Thanks Neil. I believe your last response goes directly to the issue I am experiencing.
I was aware of the offset null procedure described in the app note you attached. I now understand that the offset varies as the sample rate is changed and that there is only one set of registers containing the null values for each axis. This is a bit of a surprise as it is not well defined in the part specifications. I expected the offset to be constant across all sample rates.
We are calibrating each sensor axis in our firmware at ± 1G. The offset at 0G after calibration is very close to 0. This indicates that the slope for +1G side and the -1G side is symmetrical. Is this true? Does the slope of the sensor's output vary from range and or sample rate setting?
In our application we need to vary the ADXL345's sample rate in order to take advantage of the lower noise levels at slower user selected sample rates and the fast response at faster rates. This approach supports the various use cases our customer experiences in their science lab exercises. In one case low sample rates(higher resolution/low noise) are appropriate for incline measurement applications and in another case fast rates for collisions where noise and resolution are not as big an issue. Our customer is not in a position to re-null the part each time the sample rate is changed.
By the way I only noticed data in the 64K samples in the excel file for a sample rate of 1,600. I did not see data for any other sample rate.
At this point I have enough information regarding the offset vs sample rate to try to work around the issue in my firmware. Thanks Steve