Just by lowering current sense resistor value. If you remove the driver it has 2 big resistors with R020 written on them. So 10mOhm for 18A. From there you can calculate sensing voltage. Knowing the voltage you just change the amperage to your desired value in U=RI equation.
U=10mOhm*18A-> U=180mV.
R=U/I-> R=180mV/25A-> R=7.2mOhm.
Edit: I have tested it to function with 25A, but there’s always risk for something to break.
Thank you very much ! I have removed the switch and 2 screws but the driver still too tight can not get it out , I missed something ?
Wires to led are holding it in place. And maybe to switch also.
I removed all wires to led , switch , 2 screws but the driver still not move . Can you remember how to get it out ?
I used NCR18650GA’s for my testing but I think your explanation is correct. As the cells are depleted, the voltage will sag lower, and eventually it sags lower than the LED’s Vf so you cannot get full brightness.
That’s been my experience with all the SBT90.2 lights I’ve tested that have one cell, or multiple cells in parallel. I think SBT90.2 is best driven my multiple cells in series with a high current buck driver like in Convoy L7.
It’s probably just little stuck. There wasn’t any glue in mine. Try to wiggle it little from switch hole and usb-c port.
I couldn't agree more!
The L8 is definitely my favorite single cell sbt90 Light. I like everything about it, but the biggest drawback is the limited compatibility of a single 26800 / 26900. They're just not great cells for high amp sbt90 lights. I'd much rather the L8 head / Reflector with single 21700, 21700 in series or even 26800 in series if need be, and the appropriate buck driver of course (Side switch only). But, i still use the L8 daily and own two of them. One for each hand, of course.
Or better yet, just use higher power cells like the Samsung 30Q or Molicel P28A.
The NCR18650GA is a 5-8A cell at best honestly, and going at 10A isn’t the best idea to get out most capacity out a high power LED.
As someone mentioned above, at 18A, it should be pulling about 4.5A per cell, which may cause sag in an NCR18650GA, but should be well within its capabilities. Probably less at most voltages since it’s a buck converter and the vF is close to 3v, so at higher cell voltages, it should be pulling less than 4.5A for the driver to deliver 18A @ 3v.
Even at 25A it’s only 6.25A per battery, which should also be ok for Sanyos. Sbt90.2 is about 3.6V on 25A if extrapolating from diagram. And GA’s drop to 3,8V at 7A current draw.
4X18A’s driver is linear, not buck
Also, max achievable turbo brightness drops smoothly as the batteries drain so it makes sense that voltage sag is the cause.
Happy Birthday Simon Mao......Wishing you the biggest birthday celebration today, good health and more flashlights to build.
Happy Birthday Simon!
Have a great day, cheers!
Happy Birthday, Simon! 
Happy birthday Simon! Thanks for your continued interaction with us. It is greatly appreciated.
I didn’t know the regulator was linear, touché. Still, worst case is 4 - 6A draw per cell. Voltage sag may still be the culprit, but well within the cells’ capabilities. Only way to know for sure would be to test with different cells and compare I suppose.
About 70~80mV higher with 30Q at 5A at high SOC, not a big difference but it can help a bit.
More importantly the 180mV drop across the sense resistor doesn’t help.
With 30Qs and a more reasonable 50mV Vsense, 0.2V would be saved, that’s significant.
7135*4 is out of stock
Although today is not my birthday, I still want to thank you all. :partying_face:
There’s an interesting thing in there. China has a lunar calendar and a solar calendar. Actually my birthday is counted according to the lunar calendar. 
When will the 23.2*1.4mm ar-coated glass lens ,suitable for Convoy S21A S21B be available with the new green AR coating?