The master’s thesis addresses the problem of algorithmic correction of eccentricity in magnetic encoders, which arises due to mechanical imperfections during mounting and manifests as a sinusoidal angular position error in measurements. The goal is to estimate the parameters of the sinusoidal function, namely amplitude, phase, and offset, from noisy discrete samples. The problem is formulated as parameter estimation of a sinusoidal model from an overdetermined system. For real-time parameter estimation, the recursive least squares (RLS) algorithm is used, enabling efficient processing of sequential measurements. Due to implementation on an embedded system, the entire procedure is designed using fixed-point arithmetic, with particular attention given to numerical limitations. For the computation of the atan2 function, which is not natively supported on the target processor, the CORDIC algorithm is employed. The proposed approach is validated through numerical experiments on noisy sinusoidal signals, where the accuracy and convergence of the estimated parameters are analyzed.
|
