High-Performance Selective and Output Filter Techniques for Sensorless Direct Position and Speed Estimation

Abstract

This article introduces high-performance filter techniques for direct position and speed estimation to augment its robustness against disturbances. The direct estimation concept provides an independent position and speed estimate at each sampling instant by solving an optimization problem parameterized with the current, current derivative, and voltage of the same sample. It can operate at any speed employing a voltage injection at low speed or pulsewidth modulated current derivative. A selective filter concept is proposed that discards samples lacking robustness based on cost function properties. The concept is most effective in removing worst case errors and should always be applied. Also, output filter techniques are investigated to improve the estimates. A finite impulse response (FIR) structure is proposed that filters estimates according to a least-square criterion and is effective in reducing average estimation errors. The FIR filter is benchmarked against an enhanced dual phase-locked loop (PLL) filter, which is enabled by direct estimation. The FIR and dual-PLL filters are found to have a 6.8 and 1 kHz practical bandwidth, respectively, while achieving a $ 1% absolute mean position and speed estimation error. Hence, they perform one to two orders of magnitude better than traditional estimation schemes, which typically achieve $ 100 Hz bandwidth at similar errors.