Low-leakage PNP or PMOS

Most highside switches will have more than 5 pF of capacitance themselves.

How fast does it have to charge that cap? My first choice would be a relay!

Two sections of a BAV23 in series make a 500v diode with below 0.5 pF. MMBD5004S would be a bit better, 1KV and about 0.25 pF. But you'd need a pulldown after the big switch. Gets ugly fast.

You could possibly bootstrap the leakage. Maybe not.

Why not just leave the switch on for a second? In other words, what are you actually trying to do?

Back in 1979 I used some insulated gate FETs which relied on silicon nitride rather than silicon oxide for the gate insulator. They had very low gate leakage and could be turned off quite hard. Google doesn't show up anything helpful.

The data sheet numbers were limited by the current that could be measured in short time on the production line, They did a lot better in real life.

Assuming that 5pF and second timescale is not a typo then you are asking for nano ampere leakage at 400V reverse bias at 50degC - let alone the charge injection or switch capacitance. That is very hard. Is there any way you can bootstrap?

piglet

Picoamp.

The problem is under-specified.

If you can get a canonical photocell (vacuum type) with a UVLED cathode excitation, the OFF resistance of a dark photocell is more than good enough. Semiconductors are tested, in mass-production fashion, to loose leakage specifications, it's HARD to test nanoamps/picoamps. With tubes, though, it's guaranteed by design.

In below 1us, ~1kHz in the worst case. And you are totally right, the parasitic capacitance kills this simple idea. Now it is obvious, wasn't at the beginning.

Pulse charge a GM tube to experiment with the time-to-first-count approach. Basically, you quickly charge this GM cap and measure how long it takes to record the discharge pulse triggered by a particle. Then you apply statistics.

I have actually succeeded in Ltspice with a single pulse boost converter. I deliver a controlled quantity of energy and the GM tube charges from 0 to 400V in 330ns. The discharge pulse is sort of safe, because there is only so much energy in the parasitic and explicit capacitances. Dunno how it would work in a real case, still waiting for the tube.

Best regards, Piotr

Indeed. Now I can clearly see it is not about low leakage, it is all about rapid charging and maintaining that charge. But, to my surprise, it does not imply low leakage at all, at least not in the pA range. A series 100M resistor to V_SUPPLY will take care of any leakage.

Please consider the case closed, even if in an unexpected way. Thank you all for your input, I appreciate it very much. I'll post a Spice sim when I'm happy with its performance.

Best regards, Piotr

Picoamps of leakage will be a problem too, if you care about discharging 5 pF. A GM tube is very sensitive to voltage so you can't tolerate much droop.

You could charge the tube from regulated 400DC through a resistor, which current limits nicely and has low capacitance. Self-quenching. And simple.

A resistor and a series inductor is interesting but probably not practical.

You can get statistics from the discharge frequency. That's the traditional way.

Each shot is the start time for the next time-to-first-count. Given a random pulse, "start time" is anything you want to call it!

This is the traditional way, good for the low dose cases.

It works as long as the tube is not saturated. At high dose rates you stop getting pulses, let alone clear pulses. So the idea is to quench the tube hard by turning off the supply and periodically energize it rapidly. Rumor has it that it allows for 2 more orders of magnitude and I wanted to check it for fun. It might turn out to be simple enough to integrate it into a DIY meter.

OTOH, Phil Hobbs was right. The more I want the more it looks like a scintillation probe. Not giving up yet, though.

Best regards, Piotr

You could use a half H-bridge to power and quench the tube fast. Like an IR2213 and a couple of small mosfets and a bit of timing logic.

GM tubes are go or nogo, so can't do spectroscopy. Scintillators and PMTs are more fun.

Late to the party here, but for one-offs I'd probably use a tube for that sort of thing. (They don't come in P-channel, of course.)

Cheers

Phil Hobbs