Flashlight Firmware Repository

thank you ToyKeeper!

Sounds pretty cool; I’d love to include it if possible. DrJones hasn’t given luxdrv an open license though, so we might need to get his permission to publish derivatives.

How about attiny85? I hear Mike C has some code working on that, and I’m hoping he’ll decide to share it. :slight_smile:

Or, if I ever end up with an attiny25-based driver, I’d be happy to port something to it. I don’t really have the ability to build the driver myself though.

No, the BLF A6 firmware only has one strobe (or biking flasher) and a beacon/battcheck mode. Also, it requires an offtime capacitor and it’s designed for dual PWM, so it might not work on the same drivers as brass-edc.c.

However, I made a stripped-down version which removes the runtime config options in favor of making room for other features… so it’d be relatively easy to add other blinky modes. It’s ToyKeeper/blf-a6/tk-otc.c in my repo, and the entire purpose of it is to make it easier for people to build derivatives. I’m assuming that, if someone is modding code, they can configure it at compile time instead of runtime.

Ahh, thanks! Didn't know I could update it directly. I'll try to create an acct and directly update then...

I'm all ears on Mike's ATtiny85 version - very, very interested. Can't stand this 13A code limit.

Gotta take a look at your tk-otc version - sounds like what I need. I'd want a STAROffTime or your noinit with a strobe option. Adding strobe would be much better with a "short cycle" mode method, but looks like your A6 two method of clicks (quick and hold) would be as good or better. Right now I don't see an easy way to integrate a strobe into STAROffTime

I’ll be sharing it. It wouldn’t exist if 13A code wasn’t so freely shared by others here.

I have voltage monitoring working, and have coded a user interface so the user can re-program the amount of modes and levels. I’ll make it 105C compatible with defines, so it can be used for 105Cs out of the box. Working on another light right now, but will get back to the 85 shortly. Next on the agenda is getting to know how the temperature sensor works, but it look s fairly simple as it’s the same routine as checking voltage levels with the ADC. It’s the calibration routine that could be the hardest part.

BTW, if you use bzr at a command line instead of using a GUI front end, the commands to clone the repo, modify it, and publish a new version are: (‘%’ represents a command prompt, the lines you would type in)

% bzr branch lp:flashlight-firmware
% cd flashlight-firmware
(hack, hack, hack, maybe 'bzr add' new files)
% bzr ci
% bzr push lp:~myusername/flashlight-firmware/my-branch-name

Then either submit a merge proposal or simply let me know the branch exists, and I can merge it into trunk.

I generally prefer to keep a local copy of trunk along with a “feature branch” for each set of changes I make, to help keep things clean. This approach is pretty standard, and is discussed in more detail here, along with how the usage compares with git. But if you don’t know how to do that, I can easily accept patches or branches or merge proposals or whatever.

There are a few cross-platform GUIs available for those who want one — Bazaar Explorer, QBzr, and TortoiseBzr.

Since there are so many flashers here :bigsmile: , I thought I’d ask a question about multiple flash/test sessions. Is it OK to leave the SOIC clip attached when doing a short test of a revised firmware? In trying to learn this language (which started out as Martian, but is now more like Swahili) I’m going to be revising this or that in the code, flashing a driver on the test bench to see if it works, then repeat. It would be nice if I didn’t have to reconnect the clip for every iteration, as I have a feeling the clip’s lifespan is related to the number of connection cycles.

I can see a couple problems:

  1. Backfeeding voltage into the USBasp.
  2. PWMing a 20cm piece of straight wire, and that could cause EMI gremlins to appear and possibly affect the outcome of the test. I’m not even sure if the PWM pin(s) are connected to the clip.

Opinions?

Thanks.

Thanks for your input.

I emailed the question to the USBasp’s developer. Maybe he’ll respond.

Thanks Toy Keeper and everyone else for all the hard work. I wish I had something to contribute but I’m only just starting with the flashing but this helps for sure.

I was going through this last night and found a couple that were nice.

For us complete laymen a section with just hex files along with a brief description would really be great, but Beggars can’t be choosers right!!

I tried this a while back… and again just now. It didn’t damage anything, but it didn’t work very well either. There were two issues:

  • The clip can short some pins, resulting in strange behavior which differs from how the light behaves without the clip attached.
  • The clip doesn’t stay on very well on its own, so it’s difficult to test the light without the clip coming off.

So, it’s generally a not good idea to leave the clip on during testing. It’s possible that the clip could cause damage, and it changes the driver behavior (which kind of invalidates the test results).

TK - I pm'ed you the link to my versions (google_drive_tome_drivers).

Not sure if this is the right thread for this, but here it goes... Your blf-a6 and tk-otc drivers have a couple of features I haven't seen before:

  • The "OWN_DELAY" option to use our own timer delay rather than the delay supplied with the compiler/tool. Interesting because I just noticed on recent Nanjg 101-AK-A1 builds with STARNoInit, the two minutes for a turbo timeout is really 2 mins 20 secs - off by quite a bit
  • you are mixing FET output and the single 7135 output in some modes but not full turbo. This is interesting, and I assume you did test out these combos to find a nice spread of output levels. I notice you didn't publish what the approximate percentage of outputs are in either the source code or blf-a6.txt. Do you have that info now?
  • I assume you set the turbo mode to phase-correct PWM and not fast PWM because it doesn't really matter (PWM's are not used on max setting of 255)

I'm going to try your blf-a6 version on a A17DD-SO8 board. I'm thinking I could just zero set the mode values for the "small circuit" (MODES1x1 and MODES1x2) and set the big circuit PWM mode values to what I normally use for the A17DD-S08. Do you think this would work ok? Is this the easiest approach?

For the future, I plan on dropping the A17DD-S08 driver in favor of wight's FET+1 driver but I still have stock of drivers (most assembled) to use up.

BTW, This BLF-A6 driver is the most advanced for a clicky power switch light I'm aware of in source code. You did a great job on this!! Thanx!!!

Edit: I found one issue in using blf-a6 on a FET only board - the blink'ing is set for the 2nd PWM output port (the 7135), so I simply changed it to the primary PWM port, using a value of 5, and works like a charm now. I added a compile switch for that. From my bench test, it all looked good. Stupid me though -- put it in a Convoy C8 (old style) and it grounded a LED wire lead to the reflector after I tightened the bezel -- fried both the driver and LED Frown. It looked like there was plenty of clearance, but I guess tightening down the bezel really pushed down on the reflector, causing the short...

If i remember correctly, on fast PWM, PWM IS used on maximum setting ( ON for 255/256).
Can´t find the thread at the moment, but it was Dr Jones who discovered that in a commercial firmware.

Very nice work! Hat's off to all of you!

Thanks! I’ve been out on sort of a personal holiday (went out and played with the butterflies), but I should be able to add it soon.

Use DELAY_TWEAK to adjust the speed. The MCU clock speed is only spec’d to be accurate within about 5%, and I’ve found it’s often even worse than that. One MCU will do 70 seconds per minute while another does only 55 with the same code. So, you might need to adjust that value to get the timings right. I tried to give it a rough average as default, but I’ve seen anywhere from 850 to 1050 cycles per ms.

The output percents change based on the maximum power level — it will be different on a single XP-E2 versus XP-L versus a triple. This is because the 7135 chip runs at a fixed power level but the FET does not, so their relative contribution to the total changes with different emitters.

Because of this, the levels need to be re-calculated (and the source recompiled) for each class of light it’s used in, and a percent value would only apply to a specific combination of hardware. I calibrated the default to be visually linear (cube root model) on the BLF A6, with up to 140 lm on the 7135 and up to 1400lm on the FET. I estimated the initial values then tweaked it until my light box showed the desired results.

As for PWM on turbo, I picked those settings because they measured the highest in my light box. Specifically, FET plus 7135 together produced less light than the FET alone. And I vaguely recall someone (comfychair?) measuring max output curves on fast vs phase mode with a high-speed scope, deciding that phase was slightly better.

Yes, it should work… aside from that issue you found. :slight_smile:

Thanks for reminding me about that. I should probably release an update with a compile switch for it like what you made.

Youch, that sucks! I’ve reflector-shorted things enough times that now I always cover the bottom of the reflector with kapton tape. Then again, I also usually test out a build on a bench power supply at low amperage before I try it with a real battery.

The upcoming BLF-A6 must have the best tweaked brightness levels in the flashlight universe

This really depends on what you are doing. The power from the ISP is enough to power the MCU, so if you’re testing start up routines like off time cap testing and brown out detection you should unclip it.

It’s a pain in the * to unlcip ’em all the time. That’s why I built this setup with relays that vex_zg donated after showing me his solution. Now I just press a button to unclip:

Nice! That certainly beats my cheap setup… I just have a few different drivers attached via long wires to decapitated hosts, with extra wires soldered on to connect a battery by hand. I have to clip the clip each time, and frequently hold it in place during flashing because it won’t stay on its own.

So, what I have is cheap, compact, and only took a few minutes to make… but it’s also inconvenient. I hope you don’t mind if I drool a little over your dev setup.

Hah, we’ll see. I have to recalibrate everything when the production sample shows up.

Small through hole mount 2 pole relays are about $1 each from digikey. Anyone want to design a board with two connectors a switch and some relays?