Uncertainties associated with the use of a sound level meter

Uncertainties associated with the use of a sound level meter
NPL REPORT DQL-AC 002
Uncertainties associated
with the use of a sound
level meter
Richard Payne
April 2004
NPL Report DQL-AC 002
April 2004
Uncertainties associated with the use of a sound level meter
Richard Payne
ABSTRACT
During the years from 1997 to 2001 the National Physical Laboratory provided a sound level
meter verification service, where measurements of various parameters regarding the
performance of a sound level meter were compared to the requirements of BS 7580. These
measurements were carried out on a wide range of sound level meter types and the results are
suitable for use in an assessment of a practical value of measurement uncertainties associated
with the use of a sound level meter.
A value for measurement uncertainty associated with A-weighted noise emission levels,
based on these verification data, together with contributions based on practical consideration
of the operation and calibration of a sound level meter is proposed. This value is directly
related to the practical use of sound level meters and should be suitable for inclusion in
uncertainty budgets concerned with sound power level determination. In addition, the
following simple guidelines regarding the practical use of a sound level meter are proposed.
To reduce the magnitude of measurement uncertainty resulting from the operation of a sound
level meter, effort should be concentrated on reducing contributions associated with the time
weighting and with indicator range changing. It is recommended that the indicator range used
for measurements is the same as that used for the sound calibrator, and the slow time
weighting is used. To reduce the magnitude of measurement uncertainty resulting from the
use of a sound level meter, effort should be concentrated on reducing the effect of the
observer. It is recommended that the indication on the sound level meter is observed remotely
via an ac output (a common feature on class 1 meters).
The actual estimated value of measurement uncertainty assuming a slow time weighting is
0.39 dB. This value is reduced to 0.33 dB if there is no range change and there is no observer
present. The reported combined total standard uncertainty for measurements carried out
using the “slow” time weighting (with or without an observer and with or without a range
change) is 0.4 dB.
For the majority of ISO sound power standards the frequency range of interest is specified as
covering one-third octave bands centre frequencies from 100 Hz to 10 kHz. In this case the
reported final total combined uncertainty associated with band limited noise emission is the
same as the A-weighted value.
i
NPL Report DQL-AC 002
Ó Crown Copyright 2004
Reproduced by Permission of the Controller of HMSO
ISSN 1744-0599
National Physical Laboratory
Queens Road, Teddington, Middlesex, UK, TW11 0LW
Extracts from this report may be reproduced provided the source is acknowledged and the extract is
not taken out of context.
Approved on behalf of the Managing Director, NPL
by Dr R C Preston, authorised by Director, Quality of Life Division
ii
NPL Report DQL-AC 002
CONTENTS
1.
INTRODUCTION....................................................................................................................... 1
2.
UNCERTAINTY BUDGET FOR NOISE MEASUREMENTS WITH A SOUND LEVEL
METER................................................................................................................................................ 4
3.
2.1
UNCERTAINTY ASSOCIATED WITH A-WEIGHTING MEASUREMENTS................ 4
2.2
UNCERTAINTY ASSOCIATED WITH FREQUENCY BAND MEASUREMENTS ...... 5
VALUES OF INDIVIDUAL UNCERTAINTY CONTRIBUTIONS FOR A-WEIGHTED
NOISE EMISSION LEVELS ............................................................................................................ 6
3.1
VALUES OF INDIVIDUAL UNCERTAINTY ASSOCIATED WITH THE OPERATION
OF A SOUND LEVEL METER....................................................................................................... 6
3.1.1
The contribution associated with the A-weighting network, (δfr )................................ 7
3.1.2
The contribution associated with linearity on the reference range, δlin. ....................... 8
3.1.3
The contribution associated with linearity on other ranges, δdl. ................................... 8
3.1.4
The contribution associated with the detector characteristics, δrms .............................. 9
3.1.5
The contribution associated with fast or slow time weighting function, δtime............... 9
3.1.6
The contribution associated with the adjustment of the sound level meter during
calibration with a sound calibrator, δad ....................................................................................... 10
3.1.7
The contribution associated with the calibration of the sound calibrator, δcal ............ 10
3.1.8
The contribution associated with the resolution of the display, δres ........................... 11
3.2
SUMMARY OF INDIVIDUAL STANDARD UNCERTAINTY CONTRIBUTIONS
ASSOCIATED WITH THE OPERATION OF A SOUND LEVEL METER ............................... 11
3.3
VALUES OF INDIVIDUAL UNCERTAINTY ASSOCIATED WITH THE USE OF A
SOUND LEVEL METER............................................................................................................... 12
3.3.1
The contribution associated with microphone directivity, δdir ................................... 13
3.3.2
The contribution associated with the effect of an observer, δobs ................................ 13
3.3.3
The contribution associated with the application of case corrections, δcase ................ 13
3.3.4
The contribution associated with variation in ambient temperature, δtemp.................. 14
3.3.5
The contribution associated with variation in ambient pressure, δpres. ....................... 14
3.3.6
The contribution associated with the use of a windscreen, δws ................................. 15
3.3.7
The contribution associated with the calibration of the microphone, δmic................ 15
3.4
SUMMARY OF INDIVIDUAL STANDARD UNCERTAINTY CONTRIBUTIONS
ASSOCIATED WITH THE SOUND LEVEL METER IN USE ................................................... 15
iii
NPL Report DQL-AC 002
4.
VALUES OF INDIVIDUAL UNCERTAINTY CONTRIBUTIONS AS A FUNCTION OF
FREQUENCY ................................................................................................................................... 18
4.1
FREQUENCY DEPENDENT UNCERTAINTIES ASSOCIATED WITH THE
OPERATION OF A SOUND LEVEL METER ............................................................................. 18
4.2
FREQUENCY DEPENDENT UNCERTAINTIES ASSOCIATED WITH THE USE OF A
SOUND LEVEL METER............................................................................................................... 19
5.
4.2.1
Frequency dependent uncertainties associated with δdir. ............................................. 19
4.2.2
Frequency dependent uncertainties associated with δcase. ........................................... 19
4.2.3
Frequency dependent uncertainties associated with δtemp. .......................................... 20
4.2.4
Frequency dependent uncertainties associated with δpres. ........................................... 20
4.2.5
Frequency dependent uncertainties associated with δWS. ............................................ 20
4.2.6
Frequency dependent uncertainties associated with δmic. ............................................ 21
TOTAL COMBINED STANDARD UNCERTAINTY FOR USE IN ISO STANDARDS 22
5.1
A-WEIGHTED NOISE EMISSION LEVELS ................................................................... 22
5.2
BAND-LIMITED NOISE EMISSION LEVELS ............................................................... 23
5.2.1
The effect on the final total combined uncertainty of frequency variations in group (a)
24
5.2.2
The effect on the final total combined uncertainty of frequency variations in group (b)
24
6.
CONCLUSIONS AND RECOMMENDATIONS.................................................................. 26
7.
ACKNOWLEDGEMENTS ..................................................................................................... 27
8.
REFFERENCES ....................................................................................................................... 28
iv
NPL Report DQL-AC 002
1.
INTRODUCTION
When reporting the result of a measurement of a physical quantity, some quantitative
indication of the quality of the result must be given. Without such an indication, measurement
results cannot be compared either amongst themselves or with values given in a specification
or standard. The generally accepted method is to evaluate and express its uncertainty.
A primary need for values of uncertainty is to demonstrate compliance with noise emission
level declaration as required for numerous regulatory purposes, for example the European
Directive 2000/14/EC(1) “on the approximation of the laws of the Member States relating to the
noise emission in the environment by equipment for use outdoors”. This Directive requires that
machinery is labelled with a guaranteed A-weighted sound power level. Here, a guaranteed
sound power level means a single-number noise emission value including the uncertainties
due to production variation and measurement procedures.
Information on measurement reproducibility given in current International Standards relating
to the determination of sound power level(2 to 8) can be helpful towards the derivation of
measurement uncertainties, but it is incomplete. In particular, it does not give an analysis of
the various components of measurement uncertainty and their magnitudes. The accepted
format for expression of uncertainties generally associated with methods of measurement is
that given in the Guide to the Expression of Uncertainties in Measurement(9). This format
incorporates an uncertainty budget, in which all the various sources of uncertainty are
identified and quantified, from which the combined total uncertainty can be obtained. The
data necessary to enable such a format to be adopted in the case of the International Standards
relating to sound power are not currently available. However, an indication, taken from
ISO 3745:2003(6), is given below of the sources of uncertainty that are thought to be
associated with the methods and equipment described in the standards. The general approach
to calculation of uncertainties appropriate to International Standards, conforming to the
Guide, is illustrated below for information.
The general expression for the calculation of the sound power level, LW, is given by the
following equation:
where,
æ S
LW = L p + 10 lgçç
è S0
ö
÷÷ + d slm + d rep + d boun + d mic + d met + d angle + d imp
ø
Lp
is the surface sound pressure level,
S
S0
is the area of the measurement surface, in square metres,
= 1 m2,
is an input quantity to allow for any error in the measuring instrumentation,
d slm
d rep
d boun
d mic
d met
dB
is an input quantity to allow for any error in the operating conditions of the noise source
under test,
is an input quantity to allow for any error in the influence of the test room boundaries,
is an input quantity to allow for any error in the finite number of microphone positions,
is an input quantity to allow for any error in the meteorological conditions,
1
(1)
NPL Report DQL-AC 002
d angle is an input quantity to allow for any difference of angle between the direction in which the
d imp
sound is emitted by the source and the normal to the measurement surface,
is an input quantity to allow for any error in the impedance of the surroundings into which
the source is emitting sound energy.
The input quantities included in equation 1 to allow for errors are those thought to be
applicable in the current state of knowledge, but further research could reveal that there are
others.
A probability distribution (normal, rectangular or triangular) is associated with each of the
input quantities to allow for errors. Its expectation (mean value) is the best estimate for the
value of the input quantity and its standard deviation is a measure of the dispersion of values,
termed uncertainty. It is presumed (in Reference 6) that the mean values of all of the input
quantities for errors given in equation 1 are equal to zero. This presumption may not be valid
for some input quantities and so an allowance has to be made to account for a potential bias in
results. It is a requirement of Reference 9, that the errors considered in Equation 1 are relative
to values that have been taken to represent “true” (or correct) values. Thus, when considering
the input quantities listed above it may be necessary to take account not only of variations
about mean values (quantified as standard deviations) but also the possibility that mean
values may be different from the adopted “true” values. This is discussed later (see subsection 3.1.1). However, in any particular determination of a sound power level or sound
energy level of a noise source under test, the uncertainties do not vanish and they contribute
to the combined uncertainty associated with values of the sound power level or sound energy
level.
The contributions to the combined total standard uncertainty associated with the value of the
surface sound pressure level depend on each of the input quantities to allow for errors, their
respective probability distributions and sensitivity coefficients, ci. The sensitivity coefficients
are a measure of how the values of the surface sound pressure level are affected by changes in
the values of the respective input quantities. In the model used in equation 1, all sensitivity
coefficients have the value 1. The contributions of the respective input quantities to allow for
errors to the overall uncertainty are then given by the products of the standard uncertainties, ui
and a factor corresponding to their probability distribution.
The combined total standard uncertainty of the determination of the sound power level,
u (LW ) , from the n individual contributions is given by the following equation:
u (LW ) =
n
å (c u )
i
i
2
(2)
i =1
where ui is the ith uncertainty contribution.
The Guide to the Expression of Uncertainties in Measurement requires an expanded
uncertainty, U, to be specified, such that the interval [LW-U, LW+U] covers e.g. 95% of the
values of LW that might reasonably be attributed to LW. To that purpose, a coverage factor, k,
is used, such that U=k.u. The coverage factor depends on the probability distribution
associated with the measurand.
2
NPL Report DQL-AC 002
The standard uncertainties from the various contributions for most noise sources remain to be
established by research. There have been proposals for some contributions, in particular a
value for δslm(10). However, this value for δslm is based on the tolerances given in
IEC 60651:2000(11), which, in the absence of any actual measured data relating to the
performance of a sound level meter, is a reasonable first step in proposing a value for δslm.
However, such a value will provide an upper limit for δslm but will not be strictly applicable to
the use of a sound level meter in practice.
During the years from 1997 to 2001 the National Physical Laboratory provided a sound level
meter verification service, where measurements of various parameters regarding the
performance of a sound level meter were compared to the requirements of BS 7580(12). These
measurements were carried out on a wide range of sound level meter types and the results are
suitable for use in an assessment of a practical value of δslm.
In this paper, a value for δslm, based on these data together with contribution based on
practical considerations of the operation and calibration of a sound level meter is proposed.
Such a value will be directly related to the practical use of sound level meters and should be
suitable for inclusion in Equation 1.
3
NPL Report DQL-AC 002
2.
UNCERTAINTY BUDGET FOR NOISE MEASUREMENTS WITH
A SOUND LEVEL METER
There are many factors to be considered when assessing the measurement uncertainty
associated with a sound level meter. In this paper it is proposed that the A-weighted sound
pressure level, LA, obtained by using a sound level meter, is given by the following equation:
LA = LAmes + δfr + δlin + δdl + δrms + δtime + δad + δcal + δres + δdir + δobs + δcase + δtemp + δpres + δws + δmic
(3)
where:
LA
is the unknown A-weighted sound pressure level
LAmes
is the measured A-weighted sound pressure level
δfr
is the correction associated with the frequency-weighting network
δlin
is the correction associated with the linearity on the sound level meter reference range
δdl
is the correction associated with the linearity on other range settings of sound level meter
δrms
is the correction associated with the detector characteristics
δtime
is the correction associated with the time weighting function (fast or slow)
δad
is the correction associated with the adjustment of the sound level meter during calibration
with a sound calibrator
δcal
is the correction associated with the calibration of the sound calibrator
δres
is the correction associated with the resolution of the display
δdir
is the correction associated with the directivity of the microphone
δobs
is the correction associated with the influence of the observer
δcase
is the correction associated with the sound level meter case
δtemp
is the correction associated with variations in ambient temperature
δpres
is the correction associated with variations in ambient pressure
δws
is the correction associated with a windshield
δmic
is the correction associated with the calibration of the microphone
These corrections will be assessed in two groups:
(a) those associated with operation of the sound level meter;
δad , δcal , δfr , δlin , δdl , δrms , δres , δtime ,
(b) those associated with the sound level meter in use;
δdir , δobs , δcase , δtemp , δpres , δws , δmic.
2.1
UNCERTAINTY ASSOCIATED WITH A-WEIGHTING MEASUREMENTS
A primary need for values of uncertainty is to demonstrate compliance with noise emission
level declaration as required for numerous regulatory purposes, for example the European
4
NPL Report DQL-AC 002
Directive 2000/14/EC “on the approximation of the laws of the Member States relating to the
noise emission in the environment by equipment for use outdoors”. This Directive requires that
machinery is labelled with a guaranteed A-weighted sound power level. Here, a guaranteed
sound power level means a single-number noise emission value including the uncertainties
due to production variation and measurement procedures. So, the main body of this paper is,
therefore, primarily concerned with uncertainties associated with A-weighted noise emission
values.
Values for each of these (A-weighted) contributions are assessed in sub-sections 3.1 and 3.3
for group (a) and group (b) respectively.
2.2
UNCERTAINTY
MEASUREMENTS
ASSOCIATED
WITH
FREQUENCY
BAND
Although numerous regulatory bodies require A-weighted noise emission values, the guide to
the expression of uncertainties in measurement requires that any measured value should be
accompanied by an associated measurement uncertainty. Several of the Standards in the
ISO 3740 series include specifications for the determination of one-third-octave-band noise
emission levels and so there is a need to assess measurement uncertainties associated with
band limited values.
In group (a), the only uncertainty contribution for which there is information on any potential
frequency dependence is δfr, the correction associated with the frequency-weighting network.
In this case it will be the correction associated with a linear weighting. Values of uncertainties
associated with a linear weighting are considered in sub-section 4.1.
In group (b), information regarding any potential frequency dependence is available for all
contributions except δobs, the correction associated with the influence of an observer. Values
for uncertainties associated with these six contributions are considered in sub-section 4.2.
5
NPL Report DQL-AC 002
3.
VALUES OF INDIVIDUAL UNCERTAINTY CONTRIBUTIONS
FOR A-WEIGHTED NOISE EMISSION LEVELS
As indicated in Section 2 the uncertainty contributions may be considered in two groups, one
group, (a) associated with the operation of the sound level meter and the other, (b) associated
with its use. In this Section both groups are considered, group (a) in sub-section 3.1 with a
summary in sub-section 3.2 and group (b) in sub-section 3.3 with a summary in subsection 3.4.
LAmes is the measured A-weighted sound pressure level. No uncertainty contribution
associated with this value is provided here, as it will be accounted for separately as the
standard deviation of repeated determinations for a source under test.
3.1
VALUES OF INDIVIDUAL UNCERTAINTY ASSOCIATED WITH THE
OPERATION OF A SOUND LEVEL METER
In this group a number of contributions have been assessed by using data obtained from
measurements carried out as part of a sound level meter verification process on a range of
instruments. These verification measurements were carried out according to the tests
specified in BS 7580 : Part 1 : 1997. This Standard assesses a sound level meter with the
requirements and tolerances for a Type 1 sound level meter given in BS EN 60651:1994(13) .
The values of standard uncertainty contributions obtained in this way will represent practical
values inherent with commonly used sound level meters and not related to the tolerances that
are specified in the Standard. The five specific contributions that are obtained in this way are,
δfr , δlin , δdl , δrms , and δtime .
Data obtained over the period 1997 to 2001 from measurements during verification tests were
examined and information relating to frequency weighting, reference range linearity, other
ranges linearity, time weighting and detector characteristics were extracted. In all, data from
22 different sound level meter types involving nine different Manufacturers were used. The
Manufacturer and type numbers are listed below in alphabetical order of Manufacturer:
§
§
§
§
§
§
§
§
§
§
§
01dB
01dB
01dB
B&K
B&K
B&K
B&K
B&K
B&K
B&K
CEL
§
§
§
§
§
§
§
§
§
§
§
Symphony;
Concerto;
SIP95;
2231B;
2260;
2239A;
2238;
2260B;
2236A;
2230;
593;
CEL
CEL
Cirrus
Larson-Davis
Norsonic
Norsonic
Quest
Rion
Rion
Rion
Svantek
393;
480;
CR831A;
820;
116;
SA110;
1900E;
NL-14;
NL-15;
NL-31;
912A.
The data for each of the five uncertainty contributions is discussed in sub-sections 3.1.1 to
3.1.5.
6
NPL Report DQL-AC 002
The other three contributions δad , δcal and δres are assessed from practical consideration of the
operation and calibration of a sound level meter. The data for each of these three uncertainty
contributions is discussed in sub-sections 3.1.6 to 3.1.8.
3.1.1
The contribution associated with the A-weighting network, δfr .
BS 7580: Part 1: 1997 requires that the A-weighting network is tested over the range 31.5 Hz
to 12.5 kHz using a continuous sinusoidal signal at preferred octave-band centre frequencies.
The signal level at 1 kHz was set as an indication of the reference sound pressure level. The
response of the sound level meter was compared to that of tables IV and V of
BS EN 60651:1994. Values of the A-weighted frequency response for each sound level
meter are shown in Table A.1.1 in Annex 1 together with the ideal values.
It is acknowledged that the effect on A-weighted sound pressure levels of variations from the
ideal A-weighted function will be dependent on the frequency distribution of the noise
source. For the purposes of this examination, an A-weighted level has been calculated for
each of the 22 sound level meters assuming five noise spectra, namely; flat, rising at 3 dB per
octave falling at 3 dB per octave, rising at 6 dB per octave and falling at 6 dB per octave. In
order to assess values of standard uncertainty the resultant means and standard deviations, σm
of the data from the 22 sound level meters has been calculated for each of the five spectra.
These data are shown in Table 1 together with maximum and minimum values. It can be seen
that the mean values are not quite zero. This indicates that there may be some bias in the Aweighting assessment. In order to account for this in the standard uncertainty a value of
standard deviation, σzero has been calculated by forcing a zero mean. These data are also
shown in Table 1.
Table 1
Mean and standard deviations for A-weighted levels
Spectra
Mean (dB)
Maximum
Minimum
σm (dB)
σzero (dB)
Flat
- 3 dB
+ 3 dB
- 6 dB
+ 6 dB
-0.04
-0.06
-0.03
-0.06
-0.03
0.10
0.01
0.22
0.04
0.33
-0.13
-0.11
-0.20
-0.23
-0.24
0.05
0.03
0.09
0.06
0.12
0.07
0.07
0.09
0.11
0.12
It can be seen from Table 1 that values of σzero are slightly larger than those of σm and that
there is a variation with spectra with values of σzero ranging from 0.07 dB to 0.12 dB. For the
purposes of assessing contributions to the uncertainty budget for a sound level meter the
value of standard deviation of σzero for a flat spectrum of 0.07 dB will be assumed.
It should be noted that the measurement uncertainties associated with the data obtained from
the verification tests are of the same order as those values listed in Table 1. So, the values in
Table 1 cannot be directly attributed to variations between sound level meters with any
reasonable degree of statistical confidence. The values listed in Table 1 may, however, be
assumed to be representative of a worst case.
7
NPL Report DQL-AC 002
3.1.2 The contribution associated with linearity on the reference range, δlin.
BS 7580:Part 1:1997 requires that the linearity of the sound level meter is tested relative to
the reference sound pressure level as indicated on the reference range using a continuous
sinusoidal signal of frequency 4 kHz. Measurements were made at the level intervals
required by the standard and results were compared to the requirements of Table XII of
BS EN 60651:1994. For the purposes of assessing a value of standard uncertainty for
inclusion in a sound level meter uncertainty budget a flat spectrum is assumed and differences
between measured values and those corresponding to a perfectly linear device recorded.
Values of these differences for each sound level meter are shown in Table A.1.2 in Annex 1.
The same analysis process as described in sub-section 3.1.1 was performed and the resultant
values of mean, maximum, minimum, σm and σzero are shown in Table 2.
Table 2
Mean and standard deviations for reference range linearity.
Mean (dB)
maximum
Minimum
σm (dB)
σzero (dB)
-0.01
0.29
-0.24
0.10
0.10
It can be seen from Table 2 that the mean value is close to zero and so values of σm and σzero
are, when rounded to the second place of decimals, the same. So, for the purposes of
assessing contributions to the uncertainty budget for a sound level meter the value of standard
deviation of 0.10 dB will be assumed.
It should be noted that the measurement uncertainties associated with the data obtained during
the verification tests are estimated to be of the same order as the values listed in Table 2. So,
the values in Table 2 cannot be directly attributed to variations between sound level meters
with any reasonable degree of statistical confidence. However, the listed values of δlin may be
assumed to be representative of a worst case.
3.1.3
The contribution associated with linearity on other ranges, δdl.
BS 7580 : Part 1 : 1997 requires that measurements are made for an indication of the
reference sound pressure level on all other ranges which include it. For the purposes of
assessing a value of standard uncertainty for inclusion in a sound level meter uncertainty
budget a flat spectrum is assumed and differences between measured values and those
corresponding to a perfectly linear device recorded. Values of these differences for each
sound level meter are shown in Table A.1.2 in Annex 1. Some of the sound level meters
tested had a very large dynamic range and so there was only a need for a single range. For
these instruments a difference between measured values and those corresponding to a device
that is perfectly linear on range switching is assumed to be zero. The same analysis process as
described in sub-section 3.1.1 was performed and the resultant values of mean, maximum,
minimum, σm and σzero are shown in Table 3.
8
NPL Report DQL-AC 002
Table 3
Mean and standard deviations for other ranges linearity.
Mean (dB)
Maximum
Minimum
σm (dB)
σzero (dB)
-0.02
0.41
-0.66
0.17
0.17
It can be seen from Table 3 that the mean value is close to zero and so values of σm and σzero
are, when rounded to the second place of decimals, the same. So, for the purposes of
assessing contributions to the uncertainty budget for a sound level meter the value of standard
deviation of 0.17 dB will be assumed.
3.1.4
The contribution associated with the detector characteristics, δrms
BS 7580-Part 1:1997 requires that the rms accuracy shall be tested on the reference range for
a crest factor of 3 by comparing the sound level meter indication for a sequence of tone bursts
with that for a continuous sinusoidal signal. The frequency of the continuous signal is 2 kHz
at an amplitude that produces an indication 2 dB below the upper limit of the sound level
meter primary indicator range. Results of this test are compared to table VII of
BS EN 60651:1994. Differences between measured data and those corresponding to a perfect
rms detector were recorded. Values of these differences for each sound level meter are shown
in Table A.1.2 in Annex 1. The resulting 22 values were analysed as described in subsection 3.1.1. The resultant values of mean, maximum, minimum, σm and σzero are shown in
Table 4.
Table 4
Mean and standard deviations for rms accuracy.
Mean (dB)
Maximum
Minimum
σm (dB)
σzero (dB)
0.00
0.3
-0.3
0.19
0.19
It can be seen from Table 4 that the mean value is zero and so values of σm and σzero are,
therefore, the same. So, for the purposes of assessing contributions to the uncertainty budget
for a sound level meter the value of standard deviation of 0.19 dB will be assumed.
3.1.5
The contribution associated with fast or slow time weighting function, δtime.
BS 7580-Part 1:1997 requires that the time averaging be tested on the reference range by
comparing the indication for a continuous sinusoidal signal with that for a sequence of tone
bursts having the same equivalent continuous level. A continuous signal at a frequency of
4 kHz and an amplitude 30 dB below the upper limit of the linearity range is applied to the
sound level meter. A sequence of tone bursts as specified by BS 7580 : Part 1 : 1997 are
applied (see also Table III of BS EN 60804(14)) and the difference between the indication on
the sound level meter with that for the continuous signal recorded. Values of these differences
for each sound level meter are shown in Table A.1.2 in Annex 1. The resulting 22 values have
been analysed as described in sub-section 3.1.1. The resultant values of mean, maximum,
minimum, σm and σzero are shown in Table 5 for both fast and slow time weighting.
9
NPL Report DQL-AC 002
Table 5
Mean and standard deviations for time weighting.
Weighting
Mean (dB)
Maximum
Minimum
σm (dB)
σzero (dB)
Fast
Slow
-0.15
-0.06
0.3
0.5
-0.7
-0.7
0.22
0.19
0.27
0.19
It can be seen from Table 3 that for fast time weighting the mean is –0.15 dB and so the value
of σzero of 0.27 dB is greater than the value of σm of 0.22 dB. For the slow time weighting, the
mean value is close enough to zero for the values of σm and σzero, when rounded to the second
place of decimals, to be the same at 0.19 dB. It is not clear from an inspection of the ISO
standards relating to the determination of sound power, which time weighting should be used
so, for the purposes of assessing contributions to the uncertainty budget for a sound level
meter, both values of standard deviation, σzero will be used in calculating two values of the
final combined total standard uncertainty one for “fast” and one for “slow” time weighting.
3.1.6
The contribution associated with the adjustment of the sound level meter during
calibration with a sound calibrator, δad
Before a noise emission measurement is carried out the sensitivity of the sound level meter
must be checked using a calibrated sound calibrator or pistonphone and if necessary adjusted
according to the sound level meter manufacturer’s instructions and the data in the calibration
certificate for the sound calibrator. The potential error here is dependent on the resolution of
the sound level meter display. All the sound level meters that have been submitted to NPL for
verification according to BS EN 7580 have had a display resolution of 0.1 dB. The
contribution to an overall uncertainty budget will, therefore, be half of this resolution, which
is ± 0.05 dB.
3.1.7
The contribution associated with the calibration of the sound calibrator, δcal
It is a requirement of the ISO standards relating to the determination of sound power that the
sound calibrator shall be calibrated “periodically in a manner that is traceable to appropriate
standards”. This requirement includes the provision of an uncertainty in the calibration
supplied by the authorised laboratory. This is dependent on the quality control system of the
Laboratory or Test House carrying out the calibration. All the calibrators associated with
sound level meters submitted for verification that were used in this study were calibrated at
the National Physical Laboratory that provides calibrations with a measurement uncertainty of
± 0.05 dB. This reported uncertainty is based on a standard uncertainty multiplied by a
coverage factor, k = 2. For the purposes of assessing a value of δcal suitable for inclusion in an
overall uncertainty budget, it is assumed that the uncertainty supplied by the NPL is typical of
that supplied by other calibration laboratories and so an uncertainty of ± 0.05 dB will be
assumed.
10
NPL Report DQL-AC 002
3.1.8
The contribution associated with the resolution of the display, δres
The potential error here is dependent on the resolution of the sound level meter display. The
contribution to an overall uncertainty budget will, therefore, be the same as that assumed for
δad (see sub-section 3.1.6 above) which is ± 0.05 dB.
3.2
SUMMARY
OF
INDIVIDUAL
STANDARD
UNCERTAINTY
CONTRIBUTIONS ASSOCIATED WITH THE OPERATION OF A SOUND
LEVEL METER
All values of standard uncertainty associated with the operation of a sound level meter as
described in sub-sections 3.1.1 to 3.1.8 are summarised below together with an indication of
their associated statistical distribution. These values are then listed in Table 6 together with
their uncertainty contribution, u to the final combined total standard uncertainty, u(LW) to be
associated with the operation of a sound level meter. The value of u(LW) is given by:
u (LW ) =
8
åu
2
i
(4)
i =1
where ui is the ith uncertainty contribution.
Values of the eight contributions are listed below using the error notation from Section 2.
This is dependent on the uncertainty associated with the application of the Aδfr
weighting function and the standard uncertainty is ± 0.07 dB (see sub-section 3.1.1) with a
normal distribution.
This is dependent on the linearity of the sound level meter when used on the reference
δlin
range and the standard uncertainty is ± 0.10 dB (see sub-section 3.1.2) with a normal
distribution.
This is dependent on the linearity of the sound level meter when used on ranges other
δdl
than the reference range and the standard uncertainty is ± 0.17 dB (see sub-section 3.1.3)
with a normal distribution.
This is an assessment of the ability of the sound level meter to provide a true rms
δrms
indication and the standard uncertainty is ± 0.19 dB (see sub-section 3.1.4) with a normal
distribution.
δtime This is dependent on the ability of the sound level meter to provide time averaged data
as required by the time weighting functions described as “fast” or “slow” in
IEC 61672 part1:2002(15) and the standard uncertainty is ± 0.27 dB and ± 0.22dB (see subsection 3.1.5) for fast and slow respectively with normal distributions.
This is dependent on the resolution of the sound level meter display and the standard
δad
uncertainty is ± 0.05 dB (see sub-section 3.1.6) with a rectangular distribution.
11
NPL Report DQL-AC 002
δcal
This is dependent on the uncertainty associated with the calibration of the sound
calibrator, and the uncertainty is ± 0.05 dB, assuming a coverage factor k = 2 (see subsection 3.1.7) with a normal distribution.
As for δad above, this is dependent on the resolution of the sound level meter display
δres
and the standard uncertainty is ± 0.05 dB (see sub-section 3.1.8) with a rectangular
distribution.
Table 6
Uncertainty associated with operation of the sound level meter
Quantity
standard
uncertainty
± dB
probability
distribution
distribution
divisor
uncertainty
contribution ± dB
δad
δcal
δfr
δlin
δdl
δrms
δres
δtime (fast)
δtime (slow)
0.05
0.05
0.07
0.10
0.17
0.19
0.05
0.27
0.22
Rectangular
Normal
Normal
Normal
Normal
Normal
Rectangular
Normal
Normal
√3
2
1
1
1
1
√3
1
1
0.03
0.03
0.07
0.10
0.17
0.19
0.03
0.27
0.22
TOTAL(fast)
0.394
TOTAL(slow)
0.362
It can be seen from Table 6 that the value of the final combined total standard uncertainty for
“fast” time weighting is 0.032 dB higher than for “slow” time weighting. The value to be
used is discussed in Section 4. It is clear that in order to reduce the magnitude of the final
combined total standard uncertainty effort should be concentrated on reducing δtime , δrms and
δdl. It is possible that more recent sound level meter design may have reduced these factors
and certainly those sound level meters with a single large dynamic range will of course reduce
δdl to zero. This will reduce the combined total standard uncertainties to 0.355 dB and
0.320 dB for “fast” and “slow” time weighting respectively. It should be noted that δdl would
also be zero if the range that is used for the sound pressure level measurement were the same
as that used when applying the sound calibrator.
3.3
VALUES OF INDIVIDUAL UNCERTAINTY ASSOCIATED WITH THE USE
OF A SOUND LEVEL METER
In this group, uncertainty contributions are assessed from practical consideration of the sound
level meter in use and from manufacturers data.
12
NPL Report DQL-AC 002
3.3.1 The contribution associated with microphone directivity, δdir .
For the purposes of estimating a value of combined total standard uncertainty for measuring
instrumentation, a value for the uncertainty associated with errors resulting from sound
incident at angles other than that for which the instrumentation has been calibrated should be
included as part of δangle as described in Section 1. However, if this is not the case a value has
been estimated for a hemispherical enveloping surface as described in ISO 3744:1994. In this
simple estimation, it is assumed that the hemisphere radius is twice the largest source
dimension and that the source is radiating equally from all positions on its surface. The angles
of incidence on the microphone diaphragm resulting from this equal radiation has been
calculated as ranging from zero degrees (normal incidence) to approximately 20° and the
corresponding changes in microphone sensitivity at one-third-octave frequency intervals
estimated from manufacturers data. A flat frequency spectrum has been assumed and an Aweighted sound pressure level calculated with these sensitivity changes applied and compared
to the normal incidence case. A change in A-weighted level of up to 0.14 dB was observed.
3.3.2
The contribution associated with the effect of an observer, δobs .
This is very dependent on the frequency distribution of the source and the size and location of
the observer. Ideally measurements will be made utilising remote data logging so the
influence of an observer may be neglected. However, measurements carried out for this
investigation using a reference sound source in a hemi-anechoic room, have indicated a value
of uncertainty of up to ± 0.2 dB.
3.3.3 The contribution associated with the application of case corrections, δcase
IEC 61672:2002(15) requires that corrections be applied to account for the effect of the
presence of the sound level meter in the sound field. Whilst it is a requirement of IEC 61672
that these corrections are supplied, manufacturers do not provide any indication of the
magnitude of associated uncertainties. For the purposes of providing a contribution to the
combined total standard uncertainty, it is proposed that a value of 10% of the correction be
assumed. (It is suggested that a value of uncertainty larger than about 10% would probably
indicate poor statistical confidence in the correction). Values of corrections for the sound
level meters presented for verification purposes (see sub-section 3.1) have been applied to a
flat frequency spectrum and the difference in A-weighted sound pressure level due to the
presence of the sound level meter in the sound field has been calculated as ranging from
0.04 dB to 0.11 dB. For the purposes of assessing a value of δcase the worst-case figure of
0.11 dB has been assumed and so the value for inclusion in Equation 2 is therefore taken as
0.011 dB.
13
NPL Report DQL-AC 002
3.3.4
The contribution associated with variation in ambient temperature, δtemp.
It is assumed here that variations in sound level meter output with changes in temperature
result from changes in the sensitivity of the microphone. Microphones are calibrated at a
temperature of 23 °C and their use at other temperatures will result in small errors because
microphone sensitivity is a function of ambient temperature(16). The magnitude of the error is
dependent in the difference in temperature relative to 23 °C and on the rate of change of
microphone sensitivity, which is in turn dependent on microphone type. There is a range of
microphones used with sound level meters, but in this examination only those used with the
22 sound level meters discussed in sub-section 3.1 are considered. The rates of change of
sensitivity with temperature for the various microphones used have been obtained(16, 17, 18)
and range from 0.002 dB per K to 0.015 dB per K. The ISO series of standards relating to the
determination of sound power level provide an expression to obtain a correction factor to
account for changes in the sound power output of a noise source resulting from changes in
temperature and atmospheric pressure. In Annex E of a draft revision of ISO 3744:2003(19) for
instance, this expression is stated to be valid for a temperature range of from 15 °C to 30 °C.
So, for the purposes of analysis this temperature range will be assumed. The range relative to
the calibration temperature of 23 °C is from – 8 °C to + 7 °C. In this report, errors due to
temperature changes are assessed assuming a range of ± 8 °C in conjunction with the rates of
change above. This provides a potential error of up to between 0.016 dB and 0.12 dB
depending on microphone type. For the purposes of supplying data for a sound level meter
uncertainty budget the worst case of 0.12 dB is adopted.
3.3.5
The contribution associated with variation in ambient pressure, δpres.
As discussed in sub-section 3.3.4 above, it is assumed that variations in sound level meter
output with changes in atmospheric pressure result from changes in the sensitivity of the
microphone. Microphone calibrations are corrected to a standard atmospheric pressure of
101.3 kPa and their use at pressures removed from this will result in small errors because
microphone sensitivity is a function of ambient atmospheric pressure(16). The magnitude of
the error is dependent in the difference in pressure relative to 101.3 kPa and on the rate of
change of microphone sensitivity, which is in turn dependent on microphone type. There is a
range of microphones used with sound level meters, but in this paper only those used with the
22 sound level meters discussed in sub-section 3.1 are considered. The rates of change of
sensitivity with pressure for the various microphones used have been obtained (16, 17, 18) and
range from 0.0015 dB per kPa to 0.019 dB per kPa. Unlike the case of temperature, the ISO
standards do not provide an indication of a temperature range within which measurements
should be made. However, the ISO series of standards relating to the determination of sound
power level provide an expression to obtain a correction factor to account for changes in the
sound power output of a noise source resulting from changes in temperature and atmospheric
pressure. In order to provide an engineering grade accuracy without using this expression to
correct for changes in pressure it is necessary to limit ambient atmospheric pressure to the
range 97.5 kPa to 102.5 kPa. So, for the purposes of analysis this pressure range will be
assumed. The range relative to the calibration pressure of 101.3 kPa is from – 3.8 kPa
to + 1.2 kPa. In this report, errors due to pressure changes are assessed assuming a worst-case
range of ± 3.8 kPa in conjunction with the rates of change above. This provides a potential
error of up to between 0.0057 dB and 0.0722 dB depending on microphone type. For the
14
NPL Report DQL-AC 002
purposes of supplying data for a sound level meter uncertainty budget the worst case of
0.0722 dB is adopted.
3.3.6 The contribution associated with the use of a windscreen, δws
If a windscreen is used corrections must be applied to account for the insertion loss due to the
windscreen. Whilst the manufacturer supplies these corrections, they do not provide any
indication of the magnitude of associated uncertainties. For the purposes of providing a
contribution to the combined total standard uncertainty, it is proposed that a value of 10% of
the correction be assumed. (As for δcase above, it is suggested that a value of uncertainty larger
than about 10% would probably indicate poor statistical confidence in the correction). Values
of corrections for the sound level meters presented for verification purposes (see sub-section
3.1) have been applied to a flat frequency spectrum and the difference in A-weighted sound
pressure level due to the presence of a wind screen has been calculated as ranging from
0.07 dB to 0.19 dB. For the purposes of assessing a value of δcase the worst-case figure of
0.19 dB has been assumed and so the value for inclusion in Equation 2 is therefore taken as
0.019 dB.
3.3.7
The contribution associated with the calibration of the microphone, δmic.
The microphone associated with the sound level meter will have been calibrated and its
sensitivity as a function of frequency will be available. It is assumed here that corrections to
account for any variation of sensitivity with frequency will be applied to measured data. In
this case, only an allowance for the uncertainty involved in the calibration of the microphone
need be considered. The microphone calibration service at the National Physical Laboratory
provides uncertainties that are dependent on both microphone type and on frequency. These
uncertainties range from ± 0.03 dB to ± 0.1 dB and are based on a standard uncertainty
multiplied by a coverage factor of k = 2. For the purposes of this paper the worst case of
± 0.1 dB will be assumed to apply to all microphones and frequencies between 50 Hz and
10 kHz.
3.4
SUMMARY
OF
INDIVIDUAL
STANDARD
UNCERTAINTY
CONTRIBUTIONS ASSOCIATED WITH THE SOUND LEVEL METER IN
USE
All values of standard uncertainty associated with the operation of a sound level meter as
described in sub-sections 3.3.1 to 3.3.6 are summarised below together with an indication of
their associated statistical distribution. These values are then listed in Table 7 together with
their uncertainty contribution, u to the final combined total standard uncertainty, u(LW) to be
associated with the operation of a sound level meter. The value of u(LW) is given by:
u (LW ) =
6
åu
2
i
i =1
15
(5)
NPL Report DQL-AC 002
where, ui is the ith uncertainty contribution.
Values of the seven contributions are listed below using the error notation from Section 2.
A value of up to 0.14 dB has been estimated using a simple procedure for a
δdir
hemispherical enveloping surface as described in sub-section 3.3.1. It is assumed that over the
small range of angle of incidence considered, the change in microphone sensitivity is a linear
function of angle of incidence and so the statistical distribution associated with δdir is
assumed to be triangular.
Some measurements carried out for this study (see sub-section 3.3.2) have indicated a
δobs
value of uncertainty of up to ± 0.2 dB. The statistical distribution associated with δobs is
assumed to be rectangular.
δcase Using values of case correction supplied for sound level meter verification purposes, a
value of 0.011 dB has been proposed. The statistical distribution associated with δcase is
assumed to be rectangular.
δtemp The potential change in microphone sensitivity resulting from measurements being
carried out at temperatures removed from the calibration temperature is assumed to be
0.12 dB (see sub-section 3.3.3). The change in sensitivity is a linear function of temperature
and so the statistical distribution associated with δtemp is assumed to be triangular.
δpres The potential change in microphone sensitivity resulting from measurements being
carried out at pressures removed from the calibration value is assumed to be 0.0722 dB (see
sub-section 3.3.4). The change in sensitivity is a linear function of pressure and so the
statistical distribution associated with δpres is assumed to be triangular.
δws
Using values of windscreen correction supplied (from manufacturer’s instruction
manuals) for sound level meter verification purposes, a value of 0.019 dB has been proposed.
The statistical distribution associated with δWS is assumed to be rectangular.
δmic This is dependent on the uncertainty associated with the calibration of the microphone
and is taken as ± 0.1 dB, assuming a coverage factor of k = 2 (see sub-section 3.3.7) with a
normal distribution
Table 7
Uncertainties associated with the sound level meter in use
Quantity
standard
uncertainty
± dB
probability
distribution
distribution
divisor
Uncertainty
contribution
± dB
δdir
δobs
δcase
δtemp
δpres
δws
∆mic
0.14
0.20
0.011
0.12
0.0722
0.019
0.10
Triangular
Rectangular
Rectangular
Triangular
Triangular
Rectangular
Normal
√6
√3
√3
√6
√6
√3
2
0.06
0.12
0.01
0.05
0.03
0.01
0.05
TOTAL
0.155
16
NPL Report DQL-AC 002
It can be seen from Table 7 that the final combined total standard uncertainty of 0.155 dB is
smaller than the corresponding result for group (a). It is clear that in order to reduce the
magnitude of the final combined total standard uncertainty effort should be concentrated on
reducing δobs. If the indication on the sound level meter was observed remotely via an ac
output (a common feature on class 1 meters) then δobs will be reduced to zero and the value of
the final combined total standard uncertainty will reduce to 0.098 dB.
17
NPL Report DQL-AC 002
4.
VALUES OF INDIVIDUAL UNCERTAINTY CONTRIBUTIONS AS
A FUNCTION OF FREQUENCY
Here values of uncertainty associated with the various contributions discussed in Section 2,
are assessed as a function of frequency. Data is available for frequencies at preferred octave
band centre frequencies from 31.5 Hz to 12.5 kHz.
In group (a), the only uncertainty contribution for which there is information on any potential
frequency dependence is δfr, the correction associated with the frequency-weighting network.
In this case it will be the correction associated with a linear weighting. Values of uncertainties
associated with a linear weighting are considered in sub-section 4.1.
In group (b), information regarding any potential frequency dependence is available for all
contributions except δobs, the correction associated with the influence of an observer. Values
for uncertainties associated with the remaining six contributions, δdir, δcase, δtemp, δpres, δws and
δmic are considered in sub-section 4.2.
4.1
FREQUENCY DEPENDENT UNCERTAINTIES ASSOCIATED WITH THE
OPERATION OF A SOUND LEVEL METER
The only frequency dependent uncertainty in group (a) is δfr, which in this sub-section will be
the correction associated with a linear weighting. Data have been obtained from
measurements made during sound level meter verification as described in sub-section 3.1.
BS 7580: Part 1: 1997 requires that the sound level meter, when operating in its linear mode,
is tested over the range 31.5 Hz to 12.5 kHz using a continuous sinusoidal signal at preferred
octave-band centre frequencies. The signal level at 1 kHz was set as an indication of the
reference sound pressure level. The response of the sound level meter was compared to that
of tables IV and V of BS EN 60651:1994.
The difference between an ideal flat response and that obtained for each of the 22 sound level
meters (see sub-section 3.1) was noted and the standard deviation of these data about the
mean value for the 22 meters for each frequency band calculated. These standard deviations
(s.d.) are shown in Table 8.
Table 8
Freq (kHz)
s.d. (dB)
Standard deviations for δfr at octave-band centre frequencies
0.0315
0.21
0.063
0.16
0.125
0.11
0.250
0.08
0.500
0.06
1
0
2
0.07
4
0.10
8
0.13
12.5
0.12
The value of the standard deviation at 1 kHz is zero only because it is the datum to which the
sound level meter response at other frequencies is compared. It is assumed that its true value
is similar to the adjacent frequency bands. The value of standard deviation for the A-weighted
network is 0.07 dB (see sub-section 3.1.1). It can be seen from Table 8 that a similar result is
obtained for frequencies between 250 Hz and 2 kHz with an increase to about 0.12 dB above
2 kHz and gradually rising to 0.21 dB at lower frequencies.
18
NPL Report DQL-AC 002
4.2
FREQUENCY DEPENDENT UNCERTAINTIES ASSOCIATED WITH THE
USE OF A SOUND LEVEL METER
Information on the frequency dependence of uncertainties in group (b) exist for all
contributions except for δobs, the correction associated with the influence of an observer.
Frequency dependent uncertainties associated with the remaining six contributions, δdir, δcase,
δtemp , δpres , δws and δmic are considered in sub-sections 4.2.1 to 4.2.6.
4.2.1 Frequency dependent uncertainties associated with δdir.
Values of δdir have been estimated for a hemispherical enveloping surface according to
ISO 3744:1994, as described in sub-section 3.3.1. In this simple estimation, it is assumed that
the hemisphere radius is twice the largest source dimension and that the source is radiating
equally from all positions on its surface. The maximum angle of incidence on the microphone
diaphragm is taken as 20° (zero degrees is normal incidence) and the corresponding changes
in microphone sensitivity, from that at normal incidence, at octave-band frequency intervals
estimated from manufacturers data. Values of δdir of 0.03 dB, 0.09 dB, 0.33 dB and 0.75 dB
for frequencies of 2 kHz, 4 kHz, 8 kHz and 12.5 kHz are estimated. Below 2 kHz values are
taken as zero. The corresponding A-weighted value assuming a flat frequency spectrum is
0.14 dB (see sub-section 3.1.1). The significance of this variation with frequency and
differences from the A-weighted value is discussed in sub-section 5.2.
4.2.2 Frequency dependent uncertainties associated with δcase.
As discussed in sub-section 3.3.3, IEC 61672 requires that corrections be applied to account
for the effect of the presence of the sound level meter in the sound field and, for the purposes
of providing a contribution to the combined total standard uncertainty, it is proposed that a
value of 10% of the correction supplied by the manufacturer be assumed to be representative
of δcase. Values of corrections for the sound level meters presented for verification purposes
(see sub-section 3.1) have been examined. These corrections varied from one sound level
meter to another so, for the purposes of providing a value of δcase the largest has been
assumed for each octave-band frequency considered. Values below 125 Hz are zero, with
those above 125 Hz shown in Table 10.
Table 10
Freq (kHz)
δcase (dB)
Value of δcase at octave-band centre frequencies
0.125
0.01
0.250
0.01
0.500
0.02
1
0.02
2
0.07
4
0.01
8
0.04
12.5
0.12
It can be seen that values of δcase generally increase as frequency increases, although there is a
“peak” at 2 kHz. The value of δcase when considering A-weighted levels is 0.011 dB (see subsection 3.3.3). So, generally values corresponding to particular frequencies are similar to
those values corresponding to A-weighted levels, with the exception of the large increases at
2 kHz and at 12.5 kHz. The values of δcase are based on correction data supplied by the
manufacturers and the relatively large values of δcase are generally common to all sound level
19
NPL Report DQL-AC 002
meters. This generality is probably because most sound level meters are similar in physical
shape and size. It is likely that if sound level meters were designed to have a different
physical appearance then these large values of δcase may reduce or perhaps reappear at
different frequencies. However, if these large values of δcase are a problem they may be
reduced to zero by using a microphone on an extension lead. The significance of this
variation with frequency and differences from the A-weighted value is discussed in subsection 5.2.
4.2.3 Frequency dependent uncertainties associated with δtemp.
As discussed in sub-section 3.3.4, there are variations in sound level meter output with
changes in ambient temperature resulting from changes in the sensitivity of the microphone.
The rate of change of microphone sensitivity with temperature change is frequency
dependent, with values increasing above approximately 1 kHz. Whilst this increase for some
microphone types is quite significant (e.g. for a B&K type 4160 the increase is from 0.003 dB
per K at 1 kHz to 0.0137 dB per K at 8 kHz) the values at the higher frequencies are still
always less than the worst case values adopted in sub-section 3.3.4. Therefore, for the
purposes of this paper it is assumed that the uncertainty contributions when considering
frequency band limited data will be similar to (and certainly not greater than) the A-weighted
value discussed in sub-section 3.3.4.
4.2.4 Frequency dependent uncertainties associated with δpres.
As discussed in sub-section 3.3.5, it is assumed that variations in sound level meter output
with changes in atmospheric pressure result from changes in the sensitivity of the
microphone. Over the frequency range of interest here, the variations of the rate of change of
microphone sensitivity with changes in ambient atmospheric pressure are generally
independent of frequency. However, for some microphones there is some evidence of a slight
decrease in the rate of change between 4 kHz and 8 kHz and a slight increase at 12.5 kHz.
However, these changes are very small and are very much smaller than the values assumed
for the A-weighted values discussed in sub-section 3.3.5. So, for the purposes of this paper it
is assumed that the uncertainty contributions when considering frequency band limited data
will be similar to (and certainly not greater than) the A-weighted value discussed in subsection 3.3.5.
4.2.5 Frequency dependent uncertainties associated with δWS.
As discussed in sub-section 3.3.6, if a windscreen is used corrections must be applied to
account for the insertion loss due to the windscreen and, for the purposes of providing a
contribution to the combined total standard uncertainty, it is proposed that a value of 10% of
the correction supplied by the manufacturer be assumed. Values of corrections for the sound
level meters presented for verification purposes (see sub-section 3.1) have been examined.
These corrections varied from one sound level meter to another so, for the purposes of
20
NPL Report DQL-AC 002
providing a value of δWS the largest has been assumed for each octave-band frequency
considered. Values below 125 Hz are zero, others are shown in Table 11.
Table 11
Freq (kHz)
δWS (dB)
Values of δWS at octave-band centre frequencies
0.125
0.01
0.250
0.02
0.500
0.03
1
0.03
2
0.04
4
0.04
8
0.04
12.5
0.1
It can be seen that values of δWS generally increase as frequency increases. The value of δWS
when considering A-weighted levels is 0.019 dB (see sub-section 3.3.3). So, generally values
corresponding to particular frequencies are similar to those values corresponding to Aweighted levels, although there is a large increase at 12.5 kHz. The significance of this
variation with frequency and differences from the A-weighted value is discussed in subsection 5.2.
4.2.6 Frequency dependent uncertainties associated with δmic.
As discussed in sub-section 3.3.7, it is assumed that corrections are applied to account for any
variation in microphone sensitivity with frequency so, only uncertainties associated with the
calibration of the microphone need to be considered. A worst-case value of ± 0.1 dB was
assumed for A-weighted data. When worst case values are assumed as a function of frequency
then this value of ± 0.1 dB is reduced to ± 0.07 dB for frequencies up to 4 kHz, to ± 0.09 dB
for 8 kHz and remains at ± 0.1 dB for 12.5 kHz. The significance of this variation with
frequency and differences from the A-weighted value is discussed in sub-section 5.2.
21
NPL Report DQL-AC 002
5.
TOTAL COMBINED STANDARD UNCERTAINTY FOR USE IN
ISO STANDARDS
5.1
A-WEIGHTED NOISE EMISSION LEVELS
The combined total standard uncertainties from Tables 6 and 7 are expressed as a standard
deviation. A standard deviation is a measure of the spread of a set of sound power level
determinations, describing how values typically differ from the average. It is possible using a
set of statistical tables to estimate the possibility of a single sound power level determination
being different from the average sound power level by a given amount. It is usual to express
this as a percentage of sound power level determinations that are expected to be outside a
given range of sound power level. This percentage is called a confidence level and is an
indication of how confident you are of a sound power level determination not being outside
this range. For the purposes of inclusion in ISO standards it is suggested that a confidence
level of 95% is adopted. For the case where the number of sound power level determinations
in a set were very large (which is assumed here) it can be stated that 95% of sound power
level determinations will be within a range of the average sound power level plus 1.96 (this is
usually approximated to 2.0) times the standard deviation. This constant, 2.0 is called the
coverage factor. The coverage factor multiplied by the standard deviation is known as the
expanded uncertainty. The Guide to the Expression of Uncertainties in Measurement requires
a sound power level, LW to have a value of expanded uncertainty, U, to be specified, such that
the interval [LW-U, LW+U] covers a range of the values (this usually expressed as a
percentage and here it is assumed to be 95%) of LW that might reasonably be attributed to LW.
To that purpose, a coverage factor, k, is used (here it is assumed that the coverage factor is 2),
such that U=k.u (see Equation 2).
Using the data from Table 6 and 7, values of expanded uncertainty associated with a sound
level meter can be calculated and are shown in Table 12. In Table 12, the values listed for
sound level meter, (slm), in use, are the results of combining all the uncertainty contributions
from both group (a) and (b) described in Section 2.
Table 12
Values of expanded uncertainty associated with a sound level meter
Slm operation:
Group (a)
Slm in use:
Group (a) and (b)
Time weighting
Combined uncertainty
Expanded uncertainty
Fast
Slow
Fast
Slow
0.394
0.362
0.423 (0.406)
0.394 (0.375)
0.788
0.724
0.846
0.788
It can be seen from Table 12 that with the contribution from group (b), (the uncertainties
associated with the use of a sound level meter), increase the combined total standard
uncertainties obtained from group (a), (the uncertainties associated with the operation of a
sound level meter), by only 0.029 dB and 0.032 dB for the “fast” and “slow” time weighting
respectively. These differences are reduced to 0.012 dB and 0.013 dB if it is assumed that δobs
is zero (using the values in parenthesis in Table 12). The difference between the combined
22
NPL Report DQL-AC 002
total standard uncertainty obtained for “fast” time weighting with that using “slow” time
weighting is about 0.03 dB for all combinations shown in Table 12. In order to reduce the
magnitude of the combined total standard uncertainty the use of the “slow” time weighting is
recommended, especially when the format for reporting uncertainty values is considered (see
following paragraph). Thus the value of the combined total standard uncertainty for a sound
level meter is 0.394 dB or 0.375 dB when there is no observer present. However, this may be
further reduced assuming there is no range change to 0.355 dB and 0.334 dB when there is no
observer present. However, these reductions are not significant when the format for reporting
uncertainty values is considered (see following paragraph).
The number of significant figures in a reported uncertainty should always reflect practical
measurement capability. For the purposes of supplying a value of δslm, for inclusion in
Equation 1, it is proposed that the combined total standard uncertainty is expressed to one
place of decimals. Furthermore it is generally accepted that uncertainties should be rounded
up to the appropriate number of significant figures(20). Thus the reported combined total
standard uncertainty for measurements carried out using the “slow” time weighting (with or
without an observer and with or without a range change) is 0.4 dB. This is increased to 0.5 dB
if there is an observer and “fast” time weighting is used.
Similarly the value of expanded uncertainty assuming a confidence level of 95% will be
0.8 dB for “slow” time weighting (with or without an observer and with or without a range
change) and 0.9 dB for “fast” time weighting with an observer. However, for the purposes of
inclusion in an ISO standard only the value of combined total standard uncertainty is required
as the value of δslm will be included in the combined total standard uncertainty for a sound
power measurement that will, in turn, be expressed as an expanded uncertainty.
As discussed in sub-sections 3.1.1 and 3.1.2 the values of δfr and δlin may not be directly
attributable to variations between sound level meters. If these two input quantities were,
therefore, assumed to be negligible, the final combined total standard uncertainty would be
reduced to 0.344 dB (when δdl is zero) and to 0.312 dB (when δdl and δobs are zero). Never the
less, the reported uncertainty according to Reference 20 will still be 0.4 dB. However, it may
be that for the purposes of including a value for δslm in Equation 1 the unrounded values may
be used and the requirements of Reference 19 only considered for the uncertainty associated
with a sound power level determination.
5.2
BAND-LIMITED NOISE EMISSION LEVELS
Whilst the prime driver for uncertainty data is to demonstrate compliance with noise emission
level declaration as required for numerous regulatory purposes (see Section 2) it is of interest
to examine potential changes to the combined total uncertainty as applied to frequency band
limited data. The effect of variations in the value of uncertainty contributions for group (a) is
considered in sub-section 5.2.1 and for group (b) in sub-section 5.2.2.
23
NPL Report DQL-AC 002
5.2.1 The effect on the final total combined uncertainty of frequency variations in
group (a)
Applying values of δfr from Table 8 in sub-section 4.1 to the data in Table 6 produces the
values of the final combined total uncertainty for “fast” and “slow” time weightings shown in
Table 13.
Table 13
group (a)
Freq (kHz)
Fast
Slow
Final combined total uncertainty at octave-band centre frequencies for
0.0315
0.441
0.413
0.063
0.419
0.390
0.125
0.403
0.372
0.250
0.396
0.364
0.500
0.392
0.360
2
0.394
0.362
4
0.400
0.369
8
0.409
0.378
12.5
0.406
0.375
The value of the standard deviation at 1 kHz is not shown as the value of δfr is given as zero
in Table 8 (because it is the datum to which the sound level meter response at other
frequencies is compared) and so, it is assumed that its value is similar to the adjacent
frequency bands. The value of the final combined total uncertainty for the A-weighted
network is 0.394 dB for “fast” and 0.362 dB for “slow” time weighting (see Table 6). It can
be seen from Table 13 that a similar result is obtained, for both “fast” and “slow” time
weighting, for frequencies between 250 Hz and 2 kHz with an increase of up to 0.016 dB
above 2 kHz and increases of approximately 0.01 dB, 0.03 dB and 0.05 dB at frequencies of
125 Hz, 63 Hz and 31.5 Hz respectively. Considering the slow time weighting data (as
recommended in 5.1 above, the range of values of from 0.413 dB to 0.360 dB may be reduced
to 0.376 dB to 0.317 dB if it is assumed that there is no range change.
It may be concluded therefore that the A-weighted value for the total final combined
uncertainty calculated for contributions in group (a) may be assumed for frequency band
limited data.
5.2.2 The effect on the final total combined uncertainty of frequency variations in
group (b)
Values of uncertainty contributions for band limited data were considered in sub-sections
4.2.1 to 4.2.5. The values of their uncertainty contribution are summarised in Table 14
together with values of final total combined uncertainty (TOTAL).
The A-weighted final total uncertainty from Table 7 is 0.155 dB. It can be seen from Table 14
that for frequencies below 4 kHz the value of the final total combined uncertainty is similar to
the A-weighted value but there is an increase to 0.192 dB at 8 kHz and to 0.364 dB at
12.5 kHz.
If the group (b) data from Table 14 is combined with the group (a) data from Table 13 for
slow time weighting, values of the final total combined uncertainty suitable for inclusion in
an ISO standard may be assessed. Values calculated using all contributions and then assuming
there is no observer present (δobs = 0) and also assuming there is no range change (δobs = 0 and
δdl = 0) are listed in Table 15.
24
NPL Report DQL-AC 002
Table 14
Freq (kHz)
0.0315
0.063
0.125
0.25
0.5
1.0
2.0
4.0
8.0
12.5
Table 15
Values of uncertainty contribution for group (b)
δobs
0.12
0.12
0.12
0.12
0.12
0.12
0.12
0.12
0.12
0.12
δdir
0.06
0.06
0.06
0.06
0.06
0.06
0.01
0.04
0.13
0.31
Uncertainty contribution
δcase
δtemp
δpres
.01
0.05
0.03
.01
0.05
0.03
.01
0.05
0.03
.01
0.05
0.03
.01
0.05
0.03
.01
0.05
0.03
.04
0.05
0.03
.01
0.05
0.03
.02
0.05
0.03
.07
0.05
0.03
δWS
0.01
0.01
0.01
0.01
0.02
0.02
0.02
0.02
0.02
0.06
δmic
0.04
0.04
0.04
0.04
0.04
0.04
0.04
0.04
0.04
0.05
TOTAL
0.151
0.151
0.151
0.151
0.152
0.152
0.157
0.145
0.192
0.364
Band limited values of final combined total uncertainty
Freq (kHz)
0.0315
0.063
0.125
0.250
0.500
1.0
2.0
4.0
8.0
12.5
All contributions
.440
.418
.401
.394
.391
.392
.395
.400
.424
.523
δobs = 0
.423
.400
.383
.375
.372
.373
.376
.382
.407
.509
δobs = 0, δdl = 0
.387
.362
.343
.334
.331
.332
.335
.342
.370
.480
It can be seen that assuming there is no observer and no range change the reported value of
the final total combined uncertainty is 0.4 dB for all frequencies except 12.5 kHz where the
value is 0.5 dB.
For the majority of ISO sound power standards the frequency range of interest is specified as
covering one-third octave bands centre frequencies from 100 Hz to 10 kHz. In this case the
reported final total combined uncertainty for band limited noise will be 0.4 dB.
25
NPL Report DQL-AC 002
6.
CONCLUSIONS AND RECOMMENDATIONS
For contributions associated with the operation of a sound level meter, it is clear that to
reduce the magnitude of the final combined total standard uncertainty, effort should be
concentrated on reducing δtime , δrms and δdl. The value of δtime for the “slow” time weighting
is less than that for “fast”. Sound level meters with a single large dynamic range will of
course reduce δdl to zero. The value of δdl will also be zero if the indicator range that is used
for the sound pressure level measurement is the same as that used when applying the sound
calibrator.
To reduce the magnitude of the combined total standard uncertainty the use of the
“slow” time weighting is recommended.
It is recommended that the indicator range used for measurements is the same as that
used for the sound calibrator.
For contributions associated with the use of a sound level meter, it is clear that to reduce the
magnitude of the final combined total standard uncertainty effort should be concentrated on
reducing δobs. If the indication on the sound level meter was observed remotely via an ac
output (a common feature on class 1 meters) then δobs would be reduced to zero.
It is recommended that the indication on the sound level meter is observed remotely via
an ac output (a common feature on class 1 meters).
The contribution from uncertainties associated with the use of a sound level meter, increase
the combined total standard uncertainties obtained from the uncertainties associated with the
operation of a sound level meter, by approximately 0.03 dB.
The actual estimated value of δslm for A-weighted noise emission levels assuming a slow time
weighting is 0.39 dB. This value is reduced to 0.33 dB if there is no range change and there is
no observer present (δdl and δobs =0) and reduced further still if it assumed that δfr and δdl are
assumed negligible.
The reported combined total standard uncertainty for measurements carried out using the
“slow” time weighting (with or without an observer and with or without a range change) is
0.4 dB. This is increased to 0.5 dB if there is an observer and “fast” time weighting is used.
It is recommended that the value of δslm to be used when reporting a sound pressure
level measurement is 0.4 dB but for an uncertainty budget in ISO sound power
standards, a value of 0.33 dB should be used.
For the majority of ISO sound power standards the frequency range of interest is specified as
covering one-third octave bands centre frequencies from 100 Hz to 10 kHz. In this case the
reported final total combined uncertainty associated with band limited noise emission is the
same as the A-weighted value.
26
NPL Report DQL-AC 002
7.
ACKNOWLEDGEMENTS
The author gratefully acknowledges the financial support of the UK Department of Trade &
Industry (National Measurement System Directorate).
27
NPL Report DQL-AC 002
8.
REFFERENCES
1
Council Directive 2000/14/EC of 8 May 2000 on the approximation of the laws of the
Member States relating to noise emission in the environment by equipment for use outdoors.
Official Journal of the European Communities N°L162, Volume 43.
2
INTERNATIONAL ORGANIZATION FOR STANDARDIZATION, ISO 3741:1999,
Acoustics - Determination of sound power levels of noise sources using sound pressure Precision methods for reverberation rooms.
3
INTERNATIONAL ORGANIZATION FOR STANDARDIZATION ISO 3743-1:1994,
Acoustics - Determination of sound power levels of noise sources - Engineering methods for
small, movable sources in reverberant fields - Part 1: Comparison method for hard-walled test
rooms.
4
INTERNATIONAL ORGANIZATION FOR STANDARDIZATION, ISO 3743-2:1994,
Acoustics - Determination of sound power levels of noise sources - Engineering methods for
small, movable sources in reverberant fields - Part 2: Methods for small reverberation test
rooms.
5
INTERNATIONAL ORGANIZATION FOR STANDARDIZATION, ISO 3744:1994,
Acoustics - Determination of sound power levels of noise sources using sound pressure Engineering method in an essentially free field over a reflecting plane.
6
INTERNATIONAL ORGANIZATION FOR STANDARDIZATION, ISO 3745:2003,
Acoustics - Determination of sound power levels of noise sources using sound pressure Precision methods for anechoic and hemi-anechoic rooms.
7
INTERNATIONAL ORGANIZATION FOR STANDARDIZATION, ISO 3746:1996,
Acoustics - Determination of sound power levels of noise sources using sound pressure Survey method using an enveloping measurement surface over a reflecting plane.
8
INTERNATIONAL ORGANIZATION FOR STANDARDIZATION, ISO 3747:2000,
Acoustics - Determination of sound power levels of noise sources using sound pressure Comparison method in situ.
9
BRITISH STANDARDS INSTITUTION, PD 6461-3:1995, Vocabulary of metrology Part 3. Guide to the expression of uncertainty in measurement.
10
KLAUS BRINKMAN, Contribution to the overall uncertainty of noise measurements due to
imperfect performance of the sound level meter, 2000. Discussion paper for IEC/TC29/WG4.
11
INTERNATIONAL ELECTROTECHNICAL COMMISSION, IEC 60651:2000,
Specification for sound level meters
12
BRITISH STANDARDS INSTITUTION, BS 7580:Part1:1997, Specification for the
verification of sound level meters – Part 1, comprehensive procedure
13
BRITISH STANDARDS INSTITUTION, BS EN 60651:1994. Specification for sound level
meters
28
NPL Report DQL-AC 002
14
BRITISH STANDARDS INSTITUTION BS EN 60804:1994, Specification for integratingaveraging sound level meters
15
INTERNATIONAL ELECTROTECHNICAL COMMISSION, IEC 61672:part1:2002,
Electroacoustics – sound level meters – part1: specifications
16
KNUD RASMUSSEN, The influence of environmental conditions on the pressure sensitivity
of measurement microphones, Brüel & Kjær Technical Review No 1, 2001.
17
INTERNATIONAL ELECTROTECHNICAL COMMISSION, IEC 61094:1992,
Measurement microphones – part 2 – primary method for pressure calibration of laboratory
standard microphones by the reciprocity technique.
18
Various manufacturers instrumentation handbooks.
19
INTERNATIONAL ORGANIZATION FOR STANDARDIZATION, draft revision of
ISO 3744 prepared by ISO committee TC43/SC1/WG28, August 2003
20
UNITED KINGDOM ACCREDITATION SERVICE, M003, The expression of uncertainty
and confidence in measurement, 1997
29
-ve
+ve
-ve
+ve
Fast
Slow
Rms
Ref
range
Other
range
Time
ave
-0.1
0
0
0.06
0.1
-0.6
-0.3
1
3
4
0.3
-0.3
-0.1
-0.24 -0.02 0.04
0.04 0.01
0
-0.2
0
0
-0.01
0
-0.07
0.3
-0.1
0
0.1
0
0.5
2
6
7
8
9
14
15
16
0.3
0.2
0.2
0.1
0.2
0.20
0.20
0.2
0.04
0.03
0.05
0.03
-0.10
-0.10
17
-0.08
0.01
-0.02
0.07
-0.10
0.10
18
-0.02
0.00
-0.55
-0.05
0.00
0.00
0.01
-0.01
-0.12
-0.12
-0.10
0.00
20
22
0.10
0.29 0.03
-0.08 -0.05
-0.05 -0.04
-0.05 0.03
-0.20 0.00
-0.10 0.00
21
-0.10 -0.20 -0.20 -0.20 -0.10 -0.10 -0.10 -0.20 -0.20
-0.05
0
0.02 -0.08 -0.01 -0.16 -0.15 0.26 0.11 0.00 -0.09 0.08
-0.06 0.01 -0.16 0.11
0
0.15 0.03 0.2 -0.01 0.00 0.10 -0.03
-0.02 -0.02 -0.22 -0.02 -0.01 0.03 -0.09
0
0.03 -0.66 0.32 0.03
-0.06
0
0
0.11
0
0.16 -0.04
0
0.06 -0.08 0.41 0.08
-0.1
0
-0.1 -0.1
0
-0.10 -0.30 -0.5 -0.70 0.00 -0.10 -0.30
-0.2
0
-0.1
0
0
-0.20 0.00 -0.2 -0.20 0.00 0.10 -0.70
5
Factor (dB)
Sound level meter
10
11
12
13
19
Ideal
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
-39.4 -39.6 -39.3 -39.5 -40.3 -39.2 -39.6 -39.4 -39.6 -39.5 -39.6 -39.4 -39.5 -39.9 -39.5 -39.9 -39.9 -39.5 -39.1 -39.7 -39.7 -39.5 -39.4
-26.2 -26.3 -26 -26.3 -26.6 -26 -26.3 -26.2 -26.3 -26.2 -26.4 -26.2 -26.1 -26.4 -26.2 -26.4 -26.6 -26.3 -26.1 -26.3 -26.3 -26.1 -26.1
-16.1 -16.3 -16 -16.3 -16.4 -16.1 -16.2 -16.1 -16.3 -16.2 -16.3 -16.2 -16.1 -16.3 -16.2 -16.4 -16.4 -16.3 -16.1 -16.2 -16.2 -16.1 -16.2
-8.6 -8.7 -8.5 -8.7 -8.7 -8.7 -8.7 -8.6 -8.8 -8.7 -8.8 -8.7 -8.7 -8.7 -8.7 -9.1 -8.9 -8.8 -8.7 -8.8 -8.7 -8.6 -8.7
-3.2 -3.3 -3.2 -3.3 -3.2 -3.3 -3.3 -3.3 -3.3 -3.3 -3.3 -3.3 -3.4 -3.3 -3.3 -3.5 -3.3 -3.3 -3.3 -3.3 -3.3 -3.2 -3.3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1.2
1.2
1.1
1.2
1.2
1.2
1.2
1.2
1.2
1.2
1.1
1.1
1.1
1.2
1.2
1.4
1.2
1.2
1.2
1.2
1.1
1.1
1.2
1
1
0.8
0.9
0.9
1
0.9
1
0.9
0.9
0.9
0.8
0.9
0.9
1
1.2
1
1
1
1.1
0.9
0.8
0.9
-1.1 -1.1 -1.4 -1.2 -1.2
-1
-1.2 -1.2 -1.2 -1.2 -1.2 -1.2 -1.2 -1.2
-1
-1
-1.2 -1.2 -0.9 -1.1 -1.2 -1.5 -1.2
-4.3 -4.2 -4.5 -4.2 -4.4 -4.1 -4.3 -4.3 -4.3 -4.4 -4.3 -4.3 -4.4 -4.3 -4.3 -4.4 -4.3 -4.3 -3.5
-4
-4.3 -4.5 -4.4
Table A.1.2 Factors relating to LINEARITY, TIME WEIGHTING and DETECTOR characteristics from verification measurements
Freq
31.5
63
125
250
500
1k
2k
4k
8k
12.5k
A-weighting factor (dB)
Sound level meter
Table A.1.1 Values of A-WEIGHTING factors from verification measurements
ANNEX A – MEASUREMENT DATA
Was this manual useful for you? yes no
Thank you for your participation!

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

Download PDF

advertising