Hi @Mathias_Appel, we’re slowly learning that on the Galaxy wearable platform it may not be possible to have both accurate data and battery efficiency. We need sub-second accuracy on the sleep tracking data, which at this moment means getting data from the sensor every 500 ms.
Last year, the app has been suspended from the Galaxy store for battery inefficiency, because we were keeping the watch awake all night to be able to poll the sensor every 100 ms (which is ideal).
We then had to rewrite the app so it used sensor batching (a technique which allows the watch CPU to sleep while the sensor gathers data into it’s own internal memory and then passes it as a batch to the CPU every few seconds).
Efficiency of sensor batching heavily depends on the size of the sensor’s internal memory. We had no way to find the size beforehand for each watch model, so we wrote it in a way that the app may combine sensor batching and keeping the CPU awake if the sensor memory size is not enough for sleep tracking purposes.
Well, guess what - it’s really not enough on any Samsung watch. The internal sensor memory has only space for 300 values, which means that with using sensor batching exclusively, we can only get 2-second accuracy (mainly because we can only wake up the CPU and process the sensor data only every 10 minutes - that’s a hard restriction on the platform). We actually used the 2-second accuracy for a while, and that was battery efficient, but the data were distorted towards higher deep sleep. That’s the 10% battery drain you were seeing.
Is it correct that also the deep sleep percentage is lower for you with the new update?
So with Samsung, it’s a tradeoff between battery efficiency and accuracy, and it doesn’t really seem to be moving away
We can lower the accuracy a little within some sensible boundaries - maybe up to a second, which would halven the battery drain, but there’s not much space after that.