Bob Higgins wrote: In such cases, it is really useful to simulate the system with a model that is entirely without unknown physics and see how the model compares with observation. If it predicts the same phenomena, you can be pretty sure that the outcome was simply outside your expectation. SPICE is a wonderful first-principles tool for a lot of this with wires and magnets.
Yes, it is a great tool, but the previously unknown parameter or exotic material is the problem. Plus– SPICE, or any simulation, can be fooled by an incorrect assumption – as here. http://overunity.com/7403/this-ltspice-simulation-model-will-blow-your-mind/#.V4u7yDVgs9M SPICE works only on the known, in fact only on the well-known - and is specifically poor at modeling leakage inductance… which is a factor in EE that normally is not a good thing. Perhaps, like lemons, it can be made into lemonade (if that is what Graham is doing) but not modeled. Driving a specialty magnetic core past saturation presents problems for any model when other cores are in spatial proximity. The one-and-only acceptable recourse here, as always, is to “close the loop” in some way. If a circuit self-resonates for hundreds of hours with only microfarads of capacitance, and especially if substantial cooling of a magnetic core is seen during this time – you can throw the SPICE model out the window, until of course it is modified… but that always happens after the fact. Do you have theoretical problems (aside from natural suspicions) with the presentation of a self-powering circuit, driving a computer fan over a cold core and thus cooling a space, such that the net thermal balance is increased in entropy?