A resilient Kalman filter based servo clock

The accuracy level achievable with a given synchronization system is straightforwardly related to timestamp accuracy. Therefore, different methods have been proposed in the literature for reducing measurement uncertainty in synchronization systems. However, another important problem is deciding whether a received time information is trustworthy, particularly when timing messages generated by a supposedly reliable reference, point to significant discrepancies from the local clock. This paper presents a clock servo purposely designed for working with timestamps affected by sporadic outliers. The proposed algorithm is based on a Kalman filter (KF), whose correction phase has been modified so that the calculation of the a posteriori state vector estimate depends on the reliability of time offset and frequency deviation measurements, as assessed by an innovation-based outlier detector. To achieve greater robustness, two KF clock algorithms are run in parallel, a back-up servo providing usable state estimates whenever measurement information is withheld for too long times from the primary servo through being considered unreliable. The resulting composite algorithm features a much improved stability range, while retaining a very good accuracy.