Analysis of Short-Term and Long-Term Characteristics of GPS Satellite Clock Offsets

The GPS satellite has three or four atomic clocks that consist of cesiums and rubidiums and the NANU messages can be used to identify the kind of the onboard atomic clock because they classify the clock type on a daily basis. In this study, for long-term analysis of the GPS satellite clock behavior, we extracted satellite clock errors for every PRN from years 2001 through 2009 using the SP3 files that are provided by the IGS. As a result, the cesium clock offsets usually have a linear trend of drifting. On the other hand, rubidium offsets show curvilinear variations in general, even though they cannot be represented as anyone specific polynomial function. For short-term analysis, we extracted satellite clock errors for each PRN for a week-long period using the CLK files that are also provided by the IGS and curve-fitted them with first-order and second-order polynomial functions. In cases of cesium clock errors, they were well-represented by first-order polynomial functions and rubidium clock errors were similar with second-order polynomials. However, some of rubidium clock errors could not be represented as any polynomial fitting function. To analyze the characteristic of GPS satellite by each block and atomic clock, we applied Modified Allan Deviation criterion to the dataset from years 2007 and 2010. We found that the Modified Allan Deviation characteristics changed significantly according the block and atomic clock type.