Hello,
Here is what I did today. I connected the output to the capacitor in the input section. Then I connected two amp meters. One just before the power supply negative terminal and the other behind the capacitor. This way I could measure the real amp draw of the motor and the reduced amp draw from the power supply. I connected my function generator to the optotrigger LEDs, so I could chop the pulses just as I want. Then I watched both amp meters and adjusted the frequency till I found the biggest difference in those readings. Of course, if the frequency is high, not much current can pass through the coils, I got a high pitch noise from the coils that varies according to the LED frequency and the rotor stays in the firing position. I found that if the square waves on from my function generator are with 50% duty cycle, I can get just about 50% recovery. If i increase the duty cycle to about 60-65%, I get about 60-65% recovery. But If I increase the duty cycle more, the amp draw gets just big enough to move the rotor in a nonfire position. So far the best result I had is 75% recovery. I did not write down at what frequency that was, I will do that later. All I wanted to know so far was how much can I get back. I think that 75% is a fairly good result.

Another interesting observation. If I put the rotor so that one phase is in the firing position and then just pulse the coil with high frequencies so that the rotor can't move because of too small current, the transistor of that phase gets almost freezing cold. I found this very interesting. This is the first time I actually observed this radiant cold effect. I have heard this may happen, but now I witnessed this myself

I will play more with frequencies and duty cycles and keep you informed.
Thanks,
Jetijs.