I don’t know exactly which Nichia 219B is in use, but generally speaking they are not well suited to a poky direct-drive FET with a strong cell.
Mis-matched, and needs to be throttled back.
A linear or buck driver, even as simple as a x8 7135, or similar, would be a better match.
E.g. see Nichia 219B 4500k D220 9080 CRI Emitter output test by Texas_Ace; Amazing tint and CRI! Output is what you would expect. and study the curves.
I.e. maximum output achieved at only 3.30 Vf, after that it’s all downhill.
Edit: and increasingly inefficient as that voltage, current is approached. Then falls off a cliff afterwards.
To use these Nichias well, you need multiples, as in this one. But they also have to be driven well, with some sanity about what is possible and appropriate. Which I don’t see happening here.
Goodness knows what will happen with a strong cell at 4.2V. I think FF have already given a clue.
I.e. don’t even think about trying it.
Entirely predictable.
Sadly, just throttling back by limiting the PWM of the FET won’t alter anything, except thermal dissipation. The LED will still be being hit instantaneously with massive over-current into a deeply inefficient and stressful region of far lower output.
Maybe put in some really light gauge LED wires to try to limit peak current, if you can take it apart.
Keep it in the x3 7135 region only and it will be safe, but I don’t suppose you bought this multi-emitter torch to just tick-over at x3 7135 levels.
Well, that’s my theory, based on some actual facts and data. Once you have them in your hands, we’ll begin to know, I suspect the “Maxx Power at any cost” peeps will be the earliest to report on the 219B choice.
By the way TK, is it realistic to consider re-flashing this torch with your updated code ? It’s just that I got the impression that FF like to seal these up as tight as a very tight thing.