Performance gap between archived real-time and post-processed precise products for multi-GNSS and multi-frequency PPP-AR

Abstract Real-time Precise Point Positioning (PPP) with Ambiguity Resolution (PPP-AR) is an increasingly important Global Navigation Satellite System (GNSS) technique for high-precision positioning, navigation, and other time-critical applications. Its performance depends strongly on the availability and accuracy of real-time satellite orbit and clock products. Although PPP-AR has been extensively investigated using both real-time and post-processed precise products, direct comparisons under a common processing framework remain limited across different satellite constellations, frequency configurations, and accuracy requirements. This study quantifies PPP-AR performance differences among selected archived real-time and post-processed precise-product streams through station coordinate estimation and examines how overall product quality affects PPP-AR convergence. A 31-day dataset from 249 globally distributed Multi-GNSS Experiment (MGEX) stations is processed using archived real-time products from Centre National d’Études Spatiales/Collecte Localisation Satellites (CNES/CLS) and Wuhan University, together with final post-processed products from Wuhan University. The archived real-time streams are processed offline under continuous correction availability to isolate product-stream effects from communication-dependent factors. Single-, dual-, and triple-constellation solutions are evaluated for dual-frequency and triple-frequency PPP and PPP-AR. The results show a clear difference between single- and multi-constellation solutions. For single-constellation PPP-AR, convergence remains strongly dependent on overall product quality. For example, for the solution using only the Global Positioning System (GPS), the ensemble 68.3rd-percentile vertical convergence time at the 10 cm threshold is about 26.5 min with real-time products and 15.5 min with post-processed products. In contrast, multi-GNSS integration substantially reduces this sensitivity. For the combined solution using GPS, Galileo, and the Third-generation BeiDou Navigation Satellite System (BDS-3), the corresponding ensemble vertical convergence times are about 6.5 min and 1.5 min, respectively. However, this mitigation mainly applies to moderate positioning requirements. Under stricter thresholds, the performance gap widens markedly. For the same GPS-only solution in the vertical component, the ensemble convergence time is about 13.0 min with real-time products and 10.5 min with post-processed products at the 20 cm threshold, but increases to about 71.0 min and 24.0 min, respectively, when the threshold is reduced to 5 cm. These results provide a reference for the PPP-AR performance of the selected archived real-time product streams under continuous correction availability. Under these favorable conditions, moderate centimeter-level performance is attainable, but the gap relative to final post-processed products increases as the accuracy requirement becomes stricter. Actual real-time performance may be lower or less stable because correction latency and stream interruptions are not included in this experiment.