This is the sort of confusion I'm talking about, some see them as noisy (square wave push-pull), others clean (resonant, with a Cap between collectors/drains). I was thinking more of the latter, and with reasonable filtering on the output, the PSRR of opamps, and perhaps even linear regulators should make it fairly low noise.
As mentioned, my main concern is high current on startup, or perhaps inability to get the thing started. Again as mentioned, there are better ideas for this application but I'm doing this as a challenge and investigation. My main reason for going with MOSFET's isn't really for efficiency, but to negate the need for a feedback winding. I'm winding my own transformers anyway, so an extra winding really isn't much of a hassle, but the ZVS-royer-like-circuits seem much simpler.
Capt Bullshot: thanks for the links. I see the design in pulser.pdf just has one primary, but with two feed chokes. Thats probably the simplest design I've seen! I noticed the chokes are pretty high value. In my LTspice simulations, the higher the inductance, the lower the current spike at the start (makes sense). Most sources indicate a minimum of 4x the primary winding value, I used 10x. The isolated 12V supplies also look interesting - I have been meaning to make a fully isolated set of 9V 100mA isolated supplies for guitar pedals and that looks ideal!
Berni, thats what I'm cautious of, component selection. Generally I use off-the-shelf solutions where component selection isn't that critical (as long as the inductor has the right inductance, DCR and saturation current..) but with self-oscillating ones, the transformer specs can make or break it. I also have that Linear Tech app note on my desktop. It is partly what started this.
High efficiency isn't really required, although of course it would be nice to keep it above 70% purely to reduce the need for heatsinks. I'll get cracking with LTspice before I knock up something real. Cheers folks!