Performance of order statistic LMS equalisers on stationary channels

This paper investigates the performance on stationary channels of a linear transversal equaliser (LTE) whose coefficients are adapted by the order statistic least mean square (OSLMS) algorithm, which was originally proposed by Cowan and Hwegi (1993). We insist that the order statistics based on absolute value operations should be used to determine the coefficient elements to be updated for the OSLMS algorithm. The OSLMS-LTE is extended to its decision feedback equaliser version and the performance is studied. The results show that the OSLMS based equalisers converge slower, but provide better bit error rate performance than the corresponding LMS equalisers, if the step size is equivalently set for both equalisers.