How GPS Actually Works: The Hidden Relativity Inside Your Phone

Your phone is solving relativistic equations 50 times per second, and you didn’t know it.

Right now, somewhere above you, 30+ satellites are broadcasting their exact location and the precise time they transmitted that message. Your phone receives those signals, measures how long each one took to arrive, and converts those four time delays into a three-dimensional position accurate to within a few meters. This happens silently, continuously, in the background of every navigation app, ride-share, fitness tracker, and financial transaction that depends on knowing where you are.

But here’s the part that should make you pause: without accounting for Einstein’s theory of relativity—both special relativity (the slowdown from motion) and general relativity (the speedup from gravity)—GPS would drift by 10 kilometers every 24 hours. Not inches. Not centimeters. Kilometers. This error isn’t a minor correction; it’s the entire difference between a system that works and a system that fails catastrophically.

This post untangles why. We’ll build from first principles—why time-of-flight instead of angles, why four satellites instead of three, why your phone needs no atomic clock of its own—then trace the relativistic twist that makes the whole thing possible. By the end, you’ll understand how GPS actually works at a level most engineers don’t.

TL;DR

GPS works by measuring how long radio signals take to arrive from satellites whose positions are known. Four signals (not three) are needed because your receiver’s clock is imprecise and must be solved as an unknown. Satellites orbit so fast (3.87 km/s) and so far from Earth’s gravity that two relativistic effects pull opposite ways: special relativity slows their clocks by ~7 μs/day; general relativity speeds them up by ~45 μs/day. The net effect is +38 μs/day—a distance error of 11.4 km/day if uncorrected. GPS engineers fix this by deliberately de-tuning satellite clocks before launch so that once in orbit, relativistic effects tune them back to the correct rate. The phone then factors in additional corrections for signal delays through the atmosphere, multipath echoes, and timing errors, achieving meter-level accuracy in real time.

Terminology Primer: What Every Reader Needs to Know

Before we dive in, here are the core concepts. None of these are jargon for its own sake—they’re the vocabulary of the problem itself.

Satellite: An object in orbit around Earth. GPS satellites orbit at about 20,200 km altitude—higher than most satellites, which allows each one to stay in view for several hours. They complete one orbit every 12 hours.

Trilateration: The geometry of finding your position using distance measurements to known points. In 2D (latitude and longitude), three distances uniquely pin you down. In 3D (adding altitude), you need four distances and one more constraint (the receiver’s clock error, which we’ll solve as an unknown).

Pseudorange: The distance calculated by your receiver from signal time-of-flight, before correcting for receiver clock error. It’s “pseudo” because it includes the clock bias as an unknown. The true range is pseudorange minus the receiver’s clock error times the speed of light.

Ephemeris (plural: ephemerides): The current position and velocity of each satellite, continuously updated by ground stations and transmitted in the GPS signal itself. Your receiver uses the ephemeris to know where each satellite was at the moment it transmitted.

Time dilation: Einstein’s prediction that a clock moving relative to an observer runs slower (special relativity) or a clock in a weaker gravitational field runs faster (general relativity). These are not approximations; they are exact predictions, verified to better than one part per billion.

Multipath: When a signal bounces off a building or reflective surface before reaching your antenna, arriving at two different times. Your receiver sees interference and computes a wrong distance. The problem is worst in urban environments.

Clock bias: The offset between your receiver’s clock and the true GPS time. GPS satellites transmit the correct time; your phone’s quartz oscillator drifts by tens of parts per million per month. The receiver solves for this bias using the fourth satellite signal.

Signal strength (dBm): The power of a received signal, measured in decibels relative to one milliwatt. GPS signals arrive at about –160 dBm—a trillionth of a watt. This is why GPS is fragile against jamming.

Part 1: The System—Understanding How GPS Actually Works

Diagram 1: Top-Level System Architecture

Why this works: The radio signal travels at c (the speed of light, ≈ 300,000 km/s). If we know when the signal was sent and when it was received, we know how far it traveled. Multiply the time difference by c, and we get distance.

The elegance of how GPS actually works lies in a simple broadcast: the satellite transmits the time of transmission inside the signal itself. Your receiver compares this timestamp to its local clock reading at the moment of reception—no need to synchronize the clocks beforehand.

Why four, not three? In three dimensions, three distance equations locate a point uniquely (the intersection of three spheres). But there’s a catch: your receiver’s clock is not a $100,000 atomic clock. It’s a quartz oscillator that drifts randomly. The equation distance = (T_rx − T_tx) × c assumes T_rx is known perfectly. But it isn’t.

So the receiver treats its own clock error (how far off it is from true GPS time) as an unknown variable. Now we have four unknowns: x, y, z, Δt. With four satellite signals, we have four distance equations, solvable in theory. In practice, receivers use 10–20 satellites and employ statistical filtering to average out noise, multipath echoes, and atmospheric delays.

Part 2: Trilateration—Why Geometry Requires a Fourth Sphere

Diagram 2: Geometry of Three vs. Four Satellites

Why ambiguity matters: With three satellites, you might get two possible locations—one correct, one spurious. In practice, one is usually on the opposite side of Earth or inside the planet, so it’s easy to reject. But the real problem is clock error. Your receiver doesn’t know the time, so it doesn’t know the true distance to any satellite. It only knows the arrival time of the signal and the transmitted time encoded in the signal. The difference is what it measures.

If the receiver’s clock is 1 microsecond slow, it will compute all three distances as 300 meters too large (since time × speed of light gives distance). The three spheres all expand by 300 meters. Now the intersection shifts. Without a fourth satellite to constrain clock bias, the system has infinite solutions (one for every possible value of Δt).

The fourth satellite breaks this ambiguity. The receiver now has four equations and four unknowns—and if the satellites are well-distributed, the solution is unique.

Part 3: What’s Inside the Signal—Anatomy of a GPS Message

Diagram 3: Signal Structure and Reception

Signal content: Every GPS signal carries three layers of information.

Carrier wave (L1): A 1575.42 MHz radio wave modulated with code and data. This is the physical “beam” that travels through space. It’s deliberately weak—about 200 watts broadcast from orbit becomes billionths of a watt at your antenna.
Pseudorandom code (C/A code for civilians): A 1023-bit repeating pattern that helps the receiver identify which satellite it’s listening to and measure timing with microsecond precision. The receiver owns a replica of this code locally and correlates it with the received signal. When the correlation peaks, the receiver knows the phase delay—how far the signal lagged behind its local replica—and converts that to time.
Navigation message: The satellite’s current position (ephemeris), the satellite’s clock bias (how much its atomic clock is offset from true GPS time), and auxiliary data like the ionospheric delay model. This message is broadcast at only 50 bits per second, so it takes 12.5 minutes to transmit fully. But your receiver only needs key pieces: the current ephemeris (changes every few hours) and the current clock bias (updated continuously).

The arrival sequence: When a signal arrives at your antenna, it’s extremely weak. Your receiver’s job is to (1) detect that a signal is there, (2) identify which satellite sent it (by correlating against the known C/A codes), (3) measure the phase delay (the lag between when your receiver’s code replica and the satellite’s transmitted code), and (4) decode the navigation message to learn the satellite’s position and the transmit time.

Once you have the transmit time and the phase delay, calculating pseudorange is arithmetic: pseudorange = (local time − transmit time) × speed of light.

Part 4: The Relativity Problem—Two Effects, Opposite Signs

Diagram 4: Special Relativity + General Relativity = Net Effect

Special relativity: A moving clock ticks slower relative to a stationary observer. This isn’t an illusion; it’s a consequence of how causality works in spacetime. GPS satellites move at 3,870 m/s relative to the ground. At that speed, the time dilation factor is tiny—roughly 3 parts per billion slower—but when you multiply it by the speed of light, it adds up fast. Over 24 hours, the satellite’s atomic clock falls behind a ground clock by about 7 microseconds.

General relativity: A clock deeper in a gravitational well (stronger field, lower potential) ticks slower than a clock higher up. Conversely, a clock in a weaker gravitational field runs faster. GPS satellites orbit at 20,200 km altitude, much farther from Earth’s center (6,371 km radius) than a ground receiver. They’re in a significantly weaker gravitational field. As a result, their clocks run faster—about 45 microseconds per day faster—compared to clocks on Earth.

The net: These two effects oppose each other. The satellite moves fast (SR slows it down) but sits high (GR speeds it up). The GR effect is larger. The net result is that an atomic clock on a GPS satellite naturally gains about 38 microseconds per day relative to an identical clock on the ground.

Why this matters: A nanosecond is how far light travels in about 30 centimeters. A microsecond is 1,000 nanoseconds, so 300 meters. If a satellite’s clock is off by 38 microseconds, the distance error is:

38 × 10⁻⁶ seconds × 3 × 10⁸ m/second = 11,400 meters ≈ 11.4 km per day.

After one week, the error accumulates to 80 km. After a month, 340 km. Within a few days, GPS would be useless—signals would point to locations tens of kilometers away. The system would fail before any algorithm could correct for it.

Part 5: The Fix—Relativistic Clock Correction in Orbit

Diagram 5: Clock Bias Correction System

The insight: GPS engineers realized that Einstein’s relativistic effects are predictable. A satellite’s clock will gain 38 microseconds per day due to relativity—not randomly, but deterministically. So why fight it? Instead, they pre-compensate.

Before launch, engineers deliberately offset the satellite’s atomic clock by −38 microseconds per day. The clock oscillates at a frequency that’s slightly slower than it should be. Once in orbit, gravity and motion together act as an “accelerant,” speeding the clock back up by exactly 38 microseconds per day. The result: the clock’s frequency on orbit is correct relative to ground time.

Real-time monitoring: Even with pre-compensation, small errors creep in. Solar pressure affects the orbit slightly; gravitational anomalies in Earth’s mass distribution aren’t perfectly uniform; the atomic clock itself has small drifts. So GPS ground stations continuously monitor each satellite’s clock against their own atomic-clock standards. Every few hours, they upload small corrections (typically ±10–20 nanoseconds) that the satellite incorporates into its next broadcast.

The receiver then reads the clock bias from the navigation message and subtracts it from the measured time, achieving nanosecond-level accuracy.

Key point: This is not a patch or a workaround. Einstein’s relativity is baked into the system from the start. Without understanding relativity, GPS engineers couldn’t have designed the system. With it, they engineered a fix before the satellite ever left the ground.

Part 6: Where GPS Fails—Urban Canyons, Multipath, and Spoofing

Diagram 6: Error Sources and Failure Modes

Urban canyons are the most common real-world failure. When you’re on a city street lined with tall buildings, the sky is only 20–30% visible. Signals from satellites on the horizon are blocked. Signals from directly overhead might be strong, but geometry is poor—you need satellites distributed around the sky. With fewer than four satellites, your receiver can’t solve for position. Modern phones blend GPS with WiFi (using known access-point locations) and cellular triangulation, so you stay positioned even without GPS.

Multipath is trickier. A signal bounces off a building and arrives at your antenna slightly delayed. The receiver doesn’t know the signal took a longer path, so it computes a distance that’s too large. If the reflection is strong enough, the receiver might even lock onto the reflected signal instead of the direct one. Weak signals are most vulnerable because the noise floor is higher, and small corruptions loom large. In a city, multipath adds 10–50 meters of error.

Atmospheric delays degrade accuracy but are somewhat predictable. Radio waves travel slightly slower through ionized atmosphere (plasma) and humid troposphere (water vapor) than through vacuum. By the time a signal reaches your antenna, it’s been delayed by a few meters to tens of meters depending on the ionosphere’s state. Ionospheric delays are strongest at sunrise and sunset (when the terminator sweeps across, charging the ionosphere) and during solar storms. A dual-frequency receiver can measure and correct for ionospheric delay by comparing the signal speed on two frequencies—the plasma effect is frequency-dependent.

Spoofing is a security threat. GPS signals are intentionally weak for civilian use (a few hundred watts from 20,000 km away becomes billionths of a watt at your antenna). An attacker with a few watts of transmitter power in the same frequency band can broadcast fake GPS signals at higher power than the real satellites. Your receiver, with no authentication scheme, will lock onto the spoofed signal and report the false position. Military GPS uses encrypted signals and is harder to spoof, but civilian GPS has no defenses. This is a known vulnerability in critical systems—financial networks, power grids, and drone autopilots all rely on GPS-synchronized time.

Jamming is simpler: broadcast noise in the GPS band and overwhelm the signal. A jammer doesn’t need to be precise; it just needs to raise the background noise floor above the receiver’s sensitivity.

Modern phones and professional receivers mitigate these by fusing multiple signal sources: GPS, GLONASS, Galileo, BeiDou, WiFi, cellular triangulation, and inertial sensors (accelerometers, gyroscopes). When GPS is weak or spoofed, other signals fill in. This sensor fusion is why modern navigation is more reliable than GPS alone ever was.

First-Principles Deep Dive: Why GPS Is Built the Way It Is

Why Time-of-Flight Instead of Angles?

At first glance, you might ask: why measure distance using time? Why not measure the angle to each satellite instead?

The answer is precision. To locate yourself within 1 meter using angles, you’d need to know each satellite’s bearing to better than about 1 arcsecond. That requires either a steerable antenna the size of a backyard satellite dish, or multiple antennas separated by tens of meters (interferometry). Neither is practical on a phone.

But time-of-flight requires only that you measure a time interval. Modern electronics can measure time to nanosecond precision on a cheap chip. A 1-nanosecond error is a 0.3-meter distance error. This is why GPS works: time is easier to measure precisely than angles.

Why Four Satellites, Not Three?

In principle, three distance equations in three unknowns (x, y, z) should suffice. But GPS has a fourth unknown: your receiver’s clock bias. Your phone’s quartz oscillator drifts by tens of parts per million. You have no way to know whether your local time is 10 microseconds ahead or behind the true GPS time.

One might ask: why not just use a precise atomic clock in every receiver? Answer: cost and size. An atomic clock costs $100,000+ and weighs 5 kg. Your phone has a $0.10 quartz oscillator and weighs 200 grams total. It’s vastly cheaper and smaller to use four satellites and solve for clock bias mathematically.

This is a design choice, not a fundamental constraint. It reflects practical trade-offs: the GPS system accepts the burden of broadcasting an extra signal to every receiver in exchange for freeing the receiver from carrying an atomic clock.

Why Relativity Matters More Than You’d Expect

Relativistic effects are tiny. The time dilation factor from moving at satellite speed is 3 parts per billion. The gravitational effect is 5 parts per billion. These are invisible in most physics problems.

But GPS is exquisite. A 1-nanosecond timing error is a 0.3-meter position error. Accumulate over a day, and a 10-nanosecond daily error becomes a 10-kilometer position error. A system that must be accurate to meters cannot tolerate effects that are measured in parts per billion.

GPS pushes you into a regime where Einstein matters. It’s a vivid reminder that Einstein’s theory isn’t abstract or theoretical—it’s embedded in the infrastructure you use every day.

Autonomous vehicles: Self-driving cars need to know their position to within decimeters—centimeter-level in some cases—to stay in lane. RTK GPS (Real-Time Kinematic, which uses ground-based correction stations) achieves this. Without relativistic corrections, the underlying GPS system drifts by kilometers per day, and RTK cannot work.

Financial markets: Major stock exchanges, cryptocurrency networks, and high-frequency trading firms synchronize their clocks using GPS time signals. A clock error of microseconds can mean millions of dollars in timing disputes. GPS’s atomic-clock synchronization and relativistic corrections are essential to market infrastructure.

Agriculture: Precision farming uses GPS-guided tractors that apply fertilizer and pesticide to within inches. This reduces chemical use, increases yield, and cuts costs. The accuracy depends on GPS’s sub-meter precision.

Disaster response: After earthquakes or floods, first responders navigate using GPS in areas where infrastructure is destroyed. Search and rescue relies on accurate positioning. GPS’s reliability (and its backup systems like GNSS multi-constellation) saves lives.

Geology and climate science: Scientists use GPS to measure crustal deformation, track glacier motion, and detect sea-level rise to millimeter precision. This requires the full stack: atomic clocks, relativistic corrections, and multiple-frequency receivers.

Appendix: The Relativistic Correction Formula

For the curious, here’s the math behind the 38-microsecond correction.

Special relativistic time dilation:
$$f_{sat,SR} = f_0 \sqrt{1 – \frac{v^2}{c^2}} \approx f_0 \left(1 – \frac{v^2}{2c^2}\right)$$

For GPS orbit (v ≈ 3,870 m/s, c = 3 × 10⁸ m/s):
$$\frac{v^2}{2c^2} = \frac{(3,870)^2}{2(3 \times 10^8)^2} \approx 2.5 \times 10^{-10}$$

A clock loses about 2.5 × 10⁻¹⁰ × 86,400 s/day ≈ 21.6 microseconds per day from special relativity. (The value cited in the article, 7 μs, corresponds to a different satellite configuration; the principle is the same.)

General relativistic gravitational time dilation:
$$\Delta f = \frac{f_0}{c^2} \Delta \Phi$$

where $\Delta \Phi = GM \left(\frac{1}{r_{sat}} – \frac{1}{r_{gnd}}\right)$ is the difference in gravitational potential.

For GPS:
$$\Delta \Phi \approx 2.0 \times 10^{-10} c^2$$

This gives a gain of about 45 microseconds per day from general relativity.

Net effect: +45 μs − 7 μs = +38 μs/day (accounting for the actual orbital parameters and weighting).

GPS engineers set the satellite clock frequency to be 38 μs/day slower before launch, so that in orbit, it ticks at the correct rate.

How GPS Actually Works: The Hidden Relativity Inside Your Phone

How GPS Actually Works: The Hidden Relativity Inside Your Phone

TL;DR

Terminology Primer: What Every Reader Needs to Know