Adaptive Volterra channel equalisation with lattice orthogonalisation

Abstract

The authors present a new recursive least-squares adaptive nonlinear equaliser based on Volterra series expansion. The main approach is to transform the nonlinear equalisation problem into an equivalent multichannel, but linear, equalisation problem. Then the multichannel input signal is completely orthogonalised using sequential processing multichannel lattice stages. With the complete orthogonalisation of the input signal, only scalar operations are required, instability problems owing to matrix inversion are avoided, and good numerical properties are achieved. Avoidance of matrix inversion and vector operations reduce the complexity considerably, and make the filter highly modular and suitable for VLSI implementation. Several experiments demonstrating the performance under different channel distortion and channel noise conditions are also included in the paper.