Contents I. Introduction
As is well-known, the precise information of rotor position and speed are necessary for a PMSM servo system using vector control method. Usually these information can be measured by a sensor such as magnetic resolver, optical encoder. Unfortunately, sensors may cause undesirable, maintenance problem, besides it also add extra cost. For these reasons, many approaches for position and speed estimation have been developed over recent years. Generally the sensorless approaches can be classified into two main strategies: saliency and signal injection method [1]–[3] and fundamental excitation method [4]–[12]. Fundamental excitation methods are based on fundamental model, which detects the rotor position from the stator voltages and currents. In [11]–[13] a method using sliding mode observer (SMO) based on estimated rotating reference frame is proposed, which can estimate the rotor information with initial rotor angle uncertainty effectively. Unfortunately, the estimated speed and rotor position carries the switching noise because sliding mode method. Although we can reduce the signal to noise ratio by adjust the observer gain properly, in the practice we must set the observer gain parameter larger for robustness. In [13], the Extended Kalman Filter (EKF) method was adopted to retrieve the speed and position from the SMO, but we can find that a large lagging occurs to the estimated speed, besides the EKF algorithm is too complex. In the paper these problem will be solved. The simulation with Matlab/Simulink will give the results later.
