The ten-second version

Your phone times radio signals from satellites about 20,000 km up, converts time into distance at the speed of light, and finds where four or more of those distances cross. That’s the whole trick. It drifts in wadis because canyon walls bounce the signals — the phone times a longer, reflected path and draws your dot where that fake distance says you are.

Knowing this changes how you read your screen in the mountains: a jumping dot in a gorge is the canyon talking, not the app breaking.

Time becomes distance, distance becomes position

Every GPS satellite broadcasts two things on repeat: where I am and what time it is — kept by an atomic clock on board. Your phone notes when each signal arrives, subtracts when it was sent, and multiplies by the speed of light. One satellite’s distance puts you somewhere on a sphere; four spheres intersect at a point. Engineers call it trilateration — distances, not angles.

Why four and not three? Your phone’s clock is a few-cent quartz crystal that drifts by microseconds, and one microsecond of clock error is 300 metres of position error. The fourth satellite lets the maths solve for one extra unknown: your clock’s error. Every fix corrects your phone’s clock as a side effect.

Multipath: the wadi problem

In open desert with the whole sky visible, a modern phone is typically good to about 3–5 metres. Then you drive into a Hajar gorge and the dot goes drunk. Two things happened:

  1. The walls became mirrors. Rock reflects GPS signals. A reflected signal travels farther than a straight one, so the timing — the entire basis of the fix — is now lying. At worst the straight signal is blocked entirely and only the reflection arrives, which is when your dot teleports across the wadi.
  2. The sky shrank. A narrow canyon hides most satellites and leaves the visible ones clustered overhead at poor angles. Bad geometry multiplies whatever error exists — a few metres becomes 25 or more.

Neither is a bug you can update your way out of. It’s physics, and it affects every phone, every app, every brand of car nav equally.

What actually helps

  • Mount the phone upright with a clear view of the sky. A phone lying flat in a metal cradle can degrade to 30–50 metres even in the open. Upright by the windscreen, it sees satellites.
  • In a tight wadi, trust the track, not the dot. Your recorded breadcrumb trail averages away the jumps; the live dot doesn’t. If the track shape says you followed the wadi and the dot says you are in the cliff, believe the track.
  • If you’re buying a phone anyway, get dual-frequency. L5 reception resists multipath — the exact wadi failure. Real-world gains are smaller than the marketing centimetres, but they show up precisely where you need them.
  • Know when to stop trusting screens. In the worst gorge, the honest answer is the old one: read the ground, remember your heading, and never navigate alone.

What we do about it in the app

Since we build a navigation app for exactly this terrain, here is our honest engineering answer to the same physics: we request the highest navigation-grade accuracy the phone exposes and fuse motion sensors with GPS; we reject impossible jumps (a fix implying 200 km/h in a wadi is a multipath spike, and we drop it instead of drawing it into your track); and we show a GPS-quality indicator so a wobbling dot reads as the canyon, not a bug. The honest limit: no phone app is survey-grade. The win is offline maps that keep working, spot data, and a track you can trust when the dot lies.

Accuracy figures on this page are typical values, not guarantees — terrain, phone model, mounting and sky conditions all move them. Nothing here asserts that any navigation setup is safe on its own: screens fail, batteries die, canyons lie, and the navigation rule that has never drifted is a second vehicle and someone who knows where you went.