Fine frequency offset estimation for OFDM signals in frequency selective channels

In this paper we derive a fine frequency estimator for OFDM (Orthogonal Frequency Division Multiplexing) systems in frequency selective channels using an approximation to the ML (Maximum Likelihood) principle. Advantage is taken from the redundancy introduced by cyclic extensions -the common strategy used to avoid ISI (Inter Symbol Interference)-. In [1], Sandell, Van de Beek and Borjesson derived one such joint time and frequency ML estimator using cyclic extensions. In their work they assumed an ideal channel for the OFDM signal, so that performance is degraded in frequency selective channels. A reasonable approximation to the ML criterion is used to derive a fine frequency estimation algorithm based on the second order statistics of the received signal.