On the convergence of the order-recursive adaptive lattice equalizer

A quantitative evaluation is presented of the performance characteristics of the ORAL (order-recursive adaptive lattice) algorithm. In particular, the authors study the initial convergence, the steady-state performance, and the wordlength effect. The initial convergence analysis shows that, in contrast to the conventional MTO (mixed time and order)-recursive LS (least-squares) lattice algorithms, which are sensitive to the choice of the initialization procedures, the ORAL equalizer initialization is determined by the second-order statistics of the input data, so that there is no arbitrary choice in initializing the algorithm. The steady-state analytical and experimental comparisons show that the ORAL equalizer steady-state performance is in closer agreement with the theoretical expression than the MTO-recursive LS lattice algorithm. Simulation results demonstrate that the ORAL equalizer can work by using shorter wordlength than the recursive LS equalizers