w

FIFTH INTERNATIONAL CONGRESS ON SOUND AND VIBRATION

DECEMBER 15-18, 1997
ADELAIDE, SOUTH AUSTRALIA
Invited Paper

A CRITICAL REVISION OF 1S0 ST~DARD 9614-1:

DETERMINATION OF SOUND POWER LEK??LS OF NOISE SOURCES
USING SOUND INTENSITY

Monica Carfagni andkfarco Pierini

Diptimento di Meccanica e Tecnolog”e Industrial
University di Firenze
Via di Santa A4arta, 3
50139 Firenze, Italy

~ ABSTRACT - Experimental testing was performed to investigate the validity of ISO Standard
9614-1 for determining sound power levels of noise sources using sound intensity at discrete
points. Among the conditions varied during testing were test environment, source position,
extraneous noise, and measurement surface. On the basis of this investigation and extensive
industrial testing a critical revision of the standard has been written with the intent to eliminate
the ambiguity and repetitiveness encountered in the original version.

1 INTRODUCTION
Sound power determinatio~ of paramount importance in qual@ing an acoustic source critical,
is mandated by the EU Machine Directive CEE 392/89. Sound power is the only parameter
that enables comparing several sources and characterizing sources in any environment. In
addhio~ it can be used as an input for acoustic prediction soilware.
To uniform standards for noise measurement and abatement, the EU Standard
Committee selected the 1S0 standards for application in EU directives on the basis of their
completeness and widespread diffhsion. The ISO standards are issued according to the two
main categories of methods for measuring sound power from a source:
methods based on measuring sound pressure (1S0 Standards 3740 to 3747)
methods based on measuring sound intensity (1S0 Standard 9614, Parts 1 and 2).
1S0 Standard 9614 was selected for our investigation for two reasons: 1) because the
methods it describes can be carried out under less restrictive environmental conditions and 2)
because the uncertainty of the results can be evaluated during measurement processing. This
work is divided into two parts. In the iirst, the previsions of 1S0 9614-1 (measuring sound
intensity at discrete points) are analyzed and revisions are proposed. The revisions are
illustrated with references to the original text: Only those parts that underwent major revision
are described; minor modifications have been omitted for brevity. In the second, the reference
source for use in laboratory testing is described. The results of experimental testing conducted
in the laborato~ and in the field on actual machines are presented.
 2 PROPOSED REVISIONS TO 1S0 STANDARD 9614-1

2.1 Introduction (References: Foreword and Paragraphs 0.1,0.2,0.3, 1.1)

ISO Standard 9614 is divided into two parts:
Part 1: Measurement at discrete points
Part 2: Measurement by scanning.
The standard is based on the principle that “the sound power radiated by a source is
equal in value to the integral of the scalar product of the sound intensity vector and the
associated elemental area vector over any surface totally enclosing the source”. Earlier
standards such as 1S0 3740 to 3747 are based on sound pressure measurements. Since,
however, the relationship between sound intensity and pressure at single measurement
positions is dependent on the source and environment characteristics, as well as measurement
positio~ 1S0 3740 to 3747 specified the characteristics that ensure the measured sound power
value would be meaningfi.d.1S0 9614 differs from its predecessor in the following respects:
It can be applied with high levels of extraneous noise.
It can be applied more frequently in situ.
It requires both sound intensity and pressure measurements.
The uncertainty of the sound power determined is classified with test and calculations during
processing of the measurements.
A weighted sound power level is evaluated on the basis of constant percentage bandwidth
- levels.

2.2 Reference Standards (Reference: Paragraph 2)

2.3 Noise Source (References: Paragraphs 0.3,1.1,3.13,4.1,4.2, 5.2.1,7.1, 7.2)

2.4 Test Environment (References: Paragraphs 1.2,3.9,4.2,5.1,5.2.1, 5.2.2,5.3, 5.4)

2.5 Instrumentation (References: Paragraphs 3.10,6.1,6.2,6.2.1, 6.2.2)

Use instruments, including the sound intensity probe meeting the requisites of IEC 1043. Use
Class 1 instrumentation to comply with Grades 1 and 2 accuracy levels. In accordance with
IEC 1043, calibrate the instrumentation in compliance to national standards at least once a year
and record the results in the test report.
Before each group of measurements, calibrate and measure the p-I inde~ taking into
account the ambient temperature and pressure. If no manufacturer’s calibration instructions
have been fbrnished, proceed as follows:
Check the sound pressure level of each of the intensity probe’s microphones using a Class O, 1,
or lL calibrator according to IEC 942.
Set the probe on the measurement surface, with its axis perpendicular to the surface at a
position characterized by an intensity level higher than the surface average intensity. After
measuring the sound intensity, rotate the probe 180 deg, with the axis perpendicular to the
measurement surface and without changing the position of the probe center. Measure the
sound intensity again. The result is satisfactory if the two intensity values are of different sign
and do not differ more than 1.5 dll at the maximum band level measured in octave or one-
third-octave bands.
 2.6 Frequency Analysis (References: Paragraphs 1.1,4.3, 8.1)
Frequency analysis measurements, which can be carried out in octave or one-third-octave, are
used to evaluate the total sound power level and A-weighted total sound power level.
Therefore, A-weighted total sound power level is not measurement directly by means of an A-
weighted filter.
Current limitations of the instrumentation mandate the following fi-equency limits:
octave: 63-4000 Hz
one-third-octave: 50-6300 Hz.
The A-weighted total sound power level is to be considered correct if there are no
significantly high levels at fi-equencies beyond the measuring range. Hence, levels not more
than 6 dB below the A-weighted total sound power level after A-weighting are considered
significant. If the sound power is evaluated in a more limited frequency range, the frequency
range must be indicated in the test report.
Instruments that process signals in constant percentage bandwidths require an
averaging time that respects the relation BT >400, where B is the bandwidth at the lowest
frequency used and T is the averaging time. For instruments which synthesize octave or one-
third-octave bands fi-om narrow-bad-analysis, the equivalent averaging time must be
established according to IEC 1043. Special care must be in the case of cyclic signals.

- 2.7 Measurement Positions (References: Paragraphs 0.3,3.7,3.8, 8.2)

The-measurement positions must be distributed as unitlorrnlyas possible over the measurement
surface. There must be at least one measurement position per square meter and at least 10
positions. If the measurement surface exceeds fifty square meters, the following statements
apply:
,With significant extraneous noise, a minimum of 50 measurement positions are required and
there must beat least one measurement position per two square meter.
With insignificant extraneous noise, no more than 50 positions are required and their
distribution over the measurement surface should be as tiorrn as possible.
Extraneous noise can be considered insignificant it with the source off, the A-
weighted sound pressure levels fall by at least 10 dB at five measurement positions distributed
uniformly over the measurement surface.

2.8 Definitions (Reference: Paragraph 3)

2.9 Range Indicators (Reference: Annex A)

2.10 Testing and Degree of Accuracy (References: Paragraphs 0.3, 1.3, 4.3, 8.2, 8.3.1,
8.3.2,8.4, 10.5, and Annex B)
1S0 Standard 9614-1 sets three classes to evaluate the sound power and the degree of
accuracy:
Grade 1: precision
Grade 2: engineering
Grade 3: survey.
The uncertainty level associated with each grade is related to the random errors in the
measuring procedure and the maximum bias error (limited by the selection of a suitable bias
factor K used in Test 2 for degree of accuracy), but does not account for the instrument
tolerances and the effects of the variations in source installation and operating conditions.
 Thedegree ofaccuracy depends upon thehdof noise generated by the source, the
kind ofextraneous noise, the source absorption and the measuring and sampling procedures.
Thetests required bythe standard are designed inrelation tothese factors.
The sound power levels in several frequency bands can be ignored (and thus the
uncertainty of their determination can be considered irrelevant) provided the following
conditions are met:
If only the A-weighted determination is required, any single A-weighted band level of 10 dB or
more below the highest A-weighted band level shall be neglected. If two or more levels meet
this condition, they can be ignored if the level of the sum of the A-weighted sound powers in
these bands is 10 dB or more below the highest A-weighted band level.
If only the total sound power level is required, all levels 10 dB lower than total can be ignored.

Test 1: F1 <0.6
Test 1 is conducted at a measurement position typical of the measurement surface before and
afler all measurements have been made. The purpose of the test is to establish the temporal
variability of the sound field during measurement. If the test condition is not met in all the
bandwidths under examination (which therefore cannot be ignored according to the above
criteria), the temporal variability of the sound field must be reduced.

Test 2: F2 < Ld
- Test 2 establishes the instrumentation’s dynamic capability to perform specific measurements.
The-test is more restrictive if F’sis used in place of F2. If the test is not successfid for all the
bandwidths under exarninatio~ the procedures indicated in Table B.3 of the original standard
should be followed or else:
If only the A-weighted total sound power level is required, eliminate the bandwidths not
meeting test conditions from the calculation and indicate the effects of uncertainty in the test
report.
If only the total sound power level is required, indicate the uncertainty effects in the test
report.

Test3:Fs-Fz 53dB
Test 3 verifies the absence of highly directional and excessively reflective emissions on the
surrounding environment of the external sources. If the test fails to give satisfactory results for
all the bandwidths under examination, follow the procedures ind~ated in Table- B.3 of the
Oligiti standard.

Test 4: N > CF42

Test 4 determines whether the number of measurement positions and their locations are
suitable for representing the spatial variability of the sound fi~ld over the measurement sutiace.
If the test results are not satisfactory for all bandwidths, either modifi the number a.dor
distribution of the measurement positions or:
Indicate in the test report the estimate of the confidence interval at 95’?40for the octave or one-
third-octave bands that fhiled according to the formula
2F4
lolog(l*@ m

If only the A-weighted total sound power level is required, eliminate the bandwidths that failed
from the calculation and indicate the uncertainty effects in the test report.
2.11 Test Report (Reference: Paragraph 10)

Annex C

Annex D

3 THE REFERENCE SOURCE

A reference source was developed for laboratory testing with emissions of the following
characteristics:
. radially uniform
● stable in time
. possibility of concentrated emissions in selectable directions.
Our first step was to select the geometry of the icoshedral figure (polyhedral with 20
triangular surfaces) as the best compromise between the conflicting needs to approximate the
spherical shape and to apply flat-based loudspeakers. To enhance manageability, we used a
portion of the polyhedral figure composed of 15 tetrahedral components. Each loudspeaker
had its own tetrahedral speaker made of a 14-mm-thick, medium density fiberboard. The shape
of the speakers was especially suitable because the possibility that stationary waves could form
was minimized by the absence of parallel walls.
The source, guided by a random signal, was controlled from a console containing the
microphone switches. The loudspeakers were pretested according to manufacturer’s
instructions. To ensure stability, we developed a device that allowed a control microphone to
be accurately assembled at a given position with respect to the loudspeaker. This is critical
since, in the near field, the sound pressure can vary markedly, even when the position of the
microphone varies only slightly. To minimize the interference resulting from the others
loudspeakers, the stability measurements were made with a single pair of operating difisers,
one facing the other. The measurements of the emissions before and after each test confirmed
that the loudspeakers were capable of providing long-term stable emissions.

4 RESULTS OF EXPERIMENTAL TESTING

All tests were conducted in octave in frequencies ranging from 50-6300 Hz to Grade 2
(engineering) degree of accuracy. To cover the entire frequency range, it was necessary to use
double spacing (or microphone separation) of the sound intensity probe to carry out the
measurements:
low frequencies (50- 1000 Hz): spacing: 50 mm; averaging time: 35 s
high fi-equencies (1250-6300 Hz): spacing: 8.5 mm; averaging time: 10 s.
Testing was conducted in three locations, two indoors and one outdoors. A
parallelepipeds measurement surface was used. Conditions such as environmental
characteristics, source positio~ background noise, and measurement surface were varied.
The standard was revised on the basis of the test results. For brevity, only the main
revisions are illustrated below.

4.1 Test 4- N > CF~

The standard gives the minimum number of measurement positions distributed uniformly on
the measurement surface after Grade 2 testing and determines whether these selections fit the
sound field’s spatial variability on the measurement surface. If this condition is not filfilled, the
standard recommends either mod~g the measurement surface-which necessitates
remeasuring all the positions !-or else uniformly increasing the number of measurement
positions.
We recommend the alternative solution of gradually increasing the number of positions,
starting from the measurement segments where F’Ais highest. The test procedure should then
be repeated until the proscribed conditions are met. This procedure is optimum for minimizing
the number of positions to be added to meet proscribed conditions and is also in agreement
with Test 4 objectives.
An example is given in Figure 1: (a) shows the minimum distribution required by the
standard (16 points) and the result of the application of Test 4. If the standard’s
recommendation is followed, we need 64 points (b), whereas our proposed method requires
only 37 points (c).

[Hz] CFq2 1 CF42 [Hz] CF42

000 50 2 315 10 2000 27

r
-L-
o

0
0

00
0 0
0
0
63
80
100
125
160
200
6
5
3
8
7
21
400
500
630
800
1000
1250
10
34
25
50
98
7 2500
3150
4000
5000
6300
18
18
21
24
3
000
a) N=16 250 15 1600 43

[Hz] CF42 1 CF42 [Hz] CF42

‘b56wEL
10001000010001
50
63
80
100
6
8
6
7
315
400
500
630
7
8
7
12
2000
2500
3150
4000
15
12
17
15

F
125 14 800 32 5000 19
160 5 1000 43 6300 18
0000 200 22 1250 11
0000
0000 N=64 250 21 1600 18

[Hz] CF42 1 CF42 1 CF?

50 6 315 7 2000 15

r
1631814001712500112 I
00
00°0 18016150018131501171
00 00 I1OOI 8 1630114140001151
0 125 12 800 31 5000 19
00 0 00 160 7 1000 34 6300 18
0000 1200131112501111 I I
0000
c) +0000 N=3’7 I 250 I 18 I 16(-)0
---- I 18 I1
.-I 1 I
II

Figure 1 Example of the effectiveness of the alternative point-adding procedure.

4.2 Validity of F4
We investigated the validity of F4—which in actuality is a normalized standard deviation-by
examining the face of the parrdlelepiped surface with the highest value of Fd (nonuniformity of
the sound field). We mapped the face by carrying out 96 acquisitions over the uniformly
distributed positions. The mapping surface, measuring 1.8 x 1.30 m, was divided into 12 rows
and 8 columns. The measurement positions were 15 cm apart. Having acquired the intensities,
we used a Matlab program to analyze the values of 174.We selected an increasing number of
measurements (3x2, 4x3, 6x4, 12x8), each time calculating the value of F4 per one-third-
octave and determining its value in relation to the variations in the number of measurement
positions considered. We thus verified that, except at very low frequencies, F4 can be
considered a representative parameter.

4.3 Constant C
A.tier having calculated the sound power levels in cotigurations with different numbers of
measurement positions, we observed that the values never differed more than 2 dB. This led us
to suspect that the Test 4 constant C was overestimated. The suspicion was confirmed by
comparing an equivalent standard, S31-100 issued by the French AFNO~ which exhibits
lower values (Table. 1).

Table 1. Constant C values in ISO 9614-1 andAFNORS31-100.

. —
Octave bands 1/3 octave bands 1S0 9614-1 s 31-100
1 1 Grade 2 Grade 2
63-125 50-160 11 8
250-500 200-630 19 15
1000-4000 800-5000 29 28
6300 14 10

4.4 Hot Point Procedures

We investigated the additional hot point procedures recommended by ISO Standard 9614-1
with only some of the loudspeakers operating. The following observations are in order:
According to 1S0 9614-1, the measurement positions should be arranged in decreasing order
of magnitude on the basis of the positive partial sound powers passing through each
measurement segment and then the upper subset N. through which more than half of the total
sound power (area S.) should be determined: If N. is less than half the total number of
segments, the procedure can be applied. Conversely, we feel that it would be more correct to
refer to a part of the surface (for example, one face of the parallelepipeds)as recommended in
Paragraph 4.1 rather than to a single segment.
1S0 9614-1 recommends calculating the number of positions to add in area S. (which,,
however, might not be located on the same measurement face). According to our experimental
testing, the number of positions added by this method is tremendously overestimated. As an
example, we started with 16 measurement positions uniformly distributed with a highly
directional source. As this number did not meet Test 4 requirements, we applied the hot point
condition. The number of additional positions recommended by 1S0 9614-1 at the most critical
bandwidth is 220, whereas, if we apply the criterion described in Paragraph 4.1, only 28 are
required. The same result was obtained in testing with difYerentconfigurations.

4.5 Uncertainty in Determining Sound Power Levels

The analysis of the sound power levels obtained in testing with various bandwidths following
completion of the ISO 9614 checks exhibits much less variability than in Table 2 of 1S0
Standard 9614-1 (uncertainty in determining sound power levels). Although firther testing
must be conducted to statistically validate this result, the 1S0 values appear to be considerably
overestimated.

5 CONCLUSIONS
A revision of 1S0 9614-1 was undertaken to eliminate the problems of poor organization,
repetitiveness, and ambiguity that emerged during application of the standard. The proposed
revision is based on practical tests that account for the notable variability of the source
characteristics and test environments in relation to the huge number of existing industrial case
histories. An analogous investigation of Part 2 (1S0 9614-2) is forthcoming.

ACKNOWLEDGMENTS

The. authors are gratefbl to Sergio Pulcinelli of Bri.iel & Kjax and Fabio Miniati for their
precious help in carrying out this project.

REFERENCES
Stanahrd ISO 9414-1 Acoustics - Determination of Sound Power Levels of Noise Sources
Using Sound Intensity” - Part 1: Measurement at Discrete Points, 1993.