500 volt power supply

Isn't the RCD clamp done wrong?

Normally you do it like this:

There are at least two topologies of an RCD clamp.

The one that you show doesn't damp high-frequency ringing after the diode turns off. My diode multiplier chain actually clamps the fet drain swing in both directions, so there is no voltage overload hazard to the mosfet. All that's left is to damp the residual ringing to reduce radiated EMI a bit.

A plain RC will damp the ringing but adding the diode reduces damper power dissipation a little.

Yes, that could also be an idea. Will have some losses though, since it's switching into a capacitive load

A transformer would work better - perhaps several of them in series.

Dump a bunch of current into the primaries in parallel - perhaps time shifted to allow the high voltage pulse to propagate through the series connected secondaries - and rely on the parallel capacitances of the secondaries to shape the pulse into a half-sine.

An inductor to store a load of energy has to be a lot bulkier than a transformer which just couples it into a high impedance load.

Switching a DC supply into a capacitive load is inherently lossy, and people usually discharge the cap and waste that energy, too.

So power dissipation is C*F*V^2. The inductor trick has, ideally, zero power dissipation... it just sloshes energy around. That's how I got 4 MHz out of a tiny Pockels Cell driver.

If F is low, it wouldn't matter much.

It isn't, if the load hasn't got any resistance. Discharging the capacitor into a separate supply isn't clever.

What on earth is "F"?

Klaus seems to want to charge a 10pF load to 2.5kV peak, as a half cycle of 100kHz, getting it from 0V up to 2.5kV and back to 0V over 5usec.

If he sets up a 0.25H transformer secondary and charges it up with 10mA for about 2usec waits a usec and then discharges it at 10mA for another 2usec, he'd get close.

Figuring on a 16:1 step=up from about 100V, and 10V per turn, that would be a 160 turn secondary, which implies a core with an Al of 10uH per root turn, which is bit high.

Going to two transformers in series lets you drop the step-up to 8:1. Each secondary now needs to be O.125H which is an Al of 20uH. Three times more turns fixes that, but you are stuck with an 240 T secondary on each transformer and you don't want the parallel capacitance of the secondaries to get close to the 10pF in your load.

If you allow 20pF across each secondary - 10pF in parallel with the load your 0,25H tuning inductance has to drop to 0.18H which makes life easier.

It's actually a current snubber, not a voltage clamp.

Conventionally, for uncomplicated function, it's applied across the switch that is turning off. In Larkin's application, it will only function as well as layout strays and supply line decoupling allow.

RL

To an engineer it means, in this context, frequency. So to you it might mean something else in your fancy vocabulary.

An engineer wouldn't capitalise it. And they'd use radians per second - 2.pi.f - rather than frequency to get the numbers to come out right.

That's what I was considering doing, but still the individual transformer needs to be able to handle the 4.5kV isolation from winding to winding.

Multisection formers make that easy enough. A single layer of 25 micron kapton tape is good for 4kV, and several layers for rather more, so you can do it on a single section former.

Sloman isn't an engineer, he's a failed chemist. I don't know what symbol failed chemists use for frequency.

I'd respect his language skills more if he can ever learn the difference between

its

and

it's.

Yes, possible. Then a little dirt creeps in, poluted path, ionization, and sparks. Joerg would write "tssss, kaboom" etc ;-)

Another +1

Where is Joerg? I miss him.

John Larkin doesn't like coping with reality. Getting a Ph.D, in chemistry made me a successful chemist, not a failed one and getting a couple of patents as an electronic engineer is fairly convincing evidence that I did succeed as an engineer.

John Larkin clearly doesn't know what symbol electronic engineers use for frequency. In my experience they use "f" not "F" and in that context they'd mostly have used "2.pi..f".

A lower case omega would also work. Omega (uppercase Ω, lowercase ω) is the 24th and last letter of the Greek alphabet.

Sadly, it's (a contraction of it is) a distinction that is vulnerable to typos. Flyguy, James Arthur and John Larkin do tend to be rude when other people make them. It's (not possessive) not an attractive habit.