Re: Hybrid SG/SISG idea? "IGBT-Assist Spark Gap"?

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: "J. Aaron Holmes" <jaholmes@xxxxxxxxxxxxxxxxxx>

Thanks, Terry and Dan.  Great info!  This helps a lot.
 I like to have *some* sense of *why* something works
(or why disaster *doesn't* strike!) before I go off
and build it ;-)  If the SISG really takes off, you
oughta make friends with somebody in the right place
to conjure a one-piece SISG...something that just
looks like an IGBT brick with two leads instead of
three ;-)  Maybe the TCML collective should spring for
a PowerBall ticket every now and then ;-))

Regards,
Aaron, N7OE

--- Tesla list <tesla@xxxxxxxxxx> wrote:

> Original poster: Vardan
> <vardan01@xxxxxxxxxxxxxxxxxxxxxxx>
>
> Hi Aaron,
>
> At 02:26 PM 6/5/2006, you wrote:
> >Thanks, Terry!  Makes sense.  The illustration was
> >really only intended to suggest that the IGBTs
> would
> >be stacked with (effectively) a single gate.  I
> should
> >have realized the potential difference thing,
> though.
> >Curiously, has anybody done any IGBT "stacking"
> like
> >this for, e.g., a triggered gap?
>
> See Dan's new post on this!
>
> >I'd be interested to
> >know what's required to avoid, e.g., overvolting
> IGBTs
> >if one "fires" before the others.  In the SISG,
> what
> >prevents this from happening?  Or is the turn-on
> time
> >considered "sufficiently-identical" (provided
> you're
> >using the same parts for all modules) to avoid
> this?
>
> It is a very messy thing...  The sections have
> significant
> capacitance across them that keeps the voltage from
> instantly
> jumping.  The SIDACS turn on in the nS time frame
> but they don't
> start to "really conduct" till far latter.  The
> primary circuit will
> not deliver current either for say 20nS and then the
> current rise is
> governed the LC...
>
> If all the sections turn on but one, the voltage is
> still 900V there
> since there is no current in the string yet to drive
> it higher.  As
> the current does start up, the dV/dT is easily slow
> enough for a lazy
> SIDAC string to get the message and start up too.  I
> think the real
> key is to remember that at the instant the voltage
> across the gaps
> goes to zero, the current is also "zero".  The dI/dT
> is pretty slow
> compared to what the SIDACs can react to.
>
> That is pretty over simplified and SIDACS are
> designed with "tricks"
> to make stringing a "happy thing".  There also
> "really is" current in
> the string to charge the caps to turn on the IGBTs.
> That can't
> happen unless they are "all" on....  You can't
> charge one without
> charging all...  SIDACS are real tough too and could
> handle the first
> cycle all by themselves!!  The IGBTs would be in
> danger after 1200V,
> but the SIDACSs are way too fast to let that happen.
>  The IGBTs could
> ride the voltage far far longer than the SIDACS
> could sleep...
>
> I avoid this issue since it is such a mess to
> explain...  But it will
> not blow up ;-))  It might get to be messy if sync
> triggering or
> quenching circuits come into play...
>
> Cheers,
>
>          Terry
>
>
>
>
> >--- Tesla list <tesla@xxxxxxxxxx> wrote:
> >
> > > Original poster: Vardan
> > > <vardan01@xxxxxxxxxxxxxxxxxxxxxxx>
> > >
> > > Hi,
> > >
> > > At 01:02 PM 6/5/2006, you wrote:
> > > >No comments?  In case there was any trouble
> > > >visualizing this (I doubt it, but...), here's a
> > > sketch
> > > >done using top-of-the-line schematic-drawing
> > > software:
> > > >mspaint.exe ;-)
> > > >
> > > >http://silicon-arcana.com/SG-SiSG.jpg
> > > >
> > > >Doubtless this sketch is **WAY**
> oversimplified!!
> > > It
> > > >probably doesn't include a lot of necessary
> evils
> > > >required to place IGBTs in series (may need a
> TVS
> > > per
> > > >IGBT, divider to equalize the voltage across
> the
> > > whole
> > > >stack, etc.--thoughts?).
> > >
> > > The gates of the IGBTs are at about 900V
> potential
> > > difference each up
> > > the stack.  They cannot be hooked together.  One
> of
> > > the earlier
> > > problems was how to get an isolated power supply
> to
> > > each of the gates
> > > and how to trigger them.  Batteries and fiber
> optics
> > > would probably
> > > work and I did some tests on that years ago.
> But it
> > > was all just too
> > > complex and messy...  The SISG fixed all that.
> > >
> > > There is a problem if the spark gap does not
> fire at
> > > just the right
> > > voltage.  If it fires at too high of voltage the
> > > IGBTs could
> > > breakdown and that is a bad thing.  If the
> sparks
> > > gaps were divided
> > > among each device to get around the above, then
> they
> > > would have to
> > > fire at 900V which is not easy.
> > >
> > >
> > > >In short, the modularity of Terry's SISG is
> very
> > > cool,
> > > >but I'm wondering what simplifications one
> could
> > > make
> > > >if he/she started out assuming that they were
> going
> > > to
> > > >need a LOT of IGBTs, e.g., for use with a 15kV
> NST
> > > or
> > > >something bigger.
> > >
> > > Boards like Mike's are 3600V each.  So you can
> have
> > > "big" modules
> > > too.  If it all works out real good, someone
> will
> > > probably sell the
> > > whole thing pre-made.
> > >
> > > >Ah...if only those 6500V IGBT
> > > >bricks were cheaper!! :-)
> > >
> > > :-))))  At least they do exist!!!  Not sure they
> > > have the reverse diode?
> > >
> > >
> > > >And again, the ability to adjust the firing
> voltage
> > > by
> > > >just varying a spark gap like in a regular SGTC
> > > would
> > > >sure be neato.
> > >
> > > I just add and remove sections by moving the
> wire:
> > >
> > > http://drsstc.com/~sisg/index_html_m2dc0296e.jpg
> > >
> > > SIDACSs cost like 50 cents each so there is no
> great
> > > price
> > > advantage.  Some people have trouble getting
> them in
> > > other countries
> > > so they are trying Zener, and TVSs.  Not sure
> how
> > > well that works.
> > >
> > > Of course, all this is very young, so who
> knows...
>
=== message truncated ===