AccelSender app interval 1s instead of 10s when using sleep as android?

[ratmice]

I was connecting to the bangle.js from https://www.espruino.com/ide/# and loaded the accelsender.json into the editor, I noticed intending to set the enabled bit to true, there were a couple of odd things, presumably this file is derived from config.json

This had the contents {enabled:false,interval:1000} so no quotes, and 1000 instead of 10000.
Enabling it confirmed that I was getting updates once per second rather than once every 10s.
I don't know of any settings anywhere on the watch itself, or in gadget bridge allows you to change the update interval.

I did a test where I set this setting to 10000/disabled, disconnected from my computer, and connected to android,
and started a sleep session in sleep as android. Then I turned bluetooth off, reconnected to my computer and downloaded accelsender.json. Indeed the file was {enabled:true,interval:1000}.

I don't yet know exactly what is sending this, whether it is gadget bridge, or sleep as android itself?
Is something using centiseconds and something else milliseconds, or just using a higher frequency interval?

cc @AnotherStranger in case they might know

[gfwilliams]

As you say it looks like there's no settings page for the app which would let you change that file, or even a mention in the README.

so no quotes

Yes, that's expected. When Espruino writes JSON files using writeJSON it does so in a slightly compacted format on the Bangle (where possible) just to speed up reading/writing and save a bit of space.

I don't yet know exactly what is sending this, whether it is gadget bridge, or sleep as android itself?

I just looked in Gadgetbridge and in app/src/main/java/nodomain/freeyourgadget/gadgetbridge/service/devices/banglejs/BangleJSDeviceSupport.java there's enableAccelSender which sends a command accelsender and hard-codes an interval of 1000 (it'll be in milliseconds).

That command is then interpreted at: https://github.com/espruino/BangleApps/blob/master/apps/android/lib.js#L294-L297

So it's Gadgetbridge that's forcing the update interval

[ratmice]

I see, it's just not what I was expecting given the entry in config.js, looking at the sleep as android docs it isn't really a problem.

Sleep As Android expects to receive data aggregated per 10 seconds intervals.
...
For each sampled value, sum up acceleration change along all axes to get a single value.
You should aggregate the 10 seconds maximum of the data. For efficiency, we suggest
you do the aggregation on the watch and send only aggregated values to the phone – but that’s up to you.

So I assume that it just hasn't implemented the aggregating and sending data at different intervals.

[ratmice]

Ahh yeah, I think you're right I wasn't sure though if the interval was controlling both the updateAcceleration and the sendAccelerationData

But it looks like you are right that at https://github.com/espruino/BangleApps/blob/master/apps/accelsender/boot.js#L51-L52
only sendAccelerationData is being called at the interval, and updateAcceleration is called in on when new accelerometer data arrives.

It seems like an easy fix, but it might take me a bit to figure out how to build gadget bridge/test it.

[thyttan]

@ratmice I guess you already found https://gadgetbridge.org/internals/development/setup-environment/ ? :)

[ratmice]

Thanks, following those instructions (and modifying updateAcceleration to println) confirmed that updating and sending still occurr at different frequencies...

I went ahead and sent a PR https://codeberg.org/Freeyourgadget/Gadgetbridge/pulls/4751

I do wonder if the updateAcceleration can't be done more efficiently by using the Bangle.accelRd(reg, cnt) passing a count at less frequent intervals than Bangle.on("accel", updateAcceleration); , anyhow that doesn't require changes to GadgetBridge...

[AnotherStranger]

Hey @gfwilliams and @ratmice,

the higher send intervall was set intentionally by me because I found that for me the aggregation worked a bit more reliably.
However, I have no hard feelings for changing this. You will have to set the interval in Gadgetbridge though (Alternatively we could add a configuration for this setting and toggle only the enabled flag from Gadgetbridge).

[gfwilliams]

I do wonder if the updateAcceleration can't be done more efficiently by using the Bangle.accelRd(reg, cnt)

It's definitely more efficient the current way (using the acceleration event) rather than polling the accelerometer direct (since we already gather accelerometer data).

Internally, Bangle.js could poll less often and use the accelerometer FIFO to read more out at once (I think there might be an issue open for it), but realistically it's not going to make masses of difference to battery life.