June/July 2021

PCIM EUROPE 2021 17 www.power-mag.com Issue 3 2021 Power Electronics Europe (Vi n ) on the input DC-Bus while the current-source regulates the output current (I out ) flowing in the converter. In addition, a filter inductor (Lio) of 230 µH and an input inductor (L in ) of 1 mH are added to avoid the circulation of high frequency currents in the power supplies. Otherwise, a diode is used at the input to protect the voltage generator from any reverse currents. Thus, the two power supplies provide only the losses of the converter. The current- source imposes the output current (I out ) with a very low voltage (V opp ) corresponding to the voltage-drop of the converter ( V). The test bench allows the converter losses to be measured both electrically (measurement of the total input power) and thermally (measurement of the coolant). As shown in Figure 2, three water-cooling circuits are used in parallel (at the rectifier, transformer and inverter) to estimate the losses by calorimetry. The experimental results consider a fixed switching frequency of 15 kHz and 50 % inverter duty cycle. Experimental results Firstly, it is important to analyze the correlation between the thermal and electrical measurement methods, shown in Figure 3 (a) and (b), respectively, for 750 A and 375 A SiC-MOSFETs. In both cases, the efficiency results have an excellent correlation and the average deviation are less than 0.03 % among them. At this point, using the extra measurements from the thermal method, it is possible to separate the losses among the different converter parts (inverter, transformer and rectifier) with a proven accuracy. For the 750 A SiC-MOSFETs case, as can be depicted from Figure 3 (a), a maximum efficiency of 98.87 % is reached at the output current of 130 A. Beyond this value, the efficiency is kept practically constant till to the nominal current (Iout = 170 A). Additionally, at low output current range (less than 50 A), the efficiency curve presents a strong non-linear behavior, decreasing to 97.1 % at 22 A. In other hand, for the 375 A SiC-MOSFETs case, as can be depicted from Figure 3 (b), the maximum efficiency of 99.22 % is reached at the output current of 60 A. Beyond this value, the efficiency slightly starts to decrease, reaching 98.87 % at the nominal current. Regarding the efficiency at low output current range (less than 50 A), the converter presents a more flattened characteristic reaching 98.95 % at 26 A, approximately. The switching losses are supposed to be a function of the equivalent capacitance and the magnetizing current under DCM operation. Then, it is expected fairly constant switching losses regardless the amount of output current. In opposite, the conduction losses are a quadratic function of the output current multiplied by the device’s on-resistances. As the inverter devices have both types of losses, it is possible to deduce that the 750 A SiC- MOSFETs present considerably higher switching losses than conduction losses, due to the fairly flat linear behavior of the measured inverter losses (Figure 3(a)). Meanwhile, the 375 A SiC-MOSFETs have considerably lower switching losses than conduction losses due to the clear quadratic behavior of the measured inverter losses – Figure 3(b). Therefore, as the switching losses are supposed to be fairly constant, they are responsible for the strong non-linear decreasing of the efficiency seen for the 750 A SiC-MOSFETs case, due to their higher percentage impact at the low current range. Conclusions For the experimental tests, a water cooled 300 kW insulated R-SAB prototype rated to 1.8 kV and 170 A has been implemented using the opposition Figure 2: Schematic diagram of the circuit used for the experimental tests by opposition method Figure 3: Efficiency trajectory and losses measured of the converter using (a) 750 A and (b) 375 A SiC- MOSFETs