Performance comparison of composite clock estimates for two types of models

The paper compares the impact of two type of clock models onto the stability of the estimates and its time offset prediction error. The two types of clock models are the classical three state model and one which is extended by additional Markov states. The extension has got the capability to model Flicker noise which is dominant for averaging times longer than 1 day for different clock types, eg. Cesium clock. The performance of both models is compared by simulating a typical laboratory assemble consisting of two Active Hydrogen Masers (AHM) and three Cesium clocks. In particular, the prediction performance to time intervals up to two weeks is addressed since this is of relevance for UTC applications.