OFDM frame synchronization for dispersive channels

We address time synchronization in an OFDM system and present an ML estimator that also is unbiased for dispersive channels. The estimator uses the redundancy introduced by the cyclic prefix and knowledge of the channel. Previously presented estimators based on the cyclic prefix are either biased or have high complexity for this environment. We show that with proper choice of signal model, the complexity can be reduced considerably while maintaining the performance. Part of the novelty lies in the proper decomposition of correlation matrices that appear in the ML estimator. The new estimator structure consists of a linear filter, followed by an auto-correlating filterbank and a maximization.