The Off-Key Bard
Technology

Why Does Your Phone Sometimes Lose GPS Navigation in the Middle of a City?

Or: how a satellite thousands of miles away can find you on an empty highway but lose you between two buildings. Your phone calculates its position from signal timing, and tall buildings bend, block and delay those signals.

Or: How a Satellite Thousands of Miles Away Can Find You on an Empty Highway but Lose You Between Two Buildings

You're driving through an unfamiliar city, following directions on your phone. For twenty minutes everything has worked. The map shows your car on the right road, turns are announced at the right moments, and the arrival time looks reasonable.

Then you enter downtown. Tall buildings rise on both sides, and suddenly your phone seems confused. The blue location marker jumps to a parallel street, the app announces you're going the wrong way, and the route recalculates even though you haven't turned. A moment later, the marker snaps back to the correct road.

You haven't lost your internet connection, the receiver hasn't necessarily malfunctioned, and the satellites haven't stopped transmitting. The explanation is that GPS doesn't directly know your location. Your phone calculates its position by measuring signals from satellites, and those measurements can become misleading when buildings get in the way. In an open field the process can be remarkably accurate. Between skyscrapers, the same technology can put you tens of meters from reality.

GPS Doesn't Actually Track Your Phone

It's common to describe GPS as satellites tracking your movements. That isn't how ordinary navigation works. GPS satellites continuously broadcast radio signals carrying their positions and the time of transmission. Your phone receives those signals and calculates its own location. It doesn't have to send anything back, and the satellites generally don't know who is listening. GPS positioning is fundamentally reception and calculation. GPS tells the device where it may be, and the navigation app decides what to do with that.

Strictly speaking, GPS is the American system. Europe's Galileo, Russia's GLONASS and China's BeiDou are others, and together they're called Global Navigation Satellite Systems, or GNSS. Many phones use several constellations at once, which helps availability, including in cities, but more satellites don't guarantee a correct result if the signals themselves are distorted or reflected.

Distance From Time

Satellite positioning rests on an elegant principle. Radio signals travel at about the speed of light, so if you know when a signal left and when it arrived, you can compute how far it traveled: distance equals speed times time. NovAtel's introduction to GNSS describes exactly this: the receiver subtracts the transmit time from the arrival time and multiplies by the speed of light, and since satellites broadcast their orbits, it knows where each one was at the moment of transmission.

The intervals involved are tiny, which makes small errors expensive. A signal traveling for a millisecond covers about 300 kilometers. A one-microsecond error, one millionth of a second, corresponds to nearly 300 meters. Ten nanoseconds is roughly three meters. That's why satellites carry extremely accurate clocks and why so much engineering goes into timing corrections. Something that looks like a location problem is really a problem of measuring time.

Your phone doesn't carry an atomic clock, and that creates a wrinkle. NovAtel notes that receiver clocks are not nearly as accurate as satellite clocks, so the receiver has to solve for its own clock error along with its position. For a conventional standalone three-dimensional GNSS position fix, the receiver generally needs measurements from at least four satellites: three spatial coordinates and the receiver's clock offset must all be determined. Additional satellite measurements provide redundancy and can improve the solution, especially when some signals are blocked or unreliable. Phones can sometimes maintain a location estimate with fewer reliable satellite measurements by incorporating previous positions, motion sensors and other location sources, but that isn't the same as obtaining a fresh, independent satellite fix.

Geometry Matters Too

Satellite arrangement affects accuracy even when each measurement is equally good. Dilution of precision, or DOP, is described by NovAtel as a numerical representation of satellite geometry, with position error equal to DOP times range error. Satellites clustered together give poor geometry, and widely separated satellites give better geometry, much like estimating where someone stands from several different viewpoints rather than from almost the same direction. NovAtel lists urban environments among the places where DOP can get worse.

Now stand on a street with buildings twenty or thirty stories high on both sides. Much of the sky is blocked. Satellites you'd see from a park are hidden, and the ones that remain may sit in a narrow strip of sky. That's an urban canyon, and it hurts accuracy before the second problem even begins.

The Detour Problem

The second problem is that buildings reflect radio signals. Large structures with metal, glass and reinforced materials can create strong reflections. A signal that bounces off a building before reaching you travels a longer route than the direct one, so it arrives later. If the receiver treats that delayed signal as direct, the satellite appears farther away than it really is.

Two situations get lumped together here, and an article in Inside GNSS argues they should be kept apart. In multipath, a reflected copy mixes with the direct signal, and the interference can push the measurement either way. In non-line-of-sight reception, or NLOS, the direct signal is blocked entirely and only a reflection arrives. The article says an NLOS error equals the extra distance the signal traveled, is always positive, and is potentially unlimited, and that NLOS signals can be nearly as strong as direct ones. It adds that most multipath mitigation techniques don't correct NLOS errors, and calls NLOS one of the most important contributors to GNSS position error in urban environments.

So the receiver may be tracking a real satellite whose signal took a detour. The satellite isn't transmitting the wrong position. The receiver is measuring a signal that bounced. Because position is calculated from several measurements at once, one bad range can drag the whole solution sideways, which is why the phone can show a plausible location on the wrong side of the street.

Multipath may sound familiar from yesterday's Bluetooth article, where reflections changed signal strength. Here the same phenomenon corrupts timing instead.

A paper presented at the ION GNSS+ conference in 2025 by researchers from MITRE and NASA Langley Research Center puts it in plain terms: urban buildings block line-of-sight signals and reflect copies of them, causing much greater pseudorange errors than in open sky, and the errors are most risky when they are comparable to or larger than the width of the street canyon. The team drove two commercial receivers around a dense urban canyon, using a 3D map and ray-casting to label each measurement as line-of-sight or NLOS. The paper characterizes the errors, which is why I'm not claiming a fix from it.

From Nanoseconds to City Blocks

Suppose a reflected path is 30 meters longer than the direct one. That's about 100 nanoseconds, yet the receiver may read it as 30 meters of extra range. The final error isn't necessarily 30 meters, since the receiver combines many measurements and solves for clock bias too, but in a downtown grid tens of meters can put you on the wrong road or past an intersection. The physics happens on a nanosecond scale. The consequences appear on a city map.

Open sky is kinder. More satellites are visible, better spread, with fewer structures to block or reflect them. GPS.gov gives approximately 4.9 meters, or 16 feet, as a typical open-sky accuracy figure for GPS-enabled smartphones, while emphasizing that actual accuracy depends on satellite geometry, obstructions, atmospheric conditions, and receiver quality. Newer devices that use additional frequencies, satellite constellations, and positioning techniques may perform differently. Near tall buildings, even a capable receiver can suffer much larger errors because signals can be blocked or reflected before reaching the phone. That's why navigation can feel flawless on a rural interstate and struggle downtown. The satellites themselves haven't become less accurate; the surrounding environment has changed how their signals reach the receiver.

The Blue Dot Is an Estimate

A blue dot looks precise, but it's an estimate with uncertainty. Many apps show a circle around it, and a smaller circle suggests more confidence, but that uncertainty is itself an estimate, and when reflections introduce systematic errors the true position can fall outside it. The dot is the software's best guess at that moment, not proof of where you are.

Picture the jump. You're driving downtown with good measurements, then pass a tall building that blocks two satellites. The receiver must solve with a different combination of signals, some of them reflected, and the estimate shifts across the street. When the obstruction clears, it returns. Your phone didn't move. The measurements behind its estimate changed.

Navigation apps also don't necessarily show raw estimates. They use map matching, combining position data with road geometry and recent movement. If the phone puts you ten feet off the road at 35 miles per hour, the software can reasonably conclude you're still on the road and snap you there. That produces smoother navigation, but it opens another way to fail. Suppose two roads run parallel a block apart, and reflected signals push your estimate toward the neighbor. Both roads fit your direction of travel, so the app may decide you're on the wrong one and reroute accordingly. A moment later, better measurements arrive and it corrects itself. The map wasn't necessarily wrong. A positioning error was combined with a wrong guess about which road you're on.

Filling the Gaps: Dead Reckoning and Sensor Fusion

Navigation can also continue through a tunnel where no satellite signals reach you. The app may estimate your movement from your last known position, speed, road geometry and motion sensors, a technique called dead reckoning. It accumulates uncertainty over time, since without fresh reference measurements small errors grow.

Accelerometers detect forces on the device, but estimating position from them means integrating acceleration into velocity and then position, and even a tiny bias produces large drift. Gyroscopes measure rotation but also have noise and drift. Neither can replace satellites for long-term position.

Sensor fusion combines imperfect sources. Satellites give position that can go wrong between buildings; accelerometers and gyroscopes give continuity that drifts. Statistical filters weigh each by its uncertainty, so satellite measurements can correct inertial drift while inertial data bridges brief outages.

Other Clues: Wi-Fi, Cell Towers and Assisted GPS

Satellites aren't a phone's only location source. A phone can detect nearby Wi-Fi access points and compare their identifiers with a database of approximate locations, which helps indoors and in dense areas. That data can be stale, though: move a router and it contributes misleading evidence until the database is corrected. Cell networks give a general area, which helps when satellites aren't available but doesn't usually deliver street-level accuracy.

Assisted GPS, or A-GPS, is often misunderstood. Cell towers don't do the positioning instead of satellites. Assistance data, such as satellite orbit information, timing and an approximate starting location, helps the satellite receiver find and interpret signals faster. A phone turned on after a long trip may take longer to get a first fix when it lacks current assistance data and has poor sky view.

Satellite signals are also extremely weak by the time they reach you. Receivers can track them because their structure is known, but a roof, dense building or bad antenna position can bury them. That's why satellite navigation is unreliable indoors and underground, and why a phone can show a location from Wi-Fi, cell information or a previous position without receiving a good satellite signal at all.

Some newer phones also receive more than one GNSS frequency, such as L1 and L5, which can reduce certain errors, including ionospheric effects. But buildings obstruct and reflect multiple frequencies too, so it doesn't turn a city street into open sky.

Heading, Walking and Other Wrinkles

A phone can know where you're moving without knowing which way you're facing. Heading may come from a magnetometer, which metal and vehicle structures can disturb, or from your direction of travel, which says nothing about how the phone is oriented. At low speeds, small position errors can make the inferred direction wobble, which is why the arrow rotates when you're standing still.

Walking is harder than driving. Cars follow roads and move fast enough that direction becomes clear, while pedestrians cross streets, cut through plazas, stop suddenly and swing a phone around. Underground, ordinary GPS can't calculate from signals it isn't receiving, so a position moving along a subway line is a prediction, not a fresh measurement from space.

GPS itself doesn't need internet access. The satellites broadcast and the phone receives them directly. What needs data is the map, traffic and search, unless you've downloaded offline maps.

It's Often Not the Satellites

When navigation goes wrong, people blame the satellites. The global systems generally provide reliable signals, and the bigger problem is usually local. The archived GPS.gov page identifies blockage, reflections and receiver quality as key factors, and also says that in many cases a device's GPS hardware is working fine while its mapping software is faulty, citing incorrectly drawn maps and mislabeled points of interest. It adds that the US government cannot correct mapping errors in consumer devices. So the marker can be in the wrong place for several reasons: a correct map with distorted measurements, a correct position on a wrong map, or a map-matching step that picks the wrong road and compounds the original error.

Researchers keep working on urban positioning, including ways to identify non-line-of-sight signals and 3D building models that predict which satellites should be visible from a given spot. I couldn't verify specific accuracy gains for this article, but the point holds: better location isn't just more satellites, it's helping the receiver decide which signals to trust.

If navigation gets confused downtown, move somewhere with more open sky if you safely can, make sure location services are enabled, and keep your phone and app updated. If the map repeatedly puts you on the wrong road, check whether the map itself is right. And when driving, don't make sudden decisions based on a marker that has visibly jumped between streets. The directions are an aid, not a substitute for looking at the road.

The Bard's Take

Satellite navigation is one of the most remarkable technologies in a modern phone. A device that fits in your pocket picks up extraordinarily weak radio signals from satellites thousands of kilometers away, measures their timing, combines several of them, and works out where you are on the planet. Under open sky, that can be accurate to a few meters.

Between tall buildings, it gets complicated. Some signals are blocked completely. Others bounce off glass, concrete and metal before arriving, travel farther than a direct signal, and make satellites look farther away than they are. The receiver then calculates a position from measurements that no longer agree with reality. A few tens of nanoseconds of extra travel time becomes several meters, and several meters can be the difference between your road and the parallel street beside it.

Phones compensate with multiple constellations and frequencies, Wi-Fi and cellular clues, motion sensors and software that compares estimates against the road network. The results can be impressively stable, but not infallible: a map-matching algorithm can pick the wrong street, inertial estimates drift, and an urban canyon can deprive the phone of the measurements needed to fix them.

GPS navigation isn't a magical connection between your phone and an all-knowing satellite. It's a continuous exercise in measurement, geometry, timing and probability. The phone listens to signals from space, measures when they arrive, calculates how far they seem to have traveled, folds in other sensors, and makes its best estimate.

Most of the time, that estimate is excellent. Occasionally, a skyscraper sends the signal on a detour, and suddenly your phone thinks you're driving through the building across the street. Your phone doesn't lose GPS because it forgot where you are. It loses accuracy because the signals telling it where you are no longer describe the straight paths the calculations expect.

Sources

← Back to the Desk — Technology