Multi-Scenario Performance Evaluation of 5G NR Networks: From Physical-Layer MIMO to Deployment Density, Traffic Heterogeneity, and Self-Optimization

The performance of fifth-generation New Radio (5G NR) networks is typically evaluated layer by layer, using disconnected methodologies that make it difficult to compare the returns of design decisions across layers or to trace how deficiencies at one layer propagate to another. This article proposes a unified metric taxonomy that enables direct comparison of results spanning four network layers (physical, deployment, traffic, and autonomous management), obtained from five independently developed simulation studies that were not co-simulated. We quantify: (i) the diversity–multiplexing trade-off and massive multiple-input multiple-output (MIMO) scaling, showing a 3.5× spectral-efficiency gain from 4 to 64 antennas with diminishing returns; (ii) the density–capacity trade-off between dense indoor and sparse rural enhanced mobile broadband (eMBB) deployment, showing that the two regimes are limited by different mechanisms (link-level signal-to-noise ratio indoors, path loss and coverage radius in rural areas) and therefore respond to different scaling levers (MIMO order versus frequency-differentiated multi-band layering); (iii) massive machine-type communications (mMTC) and ultra-reliable low-latency communications (URLLC) channel-provisioning sensitivity, where increasing channels from 1 to 4 resolves most contention-driven loss while aggregate metrics can mask class-specific URLLC reliability erosion; and (iv) self-organizing network (SON) self-optimization, showing a disproportionate improvement in worst-case link quality (+9.9%) relative to already-adequate aggregate quality (+1.65%), with two key performance indicator (KPI) targets unreached. The cross-scenario comparison reveals a consistent threshold–saturation pattern across all four layers, attributable to structurally analogous mechanisms, from which we derive cross-scenario network-planning guidelines.