Question about ADXL375 noise

Hi,

I have a 2 part answer, 1 for the 3200 Hz case and the other for the 100 Hz ODR.

For the 3200 Hz ODR data:

The noise in the plots appears to be dictated by large magnitude spikes. Assuming the sensor is mounted to a stable surface, I would suggest you investigate 2 possible causes.  1. Power supply and 2. SPI communication.

Noise from the power supply can couple and affect the output noise.

Otherwise, the jumps in the data may be due to reading an MSB register for a previous sample while reading the LSB of the current sample. Are multibyte reads being done or are you using single byte reads?  Multibyte is better.  Are you left or right justified?  If you are left justified, the lower bits in the LSB register would correspond to about 1.5g steps if a transition is happening occasionally.  For example, if the previous data point has the lower bits of 1 0000 and the next sample is 0 1111.  These are only 1 bit different, but if a mis-read occurs, this could be read as 1 1111, which is about 1.5 g higher than either real reading.

For the 100 Hz ODR:

I would suggest power supply, device mounting or calculation method.  The data appears to be within +/-1lsb which would be 100 mg peak to peak.  Are you stating peak to peak noise or rms noise.  You should use rms calculations for noise.