INNOVATION ❯❯❯ TIMING & SYNCHRONIZATION

Time and Frequency Dissemination

Advances in GPS Transfer Techniques
Thomas E. Parker and Demetrios Matsakis

The atomic clock is one of the greatest inven-

tions of the 20th century. When the first com-
mercial atomic clocks based on the resonance navigation, GPS also has had a huge im- and time, International Atomic Time (TAI)
of the cesium atom were introduced in 1958, pact on the use of precise time. and UTC, which is equal in rate to TAI,
available clock accuracy jumped by orders of Each GPS satellite contains several but adjusted by an integer number of sec-
magnitude.The new clocks would not gain or atomic clocks and continually broadcasts onds to account for variations in the ro-
lose one second in a thousand years. Scientists the time and its position. A GPS receiver tation of the Earth.
used laboratory versions of the cesium atomic tracking at least four GPS satellites can Most of the contributing laboratories
clock to define the atomic second and to estab- solve for the receiver’s unknown position produce real-time realizations of UTC.
lish and maintain a time scale to which all and time at virtually any location on the The U.S. Naval Observatory (USNO) and
clocks could be set. But how do you set a globe, with a precision of a few meters and the National Institute of Standards and
clock, particularly one which is accurate to a a time error of a few tens of nanoseconds Technology (NIST) produce official real-
microsecond or better? Or once set, how do (ns), excluding receiver calibration errors. time realizations of UTC for the United
you keep track of its error? Clocks are moni- A timing user operating from a known States. These time scales are identified as
tored and synchronized using a time and fre- fixed location can derive time from GPS UTC(USNO) and UTC(NIST).
quency dissemination technique. In the past, using just one satellite, and with averag- UTC(USNO) is the external time refer-
radio signals and traveling clocks were used ing, a timing accuracy of a few nanosec- ence for GPS, and consequently the UTC
for this purpose. But GPS, itself dependent on onds is possible. time derived from GPS is considered fully
atomic clocks for its operation, has become the Time from GPS is now used for many traceable to international standards for civil
best technique for intercomparing clocks and civilian purposes, including synchroniza- and legal time, subject to user equipment
for helping to maintain the world’s time sys- tion of communications systems, cell phone errors.
tems. In this month’s column, scientists from networks, and power grids. It also is used Recent years have seen a number of im-
the United States’ two national time-keeping for many commercial applications where provements in techniques and equipment
laboratories, the National Institute of Standards accurate time tagging is becoming in- that have resulted in improved time trans-
and Technology and the U.S. Naval Observatory, creasingly important. fer. This article reviews several advances.
discuss the past, present, and future use of The international timing community
GPS for accurately disseminating time and fre- utilizes GPS to help produce the world’s Time Directly from GPS
quency. – R.B.L. coordinated atomic time scale, Coordi- The simplest and most-direct way to ob-
nated Universal Time (UTC). In some in- tain time from GPS is to use the signals
stances, GPS serves as the primary trans- broadcast directly by the GPS satellites as

G
PS is not only a high accuracy fer tool, while in others it serves as a backup illustrated in Figure 1.
navigation system, it also de- to two-way satellite time and frequency Authorized GPS users, who have access
livers time with unprece- transfer (TWSTFT). to the encrypted P-code on both GPS fre-
dented accuracy. Free to the The International Bureau of Weights quencies (L1 at 1575.42 MHz and L2 at
user, GPS is the world’s most-accurate and Measures, or Bureau International des 1227.60 MHz) can directly estimate the
globally available one-way source of time Poids et Mesures (BIPM), in France is ionospheric delay of the signals, which al-
and frequency. Declaration of full opera- charged with providing the time standard lows for a more-accurate position, veloc-
tional capability of GPS in 1995 revolu- UTC. The BIPM collects data via GPS or ity, and timing (PVT) solution. Most civil-
tionized timing as well as navigation. TWSTFT from more than 200 atomic ians can utilize only the C/A-code on L1,
Dual use (by military and civilians) of clocks and a few primary “absolute” fre- but in early 2005 GPS will introduce its
GPS was announced as a formal goal in the quency standards from more than 50 in- first IIR-M satellite, which will broadcast
wake of the 1983 downing of Korean Air- stitutions around the world. Once a month, a new unencrypted civil L2 signal, and in
lines Flight 007, and civilian users now far BIPM uses these data to produce the stan- 2006 GPS will add a third civil signal, L5.
outnumber military users. In addition to dard international references for frequency There also are several commercial code-
32 GPS World NOVEMBER 2004 www.gpsworld.com
 15
Satellite S UTC(USNO) minus GPS Time; r.m.s.=3.2 ns
10

Difference (ns)
5

-5

-10
dSA
-15
52650 52700 52750 52800 52850 52900 52950 53000

Io
Tro n 15
p
os pher
ph
os

UTC(USNO) minus GPS-delivered prediction of UTC(USNO); r.m.s.=1.1 ns

ere

10
e

Difference (ns)

5
Clock A
0

-5

-10

¶ FIGURE 1 GPS direct broadcast (one-way) -15

52650 52700 52750 52800 52850 52900 52950 53000
time transfer. The electrical path length MJD
between the satellite and the clock's GPS ¶ FIGURE 2 Daily averages of the navigation time scale, GPS Time, and the prediction of
receiver, dSA, includes contributions from the UTC(USNO) during 2003. The horizontal axis, Modified Julian Date (MJD), is the number of days
ionosphere and troposphere (not to scale). elapsed since November 17, 1858.

less and semicodeless GPS timing receivers generation system, GPS III, currently is continuing use of GPS Selective Availabil-
available to all users that can derive timing envisioned to have crosslinks enabling the ity (SA) within a decade. SA was a deliber-
information from L1 and L2. entire constellation to be uploaded at once, ate degradation of the GPS signal intended
Each GPS satellite includes in its navi- it has been argued that stronger steering to reduce the navigational accuracy available
gation message a prediction of its orbit, an is justified. In any event, the intention has to so-called unauthorized users. When SA
estimate of the difference between its in- always been that time would be derived by was set to zero in May 2000, the timing ac-
ternal clock and GPS (System) Time, and first computing the user position and time curacy from GPS improved significantly for
an estimate of the difference between GPS offset relative to GPS Time, and then ap- civilian users. A commercial product that has
Time and UTC(USNO). These predic- plying the UTC correction terms, includ- benefited greatly from this new capability is
tions are updated on the ground every 15 ing the integer seconds offset. Figure 2 the GPS disciplined oscillator (GPSDO).
minutes by the GPS Master Control Sta- shows performance of both time scales. The GPSDO is a combination of a good in-
tion Kalman filter, but each satellite’s The daily average of GPS broadcast es- ternal clock (quartz or atomic) and a multi-
broadcast information is updated approx- timates of UTC(USNO) generally are ac- channel GPS receiver. The phase (or time)
imately once per day by the Control Seg- curate to a few nanoseconds root-mean- of the internal clock is locked to the time of
ment. square (r.m.s.) and UTC(USNO) has been the broadcast GPS signal. The GPSDO may
Gentle Steering. GPS Time is the in- within 3 ns r.m.s. of UTC for the past two operate from a known position or use GPS
ternal GPS navigation time scale, which is years. Thus the UTC time obtained from to derive its position dynamically.
not adjusted for leap seconds, and is very GPS has been accurate to better than 5 ns Since the GPS signal is much noisier in
gently steered to UTC(USNO) modulo 1 r.m.s. In the future, there will be improved the short term than the internal clock, the
second. GPS Time is specified to be main- time delivery as new frequencies become signal usually is averaged for a relatively
tained to within one microsecond (µs) of generally available, more monitor stations long time. This can range from a few min-
UTC(USNO) (modulo integral seconds); are established (among them the USNO), utes to many hours, depending on the qual-
for the past seven years it has been main- and when GPS initiates crosslinks so that ity of the internal clock. The better the
tained to within 22 ns of this goal. One rea- all satellites can quickly refresh their broad- clock, the longer the averaging time, and
son advanced for such gentle steering is cast parameters in near synchrony. this generally leads to better overall per-
that it lessens the discontinuity when satel- Disciplined Oscillator. In March 1996, formance. GPSDOs with internal rubid-
lites are asynchronously uploaded with im- a Presidential Decision Directive was issued ium or quartz clocks are widely available,
proved clock information. Since the next- that outlined a path that would lead to dis- new or used, for less than U.S.$10,000.
www.gpsworld.com NOVEMBER 2004 GPS World 33
 TIMING & SYNCHR ONIZATION
❯❯❯INNOVATION

With some care in setup and measurement 50

of cable lengths, these devices will give time 40
accurate to within 50–100 ns.
30

Time difference (nanoseconds)

Figure 3 shows the time —
UTC(USNO) from GPS — delivered by 20
a quartz-oscillator-based GPSDO relative 10
to UTC(NIST) over a 10-day period. Dur-
0
ing this interval, UTC minus UTC(NIST)
was about 18 ns and UTC minus -10
UTC(USNO) was about -3 ns. The tim-
-20
ing output of the GPSDO varies over a
peak range of 20–30 ns due to noise in the -30
GPS signal and from other factors such as -40
temperature, ionosphere, troposphere,
-50
multipath, and position errors. Researchers 0 1 2 3 4 5 6 7 8 9 10
carefully measured all cable delays for this Time (days)
test, and the short-term time error over ¶ FIGURE 3 UTC(USNO) as delivered by GPS from a quartz-oscillator-based GPSDO relative to
the test period was within 6 50 ns at all UTC(NIST). Note the short-term variations due to transfer noise.
times. By averaging over 24 hours, one can
improve the accuracy to under 20 ns with
a precision (stability) of about 2 ns. A 24- clock corrections that have accumulated
Satellite S
hour average accuracy on the order of a since the satellites were last uploaded (age
few nanoseconds could be obtained with of data).
careful calibration of the receiver delay. An example of a direct broadcast cali-
The device delivers r.m.s. frequency accu- bration service is the NIST Frequency
-11
racy better than 1 3 10 for averaging Measurement and Analysis Service. It is
times greater than 1 second, and is better based on direct-broadcast GPS Time and
than 2 3 10-13 at an interval of 1 day. provides frequency traceability to NIST
Cable Delays. Virtually any GPSDO with an uncertainly of 2 3 10-13 over a day. dSA dSB
will deliver time to better than 1 µs if the
manufacturer’s instructions are followed. GPS Common-View
However, if care is taken in measuring GPS common-view is a variation on the di- Io
cable delays and choosing the antenna site rect broadcast of time from a GPS satellite, Tro n
p
os pher
ph
to be as free from multipath as possible, used primarily to compare or synchronize
os

most GPSDOs will deliver time accurate two widely-spaced clocks on the ground. ere
e

to within 100 ns over short time intervals, GPS common-view is carried out by hav-
Clock A Data Clock B
and considerably better if a 24-hour av- ing two ground stations observe the same
exchange
erage is used. Note that the user must have satellite at the same time, as illustrated in
a reliable local clock such as a high-qual- Figure 4. This technique has the advantage
ity commercial cesium standard if a 24- of removing the GPS satellite clock error,
hour average is to be used effectively. The including SA if present, and depending on
short-term stability (averaging time less the baseline length, reducing errors due to ¶ FIGURE 4 GPS common-view technique.
than 1,000 seconds) varies considerably atmosphere, broadcast orbit, and broadcast Measurements made of a commonly viewed
among products, with the best performance ionospheric corrections. GPS common- satellite at Clock A and Clock B are exchanged
to determine the relative behavior of the clocks.
generally being obtained with the more- view has helped compare clocks around the
costly, rubidium-based oscillators. world for years, and was particularly at-
Authorized users who have access to the tractive before SA was set to zero. Even years.
P(Y)-code on both L1 and L2 can achieve in the absence of SA, it can provide mod- In the days of single-channel receivers,
performance considerably better than estly better performance than time broad- a tracking schedule was required so that
L1 C/A-code-only receivers because their cast directly from the satellites for baselines both stations observed the same satellites
receivers can measure the ionospheric delay shorter than a few thousand kilometers. It at exactly the same time. With multi-chan-
correction directly. Codeless receivers also has been used by the international timing nel receivers, the tracking schedule gener-
offer some benefits. However, all users are community as a primary means to coordi- ally is not needed. However, in either case,
limited by errors in the broadcast orbit and nate time around the world for nearly 20 data must be exchanged between the two
34 GPS World NOVEMBER 2004 www.gpsworld.com
stations to calculate the time difference be- Time Service at NIST). Subscribers typi- by an unspecified time offset with every
tween the local clocks. With multi-chan- cally use the Internet or a phone service to receiver reset. However, receivers whose
nel receivers, the all-in-view data files sim- send their receivers’ observations to the na- output is insensitive to power cycling are
ply are compared to find common tional laboratory, which reduces the data to now available, and some others can be
observations. Common-view is not a real- compute the difference between the labo- modified by their manufacturers for a fee.
time system for time dissemination since ratory and subscriber clocks. Accurate time- Timing carrier-phase receivers typically
data must be transferred, but through the transfer comparisons can be made for users output data in the form of RINEX files,
use of the Internet it can be near real time. if their relative system delays are fully cal- which can be input directly into the soft-
In some respects common-view time ibrated. ware used for precise timekeeping.
transfer is easier to calibrate than direct ac- Through the codeless and semi-codeless
cess, since one does not need to know ab- GPS Carrier Phase dual-frequency GPS receivers, the user
solute delays, only differential delays, which It has long been recognized that observa- has the ability to measure the line-of-sight
can be measured with a traveling GPS re- tions of the GPS carrier frequency, in con- ionospheric delay of each observation. The
ceiver. Also, common-view has the advan- junction with the GPS spread-spectrum user potentially can combine data world-
tage for short baselines that some sources code, have the potential to provide ex- wide to solve for all relevant parameters
of error, such as unmodeled ionospheric tremely precise estimates of many para- in one grand solution.
delays and satellite position errors, will be meters. However, it took many years of Precise Point Solutions. However, it
at least partially common to both legs and development by groups such as the Jet often is faster to generate time or fre-
will partially cancel. Common-view carried Propulsion Laboratory (JPL) and the Uni- quency transfer solutions by applying pre-
out with the L1 C/A-code is about a factor versität Bern in Switzerland to develop viously determined measurements or pre-
of two better in both accuracy and stability software that optimally estimates quanti- dictions of satellite orbit, clock, and other
than direct-access time transfer with the L1 ties, including the inherent code-to-car- parameters — solving only for site-spe-
C/A-code for baselines shorter than 3,000 rier ambiguity, that allow use of the car- cific parameters, among them the site-
kilometers. rier-phase GPS measurement to more clock time offsets from the reference clock.
Baselines. Accuracies of a few nanosec- accurately estimate the GPS orbit and Such solutions are termed precise point
onds can be achieved with a 24-hour av- clock offsets. solutions. JPL, for example, offers a free
erage, whereas for shorter time intervals In 1992, a consortium of international Internet-based service entitled Auto
the accuracy will be limited by peak insta- geodetic institutions, universities, and gov- GIPSY (GPS-Inferred Positioning Sys-
bilities of 10–40 ns. The stability (or pre- ernment agencies formed a technical con- tem) that will provide precise point solu-
cision) with 1 day of averaging is about 1 federation, the International GPS Service tions using IGS final orbits.
ns for stations within the United States, and (IGS). The IGS produces high-accuracy
more like 2 ns over intercontinental base- estimates of GPS orbits and clock offsets,
lines. For baselines on the order of 10,000 with postprocessed 1-sigma accuracies bet-
kilometers, the stability of common-view ter than 5 centimeters. IGS data also yield Web Sites of Interest
and direct access are about the same. For an accurate time scale. Fundamentally, the
longer baselines and daily averaging times, comparative advantage of carrier-phase NIST Frequency Measurement
a time-transfer mode termed “melting pot,” observations is due to the fact that the and Analysis Service:
based upon differencing two averages of 600–800 picosecond period of the GPS http://tf.nist.gov/service/fms.htm
GPS Time minus the site clocks, is supe- carrier frequencies is 1,000 times smaller USNO Time Service Department:
rior to satellite-by-satellite common-view, than the 0.9766 µs chip length of a C/A- http://tycho.usno.navy.mil/
particularly if the orbital and satellite clock code bit, and 100 times smaller than the
BIPM Time Section:
corrections are applied to GPS broadcast GPS P-code, which greatly reduces mul-
http://www1.bipm.org/en/scientific/tai/
parameters. tipath sensitivity and measurement noise.
If processing delays of up to a week can However, the carrier phase has an inher- IGS Central Bureau:
be tolerated in the time comparison, im- ent wavelength ambiguity that must be re- http://igscb.jpl.nasa.gov/
proved performance can be achieved by solved in the data processing.
IGS Timing Activities:
using postprocessed ionospheric maps and The problems and principles involved
https://goby.nrl.navy.mil/IGStime/
precise orbits. Their use has the most im- in the reduction of raw GPS carrier phase
pact on long baselines and will reduce the data are largely understood at the nanosec- Auto GIPSY: http://milhouse.jpl.nasa.
instabilities of transatlantic common-view ond-level. A significant problem in the past gov/ag/agfaq.html
to about 1 ns at 1 day. was that many receiver designs did not PTTI Meetings:
A number of national laboratories pro- phase-lock their internal signals unam- http://tycho.usno.navy.mil/ptti.html
vide a time and frequency calibration ser- biguously to the input timing reference,
vice using common view (such as the Global and as a result measurements would jump
www.gpsworld.com NOVEMBER 2004 GPS World 35
 TIMING & SYNCHR ONIZATION
❯❯❯INNOVATION

Many institutions provide professional- divided by the sine of the satellite eleva- assumed site positions would lead to con-
level software capable of reducing carrier- tion angle, most of this can be removed by sistently incorrect ambiguity determina-
phase data, and some research groups have simple modeling and fitting techniques. tions in some parts of the sky and this
written their own. In the timekeeping Tests of differences between fitted tro- would affect the time measurements in
community, perhaps the best known are posphere models typically show post-fit continuously filtered solutions. However,
JPL’s GIPSY and Bernese (available from time-transfer changes of less than 10 ps. no such variations have been reported in
Astronomisches Institut Universität Bern). Discontinuities. Other forms of noise long-term comparisons between contin-
Reductions of the same data by different are the discontinuities at the boundaries uously filtered solutions and independent
institutions may vary and the differences between independent parameter solutions. day-boundary solutions. If such differences
are loosely labeled “analysis noise.” In the IGS time and frequency solutions, were observed, the optimal approach
Biases. As with any technique, when day-boundary discontinuities can be sev- would be to steer continuously filtered so-
corrections are not uniformly applied, sys- eral hundred picoseconds r.m.s., and are lutions to long-term averages of indepen-
tematic differences of a few nanoseconds mostly, if not entirely, due to the noise and dent daily solutions.
can be observed — just as they could bias in the GPS code data used to provide Another advantage of continuously fil-
occur with satellite-based biases, which the constant offset generating time trans- tered solutions is due to the temperature-
are used by the GPS receiver’s algorithm fer from frequency transfer. Although car- dependence of a system’s delay being
for measuring the L1-L2 difference. Even rier-phase time transfer is subject to the greater for code data than carrier-phase
in global solutions without such biases, same bias variations as common-view time data. Code data corrected for carrier-
time-transfer solutions may differ if the transfer, in many cases the carrier-phase phase-determined orbit, position, and
other receivers used in the solution are not technique provides some means to study clock frequency values can show up to sev-
identical. For a small array, adding sites this. eral nanoseconds of temperature-depen-
can shift time-transfer results at the sin- Important examples are the tempera- dent daily variations for systems that have
gle nanosecond level. Errors due to GPS ture-dependence and multipath of the exterior cables not temperature-compen-
satellite orbits can affect time-transfer de- GPS receiver systems. An advantage of sated. For data reduced in independent
terminations. using independent solutions from data daily solutions, the large systematic di-
For example, use of IGS rapid orbits in- parsed into one-day batches is that no cor- urnal variations in the code data result in
stead of IGS final orbits shifted time-trans- relations between fitted parameters can noticeable day-boundary discontinuities.
fer results by 200–300 ps r.m.s. for ar- grow with time. For example, if site posi- In the past several years, Natural Re-
rays of about 20 receivers. Although the tions (or any parameters that depended sources Canada and NASA/JPL have de-
entire tropospheric delay correction, for upon satellite azimuth and elevation angle) veloped real-time GPS networks that use
a near-sea-level site, is very roughly 7 ns were not solved for, then errors in the carrier phase to estimate all relevant GPS

Further Reading For information on carrier-phase GPS time transfer,see

For a comprehensive introduction to the physics of time and time mea-
“Time and Frequency Transfer: High Precision Using GPS Phase
surement,see: Measurements” by T. Schildknecht and G. Dudle in
The Measurement of Time: Time, Frequency and the Atomic Clock GPS World, Vol. 11, No. 2, February 2000, pp. 48-52.
by C. Audoin and B. Guinot, published by Cambridge University “Comparison of Continuously Filtered GPS Carrier-Phase Time
Press, Cambridge, U.K., 2001. and Frequency Transfer with Independent Daily GPS Carrier-
Phase Solutions and with Two-Way Satellite Time Transfer” by
For an introduction to time keeping and GPS,see D. Matsakis, K. Senior, and P. Cook in Proceedings of the 33rd
“Time, Clocks, and GPS” by R.B. Langley in GPS World, Vol. 2, Annual Precise Time and Time Interval (PTTI) Systems and
No. 10, November/December, 1991, pp. 38-42. Application Meeting, Long Beach, California, November 27-29,
2002, pp 63-87. This paper can be downloaded from
For additional information on GPS time transfer techniques,see http://tycho.usno.navy.mil/ptti/ptti2001/paper6.pdf.
“GPS Time Transfer” by W. Lewandowski and C. Thomas in For an overview of USNO’s current time transfer practices,see
Proceedings of the IEEE, Special Issue on Time and Frequency,
Vol. 79, No. 7, 1991, pp 991-1000. “USNO and GPS: It’s About Time” by D. Matsakis in
GPS World, Vol. 11, No. 2, February 2000, pp. 32-40.
For information on Two-Way Satellite Time and Frequency Transfer,see
For information on IGS clock products,see
“Two-Way Time Transfer Via Communication Satellites” by
D. Kirchner, Proceedings of the IEEE, Vol 79, No. 7, 1991, pp. 983- “A Global Time Transfer Assessment: New IGS Clock Products”
990. by J. Ray and K. Senior in GPS World, Vol. 13, No. 11, November
2002, pp. 45-51.

36 GPS World NOVEMBER 2004 www.gpsworld.com

parameters including clocks. They and the timing receivers produce a time output that curacy.
IGS are coordinating these networks at a is accurate to better than 1 µs, and prod- Multipath is a well known error source
higher level.A real-time continuously fil- ucts are available that are accurate to bet- for all forms of GPS observations. How-
tered solution for the timing difference ter than 100 ns, or even a few tens of ever, timekeeping has an added multipath
between the USNO Master Clock located nanoseconds. Absolute calibration can be concern due to reflections in the cabling.
in Washington, D.C., and the USNO achieved by using a GPS signal simulator Care should be taken in impedance match-
Alternate Master Clock located at the to calibrate the group delay through the ing between the elements of the user’s GPS
Schriever Air Force Base in Colorado GPS antenna, receiver, and cables. Cali- system. Failure to do so can cause large
Springs, Colorado, can be seen at bration is more commonly achieved by temperature and time-dependent varia-
http://galia.gdgps.net/igdg/demo_usn3_clock/. using a GPS receiver whose calibration has tions in the measurements (up to 10 ns).
Every second, the network’s clock differ- been previously determined. Both USNO
ences are estimated with a precision of and NIST offer calibration services that Other Transfer Techniques
20 ps. allow a receiver to be calibrated to within In addition to GPS, there are other tech-
The IGS provides a time-transfer ser- the random error of the receiver bias sta- niques for disseminating time and frequency.
vice, in which data from participating lab- bility. Using great care, the calibration bias Radio broadcasts, dial-up phone services,
oratories are made publicly available from error can be reduced to less than 5 ns. and Internet time transfer using, for exam-
their Web sites using either their final orbit Even in the best-designed system, GPS ple, Network Time Protocol are much less
or rapid orbit values. The IGS clock prod- receivers can vary by several nanoseconds precise but are used by millions of people.
ucts are averages of reductions from dif- in their calibration over time (months to However, it’s important to note that even
ferent analysis centers, which can vary years). It is important to frequently check these systems are often calibrated using
among themselves by about 1 ns peak to the calibration of GPS systems. The BIPM GPS.
peak. The clock products are referenced has begun systematic calibration monitor- The Wide Area Augmentation System
to the IGS time scale loosely steered to ing of receivers maintained at approxi- (WAAS) satellites are designed to closely
GPS Time. mately 20 timekeeping institutions, and emulate a GPS satellite. Although the prime
one important byproduct of this effort will motivation for the system is to provide
Precise Calibration be a quantification of long-term receiver safety-of-life enhancement to GPS, WAAS
A common concern for all modes of ex- calibration variations. Calibration is also provides a precise ranging (timing) sig-
tracting time from GPS is the calibration. achieved by means of a traveling receiver nal that can be used as an additional satel-
The precise calibration bias of a GPS sys- system used for parallel measurements at lite signal to form a navigation and time so-
tem (including receiver, antenna, and ca- each site visited. Based upon the long-term lution. Since the WAAS satellites are
bling) typically is one of the largest er- repeatability of these calibrations, the geostationary, they can be used by a sta-
rors in providing an accurate and traceable BIPM has ascribed a very conservative tionary timing user to obtain time without
UTC timing signal. Most commercial GPS value of 5 ns r.m.s. for the calibration ac- tracking GPS satellites. A highly directional

www.gpsworld.com CIRCLE 19 NOVEMBER 2004 GPS World 37

 TIMING & SYNCHR ONIZATION
❯❯❯INNOVATION

TABLE 1 Nominal r.m.s. values for various time-transfer error sources

Error Source Direct Common-View Carrier Phase GPS TWSTFT
Single-Frequency GPS GPS Parameter Solution
UTC(USNO)_Lab_Y Lab_X_Lab_Y Lab_X_Lab_Y Lab_X_Lab_Y
Multipath Bias 1–3 ns 1–4 ns 1–4 ns < 1 ns
(24-hour average)
Multipath Precision 1–10 ns 1–14 ns < 20 ps < 100 ps
(13-minute observation)
Model Ionosphere 3–6 ns 3–6 ns* 0 ns 30–300 ps**
Troposphere at Zenith 300 ps (no fit) 500 ps (no fit)* 10 ps after fit 0
Broadcast Orbits < 3 ns/observation < 3 ns/observation* — <100 ps*
Broadcast Clocks < 3 ns/observation 0 — —
Hardware Variations 2 ns per year 3 ns per year 3 ns per year < 1 ns per year
Receiver Noise < 1 ns < 1 ns < 20 ps < 20 ps
(averaged over 5 minutes)
Earth Tide 300 ps** 300 ps** < 10 ps* 0 ps
(at any instant)
* Baseline dependent, better for short baselines
** Assuming effect is not modeled. Baseline dependent for common-view and TWSTFT.

TABLE 2 Typical r.m.s. performance of time-transfer modes for calibrated systems

Type Time Time Time Frequency
Stability Stability Accuracy Accuracy
(1,000 Seconds) (24 Hours) (24 Hours) (24 Hours)
GPS Direct Broadcast 5–10 ns* 2 ns 3–10 ns 4 x 10-14
(GPSDO single frequency)

GPS Common-View 5 ns 1 ns 1–5 ns 2 x 10-14

(2,500 kilometer baseline)

GPS Carrier-Phase 20 ps 0.1 ns 1–3 ns 2 x 10-15

TWSTFT < 0.1 ns 0.1–0.2 ns 1 ns 2–4 x 10-15

* Equipment dependent

THOMAS E. PARKER is leader of the Atomic Frequency

antenna could be used for this purpose, Summary Standards Group in the Time and Frequency Division of
which would be much less susceptible to in- GPS provides a cost-effective method for the National Institute of Standards and Technology (NIST)
terference and mutipath. users to get high-quality time and fre- in Boulder, Colorado. He received his Ph.D. in physics
from Purdue University in 1973 and is a fellow of the
TWSTFT is another technique used by quency, and this service will improve as
Institute of Electrical and Electronics Engineers (IEEE).
some timing laboratories. It often is used as GPS advances with time. Parker’s areas of interest are time and frequency trans-
a standard for comparison against GPS The quality experienced by the user de- fer, primary frequency standards, and time scales.
time-transfer modes because it generally is pends in many ways upon the user’s sys-
DEMETRIOS MATSAKIS is head of the Time Service
more precise than direct-broadcast GPS or tem. Table 1 summarizes many of the in-
Department of the U.S. Naval Observatory in Washington,
common-view GPS, and comparable if not dividual error sources in GPS and D.C. He received his Ph.D. in physics from the University
superior in accuracy to carrier-phase GPS. TWSTFT, with the understanding that of California at Berkeley in 1978. His areas of interest
This technique requires microwave trans- quantitative values depend strongly on sys- are time generation, time measurement, time scales,
mitting and receiving equipment at both tem configuration and data-reduction tech- clock steering, and time transfer.
stations, and use of a geostationary com- niques. “Innovation”is a regular column
munication satellite (usually a commercial Table 2 summarizes the overall perfor- featuring discussions about
satellite) as a relay station. If both stations mance for these time-transfer techniques. recent advances in GPS technolo-
transmit and receive simultaneously, most The common-view numbers are for time gy and its applications as well as
the fundamentals of GPS position-
of the path delays and related instabilities transfer without using IGS ionospheric
ing.The column is coordinated by
cancel. USNO uses TWSTFT to support maps or precise orbits. For a 10,000 kilo- RICHARD LANGLEY of the
GPS by transferring time to the USNO Al- meter baseline, common-view performance Department of Geodesy and Geomatics Engineering at the
ternate Master Clock in Colorado. would be similar to that of direct broad- University of New Brunswick, who appreciates receiving
cast. Note that performance may vary over your comments and topic suggestions.To contact him,
see the “Columnists”section on page 2 of this issue.
time due to changes in the environmen-
tal conditions. c
38 GPS World NOVEMBER 2004 www.gpsworld.com