The short story: I've got a working prototype power section! It surely isn't pretty, but I tested it this evening and was charging at 22A into my 144V nom pack. It was actually charging at 155V, so that's 3.4kw!
Temperatures were looking really good:
Ambient: 35C
Mosfet: 41.5C
Diode: 44.5C
Caps: 63C
The details:
I was having problems with the mosfets getting too hot a burning up. The problems were mostly due to the electrical insulation of the mosfet to heatsink. The drain pin wasn't isolated from the back of the case; in my design, the drain changes voltages from ground to pack (-) every switching cycle and the parasitic capacitance of the heatsink caused weird voltage spikes if it's not isolated. I tried using isolation material and aluminum oxide tabs, but they only worked ok. Great electrical isolaters, but also pretty good thermal isolators. I got up to 18 amps with 2 paralleled mosfets, but they were at 80C already.
One solution would be to make this a high side configuration so that the drain stays at the same voltage (the + from the rectifier). I'd have to use some sort of charge pump high side driver instead of the simple-to-use low side driver being used now, so I didn't pursue this beyond thinking about it...
Then, I discovered isolated mosfets whose backsides are internally isolated from the pins! Unfortunately, this is the one I wanted to get and no one has it for a reasonable price:
Digi-Key - IXFB100N50P-ND (IXYS - IXFB100N50P)
I could've tried a couple of the smaller ones, but I was a little frustrated at this point and just wanted it to work. So, I ordered this one:
IXYS SEMICONDUCTOR|IXFN100N50P|MOSFET, N, SOT-227B | Newark.com
It has the same Rds_on as the ST mosfet I was initially using, but has a larger base that is isolated so it can be placed directly on the heatsink. It was a little cheaper when I ordered it, 28 I think (plus the extra 20 to ship from England).
*Note that there are lower voltage isolated FETs available, but I don't think you'd be able to use them on a 240vac line which rectifies to about 350V. These would be an option if all you were looking to use are 120vac power sources.
I also picked up one of those fancy cpu heat pumps, again mostly due to frustration. My new approach would be to get something together that worked and then optimize price/size later.
The pictures show this hacked together prototype. I used two diodes in parallel mounted on their own heatsink and again tried to make everything on the input loop close together to limit parasitic inductance and resulting voltage spikes. I also added a thermistor current limiter on the input and output so that I'd reduce the sparks when plugging in the AC and the batteries.
I think it performed really well as the temperatures were much less than I expected. This means I probably over did it with the larger mosfet and the heatsink, but maybe it'll make for a robust design. Or, maybe I can push it up to 30A... (I actually stopped at 22amps only because the output thermistor was rated for 20amps... thinkin of getting bigger ones....) I was a little surprised at the Cap temps. Maybe more smaller ones would be better? More surface area for cooling and lower ESR in parallel...
I guess costs have changed a little with these current components.
Big 400V caps: $3x 10 each
Big 30A Inductors: 2x $27 each
600V 37A diode: 2x $5
500V 60A FET: $33
600V 40A bridge: $6 (from a surplus site)
Various heatsinks: 50 + a small cpu heatsink I had.
Looks like 280 or so before any of the control hardware and things like a case and PCB. But, a comparable charger like the PFC20 is over 2k, so I think it's still good.
Speaking of, I think a PC board for this power section would be pretty easy to make...
My next steps will be to work on a PCB layout and programming/incorporating the control hardware.