Using uncombined (UC) or undifferenced original observations has become a research topic in the global navigation satellite system (GNSS) community. This far, precise orbit determination (POD) with ionospheric-free (IF) observation combination has been favored. Using the uncombined strategy, i.e., using the raw observations to perform POD, the slant ionospheric delay is regarded as an unknown epoch parameter. We assume that the slant ionospheric delay depends on the frequency having the same magnitude for pseudoranges and carrier phases but opposite sign. Since there are, at each station and each epoch, as many ionospheric delay parameters as the observed satellites, a large number of ionospheric parameters need to be estimated for uncombined POD. The usual approch is that, when the observations at an epoch are accumulated, the epoch parameters consisting of clock and ionospheric parameters are eliminated simultaneously, which results in low computational efficiency. We develop a high-efficiency method by eliminating ionospheric parameters one by one, where each of them is eliminated when the observations of a station and a satellite are accumulated and no matrix inversion is required. In this uncombined POD algorithm, the IF-combined clock datum adopted by the international GNSS service (IGS) clock products is used to keep compatibility with existing IGS products by using the re-parameterized observation equations. We deduce the ambiguity resolution strategy for uncombined POD. In the experimental part, the modified parameter elimination method is first validated. With GPS observations from 58 global IGS stations, the first iteration processing time of parameter estimation for IF POD, UC POD and modified UC POD are 3.6, 62.5 and 9.2 min, respectively. The computational efficiency of modified UC POD strategy is greatly improved compared to UC POD. Then, the IF POD and the improved UC POD strategies with ambiguity float and fixed solutions are analyzed by comparing orbits, clocks, station coordinates and tropospheric delays with respect to IGS products. Results show that the accuracy of UC and IF solutions is comparable. Afterward, the repeatability of station coordinates is analyzed. With the solutions of 13 POD arcs, the repeatability over 13 days is consistent for UC and IF strategy. In addition, the ambiguity fixing rate, data usage rate, as well as residuals of code and phase observations, is analyzed. It is concluded that the slight difference between UC and IF strategy may be related to the data usage rate.
