In response to the oscillation problem of traditional Sliding Mode Observer (SMO) in the sensorless control of permanent magnet synchronous motor (PMSM), this paper adopts an improved Super Twisting Algorithm Sliding Mode Observer (STA-SMO) to estimate the angle and speed of permanent magnet synchronous motors and reduce the oscillation phenomenon. The improved sliding mode observer is based on the super spiral sliding mode observer, using saturation function instead of switching function as the switching function, and using inverse tangent function to calculate the rotor position, which weakens the low-speed vibration phenomenon of the super-spiral sliding mode observer and improves the low-speed dynamic performance and steady-state performance. In order to enhance the characteristics of both low-speed and high-speed domain, a Sliding Mode Controller (SMC) instead of a PI speed loop based on the previous model is adopted, which improves the dynamic performance of both low-speed and high-speed domains in terms of speed overshoot and rise time.