E-switch UI Development / FSM

Nearly every project had the exact same Makefile, so I converted things to a build script per MCU type. Here’s what I do to build and flash it:

> cd ToyKeeper/spaghetti-monster/anduril
> ../../../bin/build-85.sh anduril
avr-gcc -Wall -g -Os -mmcu=attiny85 -c -std=gnu99 -DATTINY=85 -I.. -I../.. -I../../.. -o anduril.o -c anduril.c
avr-gcc -Wall -g -Os -mmcu=attiny85 -o anduril.elf anduril.o
avr-objcopy --set-section-flags=.eeprom=alloc,load --change-section-lma .eeprom=0 --no-change-warnings -O ihex anduril.elf anduril.hex
Program:    7372 bytes (90.0% Full)
Data:        288 bytes (56.2% Full)
> ../../../bin/flash-85.sh anduril.hex

Thanks, I managed to compile it now. Though these scripts use some avr-size extensions from Win-AVR, I had to remove -C and —mcu. And the following grep.

BTW, have you considered refactoring temperature control out of the UI code?
The way it is now all UIs have to duplicate it.

I think, conceptually, the UI code is probably the right place to handle signals like “low voltage” and “overheating”. However, it might be nice to put the details into a shared file so it doesn’t need to be duplicated.

The base library will probably make bad choices if it tries to handle those conditions with no knowledge of what the UI is doing. Like, low voltage in strobe mode or beacon mode probably requires different handling than a steady output mode. Or, overheating in candle mode, it’ll need to modify the UI’s internal variables in order to get dimmer, since it’s constantly changing brightness. Otherwise the candle algorithm would clobber the thermal algorithm.

So I don’t have a good and clean solution yet.

I don’t think UI should do thermal regulation. The way I see it is that UI constantly changes the desired brightness. But the actual brightness would be decided elsewhere.

UI would still need to be notified about:

  • the fact that overheating started or stopped
  • the current output ceiling
  • (?) that output ceiling changed

It would also make writing UIs easier. Forgetting about things like “I should be careful entering strobe if I’m overheating” would not lead to the light being overheated in any case. Unless UI forces something like #define NO_THERMAL_CONTROL.

As to low-voltage…I’m not sure if this is the same or different than high-temperature. It is quite unusual for UI to purposefully disable thermal regulation and allow the light to overheat. However never-leave-in-the-dark logic makes as much sense as a hard limit to protect battery from inexperienced hands.
I’m leaning to thinking that it would be best to do a PID controller that tries to keep voltage above the set floor, fluidly dimming down (or brightening up, f.e. when battery warmed up to above-freezing), just like thermal regulation. With either shutoff or not when further regulation is not possible. So it would be quite similar to thermal control when it comes to interaction with UI.

BTW there’s one low-voltage protection that I’ve never seen implemented.
When you can’t regulate down anymore, turn off. But just once. If the light is turned on again, keep working.
It will protect the battery in case of accidental activation. And will act as a signal that nothing more can be done to maintain battery life.
Certainly not muggle proof, a muggle will turn it on and think it was a malfunction.
And this is not the best way to signal low-battery as it may turn off in a bad moment, f.e. during a difficult crossing.
And the best way to protect from accidental activation is to make sure such activation doesn’t happen.
So it may make no sense. But I wanted to show the idea in case someone thinks it’s actually useful. Or gets inspired to make it actually useful.

Hi, ToyKeeper! You mentioned in the FW3A thread that you’ve implemented muggle mode into Andúril.

Is this latest Andúril version ready for release to the firmware depository? I’m itching for muggle mode and candle mode with a timer. :smiley:

Thanks again for all your hard work! :+1:

Candle timer doesn’t exist yet. Smooth ramp muggle mode works, but it needs more testing. It still feels a bit unfinished. The config mode changes need more testing too, since that got completely rewritten to save space.

If you want to try it anyway, I can publish the recent changes. It has a relatively high risk of bugs at the moment though.

Mostly, I’ve been feeling the need to rearrange the code to make it cleaner, but that’s always a high risk sort of change. And in this case, it might make the compiled code bigger. The strobes (including candle) are kind of a mess and I haven’t found a good way to untangle it yet.

That sounds like quite a task!

I’ll gladly reflash a Q8 and help with bug hunting if/when a major rewrite happens. :slight_smile:

Okay, I got Anduril (and FSM) working on the BLF GT. Here are the latest builds:

The GT build doesn’t always work in party strobe mode, due to the buck driver wanting pulses longer than 1ms, but I didn’t notice any other issues yet.

The UI diagram is updated too, showing that muggle mode can be exited with 6 clicks. The various config modes also moved to 4 clicks for consistency.

Thank you, ToyKeeper!

The GT is my ninth flashlight running Andúril. :partying_face:

Does the new muggle mode seem okay? People voted for a simplified smooth ramp, so that’s what it does.

I’m also hunting for any other bugs which might pop up.

Seems great so far. I voted ‘smooth ramp,’ so I’m quite pleased. :slight_smile:

I’ll update my Q8s and do some bug hunting. They’ll get some extensive use this weekend, so I’ll let you know if I find anything.

I really like the smooth ramping muggle mode. (Even though I voted for baton style, which i still think is easier to use and probably uses less space on the attiny)

I first tested it on the blf q8 since it comes apart so easily :slight_smile:

However, on the q8 the output does have a sort of bump in the middle of the ramp and it seems te be a bit jittery. Not as smooth as the main ramp. But I guess this is probably to reduce the size of the mode?
Also I noticed there is no click, hold to lower the output. It is more like hold , hold to lower the output. Perhaps this can be the same as the main ramp.
And finaly the ‘waiting time’ in muggle mode when you hold the button from off and the light starts to ramp seems to differ, but this can also be just me.

The four-click config modes is nice to have some more consistancy, though it will take a few days to get used to. :laughing:
The best thing in my opinion is the 6 clicks to exit, it happened way to often I lend a flashlight in muggle mode only for one person the unscrew the battery compartment and not understanding a thing how the light works.
The indicator led is also exactly how it should be in muggle mode; flashlight off, indicator led off. No questions why the led is on when the lamp is off.

TK, are the files in the depository up to date for building this latest version of Andúril? The dates seem to indicate so, but I wanted to double-check.

I like to disable boundary blinks and lower the value for HOLD_TIMEOUT.

Does this still apply to the latest version?

The repository files should be up to date, and the HOLD_TIMEOUT thing should still work.

I’ll probably merge the branch into trunk soon too, since it’s reasonably stable.

Aside from the limited range and slower speed, differences from the main ramp are there to reduce code size. The smooth ramp is, unfortunately, the largest of the muggle mode styles… and I still want to add some other things before the FW3A release.

That’s what i thought unfortunately. Maybe i’ll try to edit muggle mode to a baton style mode, however I have no programming skills so that might take a while.

Thanks! I successfully built GT and Q8 Andúril hex files without boundary blinks and HOLD_TIMEOUT set to 17. I reflashed three Q8s and my GT. :+1:

I’ll wait for timed candle mode and your other changes before digging the drivers out of my Emisars. I tried to build a hex for them anyway, but got a bunch of errors. FWIW, I was able to build one for FW3A drivers without issue.

I know you don’t use Atmel Studio, but I thought someone might have some idea what is causing these:

That looks like the driver definition line was omitted or misspelled. Check the area which says:

#define FSM_EMISAR_D4_DRIVER

Thanks, you’re right! When I deleted the slashes, I accidentally got the hash too. :slight_smile: