Atomic fountain clocks are emerging as an important new technology for the realization of extremely precise passive atomic standards. The U.S. Naval Observatory (USNO) has undertaken a project to develop atomic fountains for eventual incorporation into the USNO Master Clock.

We have recently demonstrated short term stability of 2x10^-13 at one second with white frequency behavior into the mid 10^-15 s in our first R&D device. We will report on these and any improved results and hope to have frequency measurements relative to internal timescales at the USNO. We will also discuss plans for more heavily engineered operational devices and their incorporation into the USNO Master Clock.

The preliminary design and prototype development of the Stabilized Fiber Optic Distribution Assembly (SFODA) and the Compensated Sapphire Oscillator (CSO) was reported at the 1998 PTTI. These systems were developed to enable gravitational wave searches and atmospheric occultation experiments between the Cassini spacecraft and the NASA Deep Space Network (DSN). These experiments are conducted at Ka band (32 GHz) frequencies and demand the best possible short term and long term stability. A frequency standard located 16 km from the remote antenna acts as the source of the reference frequency. The SFODA is the conveyance device to deliver the reference signal to the remote antenna while the CSO provides the short term stability and low phase noise.

This paper will provide an overview and update of the end-to-end high performance frequency and timing subsystem. Focus will be given to the final SFODA design and test results using a 16 km optical fiber under controlled test conditions. The SFODA utilizes active feedback with a temperature compensating fiber optic reel to compensate for thermally induced phase variations over the 16 km fiber cable. A reference frequency signal at 1 GHz is transmitted over the fiber link to the remote antenna site where the stability of the source frequency standard is maintained. This signal is used to steer the CSO which provides the required ultra-stable reference signal for the Cassini Ka band users. Test data show a factor of 1000 improvement in long term stability when the active phase compensator is used thus enabling degradation free distribution from the highest performing atomic frequency standards.

This work was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration.

A clock that can be used to synchronize 100 GHz digital signal processing electronics while simultaneously providing the time base for high performance phase array antennas and/or high precision Doppler radar must have very little time jitter over time scales ranging from picoseconds to milliseconds. Equivalently, the 100 GHz clock must have very low phase noise from ~10² to ~10¹² Hz. In this paper we describe an approach to satisfying these requirements that is based upon superconducting electronics. An oscillator, based upon the ac Josephson effect and which has excellent intrinsic low-noise characteristics, is used in a Phase-Locked Loop (PLL) to further reduce phase noise or timing jitter. An extremely low-noise superconducting harmonic mixer, similar to those used in state-of-the-art radio astronomy, forms the basis of the PLL phase detector. In addition to overall clock design and measurements made so far, we will discuss related issues such as superconducting device speed and noise characteristics and methods of clock signal distribution that preserve low timing jitter.

The cooperation on TWSTFT between the National Time Service Center of China Shaanxi Astronomical Observatory, the Chinese Academy of Sciences (CSAO) and the Communication Research Laboratory, the Ministry of Post and Telecommunication (CRL) in Japan are going to play a very important role in the International TWSTFT Network.

The CSAO-CRL TWSTFT data has been accumulated for more than one year since the time link was established. The routine results are acquired twice a week and each time with 30 minute time comparison. Because of some trouble with the RF Transceiver and Low Noise Converter (LNC) at CSAO, the TWSTFT had to be suspended from September 1999 to June 2000.

A new step in the project has begun with a new satellite by changing from JSAT 3 to JSAT 1. The Japanese Satellite Company set the antenna for JSAT 1 at CSAO on July 18, 2000. The TWSTFT between CSAO and CRL starts again after a 10 month absence. We have collected excellent data.

The TTR-6 GPS receivers at both CSAO and CRL have timing data for at least four years for the time link in the TAI computation. The GPS common view data are now calculated at CSAO for obtaining UTC(CRL)-UTC(CSAO). The correct coordinates for the TTR-6 antenna at CSAO were introduced recently, which improves the precision of the time comparison.

To compare TWSTT results UTC (CRL)-UTC (CSAO) [by Mr. Hirotaka Yukawa, CRL] with GPS results of UTC(CRL)-UTC(CSAO), we used 55 groups of effective data in the period from January 1999 to August 1999. The profiles of TWSTFT results and GPS results coincide well. However, the larger fluctuations can be seen in the GPS time link. Improvements are expected with the installation of GPS/GLONASS receivers at CRL and CSAO.

The Laser Interferometer Space Antenna (LISA) is a proposed mission which will use coherent laser beams exchanged between three remote spacecraft, to detect and study low-frequency cosmic gravitational radiation. The multiple Doppler readouts available with LISA, which incorporate frequency standards for measuring phase differences between the received and transmitted laser beams, permit simultaneous formation of several observables. All are independent of lasers and frequency standards phase fluctuations, but have different couplings to gravitational waves and to the various LISA instrumental noises. Comparison of the conventional Michelson interferometer observable with the fully-symmetric Sagnac data-type allows unambiguous discrimination between a gravitational wave background and instrumental noise. The method presented here can be used to detect a confusion-limited gravitational wave background.

This work was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration.

The Council for Scientific and Industrial Research (CSIR) is empowered by the Measuring Units and National Measuring Standards Act, 1973 (Act 76 of 1973), as amended, to keep and maintain all national measuring standards for South Africa. It performs this duty through the CSIR - National Metrology Laboratory (CSIR-NML).

The Time and Frequency laboratory of the CSIR-NML is responsible for the maintenance and development of the standards in time and frequency. Specifically, the laboratory is responsible for the following standards: time, frequency, phase angle, pulse rise-time and pulse characterization and time interval. Fiber optic measurements also are the responsibility of this laboratory.

The national measuring standard for time in South Africa consists of two commercial cesium beam atomic clocks. Time transfer is performed using single and multi-channel GPS and GLONASS receivers. The CSIR-NML is the only contributor in Africa of time transfer data to the BIPM.

In addition to its standards activities, the Time and Frequency laboratory also offers time services in the form of a dial-in Telephone Time Service (TTS), as well as a Network Time Protocol service on the Internet. These services are provided free to any user in Southern Africa.

Recent developments at the CSIR-NML include the development and deployment of multi-channel GPS receivers, the deployment of a multi-channel dual frequency GPS / GLONASS receiver and the re-development of the Telephone Time Service.

This paper will discuss the various activities in time and frequency in detail and briefly touch on the other activities of the Time and Frequency laboratory.

In 1998 the Time and Frequency laboratory of the Council for Scientific and Industrial Research - National Metrology Laboratory (CSIR-NML) had to decide on its replacement strategy for the Allan Osborne and Associates (AOA) TTR5 and TTR5A single channel Global Positioning System (GPS) timing receivers. These receivers would have stopped working at the GPS week rollover, which occurred on August 22, 1999 at zero hours Universal Coordinated Time (UTC). It was extremely important to replace these receivers, as they formed the primary traceability link for time in South Africa at that time.

Several options were available to the CSIR-NML. These included replacing the firmware of the AOA receivers, purchasing new timing receivers and designing and building receivers in-house. The replacement of the firmware of the AOA receivers was ruled out due to cost and the age of the receivers. The purchase of new timing receivers was ruled out due to cost. This meant that receivers had to be designed and built at the CSIR-NML. These receivers were put into operation in the week preceding the GPS week rollover.

Several excellent papers are available in the literature on this subject. The significance of this contribution is that the module used for this work is the Motorola UT+ Oncore, and not the Motorola VP Oncore. The software for the CSIR-NML receivers was written at the CSIR-NML under the Windows operating system, while the software for the receivers discussed in the literature was developed using the Linux operating system.

This paper will discuss the hardware used, the software developed and the results obtained using these receivers.

The primary mission of the U.S. Naval Observatory (USNO) Alternate Master Clock (AMC) facility, located at Schriever AFB, is to back up the critical functions of the USNO Time Service Department in Washington, D.C. The USNO AMC operates two master clocks, AMC#1 and AMC#2. Each one of these [Alternate] master clocks is ready to function as the nation's source for precise time, UTC(USNO), should the need arise.

This paper will summarize the current status of, and strategies used for, the steering of these Alternate Master Clocks. The various USNO AMC steering strategies utilize clock comparisons from Two-Way Satellite Time Transfer (TWSTT), GPS Common View (CV), and USNO AMC Timescale data. All current Alternate Master Clock steering strategies employ a combination of Kalman filtering and second-order control, first introduced into USNO operations in 1995.

The respective designs for these steering strategies are based on several factors, including goals for synchronization and stability, as well as the desire for robustness and simplicity of operation. This paper will analyze the performance of these respective designs.

On May 1, 2000, the White House issued a Presidential directive for the Global Positioning System (GPS) to turn off Selective Availability (SA) on May 2, 2000. For nearly a decade, authorized user performance of one-way synchronization via GPS has improved every single year. This paper provides an annual assessment of how well the Global Positioning System can predict and disseminate UTC (USNO) to these specified users, based on data generated and processed by the United States Naval Observatory (USNO). The recent Presidential directive now permits civilian timing users to exploit nearly the same, impressive time transfer accuracy of GPS, these annual metrics now offer a fairly representative performance assessment for both military and civilian timing users.

We have detected the gravitational red shift of a Cesium (Cs) frequency standard that has been transported from Communications Research Laboratory (CRL) Tokyo head quarter, at an altitude of 84 m, to the Ohtakadoya-yama LF standard time and frequency station, located about 250 km from CRL Tokyo head quarter and at an altitude of 794 m.

In the Ohtakadoya-yama Low Frequency (LF) station, two or three Cs clocks are used for the standard frequency reference for radio signal emission, and they are linked to UTC(CRL) by the GPS Common view time transfer method. By using this link, the link of UTC(CRL) to UTC and the published Circular T data, we can compare the frequency of any of the standards in CRL Tokyo and Ohtakadoya-yama LF station with UTC.

A Cs frequency standard, HP5071A normal tube, has been transported by car from CRLTokyo head quarter to LF station on 27 April, 2000. During the transportation, the Cs standard was kept working by the battery.

After the transportation, we have observed that the frequency of the Cs standard become higher in about 9x10^-14 with respect to UTC. According to General Theory of Relativity, 710 m altitude difference will cause a 7.7x10^-14 frequency difference.

Considering the stability of the Cs Standard and the accuracy of time transfer, the observed frequency shift shows a good agreement with the theoretically calculated gravitational red shift.

The NASA Deep Space Network (DSN) is an international network of antennas that supports interplanetary spacecraft missions and radio and radar astronomy observations. The DSN consists of three tracking stations at Goldstone, in California's Mojave Desert; near Madrid, Spain; and near Canberra, Australia. Independent master clocks at each station are synchronized to UTC using traditional GPS common view techniques. Each station contains up to 10 different antennas separated by up to 30 kilometers. Within each station, time code translators (TCT’s) are used to convert distributed frequency and serial time code signals generated from the online frequency standard and master clock into a stable and synchronous 1 pps reference signal for approximately 120 separate users. These TCT’s also compensate for distribution related timing delays.

We report the development of a multi-purpose, automated, and continuously operating time analyzer to measure and monitor distributed 1 pps signals. The instrument consists of a PC to control a VXI based time interval counter and multiplexer with a Labview based user interface and LINUX OS. The instrument performs three major functions needed for operation of the DSN Frequency and Timing Subsystem.

1. Performance tests and monitor of timing offsets and jitter of 1 pps timing outputs of 120 TCT's relative to the station master clock. Time offset data from each TCT is summarized in histogram form. Additional test channels to monitor 1 pps from any other source (e.g.; the GPS time synchronization receivers) against the master clock are also provided.
2. Captures any anomalies and alerts station operators in event of performance or operational failure. Archives timing configuration, station performance, and status.
3. Provides monitor of frequency offsets and long term stability of backup frequency standards with respect to the online frequency standard (which drives the master clock). Time and frequency offset data between online and backup frequency standards are analyzed over a twenty-four hour period and displayed numerically.

The user interface is designed to provide a high level summary of the entire timing system performance and to alert station operations in the event of an anomaly. Raw data is fully archived and network accessible if detailed analysis is warranted.

This work was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration.

The combination of Solar activity and the Earth’s own geomagnetic shielding modulates the energetic proton environment. Over the past three years, the ambient integral proton flux with energy greater than 10 MeV in the free space between the Earth and the Sun has been small and even declining from 2 to 1 particles/(cm^2×s×sr). However, during the same period, there have been at least ten times when the integral flux of protons with energy greater then 10 MeV has greatly exceeded 10 particles/(cm^2×s× sr) by two to three orders of magnitude for periods of time ranging from hours to days. These periods of increased proton flux are a concern to space missions with precision timing applications using crystal oscillators. We present a look at some crystal oscillator on orbit performance during the last three years concentrating on the known high flux times and compare to ground measurements.

Quartz crystal oscillators are arguably the most widely used timing and synchronization sources for microelectronics applications. In general, crystal oscillators have many features which make them very desirable for use in space applications including relatively stable performance, inherent rugged design, low cost, small size, and low power requirements. Over the past 30 years or so, crystal oscillators have been successfully used in many space applications, and we anticipate their use to continue. However, future space applications will demand more timing and synchronization performance to meet mission needs and requirements.

It has been observed that proton radiation introduces fractional frequency drift in measurable amounts that may exceed not only long term system timing drift requirements but also the instantaneous timing drift requirements usually expressed as the fractional frequency drift, f/f, per minute. Therefore in precision timing and synchronization space applications that use crystal oscillators, it becomes important to understand the crystal response to proton radiation.

This is a first look at on-orbit crystal oscillator performance directly correlated with known times of increased proton flux. While oscillator performance has been reasonably acceptable, we anticipate the future timing and synchronization requirements to challenge crystal oscillator performance in the proton radiation environment for space applications.

We will report the current status of millisecond pulsar timing observations at Communications Research Laboratory (CRL). We have been observing a millisecond pulsar, PSR1937+21, for three years by using our 34m antenna, and have obtained the frequency stability of 10^-13 over one year. Millisecond pulsars are known to have extremely stable pulse timing in long term, and expected to be used as new sources of frequency standards. The Communications Research Laboratory (CRL) aims to apply millisecond pulsars to the construction of a new time scale, and perform research on millisecond pulsars; theoretical research, VLBI observation, and timing observation.

As for the timing observation since November 1997, we have developed an original observation system using the 34m antenna at Kashima Space Research Center, and started the regular observation of PSR1937+21, which is the brightest millisecond pulsar in the northern sky. Regular observations have been carried out once a week, eight hours per day.

Although the observed pulse phases have shown systematic trends over a day and in the long term (reported in PTTI'98), we have checked our observation system and analysis program carefully. Through these checks, recently we have succeeded in removing the systematic trends. The phase shifts in one day were removed by a correction of the earth rotation parameter. The trend in a long term was improved by using the DE200 ephemeris for the analysis in a consistent way. Some other causes of the strange trends, such as the program bugs and the insufficient corrections of the observation time and frequency, have also been corrected.

Finally we succeeded in estimating the pulsar parameters of PSR1937+21 from our own data with the timing precision of 6.4 microseconds. In our analysis, the frequency stability of PSR1937+21 is 10^-13 at averaging time of one year, which shows the possibility of constructing the pulsar time scale with even a rather small antenna.

We will present our observation system, analysis methods, and observation results of PSR1937+21 for about three years.

For three or four years, the Motorola VP oncore GPS receiver has been utilized in several experiments that have proven its ability to achieve time transfer with a precision roughly equivalent to those offered by other commonly used time-dedicated receivers. Motorola's decision to replace the VP by the UT+, has several important drawbacks among these are:

1. Impossible to follow the standard BIPM common view schedule
2. No access to the raw navigation data

In this paper we present the results of the time transfers performed via UT+ receivers located at the Observatoire de Paris (OP) and at the Observatoire de Besançon (OB) which are 324 km apart. Both sites operate HP5071A-001 cesium clocks.

First, measurements made in a "melting pot" mode, all visible satellites being tracked at the same time by two stations will be presented.

In a second step, we will show how it is possible to overcome the first point for two stations up to around 1000 km apart.

Data collected between OP and OB will be presented; for both methods data consist in standard 13 minute measurement sessions. For the melting pot, data acquisition were synchronized with the standard BIPM schedule. For the common-view data, around 60 sessions were made daily, but at non-standard dates.

The two methods show very similar results, with a small advantage to the common-view method in terms of precision. These results are also compared to those simultaneously obtained at the same sites using time-dedicated GPS receivers.

GPS timing signals play a critical role in the modern practice of time error estimation and synchronization. The major difficulty in this practice is the large noise on the measured data provided by the GPS receiver that requires the use of the proper filtering method to achieve adequate accuracy. Statistically, the most accurate results are achieved by the use of Kolmogorov-Wiener filtering (Wiener filter) for stationary random signals and Kalman-Bucy filtering (Kalman filter) for non-stationary signals. A non-optimal smoothing filter is used as the simplest way from an engineering point of view.

Despite the differences in theoretical application of the filters (stationary and non-stationary), GPS applications require a practical method of separation. What processes may be practically considered as stationary and what processes considered as non-stationary?

In this report we answer this question and obtain a practical method of separation of the stationary and non-stationary time error signals based on the variance and the rate of the time process. Simulation shows that, for the same transient time, the following filters should be used to obtain the best accuracy, depending on a fractional frequency offset (time error rate) y₀ for times where:

We determine the critical offsets r₁ and r₂. These offsets are dependent on the filter transient.

Digital processing of a GPS-based time error signal is carried out simultaneously by all three filters. Depending on y₀, the proper estimate should be taken as the most accurate. It follows that the Smoother and Wiener filters give the most accurate estimates for the relatively stationary signals of a precise synchronization system and the time error of an atomic oscillator with a small frequency offset. On the other hand, for a large initial offset for the tested quartz oscillator, the Kalman filter provides the best results.

We present comparative results for all three filters for rubidium and crystal oscillators based on the reference timing signals of the Motorola GPS UT+ Oncore Timing Receiver. Many examples of the original and filtered processes are discussed.

Real Observatorio de la Armada (ROA) has finished the installation and set-up of a Two-Way Satellite Time & Frequency Transfer (TWSTFT) station to join the European and Trans-Atlantic Links.

In this paper we will provide a description of the main components of the station and will present very preliminary results derived from the recent incorporation of the TWSTFT system.

In this paper, the clock synchronization via GPS/GLONASS (Global Positioning System/GLObal Navigation Satellite System) is proposed. By performing the double differences on carrier phase observables, the remote Oven-Controlled Crystal Oscillator (OCXO) clock can be synchronized with the primary cesium atomic clock. Two GPS/GLONASS receivers with the external frequency option are used in our system. While the remote OCXO clock and the primary cesium atomic clock are connected to the receivers, the frequency offset of the remote clock with respect to the primary clock can be estimated via the double differences on carrier phase observables. The Proportional-Derivative (PD) controller is adopted in our system for tuning the OCXO in real time. Through the D/A converter, the remote clock is then steered to synchronize with the primary clock. For averaging times of one day under the configuration of about a 30-meter baseline, our experimental results show that the accuracy of the remote clock can be improved from about 1x10^-10 to about 3x10^-14, and the stability of the remote clock can be improved from about 2x10^-10 to about a few parts in 10¹⁴. Comparing with the GPS Disciplined Oscillator (GPSDP) system, the scheme adopted in this paper can achieve the traceability of frequency dissemination. The potential applications of the proposed architecture include the frequency sources system for calibration laboratories, telecommunication networks and power transmission system.

More than two decades have now passed since the construction of the GPS Operational Control Segment (OCS), including the Master Control Station (MCS) at Schriever AFB, Colorado and worldwide monitoring stations, and launch of the first GPS satellites. During that time, several generations of satellites have been designed, built, and launched, and operations are routinely supported by the OCS.

Today, the Global Positioning System (GPS) is entering into the new millennium with a far-reaching program to modernize the system on the ground and in space. This paper describes the goals, requirements, design objectives, and plans for implementing a $1 billion-plus modernization program in the coming decade.

The National Institute of Standards and Technology traceable measurements of a passive hydrogen maser, GPS receiver, rubidium standard and other elements of a new Primary Reference Clock being developed with European Union Assistance are reported. A new measurement system is detailed and first noise floor results are also reported.

This briefing discusses the impact of incorrect implementation of the United States of America Chairman of the Joint Chiefs of Staff timing policy. The brief systematically addresses each policy area and shows the associated level of compliance. Finally, with regard to the United States Navy, calls for the implementation of a common time reference making time a utility on board U.S. Naval vessels.

The Global Positioning System (GPS) has become the primary and most accurate means of disseminating time and frequency information. A growing and diverse mix of military positioning, communications, sensor, and data processing systems are using precise time and frequency from GPS. The precise accuracies required for their operation is also becoming more stringent. A new system architecture for providing a Common Time Reference to the operating forces and their related subsystems is being developed. This architecture is to provide a robust enhancement to implementations of GPS time and frequency subsystems. Through its implementation, interoperability between systems are to be enabled by providing a Common Time Reference and the means for systems to operate synchronously as a foundation for common interfaces and data exchange.

The Common Time Reference approach and its relationship to present GPS time and frequency usage will be described to show the concept and approach of this architecture. A robust architecture utilizing distributed time standards and existing standards within individual systems is to provide new capabilities for existing fielded systems without the impact of requiring a major retrofit. This combination of enabling existing and new resources to be used in a common reference will reduce the sensitivity to GPS anomalies and lack of continuous contact for precise updating. These systems would then be interconnected at the fundamental level of internal time and frequency generation, which would provide an inherent basis for functional interoperability. The elements necessary for implementation of this architecture with generic systems will be discussed. Technical developments necessary for implementation of the concept and the impact on interoperability will be discussed.

The current GPS has exceeded its globally averaged position and timing accuracy of 16 m (50 % spherical error) and 100 ns (one sigma) as stated in the 1990 GPS System Operation Requirements Document (SORD). The 1999 GPS Operational Requirements Documents (ORD) set a new goal for the GPS III and beyond. The 1999 ORD specifies the position and timing accuracy at any location as: range rms accuracy threshold and objective of 1.5 m and 0.5 m, respectively, and 95 % time transfer accuracy threshold and objective of 20 ns and 10 ns, respectively. This paper will evaluate how the current clocks and the clocks being developed can support the ORD threshold and objective. The paper will include the following topics: (1) Atomic clocks on the GPS Block II space vehicles, (2) Estimated accuracy of the IIF Rb clock by Perkin Elmer and digital Cs clocks by Datum-Beverly and assessment of their performance against the ORD threshold range requirements. (3) Description of the new space clocks being developed jointly by the GPS JPO, Aerospace and NRL, and evaluation of their predicted performance to see if they can support the ORD objective of 0.5 m (rms). (4) Prediction of the GPS signal-in-space accuracy, including all the space and control segments errors, using IIF Rb and Cs clocks. The predictions are based on replacing the National Imagery and Mapping Agency (NIMA) estimated GPS II/IIA/IIR clock data, contained in the actual tracking data of the GPS monitor stations and the NIMA tracking stations, by simulated IIF Rb and Cs clock data. A Kalman filter similar to that of the Operational Control Segment (OCS) then processes the resulting tracking data and the estimated results are compared with NIMA estimates treated as truth. (5) Evaluation of the various options to see whether the ORD objective can be achieved based on the predicted signal-in-space accuracy.

The Global Positioning System (GPS) signal is now the primary means of obtaining precise time to an internationally accepted standard. Precise timing applications have become dependent on this space-based source of precise time and, therefore, depend on the constellation of satellites that provide it worldwide, anytime. This paper describes the efforts by the GPS Joint Program Office within the U.S. Department of Defense to modernize the GPS signal services to meet future military and civil user requirements. GPS timing users and timing receiver developers and integrators need to be aware of these new capabilities and when they will be available. This paper starts with a brief review of the system design and an overview of the current constellation status. The GPS Modernization program to modify the current block of satellites being placed into service and the next generation currently in design to provide additional system capabilities will be described. Next, the paper discusses the GPS-III program to look at future user requirements beyond the next twenty years for precise positioning and timing services. The paper summarizes what these new capabilities will mean to the GPS timing users and provides some suggestions on what GPS timing users can do to make their future needs known. The paper concludes with some challenges to the user community to support the continued mission of GPS to provide precise positioning and time to all users free of direct charge.

Very narrow band Global Positioning System (GPS) receivers have recently been proposed in order to help mitigate the effects of scintillation. This paper will re-visit the stability requirements placed on the local oscillator of such receivers.

Internet timekeeping has come a long way since first demonstrated almost two decades ago. In that era most computer clocks were driven by the power grid and wandered several seconds per day relative to UTC. As computers and the Internet became faster and faster, hardware and software synchronization technology became much more sophisticated. The Network Time Protocol (NTP) evolved over four versions with ever better accuracy now limited only by the underlying computer hardware clock and adjustment mechanism.

Several years ago the software algorithms to discipline the Unix system clock were overhauled to provide improved accuracy, stability and resolution. In addition, a means was added to discipline the clock directly from a precision timing source, such as a cesium oscillator. The software was integrated with several operating system kernels of the day and eventually adopted as standard in Digital Alpha, Sun Solaris, Linux and FreeBSD. The performance achieved with workstations of the day was a few hundred microseconds in time and a PPM or two in frequency, so the resolution of one microsecond seemed completely adequate.

With PCs and workstations and networks of today reaching speeds in the gigahertz range, it is clear the solution of several years ago is rapidly becoming inadequate. This paper describes new algorithms and implementations providing a 1000-fold improvement in time and frequency resolution, together with a more agile and accurate precision discipline. The paper discusses the analysis and design of the algorithms, implementation details, and the results of proof-of -performance experiments. A testbed using a 450-MHz Intel PC with an uncompensated clock oscillator, but disciplined to a rubidium oscillator, has achieved RMS jitter less than 50 ns. The software has been implemented and tested in Digital Alpha and SunOS operating systems and is now distributed with Linux and FreeBSD.

Although Coordinated Universal Time (UTC) is the time scale that is transmitted by almost all time services, this scale is awkward to use in the vicinity of a leap second. Many computer systems cannot represent the epoch corresponding to a positive leap second (23:59:60), and remain synchronized to UTC by stopping the clock at 23:59:59 for 1 extra second whenever a leap second is to be added. This makes it impossible to assign unambiguous time tags to events that happen during this period. In addition, computing the length of a time interval that includes a leap second of either sign is difficult because simply subtracting the two UTC time stamps at the end-points of the interval does not account for the time interval occupied by the leap second itself.

To address these issues, we have augmented the Network Time Protocol (NTP) to allow a client system to reconstruct TAI from UTC and a table of leap seconds. This time scale has no discontinuity during the leap second. Intervals computed using TAI are unaffected by the additional time occupied by the UTC leap second, and the TAI time scale provides an unambiguous time tag to any event -- even one that happens during a UTC leap second. Although our solution is unique to servers that support the Network Time Protocol (NTP), it could be adapted to other time services and formats. Such systems could support time tags using either UTC or TAI, and would significantly reduce the problems that result from using UTC alone.

The U.S. Naval Observatory (USNO) has been contracted by the Defense Information Systems Agency (DISA) for the production of Network Timing Protocol (NTP) Stratum 1 network time servers for worldwide deployment. These will be based on the design of the present high-performance USNO NTP servers, capable of handling greater than 100 NTP requests/second. Significant improvements will include integration of the Rockwell-Collins MPE 1.12 dual-frequency keyed GPS module and a TEMEX RMO-1 internal rubidium. The P(Y) code receiver meets DoD requirements for military applications, and will be integrated into Brandywine's VME synchronized generator. The rubidium will allow sustained NTP service in the event of loss of GPS signal for up to a year with 1-millisecond flywheel accuracy.

This paper discusses a new technology of synchronizing clocks and disciplining oscillators using CDMA cellular transmissions. Like cellular telephones, these receivers will operate in most buildings without rooftop antennas. They are reported accurate to within microseconds to UTC with stable frequencies available.