New Experimental Method for Switching Noise of Motors

Transcription

1 EVS28 KINTEX, Korea, May 3-6, 2015 New Experimental Method for Switching Noise of Motors Hyunsu Kim 1, Myunggyu Kim 2 1 Research & Development Division, Hyundai Motor Company 2 Research & Development Division, Hyundai Motor Company, mgkim78@hyundai.com Abstract Among electrical noise in EV/HEV, switching noise is one of the main issues on motor noise. In evaluating switching noise level, a traditional NVH test methodology (for example, an impact hammer or shaker test) would be used but may have limitations to excite (1) very high frequency components and (2) directly coil itself. In general, a complex inverter system, which consists of inverter hardware, battery system, and control software, is required to generate the electrical noise of motors. Here, a simple and precise method to excite the switching noise is newly suggested using a function generator and an amplifier. Using this approach, there is no need for the complex inverter system, and motor coil can be excited directly with electrical current at even very high frequency range. With the proposed test setup, frequency response function of noise level per input current level is measured and newly named as Electrical Annoyance. Lastly, switching noise on the motor is illustrated with an acoustic camera, and it is confirmed that the location of the switching noise is in the vicinity of the coil. Keywords: Motor, Switching Noise, NVH 1. Introduction Electric motors have been widely used almost every industry area for a long time. Numerous studies for motors for hardware design, control algorithm, reduction of vibrations (noise), and so many applications were developed along with the motor history. Recently, however, as green vehicle market becomes larger, noise issues of electric motors appear to be spotlighted again. The overall noise level of electric vehicle (EV) is generally much lower than that of the combustion engine vehicle. Instead, particular noise due to electrical components may still make customer uncomfortable. For interior permanent magnetic synchronous motor (IPMSM), motor noise types may be divided by two based on its sources, (1) torque ripple noise and (2) switching ripple noise. Torque ripple can be caused by magnetic forces between rotor and stator and reluctance forces. These torque ripples can be controlled by optimization of topological options with slots and magnetic stick [1] [11]. Torque ripple noise and its harmonics, therefore, is decided by number of slots and poles and synchronized with motor rpm. Another noise issue is from switching ripple in motor and inverter system (including battery). To convert DC to AC voltage, an inverter switches the current with pulse width modulation (PWM). Due to PWM, AC voltage contains a tiny (compare to the level of driving frequency component) fluctuation whose the main frequency is corresponding to the PWM frequency. The fluctuation causes a high and narrow EVS28 International Electric Vehicle Symposium and Exhibition 1

2 frequency noise, and it is in general called as switching noise. To reduce the switching noise, there are known methods of using filters [12] [14] in hardware or distributing the main switching frequency randomly in software [15] [17]. These methods, however, may bring more cost (for filters) or loosing efficiency (for random switching). Experimental setup can be required in developing motor-nvh (noise, vibration, and harshness) to assess the switching noise. However, there are difficulties in order to build an experimental setup for the following reasons: (1) A complicated inverter system is required. (2) A traditional vibration testing methods, which are impact hammer test or shaker test, can be applied. Yet, the frequency range of excitations may be limited (up to around 3 khz) while the switching frequency may be very high (5 10 khz). (3) It is expected that the source of switching noise occurs mainly in the vicinity of the coils in motor, but with the traditional NVH test methods, it is difficult to excite the coil itself. amplifier, MB dynamics SS1200 VCF (max output 20 A rms ), enhance the input signals which can be considered the switching fluctuations. To measure the amount of input current level, HIOKI 3275 current sensor is used. 1/2 inch condenser type microphone senses the acoustic pressure of the switching noise radiated from motor, and the pressure would be the output of the frequency response functions. Positive and negative output lines of the amplifier are connected to any two phase of U, V, and W of the motor. Various signals such as sine, sine sweep, and random are respectively utilized as a excitation, details about these input signals will be discussed in next chapter. Laptop Data acquisition system with function generator (Pulse Type 3160-A-042) Power supply (HIOKI 3269) Amplifier (MB dynamics model SS1200VCF) Therefore, in this study, an experimental setup is newly suggested to test motor switching noise instead of using a complicated inverter system. Through the proposed method, the frequency responses of motors are easily measured and the characteristics of different motors for the switching noise are compared. Also, the location of the generated noise in motor is visualized using an acoustic camera. In Section 2, the experimental setup is described, followed by Results and Discussion in Section 3. Concluding remarks are summarized in Section 4 lastly. 2. Experimental setup Experimental setup for switching noise test is shown in Figure 1. A data acquisition system (B&K Pulse 3160, DC to 51.2 khz), which has a function generator embedded, is used to both send various signals toward motors and receive the measured signals from motors. Frequency resolution of FFT analyzer is set to 8 Hz, and spectral line is 6400 (then, automatically time period is set to s, time resolution to μs) times of spectrum averaging is applied to reduce background noise. An Motor Microphone (B&K 4189) Figure 1: Switching noise test set up Current sensor (HIOKI 3275) Frequency response function (FRF) represents the characteristics of a linear system using input-output relationship in frequency domain. FRF defined as Acceleration/Force is mostly used to illustrate the vibration characteristics, but various FRF can be defined as shown in Table 1 as other unit expression can be more appropriate for various system natures (In acoustics, for example, impedance may be used to express the characteristics of boundary conditions). Table 1: Various FRF definitions Function Description Admittance Displacement/Force (Dynamic Compliance) Mobility Velocity/Force Accelerance (Receptance) Acceleration/Force EVS28 International Electric Vehicle Symposium and Exhibition 2

3 Dynamic Stiffness (Mechanical) Impedance Apparent (Effective) Mass Force/Displacement Force/Velocity Force/Acceleration Here, a FRF may be newly defined by using electrical current as an input and sound pressure as an output. It might be useful to express how much the system can be excited acoustically per unit electrical input current. Since louder acoustic pressure per unit input current can represent how much people can be annoyed, authors would like to call it as Electrical Annoyance (EA). In decibel (db) scale, EA level can be expressed as EA (db) = 20log10 Sound Pressure (ref. 20μPa) Current (ref. 1A) khz, and the peak-to-peak value is about ± 5 Ampere. In electric vehicles, the measured current magnitude generally ranges about ± 10 ~ 20 A (peak-to-peak). For the proposed measurement setup, the amplifier used has limitation in maximum output current (max 20 Arms). The magnitude of the current level (± 5 A, peak to peak) may not provide the same maximum input level, but it could be high enough to demonstrate the switching noise. Further study to see the effect of high magnitude of input current can be performed simply using a high-performance amplifier. (1) For sound pressures level (SPL), reference pressure is 20 μpa based on the minimum pressure level with human ear sensitivity. For the input current level, 1Arms is used for the reference, which means for example 0 db for 1Arms input and -20 db for 0.1 Arms. Note that in this paper without indicating values as peak-to-peak, all units are root-mean-square (rms) values. Input signals such as Sine, Sine Sweep, Random are respectively used for the excitations and Electric Annoyance (EA) is calculated for each input. In general, Sine signal has sufficient signal-to-noise ratio (SNR), so it is often used in the situation that very high magnitude level of the signal is needed to overcome high background noise. Also, it can be utilized to verify if the system has any nonlinearity as a function of input level. Despite of these advantages, Sine signals might need long testing time, if broad band frequency range is interested. Sine Sweep may resolve the time-consuming problem, but Sine Sweep may not be suitable for nonlinear systems. Random signal may be little bit poor in SNR. However, testing speed may be fast, and nonlinear system can be measured by so called linearization of system responses. Thus, to investigate unknown characteristics of a system, those signals are applied here to clarify those uncertainties. Figure 2: Time series of electrical current (Sine, 3 khz) Figure 3 compares input current levels in db scale for Random and Sine Sweep signals. Frequency span of the random excitation is 12.8 khz, which covers entire frequency range of interest. When using Sine Sweep signal, the sweep rate is 1 khz/s (unidirectional). Volume knob of amplifier is toggled in three steps, and each volume level is notified by v1, v3, v5 as shown in the legend of Figure 3 (v2 and v4 are measured, but skipped to show). Regardless of input signal type, the magnitude of input current is almost identical for each volume level. 3. Results and discussion Input signal (electrical current), which is measured from the current sensor, is shown in Figure 2 in time domain as an example. Frequency of the input is 3 Figure 3: Comparison of various inputs (ref. 1A) Sound Pressure Level (SPL, db) measured by EVS28 International Electric Vehicle Symposium and Exhibition 3

4 microphone is shown in Figure 4. These sounds may represent the switching noise generated from motor as a function of frequency. As input level increases, SPL increases proportionally per each input current level at the frequency range over 4 khz. At low frequency range (below 4 khz), the switching noise level is not quiet correlated with input level. The reason for the different sensitivity based on the frequency range is not clear, but higher frequency range appears to be more sensitive than lower frequency range for the switching ripple. With v1 level, the measured SPL indicates that the switching noise is not emitted since the SPL is almost identical with background noise. It may mean that, the switching noise is not activated until certain level of input is provided. Criteria point of the input level for switching noise will be demonstrated later in this paper. Figure 4: Sound pressure level for various inputs (ref. 20 μpa) Electric Annoyance (EA) for each input is illustrated in Figure 5. Larger EA represents the louder switching noise per unit input. However, since input level is located in denominator in EA definition, very small amount of input could lead unrealistic results in EA. Thus, certain level of input is required to make EA meaningful. EA appears to be converged when the input is higher than v3, which the input magnitude is around -20 db. The type of input is less relative to EA, but the magnitude is more dominant factor. switching frequency can be avoided from 6 khz. The effect of modifying the motor hardware can also be investigated through the proposed test method. Figure 5: Electric Annoyance for the various inputs Effects of very high magnitude of input current on EA are illustrated in Figure 6. Using sinusoidal input, several points of frequency are excited, and magnitudes of the inputs are around 10 db. As shown in Figure 6(a), Sine (v5) provides approximately 30 db more than Random (v5), which covers very broad dynamic ranges in magnitude. As described earlier, certain amount of input level is required for EA activation, but as shown in Figure 6(b), EA is not changed any more as input level increases after certain input level, which conclude that the system response linearly (see EA values with sine input with green color). In this study, the hardware (amplifier) has been used with input limitation with ± 5 A (Sine v5) for safety reason (amplifier failure), but further cases (higher current level) should be investigated to see if nonlinear response exists. Through this measurement method, the characteristics of motor system for switching noise can be clearly demonstrated. The motor tested appears to have the highest point in EA around 6 khz. Therefore, if a vehicle uses the switching frequency at 6 khz, the noise may be the maximum. Strategically, the EVS28 International Electric Vehicle Symposium and Exhibition (a) Comparison of input levels including Sine for nonlinear check 4

5 noise with motor 2 is lower for the switching frequency component, which can be expected from EA results. (b) EA comparison for various inputs including Sine Figure 6: Input and EA comparisons (a) Comparion of input current level Figure 7 expresses the relationship between input current level and EA magnitude (5 khz of input is chosen because it is the main switching frequency in real vehicle). It shows that until about 35 db of input level, the magnitude of EA has been changed; thus, it needs larger input level. After the criteria input level, the magnitude of EA is not changed. Here, only one frequency component is compared, but as frequency changes the plot may be different. Again, the system shows the linear response after the criteria point. (b) Comparison of EA Figure 8: Comparison of EA between two motors Figure 7: EV level as a function of input current level (@ 5 khz) Figure 8 shows experimental results using the proposed method for two different motors having similar performances but made by different manufacturers. Figure 8(a) confirms that almost the same level of input current is provided, and Figure 8(b) shows EAs for two motors. As shown in Figure 8(b), except very few frequency ranges, motor 2 shows better (lower level) EA in overall. Especially the highest peak at 6 khz with motor 1 is not observed in motor 2. For confidential reason, it will not be demonstrated in this paper, but real vehicle test, cabin In Figure 9, the location of switching noise is visualized using an acoustic camera (SeeSV-S205, SM Instrument). Switching noise is generally known to be generated near motor coil, but it is not clearly demonstrated in literatures before, as far as authors know. Here, for an example of YF HEV motor, the switching noise is mostly coming from the motor coil, more specifically, near boundary between coil and rotor. Further study is necessary to demonstrate the effect of switching ripples if hardware of the coil is changed. Those investigation can be easily performed with the proposed test set up. EVS28 International Electric Vehicle Symposium and Exhibition 5

6 measured. One motor, which generate more switching noise than another in vehicle test, has higher EA value in overall. 8. The location of switching noise is visualized using an acoustic camera after exciting the motor with the proposed experimental setup. Coil area at the vicinity of rotor is the main location for the switching noise source. Figure 9: Visualization of the switching noise on motor using an acoustic camera 4. Concluding Remarks There are several points obtained through this study and they can be summarized as: 1. The relationship between current input and switching noise level is investigated as a frequency response function, and it is named as Electric Annoyance (EA). 2. Without a complicated (and expensive) inverter system, the switching noise is precisely controlled with a simple (but accurate) experimental setup. Using a function generator, various inputs are possible to generate. 3. The major factor of switching noise is the magnitude of input current. Type of the input current (Random, Sine, Sine Sweep) does not contribute to the response. 4. Until certain level of input current, the switching noise is not generated from motor (maybe the motor coil is not vibrated with the low magnitude of input current). After the criteria input level, the switching noise increases linearly proportional to input current level. 5. With higher than -35 db of input current level, EA becomes activated, which means EA does not change as the magnitude of input increases. 6. Switching noise level as a function of input current relatively increases easily at the frequency range over 4 khz. Lower frequency range than 4 khz is less sensitive for input level. 7. EAs of two distinctive motors of different manufacture (but having similar power) are This study suggests an efficient way of testing a motor performance for the switching noise compare to existing motor testing method (with converter system). The experimental setup is simple while accurate. Therefore, the proposed method can be used (for switching noise) in motor development process and testing facility for end of line in production line. Nevertheless, further study for following aspects would be necessary: 1. The effect of hardware change in motor on the switching noise can be systematically studied. For example, winding method of coil, thickness of coil, or hardening of coil may have influences on the switching noise. 2. Nonlinearity of EA for motors should be further studied. The response of the switching noise with more than 10 A current input is not studied here due to the limitation of existing experimental setup. References [1] P. Vijayraghavan and R. Krishnan, Noise in Electric Machines: A Review, IEEE Transactions on Industry Applications, 35(1999), [2] G.H. Lee et al., Torque ripple minimization control of permanent magnet synchronous motors for EPS applications, International Journal of Automotive Technology, 12(2011), [3] T.M. Jahns and W.L. Soong, Pulsating Torque Minimization Techniques for Permanent Magnet AC Motor Drives A Review, IEEE Transactions on Industrial Electronics, 43(1996), [4] L. Parsa, L. Hao, Interior Permanent Magnet Motors with Reduced Torque Pulsation, IEEE Transactions on Industyial Electronics, 55(2008), [5] R. Islam et al., Permanent-Magnet Synchronous Motor Magnet Designs with Skewing for Torque Ripple and Cogging Torque Reduction, IEEE Transactions on Industry Application, 45(2009), EVS28 International Electric Vehicle Symposium and Exhibition 6

7 [6] D. Hanselman, Minmum Torque Ripple, Maimum Efficiency Excitation of Brushless Permanent Magnet Motors. IEEE Transaction on Industrial Electronics, 41(1994), [7] H. Beroug et al., Analysis of Torque Ripple in a BDCM, IEEE Transaction on Magnetics 38(2002), [8] L. Zhu et al., Analytical Methods for Minimizing Cogging Torque in Permanent-Magnet Machines, IEEE Transactions on Magnetics 45(2009), [9] K. Atallah et al., Torque-Ripple Minimization in Modular Permanent-magnet Brushless Machines, IEEE Transactions on Industry Applications 39(2003), [10] E. Favre et al., Permanent-Magnet Synchronous Motors: A Comprehensive Approach to Cogging Torque Suppresion, IEEE Transactions on Industry Applications, 29(1993), [11] C.A. Borghi et al., Minimizing Torque Ripple in Permanent Magnet Synchronous Motors with Polymer-Bonded Magnets, IEEE Transactions on Magnetics, 38(2002), Authors Hyunsu Kim, Ph.D. received his doctor degree at the Ohio State University in 2011 after achieving B.S. from Kookmin University in 2001 and M.S. from University of Cincinnati in His research area is acoustics of intake/exhaust system especially focusing on nonlinear characteristics in silencers in the presence of flow. Since 2011, he works at Hyundai-Kia Motor Company as a senior researcher in the power-train NVH team as currently expanding his area of interest to (hybrid) electrical vehicle motor noise. Myunggyu Kim, Ph.D. received the B.S., M.S., and Ph. D in mechanical engineering from Hanyang university, Seoul, Korea in His major field of interest includes the analysis and the design of electromechanical device and noise and vibration of high speed rotating machine. After his graduation, he is a senior research engineer in ecovehicle drive train engineering design team of Hyundai motor group. His current research is focused on the traction motor system of HEV, EV and FCEV and magnetically induced noise and vibration. [12] J.K. Steinke, Use of an LC Filter to Achieve a Motorfriendly Performance of the PWM Voltage Source Inverter, IEEE Transactions on Energy Conversion, 14(1999), [13] P.A. Dahono, An LC Filter Design Method for Singlephase PWM Inverters, Power Electronics and Drive Systems, Proceedings of 1995 International Conference on, 2(1995), [14] V. Dzhankhotov and J. Pyrhönen, Passive LC Filter Design Considerations for Motor Applications, IEEE Transactions on Industrial Electronics, 60(2013), [15] C.M. Liaw et al., Analysis, design, and implementation of a random frequency PWM inverter, IEEE Trans. On power Electron, 15(2000), [16] R.L. Kirlin et al., Analysis of power and power spectral density in PWM inverters with randomized switching frequency, IEEE Trans. On Ind. Electron, 49(2002), [17] G. Yinghua et al., Study on Soft Switching Technology to Reduce Electromagnetic Interference of PWM Inverter, Energy Procedia, 17(2012), EVS28 International Electric Vehicle Symposium and Exhibition 7