The problem MLAT solves
ADS-B is a broadcast: the aircraft works out where it is using satellite navigation and then announces that position twice a second, and a receiver on the ground simply writes it down. It is wonderfully simple, and it is why a £25 dongle can plot traffic across half a continent.
But a great many aircraft never broadcast a position at all. Older airliners, most business jets of a certain age, a large slice of general aviation and almost all military traffic carry a Mode S transponder that replies with an identity and an altitude and nothing more. On a tracker fed only by ADS-B, those aircraft are invisible — not because nobody heard them, but because what they said did not include a location.
Multilateration, universally shortened to MLAT, recovers the position anyway. It does not ask the aircraft where it is. It works out where the aircraft must have been for its signal to have arrived at several ground stations at exactly the times it did.
How the timing trick works
Radio travels at very close to the speed of light: about 300 metres every microsecond. If one receiver hears a transponder reply 40 microseconds before another does, the aircraft is roughly 12 kilometres closer to the first than to the second.
That single measurement does not give you a position. What it gives you is a surface — every point in the sky that is exactly 12 km further from one station than the other. In two dimensions that shape is a hyperbola, which is why the technique is sometimes called hyperbolic positioning.
Add a third receiver and you get a second hyperbola, and the two cross. Add a fourth and the solution is pinned in three dimensions. In practice MLAT servers use four or more stations for a full fix, or three plus the barometric altitude the aircraft is already reporting, which removes one unknown and lets three stations do the work of four.
MLAT is not new and it is not a hobbyist invention. Airports have used it for years as Wide Area Multilateration to cover terrain where radar cannot see, and Airport Surface Detection Equipment uses the same principle to track aircraft and vehicles on the ground. The community version works identically; it just uses volunteers' receivers instead of surveyed installations.
Why the clocks matter more than anything
Everything above depends on knowing, very precisely, when each station heard the signal. An error of one microsecond is an error of 300 metres. An error of ten microseconds puts the aircraft three kilometres from where it really is.
Ordinary computer clocks are nowhere near good enough. A PC synchronised by NTP over the internet is typically within a few milliseconds of true — thousands of times too coarse. So MLAT clients do not use the computer's clock at all. They use the sampling clock of the radio dongle itself, which ticks at a fixed rate, and timestamp each received message by counting samples.
Those local clocks are stable over short periods but they do not agree with one another, and they drift with temperature. The server solves that continuously by watching aircraft that are broadcasting ADS-B positions. Because it knows exactly where those aircraft are, it knows exactly how long their signals should have taken to reach each station, and it can therefore work out how far each station's clock is running fast or slow. Known aircraft calibrate the network; the calibrated network then locates the unknown ones.
This is why MLAT coverage collapses in quiet airspace at quiet times. If there are no ADS-B-equipped aircraft in view to synchronise against, the clock estimates go stale, and the network cannot confidently locate anything — including the Mode S aircraft it can hear perfectly well.
Geometry decides the accuracy
Four receivers are a minimum, not a guarantee. What matters is where they are relative to the aircraft, and the effect is dramatic.
| Receiver geometry | Result |
|---|---|
| Stations well spread around the aircraft | Good. The hyperbolic surfaces cross at steep angles and the intersection is tightly defined. |
| Stations clustered in one town | Poor. The surfaces cross at a shallow angle, so a tiny timing error smears the fix over kilometres. |
| Stations roughly in a straight line | Bad, and characteristically so — position error runs mostly perpendicular to that line. |
| Aircraft outside the receiver group entirely | Worst case. Extrapolating beyond the stations rather than interpolating between them; errors grow quickly with distance. |
Surveyors call this dilution of precision, and it is the same idea that makes satellite navigation less accurate when all the visible satellites are bunched in one part of the sky. It is also why MLAT tracks often look slightly ragged compared with ADS-B tracks: each fix is an independent solution to a geometry problem, not a reading taken from the aircraft's own navigation system.
What MLAT costs you in time
An ADS-B position is close to instantaneous — the aircraft broadcasts, you receive, you plot. An MLAT position cannot be computed until several stations have reported the same message to a central server, the server has matched them up, and it has solved for the position. That round trip typically adds a second or two.
On a map showing an airliner at 450 knots, a second and a half is about 200 metres. Not enough to matter for interest, but it is the reason MLAT targets sometimes appear to lag slightly behind where a fast aircraft visibly is.
Recognising MLAT traffic on a tracker
Most trackers distinguish the two, though not always obviously. The tells are:
- No altitude type or a barometric-only altitude. MLAT aircraft report the altitude their transponder sends; there is no GNSS-derived geometric altitude because the aircraft is not sending one.
- Position updates arriving less regularly. ADS-B position comes twice a second. An MLAT fix depends on enough stations hearing the same reply, so the rate varies with coverage.
- Coverage that stops abruptly. ADS-B range fades gradually at the horizon. MLAT stops the moment the aircraft leaves the area where four stations overlap, which can be a hard edge.
- Ground traffic that vanishes. A taxiing aircraft is shielded by buildings and terrain, so the number of stations hearing it drops below the minimum very easily.
MLAT is the one part of aircraft tracking where your receiver's value depends entirely on other people's. A single station cannot multilaterate anything at all — the technique needs four. Adding a receiver in an area that already has three is far more useful than adding a fifth in a city that has four, which is why aggregators are usually keenest on stations in gaps rather than in cities.
What it cannot do
MLAT needs a transponder reply to time. An aircraft with its transponder switched off is silent, and no amount of ground infrastructure will locate it. It also needs the aircraft to be high enough to be heard by several stations at once, which is why low-level traffic — helicopters at a few hundred feet, light aircraft in a valley — is patchy even in well-covered regions.
And it is a positioning technique, not an identification one. What MLAT adds is a location; the identity still comes from the 24-bit address in the same message, looked up against a separate database.