we encountered an issue with ADIS16364 accelerometer readings in a unmanned aerial vehicle experiment (UAV; 20 m wingspan, 60 ccm combustion engine, max. propeller revolutions approx. 6000 RPM).
The sensor is attached with the standard two screws to a supporting structure, which is then rigidly attached to the aircraft frame (no damping material used). We sample the data at 50 Hz and the internal ADIS16364 digital filter is set to limit the bandwidth to 16 Hz. The problem appears, when the propeller revolutions go above approx. 3000 RPM. You can see this behavior in the PDF attached, where the propeller revolutions and norm of the accelerometer readings (x, y and z) are shown.
We would like to ask, what do you think is the reason of the behavior (marked by green rectangle - "area of interest") where the ACC norm drops from 1 g to 0.45 g due to high propeller revolutions (please note, that the UAV was not moving or changing orientation in this period; it was a propeller and engine test done on the ground; actual flight follows and is marked by black rectangle).
We didn't observe such behavior in the gyroscope readings. Even when the gyroscopes were simply integrated to obtain Euler angles, there was no visible evidence of such distortion.
Do you have any suggestions, how this can be avoided?