A phone shows a blue dot on a map within a couple of seconds, and almost everyone assumes the satellites replied to something it sent. Understanding how does GPS work starts with discarding that idea, because nothing ever leaves your device in the direction of space.

GPS satellites broadcast, continuously, to nobody in particular. A receiver listens to several of them at once, measures how long each signal took to arrive, and solves for its own position from those travel times. The whole computation happens inside the device.

The system has no idea you exist, and it cannot, because the conversation only runs one way. That single fact resolves most of the confusion around satellite positioning, including the question of who is watching.

The satellites only ever talk

Each satellite transmits a repeating message containing two things that matter: where the satellite is, and the exact moment the message left it.

That is the entire contribution from orbit. As the technical record puts it, the system does not require the user to transmit any data, and operates independently of any telephone or internet connection.

One consequence is capacity. Because receivers only listen, the number of them makes no difference at all — a hundred users and a hundred million users impose exactly the same load on the constellation, which is zero, and no amount of demand can congest it.

The other consequence is the one people get wrong. A GPS receiver in isolation is invisible. Nobody in the programme, and nobody operating a satellite, learns anything about where any device is.

Position is a timing problem

The receiver is not measuring distance directly. It is measuring time, and converting. Radio signals travel at the speed of light, so a signal that took 0.07 seconds to arrive came from roughly 21,000 kilometres away.

Knowing the distance to one satellite places you somewhere on a sphere around it. A second narrows that to a circle, a third to a pair of points, and one of those points is usually somewhere absurd like deep space or far underground.

This is trilateration, and it is a solved piece of geometry. The difficulty is not the shapes. It is the clock.

Time-of-flight measurement only works if the receiver knows precisely when the signal was sent and precisely when it arrived. The satellites handle the first half with atomic clocks. The second half is the problem.

Why three satellites are not enough

A phone does not contain an atomic clock. It contains a cheap quartz oscillator, of the kind in a wristwatch, and at the speed of light a small timing error becomes an enormous distance error.

One millisecond of clock error corresponds to about 300 kilometres of position error. A receiver accurate to the millisecond would be useless.

The fix is elegant, and it is the single cleverest thing in the whole design. Rather than trying to build an accurate clock into every device on the ground, the system treats the receiver’s clock error as one more unknown and solves for it alongside the position, which means the hardware in your pocket never has to be good at keeping time.

That is what the fourth satellite buys. With four in view the receiver has four equations and four unknowns — three coordinates and its own clock offset — and the solution corrects the cheap oscillator as a side effect of finding the position.

It is why your phone keeps better time than its hardware deserves. Three satellites will do only when one coordinate is already known, such as a receiver that can assume it is at sea level.

What is actually up there

The constellation was designed around a minimum of twenty-four satellites, arranged in orbital planes so that several are above the horizon from anywhere on Earth at any moment.

In practice more are flown than the minimum requires, with the constellation running in the low thirties, so that failures and maintenance do not open gaps in coverage.

They sit in medium Earth orbit, far above the low-orbit satellites used for imaging and broadband, and each one circles the planet twice a day. Orbits drift and clocks wander, so a network of ground monitoring stations tracks every satellite continuously and uploads corrections to the message each one broadcasts.

The programme is run by the United States, and its official government site presents it as a utility provided free to the world. Other constellations now do the same job — Galileo, GLONASS and BeiDou — and most modern receivers use several at once.

Relativity is not a footnote here

This is the part that sounds like trivia and is not. Clocks in orbit do not tick at the same rate as clocks on the ground. Special relativity slows them because the satellites are moving fast; general relativity speeds them up because they sit higher in Earth’s gravity well. The second effect is larger.

The net result is that satellite clocks, seen from the ground, run about thirty-eight microseconds fast per day.

Thirty-eight microseconds sounds negligible until it is multiplied by the speed of light. Left uncorrected it would introduce roughly ten kilometres of position error every day, accumulating.

The correction is designed in rather than patched on. GPS is one of the few pieces of consumer infrastructure that would visibly fail if Einstein had been wrong.

The precision has a second use that dwarfs navigation in importance. Power grids, mobile networks, data centres and financial exchanges all need a common time reference accurate to fractions of a microsecond, and most of them take it from GPS — which means an outage would stop far more than the map on a dashboard.

Why your phone finds you so quickly

A receiver starting from nothing has to download orbital data from the satellites themselves. The broadcast carries this at a very low bit rate, so a cold start takes tens of seconds at best and considerably longer if the sky is partly blocked.

Phones cheat, legitimately. Assisted GPS fetches the same orbital data over the mobile network in an instant, and uses nearby cell towers and Wi-Fi networks to guess a rough position before any satellite has been heard from.

That guess narrows the search enormously. It is why a phone locks on in seconds while a standalone receiver switched on in a field can sit there for minutes, and why the same phone with no signal behaves much more like the receiver in the field.

It is also where the tracking actually happens. The satellites still learn nothing, but the phone has now used the network — and once a device has computed its position, whether that position stays on the handset or is sent to a carrier, an app or an advertiser is a matter of software and policy, not physics.

The distinction is worth holding onto. GPS does not track anybody. Devices that know where they are, and have an internet connection, do.

The deliberate blurring that stopped in 2000

For most of the system’s early life the civilian signal was degraded on purpose. The military introduced errors under a policy called Selective Availability, so that public accuracy was far worse than the technology allowed.

It was switched off on 1 May 2000 by presidential order, and the change was immediate rather than phased. Accuracy improved roughly tenfold overnight, to something around five metres, and the modern consumer market — turn-by-turn navigation, ride hailing, fitness tracking — followed from that single administrative decision rather than from any new satellite or any new receiver.

Newer civilian signals have pushed this further, and receivers using the L5 band can reach the tens of centimetres, which is the difference between knowing the road and knowing the lane.

What still degrades it

The failure modes are all about the signal path rather than the maths.

  • Urban canyons: tall buildings block satellites and leave too few in view for a clean solution
  • Multipath: signals bounce off glass and concrete and arrive late, so the receiver measures a longer path than the true one
  • Atmosphere: the ionosphere slows signals unpredictably, which is the largest natural error source
  • Interference: the signals arriving at ground level are extraordinarily weak, and easily overwhelmed

That last point deserves emphasis. A GPS signal reaching the ground is weaker than the background noise around it, and is recovered only because the receiver already knows the exact pattern to look for and can pull it out by correlation.

Being that faint makes the system easy to jam with very little power, and easy to spoof by broadcasting a plausible false signal — a vulnerability that has moved from theoretical to routine in contested airspace, and one that mirrors the fragility of the physical infrastructure described in what happens when an undersea cable breaks.

What to take from this

Three things explain most of what people find confusing about satellite positioning.

The receiver is passive, so accuracy problems are never a weak uplink or a crowded network. The fourth satellite exists to fix your device’s clock rather than to sharpen the geometry. And the position on screen usually blends satellite data with cell and Wi-Fi information, which is why it can be good indoors and poor between tall buildings.

The continuous drain of listening for weak signals is also one of the more expensive things a handset does, which is part of the wider story in why phone batteries wear out.

For the official description of the constellation and how the service is operated, the FAA maintains its own explanation aimed at aviation users, where the accuracy requirements are considerably stricter than anything a phone needs.