Abstract: A PPP algorithm based on GPS dual-frequency raw observations was presented, instead of forming the ionosphere-free code and phase observables like these traditional PPP models, the effects of ionospheric delays are modeled through including the Line Of Sight (LOS) ionospheric delay parameters into the state vector which is estimated using the Extended Kalman Filter (EKF) technique. This alternative algorithm could avoid the adverse effect of amplified mutipath and measurement errors existing in the combined observations, besides, several constraints can be introduced to the LOS ionospheric delay parameters during the filtering process，to improve the convergence of the filter to more reliable and precise solutions; The LOS ionospheric delays could be obtained simultaneously with other unknown parameters, which facilitate the research on ionosphere much; Besides, the recursive DIA-procedure is adopted as quality control strategy to guarantee the reliability of the estimators from our PPP implementations. The performance of our algorithm was evaluated for different positioning applications: static, low-dynamic and high-dynamic test, a position accuracy of about 1~2 cm and 7~8 cm can be achieved respectively for static and kinematic positioning, which demonstrate the feasibility and effectiveness of this new algorithm.