Analog beamforming (ABF) is a key enabler for high-throughput mmWave communication to harvest sufficient beamforming gain from large antenna arrays. It relies on beam alignment and the radio frequency (RF) architecture. In multiuser scenarios, beam alignment becomes challenging because of the inter-beam-interference between different users. Motivated by the large number of available antennas at the base station, we propose a downlink sparse ABF architecture in which only a subset of antennas per RF chain can be scheduled for the beamforming. For this purpose, we model the RF chain-antenna association as a binary association matrix (BAM) and jointly optimize the transmit beams and the BAM subject to RF hardware constraints. To alleviate the complexity of the resulting mixed–integer programming problem, we propose an iterative beam alignment and BAM optimization where the beam alignment and BAM are successively optimized until convergence. The proposed sparse-connected RF architecture performs significantly better than the traditional fully and partially connected architectures in terms of signal-to-interference-plus-noise ratio and bit error rate. The simulation results indicate that optimizing the BAM profiles relaxes the need for optimized beam allocation beyond beamforming along the line-of-sight.

Domino-tiled phased arrays (DTPA), based on two-element single polarized sub arrays, are conceived for beyond 100 GHz multi-user multiple-input multiple-output (MIMO) communication. Compared to lower frequencies, the small wave-length poses stringent requirements on the connections between the integrated circuit (IC) and antenna array, as the antenna dimensions are comparable to IC size. To address this problem, the tiling configuration of DTPAs is optimized by a genetic algorithm (GA) while jointly maximizing the spectral efficiency (SE) and reducing side lobe levels (SLLs), considering above 100 GHz channel models for multiple users. As a proof-of-concept, optimized DTPAs for an array size of 8 × 10 are realized in full-wave EM simulator through a printed circuit board (PCB) manufacturing process. Our optimized DTPAs yield more than 16% impedance bandwidth, targeting the [120–140] GHz D-band. Moreover, the optimized DTPAs reach a SE and SLL up to 3.7 b/s/Hz and -13 dB for a two-user scenario.

In this paper we study analytically the sensitivity of analog-beamformed multiuser downlink MIMO systems to imperfect beamforming (BF) state information at the transmitter and the receiver. We consider different distributions of the transmit and receive BF errors and evaluate the end-to-end performance in terms of the statistical distribution of the minimum signal-to-interference plus noise ratio (SINR) and derive accordingly, for small fluctuation errors, analytical approximation of the average SINR and the corresponding symbol error probability (SEP). The analytical approximations of the SEPs are further benchmarked with the simulated error performance for both uncoded and coded transmission and for two BF architectures namely partially (PCA) and fully (FCA) connected architectures. Simulation results in line-of-sight propagation revealed that (i) tight estimates mainly of the transmit BF angles are required in the order of a fraction of one degree for moderate number of transmit antennas especially for FCA, and (ii) the BF accuracy requirement becomes stringent as the system load increases. We argue that with a moderate to high number of transmit antennas, multiuser beam alignment achieving sub-degree BF accuracy will be one of the main challenges of future mmWave communication systems that rely on analog BF.11This research received the support of the COREnext project, funded by the European Union's Horizon Europe Research and Innovation Actions, under Grant Agreement$N^{\circ}l$01092598.

Millimeter-wave spectrum (30 to 300 GHz) is explored in order to provide higher throughput by exploiting the large bandwidth available. At those frequencies, multiple antenna systems are essential to combat severe path loss. Fully digital (FD) architectures, where each antenna connects to its own baseband chain, are considered the most energy-efficient at low frequencies while enabling multi-user multiplexing. However, unlike lower frequencies, channel propagation above 100 GHz is heavily line-of-sight dominated and the operating bandwidth is much larger which poses different constraints on signaling schemes and hardware components. The optimal beamforming architecture configuration is still an open problem above 100 GHz. To address this problem, we compare energy-efficient beamforming architectures for the D-band (110-170 GHz). We estimate the energy efficiency of future systems by following the technology trends in circuit implementation. We show that hybrid fully connected architecture is the most energy-efficient for the 7nm technology node and size-constrained antenna arrays. Hybrid partially connected architecture is the most energy-efficient for unconstrained antenna arrays. We show that FD architecture becomes energy-efficient for technology nodes better than 2nm.

Mm-wave spectrum (30 to 300 GHz) is explored in order to provide higher throughput by exploiting the large bandwidth available. Due to those higher frequencies, future systems are expected to have larger antenna arrays with more directive beams relying on line-of-sight propagation. Besides wider bandwidth, high-throughput systems also aim at serving multiple users in parallel. When those users are located at different ranges from the base station, they put very different requirements on the base station architecture and signals. This can translate, e.g., into architectures having different numbers of components and power consumption to optimally serve those users. Finding the architecture offering the best throughput vs. power consumption trade-off while serving multiple users at different distances is still an open problem in the literature. Typically, power optimization has been done considering a fixed operating distance and neglecting the user distribution which is crucial in future communication systems. To that end, we propose a novel range distribution aware (RDA) architecture dimensioning methodology that optimizes the power consumption of a transceiver architecture over the operating range by exploiting the user distribution. We show that the RDA design methodology can reduce the average power consumption of the transceiver by 20% in a typical scenario compared to traditional power optimization.