With ADXL345 stationary observe much more variation in output at 3200Hz data rate than 800 or 1600 Hz rate.
I'm reading data via SPI with SPI Clk = 4 MHz. (I am pretty confident in my SPI code: I can write ADXL345 registers and then read back the expected values, and I ran the self-test very carefully, (at 800Hz data rate) and measured the results to be in the expected ranges.)
For both experiments shown below we have: FIFO_CTL = 0x90, DATA_FORMAT = 0x0b, POWER_CTL = 0x08. On scope measure > 5.2 microseconds between FIFO reads. I have a SW test for FIFO overflow and I am sure this is not occurring. Notice almost +- .5g variation at 3200 Hz. Everything in system is identical for both experiments except setting of BW_RATE. Note I see a similar pattern for X and Y data: much greater variation from baseline values at 3200Hz vs 800 Hz.
1.)Does output look 'reasonable/correct' for 800 Hz plot?
2.)Why so much more variation at 3200 Hz?
Experiment 1: BW_RATE = 0x0f , (3200Hz).
Experiment 2: BW_RATE = 0x0d, (800 Hz)