Application of the Normalized LMS Algorithm to MSLC

Abstract

For some conditions, the convergence time of the conventional LMS algorithm is longer than the radar dwell time so that the MSLC technique does not cancel jammers. The convergence time depends upon the power and angular locations of the jammers. The normalized LMS technique can decrease convergence times by several orders of magnitude. It causes the convergence time to be essentially independent of the jammer powers and decreases the convergence time dependence upon location. The derivation of an exact first moment analysis of the weights for the normalized LMS algorithm is given. Simulation results are also given. There is excellent agreement between the two.