Analyzing the stochastic properties of code observation using various low-cost GNSS receivers

Abstract The knowledge of stochastic properties low-of cost receivers’ observations is still incomplete. This gap motivates us to comprehensively analyse the quality of GNSS observations recorded by the current affordable receivers. Here, we focus on the GPS, GLONASS, and Galileo code pseudoranges from u-blox ZED-F9P, Septentrio Mosaic-X5, and Skytraq PX1122R, acquired with antennas of different classes. As a benchmark, we use the measurements from the Trimble Alloy geodetic receiver. The investigations are based on residual code data error extracted from detrended multipath combinations series and consist of three parts: the root mean square (RMS), detection of frequency components and autocorrelation. Based on the RMS values, we reveal the high potential of the low-cost pseudorange data from Septentrio Mosaic-X5 and u-blox ZED-F9P. The former device even outperforms the geodetic benchmark receiver. On the other hand, the level of observation noise for the Skytraq is strongly amplified and it usually exceeds several decimeters, even for the zenith-referenced data. While these patterns are observed for both antenna scenarios, the degradation of quality for a patch one is evident for all receivers and it usually reaches several dozen percent. An explanation of low pseudorange noise for the low-cost receivers is provided by the frequency analysis. The spectra depict a thermal noise reduction for most of the multi-system data. Such an effect is likely driven by internal data smoothing algorithms. A consequence of this would be an increase of autocorrelation, which was indeed observed for the low-cost receivers in our analysis. To clarify the temporal dependence of the low-cost measurements, we performed an additional zero-baseline analysis, eliminating all error sources except the ones induced by the receiver. It proved the lack of autocorrelation for the geodetic receiver and its existence for all employed low-cost instruments, which follows the suspicion of internal data smoothing.