I am using ADXL362 for a small-amplitude impact detection. I also use a PCB353B33 wired sensor as reference for comparison. I put sensors on a steel structure, and hit the structure using a hammer. The amplitude of the impact acceleration is smaller than 8g.
But the data obtained from ADXL362 is not reasonable. Here are the configuration,
ADXL362: 8g measurement range, 100Hz sampling frequency
PCB353B33: 50g measurement range, the initial sampling frequency is 8192Hz, but I then downsample data to 100Hz for comparison.
However, the data from ADXL362 has a much larger peak compared to PCB353B33 sensor at the onset of impact event (see attached figure), and we trust PCB sensor. So, I am writing to ask what are the issues behind this problem and how can we mitigate it?