How accurate is a GPS speedometer?

Put a phone with a GPS speedometer next to your car's dashboard and the two rarely agree exactly. Usually the car shows a little more. That is not because the phone is guessing: in most cases the GPS figure is the closer of the two, and the car is reading high on purpose. This guide explains how a phone measures speed, how close it gets, and when not to trust it.

How a phone measures speed

A GPS receiver locks on to signals from several satellites at once. To find where it is, it times how long each signal took to arrive. To find how fast it is going, it measures something else: the Doppler shift. Moving towards a satellite raises the frequency of its signal very slightly; moving away lowers it. Comparing those shifts across satellites gives the receiver its velocity directly.

That matters because it does not depend on two positions being exact. Working speed out from position — distance between two fixes divided by the time between them — turns a few metres of position error into a jumpy speed. A Doppler measurement does not have that problem, which is why a phone's speed reading is usually steadier than its position.

Browsers pass the device's measured speed to web pages through the Geolocation API, in metres per second, or as nothing at all if the device cannot measure it. The Speedometer.top speedometer uses that reading whenever there is one, and only falls back to working speed out from positions when there is not — which is common on laptops that find themselves by Wi-Fi rather than GPS.

How accurate is it?

For position, GPS.gov says GPS-enabled smartphones are typically accurate to within about 4.9 metres (16 feet) under open sky, and that accuracy gets worse near buildings, bridges and trees. The speedometer shows the accuracy your phone reports for each fix in the chip above the dial.

Speed generally holds up better than that, for the reason above. In practice, at a steady speed on an open road, a phone is accurate enough to check a car's speedometer against. Two things limit it:

• Update rate. Phones typically report a new fix about once a second, so under hard acceleration or braking the reading trails what is happening by a moment.
• Signal quality. Reflections off buildings and blocked satellites degrade both position and speed. When the chip says the signal is weak, treat the number as a rough guide.

One more subtlety: the Geolocation specification defines speed as the horizontal part of your velocity. On a steep road that is slightly less than your speed along the slope, but the difference is tiny — about 0.5% on a 10% gradient.

Why your car shows a higher speed than GPS

Car speedometers are built to err on one side. UN Regulation No. 39, which governs speedometers on new vehicles in the EU, the UK and the other countries that apply it, says the speed shown must never be less than the true speed, and may be higher by up to10% of the true speed plus 4 km/h. Manufacturers aim inside that window, so most cars over-read a little: at a true 100 km/h, anything up to 114 km/h is allowed. The car speed test tabulates the allowed range at other speeds, in km/h and mph.

Tyres add to it. A car's speedometer counts wheel turns, so anything that changes the tyre's size changes the reading. Worn tyres are slightly smaller and turn faster for the same road speed, pushing the reading up a little more. Larger-than-standard tyres do the opposite, and can make a speedometer read low. GPS measures speed over the ground, so tyres make no difference to it.

When GPS speed is less reliable

• Among tall buildings. Signals bounce off glass and concrete before reaching the phone, which can briefly put you somewhere you are not.
• Tunnels, car parks and indoors. No signal at all. The speedometer greys the number out and says so, and picks up again on its own.
• Just after the signal returns. The first fix or two out of a tunnel can be off while the receiver settles.
• Under heavy tree cover or in deep valleys, where part of the sky is blocked.
• On devices without GPS. Most laptops estimate their location from Wi-Fi and cannot measure speed at all.

How the speedometer handles bad readings

The speed you see is your phone's own measured speed, shown as it arrives with no extra smoothing, because smoothing would make it trail behind when you speed up or brake. The rest of the work goes into keeping your trip numbers honest:

• Accuracy cut-off. Fixes with a reported accuracy worse than 30 m are shown, but they never count towards top speed, average speed or distance.
• Impossible jumps. A reading that would need more than 15 m/s² of acceleration — about 1.5 g, beyond a road car — is dropped.
• A confirmed top speed. Top speed is the middle value of 3 readings in a row, so one stray reading cannot set it.
• Distance from speed, not positions. Distance adds up speed multiplied by time. Joining up positions instead would count the few metres each fix wanders as distance travelled, which inflates short, slow trips most.
• Standing still. Anything under 1.8 km/h reads as 0, so GPS drift while parked does not add distance.

If your device does not report a speed at all, the speedometer works it out by fitting a line through your positions over the last few seconds rather than comparing just two, which averages out much of the position error. It is still rougher than a measured speed, and the page says when it is doing this.

Checking your own car

Everything above comes down to one practical test: compare your dashboard with GPS on an open road at a steady speed. The car speed test walks through it and has a check that does the arithmetic: enter both readings and it says whether the gap is within what R39 allows.