Great driver!! I would get 1 or 2 when they are available!
I got in some new rev PC boards (that did not have the solder mask screwups that OSHPARK’s board fabricators did on the first ones… they monkeyed with several people’s boards on that panel). I built one up with some 350 mA driver chips from Fastech. Seems to work well. You lose about 60 lumens per LED going from 2.8A to 3.05A per LED…
I got a better measurement of the parasitic drain… it averages around 6 microamps… about 666 times less than the stock SRK driver (over 4 milliamps). Should drain 4 x 2600 mAh cells in around 200 years. The stock SRK will drain the cells in around 3 months.
Sounds like you have been busy getting this figured out.
Thanks for all you work 
That is awesome! Weren't 7135s supposed to leak some current when not switched on? Is that not the case, or did you add some type of switch in between the battery and 7135s?
Biggest disadvantage of this driver is that it didn’t show up about 3-4 months ago, in the golden age of SRK 
Now, with Supfire M6 and Solarstorm/FandyFire Warrior, SRK is fading away…
It looks like only some versions of the chip do that, or perhaps only chips that have been abused in some way?
I've come to the same conclusion. The chips I've gotten from FastTech and gently soldered have been fine.
PPtk
I have noticed that the lead frames on ’7135 chips are particularly fragile. I’ve left behind lots of them when trying to unsolder them. Also broken them when bending the leads down.
The behavior of the chips is not specified under the conditions of having the Vdd pin pulled to ground. Something tells me there is no pulldown on the output FET gate. Without the Vdd circuit being powered on, the gate might be floating. I suspect it is only an issue with chips from some manufacturers.
That could be the Quiescent Current specs they have at 170-200uA?
That is around 2mA for 10x or 4mA for 20x AMC7135?
That spec is a boiler plate typical spec manufacturers use when they don’t want to measure the actual value. It takes a lot of megabuck tester time to measure those low currents.
For the AMC7135, it also applies when the Vdd pin is being driven high. The current the chips draw when being operated with Vdd low (assuming the chip is not leaking current through the output pin) is actually in the nanoamps. Going from 1 chip to 24 chips the board current increased less than 0.5 microamps. If the chip is leaking current, it may be a couple of milliamps. Designing a product that depends upon unspecified chip operation is a no-no, but sometimes you gotta do what you gotta do with the parts that are available, and maybe have to live with the consequences of what you actually get.
The same thing applies to things like AVR chip ADC inputs. They spec them at microamps, but I have never seen one draw more than a few nanoamps.
I just got my new XM-L2 U2 SRK from CNQG a few days ago and the parasitic drain is only 1.42mA. Tailcap current on HIGH is only 1.52A, and only 160mA on low. This is the brightest light I have ever had, with Low mode being just a tad dimmer than my previous “bright” LED light. The driver board appears to be glued down so I may order another one when it comes time to put one of your drivers in it; I like this one too much to risk ruining it. I was going to resistor mod this one but not going to chance snapping the board. I was expecting a much higher tail current, not sure what is up.
1.42mA parasitic current is actually pretty darn high. That will essentially discharge a 2900mAh 18650 in 85 days. In just a month, you'll be down to 60% of the runtime you'd expect.
1.52A seems to be really low, I was getting 5A on high with fully charged cells.
I grabbed my 1ohm 100W precision resistor and scoped the tail current and found my earlier readings of 1.5A and 150mA to be wrong; I knew the 150mA was a RMS averaged PWM, but expected the 1.5A to be accurate. There was so much play in conductivity of the positive connection of the contact board where the cell buttons rub that it was hard to get an accurate reading. Without attempting to solder to the contact board to make connection for a test I’m not sure if I can get an accurate reading. Some spots I could only get 500mA, others up to 1.2A. The only difference from last time was I opened and closed the host twice.
I’m tempted to put a layer of solder on the board to prevent the gold plating from being completely worn away as it is already pretty scuffed and I’ve only opened the thing about four times now.
As for the parasitic drain problem; I lock out the light by unscrewing the battery bay 1/4 turn. Though with the high wear on the contact board this does worry me.
1 Ohm is Wayyyy too high of a value for a current shunt. If you push 3A through a 1 Ohm shunt, you'll burn off 3 Volts of the input. A current shunt for the types of currents we're measuring should be more like .01 to .05 Ohms.
Yeah I know. It’s all I’ve got outside of the DMM though. I normally work with HV stuff so I haven’t had reason to get a precision milliohm sense resistor. Normally I’m measuring tens of amps at hundreds of volts, so burning a couple volts here and there isn’t of consequence.
Retesting with the DMM and not using the 4x 18650 battery compartment I’ve measured 2.8A on high with a single high drain cell. Unfortunately my bench supply is out for repairs and I don’t have anything to make battery connectors with so this is the extent of my testing on this. The 1.5mA parasitic measurement does appear to be consistent so far though.
Tens of amps (Say 30) burns even more voltage.. 30 Amps over 1 ohm drops 30 Volts! And makes 900 WATTS of heat... A typical shunt for 30A would be about 3.3 MILLIOHMS, giving a drop of 100mV
Yep, and a 30V drop on a 600V line only 5, but a 3V drop on a 4.2V line is a 70 loss. I normally use the resistor to scope the DC bus current and primary current of my solid state tesla coils, there the worst side effect is the slight de-Qing of the resonant circuit.
I rigged up a PD to check the PWM of low mode and it appears to be a 500uS pulse every 4.5mS, or roughly 10% duty at 222Hz.
Don’t forget there are four cells in there, so it would take approx. 4 months to reach 60% on parasitic drain alone.
IMO that is acceptable, especially for a light with tailcap lockout. Could be better, but acceptable.
Ah yeah, forgot about the multi-cell thing.. Still awfully high though..