I must admit that my question sounds purposely a little provocative.

But I accidentally came across the circuit described below, which exclusively works in the linear part of its operational range (see the attached circuit).

The first block represents an active inverting first-order lowpass filter, where the summation can be done directly at the input node of the used opamp. In this case, the second block must be implemented as a non-inverting integrator circuit (Hint: Both polarities can also be swapped, thereby allowing the convenient MILLER integrator).

The closed feedback loop then represents a second-order lowpass function with a high pole quality factor (e.g. Qp=10).

Once triggered, this high-Q lowpass would react with a decaying oscillation - unless it receives new "kick impulses" at the „right“ time slot. Therefore, at each zero crossing of the lowpass output voltage v(out1) the comparator adds a voltage Vc to the feedback signal v(out2) for half a period with:

Vc negative for v(out1)>0 and vice versa.

(Now the analogy to the pendulum clock becomes obvious) .

Demonstration example:

Lowpass: H(s)= -10/(1+0.01*s),

Integrator: H(s)=+1/0.001*s,

Closed-loop function (2nd order): Ao=1, pole frequncy 160 Hz, Qp=10

Comparator output: Vc=(+-)1V.

The result is a sinusoidal output signal (frequency fo=160 Hz) with very good quality: Amplitude Vmax(out,2)=12.5 V (supply voltages 20V).

It is to be mentioned that the observed frequency and amplitude are in full accordance with theoretical considerations (calculation).

The frequency is set by the time constants of low pass and integrator.

The oscillation amplitude can be varied over a large range (without changing the frequency) by the comparator voltage Vc, but it also depends on the lowpass gain Ao and the closed-loop quality factor Qp (which determines the decaying time constant).

Final comment: I think, the shown circuit belongs to the class of oscillators which are stabilized by „restoring initial conditions“ (see IEE papers from Filanovski) - however, this circuit is much simpler in design compared to other related oscillator configurations.

Final question: Does the output signal contain any intentional systematic non-linearities?

Sorry - but I couldn`t manage to make the figure smaller.