Joe Clark: joeclark.org (E-mail)

You are here: Homepage > Technology > GPS explained [sidebar]


How GPS works

[Sidebar to "What's the Good of GPS? | Originally published in 1996 |
Updated here 1999.06.28]

The Global Positioning System relies on a network of 24 U.S. satellites, each of which completes an orbit of the earth in 12 hours. While all 24 satellites transmit signals on the same two frequencies, those signals are designed so that each satellite produces a uniquely identifiable message. Just as spies synchronize watches, the 24 GPS satellites sport very precise on-board clocks that are in more or less perfect synchony at all times.

GPS satellites work on a simple principle: If you know the speed of a signal, like a sound or light wave, you can easily calculate the distance of a target by timing how long that signal takes to reach you. In GPS, the satellites broadcast a signal consisting of a number of parameters, including a satellite identifier and a ranging code derived from a reading from the satellite's precise clock. Your own GPS receiver produces a facsimile of that same signal in real time, then receives the actual signal from the satellite and compares the two. (Remember, the 24 satellites are all in known orbits, so a GPS receiver can make a fair estimation of where a satellite should be at any moment and what its ranging code might be.) All a GPS unit needs to do in order to figure out the distance between the satellite in question and itself is to compare the ranging code it produced against the code from the satellite.

Receiving that information from only one satellite tells you merely how far away that satellite is, not where you are. To do that, you need readings from multiple satellites-- ideally four to get a precise fix. (With four satellites, a GPS receiver can solve for north/south and east/west position along with elevation and time.)

Even then, there are a few catches. The U.S. military has access to high-accuracy versions of the satellites' signals. For everyone else, a system euphemistically called selective availability "dithers" the satellites' clock signals. In effect, everyone other the U.S. military and its allies works with incorrect information. (President Clinton ordered that selective availability be eliminated within four to 10 years, but experts in the field wonder if that will actually happen [see 1999 article and New Scientist article.) And the satellite signals can be corrupted for other reasons-- mainly perturbations in satellite orbits and interference from the atmosphere.

A typical hand-held GPS receiver can pinpoint your location to an imaginary ellipsoid (a three-dimensional shape whose cross-section is an ellipse) with a horizontal radius of 100 metres and a vertical radius of 156 metres. (Military users can achieve accuracy in the 22-to-28-metre range.) You can improve that reading through various means, the most-widely-used of which is differential correction: Certain fixed waypoints (like radio stations) are located with extreme accuracy by means other than GPS. Then a GPS reading of those waypoints' locations is taken. The difference between the two readings amounts to the error in the GPS system.

In some differential systems, this error reading is continuously broadcast by radio (in Canada, over a CBC-FM subcarrier); if you spring for the differential option available on most GPS receivers, your machine can resolve down to 1 to 10 metres or better in a matter of seconds. Capitalism comes into play here: You have to pay more for one-metre resolution even though differential data typically are transmitted at that level of resolution-- it's selective availability all over again. If one metre isn't good enough, you can store your GPS data and postprocess it using various means to resolve to one centimetre. And there are other technologies to improve the accuracy of readings.

As a final wrinkle, all GPS measurements are expressed in a statistical confidence interval-- i.e., you can expect 100-metre accuracy 95% of the time. The other 5% of the time your readings will be quite wrong, though usually the errors are so huge they are obvious.

[With files from Elliott D. Kaplan, Mitre Corp., and Mark Corey, Natural Resources Canada.]