The dynamic Allan variance (DAVAR) is a tool that allows to understand if the stability of an atomic clock is changing with time. Since an anomaly in the clock behavior generates a change in its stability, the DAVAR can be used to monitor the performances of a clock. In this paper we present a fast algorithm for the computation of the DAVAR, which outperforms the classical computational method.
I. INTRODUCTIONThe performances of an atomic clock change throughout its life, due to aging, cyclostationary effects such as temperature and humidity, as well as sudden breakdowns. Therefore, also its stability changes with time. The dynamic Allan variance (DAVAR) [1], [2], [3] can be used to represent the change in stability of an atomic clock, and, in general, of any precise oscillator. Consequently, the DAVAR can be used to detect and identify clock anomalies [4], [5].A critical point of the DAVAR is its computational cost. The DAVAR is, in fact, a sliding version of the Allan variance [6]. To compute the DAVAR, at every time instant, the signal representing the clock phase deviation is truncated, and its Allan variance represents the clock stability at the given time instant. Therefore, if a signal has N samples, the computation of the DAVAR requires, in general, the evaluation of N Allan variances, which can turn in a computational burden if N is large.By using a recursive formulation of the Allan variance it is possible to formulate the DAVAR in a recursive way [7], [8]. Such formulation can be used to develop a fast computational algorithm, whose performances are far superior to those of the classical DAVAR algorithm. In this paper we give the recursive formulation of the DAVAR, and we show the performances of the fast algorithm.