Just saying that the diode drop doesn’t affect lumens. Any reduction in current draw to the mcu path(s) would help a tiny amount reducing sag but not sure if it would be measurable since it’s such a tiny fraction of another tiny fraction of DD draw.
Edit- what was I thinking? As Crux reminded me below while the diode isn’t in the led current path it certainly does affect the maximum voltage on the Gate which directly affects FET resistance.
Given that these cells are flat tops it’s a bit easier to mess up and drop one in backwards and this is about as “free” as anything gets.
Thanks Alex, it’s mind boggling how much time you have put into contributing to a worldwide shared resource.
This may not mean anything to anyone else but the main reason the components on the side opposite the attiny were crowded to the center was the narrow neck of a Mag Solitaire. I have plenty of the tiny10’s I can drop a 2502 on so that’s not an issue and this layout has a lot more room to work with along with a stronger B+ idea.
I've been thinking about the diode too. While it does not carry the LED current it does drop the voltage available to drive the FET, so depending on the FET and the battery voltage, the ON resistance may be higher than normal. (but maybe not enough to matter)
Aside from protecting the processor IC from reverse battery, it also isolates the power from the IC while programming it. (assuming the programmer supplies power to the chip)
What this means is without the diode the whole driver will be powered during programming, and because the PWM pin is one of the pins used during programming, the LED may follow the data on the that line. The blinking LED isn't necessarily a problem unless the power source can't handle the power draw and sags. This sag could corrupt the program process. This may all depend on the type of driver.
I'm not sure about this as I've not done any testing, but it seems reasonable to me.
Anyone have any insight on this?
Part of me wants this for the giggle factor (3 amps off of a cell that is likely good for .250 AH means that all you can do is giggle during the 5 min runtime) part of me wants to figure out a way to get a linear regulator into the mix too.
I used that java circuit sim I did the SOS disable circuit in and it looks like that capacitor that is across the LED + pad and ground helps efficiency significantly at low duty cycles. With a 47 microF capacitor and 1% duty cycle, I’m getting a peak of ~160mA and a minimum of 20mA once I add some resistor to simulate real life circuit resistances. At .5% duty cycle, I’m getting a peak of 54mA and a min of 14mA.
You’re right Crux, that part slipped my mind. It depends on how close to the elbow in the R vs Vg we’re operating since mV do count here. Not so much with purely 7135 based drivers. All the more reason to use the best FET.
Dale go ahead and try it with a jumper instead of the diode. As Crux points out it could very well change the resistance of the FET enough to result in higher amps.
Can’t get any higher than a direct connection to a cell, and that nets 3.16A, which is about where I’m at in the light. So the little Efest IMR10440 has hit the ceiling. Is there a fet out there that’s gonna boost the cell?
Think I’m just gonna stick a Luminus SBT-70 in it and run it as a mule. 
I’m going to mention this now so you have someone to blame if it ever comes into existence: XPH70 penlight. 2xIMR10400 and a reflector slightly taller than the LED.
I also released one of those today: [WIP] 10mm DD+single-7135 driver: double sided 10440 torture for Dual-PWM
XPH 50 on 10mm XM sink pad might actually fit in a 2 x 10440 host. AKA AAA Minimag.
Edit- or you could try it with 2 x 10180 in a solitaire.
A XPH70 should fit too. It fits on a 9.9mm diameter circle. Do they make IMR 10180 cells? The smallest IMR cells I can find are 10400.
The veleno 10180 at bang good is lithium cobalt. Any reason it has to be IMR? Other than don’t over discharge.
At a only a few hundred mA 2 x 10180 probably won’t be as bright as and XML with a 10440 though.
I don’t think ICR can cope with the heavy drain (that the XHP50 and XHP70 would demand to justify putting it in a light instead of an XM-L2/XP-L2) that well.
If I’m putting an emitter that expensive into a light, I’m going to drive it hard enough to justify the cost.
Think harder about what RBD is trying to tell you. 
Besides, we still have to describe the Zener mod for this chihuahua for any of that to happen.
Due to the small size of this driver, an SOD-323 Zener diode is best.
In either case the cathode of the Zener diode will face the ‘D1’ identifier (PCB marking) and the anode of the Zener diode will face the outside of the PCB.
Scrape back a bit of solder mask between “1” of D1 & “C” of C1 ?
I think that the Zener diode may be turned onto it’s side and used in the “hug” position.