GPS time accuracy vs. NIST broadcast sync

[Rubycon]

If one eliminates the possible drift/errors of standard 32.768kHz quartz "offline" timekeeping...

Is time sync from GPS any better than NIST/NBS transmitters? The ones in USA, Europe and Japan are frequently out of range for me. GPS seems accurate but I don't have a precise way to see any differences, etc.

 

[Ross Ridge]

GPS needs to have very accurate time otherwise it wouldn't work, "an error of one microsecond (0.000 001 second) corresponds to an error of 300 metres (980 ft)" according to Wikipedia. I'm guessing even it's not as accurate NIST/NBS transmitters, it's more than accurate enough for normal computing usage.

 

[Throckmorton]

http://www.nist.gov/pml/div688/utcnist.cfm

What is GPS time?

The Global Positioning System (GPS) is a constellation of satellites each carrying multiple atomic clocks. The time on each satellite is derived by steering the on-board atomic clocks to the time scale at the GPS Master Control Station, which is monitored and compared to UTC(USNO). Since GPS time does not adjust for leap seconds, it is ahead of UTC(USNO) by the integer number of leap seconds that have occurred since January 6, 1980 plus or minus a small number of nanoseconds. However, the time offset from UTC is contained in the GPS broadcast message and is usually applied automatically by GPS receivers.

 

[Mark R]

GPS is markedly better than terrestrial band radios (several orders of magnitude better precision), because the GPS receiver necessarily performs correction for signal transit time, and also because the GPS carrier frequency and chip rates are much higher than for the NIST signal (allowing phase detection to yield a more precise time). The master clock for GPS is at least as good as NIST's master clock. However, it is an independent clock, and while both NIST and GPS time are periodically adjusted to bring them towards UTC as measured in Paris, they do drift a little (up to about 20 ns). UTC itself is calculated as an average from hundreds of independent atomic clocks in dozens of countries (including both NIST and GPS time), and used to tune a master clock in Paris, giving a single universal time measure of the highest possible accuracy.

If you require a precise frequency reference, then GPS is extremely good. A low-cost GPS time reference module (<$100) will offer long-term frequency stability at least as good as a standard-grade cesium reference, with phase noise of less than 30 ns (cheap boards tend latch the pulse-per-second signal to the module CPU clock generator - although if you need better precision than this, most of these modules will transmit a digital error signal over a serial connection stating how many ns early/late the PPS signal was). Better boards with fully asynchronous PPS generators can have timing precision in the 2-3 ns range.

It has always been assumed that this precision is absolute with GPS. However, it has never been tested until recently; the recent use of GPS timing to support high-energy physics experiments has revealed some discrepant timing results. Some researchers have identified a possible error in the GPS time/position solution algorithm which may introduce errors of several 10s of ns between receivers at different positions on earth.

On a practical note, you do need to be very clear what time measure your GPS actually emits. Many report GPS time for reasons of simplicity. GPS time uses a simple rule of 60 seconds in every minute. By contrast, UTC may occasionally have 59 or 61 second minutes (in order to remain in sync with mean solar time, while keeping the duration of a second constant) - the missing or extra seconds know as "negative" or "positive" leap seconds. As a result there is an offset between GPS time and UTC (which is currently 15 seconds) due to accumulated leap seconds.

Most GPS receivers are configurable to emit either UTC or GPS time. If using UTC mode, you should check to make sure that the equipment receiving time data from the receiver will not malfunction in the event of leap seconds.

 

[Lemon law]

I look at it from the practical application viewpoint.

Its one thing to say, a not too expensive GPS system will resolve your cars position on any highway in the USA, or the world for that matter, with in about 10 meters accuracy. Making GPS turn by turn car navigators into a already practical reality.

But if we are going to take the next step into totally computer automated control driverless cars system that will place your car into a single lane and properly space any traffic volume of other cars to avoid accidents, GPS accuracy would have to increase to plus minus six inches or so.

 

[Rubycon]

GPS is markedly better than terrestrial band radios (several orders of magnitude better precision), because the GPS receiver necessarily performs correction for signal transit time, and also because the GPS carrier frequency and chip rates are much higher than for the NIST signal (allowing phase detection to yield a more precise time). The master clock for GPS is at least as good as NIST's master clock. However, it is an independent clock, and while both NIST and GPS time are periodically adjusted to bring them towards UTC as measured in Paris, they do drift a little (up to about 20 ns). UTC itself is calculated as an average from hundreds of independent atomic clocks in dozens of countries (including both NIST and GPS time), and used to tune a master clock in Paris, giving a single universal time measure of the highest possible accuracy.

If you require a precise frequency reference, then GPS is extremely good. A low-cost GPS time reference module (<$100) will offer long-term frequency stability at least as good as a standard-grade cesium reference, with phase noise of less than 30 ns (cheap boards tend latch the pulse-per-second signal to the module CPU clock generator - although if you need better precision than this, most of these modules will transmit a digital error signal over a serial connection stating how many ns early/late the PPS signal was). Better boards with fully asynchronous PPS generators can have timing precision in the 2-3 ns range.

It has always been assumed that this precision is absolute with GPS. However, it has never been tested until recently; the recent use of GPS timing to support high-energy physics experiments has revealed some discrepant timing results. Some researchers have identified a possible error in the GPS time/position solution algorithm which may introduce errors of several 10s of ns between receivers at different positions on earth.

On a practical note, you do need to be very clear what time measure your GPS actually emits. Many report GPS time for reasons of simplicity. GPS time uses a simple rule of 60 seconds in every minute. By contrast, UTC may occasionally have 59 or 61 second minutes (in order to remain in sync with mean solar time, while keeping the duration of a second constant) - the missing or extra seconds know as "negative" or "positive" leap seconds. As a result there is an offset between GPS time and UTC (which is currently 15 seconds) due to accumulated leap seconds.

Most GPS receivers are configurable to emit either UTC or GPS time. If using UTC mode, you should check to make sure that the equipment receiving time data from the receiver will not malfunction in the event of leap seconds.

Yes we use Trimble cards and the modules have UTC leap second compensation. The P-GPS output is accurate to approximately 35 centimeters on the globe.

Local time reference is handled by a hydrogen maser clock.

Thanks for that explanation - that clears up some mixed info I've been receiving.

 

[wirednuts]

I look at it from the practical application viewpoint.

Its one thing to say, a not too expensive GPS system will resolve your cars position on any highway in the USA, or the world for that matter, with in about 10 meters accuracy. Making GPS turn by turn car navigators into a already practical reality.

But if we are going to take the next step into totally computer automated control driverless cars system that will place your car into a single lane and properly space any traffic volume of other cars to avoid accidents, GPS accuracy would have to increase to plus minus six inches or so.

i dont think autodrive cars will rely on gps to drive. that would be pretty dramatic when the times that the gps system might be down or not working because someone blew a satellite or two up. could you imagine our entire highway system coming to a halt at once? cloud driving is not recommended.

i think theyll just use gps like we do now. just tells us how to get there. however... i could see gps being used to determine what parking spots are legal and whatnot (that way you dont need special sensors at every potential parking space... just upload the coordinates to the internet and the car can get those locations from a national database). and if thats true, then you are probably right we will need gps to be accurate to less then a foot.

otherwise, i it will be a mix of sensors and incredibly powerful computing that will make autodrive cars a reality. optical cameras, IR cameras, and laser imagers have already proven to be a successful combination to give autodrive cars enough accuracy to stay within a foot or so. this of course will only get better, when sensors get more sensitive for lower cost, when computer processors get about 10x faster then they are now, and also when they implement a secondary safety system that will be a wireless network between the cars themselves. so the cars can see and act independantly with any other object, but it can also talk to other cars to give confirmation on what the eachother is trying to do (like humans do today with hand signals to other cars).