Negative discharge effect

[TinselKoala]

OK, I have built the first circuit (with a few substitutions that shouldn't matter much.) IRF630 mosfet, 1n2071 rectifier diodes, 160 uF 330V fhotoflash caps (Thanks Pirate!).
I'll give a better report a little later. Scope below shows the junction of the two caps and the coil, wrt mosfet source.

[ayeaye]

OK, I have built the first circuit

Oh great, thanks

So what it shows, a 3 V voltage spikes when the gate is switched on. These are likely higher, the scope may not show the full height. What is weird is the constant coil ringing, maybe due to a slow zener diode of the mosfet. But seemingly no back-emf. As i have seen, the effect occurs only with a certain frequency and duty cycle, which may even be different with the same components. And only then there is a significant back-emf.

Oscilloscope, great thing. I cannot get one, the old oscilloscopes they sell here seem to be only the ones which are so hopelessly failed, that no one can repair them. I even tried to make an ascii oscilloscope with my microcontroller, wrote the code, and even tested it. 500 ksps and 16 bits is the best one can get with that anyhow. Because of high voltage spikes it cannot be made without an op amp. And even with an op amp there are too many interferences, so who knows what else should be made to protect against these. It's like stable, but then it jumps to some who knows what value, then jumps back. So i found it too difficult, and not good enough.

[TinselKoala]

No, I could find no trace of a narrow, higher voltage spike. This surprised me because I expected it to be there.  I'm using a yoke that I wound for another purpose, it doesn't have quite 900 turns on it. I tried another coil of around 3 H and it behaved almost the same except with a much lower ring frequency at the TP1. Still no higher voltage, narrower spike showed up. So today I'll wind a proper 900 turn yoke (I only have #27 wire though).

I deliberately used the slowest diodes in my box. I am going to start changing diodes to see if faster diodes make any difference. I have a bunch of diodes from TV and monitor chassis but I don't recall seeing your exact diode in there, I'll have to look again. I'll also change to a faster scope, in case the 60 MHz Tek 2213a is missing something.

I know money is tight these days but you really need an oscilloscope for this kind of work. You should be able to get a decent analog scope in the 60 MHz bandwidth range for 200 dollars US or so on Ebay, and I have recommended the Hantek PC-based DSO several times as a reasonable choice for a "first scope", as it is only about 100 dollars or even less and comes with 2 decent probes.

[TinselKoala]

OK, I've wound a proper yoke with 900 turns of #27 magnet wire. I wound it all on half the yoke using my recently-built coil winder. It makes quite a lump on the yoke half!

Then I reassembled the yoke and measured its inductance on the ProsKit MT-5210 RLC meter. It measures 600 milliHenry. Is it possible that you misplaced a decimal point when you calculated the inductance using your measurement method? (That method is valid, I have checked it myself and when it's done properly it produces a good answer.)

If I am understanding your negative discharge effect properly I am able to reproduce it with this setup. We are in the non-linear behaviour region of the mosfet for sure! The oscilloscope reveals so much interesting behaviour that I am going to have to make a video to show it all. I'll be putting that together over the next little while.

I'm using the second circuit, as below.

[TinselKoala]

The IRF630 data sheet from Vishay has the characteristics of the body diode.
At the moment I am tending to believe the following: The device might actually be picking up power from the house mains! Here's what I'm seeing. The 3.2V signal to the gate is indeed critical. If I go a little bit above this value the voltage does not accumulate on the capacitor and the signs of line power pickup on the scope decrease. If I go a little bit below this value the mosfet isn't stirred up enough to allow the voltage to accumulate on the cap. Pulse width is also critical but the effect is harder to explain, I'll just have to show it. The pulse width determines the shape of the spike response and if it's too wide the peak spike droops and the rate of voltage accumulation on the cap decreases or stops. I think that the coil acts as the pickup for the line power and the mosfet in its just-barely-turned on state allows the negative discharge effect to happen on the capacitor.
Maybe. I'm not sure about any explanations yet. I am sure that at low Gate drive voltages, right around and below the 3.2 V value, I definitely am picking up power from the mains somehow. I don't know if this is what is being stored on the capacitor though.