GPS-derived time baffles NOAA researcher
Recently, I discovered that time derived from GPS satellites was not quite in step with other time sources. I am presently doing research at the Climate Monitoring and Diagnostics Lab near Pt. Barrow, Alaska (located at 71° 19' N, 156° 37' W). This facility is maintained by the National Oceanographic and Atmospheric Administration and is part of the network of sites designed to monitor meteorological conditions and background atmospheric constituents for detecting changes in worldwide pollution levels and monitoring the effect of these changes on global climate.
Measuring time is a major concern for me. I need accurate time to calculate solar position for several projects. The problem of measuring and recording time may seem trivial, but in Barrow it is a continual source of frustration. Accurate clocks are available but must be synchronized to a time standard and must occasionally be checked against these standards. Delays in telephone transmissions of time signals can run as high as .25 to .5 seconds. The time tick broadcast from WWV in Ft. Collins, Colo., is not always clear, and, on many occasions, atmospheric noise completely masks the signal. An attempt to find an acceptable alternative led to the investigation of the time signal broadcasts by GPS satellites. Good handheld units can be bought for a relatively modest price from an ever-growing number of manufacturers, so I decided this was the best option for acquiring a time signal.
After acquiring the GPS receivers, I began to question whether there was any appreciable clock drift in these units. For this reason, I did a check against WWV. I found a discrepancy of up to 1.5 seconds between the time readout of a Garmin GPS-50 and time as broadcast by WWV. Upon discussion with several local GPS users it was decided that a comparison was in order to check the accuracy of the different brands in use.
Since the WWV shortwave signal is often unobtainable, an alternate standard was needed. A True Time model 468-DC GOES synchronized clock was used. The 468-DC displays a time signal it receives from a Geosynchronous Operational Environmental Satellite (GOES). It is advertised as providing time, traceable to UTC, of better than 250 microseconds (since the signal travels 23,000 miles from geosynchronous orbit, this unit corrects for propagation delay). Due to the longitude of Barrow, 156° W, the GOES west satellite was used.
The instruments in this comparison were a Garmin GPS-50, a Motorola Traxar, a Trimble TransPak II, and a Magellan 5000. Time from the Garmin, Motorola, and Trimble was examined on three occasions without regard to time of day. Morning fixes were not distinguished from afternoon or evening fixes. The Magellan was compared to the 468-DC only once.
A known survey marker is located near the observatory but position was not in question at this time. Position fixes were not compared. The only question was the difference in time from each of the handheld units compared to the GOES receiver.
After several attempts to compare time readouts, the following results were obtained. The Garmin, Trimble, and Magellan would vary between .5 and 1.5 seconds slow. All three units would start out displaying time very close to the 468-DC, and within 15 minutes each would drift to and settle at between 1 and 1.5 seconds slow. The Motorola displayed time closer to the 468-DC than any of the other units. It would start out very close to time as displayed by the 468-DC and drift to about .5 seconds slow. The Motorola was never seen to be off by more than one second and was usually within .5 seconds.
The results were surprising because the reasons for the differences between the handheld GPS units and the GOES receiver were unknown. At first, the difference was thought to be in the handheld units. A call was placed to three of the manufacturers and each gave the same answer when asked about the difference. We were told that time as displayed by the GPS should be exact because that is what the satellites broadcast.
The next thought was that perhaps the GPS receivers were showing GMT time as opposed to the UTC time derived from GOES. This brought up the somewhat complex issue of time, which can be described in many different ways.
According to Bowditch, there are three different kinds of time. These are known as Universal Time (UT), Ephemeris Time (ET), and Atomic Time (AT). It is confusing that all three time systems use the word "second" as the basic unit of time even though each defines the second differently.
UT is further divided into UT0, UT1, and UT2. UT0 is determined by direct astronomical observations. A more uniform time, UT1, is found by correcting for polar motion (a slight wandering of Earth's rotational poles). According to Bowditch this is 15 milliseconds at Washington, D.C. UT1 is the same as GMT and is used in celestial navigation. UT2 is UT1 corrected for variation of rotation due to wind, tides, etc. This correction can be as much as 30 milliseconds.
ET is a uniform time and is based on observations of the sun, moon, and planets. It is used by astronomers but not by navigators.
The third kind of time is AT, and it is based on the rate at which electrons crossover between two energy states in the Cesium 133 atom.
For scientists, simply knowing the length of a second is not enough. A way of tying an event to a particular time frame, such as the Gregorian calendar, is needed. A time scale is the system of assigning a date to an event. A1 is the time scale based on the use of AT as a clock. Time, as found by A1 and UT2, was set to be equal on January 1, 1958. Coordinated Universal Time (UTC) is time as told by A1 on a variable scale. This scale is made to agree with UT2 by adjusting the frequency of the time supplied by the atomic clocks.
In 1972 a new system was begun where, instead of adjusting time through the use of a variable frequency, a leap second would be inserted in UTC. UTC was made to agree with UT1 by the use of the now familiar leap second. Agreement between UT1 (GMT) and UTC is kept within about .9s. This is the variability that was thought to be the cause of difference in time between the GPS units and the GOES receiver.
We placed another series of phone calls to the manufacturers and were assured that the GPS units show UTC time and there was no known reason why there should be any difference. One call found us talking to a technician who finally cleared up the entire problem. We were told that yes indeed it is a problem in the software and that all the manufacturers were aware of it but not sure how to fix it, or if it was even worth fixing because of the small error. This is another caveat when using GPS. All shipboard personnel should be aware of this difference when a high degree of accuracy is needed. They should not rely on the GPS time signal without checking it against a known standard.
Something that seems as simple as measuring time can quickly turn into a complex issue. It is not enough to know what we are measuring; we also need to know how we are measuring it. For most applications knowing time to within a minute or so is more than accurate enough. For higher precision, it seems that a price is paid in confusion and complexity.
Dan Endres is a physicist and is the officer in charge of the Barrow Climate Monitoring and Diagnostics Lab. He enjoys boating in the Beaufort Sea (summer only).
Editor's note: Mention of a commercial product does not constitute an endorsement by NOAA. Use for publicity or advertising purposes of information from this article concerning products or the tests of such products is not authorized.