Optimization of MEMS-Gyroscope Calibration using Properties of Sums of Random Variables

We report a new approach for the problem of the automated calibration of offsets of a MEMS-gyrospcope. The developed method is a fast, direct and memory-efficient approach to the problem. The method uses statistical information from the time series of gyroscopic sample data and combines it with the desired fractional accuracy of the gyroscopic offsets. We show that by using the statistical information contained in the time series the process of calculating the gyroscopic offsets can be significantly optimized. We applied the method proposed in this paper to existing data used in generic algorithms. We have been able to speed up the calibration of gyroscopic offsets significantly. For the data used to benchmark against in this publication our proposed algorithm was about 50 times faster. The method also works on-the-fly in that the statistical information needed is updated every time a new measurement is available. This implies that our suggested method is also very memory efficient making it especially useful when re-calibration needs to be done often or on-chip. We have developed and published a C/C++ code which enables the application of our method to various possible application scenarios.