I'm glad people are finding this interesting/helpful.
Thank you all for your kind words.
To reply to questions so far:
Barkuti:
Any 2.5V SOT-23 size regulator would work, as long as it has the same pinout. I took one from an LD-25 driver I had spare, but there are plenty on Mouser/DigiKey/etc. No other passives required.
Whatever the specced min voltage is for your cells. It used to be 3V, but some manufacturers are going down to 2.7/2.8 nowadays.
The LVP is taken from after the regulator (no idea why), so calculating the required value of R2/R3 gets a bit weird. It might not even need changing, as it's dependent on regulator dropout.
I also don't know how the MCU will behave at 2.5V. It should be OK if LVP is adjusted, but I have since removed mine, so cannot easily test.
Input currents will remain limited to ~20A. You need to work out the input/output voltages, calculate duty cycle, and factor in efficiency.
See THIS Application Note by TI. You mostly need Page 6. At a complete guess, subtract 1-2W from my maximum for each volt out above 6V, But I'd strongly advise you to do the calculations.
Replace R16. I'm using a 220k, but you may need a different value for your emitter(s). You need a value which puts the MP3431 output voltage at half your emitter Vdrop range.
Check LED test graphs for Vdrop range of your LED up to your desired max current, use MP3431 datasheet equation to find R16. Do not change R17 or R9, you risk breaking the op-amp feedback loop.
HAHA not well. It won't be a switch for very long. You could increase the soft-start (C12) to reduce startup inrush-current, but that's not really solving the problem.
You'd need a FET switch for rear-clicky lights, or modify it to use a side switch. I'm an E-switch guy myself, so I don't have that problem...
id30209:
Change sense resistor (R10) to lower value. Please note that XHP35 Vf = 14.2V at 2.5A. That = 35.5W out, which the 10A input current limit will curtail as the battery voltage falls below ~3.5V.
To do full modification, you need to do what I did for my test (requires good SMD soldering skills).
If you are driving an XHP35, there's not much point, as over 2.5A will kill it pretty quickly.
Yokiamy:
Where does it say that? The MP3431 has internal over-temp shutdown as a last resort, but the MCU doesn't have any external components for it like with some BLF drivers.
It may be internal to the MCU, similar to Narsil/Anduril, but I have no way to check it.
Similar to the previous reply, the XHP35 requires 14.2V at 2.5A. That = 35.5W out, so your test gives 100*(1-((40-35.5)/40)) = 88.75% efficiency, which is expected.
EasyB:
Thank you!
It should be mentioned that at 10W loss, only ~3.5W is dissipated by the MP3431 itself (conduction/switching). The remaining ~6.5W is lost to parasitics in the input current path (PCB traces, inductor, vias, etc).
The inductor DCR is approx. 3mR (VERY rough measurement), giving ~1.2W loss, though I strongly suspect that it is beginning to saturate at 20A.
I will try to find a better inductor to use, and will try some serious modification of the PCB to improve current handling.
I am designing a custom boost driver for a different project, and this driver works well as a testing platform for different ideas.
The testing I did was with the driver in open air, so I kept high mode test runs to 5-10 seconds.
Mounting it properly in a flashlight body would solve that issue, but would make it much more difficult to test and modify the driver.
My future modifications should hopefully improve thermals. The 3.5W I can tolerate, the other 6.5W (that I can do something about) need to go .