HF skywave emitter monitoring from low Earth orbit
High-frequency transmissions bounce off the ionosphere, confounding ground-based direction-finding. Spaceborne receivers intercept the direct upward wave before refraction, enabling geolocation that ground networks cannot reliably achieve alone.
Sensors
- HawkEye 360 Cluster satellites: Fly in formation triads separated by tens of kilometres; use time-difference-of-arrival and frequency-difference-of-arrival across the cluster to geolocate emitters. Published geolocation accuracy is stated by the company as approximately 1 km CEP for cooperative signals at mid-band; HF performance is constrained by the long wavelengths involved and is not separately published. Revisit at a given point is irregular, typically multiple passes per day at mid-latitudes.
- Spire Global SDR payload (LEMUR-2 satellites): Software-defined radio receivers covering a wide frequency range including HF bands; primarily deployed for GNSS-RO and AIS/ADS-B, but the SDR architecture allows HF collection under specific mission configurations. Constellation of over 100 satellites gives high revisit, though tasking for HF is not a standard commercial product; Spire publishes constellation revisit as sub-90 minutes globally.
- TechDemoSat-1 SDR (historical, SSTL/UK, 2014–2018): Carried a software-defined radio experiment that demonstrated spaceborne HF signal reception and direction-finding feasibility. No longer operational, but the published results from its HF campaign remain the most openly documented LEO HF collection dataset in the academic literature, establishing that direct-path HF signals are detectable from 600 km altitude.
- NOAA/COSMIC-2 FORMOSAT-7 constellation: Primarily a GNSS radio-occultation mission, but its ionospheric electron-density profiles (published in near-real-time via UCAR) are essential ancillary data for correcting HF propagation geometry when interpreting any spaceborne HF intercept. Not an RF-intercept sensor, but operationally inseparable from HF geolocation analysis.
Why ground-based direction-finding struggles with HF
A high-frequency transmission between 3 and 30 MHz does not travel in a straight line at ground level. The ionosphere, specifically the F-layer sitting roughly 200 to 400 km above the surface, refracts the signal back to earth at distances of 1,000 to 4,000 km from the transmitter depending on frequency and solar conditions. A ground-based direction-finding network hears the signal arriving from the apparent direction of the refraction point, not the transmitter. Triangulating across multiple ground stations gives you a family of possible source locations, not a single fix. The ambiguity can span hundreds of kilometres.
The problem compounds when the emitter is deliberately operating near the skip zone, the annular region around the transmitter where the skywave has not yet returned to earth. Receivers inside the skip zone hear nothing; receivers outside it hear a bearing that points to the wrong place. Sophisticated operators know this and choose frequencies accordingly. Ground-based HF direction-finding is mature technology with a long history, but it is structurally disadvantaged against an emitter that understands propagation.
What a satellite above the ionosphere actually receives
Any HF transmitter radiating with significant vertical or near-vertical angle sends energy upward through the ionosphere as well as outward along the ground. A LEO satellite at 500 to 600 km altitude sits above the F-layer and intercepts this direct upward-propagating component before it is refracted. The geometry is far cleaner: the signal path from transmitter to satellite is essentially a straight line through the ionosphere, with predictable and correctable delay rather than a specular reflection off a curved, dynamic surface.
The practical consequence is that time-difference-of-arrival measurements across a formation of satellites, or successive Doppler measurements from a single satellite passing overhead, can be referenced to a known orbital geometry. The residual uncertainty comes not from the reflection ambiguity that plagues ground networks, but from the accuracy of the ionospheric total-electron-content model used to correct the propagation delay. That is a solvable problem, not a structural one.
Ionospheric modelling is not optional
The ionosphere is not a fixed mirror. Electron density varies with the solar cycle, time of day, season, geomagnetic latitude, and sudden solar events. At HF frequencies, the refractive index of the ionosphere is non-trivial: a signal at 10 MHz experiences measurable group delay and ray-bending even on the upward path to a LEO satellite. Ignoring this introduces geolocation errors that can reach tens of kilometres in disturbed ionospheric conditions.
Operational HF geolocation from orbit therefore requires real-time or near-real-time ionospheric electron-density profiles. The COSMIC-2 constellation provides GNSS radio-occultation profiles with latencies of a few hours via UCAR's CDAAC archive. The International Reference Ionosphere model (IRI), maintained by COSPAR and URSI, provides a climatological baseline. During geomagnetic storms, neither is fully adequate, and this is an honest limit of the method: geolocation accuracy degrades significantly during Kp index events above roughly 5. Analysts should flag storm periods explicitly rather than report nominal accuracy figures.
Over-the-horizon radar: the emitter worth watching
The most consequential HF emitters in a defence-intelligence context are over-the-horizon radars (OTHR). Systems such as the Australian Jindalee Operational Radar Network (JORN), the US AN/FPS-118, and various Russian and Chinese OTHR installations transmit at high power across broad HF bands specifically to exploit skywave propagation for long-range surveillance. Their transmissions are loud, spectrally distinctive, and operationally significant: when an OTHR is radiating, someone is watching a particular ocean or airspace sector.
From LEO, an OTHR transmission is detectable by its power level, its characteristic waveform structure (typically a frequency-modulated interrupted continuous wave), and its spectral occupancy pattern. Monitoring whether a known OTHR site is active, on which frequency bands, and at what power level gives a persistent indicator of surveillance intent that is difficult to conceal. The transmitter cannot turn off physics: a multi-megawatt HF radar leaves a clear signature above the ionosphere even when the operator would prefer it did not.
Limits that buyers should understand before commissioning
Spaceborne HF geolocation is not a replacement for a mature ground-based HF direction-finding network; it is a complement. A single satellite pass over a target area lasts roughly 5 to 10 minutes. If the emitter is not transmitting during that window, there is nothing to intercept. Formation flying constellations such as HawkEye 360 improve revisit, but their published geolocation products are optimised for VHF and above; HF performance figures are not independently verified in open literature.
Antenna design is a genuine constraint. Efficient reception of signals at 3 to 10 MHz requires antenna dimensions on the order of metres to tens of metres, which is difficult to accommodate on a small satellite. TechDemoSat-1 used a deployable monopole of modest length and accepted sensitivity penalties. Commercial operators working in HF from small satellites are trading antenna gain for platform size, and the sensitivity floor is higher than for VHF or UHF collection. Buyers should ask for demonstrated detection thresholds against realistic emitter power levels, not theoretical figures.
Satellize integrates spaceborne RF data with open ionospheric products and published propagation models to produce geolocation assessments. The Overhead analysis column has covered OTHR activity patterns as a recurring topic.
From raw intercept to actionable geolocation product
The analytic chain from a raw HF intercept to a reported emitter location involves several distinct steps. Signal detection and classification come first: identifying that a received waveform is an HF transmission rather than interference, and characterising its modulation, bandwidth, and centre frequency. Waveform fingerprinting against known emitter libraries can assign a probable emitter type, which constrains the expected power level and antenna pattern used in the subsequent geolocation calculation.
Geolocation then applies TDOA or Doppler-based methods corrected with the best available ionospheric model for the intercept time and location. The output is a probability ellipse on the ground, not a point. Honest reporting states the semi-major and semi-minor axes of that ellipse and the ionospheric model version used. Change detection across multiple passes, tracking whether an emitter is active, shifting frequency, or relocating, is often more operationally useful than a single-pass fix.
Typical figures
| Frequency coverage | 3 to 30 MHz (HF band); practical spaceborne collection most documented between 5 and 25 MHz |
| Geolocation accuracy (nominal, calm ionosphere) | 1 to 10 km CEP depending on formation geometry, signal duration, and ionospheric correction quality; degrades to 20+ km during geomagnetic storms (Kp > 5) |
| Single-satellite pass duration over target | Approximately 5 to 10 minutes at 500 to 600 km altitude |
| Revisit (formation constellation, e.g. HawkEye 360) | Multiple passes per day at mid-latitudes; irregular cadence, not geostationary persistence |
| Minimum detectable emitter power | Not independently published for HF from small satellites; TechDemoSat-1 demonstrated detection of broadcast-class emitters (tens of kW ERP); low-power tactical HF (< 100 W) remains unverified in open literature |
| Ionospheric correction latency | COSMIC-2 GNSS-RO profiles available via CDAAC with 2 to 6 hour latency; IRI model available in real time as climatological baseline |
| Archive depth | HawkEye 360 commercial archive from 2019; TechDemoSat-1 historical data 2014 to 2018 (academic access); Spire SDR HF collection not a standard archived product |
| Delivery formats | Geolocation probability ellipses as GeoJSON or KML; emitter activity timelines as CSV; waveform metadata as structured JSON; ionospheric correction logs as ancillary files |
Analytics Satellize can run
| HF emitter geolocation probability ellipse | TDOA across satellite formation or single-satellite Doppler shift analysis, corrected with IRI or COSMIC-2 electron-density profile | GeoJSON polygon with semi-axis dimensions and confidence level, per intercept event |
| OTHR activity state assessment | Spectral occupancy classification against known OTHR waveform signatures (FMICW modulation, characteristic chirp rates); comparison against published OTHR frequency allocations | Per-pass activity report flagging site, frequency band, and estimated radiated power class |
| Emitter frequency-agility pattern | Time-series analysis of centre frequency and bandwidth across successive passes; change-point detection against historical baseline | Frequency-use timeline chart and structured data feed, updated per new pass |
| Ionospheric correction quality flag | Cross-referencing intercept time with NOAA Space Weather Prediction Center Kp index and COSMIC-2 profile availability; flagging passes where correction uncertainty exceeds defined threshold | Quality metadata attached to every geolocation product; storm-period intercepts labelled with degraded-accuracy warning |
| Emitter waveform fingerprint classification | Modulation recognition against published waveform libraries for known HF emitter types (broadcast, OTHR, maritime, aeronautical); feature extraction from instantaneous frequency and amplitude traces | Classification label with confidence score appended to intercept record |
| Multi-pass emitter track | Association of successive geolocation ellipses using probabilistic data association; detection of positional drift indicating mobile or relocated emitter | Track GeoJSON with timestamped position estimates and displacement vector if movement detected |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.