Source-linked AI summary
5G 3GPP-like Channel Models for Outdoor Urban Microcellular and Macrocellular Environments
Katsuyuki Haneda, Lei Tian, Yi Zheng, Henrik Asplund, Jian Li, Yi Wang, David Steer, Clara Li, Tommaso Balercia, Sunguk Lee, YoungSuk Kim, Amitava Ghosh, Timothy Thomas, Takehiro Nakamura, Yuichi Kakishima, Tetsuro Imai, Haralabos Papadopoulas, Theodore S. Rappaport, George R. MacCartney, Mathew K. Samimi, Shu Sun, Ozge Koymen, Sooyoung Hur, Jeongho Park, Charlie Zhang, Evangelos Mellios, Andreas F. Molisch, Saeed S. Ghassamzadah, Arun Ghosh
TL;DR
Existing channel models were designed for frequencies up to 6 GHz, leaving higher-frequency propagation modeling insufficient for 5G systems up to 100 GHz. This document develops a preliminary 3D model from measurements and ray tracing, covering UMi and UMa deployments and key propagation effects.
Problem
Accurate radio propagation models for 5G operation above 6 GHz and up to 100 GHz are not addressed by existing models designed only through 6 GHz.
Method
The document combines extensive measurements and ray tracing across 6–100 GHz to define preliminary UMi and UMa 3D channel parameters and modeling procedures.
Results
The preliminary model incorporates path loss, shadow fading, line-of-sight probability, penetration, and blockage for typical UMi and UMa scenarios.
Takeaways & Limitations
The findings provide promising channel models that extend existing 3GPP models designed for below 6 GHz.
Abstract
from arXiv · showhide
For the development of new 5G systems to operate in bands up to 100 GHz, there is a need for accurate radio propagation models at these bands that currently are not addressed by existing channel models developed for bands below 6 GHz. This document presents a preliminary overview of 5G channel models for bands up to 100 GHz. These have been derived based on extensive measurement and ray tracing results across a multitude of frequencies from 6 GHz to 100 GHz, and this document describes an initial 3D channel model which includes: 1) typical deployment scenarios for urban microcells (UMi) and urban macrocells (UMa), and 2) a baseline model for incorporating path loss, shadow fading, line of sight probability, penetration and blockage models for the typical scenarios. Various processing methodologies such as clustering and antenna decoupling algorithms are also presented.
I. INTRODUCTION
The paper addresses the lack of channel models validated above 6 GHz by proposing a preliminary 3D framework for 5G frequencies up to 100 GHz. It builds on existing 3GPP models while targeting higher-frequency deployments, wide bandwidths, large arrays, mobility, consistency, and practical simulation complexity.
- 5G systems spanning roughly 500 MHz to 100 GHz require accurate propagation models above 6 GHz, where earlier channel models were not designed or evaluated.
- The framework extends 3GPP 3D UMi and UMa models to support elevation-aware modeling for two-dimensional antenna systems and future higher-frequency scenarios.
- The proposed overview derives outdoor channel properties from extensive measurements and ray tracing across multiple frequency bands and provides preliminary parameters for 5G simulations.
- Higher-frequency models must represent two-dimensional dual-polarized arrays, directional antenna-element gains, and accurately modeled azimuth, elevation, and polarization properties.
- The channel model must accommodate frequencies up to 100 GHz, channel bandwidths up to 2 GHz, large and highly directive arrays, mobility up to 350 km/hr, and smooth spatial, temporal, and frequency consistency.
- Maintaining practical computational complexity and continuity with existing 3GPP methodology supports meaningful comparative evaluations across bands up to 100 GHz.
III. TYPICAL UMI AND UMA OUTDOOR SCENARIOS
The outdoor UMi and UMa characterization summarizes measured and ray-traced propagation behavior above 6 GHz, including path loss, shadow fading, delay spread, and angular directionality.
- LOS path loss above 6 GHz appears to follow Friis’ free-space path-loss model, while NLOS conditions show a higher path-loss slope.
- Measurements show shadow fading similar to lower-frequency bands, whereas ray-tracing results can exceed 10 dB because of greater dynamic range and severe modeled losses.
- Above 6 GHz, measured delay spreads indicate somewhat smaller ranges as frequency increases, and some millimeter-wave omnidirectional channels are highly directional.
B. UMa Channel Characteristics (TX Heights ≥25 m)
UMa propagation above 6 GHz follows free-space behavior in LOS, while NLOS frequency trends remain inconclusive and may reflect diffraction, reflections, scattering, or measurement dynamic range. Building penetration varies strongly by material, facade composition, incidence angle, and interior depth.
- UMa channel characteristics: LOS path loss follows free-space behavior, while NLOS loss increases with frequency at a non-linear rate across the measured range.The lower-spectrum increase may reflect stronger diffraction, whereas reflections and scattering may dominate at higher frequencies.
- UMa channel characteristics: Higher-frequency UMa measurements may underestimate NLOS loss trends because their lower dynamic range can bias the results.The paper states that more measurements are needed to understand the UMa channel.
- Building penetration loss: Material penetration losses differ substantially: glass changes weakly with frequency, metal-coated energy-efficient glass can add up to 40 dB, and concrete or brick losses rise rapidly.Variations around approximately linear frequency trends can result from multiple reflections and interference within or between material layers.
- Building penetration loss: Effective building penetration combines transmission through multiple facade materials and can vary considerably even within a single building.Measurements and multi-material models indicate consistency between material-loss and effective-penetration results despite differing loss values.
- Building penetration loss: Grazing incidence can increase penetration loss by up to 15-20 dB, while deeper indoor propagation adds 0.2-2 dB/m over 2-60 GHz.Interior loss is weakly frequency dependent but strongly dependent on the building’s interior composition.
V. BLOCKAGE CONSIDERATIONS
The paper distinguishes transient dynamic blockage from static geometry-induced blockage in higher-frequency propagation. Moving objects add temporary path loss, while environmental obstructions produce diffraction- or diffuse-scattering-dominated channels and exceptional additional loss.
- Blockage types: Dynamic blockage is caused by moving cars or people and produces transient additional loss on paths intersecting the moving object.A 28 GHz measurement example uses continuous observations across traffic controlled by a traffic light.
- Blockage types: Geometry-induced blockage is a static environmental property caused by objects that block signal paths.Its channels are dominated by diffraction and sometimes diffuse scattering.
- Blockage effects: Diffraction-dominated blockage produces exceptional additional loss beyond normal path loss and shadow fading.Compared with reflection-caused shadow fading, its statistics may differ.
VI. PATH LOSS, SHADOW FADING, LOS AND BLOCKAGE MODELING
The section compares LOS-probability models for UMa and UMi scenarios, using map-based LOS definitions and fitted distance parameters. The UMi results find the 3GPP formula sufficient above 6 GHz, while UMa evaluation compares three models against measured and ray-tracing data.
- LOS Definition: The map-based LOS definition uses AP and UE positions and building or wall blockage, while unrepresented objects are modeled separately through shadowing or blocking terms.Because only buildings and walls determine the LOS state, the definition is frequency independent.
- LOS Models: The d1/d2 model is identified as the current 3GPP/ITU LOS-probability model, with d1 and d2 optimizable against data or scenario parameters.The distance variable d is the 2D transmitter-receiver distance in meters.
- UMa Results: The UMa analysis compares the d1/d2, NYU squared, and current 3GPP models using measured and ray-tracing data; the NYU squared model has the lowest MSE.The current 3GPP UMa comparison uses UE height 1.5 m with d1 = 18 and d2 = 63.
- UMi Results: For UMi above 6 GHz, the 3GPP LOS-probability formula is sufficient, although a fitted d1 = d2 model provides a better fit with small overall errors for the 3GPP formula.For indoor users, d is replaced by the 2D distance to the outer wall.
A. Path Loss Models
The paper considers three multi-frequency path-loss formulations for modeling propagation across scenarios. These include CI and CIF models anchored to 1 m free-space path loss, alongside the ABG model with explicit distance and frequency parameters.
- Model Overview: The three considered multi-frequency path-loss models are CI, CIF, and ABG, and they are applied to various propagation scenarios.CI uses a close-in free-space reference distance; CIF adds a frequency-dependent path-loss exponent.
- 3GPP Model: The existing 3GPP 3D path-loss formulation has ABG form without a frequency-dependent parameter, plus base-station or terminal-height dependencies and a LOS breakpoint.The choice of one recommended model per scenario and LOS/NLOS condition remained under discussion in 3GPP RAN.
- CI Model: The CI model anchors path loss to free-space path loss at 1 m, capturing frequency dependence and providing a uniform reference for measurements and model parameters.Its path-loss exponent and shadow-fading term describe propagation variation around that anchor.
- ABG Model: The ABG model uses three optimization parameters, with α describing distance dependence, β an offset, and γ the variation with frequency.Its shadow-fading term has a standard deviation expressed in dB.
- CIF Model: The CIF model extends CI by using a frequency-dependent path-loss exponent, with n and b as optimization parameters.It retains the 1 m free-space close-in reference-distance anchor.
+ XCIF
The CIF path-loss model extends the CI model across multiple frequencies by optimizing both the path-loss exponent and its frequency-dependent slope around a weighted reference frequency.
- + XCIF: The CIF model uses n as the path loss exponent and b as an optimization parameter for linear frequency dependence in the path loss exponent.A positive b indicates that path loss increases with frequency.
- + XCIF: f0 is the weighted centroid of the modeled frequencies, computed from the number of path-loss data points at each frequency.K denotes the number of unique frequencies, and Nk denotes the number of path-loss data points corresponding to frequency fk.
- + XCIF: The CIF model reverts to the CI model when b = 0 for multiple frequencies or when a single frequency f = f0 is modeled.
- + XCIF: The CI model estimates only the path loss exponent by minimizing shadow-fading standard deviation and anchors path loss to 1 m free-space path loss.The CIF model retains this 1 m close-in reference while adding the frequency-slope parameter b.
B. Fast Fading Model
The double-directional channel model represents each multipath component by its delay, departure and arrival directions, and complex amplitude, then sums the components into a channel impulse response.
- B. Fast Fading Model: Each multipath component is described by its delay, transmit and receive directions, and complex amplitude gain.
- B. Fast Fading Model: The channel impulse response is formed by summing the generated double-directional multipath components.
- B. Fast Fading Model: The model provides a complete omnidirectional statistical spatial channel model for both LOS and NLOS scenarios.
1) UMi:
The UMi analysis extracts large-scale channel parameters from ray-tracing results by examining clusters in the spatio-temporal domain and applying established extraction methodologies.
- 1) UMi:: Ray-tracing results are analyzed for cluster delays, transmit and receive angles, and received powers after clustering.
- 1) UMi:: Large-scale parameters include the number of clusters and intracluster delay and angle spreads.
- 1) UMi:: Parameter extraction follows established frameworks applied to the observed clusters in each link.
- 1) UMi:: The UMa ray-tracing study used an Aalborg environment with one 25 m AP, 1.5 m UEs, and isotropic antennas at both ends.
2) UMa:
The UMa study uses Aalborg ray tracing across six frequencies to examine frequency-dependent channel parameters and clustering, finding reduced spreads and increased depolarization at higher frequencies.
- 2) UMa:: Six frequencies from 5.6 to 73.5 GHz were considered in the Aalborg ray-tracing study.The modeled frequencies were 5.6, 10, 18, 28, 39.3, and 73.5 GHz.
- 2) UMa:: 13.87 to 7.89 dB: cross-polarization discrimination ratio decreased from 5.6 GHz to 73.5 GHz.The decrease was primarily attributed to increased diffuse scattering at shorter wavelengths, though measurements were still needed for verification.
- 2) UMa:: Delay and azimuth angle spreads decreased as carrier frequency increased.
- 2) UMa:: Average cluster count and average rays per cluster were fairly consistent across carrier frequencies.K-Means was run 50 times with different random starts, retaining the cluster set with the minimum number of clusters.
- 2) UMa:: The ray-tracing simulations limited the number of modeled rays to 20, constraining interpretation of average rays per cluster.More recent 3GPP-like statistics without this limitation are cited as an alternative.
VII. CONCLUSION
The preliminary findings and ongoing efforts support channel models that extend existing 3GPP models designed for frequencies below 6 GHz.
- The preliminary findings and ongoing efforts provide channel models that can extend today’s 3GPP channel models.These models are designed to address the limitations of models developed for operation below 6 GHz.