An Auto-Regressive Moving-Average Time Scale Algorithm (ARMA) for Synchronizing Networked Clocks

Published

January 25, 2016

Author(s)

Judah Levine

Abstract

I will report on a study of the usefulness of ARMA time scale algorithms to synchronize clocks on a digital network. The algorithm acquires periodic time differences between a local system clock and a remote time server by means of any of the standard message formats such as the format used by the Network Time Protocol. It models the current time difference as a linear combination of previous time states plus additive noise and uses the model to adjust the local system clock. The algorithm is more flexible than the traditional methods, which are based on physical parameters such as frequency and frequency drift. The ARMA model has a finite impulse response and is therefore able to cope with the non-stationary outliers that characterize the fluctuations in the message delay on a wide-area network. I will compare this method with the frequency lock loop (FLL) algorithm that is currently used to synchronize the time servers operated by NIST. Both methods take advantage of the free-running stability of the clock in the local system, which facilitates the detection of outliers without the need to query multiple remote servers in most situations. Either method is generally more efficient than the phase-lock loop process that is widely used in network synchronization applications.

Proceedings Title

2016 Proceedings Precise time and time Interval Planning and Applications Meeting (PTTI)

Conference Dates

January 25-28, 2016

Conference Location

Monterey, CA

Conference Title

Precise Time and Time Interval Meeting

Pub Type

Conferences

Keywords

Automated Computer Time Service, Internet Time Service, Kalman Filter

Citation

Levine, J. (2016), An Auto-Regressive Moving-Average Time Scale Algorithm (ARMA) for Synchronizing Networked Clocks, 2016 Proceedings Precise time and time Interval Planning and Applications Meeting (PTTI), Monterey, CA (Accessed May 8, 2024)