generate 500V "square edge" signal (into ~1Meg / 50nF) ?

[max_torque]

I trying to compare the high frequency performance of a high voltage divider circuit i've built to an off-the-shelf 100:1 probe (pico).  To do this, i'd like to generate a reasonably high voltage "square" edge signal, where this signal will be fed in parallel into the commercial probe and my probe so i can estimate the frequency response of my version.

So, best ways to simply do this, with a fast and edge as is sensible?

Wondering about taking uk mains (240Vrms 50hz) sticking it through an isolation transformer with slight step up giving a non earth reference roughly 500V peak 50Hz sine wave, rectifying and smoothing it, then sending it through a (small) capacitor bypassed current limiting resistor and into a simple mechanical relay, that is switched on and off a couple of times a sec by a micro, and routes the pulsing signal out to some insulated banana sockets.  Using a dual throw relay i can switch the output between HV and ground to get matched impedances.  Of course, there will be some contact bounce but i guess i could probably filter those out with some sort of snubber, but having them present might actually help measure the frequency response of the two probes under test??

Alternatively, i'm into some sort of semiconconductor switching arrangement, with all the necessary work to make that safe and robust at these voltages?

[wraper]

Use mercury wetted reed relay.

[Someone]

50nF or 50pF?

500V is still within the range of "normal" mosfets so you can make a very fast edge in one direction with that and have a slow pull-up for the other direction. Then you're worrying about getting the inductance of the loop as low as possible.

[max_torque]

i'd like to protect for 50nF, but i'd settle on less!  It really depends on the capacitance of the probes i'm testing, which are likely to be in the pF region, but exactly how much capacitance, i don't know!

Maybe a more sensible value would be 5 to 10nF

[T3sl4co1l]

Uh... if you've not seriously screwed things up, it'll be linear.  Do the test at 10V and save yourself the trouble.

Tim

[David Hess]

Like T3sl4co1l says, you should be able to do this at a lower voltage.

If you must have high voltage, you only need one edge so I would use either a cascode switch with a low voltage MOSFET driving a high voltage common base bipolar or for even faster operation, a series string of bipolar transistors switching in avalanche mode.

A mercury wetted relay will also work great and is probably the easiest high performance option.  This is how the Tektronix Type 109 pulse generator worked.  I would try mounting a reed relay inside of a transmission line environment.