Last night I finished the most rudimentary activity tracking proof of concept for Sensor Watch Pro: it counts the number of orientation changes the watch experiences per hour, and the number of times the accelerometer was “awake” at the top of the minute. This adds 600 nA to the power budget, and the data strongly hints that sleep tracking should be possible. Sensor Watch data at left; Oura data at right. Note that period of Deep Sleep, and that period of low motion counts from 1:00-3:00 AM…
right now an orientation change is rotation in excess of 50°, and “awake” means that the accelerometer experienced acceleration in excess of 0.75g within the last 12 seconds. For debugging purposes, I’m having the watch chirp when the accelerometer wakes up and falls back asleep (also noted by the “S”leep and “A”wake indicator at top right). Lots of dialing in we can do: we can set four different options for angle, and any acceleration threshold. We can also adjust the data rate…
Currently sampling at 1.6 Hz, which means it can miss some quick transient acceleration events. We could up the data rate to 12.5 Hz to catch more events without burning too much more power. We could also up it to 25, 50, or 100 Hz and probably still come in under 10 µA. It can go faster, and of course we could get actual acceleration values over the I²C bus for activity classification. But for now I’m trying to see how much more we can do with less.