Modified LMS Adaptation Algorithm for a Discrete-Time Edge Equalizer of Serial I/O

Abstract

Discrete-time edge equalizers can enhance symbol-rate equalizers by compensating for inter-symbol interference at data transitions (timing ISI). The reduced timing ISI improves the timing margin and provides the CDR with clean timing information. This paper shows that adapting tap weights with standard blind LMS algorithm results in a reduced eye and may not converge. A modified algorithm is introduced to maximize the data eye.