Beruflich Dokumente
Kultur Dokumente
97-109, 1997
0 1997 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
PII: SOOO3-682X(96)00043-6 0003-682X/97/$17.00+ .OO
ELSEVIER
F. J. Fahy
(Received 7 June 1995; revised version received 17 July 1995; accepted 18 July 1995)
ABSTRACT
1 INTRODUCTION
The standards which form the subject of this paper are designed to be
applied to sources in their normal operating environments, which vary
widely in terms of reverberation and background noise. This has necessitated
the incorporation of safeguards against excessive measurement error under
extremely unfavourable conditions; these safeguards may appear to be over-
conservative to those making determinations under favourable conditions.
3 STATISTICAL CONSIDERATIONS
Given the fact that measurement conditions, and, to some extent, measurement
procedures, can never be exactly reproduced in the process of application of
a standard, it is necessary to distinguish between two forms of statement of
the variability of the results of a number of determinations, namely,
‘repeatability’ and ‘reproducibility’. The following is extracted from the
section entitled ‘Measurement uncertainty’ of the draft scanning standard
ISO-DIS 9614-2. ‘The value of sound power of a noise source determined by
a single application of the procedures of this International Standard is likely
to differ from the true value. The actual difference cannot be evaluated, but
International standards 99
the confidence that the value determined lies within a certain range about the
true value can be stated, on the reasonable assumption that the values
determined by numerous applications of the procedures are normally
distributed about the true value. Where repeated applications are made to
a given source located at a given test site under nominally identical test
conditions, using the same test procedures and instrumentation, the values so
determined constitute the data set which statistically describes the repeat-
ability of the determination. Where the values are determined from tests
made on the given source at different test sites using physically different
instruments, all of which conform with the requirements of this International
Standard, the data set so obtained statistically describes the reproducibility
of the determination. Reproducibility is affected by variations of environmen-
tal conditions at the test sites and of experimental technique. The standard
deviations do not account for variations of sound power output caused
by changes in operating conditions of a source (e.g. rotational speed, line
voltage) or mounting conditions.’
There are test conditions intermediate to those specified above; for
example, different measurement teams, using different sets of equipment,
may make determinations on a given source operating in a given environ-
ment. Wherever any one component of the process of determination varies
from test to test, the appropriate statistical quantity is ‘reproducibility’. The
variance of reproducibility is likely to exceed that of repeatability.
The statements in the standards of the standard deviation of repro-
ducibility also implicitly include a factor which accounts for the fact that a
test environment actually influences the sound power radiated by a given
source, albeit generally rather weakly, and principally at frequencies below
200 Hz. IS0 9614-l explicitly states that ‘The sound power is the in situ
sound power as determined by the procedure of this part of IS0 9614; it is
physically a function of the environment, and may in some cases differ from
the sound power of the same source determined under other conditions.’
The difference between the true power of a source and that estimated by an
application of a standard constitutes an estimation error. The total error is
the sum of bias error and random error. The former originates principally
from the phase mismatch between the transducer channels of the intensity
measuring instrument: the latter is a combination of intensity field spatial
sampling error and spectral estimation error; the former of these is a func-
tion of the spatial irregularity of the distribution of the normal intensity
component over the measurement surface; the latter is determined princi-
pally by the bandwidth-averaging time product (B7’) of the intensity estimate
and the coherence between the two microphone signals. The requirements of
the standards are defined so as to restrict the maximum possible bias error to
half the standard deviation of reproducibility. Since the bias error may be
positive or negative, depending on the particular instrument and frequency
band, and may take any value between zero and the maximum allowed,
depending upon the form of the intensity field and the orientation and
location of the probe, it is not unreasonable to consider the bias error to be a
random error in terms of reproducibility: it is, of course, uncorrelated with
the other random errors, which, themselves are mutually uncorrelated.
5 INSTRUMENT PERFORMANCE
The two-microphone p-p intensity probe is the most widely used form of
transducer system for intensity measurement. The spectral density of the
time-averaged sound intensity of a time-stationary sound field is propor-
tional to the imaginary part of the cross-spectral density of signals from the
two microphones. 3 Within the frequency range of useful operation, this is
linearly proportional to the product of the square roots of the two pressure
spectral densities and the phase of the cross-spectrum. It is clear that an
instrument will only provide an accurate estimate of sound intensity if the
phase mismatch between its transducer channels is much smaller than
the phase difference between the pressures at the microphone positions, the
normalized error due to phase mismatch being equal to the ratio of the two.
The phase mismatch ‘inherent’ to any instrument (combination of probe and
processor) may be determined by placing the two microphone sensing
elements close together in a small, acoustically excited cavity: the true phase
difference is zero, and any indicated intensity is proportional to the phase
mismatch. The presence of back vents for static pressure equalization can
cause measurement errors if sound transmission through the vent passage
is not adequately attenuated. The International Standard for Intensity
Instruments4 requires tests to be made in a standing wave field to demonstrate
the adequacy of this attenuation.
International standards 101
The IS0 Standard series 3740-3748 specifies methods for the determination
of the sound power of sources in carefully controlled environments by means
db
t
6
P'O
Frequency
Fig. 1.
102 F. J. Fahy
In the early days of the use of sound intensity for determining source sound
power, there was much discussion, and even controversy, about the relative
merits of scanned and discrete point techniques. On one hand, the spectral
estimation error of an estimate of sound intensity at a fixed point may
generally be reduced to a negligible value within a reasonable averaging time;
but there is substantial uncertainty about the quality of the result as an
estimate of the spatial average of the normal intensity over the segment of
measurement surface associated with that point. On the other hand, the signals
generated by a scanned probe are non-stationary because the intensity varies
with position; spectral estimation is compromised, but the probe does pass
over, or near, all points on a segment, thereby reducing the spatial sampling
error. Theoretical analysis of the errors associated with each technique has
been attempted,7 but there is now ample empirical evidence that the results
of scanned measurements are at least as reliable as those of discrete point
measurements, and the consensus among users is that the scanning technique
is the more practicable of the two.
As with 9614-1, an initial measurement surface is defined around the
source by the user of the standard, but its minimum average distance from
106 F. J. Fahy
the source may be as small as 200 mm, compared with 500 mm for point
measurements. The reason for the relaxation of this proximity limit is that a
scan produces a degree of compensation of near field spatial alternation of
intensity direction which is not available from a point measurement. The
surface is divided up into segments according to a requirement that each
segment can be comfortably scanned by the operator: this does not apply if
mechanical scanning is employed.
8 AMELIORATIVE ACTIONS
In cases where the field conditions and/or the chosen measurement surface
do not satisfy the criteria specified in the standards, the user is advised to take
certain ameliorative actions. These consist either in moving the measurement
surface (normally toward the source surface), increasing the number of
measurement points in the discrete point method, or installing temporary
screening of extraneous sources.
International standards 107
The technical reasons for specifying the various actions are complex and
remain to be validated by practice. The sheer complexity of the sound fields
which are generated by sources in their operating environments precludes the
specification of completely reliable strategies for improving the quality of a
sound power determination; however, some general observations can be made.
The reason for increasing the number of measurement points in the
discrete point method, if the initial number is considered to be insufficient to
reduce the variance of the estimate of the sound power to within tolerable
limits because the estimated spatial variance of the normal intensity field is
excessive, is based upon a model of independent point samples of the field, as
explained above. This procedure is intuitively rational, but not necessarily
valid. However, critics of this procedure, who claim that it is impracticable,
should ask themselves what statistically valid alternative strategy could be
employed. The method based upon repeat determination with the same
number of points distributed in a different spatial array, as in ANSI S12.12,
is vulnerable to the same statistical criticism, unless it can be shown that the
two sets of samples are statistically independent: if not, good reproducibility
of an erroneous result may be fortuitous. The jury is out on this question;
only time and accumulated experience will resolve this thorny issue.
The proximity of a strongly radiating extraneous source to a source under
test often generates negative (inward-going) partial sound power through a
section of surface lying between the two. This condition increases both bias
and random errors, and is indicated by a large value of F3--F2 or of Fi . The
negative contribution can normally be reduced by moving the measurement
surface towards the source under test. If this is not effective, or practicable,
the interposition of a temporary screen to reduce the influence of the
extraneous source is recommended. This should be placed close to the
offending source, and not to the measurement surface where it may cause
local problems with S,,.
The presence of a strong reverberant sound field component on the
measurement surface increases S,, and also increases random error. If unac-
ceptably large values of F2, F3 or Fpl can be attributed to this cause, the
adverse effects can be reduced by moving the measurement surface towards
the source under test. Note that the reverberant component is not likely to
create significant differences between F3 and F2.
In cases where violation of the dynamic capability criterion is considered
to be due to strong near-field components on the measurement surface,
and where extraneous sources or reverberation allow, movement of the
measurement surface away from the source may be effective.
In cases where the reproducibility of partial power estimates produced
by scanning is unacceptable, despite a lack of evidence of changing source
outputs or environmental conditions, it is possible that very localized regions
108 F. J. Fahy
of strong radiation exist, which are not equally sensed by each of the two
orthogonal scans. In this case it is advised that the measurement surface is
moved away from the source but that the scanning line density is retained.
The rationale is that radiation normally spreads out with increased distance
from a local source.
9 AIRFLOW
The standards contain an annexe which deals with the effects of airflow on
intensity measurement. This presents a difficult problem, especially in cases
where the source under test is an air-moving device, or is cooled by an
airflow. Of course, natural wind also causes a problem. The physical problem
is that turbulence in an airstream can impose non-acoustic pressure
disturbances on microphones, which can be suppressed to only a limited
extent by the use of windscreens. The two-microphone probe suppresses the
effect of small scale turbulence on the pressure cross-spectrum by virtue of
the fact that the hydrodynamic pressures have a limited correlation length.
However, the resulting intermicrophone coherence is still decreased, with the
attendant increase in random error of an intensity estimate, which can be
reduced by increasing averaging time. Larger scale turbulence pressures are
less well suppressed. The major problems tend to occur at frequencies less
than 250 Hz, and the use of large microphone separation distances can help
considerably.
It would be helpful if a simple test for the influence of turbulence on
intensity measurements could be devised. Research on this subject is in
progress,8 but insufficient validation data was available at the time of devel-
opment of the standards. Hence, a blanket restriction on maximum airspeed
was incorporated.
10 CONCLUSION
It is the hope of the author that this paper will assist the current and future
users of the standards to appreciate the problems faced by the members of
WG25, and the reasoning behind the apparently complicated procedures
adopted. International Standards are routinely reviewed every five years: it is
the earnest wish of the national representatives on WG25 that those who
make use of the current standards will disseminate the fruits of their experience
to them personally, and to the global acoustic community, through conference
papers and other technical communications. Only by such efforts will the
standards be improved and refined, to the benefit of all concerned.
International standards 109
REFERENCES