Levenberg-Marquardt method based Precise Angle Estimation for Eccentric Magnetic Absolute Encoders*

This paper proposes an improved eccentric magnetic absolute encoder (EMAE) based on the adaptive linear neuron (ADALINE). Typically, an EMAE can determine the absolute angle by characterizing signals based on the eccentric rotation of a magnet. However, these signals are often affected by disturbances resulting from non-uniform rotation, DC offsets, amplitude mismatch, phase differences, and low- and high-order harmonics. The proposed encoder adaptively estimates and rejects these disturbances using ADALINE based on the LevenbergMarquardt method. A simple simulation and experiment are used to demonstrate the effectiveness of the proposed method.