Measurements from 6.75 to 142 GHz show a simple, physically anchored path‑loss model can extend 3GPP urban microcell rules above 100 GHz
This paper tests how well standard urban microcell path‑loss models work across a very wide range of radio frequencies, from 6.75 GHz up to 142 GHz. Path loss is the way a wireless signal gets weaker as it travels. The authors combine many real measurements to see which mathematical forms give stable and physically meaningful parameters when models are pushed into the sub‑terahertz range above 100 GHz.
The team pooled five NYU WIRELESS measurement campaigns in Manhattan and Brooklyn. The campaigns measured at 6.75, 16.95, 28, 73, and 142 GHz. They used wideband sliding‑correlation channel sounders (1 GHz bandwidth except 800 MHz at 28 GHz), collected more than 200 GB of raw data, and covered transmitter–receiver (TX–RX) separations from about 24 to 880 meters. Each campaign included line‑of‑sight (LOS) and non‑line‑of‑sight (NLOS) links and reported the number of TX–RX pairs and the TX/RX antenna heights used in the measurements.
The authors compared several common path‑loss formulas. The “close‑in” (CI) model anchors the curve to the free‑space loss at a 1 meter reference and then uses a single path‑loss exponent to capture how loss grows with distance. The “floating‑intercept” (FI) model fits an intercept and slope to the data without a physical anchor. They also tested multi‑frequency forms: CI extended with a frequency weight (CIF) and the alpha‑beta‑gamma (ABG) model, which fits frequency and distance coefficients. All fits report 95% confidence intervals from regression so the authors could test parameter stability as they added more frequency bands.
Key findings are concrete and modest. Single‑frequency fits show the CI model gives more stable and physically interpretable parameters than FI, while producing shadow‑fading standard deviations within 0.98 dB of the FI fits across the five frequencies. In multi‑frequency fits over ranges 7–24 GHz, 0.5–100 GHz, and 0.5–150 GHz, the CI and CIF models keep stable distance exponents that match the 3GPP urban microcell (UMi) interpretations. The ABG model sometimes reduces fitting error slightly, but its parameters are much more sensitive to which frequencies are included, making it less reliable for extending standards.