Uncertainty estimation on multi-channel GPS time transfer

In GPS time transfer, the linear least-squares interpolation is the usual method to estimate the daily time difference between two time laboratory clocks, as deduced from a set of up to 89 individual measurements. This daily estimation is issued with its associated uncertainty computed as the daily standard deviation of the residuals of the samples to the interpolated function. In this work, we chose to perform the Chebyshev interpolation in order to lead to a formulation of the time transfer uncertainty according to the noise levels contained in the GPS time differences. Experimental results show a good agreement with simulations when data are complete. They also proved that the noise-based uncertainty estimation over a few days is a better indicator than the daily uncertainty estimation. However, we noticed, on other hand, that in the case of real time offset samples and also in case of generated data with planned losses of tracks, the uncertainty was underestimated. This discrepancy is significantly reduced when using the multichannel mode.