The dependence of transformer sound power measurement accuracy on microphone configurations in the anechoic chamber.

Michal Kozupa 1 , Filip Kaminski 2

Hitachi Energy Research Pawia 7 St., Krakow 31-154, Poland

Robert Baranski 3 , Tadeusz Wszolek 4 , Pawel Pawlik 5

AGH University of Science and Technology Department of Mechanics and Vibroacoustics A. Mickiewicza 30 Ave., Kraków 30-059, Poland

ABSTRACT The paper is aimed at investigating the number and location of microphones required to accurately calculate the sound power level of a harmonic noise source. The measurement procedure performed is based on the IEC standard for the determination of transformer sound power level. The current procedure defines numerous measurement points around the tested transformer. Reducing the number of measurement points could significantly speed up the measurement procedure when a transformer is tested during the manufacturing process. Within the research work, estimations on the number and location of measurement points are identified to accurately calculate sound power level. The approach is tested and validated using specially developed software in LabVIEW, which allows inclusion or exclusion of measurement points from the sound power level calculations. To prevent harmonic wave interference, acoustic wave reflections and background noise, the measurement procedure took place in the anechoic chamber.

1. INTRODUCTION

When defining transformer sound power level IEC 60076-10:2016 and IEEE C57.12.90-2015 Standards are used. Both standards specify very detailed the measurement conditions and necessary corrections if needed. The corrections relate to “ambient noise” and “wall sound reflections”, parameters that represent the test environment can significantly impact the accuracy of the measurement of transformer audible sound [1][2]. Correction for ambient sound equals 0.0 dB if the difference between the ambient sound pressure level and the combined transformer and ambient sound pressure level is over 10 dB. This condition can be easily obtained in an isolated anechoic chamber but also in a well-designed test room in the factory. The wall sound reflection correction K depends principally on the ratio of the transformer sound measuring surface S to the sound absorption

1 michal.kozupa@hitachienergy.com 2 filip.kaminski@hitachienergy.com 3 robert.baranski@agh.edu.pl 4 tadeusz.wszolek@agh.edu.pl 5 pawel.pawlik@agh.edu.pl

inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS ? O 2? GLASGOW

area of the test room A. Again, in the anechoic chamber where the acoustic absorption coefficient is close to 1 and the reverberation time is close to zero the correction K would tend to zero and no environmental correction is necessary. What is interesting from the measurement point of view are the near field correction and the correlated measurement distance. The second edition of the IEC standard which replaces the first edition published in 2001 includes the following technical changes with regards to the measurement condition: standard measurement distance is changed from 0.3 m to 1 m for transformers other than distribution type transformers [1]. The reason for this change is to mitigate near-field effects and strong directivity. The sound pressure variation for large units is usually in the range of 5 dB, however a few extreme variations up to 16 dB within one meter can occur. This means that microphone location in the point-by-point procedure can also give significant variations. The near field effect covers the active and reactive acoustic field, however when applying the sound intensity method and increasing the measurement distance the impact of near field and strong source directivity can be also reduced. Sound field effects close to the test object (near field effects) can impact sound pressure measurements at a measuring distance of 0.3 m. They tend to increase the measured sound pressure level in the range of 0.5 dB to 1.5 dB [3]. It is solved in power transformer measurements by the walk-around method. When measuring distribution transformers safety conditions do not allow to be present in the nearby of the transformer therefore point-by-point measurement has to be performed. Thus, the number and location of the measurement points are critical to defining properly the sound power value. IEEE standard says, that no fewer than four microphone locations shall be used with intervals of 1 m (or less may result). In IEC standard the microphones should be equally spaced and not more than 1 m apart with a minimum of eight microphone positions along the measurement contour [1][2]. In this paper acoustic measurements results on oil filled distribution transformer in anechoic chamber are presented. The postprocessing and analysis of the acoustic measurement results focus on possible configurations when one to ten microphones are turned off and its influence on calculated averaged sound pressure level and respective sound power level values. Results of statistical analysis is presented on boxplot graph.

2. MEASUREMENT SYSTEM

The measurements to determine the acoustic power of the transformer were conducted in an anechoic chamber of AGH University of Science and Technology (Department of Mechanics and Vibroacoustics). In order to ensure the best possible acoustic conditions, all walls, ceiling and floor should be characterized by high absorption coefficient. Therefore, a steel rope net is installed in the chamber, on which one can move in order to place e.g., the tested object. The measurement procedure is point by point and 14 microphones are used to measure simultaneously in 0.3 m distance from the transformer. The prescribed contour is 6.2 m thus the distance between measurement points in 14 microphones configuration is 0.4 m.

Figure 1: Microphone position configuration

The measurement system is mainly based on the use of fourteen ½-inch condenser microphones, two PXI-4472 measurement cards, and the NI PXIe-8133 type controller located in the NI PXIe-1062Q chassis. The entire system is managed by software written in the LabVIEW 2019 environment. The 14-microphone point-by-point measurement array was achieved by using eight GRAS 46AE microphones (free-field, 50mV/Pa, 17 dB(A) to 138 dB) and six PCB 378B02 microphones (free- field, 50mV/Pa, 15.5 dB(A) to 137 dB). PXI-4472 cards were used as PXI-4472 cards were used as dynamic signal acquisition modules converting analog signals to digital (A/D). This card has 24-bit A/D converters with the ability to record a signal with sampling rates up to 102.4 kHz and ability to power the connected microphones with a constant-current source (so-called IEPE signal conditioning), thus avoiding the need for additional conditioning circuits. A single PXI-4472 card is equipped with eight A/D converters working simultaneously. Therefore, it was necessary to use two identical measurement cards. In addition, the measurement management software included a module responsible for synchronizing the triggering of measurements between the cards using the PXI Trigger Bus .

Figure 2: Measurement system configuration and signal flow

The figure above schematically shows the signal flow between the circuit components used starting from the transformer, through the microphones, measurement cards and computer. With the hardware configuration used, parallel measurements of all fourteen channels were possible. Due to the recorded acoustic phenomena occurring in the transformer in the frequency range up to 2 kHz, it was decided to sample at 10 kHz for each channel. Transformer was energized with nominal condition that is 420V and 50 Hz.

3. CALCULATION PROCEDURE

The test environment was developed in LabVIEW 2020. The software is responsible for performing calculations of the sound power level for given microphone configurations, hence for all possible combinations from one transducer being inactive up to ten (Equation 1).

𝑁 !

𝐶൫ 𝑁

𝑀 !( 𝑁−𝑀 )! , (1)

𝑀 ൯ =

where: - C – the number of combinations - N – the number of all microphones - M – the number of inactive microphones

To achieve this all of the configurations were generated automatically and then pass through the computing module which outputs the A-weighted sound power level for each case. Then the data were passed to the plotting module which prepares the data to be plotted on a boxplot graph. The plot

industrial computer with LabVIEW ax Gees GRAS 46AE 1/2" c= PCB 378802 1/2" 14x NI PXle-1062Q

is merged with the table containing statistic values: median (Q2), max (Q4), and min (Q0). In addition to min and max, information about what microphones were inactive is attached (opposite set to which of them were used to calculate LWA). The outliers were here rejected as they are irrelevant (they made up 0.7% of all cases). Due to a lack of build-in tools providing a boxplot graph, as a starting point, the available code from the NI Community forum was used. Then it was tested and adapted to fit the needs of the project mostly by altering the functionality.

4. RESULTS

Data were calculated for 0.3 m distance from the transformer. Results are presented as relative values to clearly show the deviations. Zero on the y-axis corresponds to the reference value which is calculated for all the microphones being active. Also, to clear up the graph it was decided not to include outliers as was said before. On the other hand, for min and max values outliers were included in the table below each plot.

wa x=03mlh=1/2 LWA Comparison REF = 49.36 20- 15 == 10- —— 05- > 09- |-~

A San a a -1.0- = “s- -2.0- M 3 4 5 7 10

Figure 3. Results of the A-weighted sound power level for the tested transformer for all considered cases at 1/2 of its height (h) and distance from radiating surface (x) equal to 0.3m. The REF value in the upper right corner along with the red dotted line stands for the reference value computed from all the microphones being active (M=0).

Analyzing situation where one microphone position is excluded from the sound power level calculations (M = 1) the median for all calculated combination changes only by 0.03 dB. For all other combinations, the median value is changed even less, therefore it can be said that the influence is negligible. Max values deviation including outliers no higher than 0.5 dB occurs for M=2, while obtaining values not exceeding 1.0 dB occurs for M=4, the 1.5 dB deviation appears when M=7. Measurement uncertainty evaluation is covered by obtaining Type B considering measurement setup. For Type B expanded uncertainty (p = 0.95) the assumed measurement setup accuracy equals ± 1[dB].

Mic Nums_ Mic Nums_ Mic Nums- Mic Nums Mic Nums_ Mic Nums_ Mic Nums- Mic Nums Mic Nums Mic Nums_ 6 56 25-6 256-13 256-713 ||| 25.6-7-9-13 || [2-5-6-7-9-12 ||| 1-2-5.6-7.9. ||| 1.25-6.7.9. ||(12.35.6.7-9. Max 1B 123 __|||_ 10-12-13 10-12-13 _| WA WA WA WA (WA WA LWA (WA (WA (WA 0.16 031 0.46 0.63 08 1 122 139 159 1.83 ‘Mic Nums Mic Nums Mic Nums ‘Mic Nums Mic Nums Mic Nums Mic Nums Mic Nums Mic Nums Mic Nums | 8 48 4811 48114 ||{ 348-1114 |/[3-48-10-11 || (1348-10-11. 134789. | Min 14 14 L10.11-12-14 | WA WA WA LWA WA WA WA (WA 0.25 O52 0.83 “1.02 123 145 an 214 c “4 a 364 1001 2002 3003 3432 3003 2002 1001

Table 1. Type B Expanded uncertainty results

M 1 2 3 4 5 6 7 8 9 10 U(L WA ) [dB] 0.28 0.29 0.30 0.32 0.33 0.35 0.38 0.41 0.45 0.50

Uncertainty is decreasing when number of measurement points is getting higher.

4. CONCLUSIONS

As discussed in the introduction sound field effects close to the test object (near field effects) can impact sound pressure measurements and as a result calculated sound power level in the range of 0.5 dB to 1.5 dB. The choice of the microphone positions can drastically influence calculated LWA because of the possibility of locally exiting extrema in sound level around the transformer, especially in the near field. Taking all previous into account and the settled approach it can be deducted that reducing the number of the measurement points up to 5 can be obtained in the anechoic chamber. Further research will cover analysis on more measurement distances as well as more measurement heights to show the overall propagation of sound from the transformer.

6. REFERENCES

1. IEC 60076-10, Edition 2.0 2016-03, Power transformers – Part 10: Determination of sound levels. 2. IEEE Std C57.12.90™-2015, IEEE Standard Test Code for Liquid-Immersed Distribution, Power, and Regulating Transformers. 3. IEC 60076-10-1, Edition 2.0 2016-03, Power transformers – Part 10-1: Determination of sound levels – Application guide.