Stacking Ultra-Fast Diodes Considerations 3KV 25nS 0.02A

I would stack them twice as high as I need. That way I can have 50% over-shoot and still clamp down in reverse. Diodes should not care much if the stack is operating at half the reverse voltage than the stack can handle, and as a bonus, you get the best performance possible with the diodes in the stack that are faster than the others. Series is series even if half the diodes in your stack don't switch, you still have enough of them switching to do the job.

The only drawback is that the taller the stack, the more you get a cascade effect during the clamping phase. This will widen the switching time for the whole stack and it may introduce high frequency noise into the circuit that you do not want. On a scope this will look like stepping from the peak to the trough or visa-versa, and you may also see a ringging up or down. It's like closing a set of shutters instead of a single door.

First thing to do is to test the reverse leakage current vs voltage for the desired level of leakage current. Now filter and select all those fit into specifications.

Putting them in series and parallel is just fine. Once you make an assembly then once again test its leakage current vs reverse voltage to see if it meets your requirement.

Watch out for the forward current limit or else use a series resistance to limit the forward current within the specifications of the diode.

Have you decided that you can't use common low-cost high-voltage diodes, like Vishay's GP02-40, rated at 4kV, only $0.25, qty 100 at Digi-Key? Do you have some special requirements?

BTW, 500kHz or even 250kHz seems an awfully high frequency for a 1.1kV dc supply. That's because P = f C V^2 gets pretty painful when V is higher than 1kV, usually a lower f is used (an exception would be a resonant system).

As far as avalanche is concerned, it's generally not that scary. When you examine official avalanche ratings in detail you'll see they almost always simply derive from the full thermal power ratings of the device. And pulse avalanche ratings derive from the thermal mass of the semiconductor and associated metal structure and leads, as the heat spreads from the junction. These capabilities are almost certainly far far higher than the stored-charge energy in the reverse-recovery-time of your diodes.

Finally I will run at 250KHz with an EP13 core (14x14x14mm). Size and cost are critical since Xfo will fit in the handle of a disposable surgical tool. I use discontinuous mode operation with an oversize core to minimize number of turns and hence capacitance. I found an SMT diode at 1.75$ and a TroughHole at 0.95$ that will do the job for now but for the volume production I will stack 3 to 5 cheaper 600-800V diodes based on your advice. You confirmed my belief that all diodes have an unstated avalanche and zener rating. I can not consider matching the diodes since it would be more expensive than using a single higher cost one. Many thanks for your feedback and also thanks to all.