Joule thief efficiency evaluations

[schmitt trigger]

A decade ago (yes it has been that long) I became interested in joule thieves, and their outstanding capabilities to light up LED from very low voltages.
Although the basic circuit works fine, it can be significantly improved.

To satisfy my curiosity I experimented a little, recorded my results, and wrote up a short article, to share in an online forum. Which I no longer remember which was. And I am not sure I ever shared.

Cleaning old files from my computer, I found this old file.
I am attaching it here, only FWIW and to share with whomever is interested in these circuits..

 

[betwixt]

Thank you for sharing.

I did some similar work to 'get something for nothing' when my better half wanted 'under cabinet' kitchen lights. I had a faulty 6W LED light bulb of the kind that uses thin LED candles. These candles are about 2mm diameter and 30mm long and the bulb had them wired as two series sets of four in parallel. In the original bulb, one of the candles was flashing on and off which resulted in the remaining ones flickering as well due to the change in current. I carefully broke the glass envelope and snipped out the good candles, they are incredibly fragile and it seems the failure mode is buckling as they warm up. Each candle contains about 50 tiny LEDs and the whole thing is covered in fluorescent gel.

Our mains electricity here often fails so I hit on the idea of using a battery supply so they could still illuminate the worktops in a blackout. The problem was the candles need around 65V to operate but thankfully only at about 2mA current. A high voltage battery supply was out of the question so I tried the Joule Thief circuit to see if it could produce enough voltage, the result was very disappointing, it did work but the LED was dim and efficiency (LED power/ battery consumption) was only about 20%. I experimented in many ways as you did but by far the most efficient was to abandon the simplicity and use an NE555 driving a MOSFET. The output side was the same as the Joule Thief but the drive signal was a square wave at about 50KHz. From a small 6V battery it drives five candles in parallel, each with a 1K series resistor and achieves around 95% efficiency! Even simply replacing the bipolar transistor with an enhancement MOSFET made a huge improvement.

Brian.

 

[schmitt trigger]

Something that I've noticed for the JT is that it works best with low voltages V< 3 volt

I am sure that experimenting with the winding ratio could somehow optimize the design for higher voltages, but I feel that you did the right thing: an oscillator driven Mosfet