Wideband HF Skywave Propagation: A Review with Link Modeling Corroborated by Propagation Data

I. INTRODUCTION

• HF communications regained interest as a low-cost, easily deployed complement to vulnerable modern infrastructure, with adaptive link technologies reducing the operator’s need to predict unstable propagation.HF skywave has also attracted commercial interest for limited-bandwidth transatlantic market data.Source: §I · pp. 1–2
• Conventional 3 kHz HF channels support voice and data rates up to 4.8 kbps, while scaling to hundreds of kilohertz or 1 MHz brings frequency-selective fading, dispersion, accumulated noise, and antenna challenges.HF radios have regained interest as a low-cost, deployable complement to modern infrastructure, including for disaster communications.Source: §I · p. 2

A. Ionosphere

• Solar ultraviolet radiation ionizes the D-, E-, and F-layers, whose varying electron densities refract HF waves and make usable frequencies depend on solar illumination and propagation conditions.The D- and E-layers dissipate at night, while the F-layer layers merge; D-layer attenuation can affect waves passing toward the F-layer, particularly at lower frequencies.Source: A. Ionosphere · pp. 2–3
• Ionosondes transmit vertical pulses and record reflected frequencies in ionograms; the highest returned frequency is a layer’s critical frequency, which rises with ion density.Ionograms approximate wave speed as the speed of light, while ARTIST is cited as software for more accurate ionosphere modeling.Source: A. Ionosphere · p. 3
• The maximum usable frequency depends on incidence angle and layer critical frequency, while lower takeoff angles support long-distance paths but increase travel through the attenuating D- and E-layers.Frequencies above the critical frequency can be supported when transmitted at oblique angles.Source: A. Ionosphere · p. 3
• The lowest usable frequency is set by reception becoming unusable as noise and ionospheric absorption reduce SNR; increasing transmit power or antenna directivity can lower it, but it may exceed the MUF.A path has no usable frequency when its LUF exceeds its MUF.Source: A. Ionosphere · p. 3

B. Weather

• Solar activity raises F-layer MUFs during solar maximum and lowers them during solar minimum, while space-weather disturbances can also disrupt HF propagation.The section identifies the 11-year solar cycle and space-weather effects as key sources of propagation variability.Source: B. Weather · p. 3
• The sun’s 27-day rotation adds propagation variability, and sunspot number and 10.7 cm solar flux serve as inputs for ionospheric and MUF prediction models.The 10.7 cm solar flux is described as a proxy for EUV radiation.Source: B. Weather · p. 3
• Terrestrial weather primarily drives the LUF through noise: lightning and precipitation static raise the noise floor, while atmospheric noise can travel thousands of kilometers.By contrast, space weather chiefly affects the MUF; terrestrial weather can make a channel unreadable without closing the ionosphere.Source: B. Weather · pp. 3–4

C. Noise

• The ITU-R P.372-17 model combines man-made, galactic, and atmospheric noise, but its man-made component omits impulsive noise.The model’s man-made noise estimates represent a short, vertical, lossless, grounded monopole; impulsive noise may briefly raise actual levels above model estimates.Source: C. Noise · p. 4
• Atmospheric-noise figures are read from ITU-R P.372-17 charts by season and four-hour interval, and the resulting noise power supports link-budget SNR calculations.The chart curves represent the percentage of time noise is expected to exceed a given figure; noise power depends on receiver bandwidth.Source: C. Noise · p. 5
• In the city-center summer example, man-made noise dominates the total noise figure, and moving to quiet rural conditions can reduce that component by nearly 20 dB.The example assumes daytime foF2 of 8 MHz; northern and southern link noise characteristics are largely similar, with galactic noise differing primarily with foF2.Source: C. Noise · pp. 5–6, Figure 2 · p. 6

D. Link Budget Analysis

• The simplified budget calculates received power from transmit power, transmit and receive antenna gains, additional system losses, and path loss.The analysis follows skywave path-loss or ground-wave basic-transmission-loss calculations.Source: D. Link Budget Analysis · p. 6
• The described budget excludes RF-component insertion losses, cable losses, and return loss due to system mismatches.The passage separately identifies cable, connector, mismatch, and polarization losses as possible additional system-loss terms.Source: D. Link Budget Analysis · p. 6
• Skywave path loss can come from HF propagation simulations such as PropLab or PHaRLAP, while the LFMF model supplies ground-wave basic transmission loss.Source: D. Link Budget Analysis · p. 6

• At the lower HF frequencies, long wavelengths make resonant antennas physically large, while their performance can vary substantially away from their design frequency.A half-wave dipole may approach 50 meters at the lower end of the band.Source: §III · p. 6
• Broadband antennas such as terminated folded dipoles, LPDAs, and traveling-wave structures offer wider operating bandwidth and more stable radiation patterns, with efficiency or size tradeoffs.Terminated folded dipoles can be less efficient, while LPDAs and traveling-wave antennas can require large physical dimensions.Source: §III · p. 7
• Passive linear time-invariant electrically small antennas are subject to a bandwidth–efficiency bound, while a non-LTI approach has demonstrated waveforms up to 1 MHz from an approximately 1-m monopole near 10.7 MHz.The reported waveforms include ASK and PSK variants, including BPSK, QPSK, and 8PSK.Source: §III · p. 7

A. Site Selection

• A suitable site is low-noise, flat, relatively unobstructed, and has suitable ground conductivity, infrastructure, and room for the intended antenna system.Site planning aims to maximize SNR while supporting practical installation and maintenance.Source: A. Site Selection · p. 7

B. Antenna Considerations

• Antenna choice depends on operating frequency, required bandwidth, link coverage, and desired radiation characteristics, with feed, matching, power-handling, and link-budget needs also considered.Directive antennas support long-distance reception, while omnidirectional antennas suit local or disaster-relief coverage; feed-line losses can further weaken low-SNR paths.Source: B. Antenna Considerations · p. 7

C. Wideband HF Antennas

• Wideband antennas such as LPDAs and TTFDs provide multi-octave coverage, but tradeoffs include LPDA size and TTFD efficiency.LPDAs can span the HF spectrum with low VSWR and high radiation efficiency; TTFDs use resistive termination to stabilize impedance and VSWR, dissipating some power as heat.Source: C. Wideband HF Antennas · p. 8

A. Narrowband Communication Modes

• Narrowband modes span about 60 Hz for PSK31 and 150 Hz for CW, while SSB and SSTV each remain below 3 kHz.CW concentrates energy for noise robustness; SSB filters the carrier and one sideband, and SSTV transmits static images.Source: A. Narrowband Communication Modes · pp. 8–9

B. Wideband Experiments

• Earlier wideband HF modeling lacked channel characterization and used a Watterson-model generalization that neglected dispersion, which its authors called “more conjectural than factual”.The cited empirical data had predominantly validated the narrowband Watterson model.Source: B. Wideband Experiments · p. 9
• Under optimal conditions, NRL concluded one-hop F-layer propagation could use bandwidths up to 1 MHz; Mitre equalization reduced dispersion from 95 µs/MHz to 1 µs/MHz, [67].Mitre's approximately 2,000 km experiments also found equalized channels time-invariant over 10 seconds.Source: B. Wideband Experiments · p. 9
• On an 840 km Norway path, Skaug found bandwidths of 80 kHz and below remained coherent, while performance dropped at 160 kHz.The drop was speculated to result from greater dispersion variation among the outermost spectrum components.Source: B. Wideband Experiments · pp. 9–10
• NVIS studies report challenges at wider bandwidths, while alternate waveforms established links at 200 kHz and non-contiguous 200 kHz operation improved channel availability by over 50%,,.The non-contiguous waveform divided bandwidth into 3 kHz sub-bands, allowing selection of a higher-SNR sub-band.Source: B. Wideband Experiments · p. 10
• Wideband surface-wave propagation is theoretically possible, but the author reports no publications on wideband HF ground-wave propagation beyond the standard 3 kHz bandwidth.Surface waves may experience slow fading over irregular terrain or nonhomogeneous surfaces, although guided waves have minimal dispersion at wide bandwidths.Source: B. Wideband Experiments · p. 10

V. TYPES OF LINKS

• HF propagation includes long-range skywave, ground wave, and NVIS; NVIS bridges the skip zone between ground-wave reach and the shortest long-range skywave paths.Ground waves can reach up to 500 km in favorable conditions, while long-range skywaves can exceed 4,000 km.Source: §V · p. 10

A. Long Range Skywave

• F-layer long-range skywaves typically support 10 to 50 MHz, with usable frequencies varying by time, season, solar cycle, and geography,.Higher MUFs are generally found near the equator; lower frequencies face greater D- and E-layer attenuation and radio noise.Source: A. Long Range Skywave · p. 10

B. Near Vertical Incidence Skywave

• NVIS typically uses 2 to 10 MHz and high takeoff angles, with near-vertical radiation; its usable frequencies vary by time, season, and latitude.NVIS is used for local communications across the skip zone, and wider bandwidths are difficult at its lower operating frequencies.Source: B. Near Vertical Incidence Skywave · p. 11

C. Ground Wave

• Ground waves combine direct, reflected, and surface components; the reflected wave can interfere constructively or destructively with the direct wave, while the surface wave diffracts beyond the horizon,,.Surface waves dominate when antennas are close to the ground and incur attenuation from induced surface currents.Source: C. Ground Wave · p. 11, Figure 4 · p. 12
• Ground-wave range generally increases as frequency decreases, and conductive surfaces such as seawater support longer distances; practical links typically span 200–500 km,,,.Vertically polarized antennas are required for optimal performance; range also depends on power, ground properties, and antenna heights.Source: C. Ground Wave · p. 11, Figure 5 · p. 12

D. Unique Links

• Trans-equatorial links span 5,000–8,000 km and refract twice from ionospheric slopes near the magnetic equator; they are strongest near equinoxes in late afternoon or evening under geomagnetically quiet conditions.The transmitter and receiver must be placed equidistantly from the magnetic equator.Source: D. Unique Links · p. 12, Figure 6 · p. 12
• F-layer long paths exceed 20,000 km and generally have lower SNR than shorter paths, but can be beneficial when the long route is sunlit over water or the short path has greater absorption.Gray-line links exploit the day-night transition; low-frequency links below 4 MHz can travel more than halfway around Earth.Source: D. Unique Links · pp. 12–13
• Pedersen Rays travel within the F2 layer and enable single-hop links up to 12,000 km, but require stable ionospheric conditions.They are high-angle paths that follow the Earth's contour within the ionosphere.Source: D. Unique Links · p. 13, Figure 7 · p. 13

VI. HF SKYWAVE PROPAGATION EXAMPLES

• The modeled examples use PHaRLAP for two NVIS paths and approximately 2,000 km, 3,000 km, and 5,000 km skywave links, considering time of day and location.Ground-wave modeling instead depends on path ground conditions, antenna heights, and polarization.Source: §VI · p. 13
• Most path pairs compare near-equatorial and higher-latitude locations; the 3,000 km pair instead holds path length nearly fixed to examine latitude differences.Near-equatorial paths have denser ionospheres from greater solar exposure; several modeled links are compared with measured propagation data later in the section.Source: §VI · p. 13

A. Interpreting the Ionogram

• An oblique ionogram plots refractive height against frequency, while its companion plot shows takeoff angle against frequency; lower frequencies require higher takeoff angles than higher frequencies.The transmit antenna's maximum radiation should point toward the required takeoff angle for highest link SNR.Source: A. Interpreting the Ionogram · p. 13, Figure 8 · p. 14
• Ionogram traces form three reflection-height bands whose sharp transitions identify the E-, F1-, and F2-layer boundaries, also visible in takeoff-angle plots.The modeled MUF is the highest frequency with at least three arrivals within 50 km of the destination for long-range links or 20 km for NVIS.Source: A. Interpreting the Ionogram · pp. 14–15

B. Representative NVIS Communication Links

• NVIS links use high takeoff angles; Frankfurt–Prague’s daytime one-hop returns reach its MUF, while multihop returns occur at lower frequencies and higher angles.One-hop F-layer paths experience less attenuation than multihop paths, which incur repeated lower-layer traversals and ground-reflection losses.Source: B. Representative NVIS Communication… · p. 15, Figure 9 · p. 16
• At night, Frankfurt–Prague’s NVIS MUF falls to 8.4 MHz and E-layer refractions are largely absent as the E-layer dissipates after sunset.Multihop propagation remains possible within the frequency range of the one-hop link.Source: B. Representative NVIS Communication… · p. 15
• The San Jose–Tegucigalpa NVIS MUF reaches 13.1 MHz by day and 9.9 MHz at night, above Frankfurt–Prague’s 9.5 MHz and 8.4 MHz.The near-equatorial advantage is attributed to greater ionospheric electron density; the daytime usable frequency is roughly 38% higher.Source: B. Representative NVIS Communication… · p. 15, Figure 9 · p. 16, Figure 10 · p. 16

C. Representative 2,000 km Long Range Skywave Links

• The northern 2,000 km link’s one-hop returns reach about 19.8 MHz by day at takeoff angles near 5°, motivating high gain at very low elevations.Real-earth antenna nulls, terrain, and buildings make such low-angle links difficult; an elevated LPDA is described as one possible configuration.Source: C. Representative 2,000 km Long Range… · p. 15
• At night near 5 MHz, the 2,000 km link’s one-, two-, and three-hop paths arrive at takeoff angles of about 10°, 20°, and 30°, respectively.A directive antenna can favor one hop count; the other paths’ distinct delays can cause multipath interference, though their lower power can reduce its impact.Source: C. Representative 2,000 km Long Range… · pp. 15–16
• For 2,000 km paths, San Jose–Port-au-Prince reaches MUFs of 24.4 MHz by day and 21.7 MHz at night, higher than the northern France–Helsinki path.The southern daytime MUF is about 4.6 MHz higher, attributed to greater equatorial electron density.Source: C. Representative 2,000 km Long Range… · pp. 15–17, Figure 11 · p. 17, Figure 12 · p. 17

D. Representative 3,000 km Long Range Skywave Links

• At 3,000 km, Orlando–Phoenix reaches MUFs of 25.9 MHz by day and 20.6 MHz at night, close to Moscow–Barcelona’s 25.0 MHz and 20.5 MHz.Despite a roughly 19° latitude difference, extended July daylight at the higher-latitude path offsets much of the lower-latitude electron-density advantage.Source: D. Representative 3,000 km Long Range… · pp. 17–18, Figure 13 · p. 18, Figure 14 · p. 18
• Both 3,000 km paths support low-angle one-hop F-layer propagation near the daytime MUF, but not single-hop E-layer propagation.The hop structure persists at night at reduced MUFs; a 3,000 km E-layer hop would require an unrealistically low, near-grazing takeoff angle.Source: D. Representative 3,000 km Long Range… · pp. 18–19

E. Representative 5,000 km Long Range Skywave Links

• On Newark–Glasgow, daytime propagation is predominantly multihop, and at night the lowest supported mode is two hops.At reduced nighttime electron density, a single 5,000 km hop cannot return to Earth.Source: E. Representative 5,000 km Long Range… · p. 19
• At 5,000 km, San Juan–Dakar reaches MUFs of about 26.0 MHz by day and 19.6 MHz at night, versus Newark–Glasgow’s 23.8 MHz and 16.9 MHz.The southern link spans nearly the entire HF band, while the higher-latitude transatlantic link is corroborated by FT8 operability ceiling alone because it has no single-station ionosonde reference.Source: E. Representative 5,000 km Long Range… · p. 19
• For the southern 5,000 km path, F-layer refraction is not supported below about 6 MHz even at seven hops, and E-layer paths generally require at least three hops.The stated exception is a narrow two-hop E-layer band from 22 to 24 MHz.Source: E. Representative 5,000 km Long Range… · p. 19
• Across the modeled distances, lower-latitude paths generally support higher day and night usable frequencies, while increasing distance lowers required takeoff angles and can make single-hop paths physically infeasible.Southern paths also experience greater D- and E-layer absorption.Source: E. Representative 5,000 km Long Range… · p. 19

F. Corroboration of the Simulated MUF Against Measured Propagation Data

• Five of eight links had observational coverage; three near-equatorial links remained model-only, while ionosonde MUFs scale measured foF2 by the link obliquity factor.Source: F. Corroboration of the Simulated MUF… · pp. 19–20
• At NVIS and 2,000 km, July FT8 ceilings reflect sporadic-E rather than F-layer MUF, so ionosonde measurements provide corroboration for those links.Source: F. Corroboration of the Simulated MUF… · p. 20
• For the three longest links, simulated MUFs track the measured FT8 ceiling through the diurnal cycle and sit at or above this lower bound.Source: Figure 17 · p. 21, F. Corroboration of the Simulated MUF… · p. 21
• Moscow–Barcelona values agree within a few tenths of a megahertz day and night; Orlando–Phoenix observations exceed the simulated daytime MUF (28.9, 29.8, and 25.9 MHz, respectively).The values are the FT8 noon peak, ionosonde MUF, and ray-traced daytime MUF, respectively.Source: F. Corroboration of the Simulated MUF… · p. 21
• Newark–Glasgow’s nighttime FT8 ceiling exceeds simulation because the sunlit control point sets path MUF; N. France–Helsinki’s ionosonde MUF is roughly 1.8 MHz higher near the 61°N trough.Source: F. Corroboration of the Simulated MUF… · p. 21
• Across five links at 31–56°N, simulated MUFs are corroborated by independent FT8 and ionosonde observables, and are conservative relative to measurements.Source: F. Corroboration of the Simulated MUF… · pp. 21–22, Table 2 · p. 22

G. Link Budget Analysis

• The link budgets pair PHaRLAP ray tracing with ITU-R P.372 noise modeling, assuming 1 kW isotropic transmission and including propagation losses and combined noise contributions.Man-made noise uses the worst-case urban category; feed-line, connector, and other RF-component insertion losses are excluded.Source: G. Link Budget Analysis · p. 22
• The plotted SNR and reflection-height results show the strongest supported arrival only where reference-band SNR reaches at least −10 dB, with plotted bands clipped using the reported MUF arrival criterion.The reference bandwidth is 3 kHz, and the robust edges use the same at-least-three-arrival criterion as the reported MUFs.Source: G. Link Budget Analysis · p. 22
• The San Juan–Dakar daytime usable band narrows near the top of HF, while nighttime propagation widens and shifts lower; Frankfurt–Prague shows improved low-band nighttime SNR.The plotted SNR bands cover 3 kHz to 1 MHz, with reflection-height curves for the strongest supported arrival.Source: G. Link Budget Analysis · p. 22, Figure 18 · p. 23
• Increasing bandwidth from 3 kHz to 1 MHz produces an approximately 25 dB SNR decrease across the links, primarily because noise power increases.The authors note that high-gain antennas can improve link quality.Source: G. Link Budget Analysis · pp. 22, 24

VII. DISCUSSION

• For five links with sufficient observations, simulated MUFs were conservative estimates against both crowdsourced FT8 operability ceilings and direct ionosonde measurements.The discussion attributes remaining deviations to known ionospheric structure.Source: §VII · p. 24
• Wideband HF remains constrained by narrow spectrum allocations and variable ionospheric conditions, while low-cost, easily deployed systems can serve as emergency fallback communications.The discussion suggests initially focusing experimental investigations on skywave links.Source: §VII · p. 24
• The discussion favors LPDAs for long-range skywave and TTFDs for NVIS, and identifies rural, obstruction-free sites near the equator as favorable for wideband communications.It notes that denser D- and E-layers at lower latitudes increase attenuation, which directive antennas can offset in the link budget.Source: §VII · p. 24

Appendix

• The northern France–Helsinki table exposes the per-frequency received power, reflection height, takeoff angle, and bandwidth-dependent SNR underlying the summary figure.The table covers both daytime and nighttime values.Source: Appendix · p. 24

Disclaimer

• The authors state that the paper’s views and conclusions should not be interpreted as necessarily representing official USG policies or endorsements.Source: Disclaimer · p. 25