Low Complexity Adaptive Step-Size Filtered Gradient-based Per-Tone DMT Equalisation

In this paper, a low complexity adaptive step-size filtered gradient-based scheme for per-tone equalisation is investigated for discrete multitone systems. A variant of the filtered gradient adaptive algorithms has been presented in literature for the normalised version of the orthogonal gradient adaptive algorithm which employs the mixed-tone exponentially weighted least squares criterion. A low complexity adaptive step-size approach is proposed by exploiting an estimate of autocorrelation between previous and present weight-estimated error to control step-size update. In order to ensure good tracking and convergence speed, the step-size is adjusted recursively for different per-tone equalisers. Simulation results show that the trajectories of step-size of the proposed algorithm converge to its own equilibrium despite large variations in initial step-size settings. The proposed algorithm achieves superior performance, faster convergence rate and lower computational complexity than the existing algorithms.