Parametrization of algebraic differentiators for disturbance annihilation with an application to the differentiation of quantized signals

Abstract

A method for the systematic parametrization of algebraic differentiators is introduced. It allows the annihilation of disturbances having transfer functions exhibiting repetitive peaks. Using McMahon's expansion for large zeros of Bessel functions, an approach reducing the number of free parameters of the differentiators from five to one is derived. The choice of the parameters is discussed in detail, and the existence of the parametrization for any such disturbance is proven. Error bounds for the annihilation of the repetitive peaks are also derived. The practical applicability of the approach is demonstrated in an experimental case study where the derivative of a quantized signal is numerically estimated using only an algebraic differentiator. A deterministic model for the quantization error which shows repetitive peaks in its transfer function is also proposed.