[MPaulHolmes]

One was nice and smooth for a little while and blew up earlier, but it was at around 163v, so that could have contributed to it. Another had all bars out the same side, and is behaving erratically. I think it was a design problem for the optocoupler to be on the control board, and the output to travel to the driver board. It is a signal with a tiny tiny current, and is more susceptible to noise. Even if it works perfectly well for flores, I've already made some changes and just need to get the new boards made. The new way has a much shorter path from caps to diodes to mosfets back to caps. Maybe 3 times shorter than before, and takes advantage of the backs of the mosfets and diodes for carrying current. The only time the current goes through the legs of the diodes is when it goes through 2 legs at a time. Each mosfet and diode has its own milled out little island on a 1/4" x 1" bus bar. I'm going to heat the mosfets up to around 140 degC in one oven, and the bars just to soldering temp in another oven, and then take out the mosfets and place them in their little islands, raise the whole thing back up to soldering temperature (around 180 degC) and then let it cool. I've done some tests and it works pretty well. The bus bars will squashed against a large heat sink, with the electrical isolation pad between the 2. This will mean much more heat can leave the mosfets/diodes, since there will be no more pad directly between them and the heat spreader.

It's going to use 12 gigamos mosfets. It has a driver board that's 4 homemade isolated supplies. This is based on Fran's recommendations. Each isolated supply drives 3 mosfets. There's also an RC snubber for each mosfet on the driver board. It uses ceramic 2W resistors for snubbers since their inductance is so low.

Edit: The 500amp version is so radically different, that I don't know a good way to upgrade it. I'm not sure when we'll get rid of the $500 price. Maybe in September.