Propagation & Channels

Cluster-Based Model

Q: How are clusters defined in 3GPP channel models?

3GPP TR 38.901 defines clusters through a stochastic geometry process. For a given scenario (UMa, UMi, InH, RMa), the model generates N clusters (typically 12 to 24 for urban, 7 to 12 for rural) with delays drawn from a uniform distribution scaled by the scenario's delay spread. Each cluster has a power determined by an exponential decay profile with per-cluster random variation. Angles of departure and arrival (both azimuth and elevation) are assigned based on the scenario's angular spread parameters. Each cluster contains M rays (default 20) with fixed angular offsets following a Laplacian distribution within the cluster. Two strongest clusters are split into three sub-clusters to model the typical physical behavior of dominant reflectors producing tightly spaced sub-paths. Cross-polarization ratios (XPR) of 8 to 12 dB are assigned per ray.

Q: What is the Saleh-Valenzuela model?

The Saleh-Valenzuela (S-V) model, published in 1987, is the foundational cluster-based channel model originally developed for indoor UWB channels. It describes the channel impulse response as a sum of clusters arriving at times T_l according to a Poisson process with rate Lambda (0.05 to 0.5 clusters per nanosecond). Within each cluster, rays arrive at delays tau_kl following a secondary Poisson process with rate lambda (0.5 to 5 rays per nanosecond). Cluster powers decay exponentially with time constant Gamma (10 to 50 ns), and ray powers within a cluster decay with time constant gamma (1 to 10 ns). The model captures the key physical observation that multipath arrives in bursts corresponding to distinct scattering objects, with rays within each burst having similar characteristics. Extended versions add angular dimensions (double-directional S-V model) for MIMO channel simulation.

Q: Why are cluster-based models important for MIMO?

MIMO system capacity depends critically on the angular distribution of multipath: if all energy arrives from a single direction, the channel matrix is rank-1 regardless of antenna count, limiting capacity to single-stream. Cluster-based models capture the spatial structure by assigning each cluster distinct angles of departure and arrival, creating the angular diversity that enables spatial multiplexing. The number of resolvable clusters determines the effective channel rank and thus the number of independent data streams. For massive MIMO with 64 to 256 antennas, the beamspace representation reveals that 5 to 15 clusters typically dominate the channel, meaning the practical rank is 5 to 15 even with 64 antennas. This insight drives system design decisions like the number of RF chains, codebook design, and hybrid beamforming architectures.

/klus-ter bayst mod-ul/

A cluster-based model groups multipath arrivals into clusters of rays with similar delay and angle, where each cluster represents scattering from a physical object. The Saleh-Valenzuela model (1987) established the framework: clusters arrive via Poisson process with exponential power decay, and rays within clusters follow a secondary process. Modern implementations in 3GPP TR 38.901, WINNER II, and IEEE 802.11 define 7 to 24 clusters per scenario, each containing 10 to 30 rays with Laplacian angular distribution.

Category: Propagation & Channels

Clusters: 7 to 24 per scenario

Rays per cluster: 10 to 30

Understanding Cluster-Based Models

Wireless propagation through real environments produces multipath arrivals that are not uniformly distributed in delay and angle. Instead, they tend to arrive in groups or "clusters" corresponding to distinct scattering objects: a nearby building face produces a cluster of reflected rays arriving within a few nanoseconds of each other from similar angles, while a distant hillside produces another cluster with longer delay and different angular characteristics. The Saleh-Valenzuela model formalized this observation by describing the channel impulse response as a doubly stochastic process: clusters arrive according to a Poisson process in delay, and rays within each cluster follow a secondary Poisson process with shorter inter-arrival times.

The cluster-based framework became the standard for all modern wireless channel models because it naturally captures the spatial structure essential for MIMO system simulation. Each cluster's angle of departure (AoD) and angle of arrival (AoA) define the spatial signature that antenna arrays can resolve. The number of resolvable clusters determines the effective channel rank and thus the achievable spatial multiplexing gain. 3GPP TR 38.901 defines cluster parameters for 5G NR scenarios: Urban Macro (UMa) with 12 to 20 clusters, Urban Micro (UMi) with 12 to 19 clusters, Indoor Hotspot (InH) with 15 to 24 clusters, and Rural Macro (RMa) with 7 to 11 clusters. Each cluster carries 20 rays with angular offsets following a Laplacian distribution within the cluster's angular spread.

Cluster Model Equations

Saleh-Valenzuela Impulse Response:
h(τ) = ∑_l ∑_k β_kl · e^jφ_kl · δ(τ - T_l - τ_kl)

Cluster Power Decay:
E[|β_kl|²] = E[|β₀₀|²] · e^-T_l/Γ · e^-τ_kl/γ

RMS Delay Spread:
σ_τ = √(τ²_mean - (τ_mean)²)

Where T_l = cluster arrival time, τ_kl = ray delay within cluster, Γ = cluster decay time constant (10 to 50 ns), γ = ray decay time constant (1 to 10 ns), φ_kl = random phase. UMa at 3.5 GHz: σ_τ ≈ 80 to 300 ns depending on LOS/NLOS.

Cluster-Based Channel Model Comparison

Model	Clusters	Rays/Cluster	Frequency	Application
3GPP TR 38.901	7 to 24	20	0.5 to 100 GHz	5G NR system simulation
WINNER II	6 to 20	20	2 to 6 GHz	4G LTE evaluation
IEEE 802.11 TGn/TGax	2 to 6	Variable	2.4 / 5 / 6 GHz	Wi-Fi MIMO testing
Saleh-Valenzuela	Poisson (2 to 10)	Poisson (5 to 30)	UWB (3 to 10 GHz)	Indoor UWB, 802.15.4a
3GPP TR 36.873	6 to 20	20	2 to 6 GHz	3D MIMO evaluation

Common Questions

Frequently Asked Questions

How are clusters defined in 3GPP channel models?

3GPP TR 38.901 generates 7 to 24 clusters per scenario with delays from a scaled uniform distribution and powers from exponential decay. Each cluster has 20 rays with Laplacian angular distribution. The two strongest clusters split into three sub-clusters. Azimuth and elevation angles are assigned based on scenario-specific angular spread parameters. Cross-polarization ratios of 8 to 12 dB are assigned per ray.

What is the Saleh-Valenzuela model?

The foundational (1987) cluster model for indoor channels. Clusters arrive via Poisson process (rate 0.05 to 0.5/ns), rays via secondary Poisson (0.5 to 5/ns). Cluster powers decay with time constant Γ (10 to 50 ns), ray powers with γ (1 to 10 ns). Extended versions add angular dimensions for MIMO. The framework captures the physical observation that multipath arrives in bursts from distinct scatterers.

Why are cluster-based models important for MIMO?

MIMO capacity depends on angular multipath diversity. Cluster models assign distinct AoD/AoA per cluster, creating the spatial diversity enabling multiplexing. The number of resolvable clusters determines effective channel rank. For massive MIMO (64 to 256 antennas), 5 to 15 clusters typically dominate, meaning practical rank is 5 to 15. This drives RF chain count, codebook, and hybrid beamforming design.

Related Terms

← Cloud RF Cluster Multipath →