NEXT cancellers using FDLMS filters with improved convergence rate

A low-complexity technique for improving the convergence rate in a near-end crosstalk (NEXT) cancellation system is proposed. The cancellation system, minimizes the computational load by first detecting the lines causing the major NEXT and then assigning FDLMS adaptive filters to cancel the NEXT. By initially making the step-size of each adaptive filter proportional to the magnitude of the NEXT it is supposed to cancel, the convergence rate is significantly improved at almost no extra computational cost, especially during the initial stage of adaptation. A high convergence rate during the initial stages is more critical as it reduces the latency time in xDSL systems, thereby resulting in a higher data throughput.