk 2 l xb 3 xa
rm e rt e
87 where rm is the complex reflection coefficient at the left surface of the sample. Eq. (3) shows that although
88 the output signal at position xb contains many terms, the first and second terms correspond to the incident
89 wave and the first reflected wave in the tube, respectively, and these terms can be used to calculate the
90 absorption coefficient of the sample when the length of the desired pulse and the two microphone
pr 2
92 rm , am 1 rm (4)
pi
93 where pr and pi denote the first reflected wave and the incident wave, respectively, and am is the
96 It is clear that the distance between the sensor (located at xb) and the surface of the specimen should
97 be large enough to separate the incident pulse and first reflected pulse. However, it is not enough to
98 directly crop the first reflected pulse in the time domain, as this may cause the third term in Eq. (3) to be
99 superposed on the first reflected wave (i.e., the second term in Eq. (3)). Consequently, the sufficient and
100 necessary condition to separate the incident wave and the first reflected wave from the surface of the
101 sample is that there is no superposition among the first three terms in Eq. (3). Assuming that the length
102 of the desired pulse (i.e., hi) is τ in the time domain, this condition can be expressed by the following two
103 equations:
104 2l c , 2s c (5)
105 where c is the speed of sound in air, while l’ = l−xb and s = xb−xa denote the distance from position b to
106 the surface of the sample and the distance between the two microphone positions, respectively.
107 In this situation (which was introduced in Ref. 18 and 19.), the incident wave and the first reflected
108 wave can be separated directly in the time domain using a rectangular window18, and the attenuation
111 In the first step of generating the desired signal, the reflection coefficient rt is not known, and when
112 2s’≥ τ∙c is not established, the exact reflection wave cannot be obtained. Thus, the premise to measure
113 the attenuation coefficient using the pulse method is that the first reflected wave (i.e., the second term in
114 Eq. (3)) and the third term in Eq. (3) are not superposed. However, when the incident wave and the first
115 reflected wave are superposed (i.e., the 2l’ ≥ τ∙c is not established, the position b is close to the surface
116 of the sample) because the waveform of the incident wave can be controlled and is just equal to the
117 desired signal and after the process of generating the desired signal, one can still obtain the first reflected
118 wave by subtracting the incident wave from the superimposed recorded sound pressure as follows:
H r xb , H r xb , H i e jks
119 (6)
H i rm e rt e
jk 2 l 3 s
jk 2 l s
120 The attenuation coefficient can be calculated using Eq. (4). In this situation, since position b is close
121 to the surface of the sample, the total length of the pulse tube can be shorter on the basis of Ref. 18.
123 In fact, with an unknown reflection coefficient rt, the end of the tube can be arbitrary (Sun et al.18
124 used an open end in the first step of pulse generation). If the end of the tube is the tested sample, i.e., the
125 first step of pulse generation is performed with the sample added at the end of the tube, then, rt= rm, and
jk xb xa jk 2 l xa xb xa
e rm e
H r xb , H i
1 rm 1 e
jk 2 l xa
127 (7)
= H i e jks
rm e jk 2 l s
rm e jk 2 l s 2 jk 4 l s
rm e
128 where s=xb xa and l’’=l xa. (i.e., s is the spacing between position a and position b; l’’ is the distance
129 from position a to the left surface of sample.) When 2l’ τ∙c is not established (the same as situation 2),
130 after subtracting the incident wave, one can obtain the following:
H r xb , H r xb , H i e jks
131 (8)
H i rm e
jk 2 l s jk 2 l s jk 4 l s
rm e rm 2 e
132 If the microphone spacing is small and 2s τ∙c is also not established, then the first reflected wave
133 is still superposed with the next term; however, it can be seen from Eq. (7) that the factor of the third
136 Although some new terms are introduced in Eq. (9), the delay between the second term and the first
137 term (i.e., the first reflected wave) is doubled, which makes it more convenient to crop the first reflected
138 wave. One can repeat this step if the time delay is still not enough to separate the two terms, and Eq. (10)
141 where n (n = 1, 2, 3 ⋯) is the number of iterations. It should be noted that the third term in Eq. (10)
142 cannot be eliminated; thus, for the short pulse tube method, the limit of the microphone spacing should
144 In this situation, the only size limitation is that position a and the surface of the tested sample should
145 be greater than half of the distance a sound wave travels in the time duration of the incident wave, which
146 means that the total length of the tube used for the pulse separation method can be shorter and the
147 microphone spacing can be smaller than situation 2. Thus, the superposed first reflected wave can be
148 extracted and it is possible to use the proposed pulse separation method in a commercial impedance tube.
149 4. Experiment and results
151 For the convenience of experimental verification, a homemade variable-depth (changed by moving
152 the piston) impedance tube (shown in Fig. 2) with several microphone positions is used in this experiment.
153 The downstream tube is detachable (when the downstream tube is removed, there is an open end), and
154 the sample is mounted at the rigid end of the downstream tube. Microphone position 1 (as shown in Fig.
155 2(b)) is used for signal generation (with an open end for the traditional pulse separation method; with a
156 downstream tube and a sample for the proposed short pulse tube method) in this experiment. The desired
157 impulse signal used in this experiment is a Butterworth pulse18 (as shown in Fig. 3(a)). A rectangular
158 window is used to cut the separated incident wave and the first reflected wave in the time domain, and
159 the captured pulses are zero-padded before FFT to calculate the absorption coefficient. Two different
160 polyurethane foam materials, the A1 sample and A2 sample, are used in this experiment. Taking the A1
161 sample as an example, the processing of the traditional pulse separation method and that of the proposed
(b)
FIG. 2. (Colour online) The entire measurement system: (a) photograph of the measurement system and
164 For situation 1, the traditional pulse separation method proposed by Sun et al.18,19, the depth of the
165 downstream tube is ld=180 mm, and position 4 is used for coefficient measurement (i.e., s =180 mm, l’
166 =215 mm). When the calculated driving signal is fed into the system, the responses at position 1 (with
167 an open end) and position 4 (with the downstream tube and A1 sample) are recorded separately. From
168 Fig. 3(a), one can see that because the frequency response of the loudspeaker system and that of the pulse
169 propagation in the open-ended tube are offset, the response at position 1 is very close to the single desired
170 pulse. Since the tube system is changed after adding a sample, the response at position 4 (shown in Fig.
171 3(b)) contains a series of pulses, and there is a one-to-one correspondence between these pulses and the
172 terms in Eq. (3). It is clear that the first reflected wave can be separated from both the incident wave and
173 the third pulse when the appropriate position spacing is selected (i.e., Eq. (5) is satisfied).
(a) (b)
175 0 mm, and position 5 is used (i.e., s = 180 mm, l’ = 35 mm). The measured sound pressure at position 5
176 is plotted in Fig. 4(a). According to Fig. 4(a), the incident wave and the first reflected wave are
177 superposed, but the third pulse is not superposed upon the two pulses, and then, the reflected wave can
178 be obtained by Eq. (7). Fig. 5(b) shows the process of deducing the first reflected wave in this situation:
179 first, the measured desired pulse (recorded at position 1 with an open end) is added to a time delay of
180 2s/c, and then, the desired pulse coincides with the first incident wave recorded at position 5. The
181 reflected wave can then be separated directly in the time domain after subtracting the two signals.
(a)
(b) (c)
FIG. 4. (Colour online) (a) The measured responses at positions 1 and 5, respectively, in situation 2. (b)
Adding a certain time delay (i.e., 2s/c) to the sound pressure recorded at position 1. (c) Subtracting the
182 For the general small commercial impedance tube (whose diameter is approximately 30 mm), the
183 microphone spacing is usually approximately 20 mm for a wide test frequency, and thus, both Suns’
184 method and situation 2 (when position b is close to the surface of the sample) may not apply to the
185 commercial tube. Then, a small microphone spacing should be considered (i.e., situation 3, the proposed
187 For the proposed short pulse tube method, position 2 is used, and the piston is pushed into the
188 upstream tube until the front surface of the sample is just at position 4; then, s = 20 mm and l’ = 160 mm,
189 and Eq. (5) is not satisfied. Fig. 5(a) shows the response at position 2, and it is clear from Fig. 5(a) that
190 the first reflected wave is superposed by the third pulse. As shown in Fig. 5(b), the time domain signal
191 corresponding to Hr’’(xb, ω) can be obtained by subtracting the incident wave. In addition, one can easily
192 obtain the first reflected wave (i.e., the first term in Eq. (9)) using Eq. (9). Fig. 5(c) shows the wave form
193 of Hr’’’(xb, ω). However, it can be seen from Fig. 5(c) that the first reflected wave is still superposed by
194 the next wave, and thus, to extract the first reflected wave, one should repeat this step at least three times
195 using Eq. (10) (the second term of Eq. (10) will be -rmexp(-jk0(2l’’+15s)). The spacing of positions 1 and
196 2 is 20 mm, and the time delay between the first reflected wave and the next wave is approximately 1
197 ms. As shown in Fig. 5(d), the first reflected wave can be separated directly in the time domain after
(a) (b)
(c) (d)
FIG. 5. (Colour online) (a) The measured response at position 2 for the proposed short tube pulse
separation method (i.e., situation 3). (b) Subtracting the incident wave from the measured response at
position 2 (solid line is the subtraction result) and adding a certain time delay (i.e., 2s/c) to the subtraction
result (shown as the dashed line). (c) Wave form corresponding to Eq. (9). (d) Repeating this step three
times (corresponding to Eq. (10)) and obtaining the first reflected wave.
200 Fig. 6 shows the spectrum of the incident wave as well as the spectra of the reflected waves obtained
201 in different situations (situations 1, 2 and 3). From Fig. 6 one can see that the spectra of the pair of the
202 incident and reflected waves in the three situations almost overlap, which means that the proposed short
203 pulse separation method is available. After obtaining the incident wave and the first reflected wave from
204 the surface of the sample, the complex reflection coefficient and the absorption coefficient can be
205 calculated by Eq. (4).
(a) (b)
FIG. 6. (a) and (b) show the spectra of the incident waves and those of the first reflected waves (from
the front surface of sample A) for different pulse separation situations (i.e., situations 1, 2 and 3).
206 The calculated absorption coefficients for samples A and B are shown in Fig. 7 using the traditional
207 pulse separation method (i.e., situation 1) and two other pulse methods (including situations 2 and 3 in
208 Sec. 3). The results for the traditional transfer function method (positions 3 and 4, two microphones,
209 random excitation, a Hanning window and 150 average are used) are also given. As shown in Fig. 7, the
210 general trends of the four methods are similar. However, the results for the transfer function method at
211 low frequencies show obvious fluctuations, which is caused by the spacing of the two microphones and
212 the accuracy of the analysis system10,11. For the results of the three pulse separation methods, there is
213 little fluctuation in the low frequency range, and the three resulting curves almost coincide with each
214 other.
(a) (b)
FIG. 7. The calculated results using different methods: (a) the results for sample A and (b) the results
for sample B.
215 For the pulse separation method (proposed by Sun et al.18,19), the premise is that both the microphone
216 spacing and the distance from the second microphone to the surface of the sample should be greater than
217 half of the distance a sound wave travels in the time duration of the incident wave, and this premise
218 makes the tube not very short (the length of the tube introduced in Ref. 18 is at least 400 mm). According
219 to the above analysis (Section 3), the length of the tube can be as short as 170 mm (for an incident wave
220 with a 1 ms duration and assuming the speed of sound in air is 340 m/s) for the proposed short tube pulse
221 separation method, which makes the measurement more convenient and makes it possible to use the
223 Fig. 8 shows the results measured by a B&K 4206 impedance tube (as shown in Fig. 8(a)) using the
224 traditional transfer function method and the proposed short tube pulse separation method. Good
225 agreement between the two methods can be observed from Fig. 8(b) and (d); the mean deviations between
226 the two methods are 1.05 % and 0.86 % for the two materials from 300 Hz to 6400 Hz, which means that
227 the proposed short tube pulse separation method is useful at least in the frequency range from 300 Hz to
228 6400 Hz. Additionally, the resulting curves obtained by the pulse separation method are smoother and
229 have less fluctuations in the low frequency range than those from the transfer function method (as shown
(a)
- It is interesting to say what the materials are and their thicknesses.
- Was a repeatability test or statistical evaluation of the dispersion of the results performed?
(b) (c)
(d) (e)
FIG. 8. The calculated results using different methods in a commercial impedance tube. (a) Photograph
of the B&K 4206 impedance tube. (b) The results for sample A. (c) The results for sample A (from 0
Hz to 500 Hz). (d) The results for sample B. (e) The results for sample B (from 0 Hz to 500 Hz).
231 5. Conclusions
232 This paper presents a pulse method to measure the incident sound absorption coefficient by
233 extracting the first reflected wave in a short tube. A short Butterworth pulse is generated in the tube when
234 the tested sample is added at the end (instead of an open end). Then, the first reflected wave can be
235 deduced from the superposed waveform, and the incident sound absorption coefficient can be obtained
236 from the first reflected wave and incident wave. Compared to the traditional pulse separation method as
237 well as the transfer function method, there are no limitations regarding the microphone spacing using
238 this method (the traditional pulse separation method needs a large microphone spacing for separation of
239 the incident and reflected waves; the transfer function method needs different microphone spacings for
240 different frequency ranges of interest). The only microphone position limit for this method is that the
241 spacing between the first microphone position (close to the sound source) and the surface of the tested
242 sample (added at the end of the tube) should be greater than half of the distance a sound wave travels in
243 the time duration of the incident wave. Thus, the tube used for the proposed pulse separation method can
244 be very short (approximately 170 mm when the duration of the incident wave is 1 ms), which makes the
245 measurement more convenient. The proposed method can also be used in a general commercial tube. The
246 results obtained with this proposed method are in good agreement with those obtained using the transfer
247 function method and the traditional pulse separation method, and show better quality in the low frequency
248 range.
249 Acknowledgments
250 This work was financially supported by the National Key Research and Development Program
251 of China (Grant No. 2016YFF0200902), National Natural Science Foundation of China (Grant No.
252 11474230 and 11704314), the China Postdoctoral Science Foundation (Grant No.2018M631194)
253 and Foundation of Shaanxi Educational Committee (Grant No. 2018BSHYDZZ03). We thank the
254 reviewers for their suggestions, which greatly helped us to improve this paper.
255 References
1
256 J. Y. Chung and D. A. Blaser, “Transfer function method of measuring in-duct acoustic properties.
2
258 J. Y. Chung and D. A. Blaser, “Transfer function method of measuring in-duct acoustic properties.
3
260 J. Zhao, M. Bao, X. Wang, H. Lee and S. Sakamoto, “An equivalent fluid model based finite-
261 difference timedomain algorithm for sound propagation in porous material with rigid frame,” J.
4
263 B. -S. Kim, J. Park, “Double resonant porous structure backed by air cavity for low frequency
5
265 P. H. Mareze, E. Brandão, W. D’A. Fonseca, O. M. Silva and A. Lenzi, “Modeling of acoustic
266 porous material absorber using rigid multiple micro-ducts network: Validation of the proposed
6
268 B. -S. Kim, Y. Seong, J. Park, “Modified two-thickness method for measurement of the acoustic
269 properties of porous materials,” Appl. Acoust. 146, 184-189 (2019).
7
270 Y. Salissou and R. Panneton, “A general wave decomposition formula for the measurement of
271 normal incidence sound transmission loss in impedance tube,” J. Acoust. Soc. Am. 125, 2083-2090
272 (2009).
8
273 Hans Bodén and Mats Åbom, “Influence of errors on the two-microphone method for measuring
274 acoustic properties in ducts,” J. Acoust. Soc. Am. 79, 541-549 (1986).
9
275 W. T. Chu, “Extension of the two-microphone transfer function method for impedance tube
10
277 “Acoustics—Determination of sound absorption coefficient and impedance in impedance tubes—
278 Part2: Transfer-function method,” ISO Standard 10534-2:1998 (International Organization for
11
280 “Standard test method for impedance and absorption of acoustical materials using a tube, two
281 microphones, and a digital frequency analysis system,” ASTM Standard No. E1050-98 (ASTM
12
283 D. D. Rife and J. Vanderkooy, “Transfer-function measurement with maximum-length sequences,”
13
285 A. Farina, “Simultaneous measurement of impulse response and distortion with a swept-sine
286 technique,” 108th Audio Engineering Society Convention, Paris, France, (2000).
14
287 R. D. Stevens and J. Vanderkooy, “A Novel Single-Microphone Method of Measuring Acoustical
288 Impedance in a Tube,” 115th Audio Engineering Society Convention, New York, USA, (2003).
15
289 R. Ramakrishnan, M. Salikuddin, K. K. Ahuja, “Generation of desired signals from acoustic
16
291 K. Y. Fung and X. Jing, “Characterization of impedance boundary as damped harmonic oscillators
292 via impulse reflection,” J. Acoust. Soc. Am. 119, 3831-3838 (2006).
17
293 X. Jing and K. Y. Fung, “Generation of desired sound pulses,” J. Sound. Vib. 297, 616-626 (2006).
18
294 L. Sun, H. Hou, D. L.-Y. Dong, and F.-R. Wan, “Sound Absorption Measurement in a Circular
295 Tube Using the Echo-Pulse Method,” Acta. Acust united. Ac. 96, 973-976 (2010).
19
296 L. Sun and H. Hou, “Measurement of sound absorption by underwater acousitc material using