Reduced observer for anisotropy-based position estimation of PM synchronous machines using current oversampling

Abstract

Using field-programmable gate array (FPGA) and current oversampling, a novel approach for anisotropy-based position estimation of PM synchronous machines is presented. A least mean squares regression of the current samples is performed by an FPGA during the inverter's passive switching states to compute current slopes at active voltage vectors and evaluate them for position estimation. Using the predictive dead-time compensation as in [1], the inverter's nonlinearity effects are compensated with a high dynamic. The oversampling also creates a high signal-to-noise ratio in view of current measurement noise. Compared to [2], the novel approach of sensorless control is designed as a reduced observer in the estimated reference frame. This leads to high dynamics and a reduced parameter dependency. Experimental results show the high performance of the novel approach in closed loop sensorless control.