Herrick's s.s. progress & woes

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: Ken or Doris Herrick <kchdlh@xxxxxxxxx>

Firing up the new-to-me but used CM300DY-24H IGBTs in my push-pull 
primary and at low power, I re-learned one thing I'd forgotten: keep 
high-impedance nodes far, far away from secondary coils.  I'd learned 
that when trying to use charge-pump driver-ICs years back and I 
re-learned it this time as to the opto-isolators that were to gate my 
present driver circuits.  Those little rascals have bipolar-base 
leads (or nodes, where the base is not brought out of the package) 
that can act, so one should expect, as little antennas; you have to 
leave them floating if you want the opto-part to do its thing, so the 
secondary's EMF floats right into them.
So I had to redesign the drivers a bit--and made a quite satisfactory 
(it seems) circuit using a 2nd (toroidal)  transformer + 2 bipolar 
transistors in place of each op-isolator, with negligible additional 
modification to other circuits, happily.
That seems to work like a champ: it positively clamps the IGBT gates 
to -27V between spark-events, positively drives to +/- 27V during the 
spark event, and it's all low-impedance and galvanically 
isolated.  However...as I turned up the mains power, operation seemed 
a bit anemic and then...*pop*...a failure--not unexpected in light of 
my past history, sad to say.
Out of curiosity I opened up the 2 bricks that had previously 
failed.  In my circuit I use only one of the transistors (Q2) in each 
brick (using just the diode of the other transistor as the snubbing 
diode), and...in both bricks the Q2s were burned out while the Q1s 
appeared pristine.  So at first it seemed likely that my circuit was 
at fault rather than the used bricks.
But upon some reflection and considering what I saw inside the 
bricks, I now conclude it's not my circuit per se, or bad bricks, but 
rather, that bricks are not for me--new or used.  I think what I've 
failed to understand is that if you try to run an IGBT brick at a 
frequency very much exceeding its stated upper value, it's not just 
that the circuit won't work or won't work well: it's that you will 
likely burn out the brick.  (And the data sheets don't trouble to 
mention that.)  All those paralleled internal transistors (2 sets of 
4 in the CM300DY)--with their many paralleled connections (sets of 6 
or so to each die)--will not turn on & off at the same time.  As a 
result, the load current will distribute unevenly during the too-fast 
turn-on/turn-off, and popping of those teeny little wires will occur: 
pop, pop, pop...right down the line until the whole thing is gone in 
a microsecond.
So...it's back to MOSFETs, I think...should I muster the energy to 
stay the course, that is.
Ken Herrick

A P.S:  I repeat for novice readers my take on the fundamental 
drawbacks of s.s:  1) Too many parts, a failure in any one of which 
could well lead to another, instant and expensive, failure; 2) 
difficulty in making oscilloscope measurements on high-voltage, 
mains-referred circuits; 3) difficulty in keeping out extraneous 
signals from the scope due to the large EMF generated by the (usually 
adjacent) secondary; 4) the need to avoid those pesky little 
antennas; and last but foremost, 5) the expenses that can be 
repeatedly incurred as a result of design-flubs.
KCH