Professional Documents
Culture Documents
excitation technique
A. F. Seybert, and D. F. Ross
Citation: The Journal of the Acoustical Society of America 61, 1362 (1977); doi: 10.1121/1.381403
View online: https://doi.org/10.1121/1.381403
View Table of Contents: https://asa.scitation.org/toc/jas/61/5
Published by the Acoustical Society of America
Influence of errors on the two-microphone method for measuring acoustic properties in ducts
The Journal of the Acoustical Society of America 79, 541 (1986); https://doi.org/10.1121/1.393542
Transfer function technique for impedance and absorption measurements in an impedance tube using a
single microphone
The Journal of the Acoustical Society of America 80, 555 (1986); https://doi.org/10.1121/1.394050
On the multiple microphone method for measuring in-duct acoustic properties in the presence of mean flow
The Journal of the Acoustical Society of America 103, 1520 (1998); https://doi.org/10.1121/1.421289
D. F. Ross
LIST OF SYMBOLS
•(t) amplitude
ofincident
wave estimate of the auto-spectral density of b(t)
•_(t) amplitude
ofreflected
wave estimate of the cross-spectral density be-
A Fourier transform of incidentwave ampli- tween•(t) and•(t)
tude •AB• (•AB real andimaginaryparts{of •B, respec-
Fourier transform of reflected wave ampli- tively. _
tude
estimateof theauto-spectral
densityof•
c speed of sound in still air estimateof the auto-spectraldensityof •.
cO speed of wave propagation in presence of estimate Of the cross-spectral density be-
flow
tween• and•.
frequency realandimaginary
partsof•2, respectively
wave number without flow t time coordinate
wave number in direction of flow T total length of data record
wave number in opposite direction of flow TL • transmission loss
tube length incident wave particle velocity
mean flow Mach number reflected wave particle velocity
incident wave pressure total acoustic particle velocity
reflected wave pressure Fourier transform of total acoustic particle
total acoustic pressure . velocity at x= %
Fourier transform of total acoustic pres- mean flow velocity
sure at x = xj spatial coordinate
R complex pressure reflection coefficient position of first microphone
estimate of the resistive impedance position of second microphone
estimate of the auto-spectral density of • at estimate of the impedance
x=0 estimate of the power reflection coefficient
estimate of the auto-spectral density of •,at estimate of the phase angie at x = 0
x=0 wavelength
estimate of the cross-spectral density be- p air density
tween• and • at x = 0 estimate of the reactive impedance
estimate of the auto-spectral density of a(t) angular frequency
1362 J.Acoust.
Soc.Am.,Vol.61, No.5, May1977 Copyright
¸ 1977bytheAcoustical
Society
of America 1362
1363 A.F. Seybert an•l D. F. Ross:Two-microphonemethod for determining acousticproperties 1363
_...
• v • •i(x,
t) • Zn(f)
--•
• FIG. 1. Tube with unknown termination
impedanceZn(f) excited by randomly vi-
-- •r(X,t)•/ brating piston.
i-andommotion can be regarded as harmonic motion of The impedance Z,(f) can be regarded as a linear
frequency f with a randomly varying amplitude. For system that relates the total pressure and velocity at
plane-wave propagation the incident and reflected waves x= 0, represented by the block diagram in Fig. 2.
are given by From linear theory?it canbe shownthat Z,(f) is esti-
mated by
•,(x, t) =•(t)e'½•t'•x),
•r(x,t)=•(t)e•(•f*•x), •.(f ): •,(f )/•.f ), (4)
where where S•(f) is an estimate of the cross-spectral den-
sity between the pressure and particle velocity at x = 0
co=2•r/, and•..(f) is anestimate
oftheauto-spectral
density
of
the particle velocity at x = 0. These spectra A•e esti-
/•: w/co=acousticwavenumber,
mated by
co= speed of propagation,
and •(t) and b(t) are stationary random variables. The
randomamplitudes•(t) and$(t) are not functionsof dis-
rance x since the system is nondispersive; therefore,
The auto-spectraof the pressureat • = O, S•(f), is
the wave shape does not change during propagation. estimated by
The influence of flow can be included by noting that
co: c + v for wave motion with flow,
InEqs.(5a),(5b),•d (5c),•o(4 T)•d •0(f,T)arethe
co:c- v for wave motion against flow, finite Fourier tr•sforms of the pressure andparticle
velocity time series at • = 0, respectively,
where c is the speed of propagation with zero flow. For
flow in the + x direction, the incident and reflected wave
numbers are given by Po(Z•:? •(0,t)e"•'dt, (6a)
- ei•t
•(0,t)
I
Oo(f) Zn(f) -•o(f)
•(0, t) =[•(t) - b(t)] pc . (3) FIG. 2. Block diagram showingrelationshipbetweenacoustic
impedance,pressure, and particle velocity. '
(Sa) spectral
density.Thequantities
•(f, T)and•(f, T)are
the finite Fourier transforms of the pressure time series
at points 1 and 2. By taking the Fourier transform of
(8c) Eqs. (11) and combiningwith Eqs. (12), the following
relationships are developed:
•(f) and}ss(f) aretheauto-spectral
densities
ofthe
incident
andreflectedwaves,respectively;
•,•s(f) and •::(f)=}•(f) +}ss(f)+2[&,•s
(f) cos(kt
+kr)x:
•,•s(f) aretherealandimaginary componentsofthe +•s(f)sin(k• +kr)xz], (13a)
cross-spectraldensity
•,•s(f) between
theincident
and
reflected waves: •=z(f)=• (f) +}ss(f)+2[•s(f) cos(k,
+kr)x=
•As(f)=•As(f) +i0•s(f). +Q•s(f) sin(ki+kr)xz], (lab)
Consequently,
the acousticimpedance
[Eq. (4)] is ex- Cre(f) = S• (f) eosk•(xt - x=)+ Ssseoskr(xt - x=)
pressed by +&•s(f)[eos(krx
t +k•xz)+eos(k•x
t +krxz)]
+•s(f•sin(krx1+kixz)
+sin(k{x
1+krx=)], (13e)
pc - },•(f) +•ss(f) - 2(•,•(f)
Qlz(f) = - S• sinkt(x1- x=)+ Sss(f ) sinkr(xl -
or
+C•s(f)[-sin(k•xt +krx=)+sin(krx
t +k{x=)]
/•.
(f) •;,•A(f)- }•(f)
pc -•AA(f)+•s(f)- 2•,•s(f) '
(9a) +•s(f)[eos(k•x
t +krxz)
- eos(krx
t +k•xz)].(13d)
x.(f) - 2Q,•s(f) If the quantitiesSn, Szz
, C=, •d Q= are estimated
, (Ob)
pc - from measuredtime records, the aboveequationse•
be solved for the •knowns S•(f), Szs(f), C•s(f), •d
where R.(f) and;{.(f) are the resistive and reactive im- Q•(f) from which the acoustic impeduce •d power
pedance, respectively. Also, the power reflection co- reflection coefficient can be determined us•g Eqs. (9a),
efficient •.(f) can be estimatedby (ab), =d (t0a).
For the ease with no me• flow, Eqs. (13) reduce to
•.(f ) =•ss(f )/•.•,•(f ). (10a)
Equations (9a), (9b), and (10a) are analogousto expres-
sions for impedance and power reflection coefficient
(14a)
developedby assuming harmonic excitatio•n. In the
above equations the quantities SA•(f ) and S•s(f) repre-
sent the mean-square value of the incident and reflected
(14b)
pressures in a narrow frequencyband. The phase-shift
between the incident and reflected waves is expressed
in thecrosszspectral
densitySA•(f). Specifically,
the (14c)
phase shift 8 is
(14d)
•(f) =tan'{•,•(f)/•(f)]. (10b)
The total pressure at two points xt and x,. in the tube If a third microphone is located downstream of the
is
systemunder•study,thespectraldensityof thetrans-
•(x:,t)=•,(xi,t)+•,(xiit) mitted wave, S½½(f),can be measured directly, provid-
ing an anechoic termination is used:
=[}(t)e"kixt
+•(t)e'•r•]et•' , (11a)
}½½(f)=
(1/T){•½(f,T)P•e
(f, T)}, (15)
•(xe,t)=[a(t)e-•k•e
+•(t)e'•,Xe]e
•. (11b) where Pc(f, T) is the finite Fourier transform of the
downstream microphone signal. The transmission loss
The auto- and cross-spectral densities of these two
(TL) is then given by
pressures are estimated by
TL=10log:0[
},•,•(f)/•½c(f)]. (16)
•;n(f)=(1/T){•x(f,T)• (f, •}, (12a)
},,(f)=(1/•{•=(f, •(f, •}, (12b) II. EXPERIMENTAL METHOD
•i/•ROPHONES
is x/r,.
_•ACOUSTIC
DRIVER SYSTEM
UNDER
STUDY--• One is tempted to make T, as small as possible so
that n is very large for a given length of total time T,
thereby reducing the random error to a very small val-
ue. However, this leads to a biasing of the spectrum in
AMPLIFIER
ILOW-PASS
FILTER
I RANDOM
NOISE GENERATOR which spectral peaks are not resolved because the band-
width is too wide. Consequently, the bandwidth (and
FIG. 3. Experimental set-up used in determination of imped- therefore T,) should be determined from a knowledge of
ance and power reflection coefficient using the t•vo-microphone the frequency resolution needed for the particular quan-
random-excitation method.
tity being analyzed. This is best done experimentally
by trying progressively larger bandwidths until the fre-
quency resolution is insufficient to separate adjacent
tic driver. A low-pass filter was used to attenuate the
spectral peaks (note, even if the input excitation is white
high-frequency portion of the input signal to prevent
noise, the sound spectra may be frequency dependent
aliasing of the sampled data. It should be noted that the
because, in general, the acoustic system will be re-
shape of the input spectrum is arbitrary providing there active).
is sufficient energy throughout the frequency range of
interest to guarantee an acceptable signal-to-noise ratio In evaluating the three test cases a bandwidth of 20
I .
in the sound field of the tube. Flush-mounted i-in. Hz was found sufficient to resolve the sound-pressure
microphones were used to measure the sound pressure spectra. The random error was made accep.table by
at two points in the tube. The microphone signals were averaging n= 150 individual spectra.
conditioned by microphone amplifiers, digitized by an
analog-to-digital converter, and stored in minicomputer IV. CALIBRATION
memory for processing. For evaluating the muffler
In the linear range the acoustic properties are inde-
test case (see Sec. V) an anechoic termination was at-
tached to the downstream end of the muffler and a third pendent of the sound intensity in the tube; hence absolute
calibration of the microphones is unnecessary. How-
microphone was flush-mounted just downstream of the
ever, since there may be differences between the mi-
muffler exit. The two upstream microphones must be
crophone and amplifier systems, a relative calibration
sampled simultaneously for computation of the cross-
is needed.
spectral
density
•(/), butthedownstream
microphone
can be sampled independently. Consequently, one of the Sincethecross-spectral
density
•l•.(f) is needed
for
upstream microphones can be used downstream and only the solution of Eqs. (14), a phase calibration as well
t•vo microphone systems are needed. as a gain calibration is necessary.
The location of the upstream microphones is not criti- To calibrate, the microphones were flush-mounted in
cal, within certain limitations. The microphones a rigid circular plate attached to the end of the tube.
should be located as close to the unknown system as Ideally, in this configuration both microphones would
possible so that dissipative losses in the tube are mini- measure the same pressure amplitude with zero phase
mized. The microphone spacing should be as small as shift. Differences in microphone sensitivity and phase,
possible for the same reason. However, each micro- as well as gain and phase differences between the mi-
phone is assumed to measure the sound pressure at a crophoneamplifiers, can thus be measured. The easiest
point in the tube, so that for very close microphone method is to determine a transfer function, or frequency
spacingthe effective spacing(x•- xl) is difficult to esti- response, between the two microphones with the air in
mate, particularly for large-diameter microphones. the tube excited by random noise. An estimate of the
Therefore, the microphone spacing should be much frequencyresponseH•.e(f) is obtainedby7
larger than the diameter of the microphones used so that
the effective microphone spacing can be assumed to be
=
the distance bet•veen the microphone center lines. where $n{f} is a smoothedestimate of the auto-spectral
density
ofmicrophone
I and}xe{f}is a smoothed
esti-
mate of the cross-spectral density bet•veen microphones
III. STATISTICAL CONSIDERATIONS
1 and2. Since•[•.(/) is a complex
quantity,sois
Since the sound field varies randomly with time, the •f•(/) andI•(/)1 is thegainbebveen
themicrophones.
estimates of the microphone auto- and cross-spectra Thephaseshift•l•.(/) between
themicrophones
is found
also will be random variables. It should be mentioned by
that the spectral.estimates[Eqs. (1.2)]are inconsistent
Im[• e(f)]
and some form of smoothing must be done to reduce the qb
•.e
(f) =tan'• •'
random error to an acceptable level. One method of re-
Re[•e(f)] '
ducing the random error of a spectrum is to divide the where Im and Re refer to the imaginary and real parts
+10
I I I I I I
_ +100
ø
I I I I I
0 1600 3200 4800
FREQUENCY
(HZ)
of•te(f), respectively.
Figure4 shows
thefrequency Assuming no losses, the resistive impedance is zero,
responsegain and phasebetweenthe two microphone/ the power reflection coefficient is unity, and the reactive
amplifier systems used in this study. impedance is given by
The microphone spectra for each of the test cases xn(f)/pc:- cot}œ,
were corrected using the microphone frequency re-
where L is the length of the tube. Figure 5 shows the
sponse calibration as follows:
experimentally determined values for the power reflec-
=[ tion coefficientandphaseangle[Eqs. (10)] for caseI.
Figure 6 shows the corresponding input impedance (re-
sistive and reactive) at the open end of the tube. There
is good agreement between theory and experiment at
(f): [ (f), most frequencies. Since some losses are present in the
where the subscript u refers to the uncorrected auto-
and cross-spectral densities of the microphones. Note,
since microphone 1 was selected (arbitrarily) as the
reference
microphone,
[•(f)]• doesnotneed
correcting.
z
t• 1.600 -
V. EXPERIMENTAL RESULTS
Three test cases with zero mean flow were used to o l.OOO-
tube holding the microphones and with a rigid circular FIG. 5. Power reflection coefficient and phase angle for case
plate covering the end, approximating zero particle I. Solidline: theory/ open square, opentriangle: experi-
velocity at x-0. ment.
1.000 -
wavelength apart. At frequencies
1
corresponding to
microphone spacing of •nX, n=l, 2, ..., Eqs. (14)are
.$00 -
i•11conditioned.Specifically,Eq. (14a)demands
that
1
1.000 -
data from the SWR method had to be corrected by probe- evaluated in an anechoic chamber where background
tube calibration data. This calibration was a least- noise at low frequency was minimized. )
squares fit of a curve through discrete frequency cali-
If necessary, the microphones can be placed far
bration data and the calibration curve had an error of
enough from the system under test to minimize any near-
approximately ñ 1 dB (note that the SWR transmission
field or higher-mode effects, a problem in measuring
loss data are negative around 300 Hz).
source impedances.