Nice.
Still early days... since it has the indicators on at all times, and since this thing pulls 10 or whatever amps on high, it's displaying the voltage under load which may not be all that relevant. Should the indicators only be active in the lowest mode, for a more accurate idea of battery resting voltage?
Right now the color changes happen at 3.7v & 3.4v, I'm unsure where to go from here. If those points are lowered to make it more accurate when using the higher modes, it could fool you when using the lower modes. You could have a green light that changes instantly to red just by moving up a few levels. It will also be affected by the type of cell used, one with more sag under load would give a lower reading even though it still has perfectly fine open circuit voltage. Ideas/suggestions?
Oh, and the STAR low voltage protection stuff is still in there, the indicators are for information only and aren't needed to prevent damage.
edit: And here's both LEDs on at the same time... was supposed to be yellow, but orange works too.
I got a chance to calibrate the voltages better today, and adjust the thresholds down for more even spacing. Now the levels are 3.0V, 3.3V, 3.6V, 3.9V, and 4.2V. And 3.3V, 3.9V, and 4.2V match up with my DMM in real time, so they’re pretty close (at least, on my hardware). The other two aren’t directly tested yet, just approximated based on the other values.
comfychair, I think the reason it looks orange is because the red LED shows through the window a lot more easily than the green LED. It’s just due to placement. A clear, angled plastic extension on the window should make the colors more balanced.
As for showing the real-time voltage sag under load, I’m not convinced it’s a bad thing. It matches what my DMM shows, and makes it obvious that the cell capacity actually changes based on load. I’ve watched a weak cell drop from 3.9V to 3.3V in a matter of seconds while it’s running at maximum on direct drive, and then when the load is gone it quickly goes back up near the original voltage.
So… I have this Roche F6 w/ Ferrero Rocher DD driver and a triple 219B with 10511 frosted wide-spot floody optic. Made two firmwares/interfaces for it so far; one with fixed levels and one with smooth ramping.
On this hardware, since it’s direct drive with PWM, the overall output is pretty voltage-dependent at all levels. At ~3.6V, the lowest mode is about 1 lumen. At 4.2V, the lowest mode is 3.6 lumens.
The high end is pretty impressive though, for a triple Nichia 219. At 3.6V, it maxes out at 1300 lumens on a 20R cell. But at 4.2V, it starts at 1825 lumens. Granted, it drops fast due to heat… but still. That’s a lot of light for Nichia 219s. I thought I’d be lucky to hit 1500 lumens. And that’s OTF lumens through both a TIR and a lens. The emitter lumens are probably in the range of 2000 to 2200.
This little light is achieving output almost on par with my original Skyray King (before modding), only with neutral white high-CRI output. It’s a beautiful thing.
(BTW, the firmware for this should also work fine on a modded SRK or modded Supfire M6… anything with an attiny13a and an e-switch and lots of power. Or, with some minor tweaks, it should also work on lower-power devices; just needs the lowest levels adjusted.)
I built another one, this time with triple XP-L's. The single XP-L (dedomed) in the stock OP reflector does 27-28 lux ceiling bounce, about equal to what I get with a non-dedomed XM-L2 with a BLF driver. The triple 219B F6 does around 35 lux. Triple XP-L does 73. Seventy three!
A 'SecurityIng' SRK clone with a BLF-SRKDD driver, on four INR 25R's, only does 80 lux. A totally stock CNQG SRK gets utterly trounced at just 43 lux.
That sounds about right, based on my measurements. My original CNQG SRK (stock) measured at about 2200 lumens, which is only 1.3% different than your measurement. That means the triple-XP-L F6 would be getting somewhere around 3750 to 3850 lumens! (which, BTW, is beyond the range of my light box… I need to build an integrating sphere)
No wonder the cell voltage sags so much on turbo. 
That’s a little scary.
And when I get the triple XP-L and a bit of time, I’m planning to build something that bright in a Cypreus host. Not sure I’d trust a Roche F6 to handle that much heat though; it doesn’t have much thermal mass.
It’s not scary ’till it comes on in your pocket. Think of it as a light saber with a focus problem.
With the non-ramping UI, if the button gets mashed in a bag or your pocket and stays mashed it just steps down all the way to off and stays there.
Here is my F6.
Finally it is now the perfect flashlight for me, thanks to @comfychair who helped me a lot with driver, flashing and now with the idea to fill the hole with an 3mm LED that I’ve glued there:
Out of curiousity, what firmware / UI style are you using?
He's using this: http://75.65.123.78/ramp_v07/
Color mixing from the status LEDs works better with the little optical pickup trimmed off flush with the ID of the driver cavity. I used a tiny little ball-end diamond bit, that also leaves a good scuffed-up finish that acts as a diffuser.
@ToyKeeper
Yes, I am using that one with ramping and it is the perfect one because it also has short press for on/off and memory.
Okay, I was just wondering… because I didn’t quite finish tweaking the ramping UI before I got interrupted by a work trip, and I haven’t gotten back to it yet. It’s pretty close; just needs some fine tuning on the ramp curve.
The levels I used in that one linked up above are:
0,1,1,1,2,3,4,5,6,7,8,10,12,14,16,19,23,28,34,41,49,58,68,79,91,103,116,130,145,161,178,196,215,235,255
That pattern, odd as it may be, works surprisingly well - and really well on the SRK-DD boards using the inductor.
The pattern in that one increases each mode over the previous one by:
1 1 1 1 1 1 1 2 2 2 2 3 4 5 6 7 8 9 10 11 12 12 13 14 15 16 17 18 19 20 20
Thanks. I pulled in the changes you made, and included it as a preset in my version.
That ramp pattern you made is a quadratic or “x squared” shape. It’s much closer to visually-linear than the logarithmic pattern I was using, but spends a bit less time at the lowest levels.
The new version I just pushed has a few updates:
- Made ramping UI support both fast and phase-correct PWM. (compile-time #define)
- Added explicit support to turn off PFM with a #define. Saves space but makes the bottom few levels get a stair-step effect.
- Improved instructions for compile-time configuration.
- Added more pre-defined ramps for various configs, including Comfychair’s 40-step quadratic ramp, a 64-step cubic (visually linear) ramp, and a 40-step logarithmic ramp with PFM.
- Made PWM/PFM setup code clearer and more compile-time moddable.
- Made the default ramp one which goes extremely low and uses PFM.
The default ramp now is 40 steps, uses fast PWM, uses PFM for smooth steps between PWM levels, and goes extremely low by abusing PWM=0. I’m measuring just 0.1 lumens with a Sanyo 20R at 4.18V. It can ramp to several levels under 5 lumens, yet still goes all the way up to 1800 lumens. However, this extra-low moon is extra-sensitive to voltage, so at 3.6V it’s virtually no different than “off”. And at that low a voltage, the stair-step pattern returns somewhat.
In any case, it should be a bit easier to configure now, goes much lower, shouldn’t have an audible hum if you enable fast PWM, and I’ve tested several predefined configs to make sure they work.
So are you saying the one from BG has more room in the driver compartment than the one from FT?