Propagation & Channels

Coverage Prediction

Q: Which propagation model should I choose for coverage prediction?

Model choice depends on frequency, environment, and the cell radius. For macrocell prediction from 150 MHz to 1500 MHz in urban and suburban areas, the Okumura-Hata model is standard; COST 231-Hata extends it to 2000 MHz. For irregular terrain and point-to-point links from 20 MHz to 20 GHz, the Longley-Rice (ITM) model uses a terrain profile and statistical refractivity. For dense urban microcells and millimeter-wave 5G, deterministic ray-tracing over 3D building data gives the accuracy that empirical models cannot. The Erceg model targets fixed wireless at 1.9 GHz, while the ITU-R P.1546 method is preferred for broadcast and regulatory coordination. Most planning tools let you tune a model with drive-test data to reduce the standard error from a typical 8 to 12 dB down to 6 to 8 dB.

Q: How accurate is RF coverage prediction compared to drive-test measurements?

Untuned empirical models such as Okumura-Hata typically show a location-variability standard deviation of 8 to 12 dB versus measured received signal level. After calibrating the model against drive-test data for the specific morphology and clutter set, the residual standard error usually drops to 6 to 8 dB. Deterministic ray-tracing in well-surveyed urban areas can reach 4 to 6 dB but requires accurate 3D building and material data. Because shadow fading follows a log-normal distribution, planners add a fade margin (often 8 to 10 dB for 90 percent area reliability) on top of the median prediction so that the design threshold accounts for the spread between predicted and actual signal.

Q: What input data does a coverage prediction need?

A useful prediction needs a digital elevation model (DEM) for terrain, a land-use or clutter layer to apply morphology-dependent loss, the transmitter location with antenna height, the antenna horizontal and vertical patterns with electrical and mechanical tilt, EIRP from the power amplifier and feeder losses, the operating frequency, and the receiver sensitivity or required signal threshold. Resolution matters: 30 m DEM and 5 to 20 m clutter rasters are common for macrocells, while microcell and millimeter-wave studies need 1 to 5 m 3D building vectors. The link budget then converts predicted path loss into received power so the tool can color-code where the signal exceeds the service threshold.

/KUH-vur-ij prih-DIK-shun/

Before a single antenna mast is raised, engineers estimate how strong the radio signal will be at every point across a service area, a process that drives site selection, antenna tilt, and capacity planning. The calculation feeds a chosen propagation model with digital terrain, land-use clutter, antenna patterns, and transmit power, then converts the resulting path loss into received signal level using a link budget. The output is a color-coded map showing where the predicted median signal exceeds the service threshold, typically with a log-normal fade margin of 8 to 10 dB applied for 90% area reliability. Empirical models carry an 8 to 12 dB standard error before calibration; ray-tracing over 3D building data can tighten that to 4 to 6 dB in well-surveyed urban cells.

Category: Propagation & Channels

Typical Std. Error: 6 to 12 dB

Area Reliability: 90 to 95%

How a Coverage Prediction Is Built

A coverage prediction starts with the physical environment. A digital elevation model (DEM) supplies terrain heights, and a land-use raster classifies each pixel as water, open land, suburban, dense urban, or forest so the engine can apply morphology-dependent clutter loss. For macrocell studies, 30 m terrain and 5 to 20 m clutter resolution are common; microcell and millimeter-wave work demands 1 to 5 m 3D building vectors. The transmitter is defined by its location, antenna height above ground, EIRP, and the antenna's horizontal and vertical patterns with electrical and mechanical downtilt. The engine then walks radials or pixels outward, summing free-space loss, terrain diffraction, and clutter attenuation to produce a path-loss surface.

The second half of the workflow is the link budget. Received signal level equals EIRP minus path loss plus receive antenna gain minus feeder and connector losses. Wherever the predicted median received power exceeds the receiver sensitivity (or the threshold set by the required signal-to-interference-plus-noise ratio), the location is counted as covered. Because shadow fading scatters real measurements around the median by a log-normal distribution, planners do not design to the median alone; they add a fade margin sized to the desired area or cell-edge reliability, so the served region is the area where signal exceeds threshold with the target probability.

Modern tools blend methods. Empirical models such as Okumura-Hata and COST 231-Hata are fast and adequate for wide-area screening, while deterministic ray-tracing resolves individual reflection, diffraction, and waveguiding paths through street canyons at the cost of compute time and data quality. Calibrating any model against drive-test measurements for the local morphology is the single most effective way to cut prediction error.

Governing Equations

Received Signal Level:
RSL (dBm) = EIRP − L_path + G_rx − L_feeder

Coverage Condition (per pixel):
RSL ≥ S_rx + M_fade → covered

Path Loss (Okumura-Hata, urban):
L_path ≈ 69.55 + 26.16 log f − 13.82 log h_b − a(h_m) + (44.9 − 6.55 log h_b) log d

Fade Margin (log-normal shadowing):
M_fade = Q⁻¹(1 − P_edge) × σ_SF

Where f = frequency in MHz, h_b = base height (m), h_m = mobile height (m), d = distance (km), S_rx = receiver sensitivity, σ_SF = shadow-fade standard deviation (≈ 6 to 10 dB). Example: σ_SF = 8 dB, 90% cell-edge → M_fade ≈ 10 dB.

Propagation Model Comparison

Model / Method	Frequency Range	Best Environment	Typical Std. Error	Compute Cost
Okumura-Hata	150 to 1500 MHz	Urban / suburban macrocell	8 to 12 dB	Very low
COST 231-Hata	1500 to 2000 MHz	Urban macrocell	8 to 12 dB	Very low
Longley-Rice (ITM)	20 MHz to 20 GHz	Irregular terrain, P2P	7 to 10 dB	Low
ITU-R P.1546	30 MHz to 4 GHz	Broadcast, coordination	7 to 10 dB	Low
Erceg (SUI)	~1.9 GHz	Fixed wireless access	7 to 9 dB	Low
Ray-tracing (3D)	0.3 to 100 GHz	Dense urban / mmWave	4 to 6 dB	High

Common Questions

Frequently Asked Questions

Which propagation model should I choose for coverage prediction?

Choice depends on frequency, environment, and cell radius. Okumura-Hata (150 to 1500 MHz) and COST 231-Hata (to 2000 MHz) suit urban and suburban macrocells; Longley-Rice handles irregular terrain from 20 MHz to 20 GHz; deterministic ray-tracing over 3D building data is required for dense urban microcells and millimeter-wave 5G. Calibrating any model with drive-test data cuts the standard error from 8 to 12 dB down to 6 to 8 dB.

How accurate is RF coverage prediction compared to drive-test measurements?

Untuned empirical models such as Okumura-Hata show an 8 to 12 dB location-variability standard deviation versus measured received signal level. Calibrating against drive-test data for the local clutter usually drops the residual error to 6 to 8 dB, and ray-tracing in surveyed urban areas can reach 4 to 6 dB. Because shadow fading is log-normal, planners add a fade margin of 8 to 10 dB for 90% area reliability on top of the median.

What input data does a coverage prediction need?

It needs a digital elevation model for terrain, a land-use or clutter layer for morphology-dependent loss, transmitter location and antenna height, the antenna horizontal and vertical patterns with tilt, EIRP after feeder losses, the operating frequency, and the receiver sensitivity or required threshold. Macrocells commonly use 30 m DEM and 5 to 20 m clutter; microcell and millimeter-wave studies need 1 to 5 m 3D building vectors.

Related Terms

← Coverage Planning Coverage Probability →

Link Hardware

Build the Link, Not Just the Map

A coverage prediction is only as good as the front end behind it. RF Essentials supplies the low-noise amplifiers, frequency converters, and millimeter-wave antenna assemblies that turn a predicted signal level into real-world performance.

Get in Touch