I realized that spark gap parameters are critical too in the previously linked simulations in that they affect significantly the results, but I'm not sure what would be the best to use. One could make several assumptions:
Only arc discharges are produced, which have a very low resistance.
Since the arc-discharges are basically produced under contact separation, the breakdown voltage should be very low, somewhat above the power supply voltage.
The holding current should probably be rather high, as this is far from being a vacuum environment and the process will be quickly disturbed.
I spent some time with the simulator last night, exploring the effect of the component variables on the circuit behavior. Unfortunately, I haven't yet figured out how to get a functioning link to the modified circuit for posting. I did find the conditions needed to enable the system to oscillate.There's a screen capture below for reference.
Here are my conclusions so far:
* Resonant behavior is suppressed by the 15 ohm shunt resistor, which represents conduction through the electrolyte. Resonance appears when this value is >7k ohms. I think the growth of an insulating oxide layer on the electrodes is therefore needed before the oscillation can appear.
* The spark gap breakdown voltage needs to be a bit less than twice the supply voltage for the oscillation to self-start. I used 22 V in the simulation below. The cap voltage rises well above that as the oscillation continues. I don't think this detail is accurate, since there is no diode in the model to enable charge pumping.
* The series resistance of the inductor should be smaller, perhaps 0.2 ohm, to accurately represent the copper wire actually used.
* The external pulse train and FET switch do not accurately model the dynamic behavior of the real-world system (I have removed them from my model). The output "signal" of that circuit simply follows the input pulse train. Substituting low-pass filtered noise for the square-wave also shows this behavior, though it is closer to the real-world system. There was still no sign of circuit resonance when driven that way.
I agree with can that the usefulness of this tool is in its ease of use and quick iteration. Accuracy TBD...