Sensors used in inertial navigation system needs good error modelling to eliminate the systematic error contribution as much as possible. Calibration approaches using data at multiple deterministic orientations and arbitrarily random orientation have both been developed and evaluated. One important argument for the final choice of random orientation approach is non-technical. I can easily do manual collection of calibration data when developing the parameter estimation algorithm. In mass production phase, calibration needs to be done in the out-sourcing electronic assembly factory in Shenzhen. Calibration job automation is a must for such skillful and time consuming task. I use this equipment to automatically collect sensor raw measurements, upload the data log file for post-processing parameter optimization. New firmware image customized for this particular unit is then generated, and downloaded to the device over-the-air. The workload of field worker is then minimized.