Centralized Versus Distributed Operation in Cell-Free Massive MIMO With Hardware Impairments and Prior Information

This article investigates scalable cell-free massive multiple-input–multiple-output (MIMO) systems with hardware impairments under spatially correlated channels. First, channel state information is acquired using linear minimum mean-squared error (LMMSE) and least-squares (LS) estimators. It is demonstrated that both hardware impairments and pilot contamination can cause error floors in channel estimation. Second, for centralized operation, two uplink ergodic channel capacity lower bounds are derived, and the optimal receive combining scheme is developed. Furthermore, closed-form expressions of the spectral efficiency (SE) under two channel estimators are obtained, which provides valuable insights into the effects of nonideal hardware on centralized operation without pilot contamination. For distributed operation, a lower bound of the ergodic uplink channel capacity is determined, and the optimal large-scale fading decoding (LSFD) detection scheme is obtained. Moreover, closed-form expressions of the SE for distributed operation without pilot contamination are derived based on two channel estimators. Third, energy efficiency is investigated using a realistic power consumption model, and a joint pilot allocation and access point (AP) selection algorithm for scalable systems is proposed. Finally, the effectiveness of our analysis and the proposed algorithm is substantiated through comprehensive simulations.