Frequency Offset Estimation In OFDM Systems

Due to its high spectral efficiency and robustness against the multipath fading, orthogonal frequency division multiplexing (OFDM) has been adopted in many wireless communication systems. However, a major drawback of OFDM systems is their vulnerability to synchronization errors. Both the carrier frequency offset (CFO) and sampling frequency offset (SFO) can destroy the orthogonality among subcarriers and cause intercarrier interference (ICI). This problem is even more serious for multiuser OFDM systems, because the signals from different users are mixed together at the receiver with user-dependent CFO and SFO.In this dissertation, the problem of CFO and SFO estimation in OFDM systems is discussed. Both the single-user and multi-user systems are investigated. In addition, the idea of CFO estimation is applied to the problem of phase ambiguity in blind identifications.Firstly, the problem of integer CFO estimation is discussed. Conventional methods only consider the maximum-likelihood (ML) estimation of the integer CFO under the condition of ideal fractional CFO compensation. However, their algorithm performance may degrade seriously if there is residual fractional CFO. In this dissertation, a robust integer CFO estimation algorithm is proposed by taking the residual fractional CFO into account. Simulation results show that the proposed algorithm has better performance than conventional methods.For the fractional CFO and SFO, conventional joint ML methods require a two-dimensional exhaustive search, whose computational complexity is very high. To overcome this problem, a low-complexity joint ML estimator is proposed. It is shown that the CFO can be solved in closed-form. Then an approximate ML estimation algorithm for the SFO is developed by taking the second-order Taylor series expansion. Simulation results show that the proposed algorithm achieves almost the same performance as existing ML methods, but no exhaustive search is needed.In multiuser OFDM systems, the users are driven by independent oscillators and have different CFO and SFO, so there are multiple CFO and SFO to be estimated at the receiver. The zero-forcing (2F) criterion is adopted in conventional estimation methods, which is known to suffer from the noise enhancement problem. A novel joint CFO and SFO estimation algorithm is proposed in this dissertation, where the users’ separation is achieved via the QR decomposition, so that the interferences from estimated users can be cancelled. To further improve the estimation accuracy, the received signals are properly weighted before the summation.Due to the drift of the oscillators, the CFO may be time-varying. Hence the problem of joint CFO and channel tracking in time-varying multiuser OFDM systems is discussed. Treating the CFO as latent variables, the expectation maximization (EM) algorithm is used to find the maximum a posteriori (MAP) estimation. To obtain the conditional expectation of the latent variables in the E step, the variational inference is exploited to approximate the posterior distribution. Simulation results show that the proposed algorithm can achieve high tracking accuracy with only a few iterations.At last, the idea of CFO estimation is applied to the problem of phase ambiguity in blind identifications. When tracking the CFO and the channel, the pilot overhead becomes a serious problem. The blind techniques can be used to estimate channel, but there is a phase ambiguity in conventional blind methods. Recently, some researchers find that the channel can be blindly and uniquely identified if the information of source constellation is exploited. In this dissertation, it is proved that the unknown phase can be divided into a fractional part and an integer part, and their identification results are very similar to the case of the fractional CFO and the integer CFO. Then the results are extended to the more general case where an arbitrary number of constellations are used. Simulation results show that the proposed algorithm achieves almost the same performance as pilot methods.