Abstract
As 5G networks expand, the 3.5 GHz band has become a cornerstone for urban coverage. However, the "Outdoor-to-Indoor" (O2I) scenario presents significant modeling challenges. This study presents a detailed measurement campaign and a new set of stochastic models that simultaneously analyze space, time, and dual-polarization domains. By introducing "lifted" distribution functions (LS-Laplace and LS-Normal), the authors more accurately capture the background scattered energy typical of complex urban environments, providing a robust tool for optimizing massive MIMO and beamforming technologies.
Introduction
A critical challenge for 5G network planners is ensuring reliable signal penetration from outdoor base stations to indoor users. While general channel models exist, they often lack the multidimensional precision required for advanced 5G features like massive MIMO. These technologies rely on accurate "maps" of signal behavior across space (direction of arrival), time (delay), and polarization. This paper addresses this gap by characterizing the spatial-temporal characteristics of the 3.5 GHz band in a typical Urban Macrocell (UMa) O2I environment.
Methods
The researchers conducted a field measurement campaign in a modern business district using a high-resolution 3D-MIMO channel sounder.
The experimental setup involved placing a transmitter (TX) on a 5-story rooftop to simulate a base station, while the receiver (RX) was moved along corridors on the 1st, 2nd, and 3rd floors of a nearby building, approximately 200 meters away. Both ends utilized antenna arrays with $\pm45^\circ$ polarized elements to capture dual-polarization effects.
The authors introduced two novel mathematical models to describe signal arrival:
- LS-Laplace (Lifted-Superposed Laplace): Used for the Azimuth Power Spectrum (APS).
- LS-Normal (Lifted-Superposed Normal): Used for the Elevation Power Spectrum (EPS).
The "lifted" innovation adds a "power floor" to standard models, accounting for the uniform background scattered energy common in cluttered urban settings.
Key Insights and Results
The study yielded several critical findings for 5G deployment:
The "Power Floor" and Model Accuracy
The proposed LS-Laplace and LS-Normal models significantly outperformed standard truncated models. By accounting for scattered energy arriving from all directions, the models achieved low error rates (RMSE ~0.019 for azimuth and ~0.030 for elevation). This confirms that signals in O2I scenarios do not just arrive in neat clusters but include a significant uniform background component.
Polarization and Spatial Consistency
Analysis showed that switching between $+45^\circ$ and $-45^\circ$ polarization did not significantly alter the statistical shape of signal profiles. The correlation between cross-polarized channels remained high (0.6 to 0.9), suggesting that while dual-polarization increases capacity through diversity, the physical environment shapes both polarizations consistently.
Environmental Impact vs. Height
Contrary to the expectation that higher floors always receive better signals, local obstacles were found to be more decisive. For example, heavy foliage outside the 2nd floor obstructed the direct line-of-sight path, resulting in a more complex, multi-clustered signal environment compared to the 1st or 3rd floors.
Statistical Distributions
The measurements confirmed that Azimuth and Elevation Spreads follow a lognormal distribution, while the Delay Spread follows a Rayleigh distribution. Additionally, the Power Delay Profile (PDP) was successfully modeled using a three-phase approach (rising, dropping, and trailing) to describe signal decay within a building.
Conclusion
This research provides a more realistic framework for 5G O2I simulations by incorporating the "lifted" power floor into spatial-temporal models. While the current study focused on a static environment, future work will involve integrating these models into system-level simulators and expanding measurements to include dynamic human traffic and other 5G bands like mmWave.
Summary updated