I may be a bit off topic... I took a look at aboveunity.com. Jagau shows us things that may seem surprising but remain conventional, the models giving the same results as his observations.

Moreover, one must be careful with his assertions, for example at the end of this page

https://www.aboveunity.com/thread/capacitor-recharging/, he says "the oscillation that we see at the end is caused by the spread capacitance of the inductors" whereas a model of the device shows the same oscillation without any spread capacitance for the inductor, the oscillation comes from the classical LC circuit with discrete elements.

It is a pity that he romanticizes his speech with unrealistic hypotheses, but it is still interesting because his basic setups remind us of fundamental things that we ignore or forget. Wanting to model the device that we see here in this video of Chris pointed out by Jagau (these 2 people seem close to each other in their way of thinking, maybe the same?):

https://youtu.be/-IE_UZtKr-I?t=2311, I didn't find anything particular, but by replacing the diodes by capacitors, I found a behavior that I had forgotten, that is that you can obtain a resonance by using a fictitious inductance.

In the first diagram (see attached file), each of the 2 inductances L1 and L2, strongly coupled to each other (coefficient 0.9), is coupled with a coefficient 0.5 to the inductance L0 connected to the generator. We see in the AC analysis the resonance around 13.5 KHz.

If, without changing anything in the setup, we create an imbalance of the coefficients of mutual inductance by taking 0.45 for L1/L0 while we keep 0.5 for L2/L0, we see that we obtain a second resonance.

The resonance around 13 KHz of each circuit is maintained, but there is in addition a resonance linked to the common circuit L1+L2, around 51 KHz. In fact L1 and L2 being in opposition, it is as if we had a lower inductance (up to theoretically zero, which means that in the 1st case, we do not see the second resonance). This is an old technique used for example for antenna tuning (see

http://w5jgv.com/11.7uHy_Delta_Variometer/).

We therefore understand that it becomes very complex to analyze setups like that of Kapanadze, where the degree of coupling of each coil to each other will generate multiple resonance frequencies that are difficult to control in practice. But on the other hand, we see that with these coupling coefficients, we have a large degree of freedom to vary inductors, for example to produce a parametric device with a mechanical or permeability vibration which would vary the coefficient of mutual inductance, idea on which I am at the moment.