I've decided MOSFETs aren't the way to go, because they are too easy to break! I know it is possible to use them, it's been done a million times before, but soldering the SMT mosfets to a copper buss bar with a torch takes a long time, and I've melted the plastic that the buss bars were bolted to.
IGBTs are much more rugged - basically IGBTs are the ruggedness of a BJT glued to the easy gate-drive of a MOSFET.
So I bought a pair of single IGBTs off of ebay
Mistsubishi MG400Q1US41 N-channel IGBT - 1200V 400A
That's right, 400A of switching goodness. Now I need a place to mount them!
So I went to McMaster-Carr's website to buy some metal.
IGBT bricks need a fairly flat surface to mount them on, for heat transfer purposes. So I bought a quarter-inch thick, six inch square of aluminum, precision ground to a high degree of flatness
(6.35x152x152mm for the imperially-challenged)
Any warps in the metal means there will be an air gap between the IGBT and the 'heat sink', so the IGBT will be unable to dissipate heat properly.
It is very flat, has very sharp edges, and is one stiff piece of metal - perfect, since it is the structural support for the whole controller.
Then I need some more copper, to connect the IGBTs to each-other, and to the wires. (my mosfet controller used all the copper buss bar I scrounged from home)
The main concern was being able to fit the bar inside the LEM current sensor, its opening is only 10.4mmx20.4mm, so 3/8"x3/4". Of course no-one sells this size, but 1/4x3/4 is easy to find. I got a three foot length.
My apologies for not having pictures of all these components separately, but I hope these make up for it:
The blue brick is an IGBT snubber, a 0.68uF cap, 1250V designed for this application. You mount it directly on the IGBT, across the switch. It dampens the energy spikes, preventing the voltages from getting too high (well not preventing, just ... damping) although it wastes energy every single cycle (0.5C*V^2 joules - 8.5W at 50V and 10kHz). That's a lot for a cap, but nowhere near the power getting to the motor (fingers crossed).
The silver strapping on each IGBT is shorting the Base (Gate) to the Emitter lead - thus turning off the transistor, as well as protecting the gate from ESD - I would cry if I broke one. The gate on modules as rugged as this is still *very* fragile. 20V and *poof* when it can withstand 1200 on the other two, pretty crazy.
Circuit:
Exactly the same concept as before. One transistor that pulls the M- bar to B- and one diode that allows freewheel current from M- to B+ when the transistor is off.
IGBTs have a reverse diode like a power mosfet does, but it is much a better diode, rated at the full 400A, with a pretty low voltage drop (0.3 on a meter test, but that means squat in the hundreds of amps)
Gotta run! Expect more frequent posts in the near future.
