Re: THOR episode II

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

Original poster: "Terry Fritz" <twftesla-at-qwest-dot-net>

Hi Marco,

At 09:04 AM 6/21/2001 +0300, you wrote:
snip...
>
>About the reliability, my CCPS is as much robust as it can be, resisting
>to spikes, transients, noise, EM interference. The only thing it still
>doesn't like are power injections (not only transients) on its output
>terminals, i.e. the toroid arcing to the primary.
>
>One mounth ago I had a toroid streamer curving down and hitting the RSG
>stationary electrodes -> some HV diode bridge sections had to be
>replaced. I enclosed the RSG assembly in a grounded shield assembly.
>Then, last week, I was having again 3 m long streamers when one of those
>hit the primary. This time one driver board was seriously damaged: I am
>still repairing it.
>
>Just to tell you about the effect chain:
>as the CCPS output is a floating supply (a rectified transformer
>secondary), the streamer probably induced a fat transient from the
>transformer secondary back to its primary. The primary is actually fed
>with a rectified 3-phase 230 VAC (no isolation!).
>On one driver, I had the rectified 560 VDC arcing to the 230 VAC phase,
>vaporizing big sections of PCB copper and roasting even their backplane
>connections. Every single semiconductor on the board (i.e. all IC's, all
>diodes, all regulators, all IGBT's, etc) have been damaged, including
>several 1/4 W resistors. A real disaster...
>
>I have a safety gap between the CCPS and the coil: probably it helped a
>little bit, but not enough. Also all 230 VAC safety fuses got triggered,
>but they are so slow... 
>

Conventional fuses will not help such things.  Fuses may help to put the
fire out eventually but they are useless in protecting electronics with the
exception of some big high power devices that are designed to take the
fault load until fast fuses blow.

I don't think spark gaps will help much either.  They clamp high voltages
but they require that the current falls back to zero and the gap cools
considerably before they stop conducting.  Thus, if you have a spark gap
between the 240VAC line and hit it with a high voltage.  The gap will
conduct and probably stay conducting driven by the 240VAC until it melts
down.  This is sometimes seen by coiler's that have the AC outlet in the
wall catch fire.  This sound like what happened in your DC rail at AC arc.

MOVs (Metal Oxide Varistors) my help a lot.  Unlike spark gaps, they are
voltage driven not current driven.  They will conduct away the high voltage
spike but they will cease conducting once the voltage drops below their
rating.  This allows the voltage to be clamped without the chance of
getting a continuous current arc started.  They are fast and can absorb
streamer energies.

Another possibility is to absorb the arc energy into capacitors to ground.
Since your output is DC, this may help Thor a lot.  A 500kV arc will not
raise the voltage of a large filter cap much simply since it does not have
enough energy.  Large cap values from the SMPS outputs to ground will stop
a lot of things.  The caps have to have good high frequency response which
can be easily done by mixing high-value high-ESR caps with low-value
low-ESR caps and thus getting the best of both.

Stopping the arcs hitting the primary is another possibility.  Your can use
many more strike rails.  As long as you don't make a shorted turn, you can
virtually put a cage around the primary.  E-tesla can map the fields and
give you a good idea for the best rail placement but arcs tend to go where
they want.  I would add more rails over the primary so that a direct strike
is practically impossible rather than trying to find a sweet spot to put a
single rail.

I would avoid a plastic barrier,  This just adds another dielectric layer
into the fields which will probably help attract arcs.  Even though an arc
may not be able to break through it, multiple hits may cause surface
burning and soon fail the plastic or make an arc path back into something bad.

Directing the arc away from the primary works until someone forgets to do it...

Hope something here is of use to you.  Unfortunately, trial and error is
pretty painful in this case so one cannot try a bunch of things to find the
best or optimal method.  In this case, doing a bunch of things in hopes
that at least one will work may be the answer.

Cheers,

	Terry