Over-voltage protection with series diode?

[skywalker898]

Hi, could a series diode be used to protect a power supply from a high voltage arc between electrodes in a vacuum chamber?

In the attached circuit, what would happen if the 80kV Target electrode arced to the 20kV Anode electrode? It looks like the 20kV power supply V2 would be protected, but I don't understand what would happen to the Anode electrode. If current cant flow through D1, then an arc between Target and Anode should bring the Anode electrode up to 80kV. The arc probably wouldn't be sustained because current can't flow, but there would be an initial transient as the two electrodes are brought to the same voltage. Then the arc stops. But then is the Anode electrode just going to float at 80kV after the arc event because it's in a vacuum chamber and the diode D1 prevent charge equalization? (apart from a small leakage current on the diode D1). Am I correct in this interpretation of what would happen?



The same question applies to using D2 to protect V3 from V2.

This next circuit has two MOV's to ensure that the overvoltage arc causes current to be routed to ground. The idea here is that the diode protect the power supplies from the immediate transient overvoltage, and the MOV activates after a very short delay to route the current to ground. With the MOV, it seems that the Anode Electrode would not float after the event and would go back to its intended 20kV.



Or is the diode unnecessary in this circuit and just the MOV is needed?

Most of the research I've done says that MOVs, TVS, or thyristors are used for crowbars or clamping circuits (apart from one source which says a series diode can be used: **broken link removed**

The series diode would definitely have a large forward voltage drop if they are rated to kilovolts, which is not desirable.

 

[chuckey]

I hope there is some current limiting on the 80 kV supply. Diode, .8Vf for 1kV Reverse, ~100Vf drop. Biggest problem is the construction to keep corona at bay and to keep the current paths going to earth and nowhere else.
Frank

 

[skywalker898]

The 80kV is a Glassman with short protection to limit the current to 37mA. We don't have an external series resistor on the +80kV DC supply because it is already current limited and short protected from the factory, and because the electron beam needs to draw 30mA (relatively close to the current limit). Putting a resistor in series with the electron beam would require way too high power of a resistor, and probably cause an unacceptable voltage drop across it.

What about the series diode protection? What do you think of that?

 

[Dan Mills]

I think your anode would probably flash over to the grid and the grid to the cathode, depends on tube geometry.

Also, with the grid 30kV above the cathode it seems likely you are running much grid current (and hence much power dissipated in the grid, usually not a good thing)

You may also wish to consider the fact that the grid may under some circumstances need to source current, particularly when it all starts going wrong (ion bombardment of the grid, shouldn't happen, does).

If I was you I would put a glitch resistor in series with each of those HT supplies, even a 1K job will limit the peak transient current from the output caps of the glassman to ~80A while the thing discharges and 1K at 30mA is only 30V and 1W, finding a 80kV rated 1k resistor may be slightly problematic, 10K would not be unreasonable (300V drop, 10W dissipation steady state, 8A peak glitch, 160kW until the energy in the caps is spent (Probably only a few joules)).

Personally I would probably go for a glitch resistor and spark gaps to protect the lower voltage rails from an 80kV accident.

I hope you have good health physics folks, 80kV in vacuum at 30mA = one hell of an x ray source.

Regards, Dan

 

[D.A.(Tony)Stewart]

The 80kV is a Glassman with short protection to limit the current to 37mA. We don't have an external series resistor on the +80kV DC supply because it is already current limited and short protected from the factory, and because the electron beam needs to draw 30mA (relatively close to the current limit). Putting a resistor in series with the electron beam would require way too high power of a resistor, and probably cause an unacceptable voltage drop across it.

What about the series diode protection? What do you think of that?

You may need series and reverse shunt protection to prevent reverse spikes when the arc is extinguished V=Ldi/dt, but I would confirm your layout with Glassman as I would expect their unit to be fully protected under prudent applications.

Each electrode feed and wire is an inductor, there will be coupling inductance and capacitance and you can limit the current further with external string of R's. I suspect they cannot limit th current due to your layout mutual capacitance which gets discharge during an arc.

In production, When I found HIPOT tester was destroying AC-DC PSU's with DC output grounded, thus shunting the transformer isolation, thus the 100uA current limiter was being bypassed by the external arc pulse and the failure to electronic parts was catastrophic. By simply stringing series 10M resistors to achieve the voltage breakdown level for a 4kV test, I could easily detect BDV or breakdown voltage failures in the PDU with external current limiting R's to prevent internal failures. This is what I suggest you do, unless you want high current pulses for some unknown reason.

The Glassman unit will have OCP and shutdown to quench the arc, so choose a level just above their OCP threshold level to quench the DC arc. unless you are testing with AC.... or you want to limit the current to a lower level, as I did. Be sure to compute the sustained power dissipation of the current limiting R if you intend to not rely on internal protection. Your Glassman uses a 3kW PSU.

 

[skywalker898]

Hi Dan that's a good point about the grid power. I actually simplified the circuits you saw too much - in reality the 20kV and 3kV are focusing electrodes that "ideally" shouldn't intercept the beam. The actual grid is held at about 500 V positive relative to the cathode and should only intercept about 10% of a 30mA beam, so only 1.5W dissipation in the real grid.

I had looked for high voltage resistors, but like you said they are pretty hard to find for 80kV and matching power. I'll do some more digging.

It sounds like you would avoid the series diode altogether then with the spark gap?

 

[D.A.(Tony)Stewart]

spark gaps are slower to ionize then they are faster. (than semiconductors for PIV failure)
If the grid is just DC, a LPF will reduce the peak surge easily. You might want to consult with the OEM PSU maker. you need only to reduce the rise time and current limit to allow the internal protection to work. The series R and internal C will do this using a carbon spark plug wire.

 

[D.A.(Tony)Stewart]

When I focused a beam of 60kV at 20mA on our Scanning Electron Microscope at >50k magnification onto a perfect insulating ceramic slider from a magnetic recording head for a 14" HDD in the late 70's , it burnt a crater into the surface.,

Before this it created a moving charge cloud that looked like a giant storm then lightning then the burnt a few u" crater visible to the naked eye from the surrounding black area when removed from the ultra low vacuum.

I could have drilled a via if I was more careful,,but the purpose was to qualify new heads and check the gap length and width.

 

[skywalker898]

Hi Sunny, yeah the slow response of a spark gap is why we were thinking MOV, but we might need to do a spark gap on the 20kV line because I haven't seen an MOV with an operating voltage that high.

I'm trying to find a 10k resistor rated to 80kV (with >10W) and having some trouble but I'll call around tomorrow. I wonder what the internal resistance of the Glassman supply is? It would be nice to keep our 80kV cable connection simple and avoid a series resistor on that line if the surge current from the capacitors is not absurdly high. I don't really have a good idea for what the transient short current would be from the Glassman.

I believe the 60kV beam did some damage. We are making using the electron beam for x-ray experiments. The beam is hitting a molybdenum target.

- - - Updated - - -

I think your anode would probably flash over to the grid and the grid to the cathode, depends on tube geometry.

Hi Dan, ^this is a VERY good point and I think this is the most likely scenario. The diode on the Anode would block the current so the Anode electrode would float up to 80kV momentarily, which would then likely cause an arc to the next lowest electrode or down an insulator surface.

You may also wish to consider the fact that the grid may under some circumstances need to source current, particularly when it all starts going wrong (ion bombardment of the grid, shouldn't happen, does).

I hadn't thought of this. What would be the remedy? Just removing the diodes? I'm thinking of sticking with a spark gap on the anode to protect against the 80kV overvoltage, and MOV's on the lower voltage electrodes.

If I was you I would put a glitch resistor in series with each of those HT supplies, even a 1K job will limit the peak transient current from the output caps of the glassman to ~80A while the thing discharges and 1K at 30mA is only 30V and 1W, finding a 80kV rated 1k resistor may be slightly problematic, 10K would not be unreasonable (300V drop, 10W dissipation steady state, 8A peak glitch, 160kW until the energy in the caps is spent (Probably only a few joules)).

I'm looking for a 10K >10W 80kV resistor now. I'll also pass the circuit on to the OEM to see what they think.

I hope you have good health physics folks, 80kV in vacuum at 30mA = one hell of an x ray source.

That's the goal!

 

[D.A.(Tony)Stewart]

spark plug wire is a good resistor !

 

[Dan Mills]

One trick the EHT research guys do for resistors is a length of silicone tube full of water, tweak TDS with a titration of sodium bicarbonate solution to make the conductance suit.

A Microfocus xray source running at that power level? Good luck getting the heat away from the target.

Regards, Dan.