Uncertainty estimation on multi-channel GPS time transfer

The GPS common view technique, using C/A code receivers, 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 (Comite Consultatif du Temps et des Frequences). This schedule contains 13-minute long, 16-minute spaced tracks, resulting each day in (ideally) 89 clock offset estimations. The usual method used to compute the daily average of clock offsets is a least-squares linear interpolation; the time offset uncertainty is obtained by the calculation of the daily standard deviation of the interpolated samples. This simplified estimate does not take into account the statistical properties of the different types of noise present in the measurement. We chose in this work to perform the Chebyshev interpolation in order to lead to a formulation of the time transfer uncertainty. This formula provides us with an estimate of uncertaintv as a function of the levels of noises present in the GPS tracks. The results we obtained showed that, in the case of complete data tracks, the empirical uncertainty estimation agrees well with the noise-based uncertainty calculation. In addition to validating the empirical method, our method provides a good estimation of the bounds of the daily standard deviation. However, the dispersion of the uncertainty estimation using the noise level indications shows a bias when data are not equally spaced, typically when scheduled track slots are empty. That is due to the change of the estimated noise level values when we consider the set of incomplete data. This bias can be cancelled by using a modified schedule filling all 16-minute slots, or be reduced, as we did, by using multi channel receivers with which we can take full advantage of the low dispersion property of the noise-based uncertaintv estimation method.