Grant said:
Besides, every time I looked at a datasheet for IGBT, they're too
big, too expensive, or MOSFETs seem a better fit for what I was
planning.
But they're great when you _need_ a "too big" transistor, or anywhere
you're switching a lot of voltage not too quickly.
Just one pair of those FGH80N60's should be able to handle my 10kW
induction heater, given enough cooling (i.e., heat spreaders). As is, I'm
using four, which spreads out the heat well enough not to need spreaders.
(The heatsink is water cooled, in case you're wondering.) I've ran them
at 80A peak output (sinusoidal), continuously. The only thing that's
burned out so far is one GBPC3508, which went after ten minutes, quite
understandable as it was past ratings (obviously, a 10kW induction heater
needs two in parallel, a problem which has since been, if you will,
rectified).
And I spent all of $20 on those transistors. You can't even find a module
rated the same (200A 600V, "fast" switching, with co-pack diodes) for
under $100.
IGBTs suck below 300V, and they don't tolerate funny voltages. So they
aren't great for 120VAC stuff, where conduction losses are higher, or
boost/flyback converters or motor drives necessarily, where voltages can
get peaky unless you add a lot of protection. These are all things that
MOSFETs excel at. You can run a 500V MOSFET in a 400VDC output PFC boost
circuit, and it simply works, with high efficiency over a wide range of
voltage and current.
IGBTs are intrinsically slower than MOSFETs, limited by charge effects, so
they aren't good for HF and RF type roles, where you can drive the piss
out of a MOSFET and just have it work. They are fast enough for SMPS use,
at least if you drive them well.
I've never seen an IGBT rated for avalanche. I don't know why they can't
make them rated for it. I do know they love to fail if the peak voltage
goes near ratings. So you use Vce(max) about twice the supply voltage,
which makes 600V transistors best for 240VAC, or 1200V transistors for
480V industrial stuff. They come in all sizes and shapes these days.
This one's under a buck at Mouser:
http://www.fairchildsemi.com/ds/FG/FGP5N60LS.pdf
5A at 600V with < 2.1Vce(sat), 18.3nC gate charge.
A typical 5N60 FET,
http://www.fairchildsemi.com/ds/FQ/FQPF5N60C.pdf
costs about the same and has identical Qg, but Vds(sat) is 10V at 5A! A
fairer comparison would really use a ~10N60, to get conduction losses
closer. But then Qg would be about double, as would the price.
This IGBT, using a conservative 20W rating (figure passive heatsink,
sil-pad, and the inherent limitations of the TO-220), could do 10A average
easily enough, which is 3.2kW for a pair in a half bridge forward
converter. Half bridges are great for IGBTs, since you can bypass the
rails tightly.
This IGBT is also available in D/D2PAK, which would reach ratings (~2W on
PCB) at about 1A average, which is still 320W. Not bad for SMT.
IGBTs are astonishingly good at peak current. There's a GT8G132 in my
camera's flash. It's a teeny SO-8, rated for 400V 150A (Vce(sat) typ.
4V). It carries full flash current, then turns off the tube once the
image has been exposed. IGBTs are also used in plasma TVs to drive row
and column and field and whatever other signals, with similar peak
currents and narrower pulse widths. It goes without saying, the MOSFETs
required for either of these applications would be massive and
expensive -- compact consumer electronics like these weren't even possible
until cheap, fast, reliable IGBTs arrived. The days of slow,
latchup-prone IGBTs are long gone.
Tim