John Larkin wrote...
If you look at the soar curves of a typical TO-247 fet, most of
them can dissipate kilowatts in the microsecond time range. But
I'd still expect a number of interesting blow-all-the-fets modes.
An IRFPS37N50A can dissipate about 40 kW for 10 usec.
Yes, but I'll have a stack of say 15 FETs, and so I expect a large
safety margin in the transient thermal resistance area.
Some years ago I purchased 200 IRFBG20 MOSFETs for $1.75 each
(they're still available, now at $0.96/100). These are 1kV 1A
TO-220 MOSFETs featuring moderately-low capacitance (most of the
semi houses have moved away from 1A parts for their 1kV offerings,
to 3A, etc., as the smallest ones in production). The IRFBG20's
capacitance specs are Ciss=500pF, Coss=52pF, and Crss=17pF at 25V.
See
http://www.irf.com/product-info/datasheets/data/irfbg20.pdf
The IRFBG20 is rated at 52W, and its single-pulse transient thermal
impedance (fig 11) is 0.01C/W for 1us, or 0.03C/W for 10us, so, if
we allow a 150C junction-temp jump, it can handle 15kW or 5kW, resp.
Notice the sqrt-time factor at work there, as the heat spreads out.
With 15 FETs equally sharing (another assumption), that's a 225kW
or 75kW capability for 1us or 10us, which is far more than needed.
Getting the FETs to share equally may be a difficult. My IRFBG20
FETs have their avalanche breakdown at about 1200V, so they could
theoretically handle an 5kW/1200 = 4.2A avalanche current for 10us.
Hah, that's more current than they're up for anyway! Certainly if
I'm switching anywhere from 0.5 to 2A, a short-lived FET avalanche
isn't something to fear. But if we did fear this (or, ahem, if it
failed to pass a know-nothing design review) I could parallel each
FET with a stack of say five 200V 1.5kW TVS diodes (that'd be 75
TVS for the 15 FETs, sheesh!). TVS zener diodes feature two slugs
of copper (one each side of the die) to absorb heat. The 1.5ke200
can handle 1.5kW for 1ms, or 20kW for 10us, so five in series can
handle 100kW. This is 20x more than an IRFBG20 can handle, and it
may be nearly 100x what we'll be needing for our 10kV ramp switch.
Hmm, maybe I'll look into some smaller SMD-packaged TVS parts.
One other TVS issue. The capacitance of a 200V 1.5kW TVS is 80pF,
measured at its breakdown voltage. Five in series would be 16pF,
which compares to 52pF for the IRFBG20 (I know these values change
with voltage, but this is just a quick look), so a TVS may not be
too painful in that regard. If we consider 68pF/15 for an entire
15-stage 10kV FET-switch stack, that's only 4.5pF, which doesn't
look bad either. ** Oops! ** ST's datasheet shows their 1.5ke200
to have 675pF of capacitance at 0V, which is 135pF for five. Ouch!
