This is the Class E build I made using a 1200V SiC so I have a lot more room for more voltage but I'll probably change it for a 650V version and use the 1200V on a different circuit that has high side isolation. It could probably do 100V or so before reaching about 650V drain pulses.
I'm just showing how the waveform is changing given particular voltage inputs and tuning, and how the highest resonance that equals the highest plasma discharge is drastically changing the loading in the coil at the time.
When I first tried tuning these I was doing it at low test voltages like 20-24V max not wanting to destroy the mosfet from bad switching spikes. Eventually I realized that the more power I tried getting out of it the more I'd have to change the tuning. There is a certain frequency I can set which tunes with the LC that allows for pretty good switching through a fairly large range of voltages. That frequency however is not at the coil's resonance until it has a certain amount of plasma emitting from the topload. That means at lower voltages the lower resonance is outputting less plasma than it possibly could at the time, and could also be switching poorly. This is only a problem here because you're trying to tune the LC to try to ring at the right frequency relative to that plasma loading on the secondary LC.
With the self oscillating version you have basically the same issue to where it's always self resonant even when loaded, but this resonance is going to drop in frequency the more it's loaded, so something like a 24V tuning should have it switching harder to account for the loading at higher voltage.
![](https://s2.save4k.ru/pic/7g7GWyaRmTo/maxresdefault.jpg)