Polyphase would reduce the ripple current requirements of the capacitors. For that size I would probably try it 2 phase, more phases would add more complexity with balancing the phase currents.
I think you will do better with the IGBT connected to the 200v dc rail and use high side drivers. These are a bit of a pig but so is messing about chopping the ground line at this sort of level.
The most critical area is the loop on the dc side directly at the switch, you need that to be very low inductance with a ground plane and lots of polypropylene pulse capacitors keeping the loop as short as possible. Ripple is no issue but spikes from the wiring inductance are. Copper bars and batteries might just as well be inductors as far as spikes are concerned, their low dc resistance counts for nothing in this case.
I didn't fancy pcb layouts even with the currents of my modest converter. I used a heat sink as the ground plane and mounted copper bars to this, insulated with silpad, then connected the mosfets and freewheel diodes to the copper bars with the capacitors across the top in the shortest possible route.
I used HCPL3120 opto drivers, which lets me separate the logic ground from the power ground, but the supply for the 15v to these devices needs extremely low capacitance to prevent dv/dt spikes going back into the control board and playing havoc ( especially with a 2917 tacho chip that I use for speed reference).
Flux
[ Parent ]
