Prediction Research Of Time-frequency Standard Performance On Navigation Satellite

In modern satellite navigation system, the relationship between time-frequency andlocalization is especially outstanding. In a satellite navigation system, the localization ofa user receiver relies on the high precision spatial reference and time reference. Also itslocating requires the use of high precision satellite clock prediction parameters. With thesatisfaction of the high precision time synchronization, satellite clock bias prediction isof great importance in the localization and the time service of navigation system.This paper focuses on the study of the prediction of time-frequency standardperformance (satellite clock bias). Through analyzing quadratic polynomial model, graymodel and kalman model, the application and limitations of each algorithm areproposed with comparison. The abnormal conditions of the satellite clock includingfrequency modulation, phase modulation and clock switching are emphasized. Throughthe analysis of time-frequency characteristics and the study of historical operatingparameters, the possible abnormal conditions can be effectively identified with thesignal output rules and the knowing of satellite clock’s performance. By analyzing andclassifying the abnormal situations, the rapid clock bias prediction of satellite clockafter recovery is achieved and its algorithm is completed with matlab software.According to the compass navigation system of China, satellite clock bias required forreal-time positioning can be predicted quickly after the clock’s recovery.The prediction algorithm aimed at the rapid clock bias prediction after the clock’srecovery is that analyze1hour data to build model and predict2hours clock bias within2ns precision. A new periodic linear model is built with the analysis of IGS clock data,and it is considered feasible after the analysis of its predicting accuracy. Finally, thisrapid clock bias prediction algorithm is also tested with our rubidium clock in China,which has a better predicting accuracy with an average1.8ns in two hours.