CFO Estimation for OFDM-Based Massive MIMO Systems in Asymptotic Regime

Abstract

Massive multiple input multiple output (MIMO) plays a pivotal role in the fifth generation (5G) wireless networks. However, the carrier frequency offset (CFO) estimation is a challenging issue in the uplink of multiuser massive MIMO systems. In fact, frequency synchronization can impose a considerable amount of computational complexity to the base station (BS) due to a large number of BS antennas. In this paper, thanks to the properties of massive MIMO in the asymptotic regime, we develop a simple synchronization technique and derive a closed form equation for CFO estimation. We show that the phase information of the covariance matrix of the received signals is solely dependent on the users' CFOs. Hence, if a real-valued pilot is chosen, the CFO values can be straightforwardly calculated from this matrix. Hence, there is no need to deal with a complex optimization problem like the other existing CFO estimation techniques in the literature. Our simulation results testify the efficacy of our proposed CFO estimation technique. As we have shown, the performance of our method does not deteriorate as the number of users increases.

Frequency synchronisation for massive MIMO: a survey

“…Thus, accurate and low‐complexity CFO estimation schemes for massive MIMO systems need to be sought. To this end, a number of methods have been recently proposed in the literature [12–15, 19, 20]. These CFO estimation methods can be classified into three main categories: (i) Maximum likelihood based CFO estimation; (ii) Blind CFO estimation; and (iii) Correlation‐based CFO estimation.…”

“…A CB CFO estimator for OFDM‐based massive MIMO systems was proposed in [20]. In this work, CFOs are estimated through the calculation of the spatial correlation matrix of the received signals.…”

“…Using OFDM symbols as pilots, the proposed solution in [20] extends the solution of [19] to multicarrier massive MIMO systems. The authors of [20] demonstrate that one OFDM symbol is sufficient for CFO estimation; thus, a long pilot sequence for CFO estimation is not required as in [19]. The complexity is reduced by using orthogonal pilot sequences and assuming that the number of BS antennas is infinite.…”

“…The complexity is reduced by using orthogonal pilot sequences and assuming that the number of BS antennas is infinite. Moreover, the effect of inaccurate CFO estimation can be mitigated by averaging the CFO estimates over all the BS antennas [20].…”

“…Fig. 6 compares the mean square error (MSE) performance of the CFO estimation methods for massive MIMO proposed in [12–15, 19, 20]. These results are presented for the case of 100 collocated BS antennas, a carrier frequency of 6 GHz, and CFOs (normalised by the subcarrier frequency spacing) are uniformly distributed on the range [‐0.5, 0.5).…”