1:1000 voltage divider for 40kV mishap

[Zyxel]

I just had a weird thing happen to me when trying to build a 1:1000 voltage divider for measuring the voltage on the output of a rectified LOPT (a.k.a. "TV flyback") followed by a 1nF 40kV smoothing capacitor connected in parallel with the LOPT's secondary.

I made this divider out of 10GΩ and 20MΩ resistors connected in series (the latter yields 10MΩ in parallel with the internal 20MΩ impedance of my voltmeter).
The 10GΩ resistor was rated up to 40kV and 5W.

I connected the HV DC to this voltage divider and the 10GΩ resistor got a little warm ...just like I expected.
Next, I connected the voltmeter across that 20MΩ resistor  ...and the 10GΩ resistor arced over internally and is now destroyed.

I doubt that my HV from the LOPT exceeds 40kV DC.  Anyway the 10GΩ resistor was fine after I applied the HV from the LOPT to it ...and before I connected the voltmeter.
Also, note that any accidental short across the 20MΩ resistor would increase the current through the 10GΩ resistor only by 0.2%.

Why would connecting the voltmeter to this voltage divider cause this disaster ?

[davekni]

Are you sure the resistor is rated 40kV in air?  HV parts are sometimes specified for operation under oil.

Concerning connection of the meter, my only guess would be some lead wire or other conductive object getting close to the side of the HV resistor, initiating a discharge that then bridged the resistor.

[Zyxel]

Are you sure the resistor is rated 40kV in air?  HV parts are sometimes specified for operation under oil.

I am not sure. It is a 52mm long glass resistor, though.

Concerning connection of the meter, my only guess would be some lead wire or other conductive object getting close to the side of the HV resistor, initiating a discharge that then bridged the resistor.

I am sure that was not the cause.  Everything was galvanicaly isolated and all wires were very far away from the resistor.

[davekni]

This is likely not relevant to your situation, but I'll share on the outside chance it is.  The only glass-cased resistor I have contains a long thin black element (perhaps carbon on ceramic) sealed inside a thin clear glass tube, with a wire lead at each end.  10G if I recall correctly.  Looks like it would be great for HV.  However, it arcs internally end-to-end at a few kV.  Appears to have either low-pressure and/or perhaps a noble gas inside.  It was from an ancient radiation meter - sense resistor for very-low currents.

[Duane B]

Even though you don't think so, the voltage could have been higher than 40 kV. If you build this divider again I would start with a lower voltage at first and slowly ramp it up.

A 55 mm long glass resistor is about 2 inches long. That is about how far 40 kV would arc in air (going by a rough estimate of 20 kV per inch). I think you are operating too close to the specs of the resistor. I have seen 40 kV probes and most of them have a 10G ohm resistor that is about 6" long. If you are operating close to 40 kV then your resistor should have a much higher rating for a safety margin, like at least 60 kV.

[Patrick]

I think you should clean the glass body with alcohol and submerge it in oil. Then try again and report back.

[klugesmith]

... The only glass-cased resistor I have contains a long thin black element (perhaps carbon on ceramic) sealed inside a thin clear glass tube, with a wire lead at each end.  10G if I recall correctly.  Looks like it would be great for HV.  However, it arcs internally end-to-end at a few kV.  Appears to have either low-pressure and/or perhaps a noble gas inside.  It was from an ancient radiation meter - sense resistor for very-low currents.

In my experience, high-gigaohm resistors in glass tubes are NOT for high voltage.
They are for small (nanoamp, picoamp) currents in ordinary-voltage circuits.
For example, radiation meters.

Modern ionization-based smoke detectors avoid the cost of a fancy resistor
by using a reference ionization chamber as load R.   Irradiated by same Am-241 source, but baffled to exclude the smoky air under test.