Abstract:The dynamic Allan variance (DAVAR) is a new method to analyze the random error of output signals. However, it has defects such as poor confidence on the estimate and great estimation error, due to the reduced amount and freedom of data captured by the window. To solve this problem, a dynamic total variance approach is proposed to track variations in the signal and reduce estimate error. Firstly, we truncate the original signal with a fixed window by fixed step. Then, the data in the windows is extended by an inverted mirror mapping method to improve the confidence level. Finally, the Allan variance values of the extended data are calculated and expressed by two or three-dimensional figures. Simulation data shows that the dynamic total variance approach can track the dynamic variety of signals quickly, and also overcomes the problem of low confidence level and high estimate error.