Re: Pushing the IGBT Envelop

["Tesla list" <tesla-at-pupman-dot-com>]

Original poster: "Steve Ward" <steve.ward-at-gmail-dot-com> 

Comments:


On Sun, 12 Sep 2004 16:33:26 -0600, Tesla list <tesla-at-pupman-dot-com> wrote:
 > Original poster: "Greg Leyh" <lod-at-pacbell-dot-net>
 >
 > Interesting website, and coil work!  I'm still unclear on the DRSSTC theory
 > of operation, however.  Near as I can figure, the system actually runs as
 > an oscillator, but in a quasi-CW mode for relatively short bursts, at an
 > unspecified dwell and repetition rate.  Is this roughly correct?

Yes, thats correct.  Some DRSSTCs use an oscillator to drive the
H-bridge, but i use feedback taken from the secondary coil's base
current.  Either way seems to work well, though its not entirely that
simple as you might imagine.

The coil operates for maybe 200-300uS and then is at rest for several
mS (depending on the desired BPS).  The duty cycle on my system is
only 3%.

 >
 > H-bridge drives are great for power converters to 100's of kW, and
 > especially when operation from a unipolar DC source is required.
 > However, an H-Bridge plan wouldn't be cost-effective for the ALF since it
 > requires 4X the silicon for the same primary peak current and voltage swing

Hmm, what would be the alternative?  I sort of had the idea that you
were going for an OLTC-like scheme, where you charged the tank cap and
dumped them into the primary like a normal TC, but replaced the spark
gap with IGBTs or some other thyristor.  Its interesting to make a
very crude comparison between my DRSSTC-2 and Steve Conners OLTC-2.
We are both using about the same amount of silicon (i use 4x 300A
IGBTs, while he uses 2x 600A IGBTs).  We both run just about the same
input power (around 4kW).  His coil makes a 7' spark, while mine does
11' ;-).

But another big question is, would an OLTC become more efficient at
higher voltages?  Say if it was possible to get IGBT stacks up to
15-30kv range?  On the other hand, the DRSSTC seems to work well in
all cases, even with much lower input voltage. I run 600VDC on my coil
while Conner uses 1kv if i remember correctly.

 > -- an important consideration when the material costs for the silicon alone
 > is likely to be mid 6-figures.

Yikes!

   H-bridges are also susceptable to
 > 'shoot-through' failure, where timing errors or cosmic-ray induced firing
 > of IGBTs can hard-short the DC mains, resulting in catastrophic failure of
 > the entire IGBT stack.

Yes, a full-bridge is definately more complicated and requires more
IGBTs.  But, if you weigh in efficiency... or rather the fact that the
DRSSTCs require less peak current for a specified spark length, it may
turn out to be the better option.

   For perspective, here's a photo of a catastrophic
 > H-bridge drive failure at another lab, believed to have been caused by a
 > cosmic ray induced false-triggering one of the IGBTs:
 > http://www-group.slac.stanford.edu/esd/HBridgeDebris.jpg

Impressive ;-).  Ive seen plenty of my own IGBT failures with
shattered IGBT cases spread on the floor afterward.

 >
 > The IGBT casings and structures can be seen evenly distributed on the
 > floor.  This failure occurred on a system operating at several hundred
 > kW... imagine scaling this up to a 5MW system!  It's clear that
 > partitioning the primary drive circuitry in order to isolate component
 > failures will be essential for a 5MW primary drive system.

Yeah, *complete* a failure on such a system would be a huge financial set 
back!!

Steve Ward

 >
 >
 >
 >
 > >Original poster: "Steve Ward" <steve.ward-at-gmail-dot-com>
 > >Hi Greg,
 > >
 > >Very fascinating stuff going on at SLAC.  I've also read some of your
 > >papers about IGBT module design (and the downfall in commercial
 > >modules layouts).
 > >
 > >Anyway, DRSSTC is a Dual Resonance Solid State Tesla coil.  We
 > >basically drive a resonant tank circuit with a solid state driver
 > >(using IGBTs) and pulse its operation to achieve very high peak
 > >powers.  Jimmy Hynes was the first person to have a very successful
 > >coil built under this idea.  I think my coil is currently the largest:
 > >
 > >http://hot-streamer-dot-com/srward16/DRSSTC-2.htm
 > >
 > >It uses 300A IGBT (dual modules) rated 1200V.  We typically overdrive
 > >the gates so that we can push greater currents (my system runs at
 > >about 1200A peak) through smaller IGBTs, mainly since us ametures cant
 > >afford to buy enough silicon to actually run these things in their
 > >SOA... and after all, this is tesla coiling!
 > >
 > >The DRSSTCs are, as you might expect, slightly more efficient than
 > >even the best SGTCs.  My coils seem to follow the formula of: spark
 > >length (inches)=2*sqrt(power).
 > >
 > >I was almost hoping that if the advanced lightning facility every
 > >became reality, that it would be a DRSSTC.  Mainly because you would
 > >not need extremely high voltage silicon to replace a spark gap.  With
 > >a DRSSTC, our bang energy is not limited to the storage energy in the
 > >tank capacitor, it can be much greater (usually its 2X as much).  I
 > >think the fact that my large DRSSTC can make 11' sparks from just
 > >600VDC rail voltage and only 4-5kW shows promise for even larger coils
 > >(silicon permitting).
 > >
 > >Steve
 >
 >