I recently built one of Mattaus' 32x 7135 SRK boards and, on a whim, decided to flash it with a fast-PWM build of STAR 1.0. Lo and behold, after changing the PWM values around (it doesn't start making any light until 7, and that's a reeeeally extreme low, 8 is a pretty good very low moonlight) it... works just fine. It turns completely off in the 0% mode just like it should, everything works great (and, it's silent, of course). What did I do wrong? I thought this wasn't supposed to work?
Sorry for asking this question but why is it not supposed to work? I thought fast PWM was the correct choice for 7135 drivers.
I thought it was the FET based drivers that had problems with fast PMW. 
Yes.
7135 based drivers can work just fine at 9 khz or 16 khz. DrJones sells firmware at both speeds. I think RMM does too. Both are too fast to be visible to the naked eye, but the 9 khz speed may create an audible whining sound from the driver at intermediate modes.
It’s only the FET based drivers that supposedly don’t work at the 16 khz speed.
You are correct, if the light has a clicky switch and physically breaks power to everything when you want it to turn off. The story as far as I knew was that if the light has a momentary/electronic switch, and thus power is applied to everything at all times, even when the light is 'off', then fast-PWM firmware would not fully shut off the LEDs.
Interesting. I didn’t know that. Good to know.
No. Sorry, but no. Who do you think has been doing all the mad scientist experiments with these things?
Haha yeah. Good point! I have your FET mod in several flashlights already…. keep up the good work. 
I just checked DrJones’ page giving the info on all his firmware. Two of his Nanjg 105c, 7135-based electronic switch firmwares run at 18 khz (Simodrv and Mokadrv). And these aren’t ones designed for lights with a clicky switch. So I guess electronic switches have no issue with fast PWM.
There's six variables here: 9.4kHz PWM, 19kHz PWM, clicky switch, momentary switch, 7135s, FETs. All combinations work equally well together. I had not built a 7135/momentary driver before this one, and this was the last combo I thought still wouldn't work, but it does.
Maybe it's some difference in DrJones' code and the STAR code? I don't know enough about that side of it, I'm more comfortable with a soldering iron and a hammer...
Ah makes sense, thanks for that explanation.
FET + fast-PWM = no worky myth busted here: https://budgetlightforum.com/t/-/25299?page=7#comment-558568
I think you are right about that, the PWM even at 0 still has 1 "blip" for every 255 cycles which makes it never truly 100% off. The reason they don't stay is because the STAR momentary firmware puts the driver to sleep when it's at the zero point and the switch isn't pressed, shutting off the PWM output pin. Also, you wouldn't see it here because any PWM level lower than 7 isn't enough to light the LEDs, so I doubt that the "blip" would be noticeable.
You can see that "blip" point where the PWM doesn't shut all the way off if you use the STAR momentary firmware set to fast PWM with an FET. If you set the low end PWM level to 1, you will see an "in between" point for a moment when you switch the light from off 0 to to on 1 where the light is about half as bright as it is on PWM 1. This is the PWM 0 "blip" before the driver goes to sleep.
Switch on in lowest mode, then press & hold the button. LEDs stay lit dimly for as long as you hold the button. Unless it puts the MCU to sleep instantly when you release the button, I think that must be something else happening.
That's right, the MCU is going to sleep when you release the button.
Comfy is this driver better for the SRK than the other mattaus driver here http://oshpark.com/shared_projects/JV9IGc08 which is 20A. I see that this SRK takes 32x7135 and if its using the 350 version thats only 11.2A or if the 380 version its 12.16A. Why would someone choose this over the 20A FET driver, which is alot more powerful?
Try it with some 7135s from a different order.
These 7135s have heard of you & are too afraid to misbehave.
Not better or worse, just different. The 7135's will run at a fixed current (as long as your cells have enough balls to do it), and the output is determined by the 7135s and not solely by what the cells can supply. It should do the same output on either 20Rs or Panasonic 3400s, and it will maintain a constant output until the voltage falls below the Vf needed for the LEDs to run at the current limit set by the 7135s..
The direct drive version will do whatever a particular set of cells is capable of, and 'hotter' cells will give more output at the expense of runtime. Output will also not be constant, it will be dimming from the instant you switch it on. But it will be brighter.
Thanks, explained perfectly. Seems you have a gift of translating all these complicated driver mods, and making them easy to understand for us laymen folks :) Do you have a preference for one driver over the other, since playing with them both?
Well, the 32x 7135 board has 128 little pads that have to be individually dabbed with solder paste from a toothpick. That's not much fun. The regulated vs. unregulated doesn't impress me either, a FET driver has lower modes just like the 7135 version, you rarely use the highest mode for long periods anyway (I mean, come on... five thousand lumens? lol). Regulated drivers just seem like taping a block of wood behind the throttle on your car with the rationale that most driving is done at low speed and you can get a ticket for going too fast... but you'll wish you hadn't done it when it comes time to merge into traffic on the highway. Having more on tap just in case is hardly ever a bad idea, if you don't need it don't use it.
I had a good laugh when I read your reply, thinking about all those 128 points! So seems like the FET is the way to go! Now I'm just wondering whats the best battery to push it to its limit, the 20R, 25R or Sony US18650VTC3?