I believe there is something ‘wrong’ with the ADIS 16228 data quantization, here is why.
When we record data in real time mode (dynamic range of +/-20g), we observe that numbers actually ‘abruptly’ change by counts of 30 or 32 (in unscaled integers), even when we play input vibrations that are smooth. So we performed this additional test: we recorded “background noise” only. That is, we left the device on an isolated optical table (no vibrations) and observed the recordings. The result was VERY strange: noise values only seem to take on one of 6 numbers at most, and consecutive values are separated by counts of 30.
We changed the capture mode to Manual Time Capture, and changed dynamic range settings, but observed the same issue: noise (or background) recordings only fluctuate between 5-6 different values. Statistically, this makes no sense, not for a 16 bit quantization capability.
I attached here a plot that shows you noise/background recordings for different dynamic range settings (figure title is the dynamic range). You can see (1) noise only takes on only a few values, (2) values abruptly jump by counts of 32 for +/-20g, 64 for +/-10g, and so on, and (3) changing the dynamic range does not improve the ‘resolution’ or the number of observed values.
We tested 4 different devices with same results.