Error analysis in a recursive identification algorithm

A method for evaluating the effect of errors due to A/D converter resolution and to the finite wordlength of microprocessors in digital algorithms is described with reference to a recursive identification algorithm. The proposed method is based on a theoretical error propagation analysis which furnishes useful expressions for evaluating the role that each error source plays in the overall performance of the digital algorithm. The results obtained were verified by means of two computer simulation approaches focused respectively on the validation of error propagation and of error models. Both the theoretical study and the simulation allow one to forecast the algorithm's performance, thus allowing the measurement and computation system to be optimized on the basis of both the component under analysis and the accuracy required by the particular application. © 1994.