Fractional Difference Prewhitening in Atomic Clock Modeling

摘要

A new approach to atomic clock classification has been enabled by the application of long-memory, or fractionally integrated, noise constructs. While the spectral properties of the long-memory noises are consistent with the historical 1.f(alpha) approach, they also allow a range of estimation strategies, in both spectral and time domains, for the classification of atomic clock behavior. These fractionally integrated noises are analyzed and applied to atomic timescales in this research, with particular emphasis on a prewhitening technique using fractional differencing which allows the separation of clock noise autocorrelation from clock rate and drift. Results from simulation studies show the utility of the fractional differencing approach both for simple fractionally integrated processes, and more complex processes which are more characteristic of atomic clock noise.