An improved CORDIC algorithm for PMSM position estimation based on linear magnetic sensors

Abstract Linear magnetic sensors (LMSs) are widely used for rotor position estimation in permanent magnet synchronous motors (PMSMs). Limited by non-ideal factors, additional compensation algorithms are required to improve the accuracy. However, in most existing methods, an accurate rotation speed estimation or minimal installation deviation is essential, which is usually difficult to realize in practical servo systems. To solve these problems, an improved co-ordinate rotation digital computer (CORDIC) algorithm is proposed to achieve accurate position estimation directly using raw LMS signals without requiring speed estimation or structural modification. In the proposed algorithm, orthogonal LMS signals are placed on a two-dimensional plane for decoding. Vector synthesis and rotation are employed to compensate for the signal harmonics, and the initial vectors are separated to eliminate the adverse effects of amplitude and phase deviations. This speed-independent estimation mechanism allows the improved CORDIC algorithm to be adapted to any rotational speed. To validate the proposed algorithm, experiments are conducted on a flywheel PMSM prototype with both two-dimensional and orthogonally arranged one-dimensional LMSs. Experimental results show that the proposed algorithm effectively compensates for many non-ideal factors and achieves an accuracy of ±0.005 rad (electrical angle). Compared with the commonly used phase-locked loop method, the improved CORDIC algorithm can maintain the estimation accuracy under variable-speed conditions.