PebbleOS on Bangle.js 2?

[gfwilliams]

Posted at 2025-02-07 by @gfwilliams

You may have seen in the news recently that the Pebble watch OS has now been Open Sourced and work is underway to port it to nRF52840 devices (seemingly with the hope that a new Pebble watch can be made).

Bangle.js 2 happens to also be nRF52840 and so it may be possible to run Pebble OS on it. There is some discussion on https://forum.espruino.com/conversations/374260/?offset=250#comment17663202

I have been having a bit of a play around porting the Pebble OS to Bangle.js 2, and I put a repo here: https://github.com/gfwilliams/pebble-banglejs2

Some instructions and info on the current state are at the top of the README on that repository.

It now builds, flashes, and the code kinda runs, but errors because it can't access the flash chip. I've also added code for the display but haven't been able to test it yet.

If anyone is willing to give this a try and see what they can do it'd be hugely appreciated (@fanoush? You seem to be awesome at this kind of thing)

To get debug info you'll have to crack the watch open and connect to the UATX pin, which isn't ideal - but I have a bunch of watches I can't sell here (for instance with a dead magnetometer) so I'd be willing to provide one to anyone who needs a device they can pull apart to test with.

I guess there is the potential to do debug messages over RTT serial (over SWD). It's another step, but it might make sense to allow better debug without breaking watches open.

[fanoush]

So the error seems to be that the ID read from flash does not match this predefined value
https://github.com/gfwilliams/pebble-banglejs2/blob/a137127801a5af7669580979ab7f983d9b96bf8a/src/fw/drivers/flash/gd25q64.c#L53
Did you kept old GD id or updated it for bangle2?
I never used PebbleOS before but will have a look. Also will try to add RTT since I don't have any Bangle2 opened.

EDIT: oh it gets 0 if the log is correct so indeed generic issue with the command

[gfwilliams]

Yes, that's what I thought originally, so I made the gd25q64.c file with what I believe is the flash ID and the correct commands (the ID might be wrong of course) as well as some code to try and do wakeup.

But as you say the actual command isn't working at all, and even disabling that check, it seems no other flash commands are working, so I guess it's something to do with how QSPI is setup to communicate with the Bangle - although it would appear that it works for whatever nRF52840 device the developer are using.

... I guess maybe there's a chance there is some pin I didn't manage to change in the code that overlaps with the SPI peripheral and causes problems?

[thyttan]

Pebble javascript apps on espruino.

Apps written in javascript for pebbleOS, would it be feasable to get them running on espruino? Without modification?

Maybe some could basically just be copied over and others would need some kind of compatibility layer app that routed Pebble API calls to Bangle.js counterparts.

Just thinking out loud.

[gfwilliams]

It's possible, yes. I think a lot expect some JS to be running on the phone as well, so those would have to be ruled out. I imagine there's still quite a bit of work wrapping things like the graphics API. I wonder what apps there are that would really be worth it though - whether it's less effort just to re-write those for Espruino.

... or maybe it's more sensible to write a PC-based translation layer that could run on the apps before upload to convert them? It could then warn if an app wasn't compatible, and it'd be a bit more efficient too.

There was talk of changing Gadgetbridge to allow its pebble-specific stuff to work on Bangle.js too but we never got that far.

[bobrippling]

I'd be interested in having a look at this, maybe translating on-device/adding a bit of pebble API is feasible for most apps. @gfwilliams do you have any watches you can't sell left over still, that would be ok for testing with?

[gfwilliams]

My concern with trying to translate on-device is it's going to be quite slow?

But if you develop a JS wrapper as you say (on device), then if it's too slow it might work to run a minifier/bundler over the whole thing on the PC that would inline calls? It's a tidier solution that trying to bodge something up that'll run over the parse tree.

Since you're doing so much work with Bangle.js I'll have a look around and will dig a Bangle out for you. I definitely have a few with a compass that's not working that I could send out - I'll send an email.

[bobrippling]

Yeah, I guess if the app was 100% JS, it'd execute natively on a Bangle with us just adding a few APIs, or yes, even faster if we threw a good bundler or the closure compiler at it. I'll have a tinker - and thanks!

[thyttan]

@gfwilliams I just placed a preorder for the next pebble with touchscreen and four hw buttons to arrive in december.

[jonasbark]

Agreed, though I'd love to see the Pebble OS on Bangle. I wish I could help to do that, but unfortunately, firmware development is scary.

[idavydov]

It's really an interesting discussion. I am currently not interested in pebble since the focus will be in the hardware and the key to useful smartwatch is actually software. So continuous work of @gfwilliams is highly appreciated here.

That said, pure hardware comparison is also interesting.

• I think screen of bangle js is probably more appropriate for smart watches.
• The water resistance is really an issue for bangle.js2. Mine recently died after quick submerge into the water :-( Pebble promises more.
• Having a usable mic is a great feature. Saying something to the watch so it's processed on the phone opens up a lot of opportunities!
• Hardware buttons are cool. Touchscreen on my banglejs2 was never reliable, so it was always difficult to navigate the OS. Basically doing anything on the watch was very slow, so maybe mic + buttons is a better usability.

So if there's ever banglejs3, it would be amazing to have:

• Better water resistance
• Buttons or at least more reliable touchscreen
• Mic and potentially speaker

I think it would be totally justified if banglejs with mic/speaker, buttons and water resistance will cost more.

In my case I never used GPS or heart rate sensor. But I am sure there are people who do.

[gfwilliams]

There's a thread on ideas for Bangle.js 3 but the whole idea of mic/speaker and water resistance really scares me. I'm sure it's possible for Apple/Samsung/etc, but not on the kind of R&D budget I have available! I'm really interested to see how Pebble get on with theirs.

So I think if we went for water resistance (which seems like a good plan) it'd be by having no holes whatsoever in the case. So over-moulded rubber/silicone buttons, wireless charging - but no mic/speaker or barometer. I'd like to make it extendable in some way so I can imagine there might be a different rear available with those in - but fitting it would likely mean giving up on water resistance.

[idavydov]

Thanks, @gfwilliams .

Just out of curiosity. If it's not speaker + mic, but only a mic, does that make a problem easier? I saw there are some "off-the-shelf" waterproof mics, but I have no idea how difficult is it to integrate them while keeping the whole thing waterproof.

[gfwilliams]

Well, it might make it marginally easier - but obviously every hole we put in the device is a hole that has to be sealed and has a certain percentage chance of not sealing correctly. If you're making 10,000+ of something then those chances add up to leaks and replacements/unhappy customers.

The ideal thing would be if someone made a module with barometer+mic+speaker and one single hole, and then it feels a lot more worthwhile.

[gfwilliams]

Well, it turns out it is posssible to run PebbleOS on Bangle.js 2!

Right now you've got to crack the whole thing open first and solder to UART pins, but I think it's going to be possible to at least write everything via the SWD pins

[stanelie]

What works?
What doesn't work?

Wahoo!

[jonasbark]

Awesome!

[gfwilliams]

Well, enough works to get that screen at the moment, and that's about it! It does appear on Bluetooth and is connectable but I'm not convinced there is much available there yet (I'm not sure that very much is implemented).

The main thing to do to make it usable and really see what's going on is to use the touchscreen to fake button presses - I'll try and get something going there over the next few days I hope

edit: actually I can get a console into it via the serial connection and can send it fake button presses, and you can navigate through the menus with those and get animations, so once the button faking is done it'd be good enough to have a play with at least

[gfwilliams]

The touchscreen is now working to fake button presses.

See it in action: https://youtu.be/d_U5xZiBUGc

[fanoush]

SWD dongle costing you maybe €15

Well even if you ignore cheap stuff from China (like this), you can nowadays get Raspberry Pico from many places. For UK - https://thepihut.com/products/raspberry-pi-pico - even the version with headers (for no soldering) with UK shipping is £8.80 total. Then you just plug it in and drag debugprobe_on_pico.uf2 file from https://github.com/raspberrypi/debugprobe/releases to usb drive that pops up by default and you have high quality SWD dongle.

however the nrfjprog qspi flashing is nice, too bad it is hardcoded to segger j-link. Was checking if openocd or pyocd could do it but probably not out of box. There is pyocd discussion here pyocd/pyOCD#1094 that mentions using flm file with the flashing algorithm but I guess it is not part of nordic segger installation(?), here is also something promising ARMmbed/DAPLink#1073 (comment) , actually I do have daplink build for nrf52840 host and target working that can present internal 52832/52840 flash as usb storage, maybe adding bangle2 qspi flash wouldn't be that hard and bangle2 could have separate target build configuration with that.

Anyway, just found relatively cheap 52840-DK https://www.ebay.com/itm/267170477903 that ships in EU so may be useful for someone.

[BrandonTR]

Attempting to initialize QSPI gives me this error:

nrfjprog --log --qspiini banglejs_nrfjprog.qspi.ini --qspieraseall  
[error] [ Client] - Encountered error -176: Command qspi_configure_ini executed for 25 milliseconds with result -176
[error] [ Worker] - Failed to fetch QSPI configuration: [error] key "rx_delay" not found in the top-level table
 --> unknown file
   |
 1 | unknown line
   |  the parsed file is empty
ERROR: An invalid argument was provided. Use --help for a list of valid
ERROR: arguments.

Has anyone else ran into this?

[gfwilliams]

Given key "rx_delay" not found in the top-level table it looks like a tools issue.

It might be some issue with the version of nrfjprog you have? I have:

$ nrfjprog --version
nrfjprog version: 10.15.4 external
JLinkARM.dll version: 7.62

[BrandonTR]

I downgraded to the version you're using and that solved it, thank you! For reference I was on:

nrfjprog --version
nrfjprog version: 10.24.2 external
JLinkARM.dll version: 8.26

[gfwilliams]

Great! That's interesting though - it's probably only a matter of adding rx_delay to the qspi config file to make it work with the new version.

[gfwilliams]

Update: it turns out the JavaScript TicToc clock face does run fine. There are two different ones depending on whether you have a color of black&white watch, and because right now I'm only doing black and white you're getting the simplified one (which is a digital clock).

So I think it's all working as well as can be expected now. Apparently there's still a bit of work to do on Bluetooth pairing so it's not liking pairing with Gadgetbridge as-is, so it's not really usable as a watch right now, but it should be pretty easy to follow along with development.

[thyttan]

Super cool - nicely done!

[banglejsdev]

@gfwilliams Just out of curiosity, how difficult would it be to get Espruino running on a Core 2 Duo or Core Time 2?

I rewrote an app I use frequently that I originally created for my Pebble and Pebble Time to run on my Bangle.js 1, but haven't modified it to run on my Bangle.js 2 because I think it'd be too difficult to use without multiple hardware buttons.

I really love my Bangles so am not very interested in running PebbleOS on my Bangle.js 2 but would consider buying a Core Time 2 if I could run Bangle apps on it because of the hardware buttons. Perhaps the Core Time 2 hardware could become the Bangle.js 3? Or if that's not feasible, would you consider selling a version of Espruino that would run on a Core Time 2 so you could get rewarded for your porting work?

Regardless, thank you very much again for all your dedication and hard work on the Bangle! Your efforts have made me, and I'm sure countless other Bangle users, very happy!

[gfwilliams]

Thanks!

I think porting Espruino over is probably not that hard (if you just want display+bluetooth+buttons) however Pebble doesn't use the Nordic SDK for Bluetooth like Espruino does, so you'd have to open up the Pebble and flash it via SWD to get Espruino working on it, which is going to put a lot of people off.

It's a hard one for me to justify working on though - I can't imagine folks would be willing to pay very much per device just for the software, and it's going to require a lot of support initially. The new Pebble is expensive enough there's not really enough leeway for me to be able to resell it either.

Having said that, there's kind of a bright side. A few months ago I ordered a new batch of Bangle.js but the manufacturer has told me that's the last batch they'll be able to do because they can't source the screens any more. I'm going to want to keep selling a Bangle so unless someone starts selling a PebbleOS watch that'll make a good base for a new Bangle (which is possible) I'm going to have to design my own device in the next year. And that device is absolutely going to have more buttons on it!

Given the relatively low price of flash memory, probably what I'd do is try and put more memory on it and have PebbleOS and Espruino that you could swap between without needing external hardware. It'd be a slightly more expensive device than right now but would probably sit at or below Pebble's price.

[enaon]

so unless someone starts selling a PebbleOS watch that'll make a good base for a new Bangle (which is >possible)

it looks like the moved to a SiFli chip, not NRF anymore?

[gfwilliams]

The black and white pebble is still nRF52840 I believe? But the color one is SiFli, yes...

While actually porting Bangle.js to a new architecture isn't the end of the world, the real issue is Bluetooth. We expose an awful lot of bluetooth functionality to the user/developer, and reimplementing that on a new architecture is going to be a total pain

[thyttan]

The Pebble touch screen watch shown off in Eric Migicovskys podcast: https://www.youtube.com/watch?v=pcPzmDePH3E

[gfwilliams]

How is everyone still feeling about a PebbleOS for Bangle.js port? I haven't heard a bunch recently.

I could merge the most recent changes into my fork, and could work on a small device that'd allow anyone to flash PebbleOS to a Bangle via the SWD connector.

I'm just not sure how many people are really going to care - I think those that do care have probably put themselves down for a new Pebble, which let's face it is going to be better supported and much easier to use with Pebble OS as it has 4 buttons.

[jonasbark]

I find PebbleOS the best watch OS I have encountered and a great fit for the Bangle.js as well.

I think those that do care have probably put themselves down for a new Pebble

True - I preordered one yesterday, but I also had hardware issues with the Bangle.js (plastic broke on two sides that held the watch strap pins). If that wasn't the case and PebbleOS would be supported (including Bluetooth) I'd love to stick with Bangle.js.

[enaon]

How is everyone still feeling about a PebbleOS for Bangle.js port? I haven't heard a bunch recently.

I feel a bit sad about pebbleOs on the Bangle. The UI on the pebble is pro level, one of the best if not the best for such devices, and I really enjoyed some comments he made on one of his last yt videos, he wants to make some things bigger, he said that 10 year has passed, and now he cannot see as well as he used to. He spoke to my hart :) There are some screens on the bangle, that I have to use the magnifying glass to read, I really hate it, it makes me feel very not nice.

but I will by a bangle again, because of espruino, the king of ease of use in prototying, stable, capable, flexible, amazing overall. I have no idea how many of your users feel like I do, but the bangle is not an end user smartwatch for me, it cannot compete with the big players, hrm sensors etc will just make me move to a huawei and feel like I changed century. But espruino, for me, is sui generis.

I think that if you really want the bangle to be an end user smartwatch, with added bonus of being opensource and easy to program, you should find someone that does UI, I feel you need some help on that.

[thyttan]

I'm just not sure how many people are really going to care - I think those that do care have probably put themselves down for a new Pebble

I think you are probably right.

[bobrippling]

I feel like the benefit would be running Pebble apps on a Bangle, but if we can have a thin emulation layer anyway then maybe it'll be easier sticking with Espruino underneath?