Report on the Timescale Algorithm Test Bed at USNO

A test bed is described for the evaluation of timescale algorithms. Test results for two algorithms (AT1 and KAS-1) using simulated data are presented. Both algorithms are optimum in the sense that, with the appropriate selection of clock weights, they are able to produce an output with minimum squared time prediction error in steady state for ensembles of clocks whose noises are all proportional to the noise of one of the members. The KAS-1 algorithm is also optimum during known transients such as startup, utilizes robust outlier detection, and has the property of time continuity. As a result of these improvements, KAS-1 is more suited to automated operation. Neither algorithm is optimum at all sampling times for ensembles of clocks having radically different noise processes. The KAS-1 algorithm and a clock simulator, both representing a new Kalman-filter approach, are documented herein. See also ADA224769. Presented at the Proceedings of the Annual Precise Time and Time Interval (PTTI) Applications and Planning Meeting (21st) held in Redondo Beach, CA on 28-30 Nov 1989. Published in Proceedings of the Annual Precise Time and Time Interval Applications and Planning Meeting Proceedings, p269-288, Nov 1989.