2

I had recently designed an array of GaN HEMTs to be used as a solid state relay. At low voltages (Vin = 24V) everything was working fine. However, when I bumped up Vin to 70V, after a couple cycles turning the SSR on and off to a 10A CC load, the top 4 HEMTs (Q1-4) broke down and went up in spectacular flames. The board was sized thermally for a much higher current than 10A, and at the time of breakdown the board temperature was only about 35 degrees C.

If it's relevant, all voltages labelled in red are referenced to GND. That means when on, GNDISO as labelled would be at a potential of 70V higher than GND. Any idea what could have occurred? The HEMTs are 100V rated devices.

Thanks in advance.

EDIT to add info: The GaN HEMTs used were these: https://gansystems.com/gan-transistors/gs61008p/

The gate driver datasheet is here: https://www.skyworksinc.com/-/media/SkyWorks/SL/documents/public/data-sheets/Si827x.pdf

The driver is a fully isolated driver and this gate drive circuit is recommended in an app note from GaN Systems. I didn't notice what happened to the bottom 4 transistors as I rushed to shut everything off and put out the flames. However it is clear that Q1-4 were the ones that blew.

enter image description here

Rahul
  • 23
  • 4
  • Hi! Can you please add links to the HEMT and driver datahseets? Is the driver also isolated? otherwise, its two GND pins bridge the isolation... What happened to the bottom 4 transistors when damage occured ? Were they on/off ? Please edit your question to add this info – tobalt Nov 24 '22 at 06:20
  • Have edited to include the above info, but just to repeat: the driver is indeed isolated, and I unfortunately didn't notice what happened to the bottom 4 transistors. – Rahul Nov 24 '22 at 07:32
  • Well hard to say. Next time try to scope some turn on/off transitions. I think GaN HEMT can be moody if turn-on is too slow.. And as you are driving 8 gates in parallel.. Maybe that happened – tobalt Nov 24 '22 at 18:25
  • Just out of curiosity, why GaN? What application needs such a speed and/or low capacitance? – carloc Nov 25 '22 at 06:51
  • Used GaN here for the size mostly, as well as just a general testbed to learn about these devices – Rahul Nov 25 '22 at 07:29

1 Answers1

2

Hard to tell without grabbing some waveforms, but my wild guess is MOSFET damaged by switch-off overvoltage.

Nothing is given on the load characteristics and/or wiring setup, but even with purely resistive load wiring stray inductance and very fast GaNFET behaviour can easily rise hundreds volt kickback.

On the other hand GaNFET are absolutely not avalanche rated and hence possibly destroyed even by the -relatively- small energy in the stray inductance.

Si MOSFETs are quite different on this side, they are usually capable of taking a considerable avalanche energy.

So some kind of overvoltage protection should be tried out, a couple of Transils could be quickly wired across the two GaNFET arrays. Clamping should be made to assure less than BVds at maximum load current.

Make sure clamp devices terminals are the shortest and directly across GaNFET D/S. Any stray inductance in this connection is going to severely spoil their action.

If this test actually proves my guess right, a better tuned snubber or clamp can be designed.

carloc
  • 3,174
  • 1
  • 16
  • 18
  • Thank you! I think this is indeed the problem. I did some tests on a second board, repeatedly turning on and off the devices with increasing Vds. At 48V, I did hear a tiny little arcing sound and the devices failed (couldn't turn them off after this). – Rahul Nov 25 '22 at 07:30