In order to improve the performance of precise point positioning (PPP), this paper presents a new data processing scheme to shorten the convergence time and the observation time required for a reliable ambiguity-fixing. In the new scheme, L1 and L2 raw observations are used and the slant ionospheric delays are treated as unknown parameters. The empirical spatial and temporal constraints and the ionospheric delays derived from a real-time available ionospheric model are all considered as pseudo-observations into the estimation for strengthening the solution. Furthermore, we develop a real-time computational procedure for generating uncalibrated phase delays (UPDs) on L1 and L2 frequencies. The PPP solution is first carried out on all reference stations based on the proposed scheme, undifferenced float ambiguities on L1 and L2 frequencies can be directly obtained from the new scheme. The L1 and L2 UPDs are then generated and broadcasted to users in real-time. This data product and also the performance of the new PPP scheme are evaluated. Our results indicate that the new processing scheme considering ionospheric characteristics can reduce the convergence time by about 30 % for float kinematic solutions. The observation time for a reliable ambiguity-fixing is shortened by 25 % compared to that of the traditional ambiguity-fixed kinematic solution. When the new method is used for static reference stations, the observation time for ambiguity-fixing is about 10 min in static mode and only 5  min if the coordinates are fixed to well-known values.