The estimation of rotorpositionangle was the key to the vector control of the sensorless permanent magnet synchronous motor (PMSM), and its precision will directly affect the performance of the control system. The rotorpositionangle estimation method based on the flux observer was easy to realize, but the system control performance was limited by the dynamic performance of the flux linkage estimation. To improve the running performance of sensorless PMSM vector control system, an improved lowpass filter algorithm featured in the exchange of filtering and compensation order was studied, and the feasibility and effectiveness of the improved method were proved by using MATLAB simulation.