I think I understand that the minimum of the Allen variance is the "in-run bias stability", measured with the benign conditions of ideal averaging, stable temperature, and no inertial motion. For the ADIS-16490, for example, that's 1.8 deg/hr, given as a typical spec and the average on the Allen chart.
First question: does that mean that 67% (one sigma) of ADIS-16490 devices will show a heading deviation (integrated deg/s output) of up to 1.8 deg after an hour, if they're sitting still? This would have to include bias effect too, then, not just noise.
Next question: how do I determine the real-world error over temperature which I might expect? It seems like sensitivity, nonlinearity, and bias effects like repeatability and tempco errors would have to be added to the in-run bias stability figure.
And finally: what should I do to get the very best performance? Operate in a constant-temp oven? Do system calibration for temp? For sensitivity (i.e. over the full range of input rotation speeds)? What are the repeatable error factors I can calibrate out?
Thanks very much for any help,