Standardized Channel Models

From Theory to Standards

The statistical models in Sections 6.3–6.7 describe general fading behaviour. For system design and performance comparison, the industry uses standardised channel models that specify exact PDP tap powers, Doppler spectra, and spatial parameters for defined scenarios (urban, rural, indoor, etc.). These models enable fair, reproducible comparison of different technologies.

Definition:
3GPP Channel Models

The 3GPP defines channel models across generations:

3GPP TR 25.996 — Spatial Channel Model (SCM):

6 taps per scenario (urban micro, urban macro, suburban)
Fixed PDP + per-tap angle of arrival
Used for 3G MIMO evaluations

3GPP TR 36.873 — 3D Channel Model:

Adds elevation angles for full 3D spatial modelling
Used for LTE-Advanced (4G) massive MIMO studies

3GPP TR 38.901 — 5G NR Channel Model:

Frequencies 0.5–100 GHz
Scenarios: UMa, UMi, InH, RMa (rural macro)
Cluster-based: 20–25 clusters, each with 20 sub-paths
Time evolution, spatial consistency, blockage model
Supports up to 256 antenna elements

Definition:
Geometry-Based Stochastic Channel Model (GSCM)

Modern 3GPP models use the GSCM approach:

Drop users randomly in the scenario
Assign large-scale parameters (delay spread, angular spread, K-factor, path loss) from correlated log-normal maps
Generate clusters (scattering groups) with random delays, powers, and angles of arrival/departure
Within each cluster, generate sub-paths (rays) with small angular and delay offsets
Compute the channel matrix $\mathbf{H}(f; t)$ by summing all ray contributions with appropriate antenna array responses

This is a hybrid between pure stochastic models (Rayleigh/Rice) and deterministic ray tracing (Section 5.6).

Definition:
3GPP TDL Channel Models

For link-level simulations, 3GPP TR 38.901 defines simplified TDL (Tapped Delay Line) models with pre-defined tap powers and delays:

Model	Scenario	Delay spread	K-factor
TDL-A	NLOS, clustered	$\sigma_\tau$ : scaled	0 (Rayleigh)
TDL-B	NLOS, uniform	$\sigma_\tau$ : scaled	0 (Rayleigh)
TDL-C	NLOS, extended	$\sigma_\tau$ : scaled	0 (Rayleigh)
TDL-D	LOS	$\sigma_\tau$ : scaled	13.3 dB
TDL-E	LOS, strong	$\sigma_\tau$ : scaled	22 dB

The delay values are normalised and scaled by a desired $\sigma_\tau$ to match any deployment scenario.

Historical Note: Evolution of Standardised Channel Models

1990s: ITU-R models (Vehicular A/B, Pedestrian A/B) — simple 6-tap models, no spatial information
2003: 3GPP SCM — first standardised MIMO model with angles
2007: WINNER II — extended to 100 MHz bandwidth, more scenarios
2012: 3GPP 3D model — elevation angles for FD-MIMO
2017: 3GPP TR 38.901 — extends to 100 GHz for 5G NR
Present: QuaDRiGa (Fraunhofer HHI) — open-source implementation enabling massive MIMO and 6G research

Each generation added spatial resolution and bandwidth, driven by the evolution from single-antenna narrowband to massive-MIMO wideband systems.

Why This Matters: Channel Models Drive 5G Design

The 3GPP TR 38.901 channel model directly shaped 5G NR design:

Beam management: the cluster-based model with specific angular spreads justified the beam sweeping procedures
Numerology: the delay spread ranges determined viable OFDM subcarrier spacings (15, 30, 60, 120 kHz)
Pilot density: Doppler spreads from the model set the minimum pilot spacing in time
Massive MIMO: spatial channel correlation from the model proved the feasibility of 64+ antenna arrays

Without standardised channels, comparing 5G candidate technologies would have been impossible.

Common Mistake: Using the Wrong Channel Model

Mistake:

Evaluating a 5G mmWave system using the ITU Vehicular A model (designed for 2G at 900 MHz).

Correction:

Always use a channel model matched to the frequency band, scenario, and antenna configuration. For 5G NR:

Sub-6 GHz: 3GPP TR 38.901 UMa/UMi
mmWave (28+ GHz): 3GPP TR 38.901 with specific mmWave parameters
Massive MIMO: use the full GSCM (not simplified TDL)

Quick Check

In the GSCM approach used by 3GPP TR 38.901, what is a "cluster"?

A group of base stations

A group of scatterers with similar delay and angle

A group of OFDM subcarriers

A set of channel measurements at one location

Correction:

A group of scatterers with similar delay and angle

Correct. A cluster represents a group of scattering objects (e.g., a building face) that produce multipath components with similar delays and angles of arrival/departure. Each cluster is further decomposed into sub-paths (rays).

GSCM

Geometry-Based Stochastic Channel Model: combines random cluster generation with geometric ray tracing. Used by 3GPP for system-level simulations.

TDL Model

Tapped Delay Line model with standardised tap powers and delays (3GPP TDL-A through TDL-E). Used for link-level simulations.

Discrete-Time Baseband Channel Model Summary