I think some of the confusion is about the order of the cells. I am calling the 1st cell the one on the far left of my diagram.
The body of the light carries 4.2v through it, just like a normal SRK as that single cell is all that is passing through the body.
The other 3 cells only connect to the contact plate in the tail and to the driver contact pads, which feeds into the FET on cell # 4 (the one on the far right).
I discussed this with an EE friend here @work. His concern was the gate voltage compared to the source voltage coming to the FET. He reviewed the specs on the SIR800DP: N-channel, Vgs rated at +/- 12V.
Have you tried this on the bench yet? Were you planning on using one of our standard FET's? Have you looked into the voltage delta between the gate and source of the FET?
No, I have not done anything more than the design in the OP, just a proof of concept really.
I was not real worried about the FET as I know they exist that would work. For example the old classic NXP have a gate to source voltage of 20v.
http://www.nxp.com/documents/data_sheet/PSMN3R0-30YLD.pdf
I am sure there are more options as well but the NXP is already proven and would work fine. I honestly notice very little difference in the real world between it and the sir800.
For the NXP, yes, it won't blow up anything with a range of 20V, but the source to the FET is -12V compared to the MCU's supplied gate voltage to the FET, so even for the NXP, the FET will be always be on. This is what my friend is saying.
I am not an expert on FET’s by any means.
When the MCU is not sending any voltage to the FET gate, It is an open circuit correct? I am not sure how the FET would turn on with an open circuit on the gate?
When the MCU sends the voltage output to the FET would it not turn on anyways?
Honestly I am not seeing how this setup is any different than using an LDO? Either way the gate sees +4v and the source is –12v?
I guess it is my turn to miss something?
The source is -12V, while the gate is 0 to 4V, therefore when the MCU outputs what it thinks is zero, the FET sees it as +12V, because to the FET, the -12V is it's ground. So when you think you set the gate to 0, it's really a 12V difference over the source input to the FET, so, the FET is always on.
That's the way I understand it. Think it makes sense.
Yes, it makes sense on one hand, IF the circuit was set to 0v by the MCU. The way I understand it though is that when the MCU turns the FET off the circuit is open with no connection at all. Thus how does the FET get energized if there is not a complete circuit?
Even in a zener modded 2s setup there is a 4.1v difference between the source and the gate, this is more than enough to energize the FET if what you are saying is true. Yet zener modded driver work fine?
With a zener, I think the source and MCU are both getting the same ground reference, so the MCU's gate is in the 0V to 4V range compared to the source. Batt+ of 8V goes direct to the LED, so to the FET, it's on the drain side. Dunno, I could be wrong here...
Yes, if the MCU sets PWM of 0 means an open circuit, guess you are right - should have no turn on effect to the FET.
Concerning the SIR800DP, it's been discussed and researched several times here on BLF - I don't need any more convincing. I got a few SIR400DP's that are even better. For single LED usage, probably not much benefit, but for high amps, it runs cooler, better.
There is actually an easy fix to this come to think of it, I simply run the 1s V+ through the body of the light. Then the ground will be the same but the V+ will be what the MCU needs and the FET will not see more then the 1s difference in voltage.
I didn’t want to put the V+ through the body but come to think of it there is no reason we can’t.
Hhmm, that does sound better 
Yeah, just so used to running the ground through the body I didn’t even really consider running the positive. That would eliminate any possible issues.
This is really cute. It's (almost) exactly what I started wondering if you were doing when you started talking about 4s in the other thread, because there's no headroom for a buck (I wouldn't have thought you'd raise the case voltage to do it, but that works). Are you planning to ever sell these boards assymbled, or convince someone like Mountain electronics to do it?
The good thing about all this is that it makes it all work with simple dd or linear single-cell type regulators. The bad thing is still exactly that. What would really be ideal in a perfect world might instead be to run 2s2p led's and 4s batteries with either a current controlled buck converter or an 8.4V buck converter with PWM on the output. This would allow to run higher power for much longer I would think, but.. I don't think designing the buck converter board would be nearly as easy of a task. I've been looking around for an off-the shelf one that could even fit and they don't exist, but it looks possible to fit. By the way, I'm really tempted to go in on this light, I just am having trouble convincing myself I need this light and I have my eye on this mtb gearbest thing. This driver concept however is useful in its own right.
Well, actually you would have room for a buck converter in the Q8, I actually started working on a buck driver for it but decided to wait on that for another time as it seems to be quite difficult to make work in the real world and I don’t have the resources to prototype as many as would be needed in order to get it working.
A buck driver would be ideal but sadly just not feasible at this point.
Sure, I get it. It's big job. It's exactly how I've been thinking of trying to run it with a remote 4s pack though. Then I can stick a remote buck driver anywhere I want, and the off the shelf ones (if any of them work as claimed) are still quite small, mostly just not quite the right shape to fit.
And actually an off the shelf one would fit in the battery holder, but keeping the holder and using a remote pack seems a little rediculous. It could make a bunch of sense though for say an offroad light setup on a car.
In fact if I were going to try do it (well...) I might just buy one of these:
see if it works, maybe figure how to beef it up slightly, maybe not if really handles 10A, and just try to copy it but in a 46mm diameter layout, but it's a lot of stuff crammed in. The pots can go at least. Even that plan doesn't look easy.
The parts are apparently easy, the hard part is the actual layout. At high currents there is a lot of EM fields being created and they tend to play havoc on everything in the circuit.
More than a few have tried and failed I am told.
How about the MTN-MAX drivers? Richard has shared the board on OSH Park. Maybe he would be okay with stretching the board out to the SRK size and adding your quadrant layout on the bottom. Incremental changes don’t always happen in one person’s design, they sometimes pass from person to person. Most of the BLF drivers were made by somebody looking at somebody else’s driver design and wanting it just slightly different for various reasons.
I talked to him about that and he said I was welcome to try it but also said he does not recommend it.
He said he straight up got lucky with the layout on that board and has had trouble making another one that worked. I decided to take his advice and leave it alone till I had the time to start from scratch.
He gave me a much better buck converter IC to work with that should be much more reliable but would also need to start from scratch on.
I still want to try it at some point but that will be something after the TA drivers are completely finished and when I have a lot more time to play with it.
I want to make some Avenger drivers with LDO option along with the 4S SRK driver and a Q8 driver with an extra bank of 7135’s (possibly the same as the 4s driver but not sure how feasible that will be). Also if any changes are determained to be in order with the ability to remove R1/R2 then I would make a few new drivers with those removed and pads for other options.
Minor things but still things I want to finish before moving onto other drivers.
I'm work'n on a 15 mm 85 based driver now, to update the firmware on. It's a wight driver using the small footprint FET, but didn't need any extra parts like a C2 or FET resisotrs for it work stable. Of course it was never scoped, so really dunno if I'm just lucky or not. It's in this light, the SolarStorm SC02: https://www.fasttech.com/products/1601/10005040/1669500, actually pretty good deal at that price. Has a real SS bezel.