6D Sensor Fusion Algorithm Problem

[tmctwo]

Hi all,

I did some tests with NXP Sensor Fusion Library on my custom hardware. The library works pretty well under most conditions. But when I move the board forward and backward in less than 1 sec - periodically to test the behaviour under vibration (especially the board in a diagonal position to the surface), sometimes I see abrupt changes at output. Please find attached the pictures. In the pictures, the blue line is output from NXP fusion library; the red line is output from a commercial inclinometer. The output from NXP lib. somewhat goes to a point  far away from real position. In other cases it tracks the red line closely. I tested my custom hardware with another algorithm to verify it; I had no problem with it.

I used NXP 6D Acc+Gyro algorithm for testing. The NXP lib fusion rate and sensor sampling rate is 200Hz. I also tried with 400Hz sampling rate but no change. May a software defect exist which cause incorrect output in the NXP Sensor Fusion algorithm?