DRSSTC MMC failures, Was:15kva 14.4 kV Transformer.. (fwd)

["Tesla list" <tesla@xxxxxxxxxx>]

Moderated and approved by: Gerry Reynolds <greynolds@xxxxxxxxxx>



---------- Forwarded message ----------
Date: Sat, 23 Dec 2006 14:25:40 -0600
From: Steve Ward <steve.ward@xxxxxxxxx>
To: Tesla list <tesla@xxxxxxxxxx>
Subject: DRSSTC MMC failures, Was:15kva 14.4 kV Transformer..

Hi Terry,

Yes, i "know" why they failed ;-).  The operating conditions were 6kV
peak, 90A RMS, and 700A pk.  Now for the capacitor specs: .45uF, 4kV
peak, 78A RMS, and 2532Apk.  Ok, so the problem certainly was not the
peak current rating ;-).  The capacitors run warm normally, but
eventually one or two started to get *hot*.  These capacitors have
seen years of use (though i dont really get to run my coils too much).
 Anyway, the failure was over voltage from what i can tell, possibly
aided by the heating due to the RMS currents.  When i unrolled a cap,
i found this:

http://www.stevehv.4hv.org/capfailOC/

Thousands and thousands of dielectric punch-throughs.  This capacitor
eventually measured 0nF (the 2 sections of capacitor had been
completely bi-sected!)  What finally settled the question of failure
mode for me was when i tried to run my 6' spark coil without a
breakout point (mind you, it was topped with a 6"x24" spun aluminum
toroid).  I got to maybe 75% input voltage before i wussed out (it
didnt break out).  At that level, i could hear my MMC snapping!!!  It
was undergoing a voltage failure on probably every bang (120 per
second, or more!).  Of course, you cant normally hear this over the
output sparks, so it might have been happening all along, or maybe the
capacitors took some time to wear out at 50% over voltage ;-).  By the
end of this last run, the capacitor measured like 50% out of spec over
all, some caps only read 10-15nF (150nF caps).  These are our beloved
CDE 942C .15uF 2kVDC caps.

I have since built a new MMC from metalized film GE type caps.  Their
RMS rating is nearly as good as the film/foil, but the Ipk is about
half.  In the case of a DRSSTC, i typically have many parallel
strings, so peak rating is never an issue like it is with SGTCs.  This
new cap is rated 7200VDC, hopefully thats enough to hold up for a few
years.  I havent actually been able to test this new MMC, since moving
away to school.  An interesting note: i "accidentally" stress tested 1
of these GE caps to about 2400V (they are 1200VDC rated),  I could
start to hear it snap.  At about 3000V, it was snapping all the time.
I let it run like this until it was way off spec, and the innards
looked very much like the failed CDE caps.  Surprisingly, the H-bridge
doesnt seem to care if the cap suddenly shorts out.  In fact, there is
almost no change in the primary current waveform (unless i just dont
have the bandwidth to see it).  Just a small glitch, and it keeps
going.  So apparently when these caps fail, they clear the short
*fast*.

Steve Ward

On 12/23/06, Tesla list <tesla@xxxxxxxxxx> wrote:
> Moderated and approved by: Gerry Reynolds <greynolds@xxxxxxxxxx>
>
>
>
> ---------- Forwarded message ----------
> Date: Fri, 22 Dec 2006 22:18:51 -0700
> From: Terrell Fritz <terrellfone@xxxxxxxxxxxxxxxxxxxxxxx>
> To: Tesla list <tesla@xxxxxxxxxx>
> Subject: Re: 15kva 14.4 kV Transformer.. (fwd)
>
> Hi,
>
> Of course,
>
> "Steve knows" "full well" about MMCs ;-)))))))))))))
>
> But just for "statistics", what was the failure mechanism????????????
>
> Peak current, RMS current heating, pure over voltage, Corona damage with
> old age...????????????
>
> The thermal time constant in the IGBT's thermal expansion coefficients is
> far bellow the "ripping and tearing" point at say 60Hz and above...  So
> heating and cooling is a "don't care" unless you turn it "on and off" a LOT
> (like 50,000 times ;-)))...
>
> If you really "don't know" better get hold of "ScanTesla" :o)))))))  But
> prolly RMS current and over heating then...  Polypro melts at less than the
> boiling point of water...  Cut them fried caps "open" and you will "see"
> inside...
>
> Cheers,
>
>          Terry
>
>
> >Excellent DC!
> >
> >I find that my MMC eventually fails (before the IGBTs provided no
> >"special events" occur).  I have one DRSSTC that has gone through 2
> >MMC banks now, but still has the same IGBT set.  It consistently pumps
> >out 6 foot sparks at 1300W input.  If given the chance, id also try
> >running one of my coils for extended periods of time to find out where
> >the bottleneck in the design is.
> >
> >Steve Ward