Why GPS Needs General Relativity

June 10, 2026 18 min read The Physics of Computing

gpsphysicsnetworkingtimeinfrastructure

Contents

Every time a navigation app pins your location to the correct side of a street, it is silently depending on two of the most counterintuitive results in the history of physics. The GPS satellites overhead are not simply bouncing signals back like radio rangefinders. They are flying clocks, and the accuracy of your position is a direct function of how accurately those clocks agree with each other and with clocks on the ground. Keeping that agreement requires accounting for the fact that time does not pass at the same rate at 20,200 kilometers altitude as it does at sea level — a consequence of Einstein’s general theory of relativity — and for the fact that a clock moving at orbital velocity ticks more slowly than a stationary one, a consequence of special relativity. Both effects are real, both are measurable, and together they shift GPS clock rates by 38 microseconds per day. Left uncorrected, that drift accumulates into roughly 10 kilometers of position error every 24 hours.

This is not a curiosity of interest only to physicists. The GPS system is the stratum-0 time reference for a significant fraction of the world’s telecommunications infrastructure, financial exchanges, and power grids. The GPSDO (GPS-disciplined oscillator) units that serve as stratum-1 NTP sources in well-run homelabs and data centers are directly consuming a signal whose accuracy depends on a working implementation of relativistic mechanics. Understanding why is worth the effort.

How GPS Actually Works

The Global Positioning System consists of a constellation of satellites in medium Earth orbit broadcasting continuous radio signals. A receiver computes its own position not by measuring signal strength but by measuring signal arrival times with extraordinary precision and working backward to geometry.

Constellation Architecture

The operational constellation is maintained at a minimum of 24 satellites, with the full deployed fleet running around 31 active vehicles as of 2026. The satellites are arranged in six orbital planes inclined at 55 degrees to the equator, with four or more satellites per plane. The orbital altitude of 20,200 kilometers (the semi-major axis works out to about 26,560 km from Earth’s center) gives each satellite an orbital period of approximately 11 hours 58 minutes — half a sidereal day — which means the constellation geometry repeats relative to the ground every day, making satellite visibility predictable. At any point on Earth’s surface, at any time, a receiver with a clear sky view sees between five and eight satellites simultaneously. Robust position fixes require four.

        GPS Constellation — Six Orbital Planes (top view, schematic)

                        North Pole
                            |
               Plane F   Plane A
                  \       /
          Plane E  X-----X  Plane B       Each plane: 4+ satellites
                  /       \               Inclination: 55° from equator
               Plane D   Plane C          Altitude: 20,200 km
                            |
                        (Earth)

        Ground track repeats every ~23 h 56 min (sidereal day)
        Receiver always sees 5-8 satellites with clear sky

The ground control segment — the Master Control Station at Schriever Space Force Base in Colorado, with monitor stations worldwide — continuously tracks every satellite, computes updated ephemeris (orbital position) data, and uploads corrections to the satellites. The corrections include clock drift offsets, relativistic adjustment parameters, and atmospheric modeling coefficients. Each satellite broadcasts a navigation message containing its precise orbital parameters and clock correction data so that any receiver can compute the satellite’s position at any moment.

Signal Frequencies: L1 and L2

GPS satellites transmit on two primary carrier frequencies in the L-band:

Signal	Frequency	Wavelength	Primary Use
L1	1575.42 MHz	~19 cm	Civil navigation (C/A code + P(Y) code)
L2	1227.60 MHz	~24 cm	Military / dual-frequency ionospheric correction
L5	1176.45 MHz	~25 cm	Safety-of-life, modernized civilian use

The L1 frequency carries the civilian Coarse/Acquisition (C/A) code — a 1,023-chip pseudorandom sequence repeated every millisecond — along with the navigation message. The receiver generates an identical copy of the PRN (pseudorandom noise) sequence and slides it in time until it correlates with the incoming signal. The time offset between the receiver’s local clock and the received code chip tells the receiver how long the signal spent in transit.

At the speed of light, 299,792,458 meters per second, a 1-nanosecond timing error translates to a 30-centimeter range error. Get the timing wrong by 100 nanoseconds and you have introduced 30 meters of error into every range measurement. Get it wrong by 1 microsecond and you are off by 300 meters. The entire precision of the system depends on clocks accurate to nanoseconds or better — which is why the relativistic effects, operating at the level of tens of microseconds per day, are not a small perturbation but an existential problem.

Trilateration, Not Triangulation

GPS computes position through trilateration: measuring distances from known points and finding the intersection. If you know you are exactly 20,187 km from satellite A, you are somewhere on a sphere of that radius centered on A. Add satellite B and you are on the intersection of two spheres — a circle. Add satellite C and you get two points, one of which is usually absurd (deep underground or in space). The fourth satellite breaks the remaining ambiguity and — crucially — corrects for receiver clock error.

        Trilateration (2D simplified)

        Sat A                    Sat B
          o                        o
          |`-.                  ,-'|
          |   `-.            ,-'   |
          |      `-.      ,-'      |
          |    r_A  `-.,-'  r_B   |
          |         ,-'`-.         |
          |      ,-'      `-.      |
          |   ,-'  Receiver  `-.   |
          | ,-'      HERE      `-. |
          |'                      `|
          |        r_C             |
          o
        Sat C

        Each sphere: set of all points at distance r from satellite
        Three spheres intersect at one point (+ its mirror underground)
        Fourth satellite resolves clock bias in receiver

A receiver’s internal clock is far cheaper and less stable than the atomic clocks aboard the satellites. Instead of assuming the receiver clock is correct, the system treats receiver clock offset as a fourth unknown, alongside the three spatial coordinates. Four satellites, four unknowns — the system is exactly determined. This is why GPS needs four satellites for a position fix and not three: the fourth equation exists to solve for time, not for space.

Why Clock Accuracy Is Everything

The core constraint of GPS is this: position error scales linearly with timing error at the speed of light.

    Timing-to-range error relationship

    1 nanosecond  = 0.30 meters  (30 cm)
    10 ns         = 3.0 meters
    100 ns        = 30 meters
    1 microsecond = 300 meters
    10 us         = 3 km
    38 us/day     = ~11.4 km accumulated per day (if uncorrected)

GPS satellites carry rubidium and cesium atomic clocks accurate to roughly 20–30 nanoseconds, corresponding to range errors on the order of 6–9 meters before any other error source. The Master Control Station monitors these clocks and uploads corrections, but the fundamental stability budget — set by the clocks themselves and by physics — is what determines whether those corrections are possible at all.

Each satellite’s navigation message includes clock correction coefficients: a constant offset term (a₀), a frequency drift term (a₁), and a frequency drift rate (a₂). The receiver applies these to compute the satellite’s actual clock reading at the moment of transmission. The relativistic correction is baked into this correction scheme. Strip it out and the clock model simply becomes wrong by a growing amount.

Special Relativity: The Moving-Clock Effect

Special relativity’s time dilation result is the easier of the two effects to state: a clock moving at velocity v relative to an observer ticks more slowly than the observer’s clock by a factor of 1/√(1 - v²/c²). This is not a mechanical effect from vibration or acceleration; it is a geometric property of spacetime. Moving clocks genuinely run slow from the perspective of stationary observers.

GPS satellites orbit at approximately 3.874 km/s (the orbital speed at 20,200 km altitude, derived from the gravitational parameter and orbital radius). The Lorentz factor for this velocity is small but precisely calculable:

    Special relativistic time dilation

    v = 3,874 m/s  (GPS orbital velocity)
    c = 299,792,458 m/s

    v²/c²  = (3874)² / (299792458)²
           = 15,007,876 / 89,875,517,873,681,764
           ≈ 1.670 × 10⁻¹⁰

    Time dilation factor: sqrt(1 - v²/c²) ≈ 1 - v²/(2c²)
                        ≈ 1 - 8.349 × 10⁻¹¹

    Clock runs slow by: 8.349 × 10⁻¹¹ × 86400 s/day
                      ≈ 7.21 × 10⁻⁶ seconds/day
                      ≈ -7.2 µs/day

A satellite clock, viewed from the ground, loses about 7.2 microseconds per day due to its orbital velocity. The sign is unambiguous: faster motion, slower clock. A naive GPS receiver that ignored this effect would accumulate 7.2 microseconds of timing error per day — around 2.2 kilometers of range error per day, growing without bound.

General Relativity: The Gravitational Blueshift

The general relativistic effect runs in the opposite direction and is larger in magnitude. Einstein’s general theory of relativity predicts that clocks in weaker gravitational fields tick faster than clocks in stronger fields. This is the gravitational time dilation result, and at the altitudes and accuracies relevant to GPS it is not a tiny correction — it is the dominant relativistic term.

At Earth’s surface, gravitational potential (taking the sign convention where the potential is negative and becomes less negative as you move away) is approximately -62.6 MJ/kg. At 20,200 km altitude the potential is significantly less negative — the satellite is further from the mass that curves spacetime. A clock at that altitude, compared to a clock at sea level, is in a shallower potential well and therefore ticks faster.

    General relativistic frequency shift (gravitational blueshift)

    Fractional frequency difference:
        Δf/f = ΔΦ/c²

    Where ΔΦ = gravitational potential difference (satellite - surface)
             = Φ_sat - Φ_surface
             ≈ (-GM/r_sat) - (-GM/R_earth)
             = GM × (1/R_earth - 1/r_sat)

    G = 6.674 × 10⁻¹¹ N m² kg⁻¹
    M = 5.972 × 10²⁴ kg   (Earth mass)
    R_earth = 6.371 × 10⁶ m
    r_sat   = 26.56 × 10⁶ m   (6371 + 20185 km, center to orbit)

    GM = 3.986 × 10¹⁴ m³ s⁻²

    1/R_earth = 1.570 × 10⁻⁷ m⁻¹
    1/r_sat   = 3.764 × 10⁻⁸ m⁻¹

    ΔΦ = 3.986 × 10¹⁴ × (1.570 × 10⁻⁷ - 3.764 × 10⁻⁸)
       = 3.986 × 10¹⁴ × 1.194 × 10⁻⁷
       ≈ 4.759 × 10⁷ J/kg

    Δf/f = 4.759 × 10⁷ / (2.998 × 10⁸)²
         = 4.759 × 10⁷ / 8.988 × 10¹⁶
         ≈ 5.295 × 10⁻¹⁰

    Clock runs fast by: 5.295 × 10⁻¹⁰ × 86400 s/day
                      ≈ 4.575 × 10⁻⁵ s/day
                      ≈ +45.7 µs/day

The satellite clock ticks faster than a ground clock by about 45.7 microseconds per day due to its weaker gravitational environment. Left uncompensated, GPS would accumulate roughly 45 microseconds of clock-ahead error per day — about 13.5 kilometers of range error per day.

The Net Effect: +38 Microseconds per Day

The two effects run in opposite directions. Special relativity makes the satellite clock slow (motion slows time), and general relativity makes it fast (altitude speeds time). The net result:

Effect	Cause	Daily Rate	Direction
Special relativistic time dilation	Orbital velocity ~3.9 km/s	−7.2 µs/day	Clock runs slow
General relativistic gravitational blueshift	Altitude 20,200 km, weaker gravity	+45.9 µs/day	Clock runs fast
Net combined effect		+38.4 µs/day	Clock runs fast

The gravitational effect wins. An uncorrected GPS satellite clock runs fast by about 38 microseconds every day. At 30 cm per nanosecond, 38,000 nanoseconds corresponds to 11.4 kilometers of range error per day, growing linearly. Within hours of launch, without correction, GPS would be useless for most navigation applications. Within a week, positions would be off by tens of kilometers.

This is not a theoretical concern or a historical curiosity. The effect has been measured repeatedly and independently. The GPS clocks run exactly as fast as the combined relativistic theory predicts. The theory is not an approximation that is close enough; it is correct at the level of precision the system demands.

How the Correction Is Implemented

The GPS program engineers knew about the relativistic corrections before the first satellite was launched in 1978. The correction strategy has two components: a factory offset built into each satellite before launch, and ongoing corrections uplinked from the ground segment.

Pre-Launch Clock Offset

Before a GPS satellite is launched, its onboard atomic clocks are deliberately set to run at a slightly lower rate than the standard cesium frequency of 9,192,631,770 Hz. The nominal frequency for GPS satellite clocks is set to:

    Standard cesium hyperfine transition: 9,192,631,770.000 Hz

    Relativistic rate correction factor:
        Δf/f = -38.4 µs/day / 86400 s/day
             = -4.444 × 10⁻¹⁰

    Pre-launch satellite clock frequency:
        f_sat = 9,192,631,770 × (1 - 4.444 × 10⁻¹⁰)
              = 9,192,631,770 - 4.087 Hz
              ≈ 9,192,631,765.9 Hz

The satellite clocks tick at approximately 9,192,631,766 Hz instead of the standard 9,192,631,770 Hz. On the ground, this makes them run slightly slow. Once in orbit, the combination of that deliberate slowdown and the natural relativistic speedup causes them to tick at very close to the correct rate as observed from the ground. A satellite clock that was set to the standard frequency would immediately begin drifting ahead; one set to the compensated frequency stays synchronized.

Ongoing Ground Corrections

The pre-launch offset handles the bulk of the relativistic correction but cannot account for the satellite’s actual orbit deviating from the nominal design parameters, or for small eccentricity in the orbit. GPS satellite orbits are not perfectly circular; an eccentric orbit means the satellite moves through varying gravitational potentials during each orbit, which produces a periodic relativistic correction oscillating at the orbital period.

The eccentricity correction, sometimes called the Sagnac correction or the e-sin-E term, is computed from the orbital parameters broadcast in the navigation message and applied in the receiver rather than in the satellite. Each satellite’s navigation message includes the orbital eccentricity e, and the receiver computes the additional clock correction:

    Periodic eccentricity correction (applied by receiver)

    Δt_rel = -2 × sqrt(GM × a) × e × sin(E) / c²

    Where:
        a = semi-major axis of orbit
        e = orbital eccentricity
        E = eccentric anomaly at time of transmission
        GM = Earth's gravitational parameter

    Magnitude: up to ~70 ns for typical GPS eccentricities (~0.01)

The ground control stations monitor every satellite continuously, compute residuals between predicted and measured clock behavior, and uplink updated clock correction polynomials at regular intervals. The broadcast clock corrections (a₀, a₁, a₂ in the navigation message) absorb both residual relativistic drift and any systematic drift in the atomic clock hardware itself.

A Note on GLONASS, Galileo, and BeiDou

The other major global navigation satellite systems apply analogous corrections. GLONASS satellites, operated by Russia, orbit at approximately 19,100 km altitude (slightly lower than GPS) at different inclinations; the relativistic corrections are similar in magnitude but not identical. The European Galileo system operates at 23,222 km altitude, producing somewhat larger gravitational blueshift corrections. China’s BeiDou system has satellites in three orbit regimes — geostationary, inclined geosynchronous, and medium Earth orbit — each requiring its own relativistic accounting. Every one of these systems was designed with relativistic corrections as a first-class engineering requirement, not an afterthought.

GPS-Disciplined Oscillators and NTP Stratum Hierarchies

For infrastructure engineers, the most direct consequence of GPS timing accuracy is the GPSDO (GPS-disciplined oscillator), which forms the basis of stratum-1 NTP servers in data centers, carrier networks, and well-appointed homelabs.

A GPSDO consists of a high-quality local oscillator — typically a temperature-compensated crystal oscillator (TCXO) or, in precision instruments, an oven-controlled crystal oscillator (OCXO) or rubidium standard — disciplined by a GPS receiver that continuously measures and corrects the local oscillator’s frequency against the GPS signal. The GPS receiver decodes the timing pulse-per-second (PPS) signal from the satellite, which carries the relativistically-corrected UTC-derived timestamp, and steers the local oscillator to agree with it.

The hierarchy of time accuracy looks like this:

    NTP Stratum Hierarchy (simplified)

    Stratum 0: GPS satellites (atomic clocks, relativistically corrected)
               |
               | (1-PPS signal, ~20-100 ns accuracy)
               v
    Stratum 1: GPSDO + NTP server  (data center, colocation, ISP)
               |
               | (NTP over network, ~1-10 ms accuracy depending on path)
               v
    Stratum 2: NTP clients sync'd to stratum-1
               |
               | (NTP, similar or slightly worse)
               v
    Stratum 3: Downstream clients, edge devices, workstations

When a stratum-1 NTP server reports that it is synchronized to GPS, it is synchronized — through the GPSDO, through the satellite receiver, through the satellite’s onboard atomic clocks, through the uplinked corrections from the Master Control Station — to a time standard that has correctly accounted for the relativistic behavior of clocks at orbital altitude. If the relativistic corrections were wrong, every stratum-1 GPS-disciplined server in the world would be wrong by the same amount, and every subsequent stratum would inherit that error.

The CCNA DHCP, DNS, and NTP guide covers the practical configuration of NTP synchronization in enterprise networks; the GPS timing chain is what sits above stratum 1 in that hierarchy. For homelab builders choosing time infrastructure, the homelab hardware guide discusses where a GPSDO with a small rooftop antenna fits into a serious lab build versus relying on pool.ntp.org. The former gives you sub-microsecond accuracy on-premises; the latter gives you milliseconds, which is usually sufficient but becomes relevant when you are running PTP (IEEE 1588 Precision Time Protocol) for applications that care about microsecond-level synchronization — packet timestamping, financial trading systems, 5G radio timing, or simply the satisfaction of knowing your clocks are correct.

Consumer-grade GPSDO modules (Leo Bodnar, Trimble Thunderbolt, used Symmetricom units) are available for well under a few hundred dollars and produce 1-PPS output accurate to the 10–50 nanosecond range after warm-up. That is a 30 cm timing window expressed as a distance. The relativistic correction is the reason those 30 centimeters are achievable rather than being swamped by kilometers of accumulated drift.

The Broader Lesson: Physics in the Infrastructure Stack

It is tempting to think of GPS as an application of Newtonian mechanics with a few physics-textbook corrections bolted on. The reality is that the system cannot function at all without the relativistic corrections. This is not a precision enhancement; it is load-bearing. Remove the corrections and you do not get slightly less accurate GPS; you get GPS that drifts by 10 kilometers per day and is functionally useless for navigation within hours.

The GPS story is part of a broader pattern worth recognizing for engineers who deal with infrastructure. Fundamental physics keeps showing up in systems that look, on the surface, like applied engineering. Shannon’s information theory sets hard limits on what data compression can achieve regardless of algorithmic cleverness — you cannot beat the entropy floor. The quantum mechanics of semiconductor junctions determines the characteristics of every transistor in every processor you run. The relativistic behavior of clocks at altitude determines whether your navigation and timing infrastructure is accurate.

These are not ornamental connections. The engineers who built GPS had to model general relativity correctly or the system would not work. The chip designers who shrink transistors below 5 nm have to account for quantum tunneling or the devices leak charge in ways that Kirchhoff’s laws do not predict. Physics is not something that happens in universities and occasionally informs engineering; it is the substrate on which every real system runs.

For infrastructure practitioners, the useful habit is to ask what physical constraints are load-bearing in the systems you operate. Not to become a physicist, but to understand where the actual limits come from — so that when something violates those limits, you recognize the situation rather than spending days debugging a software problem that is actually a physics problem.

Verdict

GPS requires general relativity not as a precision refinement but as a functional necessity. The two effects — special relativistic time dilation reducing satellite clock rates by 7.2 microseconds per day due to orbital velocity, and gravitational blueshift accelerating them by 45.9 microseconds per day due to altitude — combine to produce a net drift of +38.4 microseconds per day. Without compensation, that drift accumulates into roughly 10 kilometers of position error every day.

The correction is implemented at two levels: each satellite’s clocks are adjusted at the factory to run at a frequency approximately 4 Hz below the nominal cesium standard, so that in orbit they tick at the correct rate as observed from the ground; and the ground control segment uploads continuous clock correction polynomials that handle residual drift, orbital eccentricity effects, and hardware aging. The receiver applies a further periodic correction derived from the broadcast orbital parameters.

Every GPSDO-disciplined stratum-1 NTP server in the world is, in a direct and non-metaphorical sense, an application of Einstein’s field equations. The nanosecond-level timestamps your infrastructure relies on are only that accurate because the satellite clocks were engineered around a correct model of how gravity and motion affect the rate of time.

Sources

Neil Ashby, “Relativity in the Global Positioning System,” Living Reviews in Relativity, 6 (2003), 1. The definitive technical treatment; freely available at livingreviews.org.
IS-GPS-200 (Interface Specification for GPS), current revision. Published by the Space and Missile Systems Center; specifies the exact clock correction model and relativistic parameters used in the broadcast navigation message.
T.A. Herring, “The Global Positioning System,” Scientific American, February 1996 — approachable overview of system design.
Peter Misra and Per Enge, Global Positioning System: Signals, Measurements, and Performance, 2nd ed. (Ganga-Jamuna Press, 2006). The standard graduate-level GPS engineering textbook.
Richard Feynman, The Feynman Lectures on Physics, Vol. II, Chapter 42: “Curved Space” — the clearest non-mathematical explanation of gravitational time dilation in print.
GPS.gov — the official US government GPS information site — maintains current constellation status, signal specification documents, and the Interface Control Documents that define the L1/L2/L5 signal structure.
BIPM (Bureau International des Poids et Mesures), “Relativistic effects in satellite-based time transfer,” Technical Note. Covers the full treatment including Sagnac corrections for rotating Earth reference frames.