On our ADIS16490, the factory bias calibration for two of the gyro axes seems to be completely wrong. After doing a factory calibration restore, if the IMU is standing still, and we plot the outputs, one of the delta angle outputs is close to zero, and the other two are way off, apparently close to the numerical limits of the integrator, so if we move the IMU even a little bit, two of the outputs immediately wrap around (go from plus a few thousand to minus a few thousand, and sometimes back and forth).
If we now enable continuous bias estimation, wait a bit, and then do the NULL_CNFG thing to write the estimated biases to the user bias registers, then we get plausible results (all channels are almost zero if standing still and are only wrapping around during fast rotation).
Is it supposed to behave like that? Or is our IMU somehow broken?
We don't really want to do a bias calibration during each startup, since
1. customers would have to care about it and probably would have to keep the device still after turning it on, and
2. we're using it on UAVs, so if maybe the power is interrupted while in the air, it might start trying to do a bias calibration while flying around and get wrong results.