How accurate is the GPS Clock?

Question

In my company we do have critical systems that require an accurate time.

As so, we have an NTP server appliance with an outdoor GPS antenna that receives the time from the GPS satellites.

My questions are:

• How accurate is the time clock?
• Is it worth it to keep this way or use another external NTP (US-GOV, NASA, etc) ?

Thanks,

Answer 1

According to this reference, the time should be accurate to within 40 ns, which is much less than the time it would take to transmit that information to another system.

The reference also explains how GPS time might differ from UTC because of leap seconds that have been added since the deployment of GPS. The difference between the two is encoded in the GPS signal, but it is up to the GPS device to include the offset in its displayed time. Presumably an appliance that is dedicated to keeping time would do this part correctly.

Answer 2

We have to be very careful how we understand this time accuracy issue. Light travels about 1m in 3 nanoseconds. Hence if we want to accurately locate ourselves to around 1m, then we are going to require a receiver that can at least determine time to ~ 3 nanoseconds or better. Low cost GPS receivers are not that accurate however (typically ~30nsecs), but they take advantage of a trick. The GPS satellites are calibrated to send highly accurate time and location measurements....a LOT better than 30 nsecs and they are constantly monitored and corrected to maintain the accuracy. But even though the GPS receiver is measuring 4-9 different signals with poor accuracy, it is the SAME receiver measuring the signals from different satellites. Hence if there is a built in error, the measurements can be collected and adjusted so that it is essentially the same offset error on the receiver for each satellite (not quite that simple, but thats the idea) . Hence if the receiver can see enough satellites it can perform Time Difference of arrival measurements which can remove the large error. The system can then use metadata to estimate and correct for the single static offset. You may have noticed in your car, if you turn on the GPS system at a location that your nav system was not actively tracking, and you are for example in a parking garage, the system will take some time to acquire ANY signal, and then it will often show you at the wrong location on the map, and will take a few minutes to figure out where you are. The system will be constantly checking different solutions and ultimately comparing them with map and inertial navigation detectors in your vehicle until it eventually finds a solution that appears to fit the map and your vehicles motion. My car has a GPS system. However, I once had the car taken by rail from DC to LA and it took almost 30 minutes for my GPS system to figure things out.

Answer 3

Most likely extremely precise (assuming you have a decent outdoor GPS antenna). GPS distance calculations involved signals moving close to the speed of light, so the clock has to be extremely precise or you will appear to be many feet from your actual location (approximately one foot error for every nanosecond off). As far as accuracy, you would probably need to check that time against whatever time standard you want to be accurate against. But because of the precision, the amount of accuracy should not change over time (unless your time standard you're comparing to is not precise).

Answer 4

I found a good writeup here....

the answer is: very, very accurate.