Collaborative Vs. Non-Collaborative CFO Estimation for Distributed Large-Scale MIMO Systems

Carrier frequency offset (CFO) can significantly influence the performance of multiple-input multiple-output (MIMO) systems if not estimated and corrected. Most of the existing synchronization techniques are designed for MIMO systems with collocated antennas, where the same frequency error is experienced by the received signal at each antenna. In contrast, for MIMO systems with distributed antennas, the received signals at different receive antennas suffer from different frequency errors. This makes synchronization of such systems computationally complex, especially as the number of receive antennas grows large, i.e. massive MIMO. In this paper, we study the problem of CFO estimation for distributed massive MIMO and cell-free massive MIMO. In particular, we evaluate the performance of distributed large-scale MIMO systems with both collaborative and non-collaborative CFO estimation techniques based on the maximum likelihood (ML) criterion. These optimal CFO estimation results can then serve as benchmarks on the achievable estimation performance of practical collaborative and non-collaborative CFO estimation techniques for distributed MIMO systems. The proposed ML-based CFO estimation methods are tailored to work with periodic pilot sequences. We also show how exploiting the relationships which exist between the CFOs to be estimated can substantially reduce the computational load of CFO estimation in the collaborative approach.