What does the cap do? I think that because of the pwm control of the fet(or 7135) the mcu is in essence generating pulses in it’s own power supply. The cap filters that somewhat.
Comfy’s right in that less than 100% gives you a simulation of control and I would think you could slave 3 fet’s just as you would 3 boards of 7135’s.
What are the results of potting to reduce whine?
It's still completely unregulated whether it's 3 LEDs on one FET or one LED on one FET. This is a direct drive driver. If the LED's Vf falls due to heat but battery voltage remains the same, current will increase. It is the same as a battery connected straight to the LED on high mode, but with the ability to turn the FET on and off rapidly to get lower modes. It won't do anything else.
The FET is just like a relay, but with no moving parts inside.
DPAK (TO-252) package, Rds(on) = 6.5mOhms @ 4.5v, Qg (gate charge) = 17, should work though not quite as low resistance as the SUD70N02-03P, but about equal to what was used on the 'good' east-092s.
The cap has to be WAY bigger than what'll fit on the board to absorb all the noise (like, 470uF). On the 17mm drivers I think most of the noise comes from not the driver PCB itself like the big SRK drivers, but from the springs & switch, and I don't know how you go about potting those.
I didn’t think the cap eliminated the noise, just filtered the input power enough to keep the mcu happy. If it’s in the other parts then maybe fiddling with those other parts to change their resonant frequencies would help.
The original 10uF ceramic cap is enough to keep the MCU happy no matter which firmware it's using, just not enough to fully dampen out all the pulses enough to keep things from vibrating in this 9.4kHz range. A 100% stock plain unmodified 105C will whine when using the 9.4kHz version of luxdrv. You can flash back and forth between luxdrv and nlite on the exact same driver - luxdrv (9.4kHz) will whine, nlite (19kHz) does not (well, it's probably still making noise, just up where most of us can't hear it).
Playing around with a software signal generator, I can't hear a 19kHz tone (and I have speakers fully capable of going above that) but 9.4k is horribly annoying.
I was doing some general mosfet hunting while tearing down some electronic devices and it got me to thinking. First of all, I know almost nothing about choosing semiconductors for a specific electrical purpose.
This raised a question in my mind.
What if there were an N-channel mosfet, in the same SOT-89 package as our beloved AMC7135 current regulators?
I did some checking at Mouser, and found these stocked N-channel Mosfets in a SOT-89 package:
MFR: Diodes Incorporated
ZXMN6A07ZTA @ 2.2A
2.6 W Power Dissipation
Rds On: 250 mOhms
Rise Time: 1.4 ns
Fall Time: 2 ns
ZXMN6A11ZTA @ 3.6A
2.6 W Power Dissipation
Rds On: 180 mOhms
Rise Time: 3.5 ns
Fall Time 4.6 ns
Imagine being able to fill the top of a Nanjg 105c with 4 of these Mosfets, for 8.8 to 14.4 amps of near direct drive goodness. It just seems that if we can find a large size Mosfet that will work, why not find a smaller Mosfet that will fit the original footprint of our beloved 105c driver?
I don’t know if using multiple smaller Mosfets will help or hinder the max current… ie… reduce on current resistance by distriburing the load over several mosfets?
Comparing the rise and fall times of these smaller mosfets when compared to the datasheet of the 70N02 mosfets in this thread, the smaller mosfets are way faster, possibly fast enough to use fast pwm maybe?
I don’t mean to hijack this thread at all, as I also plan on trying out different large single n-channel mosfets on nanjg 105c drivers also. I was just wondering if there was a better solution to choosing a mosfet more compatible with the small size of the 105c driver.
I believe the Rds On: on comfy’s mosfets are like 4.2 milliohms @ 4.5v…way way way way lower than the ones you found
But every time I try to find a comparable Mosfet, comfy quickly informs me that I am wrong… 
[and trust me I have been looking all over the daggum place]
The pin layouts are backwards.
The datasheet says it's even worse - 350mOhms at 4.5v. Compared to 4.2mOhms at 4.5v.
The way I found the one linked in the 1st post is the exact opposite of the correct way to design electronics - I browsed ebay for hours and hours and hours, tracking down datasheets on random FETs until I found one that looked interesting, then figured out a way to use it. :|
Tivo's design uses a small FET already, up to 4 pads, and can be piggybacked to go over 4. But the specs are not as good as the one comfy found. With that said, I'm getting 6.1A with just two on a XM-L2 on copper, Sam 20R cell.
He also uses a Zilog MCU as well…completely different coding and uses a specialized flashing box that costs $30 (I think I could be backwards…I get Tivo and Pyro mixed up)
But it is a VERY cool design!
At what point will an XM-L2 on copper turn blue and go poof?
How much copper/aluminum/brass needs to be backing the copper star to push that number higher?
Does the XM-L2 continue delivering output beyond 6A even if pushed that far?
Are sausage and jalapeno and cheese kolaches better than fruit filled ones?
I keep asking myself these questions even as I try for more power in the lights I play with. Also ask myself why it’s all about the power, not better reflectors for a purer narrower or more defined beam profile. And I’ve decided, that with coffee, a cream cheese kolache is pretty dang good! Especially after a couple of sausage jalapeno and cheese ones. 
A good cup of Folgers Columbian is calling my name…
Don't know about the filosofical part, apart from that my coffee at the moment is a black Lavazzo espresso (mmm), but sadly the XM-L2 on copper will never turn blue, the bond wires will fail long before that.
Thanks for setting me straight. I knew I didn’t quite put enough effort behind my “of the top of my head” idea before posting. When I was laying in bed, it did strike me, “what if the pinout on the ones I found were backwards?”. Then I thought, NAWWWW, they must have standardized the SOT-89 package in some way! Yes, the “on” resistance is wayy lower with the mosfet you found, and that is a VERY important thing to keep the amps and voltage as high as possible. That was one thing I neglected to check when doing my quick search.
At least now, I have a little better understanding of what to look for, because it’s pretty fun disassembling scrap electronics searching for usable mosfets!
At what power level have you found the XM-L2 to fry the bond wires? Copper or no…
I wonder sometimes just how these little emitters stand up to all we give em. Rated for 3A, we far exceed what they were designed to do on a regular basis.
Just wanted it to be known that I do think of other things besides flashlights and their related tangents…food! 
I tested it a few months ago in this thread, I found (for the record, I tested just one emitter, I don't know how much it would vary of course), and found that at about 9 amps the bond wires burnt through while the colour output of the led still looked fine.
Did anyone considered AMS1117, maybe!
A... voltage regulator? A link or even a proper name might help.
SOT-223 is like the SOT-89's cornfed linebacker cousin from Topeka. Quite a bit bigger.
And besides, it's a regulator, so you'd need another component to convert the attiny's low current PWM output into a high current source to feed the regulator (like, you guessed it, a FET)... and besides, a 5v regulator won't work very well when it's fed by less than 5v.