State Estimation Using UFIR and Kalman Filtering in Applications to Local Crystal and Master Clocks

摘要

To meet the needs of the IEEE Standard 1139-2008, this paper suggests using the recently designed iterative unbiased finite impulse response (UFIR) filtering algorithm to estimate clock state via measurement of the time interval error (TIE) on an interval of N most recent past points. The algorithm has the Kalman filter (KF) form. But, unlike the KF, requires neither the noise statistics nor the initial values. Because noise in clocks has colored Gaussian components (not white, as required by the KF), the UFIR filter demonstrates higher robustness and becomes optimal when N1. Applications are given for state estimation in an ovenized crystal clock and error prediction in a masted clock. Based upon extensive experimental investigations, we show that the UFIR algorithm outperforms the KF, because the clock covariance matrix cannot be specified correctly in view of the colored noise.