I also think WaveRider's circuit worked well, and probably gets near to Class-C by just overdriving the gates, as explained by BP's diagram in the thread. Also, if you add a DC link choke and tuning capacitor, it should oscillate by itself just like the ZVS does.

I wondered about other ways I could take feedback in this circuit apart from diodes.
How would a current transformer inside the LC behave? (I need to take voltage, not current feedback).

I also noticed you used your royer without primary capacitor, which I found very destructive for mosfet's. I obviously want feedback from primary, not the secondary LC circuit.

Now I wonder, how do things look if I use primary cap, hanging coupling loops for feedback and a high-Q resonator (Tesla secondary), what am I really taking feedback from?

Coupling loops are poorly coupled to both resonators and I would expect some heterodyning mess or just cessation of oscillation. Can you clarify that a bit?

I relied on the secondary resonance to set the resonant frequency. However, as you noticed, improvement could be gained by adding the capacitor... This is because it may be compensating for phase errors that reduce efficient power transfer.....in short, you are getting a better match to the TC load (the arc). I have not checked this, so I could be
blowing hot air out my arse on this one! Certainly the shunt cap on the primary allows for phase correction on my Class-E coil..

You need to "sample" the secondary coil (resonator) field, which should be dominant near the so-called "parallel" resonance (the primary input impedance is at its highest point, like a parallel LCR circuit). These feedback loops should be small (weakly coupled) to the resonator field. Hence, they do not disturb the resonance significantly.

The drain-source capacitance of the MOSFETS that I was using was quite high... Hence VDS punch-through did not seem to be a problem . Even with capacitors across the primary, power will be dumped into the MOSFETS...unless you can switch the transistor at zero voltage, shunt caps can make the situation even worse. This requires precise phasing. Shunt caps can make things better if used with care.

I found that a small feedback "tickler loop" weakly coupled to the resonator is enough to cause stable oscillation to occur. If there is no breakout, the system Q can be quite high (depending on coupling to the amplifier or the "external world") and oscillation is easily achieved.

That's exactly what I *don't* want.. I want the thing to absolutely remain on primary resonance and in ZVS.

What assures your coupling loop will 'like' one LC more than another?

o the best of my knowledge, a set of weakly coupled transmitter and receiver resonant circuits. TThis implies relative high Q transmitter resonator.. ZVS seems to work best in tightly coupled, low-Q situations.

You are getting high values for VDS because this type of system is intrinsically hi-Q.

Your feedback circuits should not be resonant near the operating frequency. If they are, funny things will no doubt occur. The feedback coil(s) should be as far below their stand-alone (uncoupled) resonance as possible.

Of course, if coupling between the driver and driven coils is significant, your oscillator may hop between the two resonances caused by the coupling induced frequency splitting effect depending on the coupling and resonant frequency of the coils.