Techniques for Measuring the

Vibro-Acoustic Transfer Function

Andreas Schuhmacher, Brüel & Kjær Sound & Vibration Measurement A/S, Nærum, Denmark

Measurement of transfer functions is required for most appli-

cations dealing with source-path-contribution techniques often p
=
p
Qrecip Q direct Q
called transfer path analysis. Here the transfer function, typically
measured as frequency response function (FRF), takes the role of
connecting an input (for example, source position) with an output
(receiver position). In this article, we further investigate a previ-
ously described low-mid-frequency volume velocity source based
on the two-microphone method for in-situ measurement of volume
a p
velocity source strength. This investigation includes the effects of . =
Qrecip Fdirect
the acoustical environment when measuring transfer functions.
The two strategies, direct and reciprocal measurement, will also
Figure 1. Reciprocal measurement of acoustic (p/Q) and vibro-acoustic (p/F)
be compared to investigate their validity for a typical acoustical transfer functions in a vehicle.
setup. Finally, we compare the described volume velocity source
with a mid-high-frequency sound source based on the same two- hose, where the sound radiates from the orifice of an open duct,
microphone method. the two-microphone method can be used to estimate the volume
velocity output without first estimating a sound pressure to source
For applications where the acoustic radiation from a compli- strength relationship in the anechoic room. A further benefit of this
cated sound source is modeled, a series of sound pressure/volume approach is the ability to determine the output volume velocity in
velocity (p/Q) FRFs are normally measured and combined with any acoustic environment.
operating acoustic source strengths as part of a method to find the Apart from these technical requirements, one might want to add
airborne contribution of this sound source. If structure-borne noise some practical requirements. For instance, that the source should
is the main concern, the operating forces on a receiving structure be small enough to be used in confined spaces, for example, inside
are estimated, and the noise contribution at a receiver position can a tightly packed engine compartment. For airborne contribution
then be estimated from these operating forces and a set of measured analysis, a volume velocity source is used initially to provide
sound pressure/force (p/F) FRFs. Both the p/Q and the p/F transfer acoustic transfer functions between source positions close to the
functions can be measured using acoustic excitation while there surface of the actual noise source and some near-field microphones
are no other operating sources. located around the noise source. Therefore, to measure transfer
To measure vibro-acoustic p/F FRFs in a vehicle, we normally function easily, it must be possible to attach the sound source
take advantage of the reciprocity principle by placing a sound to any surface or to position it in the near field if a reciprocal
source inside the cabin at the receiver and mounting an acceler- approach is used. This supports the use of a sound source based
ometer at the input force location. Acoustic p/Q FRFs on the other on some sort of driver attached to a long flexible hose, where the
hand can be measured using either a direct or reciprocal approach, sound is radiated from the duct orifice. The duct end can then be
with a microphone measuring the sound pressure at either the attached to a surface or be positioned in free space during transfer
receiver location or at an assumed acoustic source position (on the function measurements.
engine surface; for example, see Figure 1). For practical reasons,
however, transfer functions from engine room source to cabin Two-Microphone Measurement Method
receiver are usually measured reciprocally due to the limitation The principle behind a particular volume velocity sound source
of space in engine compartments, even if the source is based on a was described in terms of how it was designed in a paper from
driver (loudspeaker) with a long hose attached. 2004.2 An omnidirectional sound source (B&K Type 4295) already
A volume velocity source has to meet some specific require- used for room acoustics applications, was chosen as the driver
ments:1 together with a special adaptor (B&K Type 4299) that measures
• Source should produce a sufficiently high sound level. the volume velocity output. A pair of phase-matched microphones
• Frequency range covered should be appropriate. is used inside the adaptor to estimate the calibrated volume ve-
• Source should behave as a monopole in the frequency range locity output spectrum in situ. Figure 2a shows the source itself
of interest. with the adaptor mounted inside an anechoic room and Figure 2b
• Output volume velocity should be measurable even when the shows a practical measurement setup where a hose (flexible duct)
acoustic environment changes. is mounted between driver and adapter for ease of use during
The acoustic source for this purpose must be powerful and om- measurements. As the useful frequency range of the driving loud-
nidirectional, and a signal related to the source strength must be speaker is 50-6000 Hz, the output will be sufficient. However, the
available if the source strength is not available directly. Also the radiation from the orifice of the adaptor becomes more directive;
frequency range covered should be as broad as possible. Most vol- that is, less omnidirectional, above 2-3 kHz. Later we will call this
ume velocity sound sources use one microphone as a reference, as- the low-mid-frequency sound source.
suming there is a fixed linear relationship between volume velocity We will now review some of the basic concepts behind the
output and reference sound pressure at the microphone. To find this two-microphone method that has been widely used to measure
relationship, the sound source is operated in an anechoic room, and acoustic properties in ducts.3 We assume that only plane waves
the volume velocity output can be estimated from a microphone are measured at two microphone locations (A and B) inside a
measurement at some known distance from the source. A transfer cylindrical duct (see Figure 3).
function between volume velocity and reference sound pressure The sound pressure p(x) in a cross-section of the duct can then
can then be calculated and stored for use when the source is used be expressed as:
in the real environment. The influence of a changing environment p( x ) = p + e - jkx + p_ e jkx (1)
on this fixed linear relationship is covered later in this article by
some real measurements. For sound sources based on driver and where p+ and p_ are the incident and reflected plane wave com-

6 SOUND & VIBRATION/MARCH 2010 www.SandV.com

 autospectra of Microphones A and B and the cross-spectrum from
Microphone A to B.2
Furthermore, since the purpose is to use this source for measur-
ing transfer functions, we can express the transfer from volume
velocity output at the source to sound pressure at a receiver mi-
crophone as a frequency response function:
CQp
H Qp = (4)
CQQ
Here CQQ is the autospectrum of the estimated volume velocity
signal, and CQp is the cross-spectrum from source volume veloc-
ity to receiver sound pressure. Both spectra can be expressed in
terms of auto- and cross-spectra among signals from Microphones
A and B plus the receiver. Additionally, we need the dimensional
parameters l and D, where l is the distance from the duct opening
to the nearest microphone (Microphone B) and D is the microphone
spacing.
In one study,4 the two-microphone method was presented and
investigated the influence of different error sources on the esti-
mated result of acoustic properties, with emphasis on measuring
reflection coefficients and acoustic impedance of materials.

Finite-Element Modelling Inside Duct

To verify some of the aspects related to sound fields in ducts, it
is very useful to perform simulation studies for optimizing design
parameters. Some studies have investigated the influence of the sur-
roundings on the duct output and the interference effect of placing
microphones inside the duct for measuring sound pressure. The
finite-element method was used to do these simulations based on
the standard Helmholtz equation.
In the two-microphone method, we need to measure the sound
pressure in two cross-sections of the duct to estimate the volume ve-
locity output. The effect of placing the two microphones inside the
duct may change the sound field locally around the microphones;
this can result in large errors in the calculated volume velocity
Figure 2. Low-mid frequency sound source (a) without and (b) with exten-
sion hose. spectrum. This will be evident when dealing with narrow ducts.
We modeled the setup from the low-mid-frequency sound
source, with the microphones placed inside the duct together with
a spacer. The two microphones measure the sound pressure on the
duct axis in order not to capture the first higher order mode. The
spacer ensures that the sound pressure is measured on the actual
center axis of the duct.
The air inside a piece of duct (inner diameter 3.8 cm) contain-
ing the two microphones and the two spacers was modeled using
acoustical finite elements. The duct is excited at one end with a
known constant surface velocity and, at the opening where the
microphones are located, an impedance boundary condition was
imposed to simplify the setup of the simulation case. The im-
pedance imposed on this boundary corresponds to the acoustic
impedance seen from a piston in an infinite baffle.5
The output from the simulation will now be the particle velocity,
integrated over the opening of the duct, which will be the simulated
Figure 3. Two-microphone measurement configuration for volume velocity
output estimation. volume velocity (also termed as exact). Furthermore, we can take
the simulated sound pressures integrated over the microphone
ponents respectively, and k is the wave number. By measuring the diaphragm boundaries, process that data using the volume veloc-
sound pressure inside the duct at the two different microphone ity estimation method (described earlier), and finally compare
locations, we can determine the unknown incident and reflected this with the exact simulated volume velocity to investigate the
plane wave components. Usually this is done by measuring the interference effect of the microphones and spacers inside the duct.
transfer function between the two microphones. Therefore, the A simulation was carried out from 20 Hz-10 kHz and, as expected,
method is also referred to as the transfer function method. at high frequencies the influence of obstacles inside the duct is
The particle velocity evaluated in a cross-sectional area is given quite significant. An example of a surface pressure distribution
by: is given in Figure 4 for a frequency of 8 kHz. Locally, around the

( )
1 position of the microphones, the sound field changes and does not
u( x ) = p e - jkx - p_ e jkx (2) consist of plane waves.
rc +
A comparison of the direct simulated volume velocity output
where rc is the characteristic impedance of air. spectrum and what was predicted based on the simulated sound
At the duct opening, x = 0, we have: pressures inside the duct at the microphone diaphragms is shown
in Figure 5 for the full frequency range considered. Moreover, the
u(0) =
1
rc +
(
p - p_ ) (3) two spectra were subtracted to provide the error made by the two-
This expression leads to volume velocity estimation if we multi- microphone method as a function of frequency. Clearly in the range
ply by the cross-sectional area of the duct. We can further express where the actual low-mid-frequency source is active, 50 Hz-6 kHz,
the volume velocity output as an autospectrum based upon the the error is always less than 1.5 dB, and for most of the frequency

www.SandV.com SOUND & VIBRATION/MARCH 2010 7

Figure 4. Setup of microphones (a) inside duct and (b) sound pressure
distribution at duct end for a frequency of 8 kHz.

Figure 6. Sound source placed in different environments: (a) inside room

away from the walls; (b) close to floor; (c) radiating into box; and (d) inside
engine compartment.

Figure 5. Comparison of simulated and predicted volume velocity output

from duct with two microphones inside. (a) Volume velocity spectra; (b)
Difference between spectra.

range, the error is actually less than 0.5 dB.

Verification of Volume Velocity Source

Verification measurements were done using the low-mid-
frequency sound source to explore some of the aspects explained
earlier. The basic parameters describing the sound source were
already presented,2 like maximum output power, directivity char-
acteristics, etc. Here we present some other measurements verifying
the concept of the volume velocity sound source. One of the ideas Figure 7. (a) Calculated volume velocity output spectrum relative to free
space for three different environments; (b) Volume velocity to reference
behind using the two-microphones for volume velocity estimation sound pressure ratio for all environments.
is that the quantity can change if the sound source is used in very
different acoustical environments. Here the two-microphone prin- rowband auto- and cross-spectra between the two microphones
ciple should always provide a good estimate of the actual volume were measured using FFT and averaging. For each of the four setups
velocity. A couple of measurements were done with the hose and in Figure 6, the volume velocity spectrum was calculated based on
adaptor placed in different environments. Photos of the different the two measured microphone signals. Calculated volume veloc-
setups are shown in Figure 6. ity spectra are shown in Figure 7a, where three of the calculated
The sound source was driven by a band-pass filtered (bandwidth spectra are shown relative to the measurement where the source
6.4 kHz) white-noise signal with sufficiently high amplitude. Nar- was placed freely in the room away from the walls. Below a certain

8 SOUND & VIBRATION/MARCH 2010 www.SandV.com

 60

50
dB, Re: 20 µPa

40

30

20

10

–10
0 1k 2k 3k 4k 5k 6k
Frequency, Hz

Figure 8. Sound pressure measured 30 cm in front of open/blocked orifice

of low-mid frequency source.

frequency (around 1.5 kHz) the different outputs from the source
are all within a few dB, but at higher frequencies the deviations
from the free space measurement become more evident especially
for the more confined spaces. The change in the output volume
velocity spectrum due to change in the acoustical environment
would require a reference signal that also changed accordingly.
Some of the volume velocity sources available are based on a
single reference microphone for calibrating the volume velocity
output from the source. The idea of such a principle is to estimate
the volume velocity under anechoic conditions using a far-field
microphone and then relate the calculated volume velocity output
to a fixed reference signal; in this case, a microphone sitting close
to the opening of the sound source. This results in a calibration
spectrum. When using the sound source in a real application, we
measure the sound pressure at the reference microphone, which
can then be translated into a volume velocity spectrum using the
calibration curve.
But the question is, what influence will the acoustic environment Figure 9. Measurement of acoustic transfer function from top engine surface
have on the volume velocity estimations we measured, since the point to left and right ear, using the mid-high frequency sound source: (a)
actual measurement environment may be very confined (like inside Hose at engine source position for direct transfer function measurement;
(b) Positioning of sound source at right ear for reciprocal transfer function
an engine compartment). In our test cases, we use the signal from measurements.
the microphone closest to the opening as a reference to examine this
ratio for our four setups. Figure 7b shows the individual curves, and transfer functions.
we see less variation in the ratio compared to the volume velocity
spectrum. However, some errors are introduced if the environment Application of Volume Velocity Sources
becomes more confined. Especially at high frequencies, there are The volume velocity source described so far has been used to
quite large differences. At the same time, we should remember that measure acoustic transfer functions between an assumed source po-
the source itself is only omnidirectional up to 3 kHz, so the largest sition inside an engine compartment and receiver positions inside
errors will occur outside this frequency range. Nevertheless, we the vehicle. Direct transfer function measurements – from source at
have seen that the acoustical environment will have an effect on engine surface to microphones inside a vehicle – were compared to
the output volume velocity spectrum and that we should measure reciprocal measurements where the source (the duct orifice), was
the actual output in situ to minimize errors on volume velocity positioned at the receiver with a microphone measuring the sound
estimation, transfer functions, etc. pressure at the engine surface position. Since the receiver positions
Another simple experiment was conducted to investigate if in the direct measurement consisted of microphones in the ears of
the sound was radiated mainly from the opening of the tube (as a head and torso simulator (HATS), the reciprocal measurement
desired) or if the driver and tube walls contributed significantly. A should ideally be made with a HATS having sound sources placed
microphone was placed 30 cm in front of the opening of the duct, at the entrance of the closed ear canals. This was not practical us-
and a narrowband sound pressure spectrum was recorded for white ing the current sound sources, so in this experiment, the orifice of
noise excitation of the sound source. Then the orifice of the duct the adapter was placed as close to the ear microphones as possible
was blocked and another narrowband sound pressure spectrum but still outside the pinna/concha.
was recorded. For normal operation with the duct open (and with The effect of the head and torso is included in the reciprocal
the opening blocked by a thick layer of damping material inside transfer functions. However, the full effect of the concha is not
the duct opening), the measured spectra in front of the opening are included, so this measurement should give an indication if it is
shown in Figure 8 and compared to the general background noise possible to measure binaural transfer functions related to an in-the-
inside this normal room. The tests show that the sound is mainly ear receiver using a reciprocal approach (sound source is simply
radiated from the opening of the open duct, and even though the attached just outside the pinna). In that case, a standard HATS
blocking of the orifice was not perfect, the levels in this case are with microphones in the ears can be used for measuring binaural
more than 20 dB lower than the case of the open duct over the transfer functions based on the reciprocal approach with one of
complete frequency range for that source. When the source was the described sound sources attached to the pinna. The validity of
blocked, some sound was transmitted through the damping at the this approach can be examined by comparing it to binaural transfer
duct opening, especially at lower frequencies. Otherwise, only functions using the direct approach, which contain the effect of the
sound coming directly from the driver itself was identifiable. concha, since the microphones are placed at the entrance of each
Altogether, we concluded that sound produced by the assembly ear canal. At the same time, we want to compare the low-mid fre-
of driver and hose is mainly radiated from the duct orifice, which quency sound source to a mid-high frequency sound source based
means it can be used as a monopole to measure vibro-acoustic on a similar principle. The mid-high frequency sound source is

www.SandV.com SOUND & VIBRATION/MARCH 2010 9

constructed out of a powerful compression driver and a long hose
made out of nylon reinforced PVC. The inner diameter of the hose
is 10 mm, and a similar set of microphones is used at the opening
to estimate the volume velocity.
The measurements were made with a vehicle standing in a nor-
mal room. Some level of background noise was expected during the
measurements. One source position on the engine top was marked
for use in all direct and reciprocal measurements. Additionally, a
HATS with two microphones was placed in the passenger seat of
this right-hand-drive car (see Figure 9).

Source Directivity
A couple of measurements were carried out with each of the
investigated sound sources for the same source position on the top
engine surface, where the orientation of the adapter or hose was
changed. When comparing transfer functions from the same posi-
tion but different orientations, the directivity of the source can be
examined with respect to omnidirectionality. Figure 10 compares Figure 10. Acoustic transfer function measured between top engine position
and HATS left ear for different nozzle orientations using low-mid frequency
a transfer function measured with the low-mid frequency sound sound source. (a) 0-2 kHz; (b) 2-4 kHz.
source for different orientations of the adapter; that is, pointing
toward the rear, front or left side of the vehicle. The measured
transfer functions are valid down to 50 Hz, where the output
power of the loudspeaker starts to decrease significantly, and we
see similar transfer functions for all three orientations up to 2-3
kHz. From that frequency on, the sound from the orifice of the
adapter becomes more directive, as explained earlier; this can be
seen from the plot in Figure 10b.
In all measurements, a white-noise signal band-limited to 6.4 kHz
was driving the sound source at a maximum level. FFT processing
and averaging were used to calculate transfer functions as FRFs
with frequency resolution of 1 Hz.

Comparison of Sound Sources

Measuring the same direct transfer function from the top engine
surface position to HATS ears was investigated using the two
sound sources. The low-mid-frequency sound source was driven
by a white noise signal band-limited to 6.4 kHz, while the mid- Figure 11. Amplitude and phase of acoustic transfer function from top
high frequency sound source was driven by a similar white noise engine surface to HATS left ear measured using low-mid frequency source
signal high-pass filtered with cutoff at 800 Hz (so not to overload (red curve) and mid-high frequency source (blue curve); 0-2 kHz.
the driver at low frequencies). Transfer functions measured with
the orifice pointing toward the vehicle rear were measured and are
compared in the frequency ranges 0-2 kHz and 2-6 kHz for both
amplitude and phase characteristics. Even though the signal for
the mid-high-frequency sound source is high-pass filtered at 800
Hz, the transfer functions obtained by this source are valid down
to 400 Hz, since sufficient sound output is produced by the source
compared to background noise levels.
Figure 11 shows that the measured transfer functions using the
two sound sources agree very well in both amplitude and phase
from 400 Hz up to at least 2 kHz. Figure 12 shows similar ampli-
tude and phase plots but now in the high-frequency range of 2-6
kHz, where deviations are in the range of 10 dB. This is expected,
since the mid-high frequency source is omnidirectional to a much
higher frequency than the low-mid-frequency source, and also the
dimensions of the sources play a role at higher frequencies together
with their different acoustic centers.
Figure 12. Amplitude and phase of acoustic transfer function from top
engine surface to HATS left ear measured using low-mid frequency source
Direct vs. Reciprocal Measurements (red curve) and mid-high frequency source (blue curve); 2-6 kHz.
Finally, we compare directly measured transfer functions to re-
ciprocal transfer functions. In the case of reciprocal measurements, Comparing direct and reciprocal measured transfer functions
the HATS was still in place inside the vehicle, but now the sound is shown for the low-mid-frequency source in Figure 13. At low
source was placed as close as possible to one of the microphones frequencies, there is very good agreement as expected. From
inside the ears. An example of locating the orifice of the hose just low frequencies and even up 3-4 kHz, the tendency remains the
outside the concha part of the pinna was shown in Figure 9. A same for both curves. Above 4 kHz, the deviations become more
small 1/4-inch microphone normally used for array applications pronounced. At these higher frequencies, the sound source is no
was placed at the top engine surface position for measuring the longer acting as a monopole, and the effect of the concha in the
blocked surface pressure. Ideally, the effect of hose and adapter on reciprocal transfer function is missing.
the sound field locally around the engine surface position should be Comparing direct and reciprocally measured transfer functions
included by having them in place during reciprocal measurements. is shown for the mid-high-frequency source in Figure 14. Some
This effect was ignored, however, since it was not practical. That deviations are expected at lower frequencies, where the output of
is, if we had wanted to include this effect, another piece of hose the source is limited. This is due to a less sensitive array micro-
with a blocked orifice would have been necessary. phone for the reciprocal measurement and also because of poor

10 SOUND & VIBRATION/MARCH 2010 www.SandV.com

Figure 13. Direct (red curve) and reciprocal (blue curve) measurement of top
engine surface to HATS left ear transfer function using low-mid frequency
source: (a) 0-2 kHz; (b) 2-6 kHz.

Figure 14. Direct (red curve) and reciprocal (blue curve) measurement of top
engine surface to HATS left ear transfer function using mid-high frequency
source; 0-6 kHz.

signal-to-noise ratio for that microphone. Otherwise, we see good

agreement between the two transfer functions up to nearly 5 kHz.
Above that frequency, other types of errors are introduced mainly
due to incorrect positioning of the sound source for reciprocal
measurements.

Conclusions
Sound sources for measuring vibro-acoustic transfer functions
have been investigated, although the emphasis has been on acoustic
transfer functions. The type of source presented here was based
on a powerful driver attached to a long hose equipped with two
microphones close to the orifice for measuring the volume velocity
source strength in situ. Transfer functions measured as FRFs can
then easily be estimated.
The principle was reviewed and some error analysis related to
the current sources was made. Acoustic transfer functions were
measured in a vehicle environment proving that it is possible to
measure binaural transfer functions reciprocally, with some confi-
dence, by placing the orifice close to the entrance of the outer ear.
In this case, a standard HATS and a volume velocity source can be
used to do all operating and transfer function measurements related
to source-path-contribution analysis (including binaural effects).
Additionally, a sound source aimed for mid-to-high-frequency
measurements was investigated and compared to the current low-
to-mid-frequency sound source.

References
1. F. H. van Tol, J. W. Verheij, Brite EuRam II: Loudspeaker for Reciprocal
Measurement of Near Field Sound Transfer Functions on Heavy Road
Vehicle Engines, TNO Institute of Applied Physics, Delft, The Nether-
lands, 1993.
2. S. Gade, N. Møller, J. Hald, L. Alkestrup, “The Use of a Volume Velocity
Source in Transfer Measurements,” Proceedings of the 2004 International
Conference on Modal Analysis, Noise, and Vibration Engineering (ISMA),
Leuven, pp. 2641 – 2648, 2004.
3. J. Y. Chung, D. A. Blaser, “Transfer Function Method of Measuring In-
Duct Acoustic Properties. I. Theory,” Journal of the Acoustical Society
of America, Vol. 68, No. 3, pp. 907-913, 1980.
4. H. Bodén, M. Åbom, “Influence of Errors on the Two-Microphone Method
for Measuring Acoustic Properties in Ducts,” Journal of the Acoustical
Society of America, Vol. 79, No. 2, pp. 541-549, 1986.
5. P. M. Morse, K. U. Ingard, Theoretical Acoustics, McGraw-Hill, New
York, 1968.

The author can be reached at: schuhmache@bksv.com.

www.SandV.com SOUND & VIBRATION/MARCH 2010 11