Attiny25/45/85 FW Development Thread

Did you still keep a version that fits on the 25? I foolishly only bought 25’s instead of larger models, but I’m looking for a good FW for my 7G3CS.

Ohhh - I'm up to bout 2500 bytes - blew past the 25... Still have the e-switch and brownout power switch support version which fits in a 25 though, just no config settings in the UI.

I’m going to have to order some 85V’s soon to try some of these firmware’s out. I’m still waiting form my 25’s to arrive.

That UI sounds great! And I like the way config works. Some will say it takes too much time. But really, how often will you config your flashlight? 8)

Ohhh - bout the time it takes... There's a shortcut built in. If you always wait for each setting to time out, yes, it would take at least 20 seconds (4secs x 5) to complete. But, instead of waiting if you click&hold, it skips to the next setting. So using the shortcut, it's pretty quick. I pretty much always use this shortcut now when I'm testing, and I've been doin lots of testing...

An issue with many firmware versions that support configuration, is the # of continuous fast clicks you have to do to get into configuration. With this, all you have to do is press and hold for 2 seconds, and you're in.

I was thinking of adding an "abort", so another words, if all you want to do is toggle moonlight mode, after you did moonlight toggle, you could bail out of configuration mode altogether, quickly. I think I can still do this with an extra long click&hold - it shouldn't interfere with anything and should be pretty quick, like 1 second or so, and is consistent with how this UI works (an extra long click in normal operation is strobe mode).

Also when you are in configuration settings, a click will always cause a blink, so you get feedback/acknowledgment that you did the click.

Biggest issue I can see with this is losing track of where you are -- in which one of the 5 settings. Originally was thinking of varying the blinks between settings, ex: entering setting #1: 1 blink, entering setting #2: 2 blinks, etc. I decided to just use 2 blinks for each setting entry because I thought the blink count itself would be hard for the user to count/track themselves, and also though one blink to start this doesn't feel right. I could of course use a different timed blink sequence to indicate you are entering configuration mode though, then do the 1 blink for the first setting - that would help.

There's lots of ways of doing all this of course...

Here's the latest doc/notes link: https://drive.google.com/file/d/0B1IxYZuk4DjcTUc5Z3ZkV3lsQnM/view?usp=sharing

This will eventually become the manual...

I spent a little time reading the manual tonight, and got temperature measurement working. I’m not 100% sure I’m doing it right, but at least it works. I get a value of 74 for room temperature and 82 after holding a hair dryer up to the driver for a while. Seems like a narrower spread than I expected, but it’s enough for my purposes.

Then I added thermal step-down (and step-up) to bistro.c. When the driver is heated, it’ll step down until the temperature is below the threshold. When the driver is colder, it will step back up until it either gets back to the original level or it gets too hot again. It currently has 64 steps in this thermal ramp.

I haven’t tested it in an actual light, but it works well with a hair dryer controlling the heat.

Let’s see… I also added a pseudo-random strobe, but it’s mostly just a proof of concept and the randomizer doesn’t seem very good because the ROM contents are less chaotic than I had hoped. And there are some other recent additions, like letting the user choose two or three levels of offtime (enable medium press?), and volts+tenths in battcheck mode.

The latest build is 1564 bytes, though some of that is extra junk which will probably be taken out later. I set 2048 as my limit so it will work on attiny25.

I didn’t feel like stopping yet, so I got the thermal calibration mode working too. Basically, go to config mode, wait for the relevant option to scroll by, turn the light off. It’ll turn on in turbo next time, measuring temperature and saving the value until you turn it off. After a power cycle, it’ll then use the new value as its threshold.

Again, only hair-dryer testing. But if I calibrate it to room temperature, it thinks it’s always over-heating so every mode steps down until it hits moon. If I heat it up during calibration, everything works fine afterward (but steps down when the hair dryer is on it, and back up when the dryer is removed).

With the work being done by both Tom E and ToyKeeper, we’ll be seeing some amazing new drivers in the (hopefully near) future!

It can be hard to follow incremental threads, so here’s a summary of what the ‘bistro’ code does now. Basically, what’s on the menu so far? These things are already working:

  • Choice of 1 to 8 regular modes, plus an optional moon.
  • Choice of low-to-high or high-to-low order.
  • Mode memory, or not.
  • Reversing via medium-press, or not. (requires OTC)
  • Thermal step-down with an easy self-calibration mode to let each person set their heat threshold. Only works in regular modes though; blinkies are not temperature-regulated.
  • 64 brightness levels available, though they’re only normally seen during thermal step-down.
  • Soft start on regular modes.
  • Hidden modes when medium-press is enabled:
    • Two-speed police-style strobe.
    • Volts+tenths style battery indicator.
    • Bike flasher.
  • Most of the code is abstracted to eventually work on either a single-channel or dual-channel (FET+7135) driver.

It’s not done yet, and I know of at least a few things which need to be fixed (like LVP), but it’s getting there. I’m basically putting in as much as I can fit in 2048 bytes. Hopefully it can cover most people’s needs for a clicky-switch light.

Sounds pretty fabulous

These are toggled through the config mode? Same 15 fast click to enter config as blf-a6?

Yes, same method for entering config mode. Fast-tap anywhere between 15 and 30 times (until it stops lighting up) then leave it on.

The config mode makes the timing and counting a little easier now though. Instead of “blink, blink, pause…” for each option it now blinks a number then kind of “buzzes” for a bit to show where the click window is. Click during the “buzz”.

Before:

  1. blink, blink, pause
  2. blink, blink, pause
  3. exit

After:

  1. blink buzz, pause
  2. bl-blink buzz, pause
  3. bl-bl-blink buzz, pause
  4. bl-bl-bl-blink buzz, pause
  5. bl-bl-bl-bl-blink buzz, pause
  6. exit

Some of these options invoke a special secondary mode. Like, to select a mode group, it will go into a beacon-like mode where it’ll slowly blink 8 times then pause, then repeat. Turn the light off after N blinks if you want mode group number N. Or, for the thermal calibration, it’ll turn on in turbo and wait for you to turn the light off to set the maximum allowed temperature.

Eventually I’ll try to make a diagram again, but I haven’t yet.

Bistro revision 185 won’t build for me.

I did try out r175 but only briefly. I’ve been sick and had to work up energy just to download, build, flash. :Sp Also I didn’t / don’t see tk-random.h in the repository so I removed random_strobe.

Oops, I completely forgot to add the random header. Fixed.

About the warning, it’s not a fatal error, is it? For me, it doesn’t prevent gcc from compiling. In any case, adding the missing ‘const’ fixed it, and I re-pushed the changes.

You’re right, it doesn’t appear fatal. :zipper_mouth_face:

I finally put one of these drivers into an actual light and did some temperature testing.

At first I nearly destroyed an LED trying to set a heat limit, since apparently the MCPCB was floating instead of having good contact with the pill. Oops. At least the emitter still works, though moon is much lower than it used to be and there are some tint irregularities.

After fixing that, I’m finding that the pill-to-MCU thermal lag makes things difficult. When setting a threshold, the MCU may not be as warm (yet) as it should be. Then when stepping down later, it can sometimes step down much farther than necessary and then take a long time to come back up. And after it comes back up, it may not go as high as it should because the threshold is too low.

… though on a second test (I left it on turbo longer during calibration), it behaves much closer to how it probably should.

It may be because I’ve got it in a big copper pill, and maybe it would work better on a smaller light. But so far, although it works, it doesn’t work as well as I had hoped. And I may need to add a special state transition to bump it back up to the original level instead of short-tap taking it to the next level.

I ditched my experiments with a temperature controlled mode. If the temperature rises above the threshold I just call the same critical routine as critical voltage and put the light in critical mode (either off or very low output depending on setting). The user has to manually switch modes or restart the light to get out of critical mode, which of coarse will only work if the temperature is lower.

I found that the different heat lag characteristics in different lights made it too much of a hassle to make a generic routine for all lights that wasn’t annoying.

After tweaking it until my battery got too low to heat the light up, I think I might have finally eliminated the thermal “bounce” effect (repeatedly over-shooting the target then rebounding).

… maybe.

I’ll find out tomorrow when the cell is recharged.

The thermal controls might actually need to be adjusted according to the physical host the driver is in, to account for varying amounts of thermal lag. It’ll be interesting to move the driver to a smaller host after I’ve got it working right, to see if it still works.

There’s also now an option to turn off thermal regulation. Just tap the button during the first second of its calibration mode, and it’ll calibrate itself to 255 (hot enough to cook pizza, I think).

For those that are interested in using it, could you add a build time option for an external thermistor to ground (as the temp increases voltage would drop on the attiny pin) instead of using the internal sensor? Attaching a thermistor to the pill with thermal compound should provide less heat lag.