【正文】 The current was measured using a current transformer placed in one of the twentyseven return paths of the load. Fig, 14. Experimentally measured output current of the PFN. CHI2700 A/div, Time base2 / i s /div. Gate Drive Circuit: There is some evidence to suggest that increasing the gate current amplitude will help increase the initial tumon area in an SCR [2],[3]. Triggering measurements have been performed [2], [3] to determine the switching dependence of thyristors on peak gate current, gate pulse width, and gate dig/amp。C per pulse. This is much lower pared to the numerically puted temperature rise of about 1100176。s performance if the rise time of the anode current is parable to the pbase transit time. IV. TRANSIENTTHERMALANALYSIS The di/dt failure is caused by the instantaneous rise in temperature in the small conduction area obtained immediately after tumon. Transient thermal analysis was performed in Fig. 8. Typical IC (CHI), ~ (CH2) and power loss (MULT) for the shorted device. C H 1 4 0 A/div, CH210 V/div, MULT2048 W/div. Time baseI00 ns/div. order to estimate the temperature rise in the shorted and unshorted devices. A generalpurpose FEM program called FIDAP39。I is to the switching delay, suggesting that the pbase transit time is responsible. Once the main cathode region tums on, the resistance of the device decreases and the anode current begins to rise again (transition from phase I1 to phase 111). From here on the plasmaspreading velocity will dictate the rate at which the conduction area will increase. The dip in the Fig. 6. Typical IC (CHI), (CH2) and power loss (MULT) of the unshorted device. C H I 4 0 A/div, CH220 V/div, MULT4096 W/div. Time base100 ns/div. gate current ( GI )and the increase in the gatecathode voltage ( GKV ),shown in Fig. 6,corresponds in time, to the pilot anode current flow. This supports the above suggestion that the anode current is initially forced to flow through a small area (high resistance) near the pilotgate contact. Therefore the current density during phase I and phase I1 is very high and leads toa considerable increase in the local temperature. It was reported earlier in [9] that failure temperature of the device is about 1200176。C). The experimental details and results are fully presented elsewhere [3]. The experimental arrangement and the results are given in brief below. The devices were electrically characterized initially and recharacterized after testing in a type E pulseformingwork (PFN) that has a total impedance of 0. This work delivers a 15 kA, 10 ps wide pulse when charged to a voltage of kV. The di/dt of this 15kA pulse is 125000 /As? . The gate trigger used for switching the SCR39。C). The rise in average temperature is therefore pletely inade quate as a measure of device applicability for pulseswitching applications. Since a simple experimental technique is not available to measure the instantaneous temperature rise, the spatiotemporal distribution of temperatures in the devices has to be estimated using puteraided techniques. In this paper, the high dildt singleshot experimental re sults are given in brief. A qualitative physical model is then proposed to explain the experimental results, which are presented in detail elsewhere [3]. Next, the results from the thermal analysis using FEM, given in detail in [lo], are briefly presented. The particulars of the experimental arrangement for the repetitive testing of the devices, results from these experi ments, and their correlation with the numerical predictions are given in the discussion. 11. SINGLESHOT EXPERIMENTS Invertergrade SCR39。s and GTO39。s with and without the amplifying gate. High dildt. highenergy single shot experiments were first done. Devices without the amplifying gate performed much better than the devices with the amplifying gate. A physical model is presented to describe the role of the amplifying gate in the turnon process, thereby explaining the differences in the switching characteristics. The turnon area for the failure of the devices was theoretically estimated and correlated with observations. This allowed calculation of the current density required for failure. Since the failure of these devices under high dildt conditions was thermal in nature, a simulation using a finiteelement method was performed to estimate the temperature rise in the devices. The results from this simulation showed that the temperature rise was significantly higher in the devices with the amplifying gate than in the devices without the amplifying gate. From these results, the safe operating frequencies for all the devices under high dildt conditions was estimated. These estimates were confirmed by experimentally stressing the devices under high di/dt repetitive operation. I. INTRODUCTION Recent innovations in semiconductor device designs and advances in manufacturing technologies have helped evolve highpower thyristors. These devices are designed to operate in a continuous mode for applications such as ac todc power conversion and motor drives. Until recently, their application to highpower pulse switching was mostly unknown. One of the main reasons that has discouraged the use of thyristors for highspeed, highenergy switching is their low dildt rating. The limiting value of the dildt before damage occurs is related to the size of the initial turnon area and the spreading velocity [I]. Recent experimental results presented in [2][4] show that with increased gate device, SCR39。s in a stack assisted by saturable inductors, have shown the potential to repetitively switch highcurrent pulses with di/dt of about 2021 Alps, on the order of kilohertz [7]. Under high dildt