Comparison of Adaptive Algorithms for Cancellation of Harmonic Noise on Distribution Power Lines

WIGGS, JAMES H. Comparison of Ad8pti ve Algorithms for Cancellation of Harmonic Noise on Distribution Power Lines. (Under the direction of Dr. H. J. Trussell.) Adaptive digital filters have been used for many years in speech processin~ echo cancellation, and other areas. The ability of adaptive filters to remove harmonic noise from a contaminated signal, especially when the noise is slowly varying with time, is of special interest in the field of distribution power line carrier communications. This thesis compares the effectiveness of three of the more well-known adaptive digital filter algorithms at removing 60 Hz harmonic noise from an actual distribution power line noise sample: the Widrow-Hopf Least Mean Square (LMS) algorit~ the Least Square Lattice (LSL) algorit~ and the Fast Kalman algorithm. The algorithms are compared in terms of convergence rate, overall noise power reduction, and the ability to reduce the bit error detection rate (BfR) of phase-shift-keyed digital data in the noise. Results indicate that the LMS algorithm. while the slowest to converge, has the best BER performance. It is shown that the performance of the LSL and Fast Kalman algorithms is strongly dependent on the value of the misadjustment parameter; a value of .01 for this parameter causes very poor BER performance. while a value of . 1 causes the algorithms to perfonm almost as well as the Lt1S algorithm, but with much faster convergence rates.