Hello,
I programmed a motion detection Application and everything seams to be running fine on a few watches.
On the Forerunner 735xt everything is running, however the motion detection gives results different from the ones expected.
Looking into the why i feal puzzled. Logging the sensor Data i find similar qualitative results as with other watches indicating that
the motion detection should work.
Looking for differences i pinned two watches together started the AccelerometerLogging on both watches on 12 samples/second and performed periodic motion.
I would love to show the graph but i am unable due to quota restrictions...
The result is that the FR735 starts falling behind in terms of the motion ( while still giving 12 datapoints to the logging)
Checking if this is a performance issue i do the same thing with other sample rates per second giving us the same result.
After several tests i see the same pattern everywhere:
When asking for AccelerometerData every second we get the correct number of datapoints however the time-intervall is shorter for the FR735 ( 12samples/0.99s). This leads to the FR735 fallling behind as it gives 0.99s of AccelerometerData to us every 1s. This cumulates over time leading to large discreapencies!
This theory is supported by the result that stretching the Data of the FR735 by mulitplying 1.01 with the x-axis we have perfect time-data agreement of the FR735 and FR935 (independent of sample rate):
The reason why i came up with this is because
a) i calculate frequencies of motion and the FR735 has shifted frequencies all the time
b) sometimes the algorithm messes up completely where i suspect that the continous shift of Data and realTime is set back to zero again.
Can anybody relate, Explain or show me a workaround?
