[Diabolical Artificer]

I use a 10k 10 turn pot to find the right R FB value; connect between IP & OPT, set at 10k, start lowering until the amp starts to get unstable, then back off a bit. To find the C, get a capacitance decade box,IP a 1khz square wave to the amp, monitor OP on a scope, play with values till you get a nice sq wave, then do the same for 10khz. This is the simple procedure.

To sort an amp properly you need to do lots of sweeps from 20hz (lower if you can) - 100khz at various OP levels with a resistive load, then a capacitive load.This reveals it's instability points, what your looking for is where the OP peaks or where THD shoots up. When you look at your results you'll see something like "peaks at 48.3khz & at 58khz" or peaks at 10hz & 42khz. You then select your FB RC values to form a HPF at the problem frequencies using 1/(6.28 * f * r or c). It's also a good idea to pop a Zobel or Bucheret network on the OPT secondary using the same data & method. What your doing is putting RC filters in various places of the amp to stomp on any instability or silly behaviour. This is why you'll see an RC in parallel with the IP stage anode resistor or small value (picofarad) caps in parallel with other parts of the circuit in an amp,broadly speaking that is.

This results in a stable amp, stable with all loads. It can be very time consuming but it's worth it.

Andy.