I did some practical tests, changed the circuit so it leaks away the gate voltage, but I could really make it lock up, so the MOSFET was always on. Does there exist a latchup effect for MOSFETS? Because when connected as here: ftp://panteltje.com/pub/power_pic/sepic_charging_self_img_1002.jpg an interesting test circuit, the SEPIC converter feeds back into the battery that powers it, via a car headlight (12V / 55W), so I can, by turning up the output voltage of the SEPIC, get a positive current back into the battery, or by turning it down, the battery feeds the light bulb directly, strange, but that is what it does! This is because the MOSFET is then still 'on' some of the time, and then a current can flow in reverse direction (so actually where the drain is positive, normally the right direction LOL), BUT when I remove the drive pulse, sometimes the MOSFET seems to say on, keeping a small gate voltage (a few volts), refusing to discharge the gate through the 4k7, and lighting up the light bulb. This I do not understand, maybe the MOSFET is damaged, but after removing all power, everything works normally again on power up.
For those who think this self-charging circuit is the answer to the quest for an unlimited energy supply, well, it puts a current back in the battery, but it draws more from the battery itself, so the difference says something about the efficiency, and that will never be greater then 100 % :-) But sure you could patent it, and try to get some investor money :-) Risk capital is hard to come by these days though...