A fast convergence algorithm for adaptive FIR filters under computational constraint for adaptive tap-position control

This paper proposes a fast convergence algorithm for adaptive FIR filters with tap-position control. The proposed algorithm consists of two stages: flat-delay estimation and constrained tap-position control. In the flat-delay estimation, the scattered coefficients are allowed to change their positions to achieve fast and correct flat-delay estimation. For constrained tap-position control, special attention is paid to a limit in computational power imposed by the hardware. By dividing a first-in-first-out queue into two parts, which store indexes to inactive taps with no assigned coefficient, fast convergence is achieved even when computation per sampling period for tap-position control is limited. Simulation results show that under the same computational limit as the conventional algorithm, the proposed algorithm reduces the convergence time by as much as 60%. The convergence speed remains unchanged for different computational limits. This algorithm is promising for echo cancellation in satellite links and in data transmission with modems.