A transponder without GPS
Older aircraft, some military ones and aircraft flying through areas where GPS is jammed don’t send a position, or send nonsense. Their transponder still answers, though. Ground radars keep asking every aircraft for its identity and altitude, the transponder replies each time, and every reply is heard by all the stations around, not just the radar that asked.
That is enough to work out where the aircraft is. The method is called multilateration, MLAT for short, and air traffic control has used it for decades: at large airports it watches aircraft taxiing between buildings where radar sees poorly, and in mountain valleys it stands in for radars that cannot see past the ridges.
From a time difference to a position
A radio signal travels at the speed of light, about 300 metres in a millionth of a second. If a reply reaches station A three millionths of a second before station B, the aircraft is about 900 metres closer to A than to B.
That alone doesn’t say where it is. All the places with exactly that difference form a curve on the map. A second pair of stations gives a second curve, and the aircraft sits where they cross. Altitude doesn’t have to be computed: the aircraft reports it in its own reply. A fourth station is still the best way to spot a crossing that is wrong.
The hard part is the clocks. A station on a balcony has no atomic clock, and the clocks of two stations kilometres apart drift away from each other all the time. The answer flies overhead. Every aircraft that does report its own position is a moving reference: we know where it was and when each station heard it, so we know how far apart the stations’ clocks are at that moment. The quiet aircraft are then measured against the talkative ones.
Why the network matters more than the maths
A millionth of a second is 300 metres, so the stations’ clocks have to agree to within millionths of a second, otherwise the error in time turns straight into an error in position. The antenna positions have to be exact too.
Geometry matters just as much. When stations stand in a row, the curves cross at a shallow angle and a tiny error moves the crossing by kilometres. Stations that surround the aircraft give a sharp crossing. So where MLAT doesn’t work, the reason is usually simple: too few receivers nearby. On the morning of 1 October, 59 of the 158 stations taking part did not yet have a neighbour hearing the same aircraft.
The station network page therefore shows where today’s stations would be enough and gives two different kinds of advice: either more stations are missing, or there are enough and one is missing on the far side.
We thought about owners making a typo, too. A station whose antenna position doesn’t match where it really is would pull every crossing it takes part in. It is left out of the calculation, its owner gets an email if they have an account with us, and it stays connected, so we can tell the moment the position is fixed and let it back in.
A clean line through the jamming
The pilot started on 7 September and regular service on 14 September. On a flight track, computed positions are drawn as hollow beads, so you can always tell what the aircraft reported itself from what we calculated.
On 13 September an Airbus over Hungary reported a position that, within five minutes, jumped almost four hundred kilometres into Romania and back. Over the same minutes, more than ninety positions from our stations formed a continuous line at normal cruising speed. On the map, in a moment like that, the line wins. The aircraft even helps: it signals when it doesn’t trust its own position, and if our stations have a fresh fix at that moment, the map believes the stations.
Speed gets the same care: for a short while after a computed aircraft reappears, its speed is left blank. It is better than the alternative: a Cessna that cruises at about 150 km/h would otherwise come back from a short gap in the data showing the speed of a jet.
The first station in a region doesn’t have a broken clock. It just doesn’t have a neighbour yet.