FreeBusy
Features How it works Blog Insiders Privacy

← Back to the blog

How GPS actually works (and why your phone sometimes thinks you're in the river)

6 June 2026 · 9 min read · By the FreeBusy team

At one minute past midnight on May 2, 2000, the world got more accurate.

That's the night the United States military stopped lying to civilian GPS receivers. For the previous decade, a deliberate jitter called Selective Availability had degraded the public signal to roughly 100 metres of error — a precaution against hostile actors building cheap precision weapons on top of free American satellites. President Clinton ordered it off. By 12:01 AM, every car nav unit, surveyor's rig, and hiking handheld on the planet was getting accuracy of about ten metres instead of a hundred. Nothing about any of those devices had changed. The truth just got cheaper.

GPS is the kind of infrastructure that disappears the moment it works. Let's pull it apart.

Four satellites and an atomic clock

The Global Positioning System is, at its core, a fleet of 31 operational satellites in medium Earth orbit, each broadcasting an extraordinarily precise time signal — accurate to about a nanosecond — along with its own position.

Your phone listens. For each satellite it can hear, it measures how long ago the signal was sent versus when it arrived. Multiply that travel time by the speed of light and you get a distance — but only a distance. The satellite could be anywhere on a sphere that far away.

With three satellites, three spheres intersect at two points — one in space, one on Earth, the one near the ground wins. With four, you correct for the fact that your phone's clock is much worse than the satellite's. That fourth measurement is what turns four nearly-good distances into one excellent position. The geometry is called trilateration; the math has been around since the 1820s; what changed in the 1990s is that we put atomic clocks in orbit.

Your phone uses four GPS systems at once

GPS is American. But it isn't the only game in orbit.

  • GLONASS (Russia) — about 24 operational satellites, full global coverage since the early 2010s.
  • Galileo (European Union) — about 28 satellites, fully operational from 2020.
  • BeiDou (China) — 45+ satellites, global coverage since 2020.
  • QZSS (Japan) — a regional constellation that augments accuracy in East Asia and Oceania.

Together, these are the Global Navigation Satellite Systems — GNSS. A modern smartphone listens to all of them simultaneously. On a clear day, your phone might be receiving signals from twenty or more satellites across four different constellations, picking the strongest, ignoring the rest, and triangulating in well under a second.

This is why your maps app gets a fix in a few seconds today, versus the two-to-twelve-minute "cold start" of a 2005 handheld receiver. The constellations got bigger, and the cellular network started helping.

A-GPS: the assist that made the smartphone

A-GPS — "Assisted GPS" — is the trick that made satellite navigation work on a phone you actually carry around.

The problem: a cold GPS receiver needs to download something called the almanac — a low-resolution map of where every satellite is going to be — before it can even start listening usefully. Over the air from a satellite at 50 bits per second, that's a 12-minute wait. From your carrier over LTE, it's a 50-millisecond wait.

A-GPS shifts that bootstrap onto the cellular network. Your phone asks the tower, "where are the satellites right now?" The tower answers in a packet. Your phone gets a fix in a second. The signal that fixes your position still comes from space; the signal that helps the phone find the right slice of space comes from the ground.

Why your map sometimes lies

Open Google Maps in midtown Manhattan and watch the blue dot drift. You're not moving. The dot is — sometimes thirty metres east of where you actually are, sometimes inside a building, occasionally in the river.

This is multipath. Your phone receives a signal directly from a satellite, and it receives the same signal a millisecond later after it bounced off the glass face of a skyscraper. The phone can't tell which is the real one. The "distance" calculation gets corrupted. The dot jumps.

Multipath is also why GPS is so bad indoors (the direct signal is gone; only reflections survive), in tunnels (no signal at all), and in dense forest (the canopy absorbs the L1 frequency). It's why an iPhone in a Brooklyn coffee shop reports your location as "somewhere on this block" rather than "table 4."

The newest phones have a partial fix. Modern receivers — Pixel 8 and later, iPhone 14 and later, recent Samsung flagships — can listen on a second GNSS frequency called L5. L5 was designed forty years after L1 and engineered specifically to reject multipath. Indoors and in cities, dual-frequency phones get accuracy of one to three metres where single-frequency phones get ten or fifteen.

For maps and ride-shares this is invisible. For an app that has to decide whether you're at home or next door, it's the difference between useful and embarrassing.

Why the dot snaps to the road

You're driving. The GPS thinks you're in the field next to the road. The map shows you on the road.

This is map matching. Every consumer mapping system runs a Hidden Markov Model over your last few seconds of position fixes, biases the answer toward the nearest valid road, and shows you the cleaned-up version. The raw GPS is always noisier than what you see. Snap-to-road is the polish.

The implication: when an app says "Akhil is at 21st Street and 8th Avenue," it almost never means the satellite measurement put him exactly there. It means the satellite measurement put him near there, and the layer above GPS guessed the rest.

What this means for a sharing app

Precision is a knob, not a number. A modern smartphone, in a city, with both L1 and L5 satellites visible, knows where it is to about a metre. The same phone in a basement knows where it is to about a kilometre. The same phone with the GPS chip off knows nothing.

A sharing app — any sharing app — has to make a choice about which slice of that range it shows to which person. Showing every contact your raw L5 fix is overshare. Showing every contact only the city you're in is undershare. The right answer is different for every relationship.

FreeBusy treats this as the central design question. Four levels — Off, Status only, Area, Precise — set per contact. Your spouse gets the metre-resolution dot. Your boss gets "at work" or "outside." Your great-aunt gets nothing unless you say otherwise. The satellites don't change. The amount of yourself you publish does.

See how FreeBusy handles location levels →

Notes & sources

  • U.S. GPS Program Office, "Selective Availability" — gps.gov/systems/gps/modernization/sa
  • U.S. GPS Program Office, "Space Segment" — gps.gov (current operational satellite count)
  • European Space Agency, "Galileo Goes Live" (2020) — esa.int
  • u-blox, "Dual-frequency GNSS: practical results" — u-blox.com
  • N. Zhu, J. Marais, D. Bétaille, M. Berbineau, "GNSS Position Integrity in Urban Environments: A Review of Literature", IEEE Transactions on Intelligent Transportation Systems, 2018
  • P. Newson and J. Krumm, "Hidden Markov Map Matching Through Noise and Sparseness", ACM SIGSPATIAL, 2009

© 2026 FreeBusy. All rights reserved.

Blog Privacy Policy Delete account Contact