After another 3 blown mosfets, and a lot of bench work over the last 2 days, I THINK I finally have my inverter remote on/off issue figured out. Previously, I manually switch on the 120VDC supply in front of a two stage, massive LC filter. The result is the h-bridges see a slowly rising input voltage. Ditto on shut down. It worked well and could even startup with load connected without a problem.
Then I tried to add a feature for remote on/off, using the existing H-bridges as the solid state switch, since they only take 5ma of 12V when idle. The processor gets powered off to save power. When I start the inverter via powering up the processor, it is starting with full voltage, and always on the positive going half wave. On shut down, it detects the shut down request after the positive going half wave and stops. This leaves residual magnetism in the 1000 watt toroidal transformer core, so that on the next start, the current surge is great enough to trip the 15 amp breaker per H-bridge ...which causes an input spike (my very slow switched design is sensitive to that) which pops those mosfets.
Two days of work to figure that one out, and a whole lot of test setups. I was just SURE the software guy screwed up...but exhaustive testing via single shot logic analyzer captures vindicated software. Removing the transformers and driving light bulbs, the processor turn on/off works flawlessly.
Now I have to think about just sticking with the slowly ramping up/down of the voltage via a solid state relay before the big dual L-C power filter, changing my on/off remote control board significantly. Plan B: add no new hardware by adding a bunch of tricky software to do soft "starting" of the transformer by a special ramp up of alternating polarity pulse widths. Damn hardware guys always expect software to solve everything.
I'll have to sleep on it to decide which way to go.