I am aware that linear oscillators must contain a non-linear part to limit loop gain (or amplitudes) after the start-up phase. Simple case: Amplitude clipping due to supply rails.
Then, using the appropriate describing function technique for sinusoidal signals, it is possible to get some information about:
* the possible existence of self-sustained oscillations, as well as
* the oscillating signal (amplitude and frequency) .
However, the following restriction exists:
The frequency dependent part of the circuit must exhibit a lowpass or bandpass response (with sufficient harmonics damping).
Thus my question is the following:
If this pre-condition is not fulfilled (harmonics damping) - e.g. for allpass oscillators, is it possible, nevertheless, to gain some information about oscillator properties by applying the technique of describing functions (as a minimum, an answer: oscillation yes/no)?