Multiscale Clock Ensembling Using Wavelets

摘要

There are currently many different timescale algorithms in use today ranging in application from scientific to commercial. While these algorithms differ in many respects and are sometimes tailored specifically for the intended application and mix of clocks involved, they all share the common goal of optimally combining the clock error difference observed or measured between a collection of clocks to form a reference timescale that is more stable than any of the constituent clocks. Of these algorithms, only a few are well suited to collections of highly disparate clocks. A new approach to forming timescales is presented here. This new multiscale ensemble timescale (METS) algorithm is based on a multiresolution analysis afforded by the discrete wavelet transform, is not dependent on a specific model for the clocks involved, optimally utilizes any mix of clocks both in terms of type and capability, and results in a reference timescale that is more stable than the constituent clocks over all scales (averaging intervals). The METS algorithm is presented in detail and is compared in a simulation study with a well-accepted timescale algorithm that uses a model-based Kalman approach.