And here ya go…
2 Thanks
You could slice some/all of the xhp50.2s if you’re trying to increase throw. I sliced an xhp70.2 in a convoy m21c-u and got great results. Also I wouldn’t worry too much about the textured reflector, I believe they only reduce throw by a few percent.
I really don’t care very much for the beam color mix from Cree flip chip emitters. (Yes, I know, dice the phosphor around the die for a cleaner output. I’m probably the one that started the slice and dice back with the big MTG2, just sayin) Nor do I like the fuzzy undefined edges of a hot spot that a textured reflector creates. Been there, all too many times. That said, the 5000K variant in my Q8 Plus are decent enough.
I have plenty of throwy lights, this one serves as a nice middle ground between power and distance. For now anyway.
1 Thank
I didn’t do spring bypasses in the tail, at this point. Should easily exceed 8000 lumens once that’s done. 
( This, in the SP36 Pro I just modified)
1 Thank
Well that’s not good
Can’t as in it just won’t work at all, or can’t as in you shouldn’t because it damages them?
Good to hear it worked out
The 7135 chips are meant to be turned on and then deliver a constant 350mA. They aren’t designed to be turned on and off rapidly for a lower average current level. So the way we use them in flashlights is generally out of spec.
Regardless, some types handle that just fine… even when given only a very short pulse and then turned off again. That’s how moon levels work on a lot of BLF lights. However, that type of use is very hardware-specific, and sensitive to tiny changes in the components or even the battery level. So the lowest moon modes typically required firmware features to help the user adjust the exact level used. Even within a single production batch, it could vary quite a bit from one light to another. That’s why NarsilM has a moon level adjustment.
Another solution is to use longer pulses, to get a more consistent response, but do them less frequently. This risks making the PWM visible, but it’s a lot less sensitive to tiny changes in hardware. It also has the benefit of using significantly less power, so moon runtimes can be like 3X as long. So that’s what Anduril does. It underclocks the MCU at low levels to improve consistency and runtime, at the cost of slower PWM.
Or people can instead switch to analog power regulators which deliver constant current instead of pulsed current. Hank has been doing that more lately. The MCU still generates pulses, but the pulses go through an analog smoothing / lowpass filter to convert it into analog control voltage… which then controls the power regulator. And this allows the firmware to do much more elaborate tricks with the pulses, to increase effective resolution beyond the usual 8 or 10 bits and make the bottom of the ramp smoother.
Anyway, 7135 chips have always been a bit finicky… and in particular, they have been one of the most frequent parts to fail. Some of the stuff we make our lights do can generate large voltage spikes, and those can damage the 7135 chips. Like, turning a DD FET on and off at high power levels… makes huge voltage spikes all across the driver circuit. The control chip is generally protected from this (or in the case of attiny13a, simply tanks it like a champ)… but other parts of the circuit are more exposed.
Those spikes were discovered when BLF started moving from attiny13a to attiny25. The 13 is amazingly tolerant, so it wasn’t an issue before that… but the 25 would generally reboot when it got spiked. So on high-power lights, circuit designers had to start protecting the MCU. But I don’t think the 7135 chips ever received extra protection.
… and, if I understand correctly, it got harder and harder to obtain any “good” 7135 chips. So the issues got worse over time.
6 Thanks
Huh. That answered a lot of things I’ve been wondering but now I have more wonderings.
So if you wanted to put as little stress on all the components on the average fet+1 with the average 3v LED(s), what hypothetically would be some things to do or some things to avoid? Any general tips? Just out of curiosity.
I mean, like stuff that is totally fine to do 1000 times but maybe not 10,000 times. Aside from stuff that isn’t realistic and the obvious kinda stuff like very high temperatures.
Would you set moonlight or the start of ramping to 350mah? Or the top of wherever that 7135 is? Would you a avoid using moonlight for an extended amount of time?Would you always ramp up and down to high/low and not just go straight to high or from high to off? At what level is the flashlight the happiest? Besides off. Low end of the fet? Top of the 7135? Depends? Would any of that change when the batteries are at a different soc, idk like >4.1V vs ~3.7v OCV for example?
Random side question, thing I’ve been wondering, what would be harder on the flashlight, using turbo at 150/150 (which Ive been told is like PWM with a 100% duty cycle?) or using a top of ramp of like 149/150? Because would that be like a 99% duty cycle? And is that easier on the components? Or is everything happier just full open instead? Happier being relative I guess because I assume both are somewhat stressful.
From what I recall, the thing which put the most stress on the light was around 50% FET power, with fully-charged high-amp cells and a fast high-powered FET. Each time the FET opens or closes, which happens at around 16 kHz, it generates a big voltage spike which rings through the entire rest of the circuit.
To reduce that effect, it was helpful to use weaker cells, a weaker / slower FET, thinner wires, really anything to reduce current or turn the edge of the square waves into more of a soft curve. Some drivers began to include dampening and other protections against ringing… though the main place they focused on was the power supply to the MCU, to make sure it wouldn’t get forcefully rebooted. But some also dampened the FET control, so it wouldn’t slam so hard when it opened or closed.
And then there’s Dale, who ran into these issues more than almost anyone else… because his high-performance modifications crank everything up to 11, including the factors which cause big spikes.
As for the 7135 chip, it’s mostly just a bystander. If you use it by itself, it should generally be fine. Just make sure the power input is 1 cell, not 2 in series. Make sure it’s not overcharged, like 4.35V. Try to make sure the driver is designed with a decent thermal path for the 7135 chips. Ideally also avoid low-Vf LEDs, since the bigger the difference between input and output voltages, the more power the 7135 chips must burn off as heat. Low output levels (like, PWM) can help too, since there is less excess power to burn off and it can cool off a bit between pulses.
It’s not the PWM which hurts the 7135 chips… it’s sustained heat and momentary or sustained high voltage.
But I’m not an electrical engineer. My understanding of these things is very limited. You’d have to ask someone with more relevant knowledge to get a trustworthy answer.
7 Thanks
Thanks for the lengthy explanations TK, much appreciated. That really helps a lot.
The new SP36 Pro driver I bought from Sofirn direct immediately showed no 7135 channel. No blink when powered up, no levels below FET engagement. When I replaced it the new ( old stock Raptor claw) 7135 did the same thing. Three new ones did. So I pulled the 7135 off the driver I damaged and it worked… for a little while. Started giving eratic behavior, then it too failed.
I would have reflashed an ATTiny85 but this new driver was different. Yes, there were flashing pads but I never used those before and am, in a way, starting over (as though I never modded a light before).
So, I piggybacked in a 26mm FET+1. With the new SIR178 FET and some (not all) the high power tricks.
I need to pull it and make a couple more ground connections to serve as thermal path.
With many of todays new offerings there’s not much reason to modify a light anyway. I know, right?
True, well appreciate it.
Btw what do you consider a low-Vf LED? Would a 519a be low-Vf? Or you mean low low?
I do seem to be doing more damage than good with the stuff I’ve tried so far lol. I am not well equipped for working on stuff this tiny and precise. Like, ya I can probably use this giant cast iron bench vise covered in motor oil to hold the pcb in place if I wipe it down a bit and put some foam in between it lol
LMAO
Yeah I can’t find any that will run lower moonlight anymore, at least without trying to slow the PWM or something. And the new ones just pop with 2S input. I have an old one “tanking it like a champ” in a 2S L6 driver
The 519A is pretty low forward Voltage. The big SFH55 too, these take huge amp loads but not sure it’s worth all the effort.
Just have to monitor what the specs are on a given emitter and try to be sure your setup isn’t going to run away and spike it out. Djozz’s tests show me what to expect as far as optimizing goes, always good to see it from other’s testing too. Average it out and plan accordingly. Not Always a great idea to use bigger leads, hotter cells.
Such great information. TK and Dale, sure is good to see you both posting again and sharing your knowledge!
TK…didn’t you have a thread back in the day that explained PWM in detail? The hardware side, not visual. Seems like a lot of people today are associating PWM with “bad” and of course that isn’t true and many don’t really understand what it is, how it’s used, or why…and I don’t understand it fully myself in order to help explain it adequately. I searched here a bit and read through some stuff from Wight and Dr. Jones and a few others but wasn’t seeing what I think I remember seeing. 
It would be great to have a bookmarked reference thread like that to share as needed.
1 Thank
Thanks Correllux, I mean, I appreciate the thought and all but ToyKeeper is the one with infinite knowledge.
I’ve said it all along, I’m just hacking stuff together and hoping nothing explodes! 
I don’t think I had any threads like that… all that comes to mind is the one I linked a couple posts ago:
This thread doesn’t really explain what PWM is… just one method for how to increase the resolution.
Ok…I may just be remembering indivudual posts and conflating users to boot. I do remember that awesome dynamic PWM thread - that was the start of great and greater things to come!
Idk if youve bought any power tools lately but everything uses brushless motors now. That’s like the thing to be. It’s all brushless. If it’s not brushless it’s trash. And brushless motors are controlled with PWM. Basically the same thing. I think.
So say if you were a marketing guy and you were tasked with improving the way flashlights with PWM are perceived you might think about calling them brushless flashlights or something. Flashlights with brushless dimming. You just gotta work the word brushless in there somewhere and it’ll sell. And give it 740lb/ft of torque too
1 Thank
At some level, basically everything digital is based on PWM… just sending square waves through electrical circuits, 1s and 0s, on and off. The important part is the details, what they do with it and how.
Most of the good stuff keeps the digital parts inside, and converts to/from analog when interfacing with the user. Typically this means turning something on and off fast enough to produce something in-between when it goes through an analog component to smooth out the signal.
For PWM style flashlights, that analog component is your eye. And … that’s not a bad thing. It works pretty well, as long as it’s done in a way which doesn’t register on maukka’s “snob index” meter. It mostly just needs a reasonably high frequency.
The “PWM is bad” thing comes from poor implementations, where the digital part is visibly digital… where the pulses are slower than the response time of an eye. It’s like the flashlight equivalent of a low-res pixelated image. It doesn’t mean all digital images are bad… just that that particular picture is. Instead of looking like the thing it represents, it just looks like a bunch of colored squares.
OTOH, it isn’t just digital systems which do this.
What people are seeing and disliking is quantization. The thing is, everything is quantized. It’s just not usually visible. At least, not without specialized equipment. Look deep enough though, zoom in far enough, and eventually you’ll see the pixels (so to speak). It seems that nothing is truly analog.
That’s one of the biggest and most fascinating discoveries of the past century or so – that everything in the universe seems to be quantized … or rather, quantum. Look close enough, and everything is digital … in a sense. Maybe not strict 1s and 0s, but at least discrete – distinct states which apparently can’t be smoothed between.
Like, make an electron go from its 4th to 3rd energy state, and it emits a single indivisible unit of energy, a photon, like a “bit” in a computer transmitting a single “1” in a sea of zeroes. It’s the universe’s smallest, highest-resolution PWM. When that bit hits something, like a particle in your eye, it briefly raises the energy state of that particle, from perhaps the 4th to the 5th level, and triggers a complex response which eventually is perceived as vision.
Do this more, and the perceived brightness goes up. But it’s not smooth, in a strict sense of the word… much like how a beach isn’t technically soft. The beach is made of many hard, sharp, tiny grains of sand… and only feels soft in aggregate. The same applies to lights, where brightness may feel smooth or steady, but only because the individual pieces are too small and too fast to perceive.
Bad PWM is like a beach made out of big, sharp rocks. It doesn’t feel soft because the pieces are too big. It’s otherwise not really all that different than a soft beach though… it just doesn’t have high enough resolution to feel good.
8 Thanks
I think I know Someone who is quantized, they lack the resolve to feel good.