Experimental Surface Roughness Issues
Validity of Acceleration Values
I'm puzzled by the vertical acceleration figures produced by my phone. They are centred around +9.8 m/s - which seems OK.
But the traces are not symmetrical - the negative excursions are much larger than the positive ones - so in the curve below there are peaks around -15.0 in the negative direction (i.e. -24.8 below the average) but only 18.0 above - (i.e. 8.2 above average).
I would have expected that when you go over a bump you would get a very large vertical acceleration which would add to g - but if you went airbourne your acceleration would fall to zero (i.e. falling weightless under the influence of gravity). I can't see why you should get large negative accelerations (i.e. downwards).
I've tried dropping my phone - and it seems to correctly measure 0g when in the air - and large positive accelerations when being stopped by a cushion.
The diagrams below show parts of the above curve expanded in time to show the individual acceleration observations (marked with crosses).
Each example shows excursions below 0g - but without any corresponding positive excursion of similar size.
Android is not a real-time operating system. Therefore at times other tasks may prevent collection of accelerometer data.
Individual measurements collected over a 2:56 (176 second) period were analysed. The adjacent table shows the number of times adjacent measurements are spaced by N milliseconds. Thus 94.6% of measurements are spaced at 8, 9, 10, 11 or 12 milliseconds.
In this period there was one outlier of 724 milliseconds followed by about 70 measurements which claimed to be at 0 millisecond intervals. I don't know whether the system is buffering the acceleration measurements and subsequently writing them out quickly - or whether there was a genuine 'gap' of 0.724 seconds.