Engineering

The Science Behind
Axiorad

Standards-based RF propagation physics with explicit fidelity labels (validated / planning-grade / approximate / idealized). Multiple propagation models, an authoritative CPU-worker path, and multi-source DEM support. See methodology.

Every calculation traces back to a published standard or peer-reviewed reference. No proprietary black-box models — the physics are auditable, the assumptions are explicit, and the results are reproducible.

01 — Propagation Models

Five Propagation Models with Fidelity Labels

Five propagation models — including ITU-R and industry standards — cover a wide range of frequencies, distances, and environments. Automatic model selection evaluates each scenario's characteristics and applies the most appropriate physics. Not every model is an ITU-R Recommendation; see /methodology for labels.

Free-Space (Friis)

Any freq / distance

Baseline line-of-sight path loss for unobstructed propagation. Establishes the minimum possible path loss for a given frequency and distance.

FSPL = 20·log₁₀(d) + 20·log₁₀(f) + 92.45 dB

Ref: Skolnik, Barton

Two-Ray Ground Reflection

< 10 km

Combines direct path with specularly reflected ground path. Models constructive and destructive interference patterns that dominate at short ranges.

Optimised for flat terrain, short range

Ref: Barton (2004)

COST-231 Hata

150–2000 MHz · 1–20 km

Urban and suburban empirical model extended from the Okumura-Hata formulation. ITU-R validated for land mobile systems in built-up environments.

150–2000 MHz, 1–20 km range

Ref: ITU-R M.1225

Longley-Rice ITM

20 MHz–20 GHz · 1–2000 km

Irregular terrain model supporting LOS, diffraction, and troposcatter modes. Industry-standard model for wide-area coverage prediction over complex terrain.

NTIA ITM v1.4 — WASM implementation

Ref: Hufford et al. (1982)

ITU-R P.1812-6

30 MHz–6 GHz

Full ITU standard path-specific propagation model. Incorporates Bullington diffraction, anomalous propagation, gaseous absorption, and location variability.

Bullington diffraction + anomalous propagation

Ref: ITU-R P.1812-6

02 — Terrain Analysis

Line-of-Sight & Diffraction

Line-of-sight analysis with Earth curvature correction, first Fresnel zone clearance checking, and ITU-R P.526 knife-edge diffraction for obstructed paths.

Earth Curvature & Atmospheric Refraction

Configurable k-factor (effective Earth radius multiplier) for atmospheric refraction. Presets: standard atmosphere (k = 4/3), no refraction (k = 1.0), super-refraction / ducting (k = 2.0). Surface refractivity (N₀) input derives k automatically. k-factor is propagated through all LOS, diffraction, clutter, and GPU calculations.

Fresnel Zone Clearance

First Fresnel zone radius computed along each path profile. Configurable clearance ratio (standard 60%) determines whether a path is LOS-clear or obstructed.

Knife-Edge Diffraction

Fresnel-Kirchhoff diffraction parameter ν computed for each terrain obstacle. Diffraction loss J(ν) applied per ITU-R P.526 Annex A.

Multiple Obstacles — Deygout

For paths with multiple significant obstacles, the Deygout construction identifies the dominant knife-edge and applies corrections for secondary peaks.

Antenna Pattern Import

Load measured or manufacturer-supplied patterns in Radio Mobile (.ant), MSI Planet (.msi/.pln), or two-column CSV/TSV format. Auto-detection identifies the format and interpolates sparse data to 1° resolution.

03 — Environmental Effects

Atmospheric, Rain & Vegetation Attenuation

Three ITU-R standard environmental models quantify signal degradation from atmospheric gases, precipitation, and vegetation.

P.676-derived approximation

Atmospheric Absorption

Planning-grade specific attenuation from O₂ and H₂O vapour using the legacy P.676-10 closed-form approximation (not the current P.676-13 line-by-line method). Critical at the 60 GHz oxygen resonance band and the 22 GHz water vapour absorption line. Significant for millimetre-wave radar at any range.

ITU-R P.838

Rain Attenuation

Specific attenuation γ = k·R^α with frequency and polarization-dependent coefficients from P.838-3. The active engine applies uniform specific attenuation across the slant path; non-uniform rain-cell reduction is not yet implemented.

ITU-R P.833

Vegetation Loss

Weissberger model and the ITU-R saturation model for signal attenuation through woodland. Seasonal variation for deciduous trees — summer foliage adds 3–5 dB relative to winter bare-branch conditions.

04 — Detection Theory

Radar Detection Chain

Complete radar detection chain from the radar range equation through Swerling fluctuation models to detection probability. Two calculation methods are available: Albersheim's closed-form approximation for fast computation, and the Marcum Q-function for high-precision results at extreme P_d values.

Radar Range Equation

SNR = (Pₜ·Gₜ·Gᵣ·σ·λ²) / ((4π)³·R⁴·k·Tₛ·B·L)

Pₜ = transmit power · Gₜ,Gᵣ = antenna gains · σ = target RCS · λ = wavelength · R = range · k = Boltzmann · Tₛ = system noise temp · B = bandwidth · L = system losses

Swerling Fluctuation Models

Five target models (Swerling 0–4) covering steady, scan-to-scan, and pulse-to-pulse RCS fluctuation statistics. Target type selection affects the required SNR for a given detection probability.

SW 0Steady non-fluctuating target
SW 1/2Many independent scatterers (χ² 2-DoF)
SW 3/4Dominant scatterer + noise (χ² 4-DoF)

Detection Probability

v1.10.0

Two detection models are available. Albersheim's approximation (default) converts SNR, number of pulses integrated, and required P_fa to P_d via a fast closed-form expression. The Marcum Q-functionuses numerical integration (Rice integral with Bessel I₀ approximation) for higher accuracy at extreme P_d values (< 0.1 or > 0.99). Both support Swerling 0–4 fluctuation statistics.

Integration gain: n·SNR₁ (coherent) or ~n^0.5·SNR₁ (non-coherent)

Firm Track Probability

v1.1.0

M-out-of-N binomial model for track initiation probability. Computes the probability that a target generates M or more detections out of N consecutive scans — a common track initiation criterion (e.g. 2-of-3, 3-of-5).

P_track = Σ C(N,k)·Pd^k·(1−Pd)^(N−k), k=M..N

Vertical Coverage (Blake Chart)

v1.4.0

Range vs. altitude analysis reveals detection probability contours across the vertical coverage envelope. Shows multipath nulls, beam elevation limits, and coverage holes — essential for siting studies and engagement geometry planning.

Pd heatmap

Color-coded 2D grid (range × altitude)

Contour lines

50% and 90% Pd threshold overlays

SCNR mode

Clutter-limited when clutter modeling active

Interactive panel

Floating, minimizable chart panel

Antenna Pattern Modeling

v1.5.0

Import measured or manufacturer-supplied antenna patterns in multiple industry formats. The engine applies the loaded pattern to the radar equation gain term, replacing the default Gaussian beam approximation with real directional data.

Two-column import

CSV, TSV, or space-delimited angle/gain pairs

MSI/Planet format

Industry-standard .msi and .pln file support

Auto-detection

Format, plane (az/el), and delimiter detected automatically

Interpolation

Sparse patterns interpolated to 1° resolution; Gaussian fallback

Clutter Modeling

v1.3.0

Real-world radar detection is limited by clutter — unwanted returns from the ground, sea, and precipitation. The SCNR (Signal-to-Clutter+Noise Ratio) model combines thermal noise with surface and volume clutter to compute realistic detection probability.

Land surface clutter

Constant-gamma model with grazing-angle-dependent reflectivity. Configurable surface types: rural, urban, forest, custom.

Sea surface clutter

Georgia Institute of Technology (GIT) model, valid 1–100 GHz. Incorporates wind speed, sea state, and polarization effects.

Rain volume clutter

Backscatter via Marshall-Palmer Z-R relationship (Z = 200·R^1.6). Clutter cell volume determined by range resolution and beam geometry.

SCNR visualization

SCNR map mode replaces SNR when clutter is active. Range resolution parameter controls clutter cell area computation.

ECM Resilience Analysis

v1.7.0

Models noise jamming effects on radar detection. Stand-off jammers at fixed positions enter through antenna sidelobes; self-screening jammers on the target enter through the mainlobe. Supports barrage (wideband) and responsive (narrowband) modes with burn-through range calculation.

Stand-off jammer

Fixed-position jammer at a configurable lat/lon and altitude. Jamming power enters via antenna sidelobes; constant power across all target cells.

Self-screening jammer

Target-mounted jammer entering through the radar mainlobe. Jamming power scales with range² (one-way path), producing a range-dependent burn-through threshold.

Barrage / responsive modes

Barrage mode spreads power across the full radar bandwidth. Responsive mode concentrates power in the radar&apos;s instantaneous bandwidth, increasing effective jamming density.

S/(N+C+J) detection

Jamming power is summed with thermal noise and clutter in the interference denominator. Detection probability is computed from the composite SCNJ ratio.

Sensitivity Time Control (STC)

v1.8.0

STC reduces receiver gain at close range to prevent saturation from strong clutter returns. The gain ramps back up as range increases, restoring full sensitivity beyond the configured STC range limit.

G_stc(R) = −k·10·log₁₀(R/R_stc), R < R_stc

k = attenuation law exponent (2 = R², 3 = R³, 4 = R⁴). Applied as an SNR penalty before detection probability calculation.

Signal Processing

v1.9.0

Waveform-based processing gain replaces the flat gain assumption with physically-derived values computed from pulse width, pulse repetition frequency, and compression ratio. Coherent integration accumulates energy phase-coherently; non-coherent integration models the square-root gain of envelope detection. Simple mode retains a single flat processing gain input for quick estimates.

Pulse compression gain

G_pc = 10·log₁₀(τ·B_w) dB, where τ is pulse width and B_w is waveform bandwidth. Derived from time-bandwidth product.

Coherent integration

G_int = 10·log₁₀(N) dB — full N-pulse coherent gain. Requires phase stability across the dwell; applicable to pulsed-Doppler waveforms.

Non-coherent integration

G_int ≈ 10·log₁₀(N^0.5) dB — square-root gain from envelope detection of N independent pulses. Conservative estimate for incoherent receivers.

Unambiguous range & blind velocity

R_ua = c / (2·PRF). Blind velocity v_b = λ·PRF / 2. Both derived from PRF; displayed as waveform performance metrics alongside processing gain.

Range resolution: ΔR = c·τ / (2·compression_ratio)

Total processing gain = pulse compression gain + integration gain. Applied additively (dB) to the SNR from the radar range equation before detection probability is computed.

05 — Calculation Runtime

Authoritative CPU Worker

Coverage runs use the browser CPU worker so the selected propagation, detector, clutter, and result-metric contracts are applied consistently. An experimental WebGL kernel remains in the codebase for research, but it is not used for authoritative analysis runs.

01

Grid Generation

A polar-to-Cartesian coverage grid is generated at configurable resolution. Each cell represents a geographic point to be evaluated. Grid density is balanced against render performance.

02

Per-Cell Evaluation

The CPU worker evaluates each grid cell with the selected terrain, propagation, environmental, detector, and output-metric configuration. Work is kept off the main UI thread.

03

Float32 Output

Results are written to typed arrays containing the calculated metrics and coverage state, then transferred from the worker to the Deck.gl rendering layer.

Experimental GPU Kernel Status

The CPU worker is the authoritative calculation path. A simplified WebGL kernel remains available to developers for parity research, but it does not yet implement the complete detector, environmental, clutter, and output-metric contract and is therefore not used for authoritative analysis runs.

06 — Elevation Data

Digital Elevation Model Sources

Multi-source digital elevation model support with automatic fallback. Choose the source that best fits your accuracy and coverage requirements.

SourceResolutionCoverageType
SRTM30 m / 90 mGlobal (±60° lat)Free
Mapbox Terrain~30 mGlobalAPI
Google Elevation~30 mGlobalAPI
OpenElevation~30 mGlobalFree API

Elevation sources are queried per-tile as the analysis area is defined. If the primary source is unavailable, the engine automatically falls back to the next configured source. SRTM tiles are served from Axiorad's own CDN for consistent performance without requiring a Google or Mapbox API key.

Ready to Analyse

Start your coverage study
Ready to analyze?

Open the Axiorad analyser and apply these physics to your own radar scenario — no installation required.

Planning-grade models · explicit fidelity labels · see /methodology