A Sliding Mode Observer for Sensorless Operation of Low Power PMS Machines

Abstract

The Sliding Mode Observer (SMO) is a non-linear estimation method that is acknowledged to be robust against parameter variations and noise that persist over a control system. In this work, the comparison of three different implementations of the SMO is examined in the case of the estimation of the back electromagnetic force (BEMF) of a permanent magnet synchronous motor (PMSM). In particular, the classical signum function of a SMO is compared to a saturated linear function with steep slope and to a sigmoidal function. These functions are chosen in order to reduce the typical chattering phenomenon. Moreover, performance of the implemented solutions is discussed in terms of maximum achievable speed and current consumption. The tests are carried out using a PIC32MZ microcontroller and a PMSM with surface mounted magnets.