search

InfiniTimeOrg / InfiniTime

Firmware for Pinetime smartwatch written in C++ and based on FreeRTOS
GNU General Public License v3.0
2.76k stars 943 forks source link

Wake on wrist raise too sensitive #548

Closed drzacek closed 1 year ago

drzacek commented 3 years ago

I think the wake-up on wrist raise is too sensitive.

While it is nice to not have to twist your arm quickly by 90 degrees to activate the screen (like on my MiBand 4), it is pointles to have it activated almost constantly during normal day activities.

My PineTime lights up even when I'm driving on a highway and barely turn the steering wheel.

I think that battery life suffer greatly from that. Maybe tuning it down a little bit might be a good idea.

Dperrigo18 commented 3 years ago

I would recommend something like a sensitivity slider in the settings for this would be good. That way the user could fine tune it to what would be acceptable for them.

atomdmac commented 3 years ago

https://github.com/JF002/InfiniTime/issues/490

hubmartin commented 3 years ago

I did some experiments with Colmi P8 LIS2DH12 (chinese copy) accelerometer filtering because it was also turning on too often. I solved that by averaging last accelerometer 4 values. I have also 100Hz sampling and 2g range (reading over I2C seems like 10 Hz over my scope) https://github.com/hubmartin/InfiniTime/blob/64758b76247ccbb25c5ee84786b60d34618aa277/src/drivers/Bma421.cpp#L126

Today the PineTime arrived so I did a test and PineTime BMA accelerometer code is really too much sensitive. Comparison video. https://photos.app.goo.gl/pEP5nXSFnpNkJMQcA

We can solve that by different finite state machines and "AI" but the simple average filter works actually good. It might turn on when you raise wrist with something in your hand, but it will not turn on when watches are flat and you bump into something. I noticed that in the BMA code there is some averaging filter in hardware. Not sure how it works or if it will need to increase that a bit.

drzacek commented 3 years ago

Averaging the last values does sound good. I only got mine last week and I'm still setting my build env. Once I got it working I'm gonna play around with stuff.

Riksu9000 commented 3 years ago

I have also 100Hz sampling and 2g range (reading over I2C seems like 10 Hz over my scope)

The sensor is polled at 10Hz in systemTask. This should be improved by enabling interrupts from the sensor, and reading the values into a small buffer. The frequency might not have to be much higher than 10Hz.

We can solve that by different finite state machines and "AI" but the simple average filter works actually good.

There's no need to make this complicated. The current wakeup code is very basic. With a small buffer and a bit higher update frequency it could easily be improved by requiring some velocity rather than looking at just the angle.

https://github.com/JF002/InfiniTime/blob/514481ef7f9c71ad816b31d979c6ab39ce9380dd/src/components/motion/MotionController.cpp#L12