A lot of people have built the circuit in this Home Power article.
There's also an active discussion forum
In my opinion it's an elegant design, very creative, and maybe it could be made even better.
Here is an equivalent circuit using an n-fet and reversing the polarity overall (view in proportional font). Nothing new there.
D1 ,------+->|-+------, | | | | | ( ) | | L1 ( ) L2 | | ( ) | | ( ) ----- | | | --- | '--+-' ----- _| | --- --||_ ---- | | ---- C4 | | | | '---------+--------'
L1 and L2 are separate inductors. They are not linked. I arranged them in parallel to illustrate something: why not use magnetically coupled windings?
,------, ,------, | | | | | *( || )* | | ( || ) | | 1 ( || ) 2 | | ( || ) ----- | | | --- | '--+-' ----- _| | --- --||_ ---- | | ---- C4 | | | | '---------+--------'
The same pulse that appears at winding 1 when the mosfet turns off will appear at winding 2. This means you can pulse the battery without having to wait for a diode to come out of reverse bias and start conducting. (Of course, in a physical circuit you could have stray inductance and imperfect coupling between the windings so I might include D1, but I left it out of the drawing to get the idea across.)
This brings up another question about an effect that I think magnetically linked windings would have. Would coupling the windings with the indicated polarity reduce ripple on the capacitor, as compared to the original circuit?
I think that when the mosfet is on, current is ramping up in winding 1, and the "back emf" in winding 1 that opposes the impressed voltage will be reflected into winding 2 and push current into the capacitor node at the same time winding 1 is drawing circuit from that node. Is my analysis correct?