linear acceleration inconsistency

[b.holzinger@gmx.at]

I was trying to measure the acceleration and deceleration of a tramway between two stops (on a straight flat route without any curves) using the "acceleration without g" experiment. From the acc data I calculated the velocity of the tram. Unfortunately, at the end of the measurement the calculated velocity is far from the expected value of zero. (The velocity v increases to about 9m/s, then drops back to around 6.5m/s)

Similarly, trying to reproduce this result in a small scale experiment, I manually accelerated and decelerated my phone along one axis (y) from standstill to some velocity back to standstill. Again, my velocity calculations dont end at a value of zero. (Velocity starts at 0, increases to about 2.5m/s, then decreases to -0.6m/s)

So, the areas under the acc-vs-time plot don't add up to zero.
Any ideas what I might be doing wrong? Thanks!

[Jens Noriʇzsɔɥ]

Numeric Integration is not a “well behaving” problem. The uncertainties in the integration constant could quickly add up and lead to the deviations that you observed, see also https://phyphox.org/wiki/index.php/Integ...celeration and the note on “linear acceleration” being a virtual sensor…

The results can be improved if you check for (and subtract) offsets at rest and apply some “knowledge” about the kinematics like velocity or acceleration being zero.