Second-order cyclostationary approach to NDA ML square timing estimation with frequency uncertainty

This paper addresses the problem of non-data-aided (NDA) symbol timing error estimation in presence of unknown deterministic carrier frequency errors. A general framework for the stochastic low-SNR ML estimation is derived considering a prior distribution for the carrier frequency error, and exploiting the cyclostationary properties of linear modulations. Finally, the well-known Oerder and Meyr (square timing) method is shown to become a particular case of this general solution.