Non-Data Aided Symbol Timing Estimation in MIMO Systems

We present two maximum likelihood (ML) based estimators for non-data-aided (NDA) symbol timing recovery in MIMO systems. These estimators are based on the classical unconditional ML and the stochastic ML (SML) methods. The proposed estimators utilize information about the particular space-time code used and give performance comparable to data aided estimators, though for a relatively higher complexity. An approximate version of the SML estimator which requires lower implementation complexity is also presented. The loss in SNR due to timing estimation error is also analyzed.