User manual | TECHNICAL REVIEW

TECHNICAL REVIEW
No. 1 – 2010
BV 0062 – 11
ISSN 0007 – 2621
ËBV-0062---'Î
Time Selective Response Method
In situ Measurement of Absorption Coefficient
Transverse Motion in Accelerometer Calibration
HEADQUARTERS: Brüel & Kjær Sound & Vibration Measurement A/S
DK-2850 Nærum Denmark · Telephone: +45 7741 2000 · Fax: +45 4580 1405
www.bksv.com · [email protected]
Local representatives and service organisations worldwide
Previously issued numbers of
Brüel & Kjær Technical Review
Previously issued numbers of
Brüel & Kjær Technical Review
1 – 2009 Use of Volume Velocity Sound Sources in the Measurement of Acoustic
Frequency Response Functions
Turnkey Free-field Reciprocity System for Primary Microphone Calibration
1 – 2008 ISO 16063–11: Primary Vibration Calibration by Laser Interferometry:
Evaluation of Sine Approximation Realised by FFT
Infrasound Calibration of Measurement Microphones
Improved Temperature Specifications for Transducers with Built-in
Electronics
1 – 2007 Measurement of Normal Incidence Transmission Loss and Other Acoustical
Properties of Materials Placed in a Standing Wave Tube
1 – 2006 Dyn-X Technology: 160 dB in One Input Range
Order Tracking in Vibro-acoustic Measurements: A Novel Approach
Eliminating the Tacho Probe
Comparison of Acoustic Holography Methods for Surface Velocity
Determination on a Vibrating Panel
1 – 2005 Acoustical Solutions in the Design of a Measurement Microphone for
Surface Mounting
Combined NAH and Beamforming Using the Same Array
Patch Near-field Acoustical Holography Using a New Statistically Optimal
Method
1 – 2004 Beamforming
1 – 2002 A New Design Principle for Triaxial Piezoelectric Accelerometers
Use of FE Models in the Optimisation of Accelerometer Designs
System for Measurement of Microphone Distortion and Linearity from
Medium to Very High Levels
1 – 2001 The Influence of Environmental Conditions on the Pressure Sensitivity of
Measurement Microphones
Reduction of Heat Conduction Error in Microphone Pressure Reciprocity
Calibration
Frequency Response for Measurement Microphones – a Question of
Confidence
Measurement of Microphone Random-incidence and Pressure-field
Responses and Determination of their Uncertainties
1 – 2000 Non-stationary STSF
1 – 1999 Characteristics of the Vold-Kalman Order Tracking Filter
1 – 1998 Danish Primary Laboratory of Acoustics (DPLA) as Part of the National
Metrology Organisation
Pressure Reciprocity Calibration – Instrumentation, Results and Uncertainty
MP.EXE, a Calculation Program for Pressure Reciprocity Calibration of
Microphones
(Continued from cover page 2)
(Continued on cover page 3)
1 – 1997 A New Design Principle for Triaxial Piezoelectric Accelerometers
A Simple QC Test for Knock Sensors
Torsional Operational Deflection Shapes (TODS) Measurements
2 – 1996 Non-stationary Signal Analysis using Wavelet Transform, Short-time
Fourier Transform and Wigner-Ville Distribution
1 – 1996 Calibration Uncertainties & Distortion of Microphones.
Wide Band Intensity Probe. Accelerometer Mounted Resonance Test
2 – 1995 Order Tracking Analysis
1 – 1995 Use of Spatial Transformation of Sound Fields (STSF) Techniques in the
Automative Industry
2 – 1994 The use of Impulse Response Function for Modal Parameter Estimation
Complex Modulus and Damping Measurements using Resonant and Nonresonant Methods (Damping Part II)
1 – 1994 Digital Filter Techniques vs. FFT Techniques for Damping Measurements
(Damping Part I)
2 – 1990 Optical Filters and their Use with the Type 1302 & Type 1306 Photoacoustic
Gas Monitors
1 – 1990 The Brüel & Kjær Photoacoustic Transducer System and its Physical
Properties
2 – 1989 STSF – Practical Instrumentation and Application
Digital Filter Analysis: Real-time and Non Real-time Performance
1 – 1989 STSF – A Unique Technique for Scan Based Near-Field Acoustic
Holography Without Restrictions on Coherence
2 – 1988 Quantifying Draught Risk
1 – 1988 Using Experimental Modal Analysis to Simulate Structural Dynamic
Modifications
Use of Operational Deflection Shapes for Noise Control of Discrete Tones
Special technical literature
Brüel & Kjær publishes a variety of technical literature which can be obtained from
your local Brüel & Kjær representative.
The following literature is presently available:
•
•
Catalogues (several languages)
Product Data Sheets (English, German, French,)
Furthermore, back copies of the Technical Review can be supplied as listed above.
Older issues may be obtained provided they are still in stock.
Technical
Review
No. 1 – 2010
Contents
Time Selective Response Measurements – Good Practices and Uncertainty........ 1
Erling Sandermann Olsen and Rémi Guastavino
Measurement of Absorption Coefficient, Radiated and Absorbed Intensity on the
Panels of a Vehicle Cabin using a Dual Layer Array with Integrated Position
Measurement........................................................................................................ 16
J. Hald, J. Mørkholt and S. Gade
ISO 16063–11: Uncertainties in Primary Vibration Calibration by Laser
Interferometry – Reference Planes and Transverse Motion ................................ 28
Torben Licht and Sven Erik Salbøl
TRADEMARKS
PULSE is a trademark of Brüel & Kjær Sound & Vibration Measurement A/S
Copyright © 2010, Brüel & Kjær Sound & Vibration Measurement A/S
All rights reserved. No part of this publication may be reproduced or distributed in any form, or by any
means, without prior written permission of the publishers. For details, contact:
Brüel & Kjær Sound & Vibration Measurement A/S, DK-2850 Nærum, Denmark.
Editor: Harry K. Zaveri
Time Selective Response Measurements –
Good Practices and Uncertainty*
Erling Sandermann Olsen and Rémi Guastavino
Abstract
Time Selective Response, TSR, is a frequency response measurement method
based on linearly swept sine signals. Because TSR can be used for free-field
measurements in ordinary rooms and is fast, accurate and relatively insensitive to
background noise, it is convenient for measurements of microphone and sound
level meter free-field responses. However, the method’s combination of time and
frequency weighting can make it complicated to estimate the uncertainty of the
measured response. This paper briefly summarizes the method and presents some
experience with and guidelines for choosing measurement and weighting
parameters and considerations on the associated uncertainty on the results. The
results are discussed on the basis of practical measurements of microphone and
sound level meter free-field responses at Brüel & Kjær.
Résumé
Une mesure de réponse en fréquence par la méthode TSR (Time Selective
Response) fait intervenir un balayage linéaire de signaux sinusoïdaux. Cette
méthode utilisable dans les pièces ordinaires pour des mesures en champ libre,
rapide et précise, relativement peu sensible au bruit de fond, est pratique pour les
mesures de réponse en fréquence en champ libre des microphones et des
sonomètres. Toutefois, l’application d’une combinaison des pondérations
temporelle et fréquentielle complique quelque peu l’estimation de l’incertitude sur
la réponse mesurée. Ces pages résument la méthode et font part de considérations
pratiques sur le choix des paramètres de mesurage et de pondération ainsi que sur
l’incertitude que ce paramétrage induit sur les résultats. Ces résultats sont discutés
sur la base de mesurages de réponse de fréquence en champ libre de microphones
et sonomètres réalisés à l’usine Brüel & Kjaer.
* First published at INTER-NOISE 2010, Lisbon, Portugal
1
Zusammenfassung
Time Selective Response (TSR) ist eine Methode zur Frequenzgangmessung, die
auf linearer Gleitsinusanregung beruht. TSR ermöglicht Freifeldmessungen in
normalen Räumen und ist schnell, präzise und relativ unempfindlich gegenüber
Störgeräuschen. Damit empfiehlt sich die Methode für die Messung der
Freifeldfrequenzgänge von Mikrofonen und Schallpegelmessern. Die verwendete
Kombination von Zeit- und Frequenzbewertung kann jedoch die
Unsicherheitsbestimmung des gemessenen Frequenzgangs komplizieren. Neben
einer kurzen Beschreibung der Methode präsentiert der Artikel Erfahrungen und
Hinweise in Verbindung mit der Auswahl von Mess- und Bewertungsparametern,
sowie Betrachtungen über die durch sie bedingte Unsicherheit der Ergebnisse. Die
Ergebnisse werden anhand praktischer Messungen der Freifeldfrequenzgänge von
Mikrofonen und Schallpegelmessern bei Brüel & Kjær diskutiert.
Introduction
In 1991, Brüel & Kjær introduced Audio Analyzer Type 2012 with its Time
Selective Response (TSR) measuring algorithm. With the TSR method, the system
responses of electroacoustic devices can be measured reliably in ordinary rooms.
Type 2012 is not produced anymore, but since the release of the PULSE™ 12
Multi-analyzer Platform, TSR has also been available on Brüel & Kjær’s present
family of sound and vibration analyzers.
The TSR method provides a fast and convenient way to perform free-field
measurements in reflective environments, but the combination of time and
frequency weighting in the method can make it complicated to estimate the
uncertainty of the measured response.
This paper presents some guidelines for choosing measurement and weighting
parameters and considerations on the associated uncertainty on the results.
TSR Method
Signal Processing of TSR
Time Selective Response, TSR, is a frequency response measurement method
based on linearly swept sine signals. The method is based on Poletti’s [1, 2] ideas
that solved some issues that could lead to erroneous measurement in earlier
methods. The method is based on an underlying assumption on linearity and
2
invariance of the object response. The impulse response of the system is
determined by combining the inverted excitation signal with the response signal.
Reflections can be excluded from the measurement by selecting the desired part of
the impulse response by weighting with a time window.
The excitation signal in TSR is a complex, linear sweep:
jkt
s t = e
2
(1)
The resulting output signal is:

y  t  = h  t  s  t  =

h   e
jk  t –  
2
d
(2)
–
where h(t) is the impulse response of the object response, the transfer function of
the complete measurement setup and the surrounding room. In the analysis, the
output signal is combined with the inverse sweep:
ut = e
– j kt
= e
2
– j kt
2
y t


h   e
jk  t –  
2
d
–
= e
– j kt
2
(3)

 h   e
2
2
jk  t +  – 2t 
d
–

=

h   e
jk
2
e
– j2kt
d
–
Inserting ξ = kt, this has the form:

u =

h   e
jk
2
e
– j2
d
(4)
–
3
The integral is recognised as the Fourier transform of the product of the system
2
impulse response and the linear sweep, h  t e jkt . Hence, using the convolution
theorem:
ht = e
– j kt
= e
2
– j kt
–1
2
F u
(5)

 u   e
j2t
d
–
From this it is seen that h(t) can be calculated for any point in time from the
complete response. In particular, the time range including the direct sound from
the measurement object can be selected for further calculation. The frequency
response function can then be calculated by Fourier transform of h(t):
H(f) = F{h(t)}
(6)
Conceptually, TSR can be understood as a combination of the swept sine signal
and a tracking filter that follows the signal with a delay so that only the
frequencies arriving at a certain delay are included in the measurement. The
tracking filter is equivalent to the weighting function that defines the selected time
range (Fig. 1).
The TSR algorithm effectively works as a zoom FFT around the centre of the
swept frequency range, i.e., the impulse response that is determined is frequency
shifted to the centre frequency of the sweep. This property means that the sweep
does not need to cover the full frequency range of the target transfer function.
The TSR method requires measurement of the complex response function. If the
measured frequency range is limited so that negative frequencies are not included
in the full sweep, the complete function can be calculated from a single sine
sweep, but if the sweep includes zero or negative frequencies, a cosine and
subsequently a sine sweep must be made in order to obtain the complex function.
Windows in TSR
In addition to the time window that is (or can be) deliberately applied to the time
response in a TSR measurement, there are two window functions inherently
applied to the signal.
4
Fig. 1. Illustration of the concept of a tracking filter, from [3, 4]
d
t = d/c
970 Hz Microphone
1000 Hz
Loudspeaker
940 Hz
920 Hz
970 Hz
Reflecting
surface
Floor
dB
Floor
Reflection
Direct
Sound
Generator
Second
Reflection
t1
t1 + 5 ms
S = 10 Hz/ms
920
940
970
1000
f [Hz]
970
990
1020
1050
f [Hz]
1070
1100
Time/Frequency
Window
t1 + 10 ms
t [s]
1020 1040
f [Hz]
100140
5
The first window inherent to the method is due to the limitation of the frequency
sweep to the specified range. This is equivalent to applying a window function to
an infinitely long frequency sweep:
s  t  = W 1  t e
jkt
2
(7)
The frequency spectrum of the finite sweep is determined by this weighting
function.
The combination with the inverse (unweighted) sweep does not by itself distort
the resulting impulse response.
The second window inherent to the method is due to the limitation of the
analysis to a certain time range, and this is also effectively an application of a
window to the impulse response, the time range mentioned above:
h  t  = W 2  t e
– j kt
2

 u   e
j2t
d
(8)
–
The second window defines the time range that is included in the final Fourier
transform so as to obtain the frequency response. Subsequent application of a
narrower time window in order to exclude reflections is effectively the same as
narrowing the time range, except that the time steps and frequency steps in the
analysis are maintained.
The windows that are applied in Brüel & Kjær’s TSR analyzers are Tukey
windows, i.e., rectangular windows tapered with raised cosine functions. The userselectable time window is a generalised form where the width of the two tapers
can be selected independently.
In order to minimise the influence of the window applied to the sweep, the
actual sweep covers a larger frequency range than that specified for the analysis.
The influence of the windows on the final result not only depends on the
windows themselves, but is also a combination of the windows and the response to
be measured. Therefore, it is not possible to predict the exact uncertainty based on
the measurement parameters alone.
6
Applications of TSR
Since the TSR method provides time selectivity and both the time and frequency
responses are immediately available, it is convenient for a large range of
applications.
The method can be used for accurate and fast comparison calibration of
microphones and sound level meters and measurement of the influence of
accessories, etc. Due to the time selectivity, no anechoic chamber is needed for the
measurements.
The method can be used for simple absolute frequency response measurements
of, for example, loudspeakers. A procedure for combining far-field measurements
with the TSR method with traditional near-field measurements for a loudspeaker is
described in [5].
Another useful application of the TSR method is to use it as an effective means
for localising reflections in a measured response. This can be used in optimising
devices to disturb a sound field as little as possible. In particular this is relevant
during the development of sound level meters, microphone holders, loudspeaker
enclosures and similar devices.
Parameter Choice
Sweep Length
As can be seen from the expressions in “Signal Processing of TSR” on page 2, the
sweep rate, and thus the sweep length, only influences the integration time, not the
resolution of the measured impulse response. This means that the sweep length
only influences the signal-to-noise ratio of the measurement. Some investigation
should be done with the signal-to-noise ratio. If lowering the signal level by 10 dB
significantly changes the measured response, the signal-to-noise ratio should be
improved. If the noise itself cannot be reduced or the signal level increased, there
are two ways to improve the signal-to-noise ratio in the TSR method. Either the
sweep length or the number of averages must be increased. The sweep length does
not have any other influence on the measurement result, provided that the
underlying overall assumption on linearity and invariance of the system is
fulfilled.
Frequency Range
The choice of frequency range in the measurement has only minor influence on
the measurement. This is due to the fact that TSR effectively works as a zoom FFT
7
around the centre frequency of the sweep. The spectrum of the frequency sweep
will be convolved with the object response, but if the weighting function that
defines the sweep is reasonably designed, this will have insignificant influence on
the valid (selected) frequency range. This is illustrated in Fig. 2. Note that the
frequency range shown in the figure includes the part that is generated outside the
specified frequency range so as to minimise the influence of the limitation of the
sweep.
Fig. 2. Simulated absolute response measurement with different frequency ranges. The time
window is the full time range of the analysis
3
500 – 22k
100 – 5k
2
100 – 2k
Response, dB
1
250 – 5k
250 – 10k
0
250 – 22k
100 – 25k
–1
Correct
–2
–3
–4
10
100
1k
Frequency, Hz
10k
100k
100141
Time Window
The time window is a critical parameter in the TSR method (as in any other time
selective method). Ideally, the time window should include the complete time
response and exclude all other information.
In order to include all the desired information, the time window must at least be
sufficiently long so as to include the fluctuations of the object response:
Tmin = 1/fc
(9)
where fc is the necessary frequency resolution. If the time window is too short, it
will behave as a smoothing function on the response. This may be acceptable in
8
some cases at high frequencies, for example in relative measurements where the
resolution is not required in the measured ratio, but if the object response rolls off
at low frequencies, too short a time window may deteriorate a large part of the
object response.
The geometry of the measurement setup should be considered carefully. Simple
measurement of the size of the measurement object helps in determining the
minimum time window, and geometrical consideration can also help in optimising
the path length difference in the setup.
In a normal building, the open height below room lighting, etc., is typically
around 2.5 m. With a measuring distance of 1 m to 2 m, this leaves a difference
between the direct path and the first reflected path of 1.2 m to 1.7 m,
corresponding to a time difference between the direct sound and the first reflection
of 3.5 ms to 4.9 ms. This limits the low-frequency resolution that can be achieved
in time selective measurements in normal-sized rooms. If a higher resolution is
desired, a room must be used where the distance to any reflecting objects is larger.
This may, however, require more bulky mounting devices that again may lead to
less stable mounting and be more susceptible to the movement of air in the room.
A time window allowing measurement of a 3.5 ms long time response will be
sufficient for most applications within microphone and loudspeaker
measurements.
It is important that the time window only covers the part of the object time
response that contains significant energy. In other words, if the complete impulse
response of the measurement object is included in the time window, no further
information can be obtained with a longer time window, and the smoothing
properties of the window do not deteriorate the object function. One important
aspect of this is that the resolution of the result cannot be improved by extending
the time window towards negative time from the impulse. This will only allow
more noise to influence the measurement. This is illustrated in Fig. 3 (note, that
the noise level is deliberately exaggerated in the example).
If it is possible within the limitations of the reflections, different measurement
distances and time windows should be tried in order to resolve whether the
complete impulse is included in the measurement. The necessary length of the
time window can be determined experimentally by gradually increasing the length
until the shape of the measured response does not change.
9
Fig. 3. Noise influence. Left: Time window covering the impulse response. Right: time window
extended towards negative time
–23
–23
–24
–24
–25
–25
–26
–26
–27
–27
–28
–28
–29
–29
–30
100
1k
10k
–30
100
1k
10k
100142
Uncertainty Estimation
Uncertainty estimation of frequency response measurements made with the TSR
method is, of course, similar to that of any other frequency response measurement
method, except for the aspects that are particular to the method. Here, the aspects
related to TSR are discussed. Note, that these aspects are to some extent the same
for other time selective methods, because the underlying principle of determining
the impulse response and selecting a certain part is common to all these methods.
As mentioned and demonstrated, it is not possible to predict the exact
uncertainty based on the measurement parameters alone. Some evaluation can be
made on the maximum variation that can be anticipated, but that is likely to lead to
overestimation of the uncertainty. Instead, the evaluation should be based on
controlled variations or realistic modelling of the actual measurement situation.
In the case of the measurement of a single response, for example, in a
loudspeaker measurement, the influence of the windows on the response
contribute directly to the uncertainty of the measurement. If some knowledge is
available of the object response, the influence of the parameters can be estimated
by modelling. Alternatively, the measurement can be repeated with variations of
the parameters. It should, however, be kept in mind that there may be systematic
deviations that are not revealed by the possible parameter variations.
In the case of relative measurements where the response of the measurement
object is compared with the response of a reference, for example, in microphone
comparison calibration or measurement, some of the deviations due to the
10
windows cancel out because they are present in both measurements. In this case
there are also considerable possibilities of making the measurements under
different conditions both in the physical measurement setup and in the
measurement parameters, thereby achieving knowledge on the uncertainty in the
measurements.
In a frequency range where the reference and the object under test have similar
and reasonably flat frequency responses, the low-frequency resolution does not
hinder an accurate determination of the difference between the two, and the
uncertainty on the determined frequency response can be relatively small if care is
taken to avoid systematic errors.
Examples
In this section, a couple of examples of these measurements are used for
illustration of the issues mentioned above. The TSR method is used extensively at
Brüel & Kjær for measurements of free-field responses. It is the company’s policy
to thoroughly document all electroacoustic devices developed, and the TSR
method has worked as an efficient and accurate tool for that purpose for several
years.
Microphone Free-field Response Measurements
Free-field response measurements of microphones at Brüel & Kjær are exclusively
made using the TSR method. Absolute free-field sensitivities are measured by
comparison with reference microphones that are traceable to Danish national
standards. Measurements of the influence of accessories such as wind screens,
protection grids, etc., and directional characteristics are relative measurements
where only the stability of the measurement object and the measurement setup is
relevant. However, the measurements are always relative measurements. The freefield measurements with the TSR method are made at frequencies above 500 Hz.
Below 500 Hz the free-field response of the microphones can reliably be
established with other methods that are not the subject of this paper.
In order to ensure high accuracy and reproducibility in the measurements, some
practices have been developed. Although these are not directly related to the TSR
method, they are mentioned here in order to demonstrate how the TSR method is
used.
Many of the measurements are carried out in a measurement room situated in a
normal office building. Some factors such as circulating air, for example, due to
ventilation systems, can disturb these measurements. Small pressure pulses are
11
generated in the building due to opening doors or the like and changing
temperature causes changing speed of sound, detectable even when the changes
are fractions of a degree. In order to minimise the variations in the measurement
results due to these factors, the ventilation system is turned off during each
response measurement that consists of six TSR sweeps. This can be done in a
minimum of time. The six sweeps allows the identification of occasional (but rare)
variations in the measurements as shown in Fig. 4. The practice also gives
supplementary information to the uncertainty estimation.
Fig. 4. Variation of six consecutive sweeps. One of the sweeps has been disturbed
0.2
0.15
Variation, dB
0.1
0.05
0
–0.05
Ignored in average
–0.1
–0.15
–0.2
100
10k
1k
Frequency, Hz
100k
100143
Sound Level Meter Free-field Response
The authors have recently worked with TSR measurements of sound level meter
free-field frequency responses in order to establish well-documented uncertainty
estimates on the free-field response of sound level meters.
Although the measurements in principle are similar to those of microphone freefield responses, the measurement of the free-field responses of large devices such
as sound level meters and outdoor microphones is particularly challenging. The
time window must be large in order not to exclude parts of the devices, and it is
often difficult to mount the devices for measurement without having reflections
from the mounting setup.
12
The authors’ measurements are carried out in different rooms, with different
loudspeakers, at different distances and with different frequency ranges. Fig. 5
shows the responses of the sound level meter and of the microphone on a long
cylinder, and Fig. 6 shows the double standard deviation of the responses. As can
be seen from the figures, it is indeed possible to achieve good reproducibility of
these response measurements. Work is still going on to improve the measurement
procedure and the data analysis, and has also been a part of the preparation of this
paper.
Fig. 5. Measured free-field responses of a sound level meter and the sound level meter
microphone carried out in widely different setups. Upper curves are for the microphone alone.
Lower curves are for the microphone mounted on the sound level meter. The orange curve
shows the low-frequency response in an enclosure
–25
Response, dB
–25.5
–26
–26.5
–27
–27.5
100
1k
10k
Frequency, Hz
100k
100144
13
Fig. 6. 2σ of measurements carried out in widely different setups. The blue curve is for the
microphone alone; the red curve is for the microphone mounted on the sound level meter.
The results shown are considered valid from 200 Hz
0.5
0.4
0.3
Respnse, dB
0.2
0.1
0
–0.1
–0.2
–0.3
–0.4
–0.5
100
1k
10k
Frequency, Hz
100k
100145
Discussion and Recommendations
The purpose of this paper is to give the reader some understanding of the TSR
method and to provide guidelines and recommendations to the use of the method.
It should be mentioned that the influence of the windows in the time and
frequency domains discussed here are of a general nature and exist in all time
selective methods. Actually, the influence of the windows in TSR is relatively
easy to understand compared to methods applying non-linear sweeps.
Our recommendations are
• Make the window as long as possible, but not longer than necessary
• Object low-frequency behaviour determines the necessary window length
• Low frequencies – go closer and increase time window
• Increasing the sweep time increases signal-to-noise ratio, not resolution
• Always try with different window sizes
• Always repeat measurements
• Large rooms are preferable, whether they are anechoic or not
14
• The air must be stable (easier in ordinary rooms than in anechoic chambers)
• Carefully consider the reflections and which to include in the time window
Routine measurements do not, of course, need to be verified each time.
Summary
In this paper, the Time Selective Response method has been summarized and the
influence of some choices of measurement parameters and their interaction with
the object response have been discussed.
The importance of the right selection of time weighting function has briefly
been demonstrated. It has also briefly been demonstrated that the actual frequency
range of the measurement is of small importance.
Some guidelines on how to evaluate the measurement uncertainty due to the
time and frequency limitations in the measurements have also been given.
References
[1]
Poletti, M. A., “Linearly Swept Frequency Measurements, Time-Delay
Spectrometry, and the Wigner Distribution”, Journal of the Audio
Engineering Society, 36 (6), 1988, pp 457 – 468.
[2]
Struck, C. J., Biering, C. H., “A New Technique for Fast Response
Measurements Using Linear Swept Sine Excitation”, 90th Convention of
the Audio Engineering Society, New York, USA 1991, preprint 3038.
[3]
PULSE™ Time Selective Response (Brüel & Kjær online documentation),
Brüel & Kjær, 1995.
[4]
Brüel & Kjær, Audio Analyzer Type 2012 Technical Documentation,
Brüel & Kjær, BE 1119 – 13, 3–6 to 3–13, 1995 (ca.).
[5]
Struck, C. J., Temme, S. F., “Simulated Free Field Measurements”, 93rd
Convention of the Audio Engineering Society, New York, USA, 1992,
preprint 3397.
15
Measurement of Absorption Coefficient,
Radiated and Absorbed Intensity on the
Panels of a Vehicle Cabin using a Dual Layer
Array with Integrated Position Measurement*
J. Hald, J. Mørkholt and S. Gade
Abstract
In some cases it is important to be able to measure not only the total sound
intensity on a panel surface in a vehicle cabin, but also the components of that
intensity due to sound radiation and due to absorption from the incident field. For
example, these intensity components may be needed for calibration of energy flow
models of the cabin noise. A robust method based on surface absorption
coefficient measurement is presented in this paper.
Résumé
Dans certaines situations, il est important de pouvoir mesurer non seulement
l’intensité acoustique totale sur un panneau de l’habitacle d’un véhicule, mais
également, du fait de phénomènes de rayonnement sonore et d’absorption
caractérisant le champ incident, les composantes de cette intensité. Ces
composantes peuvent notamment s’avérer nécessaires pour le calibrage des
modèles de flux d’énergie du bruit généré dans l’habitacle. Une méthode robuste,
basée sur la mesure du coefficient d’absorption de la surface, est présentée ici.
Zusammenfassung
Manchmal ist es wichtig, dass man an einer Innenfläche in der Fahrzeugkabine
nicht nur die Gesamtschallintensität misst, sondern außerdem ermitteln kann,
welcher Anteil der Intensität auf Schallabstrahlung von der Fläche zurückzuführen
ist und welcher auf die Absorption von Energie aus dem einfallenden Schallfeld.
Diese Intensitätskomponenten werden beispielsweise zur Kalibrierung von
* First published at JSAE Annual Congress 2010, Yokohama, Japan
16
Energieflussmodellen des Kabinengeräuschs benötigt. Dieser Artikel stellt eine
robuste Methode vor, die auf der Messung des Absorptionskoeffizienten von
Flächen beruht.
Introduction
Consider the radiation of sound from a small surface segment in a cabin
environment. Such a surface segment may radiate sound energy because of
external forcing, causing the surface to vibrate, and it may absorb energy from an
incident sound field because of finite surface acoustic impedance. When
measuring the sound intensity over the surface segment with an intensity probe,
the total intensity Itot will be measured. Assuming the radiated field and the
incident field to be mutually incoherent, the total intensity is equal to the sum of
the radiated sound intensity Irad that would exist with no incident field and the
sound intensity component Iabs due to absorption from the incident field:
Itot = Irad + Iabs
(1)
Considering the intensity in the outward normal direction on the panel surface,
the radiated intensity will typically be positive, while the component due to
absorption will typically be negative. So for vibrating panels with an absorptive
surface, the total measured intensity may be small although the radiated intensity
is relatively high. Often it is of interest to know not only the total intensity, but
also the components due to radiation and absorption. For example, this kind of
information is needed in energy-based modeling that describes the energy flow
between sub-systems [1].
The method presented here is based on separation of different sound field
components via the spatial sound field information provided by an array. The
radiated intensity is estimated as the intensity that would exist, if the incident and
scattered field components could be removed. So a free-field radiation condition is
simulated. The idea is to first separate the incident field component, then use
separately measured information about the scattering properties of the panel to
calculate the scattered field, and finally subtract the incident and scattered fields
from the total sound field. The method needs the panel geometry, which is either
imported as CAD or measured with a 3D digitizer, and then uses a measured map
of absorption coefficient. The separate measurement of the surface absorption
coefficient is obtained using loudspeaker excitation.
17
Description of Methodology
The array measurements considered here are cross-spectral measurements of the
full cross-spectral matrix between all array microphones and, based on that, a
representation in terms of Principal Components is extracted [2]. As a
consequence, no phase information is available between different array positions,
so a separate processing has to be performed for each position. For each Principal
Component a separate holography calculation is performed. Since these
calculations are identical only one will be described. We used a Double Layer
Array (DLA), and the processing was performed using Statistically Optimized
NAH (SONAH) [3, 4]. A complex time harmonic representation, with the time
variation e jωt suppressed, will be used.
Extraction of the Incident Field
A basic array processing task is that of extracting the incident sound field from the
total sound field. However, considering the sound field on a small panel segment
in a cabin, the distinction between the incident field and the radiated field is not
obvious, even when we look at the field very near the panel segment. Because of
coherent vibration components and significant mutual radiation impedances
between neighbouring panel segments, some neighbouring segments should be
included as sources of the radiated field. Fig. 1 illustrates how this distinction can
be made in practice with a DLA. Using SONAH on a DLA measurement, the
sound field components (p, u) and (p+, u+) with sources on different sides of the
array plane can be separated, [3, 4]:
(ptotal, utotal) = (p–, u–) + (p+, u+)
(2)
The array must then be used in such a way that the pressure p and the velocity
vector u represent the field incident on the source area of interest, while the
outward propagating field component (p+, u+) is the sum of the scattered and
radiated fields:
(p+, u+) = (psct , usct) + (prad, urad)
(3)
Fig. 1a illustrates the case of an isolated source object, where the incident and
outward propagating field components are well-defined. They could be
determined, for example, based on measurements across two concentric spherical
surfaces that enclose the test object. Microphones on the two surfaces would be
18
pointing to the centre, but facing in opposite directions. Fig. 1b illustrates the case
of measurement on a panel section in a cabin. In this case, the measurement plane
will define the distinction between sources of the incident field and sources of the
outward propagating field. Here a DLA (see Fig. 3) could be used, where
microphones are mounted back to back. The distinction will, however, not be
sharp due to the limited angular resolution of a practical array.
Fig. 1. Separation of inward (incident) and outward propagating field components (p–, u–) and
(p+, u+): a) Clearly separated sources. b) Smooth transition between sources
Cabin
Wall
a)
b)
Measurement
plane
(p–, u–)
DLA
(p+, u+)
Object
Mapping
area
Cabin
Wall
100146
Solution of the Scattering Problem
Provided the incident and radiated fields are mutually incoherent, then Eq. (1) is
the basis for a simple and robust energy based method. The basic assumption is
that we can measure a local absorption coefficient  at each point r on the source
surface such that the normal components I – and Iabs of the incident and absorbed
intensities are related by:
Iabs (r) =  (r) I – (r)
(4)
In general the ratio Iabs(r)/I –(r) between the two intensities will depend on the
form of the incident field. If, however, the coefficient  (r) is measured with an
incident field that is sufficiently similar to the incident field under operational
conditions, then Eq. (4) can provide good results with the operational field.
The method requires a separate set of DLA measurements with artificial speaker
excitation. Fig. 2 illustrates a setup with a set of incoherently excited speakers to
create an incident field similar to the field incident under, for example, flight
conditions in an aircraft. As mentioned already, the DLA/SONAH measurement
19
Fig. 2. Cabin panel section with surface calculation points covered by a specific array
position. Speakers for measurement of surface properties are shown
100147
can provide the total and incident components of the loudspeaker generated sound
field on the panel surface. Since in this case the absorbed intensity is equal to the
total intensity, then in accordance with Eq. (4) the absorption coefficient is
calculated as the ratio between the total and the incident intensities. Accurate
measurement of the array positions relative to the panel geometry using an
integrated position measurement system allows the absorption coefficients to be
calculated at predefined points on the panel surface, see Fig. 2.
For the case of car cabin application, the surface property measurement will
typically be performed with the car at a standstill, while the operational
measurement will be performed during driving. Often, a single surface property
measurement will be applied with several operational measurements,
corresponding to different operational conditions.
Once an operational measurement has been taken, the associated total, incident
and outgoing field components on the panel surface can be estimated using
SONAH. The absorbed intensity is estimated using Eq. (4), and the radiated
intensity is then obtained using Eq. (1).
Measurement
Measurement System
The system for data acquisition is based around the DLA shown in Fig. 3. The
array has 8 × 8 microphones mounted in 2 layers, resulting in a total of 128
20
Fig. 3. Double-layer array (DLA) mounted with six infrared (IR) LEDs
microphones. The microphones are spaced 30 mm apart in both directions and
with a spacing of 31 mm between the two layers. This results in an upper
frequency limit for the array of 5.7 kHz (spatial sampling limit). Also mounted on
the array are six infrared LEDs, which are used in connection with a position
measurement system monitoring the array position and orientation on-line. This
system is based around an integrated camera unit. The camera unit has 3 built-in
line cameras that determine the LED positions. The camera unit is connected to a
controller box which communicates with a PC via RS–232. The LEDs on the
array are also connected to this controller box. The position measurement system
also has associated with it a wireless pointing device for measuring 3D point
positions. This device can be used in connection with the measurement software to
record 3D position data on the surfaces under investigation.
The DLA is connected to two 65-channel front-ends which are connected to the
controlling PC via LAN. A third front-end with generator units is used for driving
artificial excitation speakers through power amplifiers.
The controlling PC is running measurement software with a dedicated
measurement template for acquisition of array microphone signals and
corresponding array positions in space, as well as surface point position data. As
outlined above, the analysis method requires a model of the surfaces to be
analysed. Such a model can either be imported from CAD or digitized on the spot.
Digitizing involves point measurements as outlined above, followed by curve and
21
surface construction. In both cases, the end result is a CAD-type parametric
surface description which is later used as the basis for creating a surface mesh.
This mesh is in turn used for calculation and results display.
Measurement Procedure
The DLA is placed in positions close to the surfaces under investigation to sample
the near field sound pressure. The position of the array is recorded online via the
position measurement system and microphone time histories are recorded for each
array position. The array is placed in slightly overlapping positions to cover the
investigated surfaces patch by patch. The data acquisition software displays the
current and past array positions in 3D along with the surface model. In this way
the user can identify which parts of the surface have been covered with the array
and which parts still need to be covered.
When measuring for the estimation of surface absorption, a number of
loudspeakers are distributed in the cabin interior and driven by uncorrelated noise
sources, to create a distributed and (close-to) diffuse excitation field. Measurement
for the actual estimation of entering intensity is, of course, performed in operating
conditions with these sources turned off. All recorded data are stored to a database.
Data Processing
The post-processing software application retrieves all data from the database. A
surface mesh is created based on the parametric surface description. A number of
mesh areas may be defined in this process. These can be used for averaging of, for
example, absorption coefficient.
The time histories measured in each array position are processed as shown in
Fig. 4. First the full cross-spectral matrix, CSM, for all signals is calculated by
FFT and the frequency responses corrected with response correction data from the
individual microphones using TEDS. Then a principal component decomposition
(PCD) is performed to determine the most significant incoherent components.
Each component is finally processed via SONAH to determine the corresponding
incoming and outgoing sound field quantities on the source surface.
From this point on, the processing depends on the quantity to be determined.
The case of absorption coefficient estimation is illustrated in Fig. 5. Based on the
partial sound field quantities, the total incoming and outgoing surface intensity
components are calculated for surface points corresponding to each array patch.
Results from different patches are then averaged on the surface mesh using a
scoring method, wherein each mesh node-to-array position combination is
assigned a score depending on how the node is positioned relative to the array. If
22
Fig. 4. Processing steps for each array position
(p+,u+), (p–,u–)
Array Pos #1
FFT
CSM
TEDS
PCD
SONAH
(p+,u+), (p–,u–)
Array Pos #2
FFT
CSM
TEDS
PCD
SONAH
100148
Fig. 5. Processing for absorption coefficient estimation
Surface Data #1
Intensity
Calc.
(p+,u+), (p–,u–)
I+,I–
Scored
Averaging
Surface Data #2
I+,I–
Intensity
Calc.
(p+,u+), (p–,u–)
I+,I–
Band
Synthesis
Absorption
Calc.
I+,I–
α
100149
more than one array position contributes to a node, the results for that node are
averaged using the scores. In this way, a contribution estimated with a “wellpositioned” array will be given more weight in the averaging. The surface
intensity values are then band-synthesised into, for example, 1/3-octave bands and
optionally averaged over predefined averaging areas. Finally, the absorption
coefficient is estimated in each node using Eq. (4).
23
Example: Mapping of Absorption Coefficient in a Car
Cabin
To illustrate the use of the proposed techniques in an automotive application,
measurements were made with the DLA system in the cabin of a Volvo S60
passenger car to determine the in-situ absorption coefficient of selected surfaces in
the cabin. Firstly, the cabin surfaces to be investigated were digitized using the 3D
position measurement system and dedicated digitizing software. Next, array
measurements were made with the DLA covering the surfaces patch by patch.
Four loudspeakers were distributed in the cabin and driven by white noise to
provide the acoustic excitation needed for the estimation of the absorption
coefficient.
Fig. 6. A contour plot of the estimated absorption coefficient of seat, door, window and roof in
a car cabin. Results are shown for the 200 Hz 1/3-octave band. Note that results are averaged
also over the respective areas
Fig. 6 shows a 3D contour plot of the estimated absorption coefficient on the
cabin surfaces for the 200 Hz 1/3-octave band. The absorption coefficient was
estimated by first doing 1/3-octave band synthesis of the estimated total and
incident intensities, and then doing area averaging of these quantities over, for
example, the seat or window surface before estimating the final absorption
coefficient as the ratio between the two. The figure shows that in the 200 Hz
24
frequency band, the seat has quite a high absorption coefficient compared to the
door, window and roof.
Fig. 7. Estimated absorption coefficient on the roof, seat and window of the car cabin as
functions of frequency. Results are shown in 1/3-octave bands
Fig. 7 shows the estimated average absorption coefficients in 1/3-octave bands
for the roof (A), the seat (B) and the window (C). The roof was covered by a thin
layer of foam material, which could be expected to have an absorption coefficient
increasing with frequency, as the graph shows. The seat shows an absorption
coefficient decreasing with frequency, which could be explained by the seat being
a leather type of material that is probably acoustically “hard” at higher
25
frequencies. Finally, the window shows quite a low absorption coefficient
throughout the whole frequency range under investigation, as could be expected.
Conclusion
A method for measuring the absorption coefficient, radiated and absorbed
intensity on the panels of a vehicle cabin has been described and it was shown how
the method can be used to map, for example, the absorption coefficient on the
interior surfaces of a car cabin. The method is based on a Dual Layer Array with
integrated position measurement. The method has shown good ability to determine
the actual radiated sound intensity in the presence of a diffuse field in the cabin.
Aknowledgements
This paper is based on work done in the European Union research project Cabin
noise Reduction by Experimental and numerical Design Optimization, CREDO,
which deals with noise in aircraft and helicopter cabins. Since the radiated
intensity is often seen from the perspective of the cabin, it is sometimes also called
Entering Intensity (EI) [4]. Simulations as well as results of measurement inside
aircraft can be found in [4, 7].
26
References
[1]
Hardy, P., Trentin, D., Jézéquel, L., Ichchou, M. N., “Identification of noise
sources in an in-flight aircraft by means of local energy method ”;
Proceedings of Euronoise 2006.
[2]
Hald, J., “STSF – a unique technique for scan-based Near-field Acoustical
Holography without restrictions on coherence”; Brüel & Kjær Technical
Review, No. 1, 1989.
[3]
Hald, J., “Patch holography in cabin environments using a two-layer
handheld array with an extended SONAH algorithm”; Proceedings of
Euronoise 2006.
[4]
Hald, J., Mørkholt, J., “Methods to estimate the Entering Intensity and their
implementation using SONAH ”; Report DWP 2.3 from the European
project CREDO, 2007.
[5]
Alvarez, J. D., Jacobsen, F., “In-situ measurements of the complex acoustic
impedance of porous materials”; Proceedings of Inter-Noise 2007
[6]
Hald, J., Mørkholt, J., “Array based Measurement of Radiated and
Absorbed Sound Intensity Components”; Proc. Acoustics ’08, 2008
[7]
Hald, J., Mørkholt, J., Hardy, P., Trentin, D., Bach-Andersen, M., Keith, G.,
“Measurement of Absorption Coefficient, Surface Admittance, Radiated
Intensity and Absorbed Intensity on the Panels of a Vehicle Cabin using a
Dual Layer Array with Integrated Position Measurement ”; Proceedings of
SAE, 2009
27
ISO 16063–11: Uncertainties in Primary
Vibration Calibration by Laser Interferometry
– Reference Planes and Transverse Motion*
Torben Licht and Sven Erik Salbøl
Abstract
Primary vibration calibration by laser interferometry using quadrature outputs has
been used for the last 10 to 15 years. ISO 16063–11 [1] was published in 1999 and
has increased the interest further. With the advent of new compact laser
interferometers, the difficulties of optical alignment and adjustment have been
practically eliminated, and dedicated software has made the process automatic,
which facilitates collection of much more data.
In most cases the method is applied to reference transducers, either single-ended
or those meant for back-to-back calibration. Because the laser beam cannot always
be directed towards an ideal point or surface, significant errors can be introduced.
At low frequencies this is often due to non-ideal exciter motions. At high
frequencies it is often due to relative motion between points on apparently rigid
mechanical structures, or rocking or bending motion of the combined structures.
Some examples and solutions to these problems, including uncertainty
calculations, will be presented in this paper.
Résumé
L’étalonnage primaire de vibrations par interférométrie laser avec sortie en
quadrature est utilisé depuis une quinzaine d’années. La Norme ISO 16063–11 [1]
publiée en 1999 connaît un intérêt croissant. Avec l’arrivée de nouveaux
interféromètres de laser très compacts, les difficultés liées au réglage et à
l’alignement optique ont été pratiquement éliminées, et des logiciels dédiés ont
automatisé les procédures, facilitant l’acquisition d’une plus grande quantité de
données.
* First published at IMEKO XIX World Congress: Fundamental and Applied Technology, Lisbon,
Portugal
28
Dans la plupart des cas, cette méthode est appliquée pour les capteurs de
référence, qu’ils soient accélerometre de transfert monobase ou destinés à
l’étalonnage dos-à-dos. Comme le faisceau laser ne peut pas toujours être dirigé
vers un point ou une surface idéal(e), des erreurs non négligeables peuvent
apparaître. Aux basses fréquences, cela est souvent dû à un mouvement non idéal
de l’excitateur. Aux hautes fréquences, au déplacement relatif entre points sur des
structures mécaniques apparemment rigides, ou au balancement ou fléchissement
de structures combinées. La présente communication exemplifie ces problèmes et
leurs solutions, y compris les calculs de l’incertitude.
Zusammenfassung
Primärkalibrierung von Schwingungssensoren durch Laserinterferometrie mit
Quadraturausgängen wird seit ca. 10 – 15 Jahren verwendet. Die 1999
veröffentlichte ISO 16063–11 [1] hat das Interesse weiter gesteigert. Mit den
neuen kompakten Laserinterferometern wurden die Schwierigkeiten bei der
optischen Ausrichtung und Justierung in der Praxis eliminiert. Durch
Spezialsoftware wurde der Prozess automatisch gemacht, so dass sich heute
wesentlich mehr Daten sammeln lassen.
In den meisten Fällen wird die Methode auf Bezugsnormalsensoren angewendet,
entweder single-ended oder solche, die für Back-to-Back-Kalibrierung vorgesehen
sind. Da sich der Laserstrahl nicht immer auf einen idealen Punkt oder eine ideale
Fläche richten lässt, können wesentliche Fehler entstehen. Bei tiefen Frequenzen
sind diese häufig auf nichtideale Bewegungen des Schwingerregers
zurückzuführen, bei hohen Frequenzen dagegen auf die Relativbewegung
zwischen Punkten auf scheinbar steifen mechanischen Strukturen oder auf
Schaukel- oder Biegebewegungen der kombinierten Strukturen. In diesem Artikel
werden einige Beispiele und Lösungen für diese Probleme vorgestellt, darunter
Unsicherheitsberechnungen.
Introduction
ISO 16063–11, Methods for the Calibration of Vibration and Shock Transducers –
Part 11: Primary Vibration Calibration by Laser Interferometry, does not go into
much detail when the quality of the motion is described. The standard states:
“Transverse, bending and rocking acceleration: Sufficiently small to prevent
excessive effects on the calibration results. At large amplitudes, preferably in the
29
low-frequency range from 1 Hz to 10 Hz, transverse motion of less than 1 % of the
motion in the intended direction may be required; above 10 Hz to 1 kHz, a
maximum of 10 % of the axial motion is permitted; above 1 kHz, a maximum of
20% of the axial motion is tolerated.”
About the calibration of back-to-back accelerometers it states:
“Typically, a 20 g mass is used. The laser light spot can be at either the top
(outer surface) of the dummy mass or the top surface of the reference
accelerometer.
If the motion is sensed at the top of the dummy mass, then the dummy mass
should have an optically polished top surface, and the position of the laser-light
spot should be close to the geometrical centre of this surface. In cases where the
motion of the mass departs from that of a rigid body, the relative motion between
the top (sensed) and bottom surfaces shall be taken into consideration. To simulate
a mass of 20 g of typical transfer standard accelerometers, a dummy mass in the
form of a hexagonal steel bar 12 mm in length and 16 mm in width over flats of
hexagonal faces can be used. At a frequency of 5 kHz, for example, the relative
motion introduces systematic errors of 0.26 % in amplitude measurements and
4.2° in phase shift measurements”.
The standard also states uncertainties attainable as:
“For the magnitude of sensitivity:
0.5 % of the measured value at reference conditions;
< 1 % of the measured value outside reference conditions.”
The standard was mainly written to help national metrology institutes (and
similar laboratories) to work in similar ways. These have, since the publication of
the standard, claimed far better uncertainties (0.1 to 0.5 % in the full frequency
range).
In these cases, the motion has normally been measured at the reference surface
of the reference accelerometer, and a slotted mass has been used.
This means that the point(s) at which the motion is measured by the
interferometer are positioned at a distance of some millimetres from the centre.
This makes the transverse motion, which is normally a rocking motion, a major
contributor to the uncertainty.
To avoid this, a dummy mass, as described above, can be used and the motion
measured at the centre. However, this leaves the problem of estimating (with great
accuracy) the motion at the interface between the dummy mass and the reference
accelerometer.
30
Transverse Motion of Exciters
Manufacturers of exciters do sometimes give specifications for transverse motion,
but the way these are obtained is rarely given and they are probably measured
using mostly small, carefully symmetrical loads, placed directly on top of the
moving element. Therefore, an investigation of actual transverse motion was
performed on three different exciters. One was a classical spring controlled model,
the other two were air-bearing exciters with a specified low transverse motion.
Exciters A and B are air-bearing types; exciter C is a classical spring type.
Measurements on A and B were made orthogonal to the excitation direction in two
different directions (X and Y, 120 degrees apart) at the top end of a reference
transducer (Brüel & Kjær Type 8305), with a 20 gram slotted load-mass at the
mounting surface of the accelerometer (Fig. 1). (The slotted load-mass is of PTB
design and is normally used for laser interferometry, Fig. 24.) On exciter C, Type
8305 was loaded with a cubic 30 gram block with four identical accelerometers
mounted on its surfaces to measure the transverse motion (Fig. 2). Measurements
from two accelerometers orthogonal to each other are made use of.
Fig. 1. Slotted mass on Type 8305
Fig. 2. Cubic block on Type 8305
The results are plotted in three groups: Figs. 3 – 8 are for exciter A, Figs. 9 – 14
for exciter B, and Figs. 15 – 20 for exciter C. The excitation was a random signal
and the spectrum obtained as measured by the Type 8305 is shown in the bottom
left-hand graph for each of the three groups mentioned above. The transverse
motions relative to the main direction for each shaker are shown as the first two
curves on the left-hand side of each group. The maximum on the graphs
corresponds to 100% transverse motion. From the results, it became clear that the
specifications for transverse motion are not useful for estimating the motion and
uncertainties related to transverse motion when a back-to-back reference
transducer is mounted directly on top of the exciter table. The combined structure
of the exciter table and the reference transducer shows high transverse levels in the
31
4 to 9 kHz range, probably due to bending modes. A small asymmetrical load
(introduced by the connector) can apparently create large rocking motions,
typically with peaks of 30 to 60%.
Fig. 3. Exciter A, X-direction, 8305, 20 gram
Fig. 4. Exciter A, X-direction, 8305, 20 gram,
with mechanical filter
Fig. 5. Exciter A, Y-direction, 8305, 20 gram
Fig. 6. Exciter A, Y-direction, 8305, 20 gram,
with mechanical filter
Fig. 7. Exciter A, excitation spectrum
Fig. 8. Exciter A, excitation spectrum, with
mechanical filter
32
Fig. 9. Exciter B, X-direction, 8305, 20 gram
Fig. 10. Exciter B, X-direction, 8305,
20 gram, with mechanical filter
Fig. 11. Exciter B, Y-direction, 8305, 20 gram
Fig. 12. Exciter B, Y-direction, 8305,
20 gram, with mechanical filter
Fig. 13. Exciter B, excitation spectrum
Fig. 14. Exciter B, excitation spectrum, with
mechanical filter
33
Fig. 15. Exciter C, X-direction, 8305, 20 gram
Fig. 16. Exciter C, X-direction, 8305,
20 gram, with mechanical filter
Fig. 17. Exciter C, Y-direction, 8305, 20 gram
Fig. 18. Exciter C, excitation spectrum, with
mechanical filter
Fig. 19. Exciter C, excitation spectrum
Fig. 20. Exciter C, excitation spectrum, with
mechanical filter
34
A number of experiments were performed in order to investigate this
phenomenon. The best solution was found to be a mechanical filter mounted
between the exciter and the reference transducer. To give the best results, the filter
should be optimised (preferably) for the specific type of reference transducer. The
principle of the mechanical filter is shown in Fig. 21. The filter consists of two
stainless steel parts connected by vulcanized rubber.
The resulting transverse motion is now reduced to less than 5% above 1 kHz for
all the exciters, as shown by the bottom right-hand graph in each of the three
groups. The results from using such a filter are shown as the curves on the righthand side of each group (Fig. 3 – Fig. 20).
Fig. 21. Cross-section of the mechanical filter
090131
Transverse Motion Influence on Calibration Results
To avoid, or cancel out the influence of the rocking motion, the laser beam can be
aligned with the centre line of the accelerometer (provided the rocking centre is
also aligned on that line), or the average of two or more measurements can be
made at symmetrical points about the centre line. However, rocking will always
influence the measurement due to point positioning accuracy and the unknown
centre line of the motion. To measure the importance of this, a series of
measurements close to the critical frequencies were made. The results are shown
in Fig. 22.
It can be seen that the difference between two diagonal points can be nearly
± 10%, without the filter, compared to about ±2% when the filter is used. That
makes a big difference when the average has to be found. The influence can
change the calculated uncertainties dramatically. If an off-centre position of 1 mm
is used together with a distance to the rocking centre of 50 mm, then changing the
transverse motion from 5% to 25% will increase the estimated 2 uncertainty of
0.4% to more than 0.8%.
35
Fig. 22. Measurement differences with and without mechanical filter (20 gram slotted mass)
Sensitivity [pC/ms2]
Calibrations 180 degrees apart on 8305
with and without mechanical filter
0.136
0.134
0.132
0.13
0.128
0.126
90deg_filter_PTB_20gram
270deg_filter_PTB_20gram
0.124
0.122
90deg_NO_filter_PTB_20gram
0.12
0.118
0.116
270deg_NO_filter_PTB_20gram
0
4000
2000
6000
8000
10000
090132
Frequency [Hz]
To further prove the validity of the method, a comparison was made to a wellknown reference transducer in Germany, see Fig. 23.
The figure shows the result of the comparison between a calibration performed
at PTB, Germany, and the calibration made at DPLA* using a mechanical filter
and four points with the slotted load-mass shown in Fig. 24.
The uncertainties from the PTB certificate (0.2%, 0.3% and 0.4%) were used
together with the presently stated DPLA uncertainties of 0.4% and 0.6% (above
5 kHz).
Fig. 23. En values comparing a recently PTB calibrated Type 8305 with a calibration at DPLA
using the mechanical filter
June 2009. 8305 S/N 606545 with 20 gram slotted mass
En values. Sensitivity magnitude between DPLA and PTB
PTB cal. 2009-02-18. PTB-1048-2009
1
En
0.5
0
-0.5
-1
10
100
1000
Frequency [Hz]
* Danish Primary Laboratory for Acoustics
36
10000
090133
Fig. 24. Stainless steel 20 gram slotted load-mass
20
4
Height
9.5 mm
5
10
090130
Influence of Load-Mass on Calibration Results
For the last twenty years (or more) a number of correction curves have been used
for back-to-back reference transducers. Recently, some astonishing new results
were published by PTB [2].
Fig. 25. Difference between measurement on top of load-mass and at the transducer surface
Influence Type 8305 Calibration - spot on centre on 20 g mirror Mass WS 3790
rel. spot on Type 8305 Ref. Surface with 20 g slotted load Mass (PTB design - average 4 points)
0.5
0
Dev. - %
-0.5
-1
-1.5
-2
-2.5
-3
10
1000
100
Freq - Hz
10000
090143
37
At DPLA a number of measurements were undertaken to question and verify
these new results. One of the first measurements was to compare the results on top
of a 20 gram solid stainless steel mass (as described in the standard [1]), to the
results obtained at the interface between the transducer and a slotted 20 gram
mass. The difference is shown in Fig. 25, and this shows a deviation well beyond
the value stated in the standard.
A number of measurements similar to the results reported in [2] are in the
process of being completed, but these seem to confirm the findings.
Conclusions
The transverse motion of different shakers has been investigated. The results show
that even high quality shakers designed for calibration show very high levels of
transverse motion at higher frequencies, when back-to-back accelerometers with
load-masses are mounted on them. This is detrimental to the laser-interferometric
measurements used to give the best possible primary calibrations.
A solution to this problem has been devised in the form of mechanical filters,
which modify the assembly to remove the high frequency bending modes of the
shaker plus accelerometer structure.
The use of slotted load-masses for calibration of back-to-back accelerometers
has been taken up, and it has been shown that the difference between this and the
often used mirror-masses is much larger than normally expected. This is the
subject of further work and discussion.
References
[1]
ISO 16063–11, “Methods for the Calibration of Vibration and Shock
Transducers – Part 11: Primary Vibration Calibration by Laser
Interferometry”, 1999.
[2]
Täubner, A., Schlaak, H-J., Bruns, T., “Metrological and Theoretical
Investigations of the Influence of Mass Loads on the Transmission
Coefficient of Acceleration Transducers of the Back-to-back Type”,
Physikalisch Technische Bundesanstalt, 38116 Braunschweig, Germany.
38
Previously issued numbers of
Brüel & Kjær Technical Review
Previously issued numbers of
Brüel & Kjær Technical Review
1 – 2009 Use of Volume Velocity Sound Sources in the Measurement of Acoustic
Frequency Response Functions
Turnkey Free-field Reciprocity System for Primary Microphone Calibration
1 – 2008 ISO 16063–11: Primary Vibration Calibration by Laser Interferometry:
Evaluation of Sine Approximation Realised by FFT
Infrasound Calibration of Measurement Microphones
Improved Temperature Specifications for Transducers with Built-in
Electronics
1 – 2007 Measurement of Normal Incidence Transmission Loss and Other Acoustical
Properties of Materials Placed in a Standing Wave Tube
1 – 2006 Dyn-X Technology: 160 dB in One Input Range
Order Tracking in Vibro-acoustic Measurements: A Novel Approach
Eliminating the Tacho Probe
Comparison of Acoustic Holography Methods for Surface Velocity
Determination on a Vibrating Panel
1 – 2005 Acoustical Solutions in the Design of a Measurement Microphone for
Surface Mounting
Combined NAH and Beamforming Using the Same Array
Patch Near-field Acoustical Holography Using a New Statistically Optimal
Method
1 – 2004 Beamforming
1 – 2002 A New Design Principle for Triaxial Piezoelectric Accelerometers
Use of FE Models in the Optimisation of Accelerometer Designs
System for Measurement of Microphone Distortion and Linearity from
Medium to Very High Levels
1 – 2001 The Influence of Environmental Conditions on the Pressure Sensitivity of
Measurement Microphones
Reduction of Heat Conduction Error in Microphone Pressure Reciprocity
Calibration
Frequency Response for Measurement Microphones – a Question of
Confidence
Measurement of Microphone Random-incidence and Pressure-field
Responses and Determination of their Uncertainties
1 – 2000 Non-stationary STSF
1 – 1999 Characteristics of the Vold-Kalman Order Tracking Filter
1 – 1998 Danish Primary Laboratory of Acoustics (DPLA) as Part of the National
Metrology Organisation
Pressure Reciprocity Calibration – Instrumentation, Results and Uncertainty
MP.EXE, a Calculation Program for Pressure Reciprocity Calibration of
Microphones
(Continued from cover page 2)
(Continued on cover page 3)
1 – 1997 A New Design Principle for Triaxial Piezoelectric Accelerometers
A Simple QC Test for Knock Sensors
Torsional Operational Deflection Shapes (TODS) Measurements
2 – 1996 Non-stationary Signal Analysis using Wavelet Transform, Short-time
Fourier Transform and Wigner-Ville Distribution
1 – 1996 Calibration Uncertainties & Distortion of Microphones.
Wide Band Intensity Probe. Accelerometer Mounted Resonance Test
2 – 1995 Order Tracking Analysis
1 – 1995 Use of Spatial Transformation of Sound Fields (STSF) Techniques in the
Automative Industry
2 – 1994 The use of Impulse Response Function for Modal Parameter Estimation
Complex Modulus and Damping Measurements using Resonant and Nonresonant Methods (Damping Part II)
1 – 1994 Digital Filter Techniques vs. FFT Techniques for Damping Measurements
(Damping Part I)
2 – 1990 Optical Filters and their Use with the Type 1302 & Type 1306 Photoacoustic
Gas Monitors
1 – 1990 The Brüel & Kjær Photoacoustic Transducer System and its Physical
Properties
2 – 1989 STSF – Practical Instrumentation and Application
Digital Filter Analysis: Real-time and Non Real-time Performance
1 – 1989 STSF – A Unique Technique for Scan Based Near-Field Acoustic
Holography Without Restrictions on Coherence
2 – 1988 Quantifying Draught Risk
1 – 1988 Using Experimental Modal Analysis to Simulate Structural Dynamic
Modifications
Use of Operational Deflection Shapes for Noise Control of Discrete Tones
Special technical literature
Brüel & Kjær publishes a variety of technical literature which can be obtained from
your local Brüel & Kjær representative.
The following literature is presently available:
•
•
Catalogues (several languages)
Product Data Sheets (English, German, French,)
Furthermore, back copies of the Technical Review can be supplied as listed above.
Older issues may be obtained provided they are still in stock.
TECHNICAL REVIEW
No. 1 – 2010
BV 0062 – 11
ISSN 0007 – 2621
ËBV-0062---'Î
Time Selective Response Method
In situ Measurement of Absorption Coefficient
Transverse Motion in Accelerometer Calibration
HEADQUARTERS: Brüel & Kjær Sound & Vibration Measurement A/S
DK-2850 Nærum Denmark · Telephone: +45 7741 2000 · Fax: +45 4580 1405
www.bksv.com · [email protected]
Local representatives and service organisations worldwide

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project