Bert Hickman wrote...
Winfield Hill wrote:
Bert Hickman wrote...

For can crushing, I only go to about 3500 Joules (mainly to
reduce wear and tear on the spark gap switch).

That's using 140uF? That would be two paralleled 70uF 12kV
caps from your bank, charged to 7kV, or about 60% of the cap's
faceplate rating? Implying only 35% voltage reversal while
staying under a 20% limit? Is that with a 3-turn coil, which
would be about 1uH? What's the Q of the 13kHz resonance?

Yes, with a maximum bank voltage of 7.1 kV (for can crushing). BTW,
that's the MAXIMUM energy I use - but can crushing can be done with
considerably less energy. Since the work coil remains intact during
can crushing, I assume a high Q load (100% worst case voltage reversal).
Under this scenario, the capacitor dielectric system would see a peak
voltage swing of about 14.2 kV, which is an ~18% voltage reversal
based on the 12 kV faceplate rating of the caps.

Starting at +7kV and swinging by 14kV takes the caps to -7kV,
isn't that a -7/12 = 58% voltage reversal? Or does the 20%
spec refer to rapid reverse-direction voltage swings rather
than reversed-voltage polarity?

... I recently bought a Pearson Model 301 50 kA wideband
current transformer to allow for isolated current measurements,
but haven't had a chance to hook it into the system as yet.

These can probably work to beyond-spec currents at the upper
end of their frequency range. You can test this by looping
multiple turns through the sensor, simulating "extremely-high"
currents. You can also parallel the sensor loop with alternate
wire paths, etc., to extend the range and you can calibrate the
setup at lower known currents to obtain the new ratio.

It'd also be valuable to grab the voltage waveforms, which is
easy to do with capacitive dividers. For example ~1 pF on the
HV side and 1000pF on the low side for a 1/1000 divider.

Appropriate shields are also necessary, since signal strays are
competing with the 1pF main path. The long-distance output coax
can be part of the 1000pF. You can complete the circuit with HV
resistors and zener clamps to protect the probe's opamp buffer
amplifier, which helps isolate an expensive scope. A trimpot can
be used for calibration at the output amp; a 10V cal signal gives
a 10mV output signal, enough for accurate scope readings during
the cal adjustment. I have made such dividers working to 25kV,
with a 100Hz to 10MHz bandwidth, and it shouldn't be too hard to
extend any of those parameters.

And, always remember to be afraid - very afraid - of the energy
stored in these caps. They will not give you any second chances.
=:^[

I'd like to ask a question about capacitor failure. Considering a
capacitor that's gradually degrading, I wonder if the final failure
can occur during charging, as opposed to discharge. This would mean
everyone should be far away behind shields whenever any paralleled
HV capacitor bank has a significant voltage on its caps.


--
Thanks,
- Win