Experimental investigation of possible improvements of ISO 9614

Spyros Brezas 1

Department of Music Technology and Acoustics, Hellenic Mediterranean University E. Daskalaki, Perivolia, 74133, Rethymno, Greece

Volker Wittstock 2

Physikalisch Technische Bundesanstalt Bundesallee 100, 38116, Braunschweig, Germany

Fabian Heisterkamp 3

Federal Institute for Occupational Safety and Health (BAuA) Friedrich-Henkel-Weg 1-25, 44149 Dortmund, Germany

ABSTRACT Sound power level is the key quantity to describe the noise emission of products and is needed to reduce noise at work, at home, and in the environment, e.g. for promoting and selecting low noise products (Sell and Buy Quiet). Sound power determination based on sound intensity measurements has advantages in non-ideal acoustic environments, e.g. indoors, but outside of acoustic test rooms, and is standardized in the ISO 9614 series. The current version of all three parts of ISO 9614 is their first edition. The advances of technology and knowledge indicate that the whole series should be revised. The improvement of the sound intensity method was investigated regarding a number of aspects. The use of a single spacer up to frequencies, which are outside of the currently usable frequency range, was studied. The e ff ect of electrical noise was investigated to extend the applicability of the method to low noise sources. The discretization of the enveloping surface is discussed in relation to the accuracy grade of the measurements. The possibility to measure with more than one probe is also presented. The results indicate that there is the possibility for a substantial improvement of the ISO 9614 series.

1. INTRODUCTION

Sound power level L W is the principal descriptor for noise emission from a variety of products, such as equipment and machinery. Its importance is mirrored by the number of international and national legislative documents describing the noise emissions of such products. Examples of these documents are the European Outdoor Noise Directive [1] and the Australian regulation for the protection of

1 sbrezas@hmu.gr

2 volker.wittstock@ptb.de

3 heisterkamp.fabian@baua.bund.de

a slaty. inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS O ¥, ? GLASGOW

environment operations [2].

The sound power determination can be performed by sound pressure measurements, by sound intensity measurements, and in some cases by surface vibration measurements. ISO 3740 [3] can be used as a guide for the selection of standards and technical specifications to determine the sound power level. By comparison, di ff erences among the standards result from di ff erent equipment, the acoustic environment where measurements are performed and the resulting grade of accuracy. ISO 9614 [4–6] series is used when sound intensity is measured. Although sound pressure measurements are more frequently used to determine the sound power level, limitations exist, such as low accuracy when measurements take place in di ffi cult environments. On the other hand, the sound intensity method overcomes the majority of these limitations. It can be applied in ordinary surrounding environments and is mainly a ff ected by the background noise level. Still, the sound pressure method is usually used for sound power determination. This can be explained by a number of reasons. First, ISO 9614-1 [4] and ISO 9614-2 [5] have not been revised since their publication. Second, the terminology among all parts of ISO 9614 series [4–6] is not consistent. Apparently, the standards do not follow recent progress in technology and science, and do not correspond to the state-of-the-art. Another reason is the requirement of a large number of measurement points to achieve a certain accuracy, which under certain conditions, may not be finally achieved. In this case no results are reported.

Jacobsen proposed a number of improvements for the sound intensity method [7], which were the basis for the present contribution. A study was performed for the investigation of possibilities to improve and simplify the sound intensity method. Measurements were performed at Germany’s National Metrology Institute (PTB) within a project of the Federal Institute for Occupational Safety and Health (BAuA), which initiated the subproject, aiming at a practical adaption of the sound intensity measurement method, as a part of a focus project to simplify noise emission measurement methods [8]. Results are presented concerning the use of a single spacer for the entire frequency range up to 10 kHz. The possibility to measure low noise sources was studied in terms of electrical noise. Attention was also paid to the measurement surface discretization in relation to the measurement accuracy. Finally, the possibility to measure with more than a single probe is discussed.

2. MEASUREMENTS

2.1. Sources under test and surrounding environments The ISO 9614 series [4–6] describes sound intensity measurements in broadband frequency analysis, i.e. in octave or one-third octave bands. It also includes Annexes about the e ff ects of airflow and absorption within the measurement surface. Four sources were investigated. First, a source emitting broadband noise (BS), second the BS with added absorption (BSWA), third a source emitting a signal with tonal characteristics (TS) and fourth, the TS with added absorption (TSWA). Figure 1 shows the BSWA and Figure 2 the TSWA. BS consists of two aerodynamic reference sound sources (Brüel & Kjær, Type 4204) at di ff erent heights provided by a wooden box with an open side. TS is a half-dodecahedron loudspeaker, fed with a multi-sine signal according to

31 X

s = A

i = 1 sin (2 π f i + φ i ) , (1)

where A is the signal amplitude, f i the centre frequency of the i − th one-third octave band in the range between 20 Hz and 20 kHz rounded to the nearest decade and φ i the related phase. The phase for each frequency was randomly generated once and the same values were used for all TS and TSWA

measurements to ensure repeatability. For similar reasons, a correction was applied to the TS and TSWA emitted signal provided by

C em = 10 lg Re  S V in I in

Re  S V in I in , ref ! dB , (2)

where Re  S V in I in and Re  S V in I in , ref is the real part of the cross-spectrum of the input voltage and current of the signal fed to the TS and a reference cross-spectrum respectively. The reference cross-spectrum was the mean value of all cross-spectra used.

BS and TS were the same sources shown in Figure 1 and Figure 2 having the absorption panel removed. The panel surface was 0,7 m 2 and its thickness 100 mm. It was made of a polyester fibre mat.

Figure 1: Broadband source with absorption.

Figure 2: Tonal source with absorption.

Measurements were performed in five surrounding environments, namely a hemi-anechoic room, a reverberation room (200 m 3 ), an open space in a basement, a reverberant room (50 m 3 ) and the same

room with extra absorption. To provide extra variation, measurements were performed in the hemi- anechoic room after adding 10 reflecting panels with 7 m 2 total surface. The panels were positioned in front of the wedges on two orthogonal sides of the room (five on each side).

2.2. Sound intensity probes and measurement surface According to the ISO 9614 series, measurements can be either performed by scanning [4,5] or at discrete points [6]. Three phase-matched microphone pairs (two Brüel & Kjær Type 4181 and one Type 4197) were attached to three sound intensity probes.

The scanning of the measurement surface in the hemi-anechoic room was performed both automatically and manually, and in the other environments only manually. The automated measurements utilized PTB’s scanning apparatus [9], which scanned a hemisphere of 1,7 m radius. The full scan of the hemisphere was implemented in two parts, one having covered the left quarter sphere and the other the right quarter sphere. The movement of the arc was remotely interrupted at 15° steps for the measurements at discrete points, providing 36 measurement points.

Figure 3: Holder for sound intensity measurements with three probes.

PTB’s sta ff performed manual scans and measurements at discrete points. For the simultaneous use of all three probes, a special holder was assembled (Figure 3). For the measurements at discrete points, the holder also enabled the variation in the surface sampling by selecting which probes to use. During the measurements with the holder, the measurement surface was a parallelepiped. The di ff erent surrounding environments led to variations in the total area of the surface. Each partial surface was divided in 0,5 m x 0,5 m squares. Using all three probes this led to 128 measurement points, while the use of a single probe (the middle one) generated 64 points. The measurement area was reduced for the measurements in the reverberation room and the points were 78 and 26 respectively.

The duration of the automated scanning measurements was 1200 s, while the measurement at each point lasted 180 s for the measurements at discrete points. During the manual scanning measurements, the scanning speed was kept as constant as possible fulfilling the ISO 9614-2 [5] and ISO 9614-3 [6] requirements.

2.3. Sound intensity calculation Sound intensity was calculated using the cross-spectrum of the probe microphones according to

I ( ω ) = 1 ωρ ∆ x Im n S p 1 p 2 ( ω ) o , (3)

where ω is the angular frequency, ρ the density of air, ∆ x the spacer length and Im n S p 1 p 2 ( ω ) o the imaginary part of the probe microphones cross-spectrum [10]. A multichannel analyser recorded the data and calculated the complex cross-spectra in real time. The FFT data (6401 lines, Hanning window) were also averaged by the analyser and further processed in Matlab.

The spacer length is related to the finite di ff erence error [11] through

∆ L I = 10 lg " sin( k ∆ x )

# dB , (4)

k ∆ x

where k is the wave number. For the studied frequency range two spacers were used (12 mm and 50 mm). The frequency limit for the 50 mm spacer was set to 800 Hz ( ∆ L I ≤ 0 . 5 dB).

The frequency analysis was performed in one-third octave bands resulting from the FFT analysis using

M X

m = 1 I ( ω m ) . (5)

I =

A similar analysis was performed for the sound pressure results using

" p 2 A ( ω m ) + p 2 B ( ω m ) 2

M X

# , (6)

p 2 =

m = 1

with p A and p B the sound pressure signals of the microphones. For BS and BSWA the summation included all FFT lines belonging to each one-third octave band, while this was not the case for TS and TSWA. For each one-third octave band only three FFT lines were used (the one corresponding to the centre frequency and the adjacent ones). This was due to the main concentration of energy to the centre frequency and the use of the Hanning window.

2.4. Extraneous noise The current sound intensity standards [4–6] propose actions in case of extraneous noise disturbance. To take into consideration this scenario, additional to the measurements without noise, measurements with added extraneous noise took place in the hemi-anechoic room. The noise source was a loudspeaker used for façade measurements positioned at a corner of the hemi-anechoic room pointing at the measurement surface. A noise generator fed the loudspeaker with white noise. The overall noise level was set to deviate -10 dB, -5 dB, 0 dB, 5 dB and 10 dB from the time and surface averaged sound pressure level of the source. To achieve the same level di ff erences for all sources under test a graphic equalizer was used.

The noise was both stationary and non-stationary to broaden its e ff ects. For the former, the façade loudspeaker was continuously emitting during the measurement. For the non-stationary noise measurements, the automated scanning duration was divided into four equal time intervals. During the first and third interval noise was on, while for the others o ff . For the automated measurements at discrete points, the measurements without noise were combined with measurements with noise. Non-stationarity was achieved by combining measurements without noise with measurements

of various noise levels. Half of the measurement points corresponded to measurements without noise, while the other half to measurements with noise. The same approach was also used for the manual measurements. The partial surfaces of the parallelepiped were divided to those parallel and perpendicular to the façade loudspeaker axis. For the parallel surfaces, measurements without noise were used and for the perpendicular, measurements with noise.

2.5. Indicator for the measuring capability of more than one probe According to the analysis of the report [8] that contains all measurements used for this paper, when n probes are used for the sound intensity measurements, the indicator for the adequacy of the measuring equipment is given by

n X



S 10( L p , i − L I n − δ pI 0 , i ) / (10 dB)  dB , (7)

S i

 F pI n − δ pI 0  ∗ = 10 lg

i = 1

where S i is the partial surface covered b y th e i -th probe, S the total measurement surface, L p , i the sound pressure level of the i -th probe, L I n the averaged signed normal sound intensity level [6] and δ pI 0 , i the pressure-residual intensity index of the i -th probe. The asterisk is used to distinguish the new indicator for more than one probe from the corresponding indicator used for single probe measurements. For the derivation of the equation, the reader may refer to Equation 5.13 of the BAuA report [8]. Due to the sensitivity of the measurement line to various parameters, such as the microphone preampilfier, the pressure-residual intensity index of a probe was measured each time there was a change in the corresponding connection line. For the measurement, a Brüel & Kjær Type ZI 0055 source was used, which emits signal up to 5 kHz. For higher frequencies the value at 5 kHz was used.

3. MEASUREMENT RESULTS

3.1. Influence of electrical noise Contrary to the extraneous noise, electrical noise is not taken into consideration in the current standards [4–6]. According to Jacobsen [7], if F pI n becomes large, the random error related to the finite averaging time could increase, especially at low frequencies. The limitations in sound power determination when electrical noise is present in the connection lines were studied using the TS. The electrical signal fed to the TS was initially set so that the sound pressure level measured by the intensity probe was well above the background noise level. Then, the electrical input was decreased in steps until the measured sound pressure level was strongly influenced by the background noise. During measurements, the scanning apparatus was at a fixed position with all three probes attached. The first part of the investigation focused on di ff erent measurement durations (180 s, 240 s & 300 s), which did not produce significant di ff erences. As a next step, the di ff erence between the sound intensity level for the di ff erent electrical inputs was calculated using

∆ L I , el = L I − L I , ref − C em , (8)

where L I is the sound intensity level for di ff erent inputs, L I , ref the reference level and C em the correction for the tonal sources (see Equation 2). L I , ref corresponded to the highest sound intensity.

The sound intensity level di ff erences calculated by Equation 8 were compared to the di ff erences between the sound pressure level for the electrical inputs and the lowest sound pressure level ( L el ). The comparison is shown in Figure 4. As it can be seen, the variations in sound intensity level become small when the sound pressure level detected by the pair of microphones is 10 dB above the

electrical noise level. Introducing such a criterion into the intensity standards [4–6] would increase their applicability to low noise sources, which is not the case in the current versions.

Figure 4: Sound intensity level di ff erence against the di ff erence between sound pressure level and electrical noise level.

3.2. Use of a single spacer for the entire frequency range In the current ISO standard literature [4–6], the applicable frequency range is between 50 Hz and 6300 Hz. There is no clause concerning the spacer length, which is indirectly addressed in the measurement capability paragraphs. Jacobsen has suggested the use of a single spacer (12 mm) for frequencies up to 10 kHz [7]. This possibility has been investigated, along with the use of a single spacer for frequencies down to 50 Hz as well.

For the latter, a comparison was performed between measurements with di ff erent spacers (12 mm and 50 mm). The measurements in all surrounding environments were performed using both spacers. All results were used for the determination of the related sound power level. The finite di ff erence error is lower than 0.5 dB up to 800 Hz, which was set as the upper limit for the comparison of the di ff erent spacers. The comparison was made in terms of the sound power level di ff erence as

∆ L W (spacer) = L W (12 mm) − L W (50 mm) , (9)

where L W (12 mm) and L W (50 mm) is the sound power level using the 12 mm and 50 mm spacer respectively. The sound power level after the measurement with each spacer was checked in terms of the dynamic capability criterion [6], which by applying Equation 7 is given by  F pI n − δ pI 0  ∗ > − K , (10)

where K is the bias error factor. The value of K is related to the accuracy grade and is 10 dB for precision or engineering and 7 dB for survey measurements [4–6]. The dynamic capability check was performed for both K values. The qualified levels were used for the calculation of sound power level di ff erence. Four result sets were derived, i.e measurements by scanning with K = 10 dB, measurements at discrete points with K = 10 dB, measurements by scanning with K = 7 dB and measurements at discrete points with K = 7 dB. The mean value of each set was calculated and is shown in Figure 5. As it can be seen, the sound power level using a 12 mm spacer can be within -0.3 dB and 0.6 dB from the corresponding sound power levels determined with the 50 mm spacer down

qp / ® ITV p — La) / dB L (

to 80 Hz. This may justify the use of a 12 mm spacer as long as the dynamic capability of the probe qualifies the corresponding criteria.

Figure 5: Sound power level di ff erence between results after measurements with a 12 mm and a 50 mm spacer. Continuous line: measurements by scanning with K = 10 dB. Dashed line: measurements at discrete points with K = 10 dB. Dotted line: measurements by scanning with K = 7 dB. Dash-dot line: measurements at discrete points with K = 7 dB.

Jacobsen et al. [12] proposed the use of a 12 mm spacer up to 10 kHz, by applying a correction based on the pressure response of the probe microphones at high frequencies derived using an electrostatic actuator. To avoid additional measurements, which can practically only be performed by microphone manufacturers, it is suggested that the sound intensity correction should be estimated by the available measurement results. The correction was calculated by comparing the 12 mm measurement results with a reference sound power level. The sound power level of each source under consideration was determined after sound pressure measurements in PTB’s reverberation room, according to ISO 3741 [13]. This level was used as the reference sound power level L W , ref .

The sound power level di ff erence between the measurements using sound intensity and the reference measurements were calculated according to

∆ L W (FDA) = L W − L W , ref , (11)

where the subscript denotes finite di ff erence approximation (FDA). The correction was then calculated using

C FDA = ∆ L W (FDA) , (12)

where the overline denotes mean value for all sources per accuracy grade. The related uncertainty was estimated by

u 2 ( C FDA ) = σ 2  ∆ L W (FDA)  , (13)

where σ denotes standard deviation.

Table 1 contains the correction values and the related uncertainty, which was found to be identical for all accuracy grades included in the current ISO 9614 series [4–6]. The variations in the sources,

- -Discrete, K = 10dB --Scan, K = 7dB ---Discrete, K = 7dB —Scan, K = 10dB 5 = os 800 300 Frequency / Hz 100 gp / (100eds) M77

surrounding environments and extraneous noise resulted in the correction values. Apparently, the uncertainty values of Table 1 must be energetically summed with the current ones [4–6]. Presented values are derived from only a subset of all measurement results. Further investigations will be performed on this issue, which may give di ff erent results. The applicability of the proposed correction supports the revision of the specifications of the measuring equipment for frequencies up to 10 kHz [14] and the equipment used for the sound pressure–residual intensity index.

Table 1: Correction values and related uncertainty due to the finite di ff erence approximation at high frequencies.

Frequency / Hz C FDA / dB u ( C FDA ) / dB

6300 -2.7 1.4

8000 -4.1 1.4

10000 -6.3 1.8

3.3. Minimum number of measurement points For measurements at discrete points [4], the non-homogeneity indicator F S is used to check the adequacy of the number of measurement points. It is given by

v t

N X

1 N − 1

F S = 1

 I n , i − I n  2 , (14)

I n

i = 1

where N is the number of measurement points, I n , i the normal intensity at each point and I n the normal sound intensity averaged over the entire measurement surface. The latter is calculated by

N X

I n = 1

i = 1 I n , i . (15)

N

The current standard for measurements at discrete points, ISO 9614-1 [4], states that the number of probe positions is regarded as su ffi cient when

N > CF 2 S , (16)

where C is a factor given in Table B.2 of ISO 9614-1 [4], whose value depends on frequency and measurement accuracy.

Another part of the investigation was to define the number of measurement points based on the measurements at discrete points following the above analysis. Initially, the validity of sound power levels at discrete points, which were determined after measurements in all surrounding environments, was checked by comparison to the automated scanning measurement results using

∆ L W (points) = L W (discrete) − L W (scan) . (17)

It must be noted that for each source the automated scanning measurement was the one performed without extraneous noise. The sound power level di ff erence was checked in terms of the dynamic capability criterion using Equation 10. For a global approach, the value of the factor C was set to 8, which corresponds to survey grade measurements ( K = 7 dB) [4]. Figure 6 shows the sound power level di ff erence in relation to the number of measurement points calculated using Equation 16. To restrict the deviation from the scanning measurements, an additional qualification was applied to the sound power level di ff erences excluding those outside ∆ L W (points) ± 0 . 5 dB. Results are shown in Figure 7.

Both Figure 6 and 7 do not reveal any correlation between the number of measurement points and the deviation from the reference sound power level (after measurements by automated scanning without extraneous noise). The lack of correlation leaves the necessity of indicator F S requiring further investigation. Another point of revision would be the common calculation of the field non- uniformity indicator for both measurements at discrete points and by scanning. This is not true today, because concerning discrete point measurements, the indicator is calculated based on the number of points, while for scanning N is the number of partial surfaces, which can significantly alter the F S value. The variations in F S due to di ff erences in N are apparent in Figure 7, which includes measurements using one and three probes.

Figure 6: Sound power level di ff erence between results after measurements at discrete points and by scanning vs the theoretical minimum number of measurement points. Dynamic capability criterion applied.

4. CONCLUSIONS

The paper presents a study in order to investigate potential improvements of the sound power determination based on sound intensity measurements. Despite the advantages of the sound intensity method, the current standards [4–6] have not been revised since their publication. A number of possible revision topics have been investigated and the results have been presented. The study included both broadband and narrow band noise sources and the possible use of more than one probe. The measurements were performed in a variety of surrounding environments by measuring at discrete points or by scanning. Measurements also included the introduction of stationary and non-stationary extraneous noise.

=) Yor) So Kor) =) re re N ap / (x04) Ary 104 10° 10? 10!

Figure 7: Sound power level di ff erence within ± 0 . 5 dB between results after measurements at discrete points and by scanning vs the theoretical minimum number of measurement points. Dynamic capability criterion applied.

The first revision topic could be the consideration of electrical noise along with extraneous noise, because this would broaden the applicability of the sound intensity method to low noise sources. According to the results, the lowest limit to measure is set when the emitted sound pressure is at least 10 dB higher than the electrical noise. The next revision aspect could be the use of a single spacer for the entire frequency range. At low frequencies, measurements with two spacers showed that if the dynamic capability criterion is met, a single spacer can be used. For the extension of the frequency range to higher frequencies, a correction has been derived along with the corresponding uncertainty. For the extension to apply, revisions on the requirements of the measuring equipment [14] are complementarily necessary. The correlation of the number of measurement points with the deviation from a reference sound intensity level has not revealed conclusive findings, leaving this topic open for further investigation.

The measurements presented in this paper were the basis for two revision proposals, one for ISO 9614-1 [4] and another one for a merger of the two scanning methods ISO 9614-2 [5] and ISO 9614- 3 [6]. The revised ISO 9614 series will help machinery manufacturers and other users to better determine the sound power level of machines and other products. As a result, the reliability of noise emission data of machines will be improved, so that there is a fair competition towards quieter machines. Reliable noise data will help employers choose quieter machines to better protect their workers from noise hazards.

ACKNOWLEDGEMENTS

The authors would like to thank the Federal Institute for Occupational Safety and Health (BAuA), funder of this research project (F2450), Heinrich Bietz and Kevin Picker for performing measurements.

REFERENCES

[1] European Parliament and European Council. Directive 2000 / 14 / EC of the European Parliament and of the Council of 8 May 2000 on the approximation of the laws of the Member States relating

10?

to the noise emission in the environment by equipment for use outdoors. O ffi cial Journal of the European Communities , L 162:1–78, 2000. [2] State of New South Wales and Environment Protection Authority. The protection of the environment operations (noise control) regulation 2017 – Approved methods for testing noise emissions, 2017. ISBN: 978–1–76039–353–3. [3] ISO. Acoustics – Determination of sound power levels of noise sources – Guidelines for the use of basic standards. Standard 3740:2019, International Organization for Standardization, 2019. [4] ISO. Acoustics – Determination of sound power levels of noise sources using sound intensity – Part 1: Measurement at discrete points. Standard 9614–1:1993, International Organization for Standardization, 1993. [5] ISO. Acoustics – Determination of sound power levels of noise sources using sound intensity – Part 2: Measurement by scanning. Standard 9614–2:1996, International Organization for Standardization, 1996. [6] ISO. Acoustics – Determination of sound power levels of noise sources using sound intensity – Part 3: Precision method for measurement by scanning. Standard 9614–3:2002, International Organization for Standardization, 2002. [7] F. Jacobsen. On the uncertainty in measurement of sound power using sound intensity. Noise Control Engineering Journal , 55(1):20–28, 2007. [8] S. Brezas, F. Heisterkamp, and V. Wittstock. Practice-oriented simplification of noise emission measurement methods, subproject 2: Practical adaptation of the sound intensity measurement method. Report, Federal Institute for Occupational Health and Safety (BAuA), 2021. DOI: 10.21934 / baua:report20210914. [9] S. Brezas, C. Bethke, and V. Wittstock. A new scanning apparatus for the dissemination of the unit watt in airborne sound. In Proceedings of DAGA 2016 , page 1270–1273, Aachen, Germany, March 2016. [10] F. Fahy. Sound Intensity . E & FN Spon, London, U.K., 1995. [11] F. Jacobsen. Sound Intensity Measurements in Handbook of Noise and Vibration Control edited by Malcolm J. Crocker. John Wiley & Sons, New Jersey, U.S.A., 2007. [12] F. Jacobsen, V. Cutanda, and P. M. Juhl. A numerical and experimental investigation of the performance of sound intensity probes at high frequencies. Journal of the Acoustical Society of America , 103(2):953–961, 1998. [13] ISO. Acoustics – Determination of sound power levels and sound energy levels of noise sources using sound pressure – Precision methods for reverberation test rooms. Standard 3741:2010, International Organization for Standardization, 2010. [14] IEC. Electroacoustics – Instruments for the measurement of sound intensity – Measurements with pairs of pressure sensing microphones. Standard 61043:1993, International Electrotechnical Commission, 1993.