Transputer/A100 adaptive signal processing systems

This paper presents implementations of adaptive signal processing applications on the noncanonical A100 FIR filter using controlling transputers. These applications when performed using the least mean squares (LMS) algorithm, generally use a canonical transversal filter with feedback of an updating error signal to iterate the filter weight vector according to the LMS algorithm which, in the mean, yields the Weiner-Hopf solution. This paper shows that it is not possible to implement LMS algorithm for the noncanonical FIR filter. Using both mathematical analysis and cut-set graph theoretic techniques it is shown that, although similar to the LMS, what is realised is a variant which is labelled the noncanonical LMS (NC-LMS). Inverse system modelling and beamforming using the NCLMS have been implemented on the transputer/A100 system and are presented in this paper