Measurement of Aerodynamic Noise using STSF

by M. Nakamura, T. Komine, M. Tsuchiya: Honda R&D and J. Hald: Brüel & Kjær

Abstract
Until now it has been fairly difficult
to measure accurately automobile
aerodynamic noise in a wind flow us-
ing a microphone, because of distur-
bances such as the microphone self-
generated wind noise and back-
ground noise in a wind tunnel.
This Application Note presents the
results of checking the validity of the t
Spatial Transformation of Sound t --
f
H i
Fields (STSF) technique for noise
measurements in a wind flow.
First, the basic validity is demonstra-
ted by a measurement on a known
source consisting of two loudspeak-
ers and then the application of the
method for automobile aerodynamic
noise measurement is demonstrated.

Below, the interior of Honda’s low noise wind tunnel. Above, the STSF scan array in use in the wind tunnel

Brüel & Kjær

1 Introduction
There are three major sources of noise cant reductions in the interior noise, urement technique applied in STSF
inside cars: engine noise, tyrelroad the aerodynamic noise must be re- can be used to reduce the influence of
noise and aerodynamic noise. The en- duced. Also, thinking about the envi- the self induced noise, and the holo-
gine noise is currently being reduced ronmental noise around highways, graphic calculations of STSF can be
by investing efforts in many noise re- the external aerodynamic noise must used for precise localization of the tur-
duction techniques. Therefore, the be considered. Until now it has been bulence noise sources from measure-
relative contribution of the aerody- difficult to measure aerodynamic ment taken outside the turbulence re-
namic noise to the interior noise un- noise using a microphone, primarily gion.
der high speed driving conditions is because the microphone generates
increasing. To achieve further signifi- noise itself. The cross spectral meas-

2 Measurement Principle
2.1 Basic sound field transformation principles

The basic principle of STSF is to meas-

ure cross spectra of the sound pres- Spatial domain Spatial frequency domain
sure over a plane close to the sound z=0 z=0

Oi
source, and then apply Helmholtz’In-

O
tegral Equation and Near-field Acous-
tic Holography to calculate descriptors
of the sound field at other points.
6
Helmholtz’Integral Equation is used
for calculation of the Sound Pressure 2
Level at rather large distances, while
Near-field Acoustical Holography ap-
\-’
Noise source J-J

plies for calculation of pressure, parti-

cle velocity, active and reactive acous-
tic intensity in the near-field region,
both closer to and further away from Fig. 1 Noise source and measurement
the source than the measurement plane
plane (Fig. 1).
In Near-field Acoustic Holography
the spatial distribution of (complex) Fig. 3 illustrates the use of the spa-
pressure on the measurement plane tial frequency domain in the hologra-
z=O is transferred to the spatial fre- phy calculations. Plane waves and
Fig. 2 Principle of Near-field Acoustic
quency domain using a two-dimen- evanescent waves in the spatial do- Holography
sional FFT. (Fig.2). The spatial fre- main are located inside and outside
quency domain representation can respectively of the so-called radiation
then be easily “moved” to a parallel circle in the spatial frequency domain.
plane z=z’by multiplication with a The evanescent waves contain the pick up as much as possible of the
transfer function, and the acoustic high resolution information and evanescent wave information, the
field reconstructed by taking the in- therefore the capability of precise lo- measurement plane must be as close
verse two-dimensional FFT. calization of noise sources. In order to as possible to the sound source.

Spatial domain Spatial frequency domain

Evanescent wave F-‘

__________
________ 4
Propagating wave

kz= d($,‘- &+$)I

Fig. 3 Relationship between spatial domain and spatial frequency domain

2
2.2 The cross spectral measurement technique Array microphone

In the general case where no assump- every position to each of a set of q +:-P
tions are made about the coherence of references (scan measurement) and
the sound field, the cross spectrum the cross spectrum between every pair
must be measured between every pair of references (reference measure-
of measurement positions in the scan ment).
plane in order to obtain a complete In this case we need to measure
cross spectral model of the sound field. only qN2 + q2 cross spectra which is a
For example, with N rows and N col- very significant reduction when q is
umns of measurement positions, the small (Fig. 4)
number of cross spectra to be meas- In order to avoid the need for si-
ured will be N4. multaneous measurement of all cross
But in STSF the noise source is spectra, the source is assumed to be
1 Reference microphone
assumed to have only a limited stationary stochastic, and all positions
number (q) of independent parts. Un- are covered by traversing a column
der this assumption we need only array or a single microphone across
Fig. 4 Reduced cross spectral measure-
measure the cross spectrum from the measurement area (Fig. 4). ment

2.3 Rejection of wind noise generated by the microphones

In general it is difficult to measure The cross spectra applied in the STSF 80 -

100 km/h
aerodynamic noise generated by the technique are the cross spectra GRR
surface of a car using a microphone between the references, and the cross 70 -
0
located in the flowing air because the spectra G, between reference and Car in wind tunnel
microphone itself generates noise array microphones. If we want to re- A
even when a wind shield is applied. move the influence of the self induced
The cross spectral measurement tech- wind noise from the STSF model of
nique applied in STSF provides a the sound field, it is important to
means of reducing the self generated avoid the generation of wind noise in
noise. Provided the reference trans- the references, because this kind of
ducers are not in the flow (and thus do noise will contribute to the reference 30 -
not generate noise) and provided they auto-spectra GRR. We shall therefore
do not pick up the noise generated by assume that the references are out- zo- , , , , , , , ,
the scan microphones, then this self side the air flow and thus do not gen- 63 125 250 500 lk 2k 8k Lin.
generated noise will be averaged out erate significant wind noise. Frequency (Hz)
in the cross spectra between the refer- Denoting by “*" a cross spectrum Fig. 5 Background noise level in wind
ences and the scan microphones. This operation between two signals, the tunnel
is described mathematically in the fol- cross spectra G,, can then be ex-
lowing. We shall apply the symbols pressed as For the cross spectra G, between ref-
defined below: erences and array microphones we
PR total reference microphone sig- G,=P, * P, obtain similarly
nal
p.4 total array microphone signal =(Paw+PsM+P,)*(Psw+P,+Pss) GRA = PRw*PAw+PRM*~M+~as*~AB (5)
PRW aerodynamic noise part of ref-
erence signal = PRw*PRw+Pm*PRM+Pas*PRs (3) For the same reasons as above, we
PAW aerodynamic noise part of array shall disregard the contribution from
microphone signal where we have assumed that aerody- the background noise:
wind noise generated by array namic noise, self induced noises and
microphones and measured by background noise are mutually uncor- GM = PRw*PAw+PRM*PAM (6)
a reference microphone related. Clearly, the wind noise PRM
wind noise generated by array created by the array microphones However, the wind noise generated
microphones and measured by must be negligible compared to the by the array microphones will nor-
an array microphone aerodynamic noise P,, at the refer- mally have a very high level as meas-
P RB background noise measured by ence microphones and it can then be ured by one of the array microphones
reference microphone disregarded in eq. (3). If the reference because a turbulence source is just in
PAB
background noise measured by microphones generate wind noise, this front of the microphone. In order that
array microphone noise must also be small compared to the second term in eq. (6) can be ne-
PRw. In the Honda wind tunnel, the glected, the same array wind noise
Here, the total microphone signals can background noise is very low (less must be very small as measured by
be written as than 60 dB(A), see Fig.5), and it can the reference microphones. Effec-
therefore be disregarded. Conse- tively, the following inequality must
‘
R= ‘
RW + ‘
R&l + ‘
RB (1) quently, we obtain from equation (3) hold

PA=Paw+PAM+PAB (2) GRR = ‘

RW*‘
RW (4) IP,*P,,I<< IPRw*PAwI (7)
3
in order that we can obtain the de- sible from the array. The desired aero- Although one scan microphone is
sired expression dynamic wind noise contribution to the best solution as far as only the
the same cross spectrum can be maxi- suppression of self induced noise is
GR4 = PrW*P*w (8) mized by positioning the references as considered, an array of scan micro-
close as possible to the aerodynamic phones is chosen in order to reduce
The undesired array wind noise con- noise sources and by scanning the ar- the measurement time. By the use of
tribution to the cross spectrum GRA ray as close as possible to these a rather large spacing in the array,
can be reduced by selecting optimum sources. the noise from the other microphones
sized and shaped wind screens on the Positioning of the references close at a given array microphone position
array microphones, by the use of only to the noise sources of interest will in can be minimized.
a single array microphone and by po- general reduce the influence of other
sitioning the references as far as pos- undesired sources.

3. Measurement system
Fig. 6 shows the configuration of the
measurement system. A car is located
in the low-noise wind tunnel. Several
reference microphones can be set up
Channel
selector
+-ijzJ
inside and/or outside the car. A verti-
cal array of microphones with small
wind shields are scanned along the
Reference microphone
$
‘
I
side of the car. Both the reference and

CSL
Computer
the scan microphones are pressure
microphones connected to a Brüel &
Kjær Dual Channel Real Time Analy- Printer Plotter
zer Type 2133 through Brüel & Kjær
Type 2811 multiplexers. Analyzer,
multiplexers and a microphone tra-
versing system are controlled by a Fig. 6 Configuration of measurement system
HP9000 series computer. During the
scan of the microphone array, a cross
spectral model of the sound field co- Microphone
herent with the reference signals is traverse
acquired.
Afterwards, this model can be used Microphone
Microphone
for mapping of sound pressure level,
(active and reactive) sound intensity
and sound power.
+E!

3.1 Selection of wind shields

A series of measurements were taken

over a loudspeaker sound source in a
wind flow in order to identify the best
way of reducing the self induced noise
picked up by the array microphones.
In order to avoid too much influence
of the array on the air flow in the
Fig. 7 Microphone array with windscreens Fig. 8 Test setup for windscreen selection
measurement region, it is very impor-
tant to keep the dimensions of wind
shields small. The standard B r ü e l & tion signal and the signal from the 100 km/hour was applied. In case (b)
Kjær wind-screens were too large, and array microphone is measured with a the microphone was placed in the tur-
therefore a set of smaller screens were finite averaging time of 4 seconds with bulent flow behind a side mirror of a
made and their noise reduction was and without wind flow. The best wind- car.
compared with that of other typical shield will give the smallest deviation Three different shielding methods
shielding methods. between .the two cross spectra. are investigated:
Fig. 7 shows the array with its car- Two distinct cases are considered
riage structure and with the small in Fig. 9: (1) small wind-screen
wind-screens fitted on the micro- (2) nose cone
phones. Fig. 8 shows the test setup for (a) steady state wind flow (3) no wind shielding
wind-screen selection and the results (b) turbulent wind flow
are presented in Fig. 9. The cross spec- For the case of steady state wind flow
trum between the loudspeaker excita- In both cases a wind speed equal to (a), the cross spectral averaging is al-

4
most sufficient to suppress the self 120 120
induced noise with all three shielding 100 km/h ,.... 100 km/t
110-
methods. The small wind-screen is,
however, the best. For the case of tur- ~loo-
bulent wind flow, only the small wind- D
screen provides acceptable results - E go-
within approximately 1dB from the : 80_ 5 80s
measurement with no wind flow. 0 :
70_ :’ 70-
$
-No wind flow 6 60_ - No wind flow
6 60_
- - - Small windscreen -- - Small windscreen
50 _ --- Nose cone 501 --- Nosecone
Without windscreen
40
125 250 5tiO lit 2j( 4i(
Frequency (Hz)
(a) Steady State Wind Flow

Fig.9 Test results of windscreen selection

4. Verification of the Measurement Technique by a Loudspeaker Measurement

4.1 Verification method

It is impossible to verify the ability of wind induced noise. The scan array and the height of the loudspeakers
STSF to identify noise sources in a consists of 8 microphones with 56 mm above the floor is relatively large to
wind flow by taking measurements on spacing. The distance from the loud- approximate free-field conditions in
aerodynamic noise sources, because speakers to the scan plane is 100 mm the measurement region.
the localization of these noise sources
and their acoustic power are un-
known. Instead, we shall compare two
measurements on a set of loudspeak- 4.2 Results of the verification
ers: one with wind flow and one with-
out. Fig. 11 shows the calculated acoustic intensity levels in front of the loud-
Fig. 10 shows the experimental set- intensity distribution on the surface speakers agree within f 1 dB with and
up. Two loudspeakers excited by pink of the loudspeakers. The two plots without wind flow, indicating that
noise are located one after the other show good agreement between the noise sources can be localized and
in the flow direction. case with presence of wind flow and quantified with high accuracy in a
The sound pressure level from the the case without wind flow. The sound wind flow.
speakers is adjusted to equal the aero-
dynamic noise level just in front of the
speakers, and the electric signal from
the generator is used as reference sig-
nal. This signal has high coherence
with the sound field from the loud-
speakers and no coherence with the

Y +Y

rind velocity:Okm/h Wind velocity:lOOkm/h

Noise
Generator Sound power (+)77. ‘2dB Sound bower (+)77,5dB

l- To Analyzer (a) N O Wtnd Flow (b) Wind Flow

930%~

Fig. 10 Speaker test setup Fig. 11 Sound intensity distribution

5
 5. Application for Automobile Aerodynamic Noise Measurement
5.1 Measurement set-up

The vehicle is fixed on the turntable spacing constitute the array, which is passenger, 100 mm from the side win-
in the wind tunnel, and the vertical scanned from the front of the vehicle dow, is used as reference.
linear microphone array is moved to the rear in 112 mm steps. Two Since the floor is perfectly reflec-
along the side of the vehicle. sweeps (traverses) are performed to tive, a mirror ground type of measure-
The distance from the centre of the cover the vertical extent of the scan ment is selected, with the lowest scan
side window to the scan plane is 250 area. A sound pressure microphone microphone very close to the floor.
mm. Eight microphones with 112 mm located at a typical ear position of a

5.2 Results

Fig.12 and Fig.13 show the pressure mogeneous average. Small inhomo- well with the region where flow re-
distribution on the surface of four dif- genities will cause a distortion (blur- joins the car body after a separation.
ferent vehicles. Although the NAH ring) of the pressure map, while Recall that the mapped sound field
calculations require a homogeneous sources between the measurement is the part of the total sound field
source free medium between the and calculation planes will be repro- which is coherent with the reference
measurement plane and the calcula- duced in a defocussed form. signal. Therefore, Fig.12 and 13 show
tion plane, reasonable results are The pressure level is high behind the pressure field which is coherent
achieved because the inhomogenities the side mirror and behind the front with the sound perceived by the pas-
are either small or correspond to tyre. Otherwise it is low. These re- senger.
rather small deviations from the ho- gions with high pressure coincide very

CAR 4 Pressure 194 Hz to 729 Hz

142 wt. bandwidth Z-courd. -,258 m A-Weighted
No Eq, Smooth EV 10 d B

Fig. 12 Outside Sound Pressure Distribution on the Side Surface of a Car measured with 100 km/hour Wind Speed

.
6. Conclusion
The application of STSF for localiza- of interest and low coherence with ments taken at a certain distance
tion of aerodynamic noise sources has the self induced wind noise, then on the surface of the source.
been described and tested. The main the noise sources can be identified
conclusions are: in a wind flow. 3) Experiments indicate that it is
possible to locate the aerodynamic
1) If the reference signals have high 2) Holography can do a precise noise noise sources on a car body.
coherence with the noise sources source localization from measure-

7. References
[l] J. D. Maynard et al: “Near-field acoustic holography. I: Theory ofgeneralized holography and the development of NAH”,
J. Acoust. Soc. Am. 78 (4), 1395 (1985).

[2] J. Hald & K. B. Ginn: “Spatial Transformation of Sound Fields: Principle, Instrumentation and Applications”, Acoustic
Intensity Symposium, Tokyo (1987).

[31 J. Hald: “STSF -a unique technique for scan-based Near-field Acoustic Holography without restrictions on coherence”,
B&K Technical Review No. 1 (1989)

141 T. Komine et al: “Aerodynamic Noise Measurement using Near-field Acoustic Holography”, JSAE, May (1992).

7
 75 85 35w
A?’ .ll2m

33 dtl
AY S12m

95 d8
AY .Il2m

Fig. 13 Outside sound pressure

distribution on the side surfaces
of four different cars.

Wind speed: 100 km/hour

Frequency range:
194 Hz to 729 Hz A-weighted

Brüel & Kjær

WORLD HEADQUARTERS: DK-2850 Nzerum Denmark Telephone: +45 42 80 05 00 Telex: 37316 bruka dk Fax: +45 42 8014 05
Australia (02)450-2066 Austria 0043-l-8167400 Belgium 016/449225.Brazil (011)246-8166 Canada: East (514)695-8225 West (604)591-9300 China 1-8419625
Czech Republic 02-67021100.Finland (0)1481577France (I)64572010 Germany 06151/8149-O. Great Britain (081)954-2366.Holland 03402.39994
Hong Kong 25487486. Hungary (1)2158305.Italy (02)57604141Japan 03-5420-7302.Republic of Korea (02)554-0605.Norway 66904410. Poland (O-22)409392
Portugal (1)3859256/3859280Singapore 7358933. Slovak Republic 07-497890 Spain (91)3681000. Sweden (08)7112730. Switzerland 01/9400909
Taiwan (02)7139303Tunisia (01) 750400 USA (800)332-2040
Local representatives and service organisations worldwide
PRINTED IN DENMARK BY N&=iUM OFFSET
English BO 0392-I 1