Snubbing an IGBT in a sstc

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: "K. C. Herrick" <kchdlh@xxxxxxx>

I'll ask Chip to post 
<http://www.pupman.com/current/kcherrick/TCH_p-p4.jpg>http://www.pupman.com/current/kcherrick/TCH_p-p4.jpg, 
a simulation-schematic illustrating what I have in mind.

I'd like to ask if anyone has had experience with using a 
power-MOSFET in its avalanche mode in an active snubber circuit.  In 
my posted circuit, Q4 is a power-IGBT prospectively driving an 
untuned Tesla-coil primary, with secondary-return feedback to sustain 
oscillation.  In hardware, I would implement a push-pull 
configuration, with a center-tapped primary and a pair of IGBTs.

The snubber circuit involves Q1, Q3 and the associated 
components.  Absent snubbing, negative-going Q4 drive passes through 
D9 and turned-on Q3 while positive-going drive passes through R1, 
C4/R4 and D5.  (In simulation, C4/R4 was necessary to preclude 
oscillation during the snubbing.)  Thus negative-going drive is 
relatively fast while positive-going drive is relatively slow--as 
necessarily implemented in push-pull or H-bridge configurations.  But 
when Q4 cuts off every half-cycle, TX1's resultant overshoot reaches 
~800 V and Q1 avalanches.  Q1's current then passes thru D8, C4/R4 
and D5, shutting off Q3 and thus the negative-drive, and rapidly (and 
briefly) turning on Q4 to snub the overshoot.

The circuit of D6, C3, D1 and R9 acts to divert Q4's current during 
each turn-off so as to diminish the power dissipation in Q4, and also 
to slow the rise of TX1's overshoot to give time for the Q1 circuit 
to operate.  In effect, turn-off power dissipation is transferred 
into R9 each time Q4 turns on, easing the stress on Q4.  (In hardware 
for D6, I'd plan to utilize the otherwise-unused reverse-diode of a 
half-H-bridge "brick" of which Q4 is a part.  The diode needs to be 
fast and it needs to have a large reverse-voltage-withstanding.)

Simulation shows that Q1, in avalanche, needs to conduct ~2 A--which 
goes thru R1--while the voltage across it is 800 V.  That's 1600 W, 
but only for ~1 us out of 8 (for a ~120 KHz Fr).  That cuts the mean 
dissipation to 200 W during the pulse burst, and with a 1% sparking 
duty-cycle, that's cut further to only 2 W.  So the question is...can 
a IRFBE30 MOSFET or the like withstand the 800 V and at the same time 
the 2 A--even at only a 12% x 1% = 0.12% duty cycle?

I'd thought of using an MOV in the same circuit, or perhaps a string 
of them, instead of an avalanched MOSFET, but I'd be worried about 
the capacitance of the MOV(s).  Any ideas about that, as well?

Ken Herrick