Run 1, 8 Dec. Driver update
As i promised yesterday here are some details about the joule thief driver.
The schematic is still the same with an added X7R 10uF decoupling capacitor at the input. I am not sure if it's really required here but i am used to always have at least some decoupling and it might help when the cell internal resistance increase.
My goal for this driver was :
- output at least 5 lm at 1.5V
- still about 3 lm at 1.0V
- at least 1 lm down to 0.7V
- overall efficiency better than 65 lm/W at maximum output, going higher for lower output
These goals were copied from the Sofirn C01 3200K which was my reference while designing this little light. The C01 driver use an integrated circuit with output regulation and the single transistor JT can't really compete with that but i tried my best and here are the results :
Light output (lm) vs input voltage (V) curve :

Efficiency (lm/W) vs light output (lm) curve :

The green curves (JT 1) are from the first JT prorotype.
The orange curves (JT 2) are from the final JT prorotype.
The blue curves (C01) is from the Sofirn C01 3200K which was my reference while designing this little light.
As you can see, working on the JT allowed me to increase both of the light output and efficiency compared to the first prototype but there is a limit after which i wasn't able to increase efficiency without decreasing light output or vice-versa so i choose to keep the output.
Conclusion :
As can be seen on the curves the ouptut light goals are totally met and while i would have liked the efficiency to be a bit more on par with the C01 i am still very happy with the (about) 10% difference considering the very simple schematic !