[Amps] Glitch resistor + fuse?

[John Lyles]

In very high power amplifiers for pulsed RF service, I use active 
crowbars. With 15-25 kV plate voltages, I use triggered mercury 
Ignitrons. Right now we are evaluating a stack of big thyristors, 85 mm 
diameter hockey puck devices. In either case, these crowbars fire in 
less than 5 microseconds and completely protect the tube(s) by diverting 
the current from stored energy away from the tube spark. I always test 
the crowbar after repairs and at least once a year. A 29 AWG magnet wire 
of some length is shorted across the power supply output terminals, by a 
remote Ross HV relay. I have measured 32,000 amps in the crowbar device 
when it fired across a 30 kV supply with 240 uF. Its not pleasant when 
the crowbar does not trigger when the test wire is applied. Deafening 
sound and bright flash through the window.
The crowbar logic must also turn off the power supply as quickly as 
possible to prevent blown mains fuses and excess fault currents on the 
rectifiers. Relays delay this at least mS, but we try to open a Jennings 
vacuum contactor (three phase) in just over a powerline cycle.

Applying such experience to amateur radio power supplies, if you put the 
shunt thyristor device(s) in between two 'glitch' resistors, (break 
value in two) so that it is in middle of a voltage divider where the 
tube spark has ~50 volt drop, and the power supply voltage is at the 
other end, the crowbar device will continue to remain clamped until 
voltage is down to 50 volts. Worse thing is to put the crowbar device at 
the load end of the glitch resistor where voltage is only about 50 volts 
during a spark.

If you can find a fast current transformer like a Pearson from surplus 
or Ebay, they are typically used to sense the current in the B+ line and 
fire the logic that triggers the crowbar. The CT will put out a large 
spike of voltage when 60 Amps flows, so best to look for something with 
a 0.1V/A sensitivity. You can also build something. Can also use a low 
value resistor shunt in the bottom of the power supply to ground, to 
sense the overcurrent and fire a crowbar.

If your glitch resistor is good enough, then you don't need a crowbar. 
Best to test it either way, using the magnet wire approach recommended 
in the Eimac note (and other tube manufacturers supply similar 
guidance). I would wonder if a HV fuse breaks the plasma when it blows, 
fast enough.
73
John
K5PRO

> Date: Sun, 19 Nov 2017 00:26:36 +0000
> From: Alan Ibbetson <alan at g3xaq.net>
> Subject: Re: [Amps] Glitch resistor + fuse?

> Eimac's recommendation for fault protection
> http://www.cpii.com/docs/library/9/AB17%20Fault%20Protection.pdf
> suggests fragile tubes (I guess 8877?) need the energy in a flashover to be
> limited to 4 Joules. Eimac's figure of 50V across the arc and Jim's 2ms for
> the HT fuse to open with a 3KV PSU and 50 ohm glitch resistor gives 60A
> fault current and  50V*60A*2ms = 6J, so a bit too much. However the 3-500Z
> being discussed in this thread is very probably more robust than an 8877.
>
> Jim: how did you measure the 2ms fuse opening time? I can't see it is all
> that easy with a multi-kilovolt supply dumping many tens of amps through
> the fuse and glitch resitor!
>
> I wonder if any US readers use a crowbar circuit as described in the Eimac
> paper to really cut down on the energy a valve has to endure during an arc?
> There is the outline of a circuit using a string of thyristors (SCRs) in
> figure 2 here
> http://www.qsl.net/oe5jfl/flashover.htm
>
> although the circuit does not offer protection during the first half
> second, until the gate trigger reservoir capacitor has charged up. It also
> may not keep the thyristors triggered for 20ms or more, until after the
> primary-side relay/breaker/contactor has dropped out. A better approach
> might be to use transformer coupled gate triggering, driven from the low
> voltage DC supply that is already provided for the control logic. Some kind
> of pulse train circuit (NE555?) would be needed to keep the thyristors
> triggered during the "follow on" phase.
> 73, Alan G3XAQ