Framework for sensitivity analysis of industry algorithms for sensorless PMSM drives

The Permanent Magnet Synchronous Machine (PMSM) is gaining popularity in many applications due to its high efficiency and high power density. In order to achieve efficient and robust motor operation, motor position information is required. If active compensation of motor torque ripples is to be achieved in addition to simply rotating the motor the quality of the position information becomes even more important. Traditionally this position information is acquired through the use of a position sensor, however so called `sensorless' control schemes are now becoming a popular alternative. This paper presents a review of the sensorless position estimation algorithms which are being adopted by industry. A framework is developed to assess the sensitivity of these algorithms to errors in motor parameters and non-ideal aspects of the motor and controller. This framework is based around a MATLAB®/Simulink® simulation which includes these non-ideal aspects. The simulated control loop is designed to be very close to the control loop implemented on a DSP and used to run a real motor. Results for a simple back-EMF estimator show that a 10-bit or better analog to digital converter (ADC) gives good performance, the position estimation error is almost insensitive to resistance variation and a 20% change in inductance results in a 0.7% offset error in the position estimate.