Downlink Performance of CF Massive MIMO with Hybrid MmWave/Microwave Fronthaul Network

This paper investigates the downlink (DL) performance of cell-free (CF) massive multiple-input multiple-output (mMIMO) systems under limited-fronthaul capacity. We consider a practical fronthaul deployment where the fronthaul network is modeled as a hybrid millimeter wave (mmWave) /microwave network. In particular, we assume the presence of multiple edge-cloud processor (ECP)s to which access points (AP)s are associated in a distance-based criterion. The APs are supported by mmWave fronthaul link given that line-of-sight (LoS) links exist with their associated ECPs, otherwise, microwave fronthaul links will be used. We analyze the achievable DL data rates under distributed and centralized system operations. For the distributed system operation, APs apply local conjugate beamforming (CB) precoding along with beamforming training. On the other hand, ECPs apply zero-forcing (ZF) precoding in the centralized system operation. Results reveal that the centralized system operation remarkably outperforms the distributed counterpart under perfect fronthaul links. However, the impact of limited capacity fronthaul links is more prominent on the centralized system operation with ZF precoding. Moreover, we show that the network should be designed according to the adopted system operation and the capacities of fronthaul links.