RF emitter geolocation from space
Clusters of LEO satellites can fix the position of any radiating emitter using time and frequency differences of arrival, no cooperation required. Accuracy depends on constellation geometry, ephemeris quality, and signal duration.
Sensors
- HawkEye 360 cluster constellation: Tri-satellite clusters flying in formation separated by tens of kilometres; uses TDOA and FDOA across VHF, UHF and S-band (roughly 100 MHz to 6 GHz); published geolocation accuracy of 1–3 km CEP under good geometry, degrading with short signal duration or poor baseline
- Kleos Space Scouting Mission: Four-satellite clusters in 37-degree and sun-synchronous orbits; TDOA-based geolocation across HF through UHF; designed for maritime and border-surveillance tasking with revisit improving as cluster count grows toward planned 20-satellite architecture
- Unseenlabs BRO constellation: French commercial constellation specialising in radar emitter detection; passive RF sensing of maritime radar bands; each satellite operates independently rather than in formation, using Doppler-based methods to characterise emitter type and bearing, with geolocation precision improving when multiple passes are combined
- ORBCOMM / legacy VHF LEO assets: Older VHF-band LEO networks have demonstrated opportunistic TDOA geolocation of VHF emitters as a secondary product; accuracy is coarser than dedicated cluster missions but useful for corroborating fixes
Why a radio signal is harder to hide than a ship
A vessel can switch off its AIS transponder. A radar cannot switch off and remain a radar. Any emitter that radiates, whether a navigation radar, a jammer, a satellite uplink terminal, or a push-to-talk radio, produces a wavefront that propagates at the speed of light in all directions simultaneously. From space, that wavefront is an observable fact.
The physics exploited here is elementary: a signal reaches two spatially separated receivers at slightly different times. That time difference, measured in microseconds, maps to a hyperbolic surface on which the emitter must lie. A third receiver adds a second hyperbola. Where the surfaces intersect, you have a position fix. This is time-difference-of-arrival (TDOA), and it requires no cooperation from the emitter whatsoever.
Geometry is everything: why clusters fly in formation
A single satellite can detect a signal and measure its frequency. It cannot produce a TDOA fix alone. You need at least three receivers with known, well-separated positions to generate two independent hyperbolic surfaces and resolve a 2D ground position. Four receivers allow a third measurement that cross-checks the solution and reduces ambiguity.
This is why HawkEye 360 and Kleos fly satellites in clusters rather than in independent orbits. The baseline between cluster members, typically tens of kilometres, sets the geometric dilution of precision. Too small a baseline and the hyperbolas intersect at a shallow angle, producing a long, thin uncertainty ellipse. Too large and synchronisation between receivers becomes harder to maintain. The cluster geometry is a deliberate engineering trade-off.
Frequency-difference-of-arrival (FDOA) adds a second observable. Because each satellite moves at a slightly different velocity relative to the emitter, each sees a slightly different Doppler shift. That difference constrains the emitter's position and, usefully, its velocity. FDOA is particularly valuable for moving emitters such as vessels or aircraft, where TDOA alone would smear the fix across the emitter's track during the observation window.
Frequency coverage and what each band reveals
Current commercial RF-geolocation constellations cover broadly from HF (shortwave, roughly 3–30 MHz) through VHF, UHF, and into low microwave bands around S-band. HF propagates via ionospheric skip, so detecting it from LEO is possible but complicated by multipath; most published cluster work concentrates on VHF and above. Maritime navigation radars typically operate in X-band (9–10 GHz) and S-band (2–4 GHz). Unseenlabs targets these specifically, characterising emitter type from pulse repetition frequency and waveform signatures before attempting geolocation.
Jammers are a particular target of interest for defence customers. GPS jammers operate near 1.2–1.6 GHz. Communications jammers span a wide range. Their defining characteristic from a geolocation perspective is that they are often high-power and continuous, which makes them easier to detect but also means they may be deliberately mobile. Combining TDOA fixes across multiple passes can track a moving jammer's route over hours.
The honest accuracy budget
Published figures from HawkEye 360 cite geolocation accuracy in the range of 1–3 km CEP (circular error probable, the radius containing 50% of fixes) under favourable conditions. That figure deserves unpacking. It assumes the satellite ephemeris is known to better than roughly 10 metres, the receivers are synchronised to sub-microsecond precision, and the signal duration is long enough to accumulate a clean cross-correlation peak.
Ephemeris uncertainty is the dominant error source at current accuracy levels. A 10-metre position error in a satellite moving at 7.5 km/s translates directly into a timing error that degrades the TDOA solution. GPS-based onboard positioning and precise orbit determination from ground tracking reduce this, but residual errors remain. Short signals, below roughly one second, produce poor cross-correlation statistics. Signals that are frequency-hopping or spread-spectrum are harder to process and may not yield a fix at all. Multipath from terrain or sea clutter adds noise in coastal environments.
For most intelligence applications, 1–3 km CEP is operationally useful. It is not sufficient to designate a precise target for kinetic action, but it is sufficient to cue a vessel intercept, identify a region for further collection, or track a pattern of life over days.
Combining passes and fusing with other layers
A single cluster overpass lasts roughly two to four minutes. In that window, a stationary emitter may receive one fix with the accuracy described above. Revisit rates for current commercial constellations range from a few hours to once or twice per day at mid-latitudes, improving toward the poles where orbital tracks converge. As constellation sizes grow, revisit shortens.
The real analytical value emerges when RF fixes are fused with other data. An RF geolocation placing a vessel at a position inconsistent with its last AIS broadcast is a dark-ship indicator. An RF fix on a radar emitter correlated with a SAR image of a vessel at the same location and time provides a much stronger identification than either alone. Satellize builds these fusion workflows for clients who need to move from raw fixes to assessable intelligence products. The same approach underlies the kind of sovereign maritime-domain-awareness programmes we support for smaller island and coastal states, though the specifics of each engagement remain confidential.
What this technique cannot do
RF geolocation from LEO does not work against emitters that are off. That sounds obvious, but it has a real operational implication: a disciplined adversary who transmits only briefly and unpredictably is harder to fix than one with continuous emissions. Burst transmissions shorter than the cross-correlation integration time may not produce a usable TDOA peak.
Coverage is not global-continuous. Unlike SAR or optical sensors, which image a swath regardless of what is below, RF geolocation requires the emitter to be transmitting during the specific window when the cluster is overhead and the geometry is adequate. This creates collection gaps that a sophisticated operator can exploit. Finally, the technique identifies a location, not necessarily an identity. Emitter fingerprinting from waveform characteristics narrows the field, but positive identification usually requires corroborating intelligence from other sources.
Typical figures
| Geolocation accuracy (CEP 50%) | 1–3 km under good geometry and ephemeris; degrades to 5–10 km with short signals or poor baseline |
| Frequency coverage | HF (3–30 MHz) through S-band (~4 GHz) for cluster TDOA; X-band (9–10 GHz) for dedicated radar-emitter sensors |
| Minimum signal duration for reliable fix | Approximately 1 second continuous emission; shorter bursts reduce cross-correlation quality |
| Revisit rate (mid-latitudes) | 2–6 passes per day per constellation at current fleet sizes; improves with additional clusters |
| Satellites required per fix | Minimum 3 (two TDOA surfaces); 4+ preferred for cross-check and FDOA |
| Ephemeris accuracy required | Better than ~10 m satellite position error to achieve sub-3 km CEP |
| Latency from collection to fix | Typically 30 minutes to a few hours depending on ground-station contact and processing pipeline |
| Archive depth | HawkEye 360 operational since 2019; Kleos since 2020; historical tasking data available on request |
| Deliverable formats | Geolocated point with uncertainty ellipse (GeoJSON, KML, shapefile), emitter characterisation report, time-stamped event feed |
Analytics Satellize can run
| Emitter fix with uncertainty ellipse | TDOA cross-correlation across cluster baselines, with FDOA applied where emitter or satellite velocity differential is sufficient | GeoJSON point feature with semi-major/minor axes of 1-sigma uncertainty ellipse, timestamp, frequency band |
| Dark-vessel RF indicator | Spatial and temporal fusion of RF fix with AIS track gaps; positional discrepancy flagged when RF-derived position exceeds AIS-projected position by more than fix uncertainty | Alert report with vessel candidate list, AIS gap duration, and RF fix confidence score |
| Jammer pattern-of-life track | Sequential TDOA fixes across multiple passes aggregated into a time-ordered position series; linear or kinematic interpolation between fixes | Track shapefile with timestamps and per-fix accuracy estimates; PDF summary for non-GIS users |
| Radar emitter classification | Pulse repetition frequency, pulse width and frequency agility analysis against published radar-type libraries; Unseenlabs-style waveform fingerprinting | Emitter type probability table (navigation radar, surveillance radar, jammer, communications) attached to each fix |
| Maritime domain awareness layer | RF geolocation fused with SAR vessel detection and optical imagery; cross-cued collection scheduling to maximise coincident coverage | Fused activity layer (GIS-ready) updated per constellation pass cycle; weekly analytical summary |
| Spectrum intrusion report | Detection of emissions in licensed or protected bands (e.g. GPS L1/L2, aeronautical VHF) with geolocation and power estimation from received signal strength | Incident report with fix, estimated power, affected band, and recommended follow-up collection |
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.