Abstract:It is imperative to consider analysis accuracy and computational time of every algorithm. When it is used to analyze random error, the computational time of total variance is very long. Aimed at this problem, this essay analyzes the Allan variance and the total overlapping Allan variance approaches. Three sampling modes are put forward. We find the origin of long computational time of total variance, based on which half overlapping total variance is proposed. Finally, the new algorithm is tested with a static experiment of inertial measurement unit including optic-fiber gyro. The result shows that analysis accuracy of half overlapping total variance is the same as with total variance, but computational time of half overlapping total variance is less than total variance.