Performance Analysis of 5G Massive MIMO Networks Using Hybrid Beamforming and Geometric Channel Models

This paper analyzes the performance of a 5G massive MIMO network based on hybrid beamforming (HBF) on geometric channels of 3GPP TR 38.901. The evaluation is carried out through link/system-level simulations at 28 GHz (OFDM 400 MHz), BS 128 antennas, 8 UEs, HBF architecture with N_RF.∈{12,16} and a 3–4 bit phase shifter, covering both UMi street canyon and indoor office (LOS/NLOS) scenarios. Channel estimation and precoder/combiner design follow the framework of compressed/hierarchical beam training and factorization of fully digital solutions; each configuration is evaluated on ≥1000 Monte Carlo drops. Results show that HBF (N_RF=16, 4-bit) achieves ≈90–95% sum spectral efficiency (SE) over fully digital with an average gap of 3–6 bps/Hz, while N_RF=12 reduces SE by an additional ~1–2 bps/Hz but provides the highest energy efficiency (EE). Compared to fully digital, HBF improves EE by ~4–5× due to the RF-chain reduction despite a slight SE decrease. Robustness tests against angle misestimation show SE degradation slopes of approximately 0.5%/° (fully digital), ~1.2%/° (4-bit HBF), and ~1.6%/° (3-bit HBF). The differences between scenarios highlight the sensitivity of HBFs to wider angular spread indoors. Overall, HBFs with ≥4-bit and N_RF selection proportional to the number of user flows provide the best SE-EE compromise, making them feasible for 5G mmWave implementations. Future directions include beam-squint mitigation in broadband, more detailed hardware impairment modeling, and robust machine learning-based HBF design.