I am trying to determine the theoretical accuracy of an angle calculated with the lsm6ds3 accelerometer. From the spec I see there are a number of sources of inaccuracy/noise. There is noise density that is dictated by range and sampling frequency, zero point change due to temperature and zero point accuracy. Of these, zero point accuracy is listed as a constant value of +/- 40mg. There is no mention of any factors that can be used to change this value for the better (e.g. lowering the range reduces the noise density).
When we determine the angle we use the arctan of two of the readings. For the 45-degree worst case that means we would get a theoretical measurement of x = 0.707g and y = 0.707g. If I assume worst case zero point accuracy only on both of these values (+/- 40mg), one positive and one negative, I get x = 0.747 and y = 0.667. That gives me a calculated angle of 41.76 degrees which is an error of -3.24 degrees. For my application this is unacceptable.
Am I correct to assume that we can calibrate the zero point accuracy out of the part after production? How accurate should we be able to get the zero value in mg?
Thanks in advance for your help.