The following discussion is in regards to the schematic in my Supercharged Joule Thief blog dated 2011-12-04. Flagiusz had some different ideas as to how it works. I would like to say that I totally respect his opinions, but I have no way of offering a firm explanation of how it works because I have not taken the time to do any analysis of the circuit, only my observations during my experiments with more than a dozen that I’ve built (I can only say that the circuits all work consistently and very well, and have no quirks that would cause it to give unusual results). The measurements (taken with a luxmeter) that I included in the schematic show that it is much more efficient than a conventional JT.
Some points in his opinion that I would like to discuss follow.
Flagiusz believes that C2 speeds up the switching of the B to E junction.
The C2 value was chosen because it seems to be a good optimum point. If I went below 500 pF or over 1000 pF, the output would drop. Flagiusz believes that it should be close to the B to E junction capacitance. As far as I know, most small signal transistors have a junction capacitance somewhere around 10 pF. The value of C2 is between 50 and 100 times that amount, so it is much greater than the junction capacitance.
As I said above, if I reduce it to only 10 pF, the LED’s light output would drop to a much lower value. So I think there is much more to this than speeding up the switching of the B to E junction.
When the switch is switched to the conventional JT, the switching rate might be around 80 kHz. When the switch is switched to the Supercharged JT, the switching rate might be 240 kHz, or three times as fast. I don’t think that if a small capacitor was speeding up the B to E junction switching time, it would make this great a difference in the switching rate.
I think (but haven’t confirmed) that the C2 capacitor interacts with the feedback winding to become a series tuned circuit. Series tuned circuits have the broad peak like the Supercharged JT has.
I have confirmed what others have said about rectifying and filtering the JT output before it gets to the LED. This makes the LED brighter with no increase of battery current. But there is a half volt or more drop across the rectifying diode, and this wastes several tens of milliwatts. When designing the Supercharged JT, my intention was to eliminate this high LED current from the rectifying diode, so the power was not wasted. The Supercharged JT accomplishes this very effectively.
I have not tried Flagiusz’ suggestion to use the conventional JT with the C2 similarly connected. So I do not know if it works better, worse or equally well compared the Supercharged JT.