glob-top epoxy

Does that work as a heatsink? Or is it mainly to prevent contact with huma ns and other objects?

At my last job I worked for a while on a converter using quite a few EPC Ga N FETs; the half-bridge ones as well as a couple of the 1mm guys. Our tec hs got pretty good at re-working them, and I did a few as well. You gotta be able to put tiny amounts of solder paste... I'm guessing it's a waste o f your time though.

Mechanical protection. That baby board will solder to the bottom of a bigger board, and the gan fets will be the things that poke out. They are really fragile, both the balls and the thing itself.

The main board is a big, 4-layer thing. We wanted 2 oz copper around the gan fets for cooling, but prefer to not make the main board 2 oz, and expecially didn't want to rework gan fets on a complex, dense

We might use this little module in other places.

Got it. I'm curious what your effective theta_JA will end up being with that setup; will be a good data point to have. The board looks like it's around a D2PAK in footprint size.

I guess I should measure theta somehow. The baby board is 32 mils think, 2 oz copper, lots of vias. It will be soldered to the bottom of my main board, which is bolted to a cold plate. I have gap-pads between the baby board and the cold plate. As you might surmise, I'm pushing these tiny parts on power dissipation.

Thermal putty would be better than the gap-pads, by about 2:1, but it would be messy.

I might drill a small hole in the cold plate and thermal image the GaN. I did that on an earlier version.

Stupid and crazy idea that might not be acceptable: Mount the GaN Fet on a OFHC copper plate, biggest desirable size and as thick as practical. "What about insulation for leads"? Answer: use solder mask or silkscreen, or an epoxy coating.

Call the idea an inside-out PCB...

You can get solid-copper PCBs arranged in whatever layers you like. Don't ask about price...

Tim