A Simple and Effective Method for Abnormal Data Processing of Atomic Clock Time Difference

Atomic clock time differences are most important original measurements for atomic time calculations of time-keeping system, whose qualities affect the continuities and stabilities of time scales directly. In actual measurement, the original data often appear to have gross errors such as phase jumps, frequency mutations influenced by measurement equipments and environment. How to find out and eliminate the abnormal data is essential. A simple and effective abnormal data processing method is proposed in this paper. The original data is one or two rank fitting according to its characteristics, then the fitting residual error can be obtained after deducting one or two ranks. The residual data show normal distribution, which is mainly composed of white noise. The 3σ method cannot effectively eliminate gross error when the distribution is very scattered. So this paper proposes an iteration method for gross error elimination using residual data. The phase jump and frequency mutation can be detected by the standard deviation of clean residual data and statistic value, it can be verified by atomic clock time difference predicting data. The method has stronger specific aim and practicability comparing with some theory of error processing method, and has been achieved good results in standard time-keeping and on-board atomic clock data processing.