Uncertainty Estimation on GPS Time Transfer

The traditional GPS common view technique, using C/A code receivers, is the main time transfer method used by various timing laboratories over the world. Moreover, this method is used to realize the TAI (Temps Atomique International) and the TA (F) (Temps Atomique Francais). Clock offsets between laboratory clocks are determined according to a fixed procedure defined by the CCTF (Comit Consultatif du Temps et des Frequences). Using this procedure, one can perform on average 54 tracks per day (theoretically 90 tracks per day), providing clock offsets. Each of these clock offsets results from one 780's track and is obtained as a result of a quadratic regression, followed by various model-based corrections and finally a linear regression. The clock offsets are then issued with their standard deviations. However, this simplified estimate does not take into account the statistical properties of the different types of noise present in the measurement. We propose here to rigorously estimate this time uncertainty for various types of noise that characterize the transmitted GPS time offset data. This is achieved by the calculation of the covariance matrix of time samples. This method provides us with the variances of the drift coefficients and of the residuals, in the case of a linear drift model for 1-day sample sets, taking into account the different types of noise. In this paper, we compare the obtained results with simulated and real data over several days.