“If timing is to become mission critical, redundant means of distributing timing information is essential,” according to NIST.
NIST hosted the “Time Distribution Alternatives for the Smart Grid Workshop” at its Gaithersburg, Maryland, campus on March 21. The information gained will inform future NIST, U.S. Department of Energy, national laboratories and private sector technical programs and strategic planning.
The workshop consisted of experts on both electrical power and wide-area time distribution. The experts came from industry, utilities, academia and government.
The findings cover desired future characteristics, targets, challenges and barriers to adoption of time distribution alternatives; and priority R&D areas for time distribution alternatives.
Potential alternatives to wide area distributed time synchronization include Enhanced WWVB (radio signal broadcasting), eLoran (hyperbolic radio navigation) and the IEEE Wide Area Precision Time Protocol (PTP – master slave clock synchronization).
Results of the workshop illustrate the need for alternatives to existing GPS timing systems as well as backup systems and many of the challenges that need to be addressed to develop and implement alternatives. Some of the overarching themes that emerged include the following:
While a number of potential alternative exist, they will require further infrastructure, research and concerted investment to implement and demonstrate their potential to replace, supplement, back up, or fill gaps in existing GPS systems.
Potential alternatives may need to be combined in ensembles to fill gaps, create the needed redundancies, and supplement GPS-based timing.
Future alternatives to GPS will need to have the same or better levels of accuracy, resilience, security, trustworthiness, and availability to supplant existing systems; a diversity of timing distribution systems may be needed (terrestrial, communication-based, wireless, etc.).
Dependency on space-based systems is currently strong due to their perceived reliability; there is limited awareness of the possible adverse impacts of timing failure events in such systems (and few backups exist).
Developing and using existing alternatives and new technologies, and integrating these with legacy systems will require standards and use cases to enable new technology, architectures, and interoperability among systems.
Better understanding of attack and failure threat modes is needed to estimate and demonstrate the true consequences of timing failures in systems based entirely on GPS.
Wide-Area Wireless Network Synchronization with LocataNets
The United States Naval Observatory conducted several independent frequency synchronization experiments in Washington, D.C., using an alternative PNT technology in multiple network configurations. The results suggest that sub-nanosecond time transfer using this technology may be possible over wide urban areas, and that it could thus serve as a GPS augmentation or back-up solution over wide areas for critical applications that depend on precise time.
By Edward Powers and Arnold Colina
Because of the great responsibility of being the prime source of time for many critical national systems, the United States Naval Observatory’s (USNO’s) clock system must be at least one step ahead of the demands expected to be made on its accuracy. Therefore, innovative methods of transferring precise time and frequency must continually be anticipated, investigated and supported.
The USNO has developed one of the world’s most accurate and precise atomic clock systems, used by many systems requiring highly precise time. The USNO operates the U.S. Master Clock, which provides the precise time source for the GPS satellite constellation run by the Air Force; it is also the time standard for the U.S. Department of Defense. Along with its sister organization, the National Institute of Standards and Technology (NIST), it provides the official time for the entire nation.
To investigate new precise time transfer methods, the USNO desired to independently test Locata’s TimeLoc methodology as a possible technology for maintaining precise frequency synchronization across an urban or wide-area network — the foundation for supporting precise time transfer.
Internet of Everything Ups Timing Requirements
Many critical modern systems such as 4G mobile phone networks, banking, and electricity grids demand high-accuracy time and frequency stability across specified areas. Precise network synchronization is critical for nearly all digital networks, and more stringent network stability requirements are expected to emerge as the user base for these applications continues to grow. To date, the preferred method to achieve this performance is via synchronization from GPS. However, the vulnerability of GPS signals causes growing concern among industry experts. Many actively seek alternative means of precise time transfer and frequency stability across wide areas.
Alternative position, navigation, and timing (PNT) technologies such as chip scale atomic clocks (CSAC), precision time protocol (PTP), and enhanced long range radio navigation (eLoran) are proposed or operational today, with each serving different markets.
Meanwhile, timing needs for wireless protocols continue to increase with the proliferation of mobile phones and other wireless communication devices. To accommodate a booming user base, wireless spectrum must be carefully managed to improve bandwidth and channel efficiency. Wireless communication performance is fundamentally dependent upon precise time and frequency, so improvements in highly accurate timekeeping methods will permit better spectrum utilization, which in turn permits more users and more bandwidth per user.
Clearly, synchronization is a core enabling technology for modern digital systems, both for radiopositioning and the world’s telecommunications highways. But synchronization is taken for granted because, when it works well, it is effectively invisible. Without it, however, everything is likely to fall apart.
Synchronization will become even more crucial for the next generation of digital systems. A recent paper by the U.S. National Institute of Standards and Technology (NIST) states that we stand at the advent of a revolutionary new economy fueled by a global Internet of Everything (IoE), in which 37 billion new things will be connected to the Internet by 2020.
This NIST paper adds that “One fundamental enabler of this revolution will be the marriage of timing signals and data that breaks through the existing barriers. Timing is critical for future development and improvements.”
Improved wireless synchronization has proved very challenging to realize, as the timer in each network node is derived from an independent oscillator that is affected by long/short term frequency drifts and jitter. Many alternative timekeeping methods present serious limitations in terms of precision or network size.
Encouraged by earlier published results showing that Locata Corporation’s radio-based PNT technology enables network synchronization at the nanosecond level, and suggesting that it could perform comparably across large urban areas, the United States Naval Observatory (USNO) conducted its own synchronization experiments on Locata technology.
Real-World Challenges
The USNO campus is situated about 4 kilometers northwest of the White House in Washington, D.C. The grassy tree-lined campus is, unfortunately, a relatively small area for testing wide-area synchronization capabilities. It became apparent that realistic long-distance tests would necessitate extending the LocataNet outside USNO boundaries. This meant coordinating access to other facilities in theWashington, D.C., area to allow remote housing of LocataLites and their antennas. As many researchers will confirm: when real-world testing requires access to multiple external sites and their disparate administrations, the coordination required to keep everything on track can quickly become the most daunting challenge of the exercise. We needed to find cooperative facilities, preferably with line-of-sight (LOS) to the USNO and its Master Clock in order to establish the best TimeLoc link between facilities. As we also wanted to exercise TimeLoc’s ability to cascade its synchronization through multiple LocataLites, ever more D.C. facilities would need to be involved. Predictably, it transpired that not many facility managers in the Washington district were eager to help the USNO broadcast and receive new and unknown signals in or around their government buildings! And those who were amenable to support the demonstration either lacked a LOS, or were not willing to assist without considerable monetary compensation.
Step 1: LocataLite A transmits a unique signal (code and carrier). Step 2: LocataLite B acquires, tracks and measures the signal generated by A. Step 3: LocataLite B generates its own unique signal (code and carrier) which is transmitted in the normal manner. Importantly, the transmitted signal is received by the receiver section of LocataLite B as well. Step 4: LocataLite B calculates the difference between the signal received from LocataLite A and its own locally generated and received unique signal. Ignoring propagation errors, the differences between the two signals are due to the difference in the clocks between the two devices, and the geometric separation between them. Step 5: LocataLite B adjusts its local oscillator to bring the differences between its own signal and LocataLite A’s received signal to zero. The signal differences are continually monitored and adjusted so that they remain zero. In other words, the local oscillator of B follows precisely that of A. Step 6: The system corrects for the geometrical offset (range) between LocataLite A and B, using the known coordinates of the LocataLites’ antennas. When this step is accomplished, TimeLoc has been achieved.
After months of attempts to secure appropriate partners for this demonstration, we finally found some supporters in the shape of the Federal Aviation Administration Building in Rosslyn, Va., and the National Cathedral in Washington, D.C. Regrettably, it turned out that these facilities were not going to be available at the same time! Logistic challenges never end. This scheduling reality necessitated spreading the TimeLoc demonstration over several months in three different blocks of trials. Nevertheless, we were eventually able to devise a plan which leveraged access to the USNO and still accommodated the timetables of the supporting external facilities.
A series of experiments were planned to measure and evaluate the stability between master and slave LocataLite 1-pulse per second (PPS) signals in several urban LocataNet configurations. Many of the trials were specifically designed to measure TimeLoc’s ability to cascade multiple times through multiple LocataLites, exercising the technology’s capabilities over increasing distances and hence correspondingly larger notional coverage areas.
Figure 2. LocataLites under test at USNO.
Locata signals were broadcast in the Industrial, Scientific and Medical (ISM) 2.4 GHz radio band, commonly known as the Wi-Fi band, with a total radiated power of 200–500 mW. LocataLites and their respective antennas were installed at locations that permitted LOS between units, according to whichever specific LocataNet configuration was being evaluated at the time. In each configuration, the master LocataLite, designated as LocataLite 1, was synchronized to the USNO Master Clock so that the Master Clock’s time would be propagated through the LocataNet. Both the master and slave LocataLite 1PPS signals were collected into a time interval counter and the time difference between their rising edges was measured.
When tracking radio frequency signals over a significant distance, tropospheric delay becomes an important error source for measurements used in timing solutions. The speed of light can only be assumed to be universally constant in a vacuum, so atmospheric temperature, pressure and humidity materially changes the speed of light when propagating through air. In fact, using standard atmospheric parameters, the unmodeled tropospheric delay is surprisingly large — approximately 280 parts per million (ppm), which equates to slowing down almost one nanosecond over each kilometer of radio transmission. Obviously, as transmission distances increase, tropospheric error becomes a substantial factor which must be accounted for in hyper-accurate timing systems. Devising methodologies that effectively mitigate large tropospheric errors becomes essential.
To help solve this problem, Locata developed new tropospheric models that use relatively inexpensive meteorological (MET) stations which measure temperature, pressure and relative humidity at the LocataLite sites. This modeling alone is able to mitigate the tropospheric effects to within just few parts per million. This proved to be an essential feature, as the weather during the course of the entire months-long testing campaign varied significantly among the separate trials.
The TimeLoc Process
LocataNets function as local ground-based replicas of the satellite-based GPS position and timing networks. A LocataNet can be designed and configured by the user to deliver a powerful, local, controllable, tailored signal as required by different applications.
The easiest LocataNet layout to describe is a hub-and-spoke model consisting of a single master LocataLite transceiver and one or more slave LocataLites. More complex network configurations have been deployed in many commercial systems in use today. The patented process by which slaves are synchronized to the master (or other slaves) is known as TimeLoc.
In 2013, a University of New South Wales team demonstrated that Locata’s radio-based TimeLoc technology provided accurate time transfer (~5 ns) and frequency stability (~1 ppb) across a large distance of 73 kilometers (45.4 miles). This significantly outperforms GPS for wireless time transfer. Given this demonstrated radius of transmission in a rudimentary configuration, Locata was shown as being able to supply nanosecond-accurate time to a 146 km diameter circle, which would cover 16,750 km2 — almost 200 times the size of Manhattan. Ranges greater than this can be deployed if required for safety-of-life, military or government-mandated systems.
As TimeLoc is accomplished without the use of atomic clocks, this represents a new level in precision network synchronization of this scale. It could conceivably serve as a GPS augmentation or back-up solution over wide areas for critical applications that depend on precise time.
Since Locata technology was originally developed as a high-accuracy non-GPS-based positioning and navigation solution, the time synchronization accuracy requirements for a LocataLite transceiver are very high. If sub-centimeter positioning precision is desired for a Locata receiver, every smallest fraction of a second is significant; for example, a 1-nanosecond error in time equates to an error of approximately 30 centimeters.
TimeLoc wireless synchronization enables LocataLites to achieve high levels of synchronization without atomic clocks, without external control cables, without differential corrections, and without a master reference receiver.
In theory, there is no limit to the number of LocataLites that can be synchronized together. TimeLoc allows a LocataNet to propagate into difficult environments or over wide areas. For example, if a third LocataLite C can only receive the signals from B (and not A) then it can use these signals from B for time synchronization instead. The only requirement for establishing a LocataNet using TimeLoc is that LocataLites must receive signals from one other LocataLite. This does not have to be the same central or master LocataLite, since this may not be possible in difficult environments with obstructions, or when propagating the LocataNet over wide areas.
This method of cascading TimeLoc through intermediate LocataLites has been proven in a growing number of real-world operational LocataNets, including a network in use today by the U.S. Air Force which is configured to cover up to 2,500 square miles (6,500 square kilometers) of the White Sands Missile Range in New Mexico.
In large networks where extremely high synchronization accuracies are required, it is useful to incorporate a meteorological sensor at each LocataLite to monitor the change in weather over considerable distances. This is certainly the case for long-range systems such as the USAF LocataNet installed at the huge White Sands Missile Range, where distances of over 50 km can be found between LocataLites. However, for the purposes of these USNO Washington experiments, where the longest point-to-point transmission distance was 2.9 km, it was assumed that weather parameters would be virtually identical at all LocataLite locations. Therefore only one MET station was employed within the entire network, which for these trials was collocated with the master LocataLite.
The very first experiment conducted by the USNO to gain some familiarization with TimeLoc was run entirely within the grounds of the USNO campus. It employed two LocataLites with their respective antennas on the roof of USNO Building 78. In this initial configuration the antennas were positioned 15.24 m apart. It was intended to use the measured result as a baseline against which TimeLoc synchronization over longer distance could be compared. This first arrangement is referred to as the two-node setup. A diagram of this configuration is shown in Figure 3.
Second and third experiments demonstrated Locata’s ability to cascade the master 1PPS signal to an intermediary slave LocataLite, which in turn transmits a signal to which a third LocataLite can TimeLoc. This LocataNet configuration is referred to as the three-node setup (Figure 4).
Figure 4. Three-node setup (total range: 5.794 km/3.6 mi and 2.401 km/1.49 mi).
This experiment was conducted twice using two different intermediate LocataLite locations. The first intermediate location was indoors on the top floor of the FAA Building in Rosslyn, Va. (Figure 5). The distance between the master/slave antennas to the intermediate antenna in the FAA building was 2.897 km, but since the signal was propagated through a tinted window, the received signal strength inside the building was greatly attenuated, effectively simulating a much longer transmission distance. The second intermediate (LocataLite 2) location was from the balcony of the National Cathedral’s Ringing Chamber. In this case the distance between USNO LocataLite 1 master/slave to the intermediate antenna in the National Cathedral was approximately 1.183 km.
Figure 5. Intermediate LocataLite 2 antennas inside FAA building. In the distance, both the USNO and the National Cathedral are visible.
In both cases in Figure 4, the distance between master and terminal slave antennas was 3.048 m on the USNO building, but they were intentionally not TimeLoc’d to each other. The timing signal was therefore forced to route through the intermediate LocataLite 2 at either the FAA Building or the National Cathedral.
A fourth experiment included yet another intermediate cascade where the TimeLoc signal was transmitted from the second to a third LocataLite/antenna before arriving at the fourth LocataLite in the chain. This LocataNet configuration is referred to as the four-node setup. A diagram of the setup is shown in Figure 6, and it now added a LocataLite on USNO Building 1 to expand the set-up, along with the intermediate LocataLite installed at the National Cathedral (Figure 7).
Figure 6. Four-node setup (total range: 2.413 km/1.5 mi).Figure 7. LocataLite antennas outside the Ringing Chamber of the National Cathedral Spire.
Referring to Figure 6, the distance between the master LocataLite (antenna 1) at USNO Building 78 and the LocataLite (antenna 2) at USNO Building 1 was approximately 42.672 m. The distance between the USNO Building 1 (antenna 2) to the Washington National Cathedral (antenna 3), was approximately 1.144 km. The distance between the Washington National Cathedral (antenna 3) back to antenna 4 on USNO Building 78 was approximately 1.183 km. The total range in this four-node chain was 2.413 km. In this configuration, LocataLites 1 and 4 are intentionally not TimeLoc’d to each other, forcing the 1PPS signal to be routed through LocataLites 2 and 3.
A fifth experiment included yet one more LocataLite and antenna at USNO Building 1 (Figure 9), totaling cascaded TimeLoc among five LocataLites and their respective antennas: the five-node setup, shown in Figure 8. In this configuration LocataLites 1 and 5 are intentionally not TimeLoc’d to each other, forcing the 1PPS signal to be routed through LocataLites 2, 3 and 4.
Figure 8. Five-node setup (total range: 2.427 km/1.51 mi).Figure 9. LocataLite antennas on USNO Building 1. National Cathedral in background.
Measurement Methodology
A measurement of time difference between master and slave LocataLite 1PPS readings was done using a Stanford SR620 universal time interval counter. The rising edge of the 1PPS signals were inspected at 1-Volt trigger level. A 10 MHz reference was provided to the counter from the USNO’s Master Clock. Channels A and B on the counter were designated to the master and slave 1PPS signals respectively. Data were collected from the counter through serial connection to a PC. The length of each experiment was time-limited in some way because of limited access to facilities, such as the FAA building or National Cathedral. However, a minimum of at least 30,000 seconds (8.33 hours) of data were collected for each test to characterize the overall stability of the 1PPS signals between master and terminal slave LocataLites.
Collected Data
Figures 10 to 14 show the normalized 1PPS time difference between the master LocataLite and the terminal slave LocataLite. Normalization effectively removes errors due to unsurveyed antenna locations and uncorrected cable delays; hence it highlights the frequency coherence of the network.
Figure 10. Two-node setup on USNO rooftop, collected for slightly more than 1 day. Distance: 15.24 m/50 ft. Synchronization Standard Deviation (SSD) = 51.095 picoseconds.Figure 11. Three-node setup at USNO and FAA Building, collected for over 12 hours: 5.794 km/3.6 mi. SSD = 127.333 picoseconds.Figure 12. Three-node setup at USNO National Cathedral, collected for over 12-hours: 2.401 km/1.49 mi. SSD = 171.325 picoseconds.Figure 13. Four-node setup at USNO with cascades at National Cathedral and USNO building 1, collected for over 17-hours: 2.413 km/1.5 mi. SSD = 145.247 picoseconds.Figure 14. Five-node setup with cascades at National Cathedral and 2 hops at USNO building 1, collected for over 8-hours: 2.427 km/1.51 mi. SSD = 197.766 picoseconds.
Results
The results in Table 1 show the 1PPS signal variability for each LocataNet under evaluation. These values represent the frequency coherence between master and terminal slave LocataLite 1PPS signals for each experiment.
Table 1. LocataNet frequency stability.
The two-node setup used two LocataLite antennas located within 15.24m of each other. The measured precision standard deviation was 51.095 picoseconds. This value is a culmination of the total Locata noise budget, which is expected to consist of TimeLoc noise, residual tropospheric error, multipath change (signal scattering/diffusion), PPS generation, and PPS measurement. This two-node result can be used as a baseline for Table 1 measurement results over longer distances. The differences are shown in the last column of Table 1. For example, cascading TimeLoc over the 5.794 km three-node setup introduced an additional deviation of 76.238 picoseconds, compared to the two-node set-up.
The three-node setup tested the effect of adding a TimeLoc cascade wherein the Locata signal from the master is routed to an intermediate LocataLite, and then to the terminal slave. When the master LocataLite signal was cascaded through the intermediate LocataLite at the FAA Building, the configuration showed a standard deviation of 127.333 ps across a total signal path length of 5.794 km. Alternatively, when the master LocataLite signal was cascaded through the intermediate LocataLite at the National Cathedral, that three-node configuration showed a standard deviation of 171.325 ps across a total signal path length of 2.401 km.
Interestingly, it appears that in the two different three-node setups, the intermediate cascade to the FAA building (2.9 km from the master and terminal slave LocataLites) delivered slightly better time transfer performance than the configuration which leveraged the closer (1.183 km) National Cathedral intermediate cascade. We believe this is attributable to the fact that the line-of-sight between USNO Building 78 (the site of the master and terminal slave) and the National Cathedral (the intermediate cascade) was completely obscured by heavy foliage seen in Figure 9, and that this particular configuration required the signal to pass through the foliage twice when being transmitted back and forth. Not only does foliage introduce multipath, but the properties of this foliage also changed regularly according to wind and moisture — two weather attributes that varied significantly over the course of the week in which those particular experiments were set up and run. This theory seems reasonable, since the four-node setup only required the signal to pass through this foliage once, and the recorded performance was better than the three-node setup — despite the fact that an additional TimeLoc cascade point was introduced.
The four-node setup included TimeLoc cascades at USNO Building 1 and the Washington National Cathedral. In this configuration, the data from Table 1 shows a standard deviation of 145.247ps across a total signal path length of 2.413 km.
The five-node setup included yet another TimeLoc cascade between the National Cathedral and Building 1 at USNO before reaching the terminal LocataLite slave. In this configuration, the data from Table 1 shows a standard deviation of 197.766ps across a total signal path length of 2.427 km.
Frequency Stability
Frequency stability is best measured over long periods. Because all of the equipment in the two-node setup was located on USNO premises, it could be run undisturbed for a longer period of time than configurations which required access to external facilities outside of the USNO’s control. Data obtained from the two-node setup were used to calculate the frequency stability between the two TimeLoc’d LocataLites. The length of this data set was 28 hours, 22 minutes, and 40 seconds. During this period, the approximate one-day frequency stability was measured as 1×10-15 (1 part per quadrillion).
To put this measurement into a more practical context: Stratum 1 is defined as a source of frequency with an accuracy of at least 1×10- 11, hence Stratum 1 performance generally originates from an atomic standard. For example, Cesium beam atomic clocks typically provide better performance than this, with one day Allen deviation stabilities in the mid- Ex10-14 (usually stable to between 3×10-14 to 6×10-14). Rubidium clocks are typically never more stable than 1×10-13 and Maser clocks are typically stable to mid-to-low Ex10-15 over one day.
Locata’s link stability — achieved without the use of atomic clocks — is clearly capable of distributing Stratum 1 frequency and precise time without substantially degrading the reference clock stability. This measured performance is significant, because a stable network is an essential prerequisite for precise time and frequency transfer. Moreover, for many traditional timing applications and developing digital and IoE applications, stability is more important than accuracy; just as for most advanced technology applications, frequency is more important than time of day.
Conclusions
The five USNO experiments suggest that the variations of the measured frequency synchronization between master and terminal slave LocataLites were not inevitably attributable to the distance between LocataLites, but rather governed by the number of nodes or cascade points in the LocataNet configuration, and LocataLite signal quality. Each signal cascade through an intermediate LocataLite introduced ~25ps of jitter into the solution.
Additionally, it was noted that transmitting TimeLoc signals across an urban environment did not always allow for unobstructed line-of-sight or completely open-sky environments. For instance, some of the LocataNet configurations required the signal to travel through dense, leafy trees which appeared to slightly affect overall frequency stability. Additionally, one FAA configuration required the signal to travel indoors through a tinted window which ultimately affected received signal strength.
These USNO tests highlighted the capability of the LocataLite as a viable option for a stable 1PPS distribution setup within an urban environment in support of applications like cell tower synchronization in “GPS-challenged” environments. All tested configurations demonstrated frequency synchronization of less than 200 picoseconds. This is significantly better than any other known wireless network synchronization methodology, including GPS. Furthermore, if clear line-of-sight is available between a master and slave LocataLite, the 2-node precision has been shown to be on the order of 50ps, and has one-day stabilities to 1×10-15.
These results, reinforced by those previously reported in University of New South Wales tests over a very large area, suggest that distance between nodes is not a significant factor, provided that sufficient signal quality is maintained. Thus, there are no theoretical or technical problems with scaling LocataNets to very large areas. In fact, this has already been demonstrated at the White Sands Missile Range where the USAF has now deployed a fully-operational Locata network that covers up to 2,500 square miles (6,500 square km), about 80 times the size of Manhattan.
The USNO trials reported here have clearly demonstrated TimeLoc’s relative picosecond-level synchronization of independent Locata networks. If this highly-stable network capability were not in place, precise time transfer would not be possible. The next step is to demonstrate how well a LocataNet can deliver absolute time transfer of the USNO’s Master Clock time to any other network node across similar areas of Washington, D.C.
Acknowledgments
The authors would like to thank James Shepherd of the National Cathedral and Paul Worcester of the FAA for use of their respective buildings. The authors would also like to thank Locata personnel for the use of their equipment and technical assistance in setting up the LocataNets under evaluation.
This article is based on a paper presented at ION GNSS+ 2015.
Disclaimer
Though particular vendor products are mentioned, neither official USNO endorsement nor recommendation of any product is herein implied.
Edward Powers is the GNSS and Network Time Transfer Operations Division Chief at USNO. He also serves as an advisor to the USAF GPS Directorate supporting space atomic clock development, modernized GPS III navigation message design, GPS accuracy improvement studies, and GPS UE development.
Arnold Colina is an electronics engineer in USNO’s GNSS and Network Time Transfer division, tasked with providing accurate UTC reference through GPS and performing calibration tests on GNSS receivers.
Timing Versus Synchronization
“They say “timing is everything” but nowadays it’s probably more correct to say “synchronization is everything”. There is a significant difference, yet many are surprised to learn they are not the same thing.
“Time dependent” applications rely on their clocks being close to the “real time”, as defined by a consensus of super-accurate atomic clocks managed by national bodies like the USNO. Once agreed upon by the labs, this “real time” can be distributed to various “time dependent” networks as a reference time to drive their operations.
“Time synchronized” applications, on the other hand, employ a methodology in which a common network time can be transferred to each network node. In other words, often the real technology enabler is that all the clocks in a defined network are synchronized to each other, even if they all run to what is any arbitrarily defined time-base. The “real time” doesn’t matter as much as how closely the node times agree with each other. As Einstein famously taught us: “Everything is relative.”
For example, accurate synchronization enables GPS positioning to work because a user’s GPS receiver relies on time-of-arrival comparisons from four or more satellites transmitting their signals at the same instant. But — even in this GPS paradigm where atomic clocks are always touted to be the most fundamental of requirements — it is important to appreciate this: A GPS user’s receiver does not care how, or to what “time,” the satellites are made synchronous. The only things the user receiver needs to know is where the satellites are, and that the satellites are synchronized when they transmit their signal.
Unfortunately sustaining high-precision, reliable time synchronization of multiple network nodes is a mammoth engineering task. Just ask the US Air Force! All clocks, no matter how accurate they are, eventually drift, so they cannot remain synchronized without comparison and adjustment.
Given the world’s exploding, insatiable demand for more data transmitted via ever-faster wireless systems, synchronization will become even more important than it is today. More wireless users and more bandwidth per user means that nanosecond — or even picosecond — network synchronization is one of the emerging engineering challenges of the 21st century. There are few resource on earth which are as scarce, or more precious today, than spectrum. So there is no question that better or cheaper ways to greatly improve network frequency and synchronization will translate directly into better use of the world’s exceptionally valuable, extremely limited spectrum resource.
40th Annual NIST Time and Frequency Metrology Seminar
There were four of us, mature males who all remember having a crush on Annette Funicello, were seated around a table avidly discussing deviant behavior with a sometimes rapt mixed-gender audience. The four of us, loudly discussing deviant, and only occasionally aberrant behavior, were doctors: David Allan the world renowned creator of Allan Deviation or variance fame, Judah Levine, world renowned nuclear physicist and Father Time of NIST (National Institute of Standards and Technology), Neil Ashby, former chair and currently Professor Emeritus of Physics at UC Boulder, also from NIST, along with yours truly representing GPS World magazine and the Institute for Defense Analyses. Our ever-changing audience was composed of the 40+ members from around the globe attending the 40th Annual NIST Time and Frequency (T&F) Metrology Seminar, held June 2-5 in stunningly beautiful Boulder, Colo.
Of course, the numerous deviant behaviors under discussion had more to do with the sometimes-fickle performance of various atomic reference systems than they did anatomy. And we were speaking loudly because that is what most men of our age do. Dr. David Allan frequently threw in quotes and anecdotes from his recently published book on time, It’s About Time, about which you will read more later.
The NIST T&F Metrology Seminar is truly one of a kind, easily the best in the world for time and frequency metrology. I have been fortunate enough to attend numerous times. I can truly say I have never found it repetitive or boring. There are so many exciting discoveries concerning time, which David Allan staunchly maintains is a purely human construct, and how time applies to our everyday lives, especially to GPS — all PNT systems actually — that it is impossible not to be constantly fascinated.
NIST Mission
NIST Boulder is all about research and development for timing standards, which is a benign way of saying NIST SMEs (subject matter experts) are the world’s foremost authorities on time and metrology. To wit, NIST has produced no less than four Nobel Prize winners in metrology, the last being awarded in 2012. The atmosphere at NIST and the University of Colorado Boulder campus is such that you can’t help but feel certain there are more Nobel Prizes for NIST on the horizon.
David Howe (Ph.D.), my NIST host and group leader of the Time and Frequency Metrology Division, explained that his organization, which sponsors the seminar, is an operating unit of the Physical Measurement Laboratory of the National Institute of Standards and Technology (NIST), an agency of the U.S. Department of Commerce. The NIST T&F Division is located in Boulder at the NIST Boulder Laboratories, just across from the street from the University of Colorado. Many of the NIST researchers are also University of Colorado professors, adjuncts or graduate students.
The NIST mission includes:
Maintaining the primary frequency standard for the United States
Developing and operating standards of time and frequency
Coordinating United States time and frequency standards with other world standards
Providing time and frequency services for United States clientele
Performing research in support of improved standards and services
Precise time and frequency information is required by electric power companies, radio and television stations, telephone companies, air traffic control systems, participants in space exploration, computer networks, scientists monitoring data of all kinds, and navigators of many types. These users need to compare their own timing equipment to a reliable, internationally recognized standard. NIST provides this standard for the United States.
Of course one of the largest distribution networks for timing data is through the Global Positioning System (GPS), which provides this data globally to more than 4+ billion users and millions of timing systems everyday, numerous times per day. The number of times GPS time is utilized per day is almost impossible to calculate, but most certainly resides in the trillions.
The NIST Time and Frequency distribution system delivers NIST Internet time over the Internet at the rate of 8 billion requests per day from servers at 25 locations across the United States.
The frequency stability provided by classic Cesium and Rubidium atomic reference systems onboard GPS payloads have historically been on the order of 1 x 10-14. While this is the stability provided by the GPS IIF rubidium clocks, currently the rubidium clocks being prepared for GPS III are achieving frequency stability on the order of 1 x 10-15 under laboratory conditions, an order of magnitude better than the current on-orbit clocks.
This is actually an amazing feat. For those of you who don’t deal in scientific notation on a daily basis, this means — since it is on a logarithmic scale — that the frequency stability of GPS III’s atomic clocks have the potential to be 10 times as stable as the IIF clocks, which are currently the most stable and accurate GPS clocks on orbit to date.
Where atomic reference systems are concerned, we routinely speak of frequency stability and not clock accuracy. It is the stability over measured epochs, short and long, that matters most. Indeed, it is the oft-misunderstood frequency stability uncertainty expressed as delta f/f = 1 x 10-16 that produces the clock accuracy to within one standard (SI) second in three hundred (yes, 300) million years — a statistic that is obviously not directly observable, but reasonably predictable. Hence, as Judah Levine often says, where stability is concerned you are an historian, but where accuracy is concerned you are a prophet. NIST defines an SI second as the duration of 9,192,631,770 cycles of the cesium hyperfine transition.
Tom O’Brian, the current chief of the NIST Time and Frequency Division, explained that this level of precision is equivalent to measuring the distance from the Earth to the Sun, a distance of 150 million kilometers, to the uncertainty of 15 microns or 1/10 the thickness of a human hair. While that is impressive, the best is yet to come. NIST is currently working on research-grade optical clocks, which we could reasonably expect to see on orbit one day in future GPS payloads, with an optical frequency stability equivalent to delta f/f = 2 x 10-18 or accuracy equal to 1 second in 15 billion years. Again this is the equivalent of measuring the distance from the Earth to the Sun to an uncertainty of 0.3 micron or the size of a virus.
So What?
Many of you may be asking why, as a GPS user, or merely as a user of technology, you should care about accurate and stable time reference systems. Marc Weiss, a long-time acquaintance and noted researcher at NIST (now in semi-retirement), very eloquently put his thoughts about time in an introduction to a recent timing white paper*, which has been slightly edited for length, current trends and readability. [Ed. So as to not be accused of putting words or opinions in the authors’ mouths, we have provided a reference for the unedited paper at the end of the referenced section]. Marc and several other metrology luminaries express their feelings concerning the future of time and why we should all care:
We stand at the advent of a revolutionary new economy fueled by the global Internet of Everything (IOE). The IOE is a combination of traditional telecom systems with a growing need for wireless technology, and the emerging Internet of Things (IOT) including Machine-to-Machine (M2M) technology. Several current technology providers predict there will be a trillion global endpoints connected to the Internet by 2022, with $14.4 trillion in value at stake.
One fundamental enabler of this revolution is a marriage of timing signals and data that breaks through existing barriers. Currently, optimal use of data in computing and networking is anathema to optimal use of timing signals. Computer hardware, software and networking all isolate timing processes, allowing the data to be processed with maximum efficiency due in part to asynchrony. Yet, the coordination of processes, the time stamping of events, latency measurements and optimal use of precious spectrum are all enabled by ever more accurate and stable timing.
Timing is critical for the future development of and improvements to several high-value applications. For example, in smart transportation systems the exchange of information between vehicles, highways, and civil authorities depends on a robust ubiquitous timing system to ensure the rapid, accurate synchronization and provenance of data. Similar requirements are found in the operation of power grids, especially now that wind farms, solar arrays and the like require different control strategies, which are a critical part of the system. Modern medical applications such as tele-surgery and real time integrative online medical conferences, as well as applications in financial systems are all important examples that require accurate and stable timing signals and may well affect us all.
There are three different types of timing signals for dependable synchronization: frequency, phase, and time. Frequency can be supplied by an individual clock, such as a commercial (atomic) Cesium or Rubidium standard, though practicality drives the use of local oscillators that require calibration and active reference signals. [Ed. Many of these local reference systems and oscillators are routinely updated by GPS signals.] By contrast, phase and time synchronization always require the transport of timing signals plus data. Timing signals are physical, they occur on the physical layer of networks. Indeed the IoT has many devices and applications that require frequency, time and/or phase synchronization. Frequency, time and phase all need to cross layers, boundaries, and networks from their sources in accurate clocks. Requirements for these transfer systems include parameters that create different, perhaps orthogonal, demands on systems. Accuracy, stability, integrity and even non-repudiability requirements are realized with varying demands on different systems….
To facilitate the massive growth of the IoE, data processing and networking require new designs at fundamental levels, allowing integration with precise and verifiable time, frequency and phase signals.
Timing performance is fundamentally dependent upon an underlying oscillator, or ensemble of oscillators and the clocks constructed based on these oscillators.
However, it is apparent to us that many of the researchers and developers of the various time aware systems operate independently of each other. They attend different conferences, read different literature, and in general do not interact sufficiently to achieve the breakthroughs needed. In our minds this calls for a dedicated and collaborative “across the stack” research collaboration focused on two or three comprehensive challenge problems.
Fortunately, this is what researchers, scientists, analysts and metrology experts do at NIST and what we learned about during the T&F Metrology Seminar. The bottom line is many perturbations affect timing signals from atomic reference systems and even local quartz oscillators (clocks). The more these perturbations are understood, the easier they are mitigated and the more stable and accurate our timing signals will be and the faster technology — PNT (position, navigation and timing), clock and otherwise — advances.
For many traditional timing applications and developing “post-timing” applications, stability is more important than accuracy; just as for most advanced technology applications, frequency is more important than time of day.
NIST clearly states its Time and Frequency Metrology Group has the world’s most advanced measurement and calibration facilities for characterizing noise components in oscillators and frequency synthesizers. NIST engages in numerous research and development activities to determine the cause of various types of noise for the purpose of isolating and reducing it, leading to improved components, instruments, techniques and results that are often critical in modern applications. In other words, you have to thoroughly understand a clock issue before you can begin to mitigate issues affecting it. NIST, a synecdoche for understanding time, does that better than any other metrology laboratory in the world today when it comes to atomic reference systems.
What Is Time and Why Does It Matter?
Accurate timing and synchronization are a crucial part of the world’s critical national infrastructure and of modern technology in general, especially the timing signals from GPS satellites, which are used by billions of users continuously every day — although most users remain unaware of the importance and impact that accurate and stable timing has on their everyday lives.
Tom O’Brian reminded us that even St. Augustine of Hippo wondered about time. In circa 400 he wrote:
“What then is time? If no one asks me, I know. If someone asks me to explain, I know not.”
Then, just 1500 years later in 1930, Albert Einstein had this to say about time:
“Space and time are modes by which we think, not conditions under which we live.”
Therefore, I agree with David Allan when he posits that time is a human invention with which only humans struggle. Be that as it may, it is still a condition we live under, and when you consider all the forces, minute to infinite, that affect atomic reference systems and clocks in general, it is amazing our clocks function as well as they do.
Consider that atomic clocks, and even quartz clocks to some extent, are affected by the following elemental and environmental forces and more in the laboratory:
Motion
Acceleration
Gravity – Earth, Moon and planetary
Changes in elevation
~23 different types of noise
Temperature
Magnetic fields
Earth’s Poles
Tides
Light (including lasers)
Electricity
General and Special Relativity
Radiation
The United States Air Force then takes these delicate clocks, atomic (Rubidium and Cesium) as well as quartz VCXOs and OXOs, and launches them (with violent maneuvers) into space in a Medium Earth Orbit that regularly intersects the Van Allen radiation belt. Once on orbit, the clocks routinely experience every one of the listed forces and more on both a regular and changing basis. Of course, we expect the GPS clocks to operate at the same standards and with the same stability and accuracy they displayed in the laboratory. Not asking much are we?
The amazing fact is that thanks to the dedicated scientists and physicists at NIST and other timing laboratories, the clocks work as advertised and continue to do so sometimes for more than 20 years. The current GPS III Rubidium clocks being tested and aged at NRL (Naval Research Laboratory) and other locations around the U.S .are posited to be the first 30-year Rubidium standards with nominal frequency stability of 1 x 10-15. This should provide GPS with another nanosecond of timing accuracy and another 12 inches of positioning accuracy. There will be three of these extremely stable Rubidium clocks on board each GPS III satellite — no Cesium clocks for this family of satellites. Horologists around the world are hoping it is truly a 30-year tube and that only one Rubidium will be required. Only time will tell.
Little Known Factoid (LKF): The first family of GPS satellites on orbit made use of a General and Special Relativity switch that could be set in one of three positions: neutral, plus or minus, depending on whether the universe was relatively static, expanding or shrinking in size. Guess where the switch was set initially and (hint, hint) it could be changed via software from the ground. Drop me a line @ [email protected] and let me know what you think — posit or know, as the case may be.
Thanks
My thanks to David Howe, Judah Levine, Neil Ashby, David Allan (Ph.D.s all) and Danielle Lirette, who made my visit to NIST such a wonderful experience.
It’s About Time
Earlier I mentioned physicist David Allan’s wonderful book, published in 2014. It’s About Time: Science Harmonized with Religion. Allan is about science harmonized with religion and where we are in God’s time. I am halfway through the 402-page tour de force on time, and it is a fascinating read. It is a 50-year biography and history of atomic reference systems, since the first atomic clock only came about in 1949. You’ll be amazed how that happened. Based on what I have read so far, I highly recommend this scientific tome, which is very readable and understandable even for the lay reader. I promise a full review in a future column.
Until then, Happy Navigating! I hope to see many of you at ION JNC (Institute of Navigation Joint Navigation Conference) in Orlando, Fla., June 21-26. There will be a classified day on Thursday, June 25, and a Warfighters Panel as well. Hope you can join us. Remember, GPS is brought to you courtesy of the United States Air Force.
By bouncing eye-safe laser pulses off a mirror on a hillside, researchers at the National Institute of Standards and Technology (NIST) have transferred ultraprecise time signals through open air with unprecedented precision equivalent to the “ticking” of the world’s best next-generation atomic clocks.
Described in the April 28 issue of Nature Photonics, the demonstration shows how next-generation atomic clocks at different locations could be linked wirelessly to improve geodesy (altitude mapping), distribution of time and frequency information, satellite navigation, radar arrays and other applications. Clock signals of this type have previously been transferred by fiber-optic cable, but a wireless channel offers greater flexibility and the eventual possibility of transfer to and from satellites.
NIST researchers transferred ultraprecise time signals over the air between a laboratory on NIST’s campus in Boulder, Colorado, and nearby Kohler Mesa. Signals were sent in both directions, reflected off a mirror on the mesa, and returned to the lab, a total distrance of approximately two kilometers. The two-way technique overcomes timing distortions on the signals from turbulence in the atmosphere, and shows how next-generation atomic clocks at different locations could be linked wirelessly to improve distribution of time and frequency information and other applications.
The stability of the transferred infrared signal matched that of NIST’s best experimental atomic clock, which operates at optical frequencies. Infrared light is very close to the frequencies used by these clocks, and both are much higher than the microwave frequencies in conventional atomic clocks currently used as national time standards. Operating frequency is one of the most important factors in the precision of optical atomic clocks, which have the potential to provide a 100-fold improvement in the accuracy of future time standards. But the signals need to be distributed with minimal loss of precision and accuracy.
The signal transfer demonstration was performed outdoors over a two-way wireless link using two laser frequency combs. A frequency comb generates a steady stream of ultrashort optical pulses with a spacing that can be synchronized perfectly with the “ticks” of an optical atomic clock. (Click here for more on how frequency combs work.) In the experiment, the two combs were synchronized to the same stable optical cavity, which serves as a stand-in for an optical atomic clock. Each comb pulse was sent from one of two locations on NIST’s campus in Boulder, Colorado, reflected off a mirror on a mesa behind the campus, and returned to the other site, traveling a total distance of two kilometers.
Researchers measured travel times for pulses traveling in opposite directions between the two sites. The cumulative timing differences and frequency instabilities were infinitesimal, just one million-billionths of a second per hour, a performance level sufficient for transferring optical clock signals.
The transfer technique overcomes typical wireless signal problems such as turbulence in the atmosphere—the phenomenon that makes images shimmer when it’s very hot outside. Because turbulence affects both directions equally, it can be cancelled out. The transfer technique can also withstand signal losses due to temporary obstruction of the light path. The method should be able to operate at much longer distances, possibly even over future ground-to-satellite optical communication links as an added timing channel, researchers say.
The combs potentially could be made portable, and the low-power infrared light is safe for eyes. The research is funded in part by the Defense Advanced Research Projects Agency.
The National Institute of Standards and Technology (NIST) has published a 33-page special publication reporting on the results of a workshop convened to recommend research and development priorities for alternatives to GPS time distribution in electrical power systems.
