Data Aided ML Symbol Timing Estimation for MIMO Systems Using a Polyphase Filterbank

Abstract

This paper derives the maximum likelihood (ML) symbol timing estimator for a frequency non-selective multiple-input, multiple output (MIMO) channel assuming linear modulation and known data and known channel gains. The non-frequency selective fading assumption means the multipath delay spread is small so that the symbol time delay is the same across all receive antennas. Both a closed-loop and an approximate open-loop architecture are demonstrated. The open-loop or “batch” estimator is based on a polyphase filterbank implementation of the derivative matched filter.