February/March 2021

INDUSTRY NEWS 17 www.power-mag.com Issue 1 2021 Power Electronics Europe within the die leads to a rapid rise in junction temperature and the formation of strong thermal gradients. For sufficiently high power or pulse duration, the device simply overheats and fails catastrophically. This is known as thermal overload failure. In Si MOSFETs, another failure mechanism known as secondary breakdown (or Spirito effect) has been observed in SOA testing. This failure mode, which occurs at high VD and low ID, is caused by an unstable feedback between junction temperature and threshold VTH. As the junction temperature rises during a pulse, VTH drops, which can cause pulse current to rise. The rising current, in turn, causes temperature to rise faster, thereby completing a positive feedback loop that leads to thermal runaway and ultimate failure. A goal of this study is to determine if the Spirito effect exists in eGaN FETs. EPC designed and built a custom SOA test system which works similar to a curve tracer. The gate bias on the device under test (DUT) is set before the pulse and is used to modulate the ultimate pulse current. The drain voltage is then pulsed onto the drain by means of a p- channel control FET for a specified pulse duration. For DC, or long-duration pulses, the SOA capability of the FET is highly dependent on the heatsinking of the device. This can present a huge technical challenge to assess the true SOA capability, often requiring specialty water-cooled heatsinks. However, for short pulses (< 1 ms), the heatsinking does not impact SOA performance. This is because on short timescales, the heat generated in the junction does not have sufficient time to diffuse to any external heatsink. Instead, all of the electrical power is converted to raising the temperature (thermal capacitance) of the GaN film and nearby Silicon substrate. As a result of these considerations, SOA tests were conducted at two pulse durations - 1 ms and 100 s. Figure 6 shows the SOA data of 200 V EPC2034C. In this plot, individual pulse tests are represented by points in (I D , V DS ) space. These points are overlaid on the datasheet SOA graph. Data for both 100 s and 1 ms pulses data are shown together. Green dots correspond to 100 ?s pulses in which a part passed, whereas red dots indicate where a part failed. A broad area of the SOA was interrogated without any failures (all green dots), ranging from low V DS all the way to V DS(max) (200 V). All failures (red dots) occurred outside the SOA, indicated by the green line in the datasheet graph. The same applies to 1 ms pulse data (purple and red triangles); all failures occurred outside of the datasheet SOA. Literature EPC Reliability Report - Phase 12: Alejandro Pozo, Shengke Zhang, Gordon Stecklein, Ricardo Garcia, John Glaser, Zhikai Tang and Robert Strittmatter, Efficient Power Conversion Corporation www.epc-co.com Figure 4: EPC2212 MTTF vs. V GS at 25°C MTTF (and error bars) are shown for four different voltage legs. The solid line corresponds to the impact ionization lifetime model. Extrapolations of time to failure for 100 ppm, 10 ppm, and 1 ppm are shown as well Figure 5: Measured MTTF for EPC2212 (25°C) measured at four different gate biases. Blue line is lifetime model. Red and green lines are predictions of the lifetime model at 125°C and ?75°C respectively Figure 6: EPC2034C SOA plot. The “Limited by RDS(on)” line is based on data sheet maximum specification for R DS(on) at 150°C. Measurements for 1 ms (purple triangles) and 100 s (green dots) pulses are shown together. Failures are denoted by red triangles (1 ms) or red dot (100 s). Note that all failures occur outside the data sheet SOA region