⚡ electrical engineering

Quasi-Deterministic Modeling of Sub-THz Band Access Channels in Street Canyon Environments

This paper presents a quasi-deterministic channel model for sub-THz frequencies (154 GHz and 300 GHz) in outdoor street canyon environments, derived from extensive double-directional measurements that characterize multipath clustering and large-scale parameters under both line-of-sight and non-line-of-sight conditions to support future 6G network development.

Original authors: Minseok Kim, Masato Yomoda, Minghe Mao, Nobuaki Kuno, Koshiro Kitao, Satoshi Suyama

Published 2026-03-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Minseok Kim, Masato Yomoda, Minghe Mao, Nobuaki Kuno, Koshiro Kitao, Satoshi Suyama

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to have a conversation with a friend in a long, narrow alleyway made of tall, glass-and-concrete buildings. Now, imagine you are trying to do this using a super-high-pitched whistle (Sub-THz waves) that travels almost like a laser beam.

This paper is a report on exactly that kind of experiment. The researchers wanted to understand how these "laser-like" radio waves behave in city streets to help build the 6G internet of the future, which promises to be incredibly fast.

Here is the story of their findings, broken down into simple concepts:

1. The Problem: The "Laser" Problem

Current 5G networks are like a floodlight. They shine light (signals) in all directions, bouncing off walls, cars, and trees to fill a whole room. This works well, but it's slow.

The new 6G networks want to use Sub-THz frequencies (154 GHz and 300 GHz). Think of these not as floodlights, but as laser pointers.

The Good: They can carry massive amounts of data (like a firehose of information).
The Bad: Because they are so focused, if you block the laser with a person or a wall, the signal dies instantly. They don't bounce around as easily as older signals.

The big question was: If we use these laser-like signals in a city street, will they work? Or will the buildings block them all the time?

2. The Experiment: The "Street Canyon" Test

The researchers set up a test in a real street at Niigata University in Japan.

The Setup: They placed a transmitter (the "talker") on a pole and a receiver (the "listener") on a cart, moving it up and down the street.
The Tools: They built a special device that could listen to two different "whistles" at the same time: one at 154 GHz and a higher-pitched one at 300 GHz.
The Conditions: They tested two scenarios:
1. Line-of-Sight (LoS): The talker and listener could see each other directly.
2. Non-Line-of-Sight (NLoS): A building corner blocked the direct view, forcing the signal to find a way around.

3. The Discovery: The "Hall of Mirrors"

When they analyzed the data, they found something fascinating. Even though these waves are "laser-like," the street acted like a Hall of Mirrors.

The Main Path: The direct signal (the laser beam) was the strongest.
The Bounces: The signal didn't just stop; it bounced off the North Wall and the South Wall of the street.
The Surprise: At the higher frequency (300 GHz), the signal was a bit more "picky." It bounced off the walls, but the bounces were weaker and fewer than at 154 GHz. It's like trying to bounce a super-precise laser off a slightly rough mirror; some of the light scatters and gets lost.

Key Finding: The street isn't a chaotic mess of signals. It's actually very organized. The signals mostly follow a few predictable paths: Direct or Bounce-Off-One-Wall. This is great news because it means we can predict where the signal will go.

4. The Solution: The "Quasi-Deterministic" Model

The researchers realized that old computer models (which assume signals bounce everywhere randomly like a pinball) were wrong for these new frequencies.

So, they created a new model called Quasi-Deterministic (Q-D).

Deterministic (Predictable): They programmed the model to know the "main actors": The Direct Path and the two main Wall Bounces. These are like the main characters in a play; you know exactly where they will be.
Random (Unpredictable): They added a "chaos factor" for the tiny, messy signals that bounce off windows, doors, or rough concrete. These are like the background extras; they are there, but they are random and hard to predict.

The Analogy: Imagine trying to predict the weather.

Old Model: "It might rain anywhere, anytime, randomly." (Too vague).
New Q-D Model: "It will definitely rain on the mountain (Direct Path) and the valley (Wall Bounce), but there might be random drizzle in the forest (Random scatter)."
This new model is much more accurate for planning 6G networks.

5. Why This Matters for You

This research is the blueprint for the 6G "Hotspots" of the future.

Immersive Reality: Imagine wearing VR glasses in a busy city and having zero lag. This technology makes that possible.
Smart Cities: Self-driving cars and robots need to talk to each other instantly. This model helps engineers place the "laser towers" in the right spots so the signal doesn't get blocked by a bus or a pedestrian.
Efficiency: Because the signals are so predictable, we don't need to build a tower on every single corner. We can use the buildings themselves to bounce the signal to where it's needed.

Summary

The paper proves that while Sub-THz signals are fragile and "laser-like," they behave in a very predictable way in city streets. They mostly travel straight or bounce once off a wall. By understanding this "Hall of Mirrors" effect, engineers can build a new generation of super-fast, ultra-reliable mobile networks that will power our future digital world.

1. Problem Statement

Sub-terahertz (sub-THz) frequencies (100–300 GHz) are critical for achieving the ultra-high data rates (>100 Gbps) and ultra-low latency required for 6G networks. However, existing channel models (e.g., 3GPP TR 38.901) rely on extrapolations from sub-100 GHz data and assume rich multipath scattering.

The Gap: Sub-THz channels exhibit "quasi-optical" propagation characteristics, meaning they are dominated by Line-of-Sight (LoS) and sparse multipath components due to high attenuation and blockage. Current models fail to accurately capture this sparsity and the frequency-dependent behavior of multipath clusters in urban street canyons.
The Need: There is a lack of extensive, dual-band measurement data and accurate channel models that account for the specific propagation mechanisms (e.g., single-bounce reflections vs. diffuse scattering) in street canyon environments at 154 GHz and 300 GHz.

2. Methodology

A. Measurement Campaign

Environment: An outdoor street canyon at Niigata University, Japan, with a width of 20–25 m and high-rise buildings (>40 m) on both sides.
Frequencies: Simultaneous measurements at 154 GHz and 300 GHz.
Scenarios:
- LoS: Transmitter (Tx) at 3m height, Receiver (Rx) at 1.2m height, distances from 10–80 m.
- NLoS: Tx positioned behind a building corner, blocking the direct path, distances 10–40 m.
Equipment: An in-house developed dual-band multi-tone frequency-domain channel sounder.
- Bandwidth: 4 GHz (154 GHz) and 8 GHz (300 GHz).
- Delay Resolution: 250 ps (154 GHz) and 125 ps (300 GHz).
- Antennas: Directional horn antennas (26 dBi gain) with narrow beamwidths ( $8^\circ \times 9^\circ$ ).
- Scanning: Full azimuth sweep at both Tx and Rx (1,600 angular combinations).

B. Data Processing & Analysis

Noise Filtering: A novel dual-band joint noise filtering algorithm was developed. Instead of using a static threshold, it cross-references the two frequency bands. If a path is detected in one band but missing in the other due to attenuation, the algorithm checks for spatial consistency (angle alignment). If angles match, the path is "rescued" in the weaker band; if not, it is removed as a false alarm.
Clustering: A K-PowerMeans algorithm was applied to merged datasets from both frequencies to identify common and distinct multipath clusters. Power scaling was applied to the 300 GHz data to account for free-space path loss differences before clustering.
Ray Tracing: Used to validate dominant propagation mechanisms (e.g., single-bounce reflections from North and South walls) against measured power delay profiles (PDP).

C. Modeling Approach

Quasi-Deterministic (Q-D) Model: The proposed model combines:
1. Deterministic Components: The LoS path and single-bounce (SB) reflections from the North and South walls. The presence/absence of these reflections is modeled using a two-state Markov process to account for surface roughness and inhomogeneity.
2. Random Components: Higher-order reflections and diffuse scattering are modeled as random clusters. Their arrival times follow an exponential distribution, and their power follows a linear decay relative to the LoS path with Gaussian shadowing.

3. Key Contributions

Extensive Dual-Band Measurements: First extensive double-directional (D-D) channel sounding campaign in a street canyon at both 154 GHz and 300 GHz under consistent LoS and NLoS conditions.
Advanced Signal Processing: Development of a cross-frequency noise filtering and clustering technique that leverages physical consistency between bands to improve multipath detection reliability.
Comprehensive Characterization: Detailed evaluation of Large-Scale Parameters (LSPs) including Path Loss (PL), RMS Delay Spread (DS), Angular Spread (AS), and Rician K-factor.
Proposed Q-D Channel Model: A novel model that extends corridor-based frameworks to outdoor street canyons, explicitly separating deterministic wall reflections from random scattering, validated against measurement data.

4. Key Results

A. Propagation Characteristics

Sparsity: Sub-THz channels are highly sparse. In LoS scenarios, the channel is dominated by the direct path and two strong single-bounce reflections from the side walls. Double-bounce reflections are rare.
Frequency Dependence:
- Cluster Count: The 300 GHz band exhibits slightly fewer clusters than 154 GHz due to higher interaction losses.
- Path Loss: 300 GHz experiences slightly higher degradation (~1.2 dB more than free-space prediction) compared to 154 GHz, attributed to increased surface scattering losses.
- LSPs:
  - Delay Spread (DS): Measured DS is smaller than 3GPP extrapolations (approx. 6.5 ns at 154 GHz vs. 4.7 ns at 300 GHz), confirming lower dispersion.
  - Angular Spread: Measured angular spreads are smaller than 3GPP predictions, highlighting the directional nature of sub-THz channels.
  - K-Factor: Average K-factors are ~7 dB, indicating a strong LoS dominance that is consistent across frequencies.

B. Model Validation

Path Loss Accuracy: The proposed Q-D model achieves a Root Mean Square Error (RMSE) of ~1.6–1.7 dB when predicting path loss against measured data for both frequencies.
Statistical Consistency: Kolmogorov–Smirnov (KS) tests confirmed that the statistical distributions of Delay Spread and K-factor generated by the model are statistically indistinguishable from the measured data (p-value > 0.05).
Markov Process: The model successfully captures the random "on/off" nature of wall reflections, with the probability of reflection absence increasing from ~19% at 154 GHz to ~30% at 300 GHz.

5. Significance

6G System Design: The findings provide critical insights for designing 6G access networks in urban environments, emphasizing the need for beam management strategies that exploit dominant LoS and single-bounce paths rather than relying on rich scattering.
Model Accuracy: The proposed Q-D model offers a more accurate alternative to standard 3GPP models for sub-THz frequencies, reducing the error in link budget calculations and coverage planning.
Frequency Scaling: The study demonstrates that while geometric dominance (street layout) dictates channel structure, higher frequencies (300 GHz) suffer from increased interaction losses, necessitating specific modeling adjustments for surface roughness and reflection availability.
Future Research: The work establishes a foundation for developing intelligent reflecting surfaces (IRS) and multi-beam management techniques tailored to the sparse, quasi-deterministic nature of sub-THz street canyons.