The OPT's do not have 16ohmm output taps, so the C12/13 network will have to adjusted. I have 4 ohm and 8 ohm though. Do you have any idea how to adjust R22 and R23 accordingly?

Karel,

Apology! I seem to have missed this question.

The winding ratio between 16 ohm and 8 ohm is 1.414, thus R22 + R23 will have to be decreased to 70,7%, giving 4,2K total resistance. If one stays with the feedback via C12 to the 4 ohm output winding, then R23 becomes about 2,2K, with R22 remaining at 3K. (R23 now connected to your 8 ohm output.)

But the values of C12, C13 are entirely a function of the OPT characteristics. They work in the supersonic region, to compensate for OPT internal capacitance and leakage reactance (thus phase correctors.) Their effect for that particular OPT is somewhere >60 KHz and certainly cannot be trimmed by hearing. Thus I cannot recommend values for your OPT; to get those right one will need a square wave generator and oscilloscope to check.

Also the necessity of C12 might indicate some inequality between secondary sections of the particular OPT; ??? I have always just used a single RC feedback combination from the whole secondary, but that is with using all secondary windings for both 8 ohm and 4 ohm outputs (for three secondary sections the usual 30% - 40% - 30% secondary winding segmentation).

P.S: Despite Dr Gillespie's remark that his design is stable, I would still add a 680 ohm 5W resistor over the secondaries just to assure stability without a load. (I do this in all my valve designs. Stray capacitance for point-to-point wiring will always differ from model to model.)