The integrated Nav systems in cars typically use dead reckoning via wheels sensors, etc to supplement the GPS accuracy. Plus with the map data, they “know” where the roads should be.
I have a friend working in GPS navigation right now and I’ll see if I can get him to chime in on his thoughts. He seems a lot more pessimistic about state-of-the-art than most of the posters here but that’s possible because he’s using them in adverse conditions (mainly urban high rise enviroments).
From what I gather, the absolute most accurate super dooper system you can get is carrier wave GPS which is supposedly centimeter-accurate but costs in the order of $100,000. Next best is DGPS which relies on ground based correction systems and can do a few meters. Ordinary GPS is in the order or tens of meters, even for some of the best units.
(Note: this is talking about typical response, “best” response is usually much lower but is useless as anything but a marketing figure).
Well I understand what you are saying, but it would seem if the GPS needs the sat signal to get it’s postion and that signal contains atomic clock info, if it doesn’t get that atomic clock signal, it doesn’t get it’s position signal either, so it can’t really wander at all with any ability to track. Perhaps the atomic clock signal is not sent every time.
A good example, but if you only need the time when your computer gets the time from the network, your computer’s internal clock doesn’t really matter.
Each GPS satellite broadcasts a signal that contains the current time on board the spacecraft and its current position in space. The unknowns are the location of the GPS receiver and the time at that location. A rough estimate of the local time can be obtained by setting the local time equal to that from one of the GPS satellites, plus a fudge factor. Now we can start measuring the propagation delay from each of the GPS satellites to the GPS receiver. Using the propagation delay measurements, we can plot a sphere in 3d space for each satellite, with the radius determined by the propagation delay times the speed of light. Let’s assume that we are receiving signals from four satellites. If our propagation delay measurements were perfect, all four spheres would intersect at a single point, the position of the GPS receiver. The problem is that we know that our local clock isn’t accurate, which means that our propagation delay measurements are also in error. The four spheres do not intersect at a single point. We can solve this problem by iteratively adjusting the local clock’s time, recomputing the propagation delays, reploting the spheres, and measuring how close we are to having the four spheres intersect at a single point. When the local clock has been adjusted to a value that produces a perfect, or near-perfect, solution, we know that the local time is accurate and we have an accurate position fix.
The next problem is that our GPS receiver does not have a high-quality clock, like the atomic clocks used in the GPS satellites. It’s probably a cheap quartz crystal oscillator with a substantial error in its frequency and a tendency to change frequency as the temperature changes. That means that even though we set it accurately by measuring the propagation delays to the GPS satellites and doing some math, it is going to run a bit slow or fast. As the error accumulates, our propagation delay measurements become less accurate, and the quality of our position fix suffers. At some point, we have to reset our local clock by recomputing the local time from the current GPS data. This problem can be alleviated by using the clocks in the GPS satellites to measure the frequency of our clock oscillator. If we have a 1 MHz oscillator, and we measure its actual frequency to be 999,990 Hz, we can correct for the 10 Hz (10 PPM) error in the clock rate in the GPS receiver’s software. By periodically comparing the local clock to GPS time, we can characterize the behavior of the local clock’s oscillator and compensate for its gross errors. Other errors are harder to deal with. The oscillator’s frequency isn’t constant. It varies over short periods of time, sometimes a bit slower and sometimes a bit faster than its average frequency. It is also affected by temperature, physical orientation, age, instabilities in the values of other electronic components, and changes in power supply voltage. These errors place limits on the accuracy of the local clock and the quality of the navigation solutions provided by the GPS receiver.
Even if the local clock was perfect, we would still have errors introduced by multipath and atmospheric effects that are hard to predict, such as weather and activity in the ionosphere.