Repetitive Partial Discharge Phenomena on Electrical Motor Coil Windings Under High-Repetition Nanosecond Pulsed Voltages Driven by SiC MOSFET Inverter

This paper presents new findings on partial discharge (PD) phenomena under high-repetition nanosecond pulsed voltages, essential for the insulation design of electrical motors fed by silicon carbide (SiC) inverters with high-switching frequencies. Here, a twisted pair sample simulating the turn-to-turn insulation of a random-wound motor coil winding was subjected to high-repetition nanosecond pulsed voltages generated by a SiC inverter-based pulse generator. In this experiment, the rise time at a pulsed voltage of 1.5 kV is 23 ns, the pulse width is 500 ns, and the pulsed repetition frequency (PRF) ranges from 60 pps to 1 Mpps. In this PRF range, the partial discharge inception voltage (PDIV) was found to be almost constant. Meanwhile, the repetitive partial discharge inception voltage (RPDIV) significantly reduced when the PRF exceeded several tens of kpps. Under high PRF conditions, the high-density ions produced by the preceding PD remain in the gap space between enameled wires during the pulse-off period and contribute significantly to the initiation of subsequent PD. The resulting increase in the probability of PD occurrence reduces the RPDIV. Unlike previous reports, this study clarified the difference in the frequency characteristics between PDIV and RPDIV, providing valuable reference data for the insulation design of SiC inverter-fed motors. Based on the obtained results, we propose revising the IEC60034-18-41 standard, which verifies a PD-free system, to incorporate the effect of the pulsed voltage waveform of SiC inverters.


I. INTRODUCTION
Wide bandgap (WBG) power semiconductor devices such as silicon carbide (SiC) metal-oxide-semiconductor field-effect transistors (MOSFETs) and gallium nitride (GaN) field-effect transistors (FETs) offer higher breakdown voltage, faster switching speeds, higher switching frequencies, and higher The associate editor coordinating the review of this manuscript and approving it for publication was Guillaume Parent . temperature operation than the conventional Si insulated gate bipolar transistors (IGBTs) [1]. The switching frequency of Si IGBT inverters is typically less than 20 kHz, whereas that of SiC MOSFET inverters can reach several hundred kHz. The voltage rise time for Si IGBT inverters is several hundred ns or even longer, while that for SiC MOSFET inverters is less than several tens of ns [2]. Downsizing of cooling components due to lower-loss power semiconductor devices and downsizing of passive components due to higher frequencies enable smaller and lighter inverters for motors. These features are suitable for electric vehicles (EVs) and electric aircraft that require inverter-fed motors with high efficiency and high power density [3].
However, inverter-fed motors are associated with technical issues, such as overvoltage at the motor terminals, bearing currents, electromagnetic interference, and iron losses [4]. Voltage reflection due to surge impedance mismatch between the inverter, the cable, and the motor causes overvoltage at the motor terminals. When the overvoltage stress at the motor terminals exceeds a certain level, partial discharges (PDs) may occur in the motor coils, leading to premature failure due to electrical insulation breakdown [5]. The cable length in EVs is about 3 m at most, so voltage reflection has limited effects, but surge voltages occur due to circuit resonance [6]. Note that surge voltages are unevenly distributed across the motor coil windings and tend to be concentrated in the first coil [7]. Thus, if the enameled wires at the beginning and end of the first coil are in contact, the risk of PD in the air gap between the enameled wires must be considered. The insulation between the enameled wires is called turn-to-turn insulation. Moreover, since the shared voltage of the first coil is more significant with a shorter rise time of the surge voltage, higher voltage stresses are expected to occur under the fast switching of WBG inverters. In comparison, recent EVs use form-wound motor coils with hairpin technology, which has a different structure than random-wound motor coils. However, the risk of PD occurrence must be fully verified since inverter surges generate high voltages between the coils [8].
The issue of inverter surge has already become apparent in the past 30 years [9]. In standard low-voltage motors (typically 700 V or less), enameled wires are adopted for the motor coils. However, PDs easily damage the enamel coating of organic insulating materials, such as polyamide-imide. Thus, it is essential to design low-voltage motors that use enameled wires as motor coils to prevent the occurrence of PDs during the entire service life. This PD-free electrical insulation system is called Type I. The International Electrotechnical Commission (IEC) has issued IEC60034-18-41, which specifies an insulation qualification method for inverter-fed motors employing Type I insulation systems [10]. This standard refers to IEC/TS 61934 [11] as the PD measurement method under repetitive voltage impulses with short rise time to verify PD-free insulation at the specified test voltage. Under conventional AC voltage, partial discharge inception voltage (PDIV) is defined by the peak voltage value when PD occurs for the first time. PDIV fluctuations are more significant under a fast-rising impulse voltage than under conventional AC voltage application because of complex PD phenomena, such as the delay in PD generation. Thus, repetitive partial discharge inception voltage (RPDIV) is introduced as the peak voltage at which the probability of PD occurrence under the same peak value of the repetitive impulse voltages exceeds 50% [11]. However, it cannot be said that RPDIV is being applied based on a sufficient understanding of repetitive PD phenomena.
The current version of IEC60034-18-41 assumes Si IGBT inverters and does not specify the effects of high-switching frequency in WBG inverters on PDIV and RPDIV. Sufficient academic experimental data is needed to incorporate the effects of WBG inverter voltage waveforms into the IEC standards. Several results using SiC inverters have been reported in recent years [12], [13], [14], [15], [16], [17]. One study [12] has reported RPDIV measurements under SiC inverter voltages with a rise time of 14 ns and a maximum repetition frequency of 200 kHz using twisted pair samples to simulate turn-to-turn insulation in random wound motor coils. The paper also stated that the inverter switching noise overlaps the PD signal, making it challenging to measure PDs with high sensitivity in the case of the SiC inverter with fastrise time. In such an environment, it is reasonable to assume the existence of PD below the detection sensitivity even at voltages lower than the PDIV obtained in the measurement. Therefore, it is challenging to evaluate RPDIV according to its original definition with the probability of PD occurrence. Moreover, RPDIV may be affected by overshoots and subsequent high-frequency damped oscillations (i.e., ringing waveform) in the pulsed voltage applied to the twisted pair sample. Therefore, RPDIV should be measured in a pulsed voltage environment where overshoot and ringing are suppressed to the maximum extent possible.
This paper reports the results of an investigation into the effect of pulse voltage repetition frequency on PDIV and RPDIV in twisted pairs using a new SiC inverter-based pulse generator for electrical insulation tests. In this study, the experiments performed under pulsed voltages with minimum overshoot and ringing provide a better understanding of the effect of pulsed voltage repetition frequency on PDIV and RPDIV. Moreover, a quantitative measure of the intermittency of PD occurrence under a pulse train of the voltages is introduced, and the effect of the intermittency of PD occurrence on the determination of RPDIV is discussed. The effect of the repetition frequency on RPDIV is also examined in terms of memory effects, where residual particles generated by the preceding PD affect the subsequent PD. Based on the findings of this study, recommendations will be made for revisions to IEC60034-18-41 to incorporate the effects of SiC inverter voltage waveforms.

A. EXPERIMENTAL SETUP
In this study, the polyester-imide overcoated with polyamideimide enameled round copper wire used for the twisted pair samples has a conductor diameter of 1 mm and an enamel coating thickness of 0.033 mm. The SiC inverter-based pulse generator is a prototype by NexFi Technology, Inc. developed for electrical insulation tests. Owing to superior device characteristics and reduced parasitic inductance, the switching VOLUME 11, 2023 waveforms exhibit fast-rise times of 48 ns under 5 kV and 6.2 ns under 400 V with a low-voltage overshoot and ringing. Here, the output voltage waveform is positive unipolar. In addition, the pulse generator can vary the pulse width from 150 ns to DC and increase the pulse repetition frequency (PRF) up to 1 MHz by changing the control signals of the SiC MOSFETs. The pulse generator is described in detail in a previous study [18]. In this experiment, the pulse width of the voltage was set to 500 ns, and the PRF was varied from 60 pulses per second (pps) to 1 Mpps. Figure 1(a) presents a schematic of the experimental setup. So far, several studies have examined the effect of ambient humidity on PDIV in twisted pairs [19], [20], [21], [22]. A high-humidity environment strongly reduces PDIV under impulse voltages. In other words, PDIV decreases with the increase in the initial electron supply from the negative ions related to clusters of water molecules in the humid air [22]. Thus, in this experiment, the twisted pair sample was placed in a constant temperature and humidity chamber (ETAC, SXN401-E), and the ambient temperature and humidity were controlled at 25 • C and 50%, respectively. When the pulsed voltage is supplied from the pulse generator via a cable to the twisted pair sample in the chamber having constant temperature and humidity, the overshoot at the rise of the pulsed voltage becomes large. Thus, the twisted pair sample was directly connected to the output terminal of the pulse generator without a cable, as shown in Fig. 1 The pulsed voltage applied to the twisted pair sample was measured using a high-voltage probe (Tektronix, P6015A). A photomultiplier tube (PMT, Hamamatsu Photonics, R5070A) having a measuring range of wavelength from 300 nm to 900 nm was used to observe PD emission. The anode-to-cathode supply voltage for the PMT was −900 V. The PMT output current was converted to a voltage signal using a 50 -resistor. The PMT signal was recorded using a digital oscilloscope (Tektronix, MSO54), which has a maximum sampling rate of 6.25 GS/s and a bandwidth of 500 MHz and is installed outside the temperature and humidity chamber.
The PMT was placed in a grounded metal case to suppress the PMT signal disturbance due to switching noise. In addition, the BNC coaxial cable to the digital oscilloscope was covered with a copper foil shield. The PDs generated at high repetition frequencies may increase the surface temperature of the twisted pair sample [23]. Therefore, the surface temperature of twisted pair samples under repetitive PD generation was measured using a radiation thermometer (KEYSIGHT, U5855A). Figure 2(a) shows the repetitive pulsed voltage application pattern for PDIV and RPDIV measurements. A pulse train comprises N voltage pulses having the same amplitude. Here, N was set to 10 or 100. After a pause of 1 second, the voltage amplitude increased by 10 V, and the pulse train was generated again. This pause time is required to change the voltage setting of the DC power supply. In the case of N = 10 (100), the RPDIV is defined by the peak voltage (V p ) at which 6 (51) or more PDs are detected out of 10 (100) pulsed voltages. Note that the number of PD occurrences is counted as one even if multiple PDs occur under one voltage pulse. As shown in Fig. 2(a), the applied voltage increases from 510 V to 2200 V and then decreases to 510 V. Thus, in the case of N = 100, the RPDIV measurement requires data acquisition of up to 34,000 pulses at a sufficiently high sampling rate. To achieve such data acquisition, this study adopted the segmented memory technique shown in Fig. 2(b). The internal memory in the oscilloscope was evenly divided by a length slightly longer than the pulse width of the voltage applied to the twisted pair sample. The control signal to the SiC MOSFETs triggers the oscilloscope, and all waveforms of the pulsed voltages and PMT signals in the 'pulse-on' period were sequentially recorded. Thus, the memory capacity of the oscilloscope was effectively utilized by omitting the data for the pause time before the 10 V increase and the 'pulse-off' period. All necessary data in this measurement were acquired at a sampling rate of 3.125 GS/s.

B. RPDIV EVALUATION METHOD
As mentioned earlier, there is variation in the presence or absence of PD inception under a short voltage pulse. Although the RPDIV was introduced based on the probability of PD occurrence, it cannot express the intermittency of PD occurrence under repetitive pulse voltages. Thus, the intermittency of PD occurrence was quantified using PD block number (N b ). Figure 2(c) shows the definition of N b schematically. Here, the PDs in the voltage falling edge are related to the back-discharges, which will be discussed in Section III. First, the PDs are depicted to occur at all repetitive pulse voltages; this is called the successive mode. Next, two examples are shown in which 5 out of 10 PDs are detected; this is called intermittent mode. Both cases have the same probability of PD occurrence at 50%, but they have different occurrence patterns. Successively occurring PDs are represented as a single PD block, shown by the blue dotted line in Fig. 2(c). These are expressed as N b = 3 and 5. Together with the PD occurrence probability, the intermittency of PD occurrence can be discussed with N b . If the PDs occur at all voltage pulses, N b = 1. For N = 100, the maximum value of N b is 50, where the PD occurrence appears alternately for each pulsed voltage application. Figure 3 shows the pulsed voltage waveforms when the PRF is 60 pps and 1 Mpps. Note that the waveform of the pulse-off period is omitted in Fig. 3(a). The overshoot rate of the pulsed  voltage is defined as the ratio of V p to the DC component of the pulsed voltage and is found to be about 1.1, which is very low compared to that reported in a previous study [12].  The variation in V p in a pulse train was within ±2% at most, indicating that the pulsed voltage was applied with high reproducibility. In addition, the pulsed voltage waveforms are almost identical even when the PRF changes significantly. Figure 4 shows the pulsed voltage and PMT signal waveforms in the case of PDIV. As shown in the figure, the PMT output waveform, which is a negative signal, is inverted. The threshold for PD detection was set at 30 mV of the PMT signal intensity. The switching noise level is less than 5 mV. Here, the PD delay time (t d ) is defined as the time between the pulsed voltage application and PD generation. Figure 5 shows the results summarizing the occurrence or absence of PD for each voltage pulse for one measurement with N = 100. The horizontal axis of the figure corresponds to the 1st to 100th shots in the pulse train, and the vertical axis corresponds to the value of V p . As shown in Fig. 5(a), at the PRF of 60 pps, the PD is detected for the first time when V p is about 1150 V. The probability of PD occurrence gradually increases with the increase in V p , but the PDs are found to be intermittent. The intermittent occurrence of PD would be due to the lack of initial electron supply. Moreover, although V p is the same, the last 10 or 20 pulses have a slightly lower probability of PD occurrence than the first 10 or 20 out of 100 pulses. This repetitive effect on a slightly longer time scale may be a surface memory effect due to the surface charge of the enameled wires, as will be discussed in Section IV. The probability of PD occurrence reaches 100% with a further increase in V p .

B. RPDIV MEASUREMENT
The result for the PRF of 1 Mpps is shown in Fig. 5(b). As shown in the figure, PDIV is approximately 1150 V, which is similar to that at the PRF of 60 pps. When V p is slightly larger than the PDIV, the probability of PD occurrence increases rapidly. The PDs occur successively in a single pulse train at the voltage where RPDIV is observed. Thus, the pattern of repeated PD occurrence varies significantly depending on the PRF. Figure 6 shows the PRF dependences of PDIV and RPDIV (described in detail in Section IV). Here, the mean values of the three measurements are plotted, and the standard deviation is shown as error bars. As shown in the figure, both PDIV and RPDIV are almost the same at N = 10 and N = 100. Moreover, PDIV is independent of PRF. RPDIV decreases significantly when the PRF exceeds several tens of kpps. From another perspective, the difference between PDIV and RPDIV becomes smaller when the PRF is above several tens of kpps. 68830 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. Figure 7 shows the frequency characteristics of PD occurrence probability and N b . Here, the horizontal axis is V p with respect to RPDIV. The results are for N = 100. As shown in Fig. 7(a), at the PRF of 60 pps, the PD occurrence probability and N b gradually increase with the increase in V p , and when V p reaches RPDIV, N b ∼ 25. Thus, this observation indicates that PD occurs intermittently. With further increase in V p , N b decreases, and the probability of PD occurrence reaches 100% (N b = 1). In addition, as shown in Figs. 7(b) and (c), the behavior of PD occurrence probability and N b is similar to that of 60 pps when the PRF is 1 kpps and 10 kpps. On the other hand, when the PRF is 100 kpps and 1 Mpps, N b ≤ 3 in the voltage range from PDIV to RPDIV, as shown in Figs. 7(d) and 7(e). This result indicates that the PDs are more likely to occur successively when the PRF is above several tens of kpps, resulting in lower values of RPDIV. Figure 8 shows the waveforms of PD optical emission and pulsed voltage when V p = RPDIV at RPF of 60 pps and 1 Mpps. As shown in the figure, PD occurs at three locations: the rising edge, the flat part, and the falling edge of the pulsed voltage waveform. In addition, with a PRF of 1 Mpps, the PDs are localized to the rising and falling edges of the pulsed voltage and occur reproducibly. As shown in Fig. 8(c), the optical emission from the PD generated at the rising edge of the pulsed voltage disappears in 20-30 ns because the internal electric field created by space charges reduces the electric field strength in the discharge space. After the termination of the pulsed voltage, the external electric field becomes zero, and the internal electric field created by the space charges causes the back-discharge. Figure 9 shows the histograms of t d obtained in a voltage pulse train with V p = RPDIV in each PRF. As shown in the figure, t d is randomly distributed for PRFs of 60 pps, 1 kpps, and 10 kpps. Meanwhile, the spread of t d distribution becomes narrower for PRFs of 100 kpps and 1 Mpps, with most PDs occurring at t d of 40-60 ns. The delay in discharge generation is expressed as the sum of the discharge formation delay time and the statistical delay time due to the initial electron supply probability [24]. Therefore, t d of 40-60 ns is probably because of the sum of the delay time for PD formation and the rise time of the pulsed voltage (20 ns). This result indicates that the PD formation delay time is about 20-40 ns.

A. DEPENDENCES OF RPDIV ON PRF
PDIV under AC voltage in twisted pairs can be explained by analyzing the electric field strength and the Paschen curve in the air gap between the enameled wires [19]. On applying the same analysis to the enameled wires in this experiment, a PDIV defined by the peak voltage value is obtained as 875 V. The PDIV on the twisted pair sample under AC voltage at a frequency of 2 kHz measured using the same PMT was about 850 V. Therefore, in this study, the analytical and experimental values of PDIV under the AC voltage are in good agreement. As shown in Section III, the PDIV under the high-repetition nanosecond pulsed voltage is about 1150 V, which is considerably higher than PDIV under the AC voltage. This difference in PDIV can be explained by the lack of initial electron supply under a pulsed voltage with a fast-rise time and a short pulse width [24]. This study discusses the decrease in RPDIV under the high PRFs as follows.
First, the effects of the temperature rise of enameled wires on RPDIV are considered. With the increase in temperature of the enameled wire due to repeated PD generation, the changes in electron emission from the enameled coating surface and a decrease in the air density may affect PD generation. However, with a PRF of 1 Mpps, the surface temperature rise of the enameled wire during the PD generation was measured to be less than 3 • C. In this experiment, the number of repetitive pulsed voltages is the same for each PRF, and the pause time between pulse trains is as long as 1 s. The heat input from the PDs to the enameled surface diffuses sufficiently during the pulse-off period, so the temperature rise can be considered to be negligible.
Next, the effects of charged particles and radicals produced by PDs on the subsequent PD are discussed. The agents of volume memory effects, such as residual charges and radicals in the gap space, have relatively short decay time constants (several µs to several hundred µs). In contrast, the surface memory effect of deposited charges on the enameled surface usually has longer decay time constants (several ms to even several hours) [25], [26], [27], [28]. As shown in Fig. 6, the decay time constant of the memory effect agents will be several tens of µs since the RPDIV decreases significantly when the PRF exceeds several tens of kpps. Therefore, the volume memory effect is the main factor determining the decrease in RPDIV at high PRFs.
The discharge initiation mechanism for plane-to-plane dielectric barrier discharges (DBDs) with a gap length of 0.85 mm under nanosecond pulse voltages in atmospheric pressure air has been reported [29]. The repetition frequency is 10 kpps. The ions produced by the preceding discharge remain in the gap space and play an essential role in triggering secondary electron emission from the cathode at the precursor to the onset of the discharge. Studies have also reported the numerical simulations of similar DBDs that reproduce the experimental results well and discuss the following discharge phenomena [30]. The electron energy during the pulse-off period is low, and the electrons preferentially attach to molecular oxygen (O 2 ). Therefore, the remaining electrons convert to O − 2 ; the most positive ions are O + 2 due to the ion conversion between molecular nitrogen ions (N + 2 ) and O 2 . Since the rate coefficient of recombination reaction between O + 2 and O − 2 is 10 −12 m 3 s −1 , the ion density of 10 19 m −3 generated in the discharge decays to 10 16 m −3 during the pulse-off time of the 10 kpps repetition cycle. The high-density ions remaining in the gap space probably drive the discharge initiation mechanism as described above. In our experiment, high-density ions originating from the preceding PD are expected to remain in the gap space as the PRF increases above several tens of kpps. The increase in initial electron supply due to those ions is thought to decrease RPDIV by facilitating the transition to a successive mode of PD occurrence.

B. MODIFICATION OF THE IEC60034-18-41 TO CONSIDER HIGH-REPETITION PULSED VOLTAGES IN SIC INVERTERS
As mentioned in Section I, the future revised version of IEC60034-18-41 should include the effect of the voltage waveform of the SiC inverter. First, it is clear from the actual measurements and circuit calculations that the rise time of the pulse voltage is reduced to several tens of ns, and the shared voltage of the first coil increases, and it is relatively easy to incorporate this effect. Meanwhile, it is necessary to consider how the PRF dependence of RPDIV clarified in this study can be incorporated into IEC60034-18-41.
The critical point is that RPDIV, which accounts for variations in measured data, overestimates insulation performance, particularly when the PRF is low. Thus, it is desirable to apply a pulse generator with the highest possible repetition frequency in electrical insulation tests. However, it is not always possible to provide a high-power, high-repetition pulse generator that can be applied to test actual motor coil windings. Therefore, we propose the following revised definition of RPDIV so that a reasonable RPDIV can be obtained even at low PRFs. Figure 10 shows the results of PRDIV defined as the voltage value when the probability of PD occurrence is greater than P% using a pulse train with N = 100. Here, the data for P = 50% corresponds to RPDIV shown in Fig. 6, and that for P = 1% corresponds to PDIV. As shown, at P = 5%, the difference in RPDIV at each PRF is within 100 V. When P is higher than 10%, the RPDIV at the low PRFs is significantly higher than those at the high PRFs. Thus, we propose to define RPDIV at P = 5% to account for the effect of PRF on RPDIV while ensuring data variability and measurement accuracy.
Finally, the effect of high-frequency switching of the SiC inverter on the insulation performance of the motor is mentioned. First, no changes in PDIV are observed when the PRF increases to 1 Mpps. This finding indicates that the high-frequency pulsed voltage of the SiC inverter does not degrade the motor coil insulation performance. However, the RPDIV decreases significantly when the PRF is higher than several tens of kpps. This indicates that the subsequent PDs are likely to occur successively if a sudden large surge voltage enters the motor coil and PDs occur.

V. CONCLUSION
This paper presented the measurement results of PDIV and RPDIV under high-repetition nanosecond pulsed voltages generated by the SiC inverter-based pulse generator using the twisted pair sample that simulates the turn-to-turn insulation of a random-wound motor coil. Here, compared to previous reports, the overshoot and ringing waveforms at the rising edge of the pulsed voltage waveform applied to the twisted pair sample were successfully suppressed by optimizing the test system, including the pulse generator. Previous reports did not clarify whether the fundamental switching frequency or the frequency of the ringing waveform was the main factor determining the frequency characteristics of RPDIV. However, the present study clarified that RPDIV is significantly reduced when the fundamental frequency of the pulsed voltage exceeds several tens of kpps. Moreover, the PDIV was measured with high sensitivity using PD emission and was found to be almost constant in the PRF range of 60 pps to 1 Mpps. The intermittency of PD generation was quantitatively evaluated under a pulse train consisting of 100 pulse voltages. The results show that the PDs occur intermittently when the applied pulsed voltage is in the range of PDIV to RPDIV and when the PRF is less than 10 kpps. On the other hand, PDs were found to occur successively at PRFs of 100 kpps and 1 Mpps. At higher PRFs, the ions produced by the preceding PDs remain in the gap space, providing a sufficient initial electron supply for subsequent PD. Such volume memory effects cause successive generations of PDs, resulting in a smaller difference between PDIV and RPDIV. This study clarified that the RPDIV decreases at conditions above several tens of kpps, i.e., the switching frequency range of SiC inverters.
Based on the results obtained in this study, a revision of IEC60034-18-41 considering the voltage waveform of SiC inverters is discussed. Our results indicate that PD-free operation should be demonstrated at the high PRFs where the difference between PDIV and RPDIV is small. By assuming a low PRF pulse generator, we proposed that defining RPDIV as the voltage value at which PDs occur more than 5 times out of 100 pulsed voltages would result in a value close to the RPDIV with the high PRF.
The obtained findings are beneficial and serve as timely reference data for ensuring the insulation reliability of SiC inverter-fed motors. These results can also be applied in the evaluation of the insulation performance of high-power capacitive coupling wireless power transfer systems driven at high frequencies around 1 MHz. TAKAFUMI OKUDA (Member, IEEE) received the Ph.D. degree from Kyoto University, in 2015, with a focus on the characterization of SiC crystals and the fabrication of SiC power devices. He joined the Department of Electrical Engineering, Kyoto University, where he focuses on the study of power electronic circuits utilizing SiC power devices. In 2020, he joined Osaka University, where he focuses on ultra-high-voltage circuits using SiC power devices. Currently, he holds the position of specially-appointed Associate Professor with Osaka University. He is also actively engaged with NexFi Technology Inc. He is a member of the Japan Society of Applied Physics (JSAP) and the Institute of Electrical Engineers of Japan (IEEJ).