Gate drivers? We don't need no steenkin' gate drivers!
Works beautifully. Everything - absolutely everything - except the three FETs, can be removed (well, maybe keep the cap where the 105C is picking up the BAT+). Attiny PWM-out is tied straight onto the gates, with the 3 gate drivers removed everything else on the board is inert and not needed.
It does 4.18A, one channel into a single XML2/Sinkpad. All 3 channels tied together to the same LED, around 4.80A. The stock FETs (AO4468) are not very good, at least according to the datasheet, 18 milliohms.
The 3 channels/1 LED thing isn't really representative because of Vf issues, more LEDs will increase the output. I imagine in a 6x light, with 2 LEDs per channel, it'd do around 5A x 3.
That is impressive. And some sweet looking soldering. Wish I could do that without bweakin it.
Still following this closely.... :-)
Lol what is/was this copperpipe?
I guess there is no longer a problem with the ground connection :bigsmile:
The ground... thing is just 12AWG wire, with some copper bits to connect to the FETs' source legs. Originally they got their ground through the sense resistors and the funky-shaped traces, no need for all that anymore.
Digging the look of that driver. Kind of Borg meets flashlight driver. Your work is a little over my head right now, but it seems to be going in a great direction.
Works fine this time. I guess the first try with the stock MCU(-thing) still on the board did something weird to the signal. Just bridging the big sense resistors gets over 4A per channel, I may be sane and boring on this one and just add resistors to get it up around 3.2-3.3A each.
PWM wire connection onto the SRK board:
This is just like the master-slave thing with multiple 105C boards, only using various different hardware as the slave instead. And the momentary switch firmware is what I didn't have to do this stuff before.
I'm afraid to try one and find out it's not as good as the ones I already have... :~
HA, 70N02 does 4.34A, single XML2, single 20R (two 20Rs=4.62A, 3=4.78A, 4=4.86A). Works fine connected straight to the attiny. Bypassing the FET (short drain to GND) with the same single cell/LED gives 4.45A. So it's a good'un. :party:
-----
And YES, adding the toroid makes the low much much lower. I checked by jumping across it while running in various modes. Makes barely any difference in higher modes but affects lower modes A LOT.
Good info in this thread comfychair. 
I have some of the 70n03 FET’s from this thread recommended by moviles2. East-92 4amp 17mm driver.
I have tired to keep up with this thread from day to day and I guess I keep forgetting the day before. Getting old sucks.
I’m going to have to go back and read this all from the start. Maybe once your done comfychair, maybe you could make a “How to” or a “guide” to help forgetful people like me. 
Nice work you have done here, thanks for sharing.
Sense resistors bridged, it does ~4.3A per channel. The copper bar isn't a ground bus like the stripped-down version, it's just there to mount & ground the 105C.
1. The toroids help achieve a lower "low" mode, correct? Are you using lower driver PWM settings to get this, or with the same PWM setting is the low mode lower?
2. Where can you buy toroids like these? (Besides harvesting off of existing drivers ;-) )
Love the way you modded that traditional SRK driver. Do you plan on using DrJones MiniMo on it? What is the lowest PWM value you think you can get to work on this rig?
PWM of 1, with a toroid, gives a hilariously low low. Won't even register on the meter. I haven't done a new build of the firmware to compare the low numbers, but Werner's is using 7 for the low, and even with the toroid it's too much light to really be called moonlight. 1 without the toroid is fairly low, but still not moonlight really.
I don't think the uH rating on the toroids matters much, just that there's something there, and is able to carry enough current to not be a restriction in the higher modes. A little bit of resistance drops the current a lot with the single cell input voltages. To be small enough and carry enough current, it's almost got to be one with multiple conductors, I think. Small number of turns you get with multiple conductors on such a small core seems to work fine. These were harvested from old motherboards.
We're working on nailing down a readily available toroid that ticks all the boxes.
You have any thoughts on 70N02 vs 70N03?
Yes, it has higher on-state resistance (.0051 vs .0042) and much much higher gate charge (90nC vs only 40nC), which means a tiny13 would probably not be happy driving it without extra components added in that nobody wants to mess with (gate drivers).
If its easy and simple to explain, what does the Torroid do?
Magic, would be my guess.
According to Wikipedia, it's like a resistor when current is changing (like, a PWM signal that's on-off-on-off), but a plain conductor when the current is steady. Or something. I think magic is probably a more plausible explanation.
This is the old-style SRK driver, MiniMo, PWM of '1':
That's much brighter than in real life, the camera's trying its best to find any available light.
Frickin cool. I haven't seen any srk or triple emitter deliver a decent low, let alone a moonlight. That looks like a firefly mode or something. Very awesome.
Thanks for the responses above Comfy. Much appreciated.