A performance evaluation of an IIR adaptive smoother on a dual transputer system

Abstract

This paper describes the performance evaluation of an adaptive smoothing filter implemented on a dual transputer system. The smoother was used for the adaptive smoothing of narrow band signals corrupted with additive white noise. The topology of the transputer network is such that a transputer handles the data I O by interfacing to an IBM-AT host computer and the adaptive smoothing filter is run on two target transputers—one to handle the digital smoothing filter and one to run the extended least squares (ELS) parameter identification algorithm. By careful arrangement of the task structure within the transputers the three processors can be run in parallel. The code for the transputers was written in occam. A comparison is made between implementations using T414 fixed point processors and T800 floating point processors.