Robust frequency offset estimator for OFDM over fast varying multipath channel

Introduction: Frequency offset estimation is one of the most important task in OFDM receiver. Various frequency estimators have already been developed in the literature, including blind 、 semi-blind and data-aided algorithms. Showing great superiority in estimation accuracy、speed and complexity, data-aided algorithms attract much interest in recent years. The normalized frequency offset is usually divided into integral frequency offset(IFO) and fractional frequency offset(FFO), which may be estimated separately or jointly. Most data-aided algorithms have a limited estimation range and can only be used to recover FFO. Morelli proposed a time-domain preamble based algorithm, which utilized an OFDM symbol with multiple identical parts and can estimate the integral and fractional parts simultaneously [1]. However, the algorithm’s estimation range is restricted when a satisfactory accuracy is required. In [2]~[4], the differential information between two consecutive OFDM symbols is used to estimate IFO by assuming that the channel is invariable over at least two OFDM symbols duration. This condition is not met any more over fast varying multipath channel. In this Letter, two special CAZAC sequences are firstly designed for preamble purpose. The fractional part of CFO is then estimated and compensated in time-domain exploiting auto-correlation property of received preamble. Correlating two compensated preamble with local referential preamble in frequency-domain and making full use of the information of the locations corresponding to two correlation peaks, we may obtain the estimation of IFO. Due to the independence in searching for the peak location, the channel is permitted to vary independently