1. What the engine computes
Given a transmit/receive pair, date/time, and space-weather indices, the engine produces the frequency window that will support the link (basic and operational MUF50, FOT, LUF), an expected time availability (reliability), and field-strength / SNR predictions per candidate frequency. The computation follows ITU-R P.533-14 as its top-level algorithm, with ionospheric inputs from ITU-R P.1239-4 (the CCIR maps) and noise from ITU-R P.372. P.533-14 specifies 2 to 30 MHz. Any result below 2 MHz is an extrapolation beyond that scope.
- Frequency window: the path basic MUF of ITU-R P.533-14 §3.1 (the higher of the lowest-order E and F2 modes' basic MUFs; fBM of §5.3.1 beyond 9000 km) and its decile envelope (10 / 50 / 90 %); operational MUF50 = basic MUF50 · Rop for F2 (ITU-R P.533-14 §3.7, P.1240-2 Table 1; fM of §5.3.1 beyond 9000 km); FOT = OWF, the operational MUF exceeded on 90% of days (P.533-14 §3.7, P.1240-2 §7: Fl · OPMUF, not capped, with Fl 0.95 where an E mode sets the path basic MUF and the P.1239 lower decile where an F2 mode does), so it can sit above the basic MUF50 and is always below the operational MUF50; and the LUF, the lowest frequency whose P.842 basic circuit reliability meets the required reliability at the required SNR.
- Field strength and SNR: ITU-R P.533-14 Part 2 by path length. Up to 7000 km, the §5.2.2 field of every mode §5.2.1 selects, each on its own ray (§5.1: eq 13 at the reflection height of eqs 14 to 16 for the operating frequency), with Lz = 11.39 dB (section 3.14) and the transmit and receive antenna gains read at that mode's own takeoff angle, summed by power (eq 28 for the field, eq 44 for the received power the SNR is worked from). From 7000 to 9000 km the §5.2 and §5.3 results are interpolated (§5.4, eq 42); beyond 9000 km the §5.3 composite field is used alone (eq 39, Ly = −0.39 dB, with the largest antenna gain between 0 and 8 degrees). The LUF is searched on the same field.
- Ionospheric layers: foF2, foF1, foE, foEs, M(3000)F2, hmF2 at path midpoint and per-hop reflection points.
- Absorption: P.533-14 equation 20 for each mode, averaged over the Table 1d control points along the great circle and scaled by that mode's number of hops.
- Mode selection: the field strength sums the modes P.533-14 §5.2.1 selects: up to three E modes on paths up to 4000 km and six F2 modes from the lowest-order one. F1 and sporadic-E modes are not in that sum. The engine also reports a controlling mode (the fewest hops at an elevation of at least 2 degrees), which P.533 does not define.
- Reliability: ITU-R P.842-5 Table 1 Step 11 basic circuit reliability, the probability that the predicted SNR meets the SNR the circuit requires, given the upper and lower deciles of that SNR. It is quoted against a required SNR that you set, because a reliability figure without its threshold cannot be checked by the person reading it. The day to day spread of the maximum usable frequency is already carried inside the P.842 deciles and is not applied a second time.
- Antennas: horizontal dipoles, inverted Vs, LPDAs and Yagis use the image method over a perfectly conducting ground, with height and frequency, less a flat 1.5 dB allowance for average ground; rhombics and Beverages use empirical elevation patterns; imported patterns are used as supplied.
- Path routing: short-path (great-circle) and long-path (antipodal) both supported.
2. ITU-R recommendations implemented
| Recommendation | Scope | Where |
|---|---|---|
| P.533-14 | HF propagation prediction (master algorithm) | HFPropagationEngine.ts |
| P.1239-4 | Reference ionospheric characteristics: CCIR foF2 & M(3000)F2 from shipped 3° grids derived from native 1.5° reference maps, with the Fourier coefficients (76/49 terms) and the required epoch-1960 modified dip as the fallback while a month's map loads. Includes the 2025 editorial amendments to P.1239-4 | ccirGrid.ts, ionosphericModel.ts, ccirEvaluation.ts |
| P.371-9 | Sunspot number: the one you enter or the planner fetches is the published number (version 2). The P.1239 maps and P.533 take R12 on version 1, so the engine uses 0.6 times it (Annex 1 §2, equation 2). A solar flux on its own gives R12 by equation 4. | sunspotScale.ts |
| Engineering peak-height estimate | hmF2 proxy from M(3000)F2 and foF2 / foE, with bounds. This is separate from P.1239 §7 mapped minimum virtual heights h′F and h′F,F2, which are not implemented | ionosphericModel.ts::calculateHmF2 |
| P.533-14 §5.2.2 | Non-deviative absorption (D & lower-E region): equation 20 at the two 90 km penetration points of each hop, with the absorption factor of Figures 1 to 3 and the ½ the ITU reference implementation applies (section 3.8) | absorptionModel.ts::p533NonDeviativeAbsorption, p533AbsorptionTerm.ts |
| P.1240-2 | F2/F1/E numerical kernels for diagnostic modes, Table 1 operational factors, and OWF/HPF factors selected by the path basic-MUF layer. The headline path follows P.533, whose E and F2 range rules differ from the standalone P.1240 method (section 3.17) | p1240Muf.ts, p1240Opmuf.ts, p1240Eirp.ts |
| P.533-14 §5.1 to §5.4, §6 | Median field strength and received power: the §5.2 mode sum (each mode on its §5.1 ray, with its own antenna gains, Lz) up to 7000 km, the §5.4 interpolation to 9000 km, the §5.3 composite field (with Ly) beyond. Lz (11.39 dB) and Ly (−0.39 dB) are re-determined against the D1 databank as the Recommendation requires (section 3.14) | multiModeField.ts, fieldStrength.ts::makeP533SignalModel, longPathField.ts |
| ITU-R F.1821-1 | Representative digital-HF station characteristics for sharing studies. Mobile-fleet powers and noise figures are labelled illustrative inputs; no system conformance or automatic range check is claimed. | Example station inputs |
| NOAA SWPC D-RAP (not ITU-R) | NOAA X-ray equations with supplied flux and an application snapshot-time window (section 3.9); P.531 is trans-ionospheric context | absorptionModel.ts::drapFlareAbsorptionDb |
| P.533 §5.2.2 Table 2 | Auroral and other signal losses (Lh), averaged over the Table 1d control points as the Recommendation requires, with no geomagnetic storm term, as Table 2 has none | p533Auroral.ts |
| P.534-6 | Conditional monthly foEs exceedance proxy: the equation-6 log-linear form fitted to P.1239 quantiles. The full P.534 annual-map field-strength method is not implemented. | p1239SporadicE.ts, HFPropagationEngine.ts |
| P.372-13 map; P.372-18 | Radio noise. Atmospheric noise uses the real worldwide numerical map (CCIR Report 322), ported from the ITU reference implementation, which states P.372-13, and reproducing it to within 0.003 dB. Galactic and man-made noise are the P.372 curves, with the man-made noise deciles of P.372-18 Table 2; galactic noise is counted only above foF2 over the receiver, since it is not observed below it (P.372-18 §4.1). All of it is taken at the receiving site, as P.533-14 §7 does. The system noise is referred to the receiving antenna's terminals (P.372-18 eq 5), so the loss of a lossy receiving antenna, such as a short whip, lowers the external noise along with the signal, and the receiver's line loss and noise figure are added to it. | p372Atmospheric.ts, noiseModel.ts, rfChain.ts |
| P.842-5 | Basic circuit reliability, Table 1 Step 11 piecewise BCR, computed against the required SNR below. The SNR deciles are P.842-5 Steps 4 to 9: Table 2 for the signal, and the noise deciles of each source weighted by its power (Steps 6 and 9). Measured against ITURHFProp: mean error 8.4 points, bias +1.9 points (section 4). | statisticalReliability.ts |
| F.339-8 | Required SNR for SSB, AM, CW and RTTY: Table 1, with the grade of service you pick, converted to the link bandwidth. Sky wave takes the fading column and a ground wave the stable one (F.1487 §3). FT8, FT4, JT65 and WSPR take their authors' published thresholds. A mode with no published figure needs one entered from the equipment's specification, and the source is printed beside every result. | requiredSnr.ts, hfDataModemStandards.ts |
| MIL-STD-188-110D, 141D | Data services and ALE linking: each 110D serial, Appendix C and Appendix D (3 kHz) rate at the bit error rate the standard prints, on the printed test channel that covers the historical P.842-4 §9 channel-spread approximation at the operating frequency, with 110D's allowances for a modem built into the radio; 2G ALE linking from 141D Table A-II. | dataService.ts, hfAleStandards.ts |
| P.368-10 | Ground wave, for coverage and along a route, on smooth earth with Millington's method over mixed land and sea, with P.368's own reference antenna (a short vertical monopole, 4.77 dBi) at each end plus your gain offsets, less each real antenna's circuit loss. Ends of different polarisation get no ground-wave figure. Reported beside the sky wave, never added to it, since no ITU-R Recommendation combines them: along a route each sample gives both, and the best-of coverage map shows, cell by cell, whichever of ground wave, sky wave or NVIS reaches the higher reliability, with that mechanism's own figures. | groundwave/p368.ts, groundwave/millington.ts |
| Report M.2412-0 | Doppler on a moving mobile, eq (28) and (30), for each P.533 §5.2.1 mode at its own arrival angle; level motion reduces to the P.681-11 eq (27) direct-signal projection (supporting identity only, not the satellite channel models). Checked against the offsets 110D and 141D test equipment at, with endpoint shift changes divided by 110D's 3.5 Hz/s sweep as duration lower bounds only; this does not bound peak turn slew or establish equipment conformance and is not used in reliability. | doppler.ts, routeRunner.ts |
| P.525-5 | Standalone vacuum free-space reference; stated field/power units and applicability | p525Conversions.ts |
| P.341-7 (2023 amendments) | Antenna-terminal system loss, total link loss including feeders, and grounded reference receive capture gr=3/4 | p341ReferencePlanes.ts |
| IGRF-14 (epoch 2025) | Modern MODIP used only for the separate PyIRI comparison. The production coefficient maps use the P.1239 epoch-1960 field | modipCalc.ts |
3. Documented deviations from ITU-R literal
These are the places where the engine departs from, or adds to, the printed Recommendations, each with its rationale. Deviations the engine no longer applies are kept below, marked removed, so that a result from an older engine version can still be read.
3.1 Equatorial Ionisation Anomaly (EIA) damping (REMOVED)
This engine used to damp predicted foF2 in the daytime on paths whose midpoint fell in the anomaly crest region. That damping has been removed. It is not part of ITU-R P.1239, and the engine agrees better with the ITU reference implementation without it. foF2 is now the CCIR map value everywhere.
3.2 Tropical reliability ceiling (REMOVED)
This engine used to cap reliability between 75 and 95 per cent, graded on latitude, for paths whose midpoint sat within 20 degrees of the equator. That cap has been removed.
It was not a refinement, it was a defect. ITU-R P.533 Attachment 1 correctly returns zero probability of equatorial scattering in daylight, and the engine was reading that correct zero as "no value supplied" and falling back to the empirical ceiling. On a daytime equatorial path the reliability the tool reported was the ceiling itself, not a prediction: it did not respond to power, antenna, frequency or noise. Equatorial scattering is now applied only where P.533 says it occurs.
3.3 Greyline reduction (REMOVED)
This engine used to take 8, 5, 4 or 2 dB off the absorption, and so off the LUF, on paths that cross or touch the solar terminator. That reduction is no longer applied: it is not in P.533-14, and the field now follows the Recommendation. The planner still marks a greyline path, and the figure shown beside that mark does not enter the prediction.
3.4 SNR spread σ and the stability label
The default reliability takes its spread from P.842-5 itself: the signal, within-the-hour and noise deciles it prescribes. The SNR σ shown with each prediction is that spread, the upper and lower deciles taken together as σ = (DU + DL) / (2 × 1.2816).
The stability label is the engine's own operational indicator, not an ITU-R quantity: no Recommendation defines one. It weighs a spread of the engine's own (a 4 dB base with terms for the K index, hops, night, latitude, closeness to the MUF, multipath and the two of §3.12, held between 3 and 20 dB) with the number of active modes, the headroom to the FOT and the day's MUF range. No published standard gives these values; that spread also sets the reliability in the VOACAP-compatible calibration mode, and nowhere else.
3.5 Sporadic-E
foEs and its deciles come from the ITU-R P.1239 numerical maps. P.533 has no sporadic-E model, so the sporadic-E modes are additions to it, on P.533's E-mode mirror height of 110 km. When a sporadic-E mode controls the path, the reliability at a frequency is multiplied by the monthly critical-frequency exceedance proxy at vertical incidence, f·cos(i). It fits the log-linear form of P.534-6 equation 6 through the P.1239 median and upper decile and caps it at 1. A one-hop estimate samples the midpoint; a two-hop estimate samples both quarter points and uses the smaller marginal probability. This assumes common quantiles between the reflection points; it is not a measured joint probability. There is no sporadic-E blanketing term: P.533's median field, fitted to the CCIR D1 measurements, already holds the median screening, and a blanketing loss built from published sources (Sinno et al. 1976 blanketing ratios, VOACAP's obscuration term) moved the field further from those measurements (rms 9.69 to 9.82 dB).
P.534-6 (September 2021) remains current. Its empirical Phillips rule supports mid-latitudes, frequencies above normal foE and time percentages below about 30. Using a 50% anchor, extrapolating below the median, or applying it at equatorial/high latitudes is a planner approximation. The proxy estimates critical-frequency exceedance, not independent Es presence: monthly foEs reductions can include normal E when no sporadic layer is observed. Coincident quantiles cannot define a finite Phillips slope; the planner uses a limiting step of 1 below the median, 0.5 at it and 0 above it. Zero anchors use a separate no-layer policy. Invalid anchors, thresholds and Es ray angles are refused. Applied results retain their quantiles, locations and policy in the frequency rows and PDF.
Equation 7 instead interpolates the separate annual 0.1%, 1%, 10% and 50% maps. Those maps and the full P.534 field/loss methods, effective Earth radius 8500 km, Es height 120 km and horizon diffraction are not implemented. P.534's full method covers distances up to 4000 km at VHF and potentially upper HF, with latitude and regular-layer cautions. This planner's one/two-hop geometric filters and 110 km rays are separate policies, not that field model's validity limits. Multiplying regular-layer SNR availability by the proxy does not calculate an additional Es signal or demonstrate end-to-end Es service availability. Regular-layer predictions are unaffected by this factor.
3.5a Ground-wave source and applicability
P.368-10 (August 2022) remains current. Its integral C++ v1.1 software is authoritative for the homogeneous solver. Both the flat-earth/curvature and Airy residue-series branches are implemented; invalid inputs and solver failures do not select another model. Numerical agreement on 3575 published vectors is checked at their six-decimal precision, with a fresh official C++ comparison. It is not bit identity or a measured physical-accuracy guarantee.
The integral-software bounds are 0.01–30 MHz, 0.001–10000 km, terminal heights 0–50 m, Ns 250–400, positive power and conductivity, and relative permittivity at least 1. Coverage and routes retain supported requested values. A route ground wave outside these bounds is unassessed with a reason; its sky-wave result remains separate. Coverage grids support at most 2048 frequencies and mixed paths 4–4096 segments, with 24 segments by default. These workload limits are software policies.
P.368 assumes a smooth surface, negligible ionospheric contribution and antennas on or near the ground. Configured heights are used without silently clipping them. Height above the average path terrain and platform altitude relative to that terrain are not derived; the 50 m software ceiling alone does not prove physical applicability. For kr at or below 10, results carry a radiation-field qualification. The optional total near-field correction in Annex 1 Note 3 is not applied. The integral software's horizontal branch is available, but cross-polar coupling is not computed and there is no fixed 30 dB horizontal-polarization penalty.
Annex 2 averages forward and reverse field differences in dB. Classification uses spherical segment midpoints, including at the dateline. Its Natural Earth/climate-proxy classification and fixed ground presets are coarse planning assumptions, not a measured conductivity raster. A single Ns from the arithmetic mean endpoint latitude is a latitude-only proxy, not a P.453 map or path profile. Branch metadata describes dispatch at the total path distance; mixed paths can evaluate both branches internally. Terrain roughness excess remains a separate engineering approximation, without a P.526 conformance claim. Weak coverage SNRs are retained exactly; the below-noise label at -30 dB changes no stored SNR or reliability.
3.5b Ground material assumptions
P.527-6 remains current; this review uses the English text posted in July 2026 with editorial amendments. Its Attachment Figure 24 supplies typical reference curves. The fixed HF approximations used here are sea water (εr 70, σ 5 S/m), fresh water (80, 0.003), wet ground (30, 0.01), medium dry ground (15, 0.001) and very dry ground (3, 0.0001). They are not full frequency-dependent curves or the complete Annex 1 material models. Fresh water on curve C differs from pure water on curve F; the old claim of negligible material error throughout HF is not supported by this review.
The saved fresh_water selection keeps its historical εr 80 / σ 0.03
brackish assumption and carries a legacy label. New freshwater selections and mapped
lake paths use the separate typical-freshwater choice. The earlier εr 15 / σ 0.003
inland default remains a declared assumption, distinct from the P.527 medium-dry choice.
Brackish water and urban ground are engineering assumptions. Ice retains a dry-ground
geographic proxy; no temperature-dependent ice model is implemented.
Lake selection uses the bundled Natural Earth polygons, with a named saline-lake exception. These coarse and sometimes historical outlines do not determine current shorelines or salinity. Humid-region boxes do not calculate soil moisture or layering. Reflection coefficients use elevation complementary to P.527's normal-incidence angle, with the existing BS.705 rounded 18000 σ/fMHz complex-permittivity convention. Antenna pattern directivity and separately charged efficiency/circuit losses retain their existing accounting. The legacy geographic per-hop loss policy is empirical; it does not evaluate P.527 Fresnel or material equations.
3.5c Conductivity-atlas scope and geographic defaults
P.832-4 (July 2015) remains current. Its atlas is not loaded by this planner. Table 1 lists nine conductivity standards in S/m for presenting maps; matching a listed value does not make a preset an atlas-derived local estimate or require every standard to be a software preset. The maps use mS/m: 1 mS/m equals 0.001 S/m. VLF maps apply up to 30 kHz; MF maps are standardized to 1 MHz. Seasonal variation is omitted, some geographic coverage is provisional or unavailable, and effective conductivities can already include terrain effects. These maps do not establish an HF accuracy bound for the planner's different geographic proxy.
Automatic selection uses retained Natural Earth coastline/lake geometry and declared climate/material assumptions. Caspian membership uses the interior ring from pinned Natural Earth v5.1.2 land data, restoring water omitted by the older packing. Aral membership uses the retained historical lake polygon. Water rectangles no longer extend across nearby land. The sea/brackish conductivity choices for those waters retain their legacy meanings; neither chemistry nor current shorelines are inferred. Geometry is distinct from the ITU conductivity atlas. The pinned Natural Earth land source supplies the Caspian ring, without a conductivity or salinity assignment.
Ground-wave footprint shape depends on assumed material paths, distance, refractivity, receiver noise, selected frequency and optional terrain excess. Range-curve bends can also arise from radial aggregation; they do not locate measured conductivity boundaries. A named antenna's azimuth pattern is not applied to the surface-wave footprint. The field at a fixed frequency is static here, while SNR and reliability can vary with noise and frequency selection.
3.5d Refractivity source and surface Ns
P.453-14 (August 2019) remains current; the June 2023 text includes 2022/2023 editorial amendments. Figure 1 is February monthly mean sea-level N0; Figure 2 is August. They are not an annual latitude-only surface-Ns curve and are unavailable as numerical maps. Section 2.1 distinguishes sea-level N0 from surface Ns using site altitude. Its terrestrial reference profile has N0 = 315 and scale height 7.35 km; the older Figures 1/2 used 9.5 km. The section 2.2 digital maps contain only the wet term Nwet, which cannot replace total Ns.
The HF planner evaluates none of those maps, meteorological equations, altitude profiles or gradient statistics. Its retained engineering curve uses absolute latitudes 0, 20, 40, 60 and 90 degrees with 360, 340, 315, 295 and 280 N-units. No fit or measured accuracy bound is established. A path takes one Ns from the arithmetic mean endpoint latitude; both latitudes must be finite and within −90 to 90 degrees. This is not a spherical midpoint or a refractivity profile. Newly computed ground-wave coverage and route results record their actual Ns and proxy identity; older missing values stay unrecorded. The native P.368 effective-radius expression consumes that assumed Ns, independently of a P.453 gradient-derived k factor. Standalone wrapper 314 and WSPR 315 defaults remain separate assumptions. Existing valid-input predictions are unchanged.
3.5e P.525 field and available-power references
The current P.525-5 (November 2024) is a free-space and plane-wave reference. Equation 5 uses vacuum wavelength c/f (c = 299792458 m/s); the standalone km/MHz helper uses 32.4477832219 dB. Equation 6 prints rounded 32.4. Equation 9 gives matched available power in dBW with frequency in GHz and constant 167.2; MHz changes that constant to 107.2, and dBm to 77.2. The ground-wave application keeps its declared two-decimal vacuum-aperture convention 77.22 (unrounded 77.2189957691). Native P.368 and P.533's prescribed 136.6/32.45/107.2 constants are retained. P.525's transmitting ground half-space effect is already included in P.368; no extra 3 dB is added to the field. Receiving aperture is treated separately by P.341. These are not general near-field or measured-accuracy claims.
New sky-wave fields are incident RMS values, including transmit losses. Receiver efficiency, polarization coupling and noise affect available receiver power and SNR, not that field. P.533 sums fields and received powers with different modal antenna weights; a displayed dominant-mode receive gain cannot generally convert between the combined quantities. Directional budgets and previews use the actual reverse row at the forward channel, including its field and receiving-system noise. New results and CSVs record the incident-field identity; older missing references remain unrecorded. Reports do not fill missing propagation quantities with a free-space prediction, and MDS uses the recorded required SNR.
3.5f P.341 terminal planes and ground-wave capture
P.341-7 with its 2023 editorial amendments distinguishes antenna system loss Ls (section 1.4, includes circuit losses and excludes feeders) from total link loss Ll (section 1.5, transmitter output to receiver input). Current budgets record both power differences from the actual directional signal. Receiver NF changes noise only. The retained P.533 equivalent field loss is a separate diagnostic, not Ll or a unique modal basic loss. Missing older terminal references stay unrecorded.
Annex 3 equation 12 gives gr=gt/4=3/4 for the short vertical monopole surface reference; transmit directivity is gt=3. Ground-wave coverage and routes use this capture factor plus user offsets: signal and SNR are 6.02 dB lower than when transmit directivity was incorrectly used for reception. P.368 field, native software diagnostics and P.372 reference noise are unchanged. This reference-antenna approximation does not model the in-situ receiving aperture of arbitrary elevated antennas over real soil. Saved results retain their actual engine and receive-reference evidence; recomputation applies the correction.
3.5g M.2412 motion Doppler
Report M.2412-0 remains current (register November 2017; cover October 2017; current PDF posted January 2021). Only Annex 1 equations 28/30 and the zenith coordinate convention are reused. The incoming-ray vector dotted with total 3D velocity, divided by vacuum wavelength, is positive on approach. Climb includes pure vertical motion. Ground-wave horizon bearing follows the short route even when sky routing is long. Modeled above-basic-MUF rays are retained; the transition resolves only part of the sky-ray set, and P.533 section 5.3 provides no resolved mode span. This unweighted motion span is distinct from ionospheric Doppler width and does not alter reliability.
A turn's endpoint frequency difference divided by the MIL test sweep of 3.5 Hz/s is a duration lower bound, not proof of peak slew or equipment conformance. Frequency/elevations come from the nearest route sample and the turn trajectory is unknown. The full IMT channel model, ionospheric dynamics and two moving terminals are outside scope. Missing older motion references stay unrecorded in reports and CSV.
3.5h P.681 supporting level-motion identity
The current P.681-11 (August 2019) text was posted in March 2021. Annex 1 section 6.2 Step 3 equation 27 describes a signed direct-signal Doppler spectral line. Its relative-azimuth and elevation cosines give the zero-climb reduction of the planner's M.2412 vector term. Here theta is elevation; in M.2412 it is zenith. Carrier frequency is Hz when speed and c are in m/s. Climbing motion keeps the full three-dimensional term.
P.681 concerns land mobile satellite channels, with a measurement context from 870 MHz to 20 GHz. The planner reuses only this generic kinematic identity for fixed apparent HF arrivals. It does not implement or validate P.681 satellite channel statistics at HF, satellite/orbit motion, Loo/Jakes fading or the supplied satellite-to-indoor simulator. The latter's README states L-band-only validation. A spectral line and an unweighted range of modal shifts are distinct from ionospheric PSD width and do not add a reliability penalty. This supporting review changes no numerical results.
3.5i Adaptive frequency planning (F.1611-0)
F.1611-0 remains current; the English PDF posted May 2010 includes December 2009 editorial amendments. Annex 1 sections 2–4 use predictions to shortlist channels, then sounding or real-time channel evaluation (RTCE) to choose a live channel. Section 4 gives a 90% hourly reliability design goal and three to five frequencies meeting the median SNR requirement. These are guidelines, not universal rules.
Daily plans and ALE pools retain exact evaluated channels. Each candidate uses its own SNR-based reliability; talk-back applies the weaker direction at that same channel. A nearby channel or LUF/FOT window overlap supplies no substitute result. Missing older assessments remain unavailable beyond their recorded operating channel. Coverage pools retain the computed candidates, including after edits to a frequency plan.
Four-hour blocks, greedy ranking, channel caps and stopping rules are planner policy. The selected service grade remains distinct from the 90% design goal. A pool’s share of eligible cell-hours and a daily mean do not establish per-hour availability. Channel failures can be correlated; the independent four-channel probability example in F.1611 is not used as a prediction. Seasonal and solar-extreme lifecycle studies, receiver-specific sensitivity checks and operational RTCE are separate design work. Propagation equations, measured reference sets and the fixed calibration remain unchanged.
3.5j Observed foF2 and the withdrawn P.434 reference
P.434-6 was withdrawn on 24 October 1997 and replaced by P.1239 and P.1240. Its Annex 3 predicts ray paths; it does not prescribe live-ionosonde assimilation. The current replacements do not prescribe this blend either.
Optional observed foF2 blending is a noIM3 extension. At each evaluated ionospheric location it blends up to three distinct stations with inverse square distance weights (a 1 km floor), full measurement-blend weight within 250 km of the nearest station and a linear taper to zero at 1500 km. Those limits, the station cap and the 60-minute model-time window are planner policy. Untimed observations are excluded by default. Live feed ingestion rejects invalid coordinates/values, future timestamps and readings older than 60 minutes. Reanalysis may use timed observations on either side of its model time within the specified window.
Feed availability does not establish use in a particular prediction. Current link and coverage traces retain the observations actually blended into evaluated model inputs, with station identity, coordinates, measured foF2 and observation time. Their evaluation count is not a count of unique cells or locations. Older proximity-only flags remain unverified. Traces survive worker merging and saved cubes, including pinned and storm copies. The P.1239 climatology and P.533 calculations retain their separate source basis; using observations with monthly reliability statistics is a planner extension, with no new nowcast or real-time accuracy guarantee.
3.5k Selected BS.705 antenna equations and generic models
BS.705-2 (November 2025) supplies the ground-reflection method and shape-only directivity in Annex 1, Part 1 sections 3.3 and 3.4, and the ground-mounted monopole equation (19). Its average ground is relative permittivity 4 and conductivity 0.01 S/m. The vertical half-wave dipole applies that reflection method to another element. Directivity assumes 100% efficiency; earth-system and antenna losses are separate. Datasheet efficiency replaces the generic whip loss. A 7 m marine whip above about 21 MHz extrapolates beyond the below-half-wave monopoles considered in section 7.1.
Raw patterns retain physical nulls. The Part 2 section 5.3 planning envelope is applied afterwards: maximum gain minus 25 dB below a 25 dBi peak, and 0 dBi for peaks of 25 dBi or more. Generic horizontal dipoles and beams use a perfect-ground image factor, fixed 1.5 dB allowance and empirical tapers. The terminated rhombic includes back rejection. These are generic proxies, without a held NEC accuracy bound or the full BS.705 geometric array catalogue. Imported antenna patterns replace the generic patterns.
Annex 2 defines a different relative receiving-whip environment, with permittivity 10, conductivity 0.01 S/m and the same response applied to wanted signals, interference and noise. Its urban attenuation median 11 dB and standard deviation 7 dB are not implemented. The current P.372 noise model is not a full directional antenna/noise-sky convolution. These selected methods do not establish broadcasting-array conformance.
3.5l P.373 frequency definitions and selected channels
The current P.373-9 (September 2013) distinguishes basic MUF, which concerns ionospheric refraction alone, from operational MUF under specified circuit conditions. Annex 1 defines OWF/FOT as the lower decile of daily operational MUF at the same UTC time, usually over a month: 90% of those daily MUFs exceed it. HPF is its upper decile, exceeded on 10% of days. These differ from basic-MUF deciles and do not establish end-to-end circuit availability. FOT means optimum traffic frequency.
The selected operating frequency is a separate planner choice. It can equal raw OWF, be raised to LUF or be a licensed channel. Closed-window placeholders do not identify usable frequencies. Current link results retain raw OWF and HPF separately; missing historical values stay unrecorded. Charts label the selected value as Operating, and their tooltips refer to actual channel results without assigning availability from its position relative to that marker. Known closed placeholders are omitted from frequency curves and CSV limits.
P.373 gives no universal LUF percentage or 0.85 factor. The planner uses the declared SNR, bandwidth and requested P.842 service grade; default 50% is the P.533 median-SNR criterion. When a requested grade cannot be met but median-day service exists, the reported LUF is a declared 50% fallback. Older records without grade/fallback metadata remain unknown. Coverage records do not store every cell's effective fallback grade, so a requested percentage does not prove a uniform effective grade. Ionospheric terms do not apply to ground-wave limits. Annex 1 notes 1–4 give mode, component, range and percentile notation; no separate extraordinary-wave MUF(X) is resolved by this planner.
3.6 LUF ≥ 25 MHz semantic
VOACAP emits LUF ≥ 25 MHz as a "no frequency meets the required reliability" sentinel. Our LUF is defined differently (the lowest frequency whose P.842 basic circuit reliability meets the required reliability at the required SNR, as section 4.2 states), so the two are not directly comparable above about 25 MHz. The VOACAP check in §4.1 drops LUF comparisons above that threshold rather than pretending they're comparable.
3.7 Above-MUF experimental extrapolation (REMOVED)
This engine used to estimate availability above MUF50 with a scattering-loss model of its own. That estimator is no longer used. Above each mode's basic MUF the field now carries the loss ITU-R P.533-14 §5.2.2 prescribes there (equations 24 to 26, as printed) on paths up to 9000 km; beyond 9000 km the §5.3 method is used alone. The frequency chart still hatches the band above the basic MUF10 and labels it experimental, as a caution: the basic MUF reaches that band on fewer than 10 per cent of days.
3.8 D-region sec(φ) cap and cosp exponent (REMOVED)
This engine used to work the non-deviative D-layer absorption from a fitted constant, with sec(i) capped and the cosp exponent fixed at p = 1.3. Those are gone. The absorption is now equation 20 of ITU-R P.533-14 §5.2.2: the absorption factor varies with latitude and month (Figure 1), its daily exponent p with magnetic dip and month (Figure 3), and sec(i) is not capped. It is evaluated, as equation 20 says, at the two 90 km penetration points of each hop (from a fixed 300 km reflection height), each with its own absorption term and gyrofrequency. The printed sum carries no ½, which doubles P.533-12's per-hop absorption for the same Figure 1 factors and reads 4.7 dB low against the CCIR D1 measurements (rms 14.1 dB); the ITU reference implementation halves it, and so does this engine (D1 rms 9.69 dB).
In one place this departs from ITURHFProp, on purpose. Figure 1 prints a southern-hemisphere month axis six months on from the northern one, and the engine enters the figure that way; the reference implementation does not. On southern-hemisphere paths the two differ by design. Figure 3 prints the same southern axis, and the engine enters it the same way.
3.9 Solar-flare absorption: NOAA D-RAP
P.531-16 section 6 gives trans-ionospheric absorption context, not the HF X-ray algorithm used here. The engine uses the X-ray part of the NOAA Space Weather Prediction Center model, D-RAP ("Global D-Region Absorption Prediction Documentation", 2010): the highest affected frequency is 10·log10(F) + 65 MHz at the sub-solar point for the 0.1 to 0.8 nm flux F in W/m², tapering as (cos χ)0.75 to nothing at the terminator, and the loss at a frequency f is (HAF/f)1.5 dB for a vertical round trip. The engine takes half of that at each of a mode's 90 km penetration points, at that point's solar zenith angle, made oblique by sec(i) at 110 km as equation 20 makes the D-region absorption oblique (D-RAP's 1/sin(elevation) is the flat-Earth form of the same factor). With the sun overhead at 10 MHz, one vertical round trip: C1 0.35 dB, M1 1.8 dB, X1 4.0 dB. This multi-hop obliquity treatment is a planner extension; SEP/PCA and recovery modelling are absent. For timed supplied flux, the planner applies a symmetric model-time window strictly shorter than one hour. This is application policy; snapshot fetch time does not establish GOES measurement time. Untimed supplied flux is an explicit scenario. The current loader does not automatically fetch X-ray flux.
3.10 Auroral storm increment (REMOVED)
This engine used to add a K-index increment of its own to the auroral loss. It is gone. No published model gave it, and the published ones found do not reach standard grade (Foppiano and Bradley's model understates the Kp effect against riometer data; Kavanagh et al. 2004 is one station; ICEPAC's auroral absorption reads 100 to 200 dB high). The auroral loss is P.533-14 Table 2 Lh alone, a monthly median with no geomagnetic-activity term.
3.11 P.842 reliability: piecewise BCR and multiplicative decomposition
The SNR term is computed by default from the P.842-5 Table 1 Step 11 piecewise basic-circuit-reliability (BCR) curve: the literal asymmetric sigmoid, not a symmetric Gaussian CDF. The Gaussian-CDF form is used only in the VOACAP-compatible calibration mode. That BCR term is then combined multiplicatively into R = BCR · P(Es open), where the sporadic-E factor applies only when the controlling mode is sporadic-E (§3.5). Where P.533-14 Attachment 1 predicts equatorial scattering, R is limited by its probability (§3.2). The day to day spread of the MUF is not a separate factor: it is already carried inside the P.842 deciles. The sporadic-E factor is an engine extension beyond P.842's single-circuit formula.
3.12 Additional σ contributions: TID and EIA spread-F
The empirical standard deviation in dB (§3.4) includes assumed independent TID and geographic equatorial terms: 2 dB plus 0.3 dB per K above 3, and a 5 dB maximum falling linearly to zero at ±35 degrees. These coefficients are application policy, with no held measurement validation or quantitative P.531 attribution. They feed stability and the Gaussian compatibility path; native P.842 BCR uses its separate signal/noise deciles when those inputs are available. P.531-16 section 5.1 equation 8 estimates peak-to-peak intensity fluctuation from S4; it does not define a universal conversion to Gaussian SNR sigma. No S4, NeQuick2 or GISM model is implemented. Series CSVs name this legacy snrVariance field empirical_snr_sigma_db; its values are standard deviations, not variances.
3.13 Where P.533-14 leaves the field geometry open
Equation 14's E1 is a cubic in f / foF2. On a hop longer than ds at a frequency many times foF2, far above every mode's basic MUF, it gives a reflection height of zero or less, where equation 13 has no ray. There the engine uses the mode's ray at its basic MUF, the ray §5.2.2 already prescribes for the absorption above the basic MUF. Such a mode carries the full above-the-MUF loss and adds nothing to the sum. The largest antenna gain between 0 and 8 degrees (§5.3) is found on a 0.25 degree grid.
3.14 Excess losses Lz and Ly: re-determined against D1
P.533-14 prints Lz = 8.72 dB (§5.2.2) and Ly = −0.14 dB (equation 39). It requires revision of these terms when the prediction elements change, using predicted minus measured field strength against the D1 databank. With this engine's documented choices, including the equation-20 half factor in section 3.8, engine 2.25.0 gives Lz = 11.39 dB from the 11 620 D1 hours below 7000 km and Ly = −0.39 dB from the 4 589 hours beyond 9000 km, where §5.3 applies alone. Their residual group means are +0.0030 and +0.0029 dB; the existing 0.05 dB calibration limits are retained. Engine 2.24.0 used 11.61 and −0.37 dB before the P.1239 input corrections described below.
D1 is calibration data. D2 is an additional measurement check, is not a direct fitting target and is never used to choose these terms. Its current mean and RMS errors are published in section 4.4, including the slight deterioration after this review.
3.15 Solar geometry and long-path interpolation
Engine 2.24.0 uses UTC day numbers throughout and includes the monthly equation-of-time correction in the absorption zenith angles required by P.533-14 section 5.2.2. The correction uses the NOAA GML approximation at noon on the 15th of the month, including leap years; the existing daily declination approximation is retained. Section 5.3.2 still uses the approximate hour angle of equation 35. The winter factor uses the selected arc midpoint, the long-path geometry consistently uses the specified 6371 km Earth radius, and Table 3 interpolates the azimuth angle linearly.
3.16 P.1239 ionospheric inputs and their scope
Engine 2.25.0 was checked against the current official P.1239-4 (08/2023), including the 2025 editorial amendments. The review corrected 78 of the 8 208 foF2 decile-table cells, the Fourier fallback's magnetic coordinate and 160 limit on R12 for both foF2 and M(3000)F2, and the magnetic field's reference radius, Legendre terms and polar calculation. The foE nighttime decay now uses hours after actual sunset; the nighttime minimum is not imposed in daylight. F1 follows equations 19 and 20 without the previous solar fade, twilight taper or frequency cap. Solar zenith uses apparent solar time with the declared daily-declination and monthly-equation-of-time approximations.
The shipped 3° grids preserve every second node of the held 1.5° maps, with quantisation of 0.001 MHz for foF2 and 0.0001 for M(3000)F2. Interpolation between retained nodes is an approximation. The fallback is another P.1239 map representation and can differ from the grid. Invalid or mismatched downloaded grid metadata is rejected before caching. Sporadic-E coefficients are checked against the official maps, but the planner's R12 limit of 150, positive floor and decile ordering remain engineering policies. fbEs, mapped virtual heights and spread-F maps are not implemented. The hmF2 estimate does not substitute for those maps. This review establishes the checked scope, not complete implementation of every characteristic in P.1239.
3.17 P.1240 numerical methods and operational factors
The current P.1240-2 (July 2015) was freshly checked for engine 2.26.0. Its F2 kernel now uses the uncapped distance formula unless the caller explicitly requests P.533's 4000 km limit. F1 no longer borrows the F2 solar saturation at R12 = 160. The 2E diagnostic for paths of 2000–4000 km uses E(2000)MUF with midpoint foE. Diagnostic F2 modes use gyrofrequency at 300 km; absorption retains its 100 km value. E screening is restricted to the P.533 path range of 4000 km or less.
Table 1 uses the midpoint's season and day/night state. The planner defines daylight as apparent solar zenith below 90°, using the astronomical approximations already stated. Its EIRP includes modeled antenna gain, gain offset and transmit-side feedline, mismatch and efficiency loss. For a frequency-dependent antenna or feedline, the reference is the path basic MUF limited to 2–30 MHz, with maximum modeled gain in the path's bearing sampled every 0.25° of elevation. This reference choice avoids feeding OWF back into its own power tier. It is an engineering approximation; the frequency and EIRP used are retained with the result. Receive loss and receive polarization mismatch do not reduce transmitter EIRP. The high-power tier starts strictly above 30 dBW.
The headline path remains the P.533 method, including its secant-law E modes and long-path K factor. It is not a separate implementation of MUFFY's complete path algorithm. The P.1240 diagnostic F1/E polynomials and shared Annex-2 mirror-height equations have been checked independently. The extra NVIS/Es modes, fixed F1 reporting height and mode-list elevation filters are planner policies; the latter can exclude an F1 diagnostic inside the polynomial's stated distance range.
P.1240's OWF and HPF multiply the operational MUF by the lower and upper factors selected by the layer giving the basic MUF. The raw OWF remains a distinct result. The legacy internal fot field holds the selected operating frequency: it can be raised to LUF, replaced by a licensed plan channel, or collapsed to a closed-window sentinel. Those operating choices do not change the theoretical P.1240 OWF. The text summary now prints the raw OWF and selected operating frequency separately.
3.18 P.371 solar indices and forecasts
The current P.371-9 (August 2023) was checked for engine 2.27.0. Its equation 3 defines a weighted running mean of 13 monthly means: half weight on the two endpoints, unit weight on the 11 interior months, divided by 12. This is a mean, not a median. An observed centered value needs six subsequent months. NOAA's current-month value is therefore a smoothed forecast, not an observation of the current month's R12. The service requires an explicitly smoothed field for the current UTC month, retains that epoch with each run, and refreshes the index when the UTC month changes. A missing current month requires an entered smoothed index; a nearby month is not silently substituted.
Equation 2 converts published version 2 sunspot numbers to version 1 by a factor of 0.6, once at the model boundary. Explicit version 1 R12 values are used as supplied. Equation 4 relates smoothed F10.7 (Φ12) to version 1 R12; flux-only overrides use its positive inverse. Observed daily F10.7 is displayed separately and cannot supply R12. The software permits R12 v1 from 0 to 300 (equivalent published v2 from 0 to 500), floors the inverse at zero below 63.7 sfu, and offers more conservative limits in some input controls. These are software policies, not P.371 limits. Other models' solar saturation rules still apply to their own equations. Saved coverage records and reports retain the actual R12 and index provenance; older records can only reconstruct an index from the inputs they saved.
The planner consumes NOAA's externally produced smoothed predictions. It does not implement the modified McNish-Lincoln forecasting method recommended in section 6, or infer daily ionospheric conditions from these long-term indices. Forecast uncertainty, especially beyond 12 months, and P.371's high-latitude cautions remain relevant. NOAA's separately predicted sunspot and flux series need not satisfy equation 4 exactly; the planner does not force them to agree. These checks establish input handling and the stated equations, not measured propagation accuracy or implementation of every recommendation in P.371.
3.19 P.372 radio-noise references and receiving systems
The current P.372-18 (September 2026) was checked for engine 2.28.0. All 39,168 atmospheric coefficients were compared with freshly obtained Study Group 3 supplemental files; the generator now retains their full source precision. The atmospheric calculation keeps that software's monthly tables, integer local-time convention and four-hour power interpolation. These are declared numerical choices. The PDF's atmospheric charts describe seasonal four-hour blocks; agreement with the supplemental code does not establish complete literal conformity.
The external noise used by the HF reliability chain is the power sum of component medians required by P.842 Table 1 Step 3. It is distinct from the moment-matched noise distribution in P.372 Part 7. P.372 equation 7 converts this noise figure to a short vertical monopole reference field using bandwidth in Hz. A 1 kHz bandwidth contributes 30 dB; the former conversion omitted that term. Equation 8 uses a different constant for a free-space isotropic reference. Neither conversion replaces the practical receiving-antenna calculation used for signal power.
Receiving-system noise uses P.372 equation 5 at the declared 290 K circuit and feeder temperature, referred to the terminals of the realized receiving antenna. The same affine power transformation now acts on upper and lower noise deciles before they enter P.842. Fixed receiver and feeder noise therefore reduce external-noise variability when they dominate; they are not an extra independent day-to-day fading process. Each link direction and ground-wave coverage use the actual receiving end's losses.
Galactic noise is excluded below foF2 over the receiver. Above that threshold the full curve is an upper-bound approximation through the partially shielded region up to about three times foF2. With no ionospheric index the unscreened curve is used. Quiet-rural curve D describes reflected man-made noise below the extraordinary critical frequency; applying its fit outside that regime is an extrapolation, and its rural time-decile proxy is a software policy because Table 2 has no quiet-rural row. City, residential and rural fits below 0.3 MHz are also extrapolations. Location deciles are not added to one circuit's time variability. Atmospheric impulsiveness/APD, full antenna-sky convolution and P.372 Part 7 statistical combination are outside this implementation.
3.20 P.842 reliability scope and path latitude
The current edition is P.842-5 (September 2013), freshly checked on 1 November 2026. Table 2 selects its high-latitude signal deciles when any point between the control points 1000 km from each end of the selected great-circle arc reaches 60° geomagnetic latitude, in either hemisphere. The engine finds the analytic maximum over that interval, including long paths. Coordinates retain the centred-dipole pole 78.3° N, 69° W used with P.1239 section 5; this is a coordinate approximation, separate from the exact arc search. At distances up to 2000 km, the midpoint tier is a stated software continuation because the ordered internal interval does not exist. Interpolation between Table 2 knots remains a linear software policy; the forty printed values are unchanged.
Table 1 Step 11 retains the full 0–100% range without a minimum decile spread. Zero spread uses the one-sided limit, with 50% at zero SNR margin. The inverse reaches 100% at its finite saturation knee; a zero requirement imposes no minimum SNR margin. The planning grade itself must be greater than zero. The LUF search shares the path tier and receiver-transformed noise deciles used in the frequency table. Route ground-wave deciles use that same receiver reference; their signal fading terms remain zero. This is a stable-signal ground-wave application of the reliability curve, not P.842's full sky-wave signal model.
The digital overlay retains historical P.842-4 section 9: the product of path availability and independent delay/Doppler tolerance estimates. P.842-5 removed that procedure and refers digital reliability to P.533. Current P.533-14 section 10.2 needs modal wanted/interfering power, time/frequency windows and amplitude ratio A; this overlay does not implement it. Above basic MUF the historical squared-ratio spread is an extrapolation, and zero spread with zero tolerance uses a deterministic passing convention. Data-service channel matching also uses this historical spread estimate. Fixed modem tolerances are separate inputs, not prescribed by P.842-5.
Sporadic-E presence, the equatorial-scatter ceiling and Gaussian compatibility remain separate planner estimates. The lower of the two directional availabilities is a planning comparison, not a calculated joint two-way probability. Interference, site/frequency/time diversity, service-area and network reliability clauses are outside this implementation. No claim of complete P.842 or current digital-procedure conformity is made.
3.21 F.339 required SNR and source conditions
F.339-8 (February 2013) remains current, checked on 1 November 2026. The stored Table 1 subset, all Table 2/3 points and Tables 4a/4b match the fresh official PDF, including its blank cells and unusual printed values. These are recommended estimates for specified systems, not universal receiver or modem specifications. F.339's stated scope is 3–30 MHz; using these thresholds below 3 MHz is an extrapolation.
Table 1 supplies RF signal-to-noise-density ratios, printed as dB/Hz and referenced to noise in 1 Hz. The displayed threshold is that ratio minus 10 log10(B), where B is the run's noise bandwidth in hertz. The predicted noise uses the same B. This is a change of SNR reference; it does not change the source's pre/post-detection receiver bandwidths or establish that a waveform fits another channel width. Audio SNR is not substituted for the RF ratio. Signal power is PEP except A3E AM, which uses carrier power; entered transmitter power must use the same reference. No automatic carrier/PEP/mean-power conversion or extra speech crest factor is used.
Sky wave retains the fading non-diversity column; stable ground wave uses the stable column. Table 1's service conditions remain part of the result: just-usable speech uses 90% sentence intelligibility, commercial public-network grades carry their specified protection, good commercial assumes 10 dB improvement from noise reducers, and the selected RTTY row assumes atmospheric noise Vd = 6 dB. These conditions are distinct from the requested monthly P.842 availability. Enter an equipment threshold when its performance or receiver conditions differ. No extra provisional 11.5 dB day-to-day allowance from note 4 is added to the separate P.842 variability model. No Lincompex improvement, diversity benefit or alternative protection figure is invented for a mode without a supported row.
A finite entered SNR takes precedence; otherwise a selected supported data target supplies its own standard's threshold, followed by a published mode threshold when no target was selected. Invalid entries and unsupported selected targets are refused. The target is an SNR planning criterion conditional on its printed test channel; service/channel assessment remains a separate result. Table 4a's 3 kHz waveform rates label the MIL-STD data-service rows; the MIL-STD supplies their SNR values. F.339 Tables 2, 3 and 4b are verified source data, not additional active modem models. Missing cells, unsupported channels and arbitrary rates are not interpolated.
3.22 F.1487 channel quantities and capability proxies
F.1487-0 (May 2000) remains current, including its December 2009 editorial amendments. Annex 3 describes two independently fading equal-power Rayleigh paths with equal Gaussian frequency spreads and no frequency shifts. Its delay is their differential separation and its frequency spread is the full 2σ width of the Gaussian power spectrum. All ten native channel pairs are retained. They are laboratory channel parameters, not modem limits or a prediction that a particular path has those conditions.
The historical P.842-4 overlay uses a −10 dB width. For a Gaussian power spectrum, that full width is sqrt(2 ln 10), approximately 2.145966, times the 2σ width. The selected channel proxies convert their frequency tolerance to that reference; predicted historical widths are converted back once before data/ALE channel screening. This conversion assumes a Gaussian spectrum and does not apply unchanged to arbitrary shifted or multicomponent spectra. Motion-induced frequency offset is a separate quantity. Differential delay is used as a two-path time-span proxy, not RMS delay.
Selecting a channel profile is an explicit assumption about modem capability. The Annex 2 2400 bit/s 8-PSK example is a performance surface for one particular modem; its approximate 10 ms and 4 Hz edges are an illustration, not a generic hardware specification, a guaranteed joint limit or a measured 50% success boundary. Data rate, FEC and interleaving matter. Removed STANAG/MIL-STD names require an explicit replacement and receive no silent example substitution. Custom T0/F0 require both finite nonnegative values at −10 dB; no 2 ms/1 Hz defaults are invented. Entered sources are carried into the computed result, input fingerprint and PDF.
Unknown fading conditions remain unassessed while conditional planning thresholds can still be listed. A nonzero-spread prediction does not match the single non-fading AWGN tap. The rectangular delay/Doppler screen is an engineering envelope assumption, not a BER or conformance guarantee; representative channels do not establish a universal ordering of modem performance. MIL-STD/ALE test channels and authored FT8/FT4 thresholds retain their own source values and conditions.
F.520 is withdrawn. F.1487's former Poor-channel note is ambiguous: low-latitude moderate gives 2 ms/1.5 Hz, while mid-latitude disturbed gives 2 ms/1 Hz and points back to that note. Both native rows are preserved without an unconditional historical equivalence claim. A separate non-fading ground-wave tap may be required for the NVIS test profile; the proxy does not synthesize that mixture.
No Watterson bitstream simulator or Annex 2/3 modem test protocol is implemented. In particular, performance surfaces, test duration and measured BER are not generated. The source validates the Gaussian model for 3 kHz and describes possible use up to about 12 kHz; it does not establish a general wider-band model. Noise-bandwidth normalization is distinct from actual waveform bandwidth. Current P.533 digital DCR remains outside the historical overlay, and no complete F.1487 conformity is claimed.
4. Compliance with the ITU reference implementation
The engine is measured against ITURHFProp, the reference implementation published by ITU-R Study Group 3 (source). This finite comparison helps detect implementation errors. The current Recommendation is authoritative where reference software differs; passing these vectors does not establish conformity to every clause. The independent clause checks and documented departures remain part of the assessment.
30 international paths across 29 frequencies (1 MHz grid, 2 to 30 MHz), isotropic antennas at both ends so the antenna model is removed from the comparison entirely, 1 kW, 3 kHz, rural noise, required SNR 10 dB. Measured on the 624 vectors where the Recommendation predicts a real signal, with engine 2.37.0 on 1 November 2026, each at the instant the reference evaluates it. The noise is P.372's at the receiving site, as §7 of P.533-14 and the reference take it, with galactic noise only above foF2 there (P.372-18 §4.1), where the reference adds it at every frequency.
| Quantity | Mean abs error | Median | P90 | Bias |
|---|---|---|---|---|
| Basic MUF | 0.93 MHz | 0.47 MHz | 2.83 MHz | +0.74 MHz |
| Field strength | 2.83 dB | 1.48 dB | 7.45 dB | +0.66 dB |
| Noise (P.372) | 0.36 dB | 0.25 dB | 0.77 dB | +0.11 dB |
| SNR | 2.85 dB | 1.51 dB | 7.55 dB | +0.62 dB |
| Basic circuit reliability | 8.4 pts | 4.4 pts | 22.0 pts | +1.9 pts |
Read the bias column. Scatter is expected on a model with a cliff in it: close to the MUF, a fraction of a megahertz of disagreement swings field strength by tens of decibels. A systematic offset is the thing that reveals a missing, doubled or misapplied term. Part of the offset above follows documented differences from ITURHFProp, with their scope and retained departures stated above. Since engine 2.19.0: the above-the-MUF loss of equations 25 and 26, F2(dmax)MUF on paths longer than dmax, and the lowest-order mode with no minimum elevation, each of which reads higher than the reference (with those three set to ITURHFProp's forms, the 2.19.0 engine's basic MUF sat a median of 0.06 MHz from the reference and its field strength bias was −0.48 dB). Since engine 2.20.0: the §5.1 reflection height (equations 14 to 16) for each F2 mode's field, where the reference uses equation 2, which moves single paths either way. Since engine 2.21.0: Lz of 11.61 dB and Ly of +0.38 dB, re-determined for this engine against the D1 databank as the Recommendation requires (section 3.14), where the reference uses 9.14 dB and −0.17 dB; and the ionospheric maps and the sun are taken at the same instant. ITURHFProp evaluates its maps for an input hour at that hour but its short-path sun (absorption and foE) an hour earlier, from its 0-based hour index; taking both at the same instant agrees better with the CCIR D1 measurements (rms 9.69 dB against 10.18 dB the reference's way). With the reference's sun reproduced, field strength reads 3.08, 1.67, 7.77 and +0.11 dB in the columns above, SNR a bias of +0.07 dB and reliability a mean error of 9.4 points (+0.0). Since engine 2.24.0 the selected-arc winter midpoint, UTC dates, monthly equation of time and Table 3 azimuth interpolation are corrected. Since engine 2.25.0 the P.1239 inputs in section 3.16 are corrected and the current calibration is Lz = 11.39 dB, Ly = −0.39 dB. Since engine 2.23.0: galactic noise is left out below foF2 over the receiver, as P.372-18 §4.1 says, where the reference adds it at every frequency; with rural man-made noise dominating here, it lowers the noise bias by 0.01 dB. And the OWF (FOT) takes the factor of P.1240-2 §7 by the layer that gives the path basic MUF (0.95 where an E mode does), where the reference applies each layer's decile to that layer's operational MUF and takes the larger. The FOT is not in the table above. Engine 2.23.0 also combines the noise deciles as P.842-5 Steps 6 and 9 do, by power, as the reference does: the reliability bias came down from +6.0 to +3.9 points. On paths of 7000 km and more, the local noon of §5.3.1 is the nearest hour, where the reference truncates to the hour at or before it. On the open or closed verdict itself, all 30 paths now agree with the reference: the LUF is searched on the field P.533-14 prescribes for the path length, and so is the table. Reliability is compared only where the reference figure sits between 10 and 90 per cent, because vectors on which both models say zero agree trivially and would flatter the average.
Of the 246 vectors on bands the Recommendation calls closed, the number on which this engine claims the circuit meets its required grade of service is zero, and that is gated in the test suite. Recommending a frequency that does not work is the worst thing a planning tool can do.
4.1 Secondary check against VOACAPL
VOACAPL is not P.533. It is the Lucas-Haydon and ICEPAC lineage, a different model with a different reliability definition. Divergence from it is expected and is not, on its own, evidence of a defect. It is kept only as a sanity check, and the figures below should be read in that light rather than as an error budget.
| Quantity | Mean abs error | Median | P90 | Max |
|---|---|---|---|---|
| MUF50 | 2.1 MHz | 1.8 MHz | 4.5 MHz | 5.4 MHz |
| FOT | 3.1 MHz | 1.9 MHz | 8.7 MHz | 9.9 MHz |
| Reliability (peak-across-freqs) | 18 pts | 18 pts | 41 pts | 50 pts |
4.2 What is not validated here
- LUF has a standard definition and a planner extension. P.533-14 section 9 uses the P.373 definition and the monthly median SNR criterion, reported to the nearest 0.1 MHz. Here the criterion is the required P.842 basic circuit reliability at the required SNR; at the default 50% reliability, this is the median SNR criterion. The search scans at 0.25 MHz, refines the first crossing to a 0.1 MHz interval, and rounds upward to 0.1 MHz. A median-day fallback is flagged when the requested grade cannot be met. ITURHFProp's report has no LUF output, so this benchmark does not directly validate LUF. Both thresholds are inputs you set.
- Sporadic-E, the solar-flare absorption (NOAA D-RAP) and the other additions in section 3 are documented extensions. They are not part of P.533 and are not claimed to be.
4.3 Secondary comparison against PyIRI / IRI-2020
PyIRI uses a modern magnetic epoch; P.1239 requires the epoch-1960 coordinate used to fit its maps. The following comparison uses 384 held PyIRI vectors and the production Fourier fallback with no grids loaded. Differences combine magnetic epoch, model and numerical effects; they are not measurement errors or a P.1239 conformity test. The F1 row includes only the 195 vectors where PyIRI reports F1 present.
| Quantity | Mean abs difference | P95 | Max |
|---|---|---|---|
| Production foF2 | 0.325 MHz | 1.86 MHz | 5.61 MHz |
| foE | 0.155 MHz | 0.31 MHz | 0.38 MHz |
| foF1 | 2.196 MHz | 4.93 MHz | 5.24 MHz |
| M(3000)F2 | 0.028 | 0.16 | 0.38 |
| Historical MODIP | 2.97° | 15.94° | 16.04° |
| foF2 kernel with matching modern MODIP | 0.008 MHz | 0.03 MHz | 0.17 MHz |
| Modern MODIP comparison | 0.066° | 0.29° | 0.40° |
The two modern-coordinate rows check comparable inputs and retain the existing PyIRI test limits. They do not validate the production historical coordinate. That coordinate, the field and map equations, foE/F1 rules and Es maps are checked separately against independent calculations from the current Recommendation and fresh official coefficients. The former F1 truncation below foE has been removed.
4.4 Accuracy against measured signals
Predicted minus measured median field strength, every measured hour, at 1 kW EIRP with isotropic antennas, for the two ITU-R Study Group 3 databanks. Lz and Ly are fitted to D1 (section 3.14). D2, measured in Germany from 1969 to 1993 under P.845, is an additional check and not a direct fitting target. Matching named circuits, frequencies and observation periods occur in the collections; strict observational independence is not established.
| Databank | Hours | Mean | Sample standard deviation | RMS |
|---|---|---|---|---|
| D1 (CCIR, worldwide) | 16 268 | +0.01 dB | 9.62 dB | 9.61 dB |
| D2 (Germany) | 103 792 | +1.87 dB | 11.22 dB | 11.37 dB |
The table uses the shipped CCIR grids in the final review. Standard deviation uses n−1 (P.1148-1 equation 2); it is not RMS or an individual-day error bound. Source records are monthly medians normalized to 1 kW EIRP. P.845 measurement/instrument conformity is not independently certified from those monthly records. The P.1148 subset analysis uses a qualified centred-dipole midpoint proxy, without conditional-day conversion or a new minimum-elevation constraint. No complete-procedure conformity is claimed.
Earlier coefficient-fallback audits, retained as historical comparisons: on the long way round, the D1 circuits from Australia and Japan read an rms of 12.8 dB, and ITURHFProp 12.3 dB on the same 666 hours. These empirical errors remain substantial. After the P.1240 corrections, D1 RMS changes from 9.62168 to 9.62169 dB and D2 RMS from 11.37450 to 11.37455 dB; D2 deviations over 50 dB remain 196. D1 supports retaining the existing Lz and Ly to the stated 0.01 dB precision. The existing test limits are retained, and D2 is not used to fit the calibration terms.
5. Known outliers
These are the paths that sit furthest from VOACAPL in the secondary check (§4.1), as the benchmark run of 5 October 2026 on engine 2.20.0 measured them. Across all of its paths the engine's MUF50 sits 1.3 MHz below VOACAP's on average and its reliability 10.6 points above. Against the ITU reference implementation (§4) the basic MUF bias is +0.74 MHz and the reliability bias +1.9 points, so a gap to VOACAP is not, on its own, evidence of a defect.
- Hawaii → Auckland (MUF −4.5 MHz) and Singapore → Nairobi (MUF −5.4 MHz, REL +19 pts): trans-equatorial paths. The CCIR foF2 maps do not resolve the equatorial ionisation anomaly crests, and the two models handle that region differently.
- Hanoi → Manila (MUF +4.5 MHz): tropical daytime over the EIA crest region, the same cause as Singapore → Nairobi. The engine has no EIA model beyond the CCIR maps (the damping of §3.1 was removed).
- Reliability: Moscow → Vladivostok (+50 pts), London → Sydney at solar minimum (+42 pts), London → Sydney (+41 pts) and at solar maximum (+36 pts). VOACAP bakes a man-made noise margin into its REL figure; this engine takes the required SNR as an explicit input. The two are not the same quantity, and basic circuit reliability is compared against ITURHFProp in §4, where its mean error is 8.4 points.
- Without a documented cause: Reykjavik → Wellington (REL −30 pts), Cape Town → Perth (REL −27 pts) and London → Sydney at solar maximum (MUF −4.3 MHz) are among the larger gaps and are listed here rather than explained.
6. When to trust this tool (and when not to)
An honest three-tier answer for an RF engineer evaluating whether to use this in a workflow:
| Use case | Suitable? | Why |
|---|---|---|
| Pre-planning and band selection (which bands to try, at what hours, which season) | Yes, confidently | Basic MUF sits within a median of 0.47 MHz of the ITU reference implementation, well inside any allocation width. The frequency window recommendation is the strongest thing this tool does. |
| Engineering design with documented uncertainty (antenna sizing, power budget, availability targets) | Yes, with the envelope shown | Use MUF90 and MUF10 as design bounds rather than MUF50 as a point estimate. Quote reliability with the required SNR it was computed against, and with its envelope: "80 per cent at a required SNR of 10 dB in 3 kHz, plus or minus 10 points". |
| Regulatory and tender submissions (spectrum licensing, ACMA coordination, client deliverables) | Yes, with the PDF report attached | The engine implements ITU-R P.533-14 and is measured against the ITU's own implementation of it, with the error and bias of every quantity published in section 4: basic MUF within a median of 0.47 MHz (bias +0.74 MHz), field strength bias +0.66 dB, part of it by design where this engine follows the printed Recommendation. The scope is traceable: the implemented method, finite reference comparison, documented departures, published differences, and a run hash and engine version stamped on every result. The PDF report's Validation and quality assurance section states the bounds. Two caveats belong in the submission: reliability carries roughly 8 points of scatter, and the non-ITU extensions listed in section 4.2 are extensions, not the Recommendation. |
| Safety-of-life circuits (aviation HF, maritime distress) | No | No monthly median HF prediction, this one or any other, should be the sole basis for a safety-of-life circuit. Day to day conditions routinely depart from the median by more than ten decibels. Design with real margin and a fallback band. |
Questions or corrections? Contact us. If you find an outlier with a physical explanation we haven't documented, we'll credit the fix.