I got a Google Alert that pointed to a facebook page for a Joule Thief. In that page I found a link to this UK blog about using a Joule Thief as a battery charger (click on the link to the .PDF schematic).
The Joule Thief circuit is a typical example of a conventional JT. Instead of a 1k resistor, the author uses a 50k pot. He should also include a 1k resistor in series with the pot, so if he turns it to minimum it will prevent the transistor’s base current from becoming excessive and damaging the transistor.
The author winds a separate winding on the toroid core and connects this to the AC inputs of a full wave bridge rectifier (FWB). Then the rectified DC goes directly to a LED light (and also through a 10 ohm resistor). One characteristic of a full wave bridge is that it has two diodes in series with the current, so each time it rectifies there is 0.6 to 1 volt drop for each diode, or 1.2 to 2 volts drop total. The power wasted in the FWB is more than the power that is going to charge the battery. Combine this with the approximately 50 percent losses in the Joule Thief itself, and the charging battery gets less than 25 percent of the power that is drawn from the “run” battery. This is very wasteful.
Most people don’t remember that the LED is a light emitting DIODE. This means that it will also rectify the AC into DC, so actually the FWB is not needed at all. The only thing that one has to do is try connecting the winding to the LED first one way, then the other way (swap the wire ends). The pulses from the winding are not symmetrical and will light the LEDs brighter one way than the other. Leave the leads connected whichever way the LEDs are the brightest.
The idea of using the Joule Thief or a similar circuit called a Bedini SSG came from those experimenter who were trying to ‘harvest energy’ and make a circuit that has a a COP (coefficient of performance) of greater than unity. More can be read about this at overunity.com.