The motto of this year's PTTI meeting is "31 years of progress". In a certain sense, there has been astonishingly little progress, namely in the areas of practical clocks and the basic ideas about their use. Although much science advance has been reported in regard to novel quartz crystals, trapped ion and atom configurations, and lasers; the practical clocks in use are still the cesium, rubidium, hydrogen and crystal clocks already in use 31 years ago, albeit better engineered. Furthermore, the ideas of digital communication and time and frequency based global navigation had been well-developed 31 years ago. However, the combination of what was known and largely dreamed about 31 years ago with the advances in microelectronics and computing of recent times, has catapulted time and frequency into a sweepingly pervasive role in today's communication and positioning (navigation) systems.

  What a revolution it has been! Our daily lives have become richer and more convenient; local, regional, and global information, flowing at incredible speeds, is now routine. Political decisions and military threats, and even war itself have fundamentally changed.

  This talk will paint a picture of these changes that were enabled by time and frequency technology, and provide examples ranging from clocks in the home, to hiking and driving in uncharted areas, to the reduction of the threat of weapons of mass destruction.

  In the 1990's, the number of military and civilian government agencies who have become dependent on GPS for accurate timing and navigation has consistently increased. The GPS Operational Control Segment, using information provided by the United States Naval Observatory, maintains the GPS timing signal well within specifications. This paper summarizes the 1999 performance of One-Way (Direct-Access) GPS time transfer for authorized users. Data from previous years will also be presented as a means of comparison. Additionally, the paper briefly covers some recent GPS Master Control Station (MCS) activities affecting the GPS signal.

  Time Transfer Users who make use of GPS for one-way synchronization benefit from being able to operate autonomously and in anonymity. Though autonomous, anonymous operations can prove advantageous, in particular, for activities associated with the military, their anonymity can result in the undesirable side effect of lost connectivity with the GPS community. Recent GPS satellite maintenance activities have exposed some unusual characteristics of certain, but not all, types of GPS time transfer receivers. As a result, many Department of Defense and other government agency users affected by these satellite activities have emerged to provide the GPS community (2 SOPS in particular) with feedback concerning their experiences with GPS Time Transfer. This paper presents general descriptions of the types of users who have emerged, as well as general observations on the received feedback.

Wilson G. Reid
SFA, Inc.
Naval Research Laboratory
Washington, D.C.

ABSTRACT

  Analysis of the on-orbit Navstar clocks and of the GPS monitor station reference clocks is performed by the Naval Research Laboratory [1] using both broadcast and post-processed precise ephemerides. The precise ephemerides are produced by the National Imagery and Mapping Agency (NIMA) for each of the GPS space vehicles from pseudorange measurements collected at five Air Force and at ten NIMA GPS monitor stations spaced around the world. The time reference at the NIMA Washington, D.C. site, co-located with the U.S. Naval Observatory precise-time site, is the DoD master clock. Hence, it is possible to transfer time via linked common-view every fifteen minutes from the DoD master clock to the remaining fourteen Air Force and NIMA GPS monitor stations. Linking stations serially, although computationally efficient, was found to suffer several disadvantages. The most serious of these was the accumulation of gaps in the time transfer to a remote site. A secondary disadvantage of the serial linking was the dominance in the linking process of the short-term noise from a single noisy site. Using multiple common-view paths overcomes both of these defects by supplementing the link that propagated the gap in the data and by providing multiple independent measurements that can be averaged to reduce the measurement noise. An additional benefit of the multiple-path method is an improvement in the continuous coverage of the Navstar space vehicle clocks; i.e., in referencing the observations of a space vehicle clock by each of the monitor stations back to the DoD master clock. Improvement was manifested by an increase in the number of estimates obtained of the offset of the space vehicle clock from the DoD master clock and by a decrease in the short-term noise of these estimates.

  Since the launch of the first GPS Block IIR SV (SNV43) on July 22, 1997, there have been four reported Block IIR Timekeeping System (TKS) anomalies in which non-standard codes were transmitted by the Space Vehicle (SV). ITT analyzed the first two events, which occurred on December 10, 1997 and February 9, 1998, and conjectured with limited data that large frequency jumps in the Rb clock output were the cause of both problems. These two anomalies occurred outside of the normal daily contact windows by the ground control, so no recorded telemetry is available to accurately identify or simulate the cause of these. To remedy this shortcoming, the satellite software was modified in May 1998 to store key data in several circular buffers in a plus/minus 150-second window around a TKS transient. These buffers in the SV computer memory can be dumped at a subsequent SV contact time, to the ground, for analysis. Since the upload of the modified SV software, two TKS anomalies were reported on August 16, 1998 and January 29, 1999. This paper presents the TKS simulations performed to reproduce these last two events, and will show that these two events were most likely caused by VCXO transients. Recorded data will indicate that VCXO transients are highly correlated with eclipses. The paper will also suggest ways to prevent or curtail anomalies caused by VCXO transients.

  Since its activation in November 1995, we have been following the performance of the rubidium (Rb) atomic clock carried onboard the Milstar FLT-2 satellite. The Milstar communications system is meant to provide secure antijam communications for DoD operations into the next century, and to accomplish that mission Milstar employs precise timekeeping onboard its satellites and at its mission control ground stations. FLT-2 is the second Milstar satellite to become operational, and is the master clock for the crystal oscillator clocks carried onboard Milstar FLT-1, the first Milstar satellite. The FLT-1 clocks tie their time and frequency to FLT-2 via timekeeping information passed along satellite cross-links [1].

  Using frequency measurement data, we demonstrate the excellent performance of the FLT-2 clock in its geosynchronous orbit. Specifically, we have measured the clock's linear frequency aging rate and its random-walk Allan variance at +7.0 x 10^-14/day and 1.0 x 10^-15 t ^1/2, respectively, where the Allan variance measurements extend to t ~10⁷ seconds. For roughly the first six months of the Rb clock's operation, the deterministic changes in the clock's frequency (i.e., its frequency drift) were not truly linear. Though the frequency drift appeared to be linear over periods of weeks, when viewed over periods of months it was clear that the linear drift coefficient was slowing down in something like an exponential fashion. The exponential time constant for this frequency drift "slow-down" was about two months, so that after roughly half a year of operation the Rb atomic clock achieved its "true" quiescent linear frequency aging rate noted above. We believe that this drift slow-down may be typical of all Rb clocks, and related to the device’s light shift effect.

  Analysis of the performance of all on-orbit Navstar Space Vehicle (SV) clocks and Global Positioning System (GPS) monitor station reference clocks is performed by the U.S. Naval Research Laboratory in cooperation with the GPS Master Control Station. The measurements are collected by multi-channel GPS receivers located at the Air Force and National Imagery and Mapping Agency (NIMA) GPS monitor stations. The offset of each Navstar SV clock, computed every 15 minutes, is referenced to the Department of Defense Master Clock. The resultant Navstar SV clock offsets are then used to compute frequency offset, drift, frequency stability profiles, and frequency stability histories. Frequency offset histories, with superimposed Navstar eclipse seasons, are presented to analyze the Navstar clocks for possible temperature effects. The beginning-of-life, steady state and end-of-life performance of selected cesium and rubidium atomic clocks is presented. Frequency stability results are presented using sample times that vary from 15 minutes to several days. The stability results for sample times of less than one day are used to characterize the measurement noise and periodic effects in the on-orbit data, while sample times of one day, or more, are used to report the long-term performance of the Navstar atomic clocks.

  Satellite-based navigation would be impossible without the synchronization of the system inherent clocks. Obviously one can find high accurate precision standards (cesium and rubidium standards, H-maser) onboard the satellites and in the overall ground segment. This paper outlines some basic considerations for the conception of the timekeeping system of the GALILEO satellite navigation system to be developed. This European contribution towards a 2^nd Generation Satellite Navigation System (GNSS2) is required to be compatible and interoperable with GPS but independent of it. These features have to be considered in the overall planning from the beginning.

  This paper reflects some architecture aspects as to the number of GALILEO satellites and ground clocks as well as clock types which are foreseen. One of two implementation strategies on this way has to be selected for the system time: a master or composite clock, the latter having some certain advantages. Due to international recommendations the new GALILEO time has to be in close agreement with UTC, which requires a steering procedure. The offset versus UTC has to be broadcast to the users by means of the navigation message. For the system provider this offset is also requested. Similarities and differences versus GPS-time are discussed. The establishment of GALILEO can lead to some qualitative improvements with respect to GPS, which is discussed here. The objective for GALILEO is to implement, at least, an equivalent or a better timekeeping system compared to GPS to allow for a comparison with the present worldwide accepted GPS standard. It is well understood that this is a great challenge for this European-based development. The paper aims to give some insight in the present planning to the world-wide concerned timing community and invites inputs to be given to the European system design team.

  This paper describes a redundant Timekeeping System (TKS) that utilizes multiple Atomic Frequency Standards (AFS) to provide an adjustable 10.23 MHz frequency output derived from one of four 5, 10, or 13.4 MHz reference AFS's. Phase and frequency adjustment is accomplished without disturbing the input AFS. Other reference and output frequencies can be accommodated with minor modifications to the design. The TKS has the capability of seamless switchover in case of primary AFS failure. Seamless switchover is defined as the maintenance of both output phase and frequency continuity while the unit: (1) detects a primary AFS failure, and (2) switches-over to a hot spare AFS. This feature has applications both in space and ground timekeeping systems where continuous predictability after an AFS failure is desirable. The paper also describes a simulation that demonstrates the dramatic improvement in predictability after an AFS failure when seamless switchover is utilized in the GPS system.

  This paper first describes the structure and theory of the TKS and provides measured results. The TKS consists of four AFS's and a redundant Frequency Synthesizer Unit (FSU). In each FSU, a high isolation switch selects a primary and hot back-up AFS from the four AFS units. The selected AFS’s are routed to two programmable downconverters, which heterodyne the AFS frequencies to 100 Hz IF's utilizing the 10.23 MHz output of a precision voltage controlled crystal oscillator (VCXO). A microprocessor, uses the 100 Hz signals to determine the phases of the two AFS's relative to the VCXO to a resolution of 1 ps (1.3 ps measured jitter). The processor then utilizes this information to digitally phase lock the VXCO to the primary AFS while allowing the VCXO to be offset in phase with 1 ps resolution and in frequency with 0.02 m Hz resolution. The microprocessor also utilizes the measured hot spare AFS phase difference to model the hot spare's phase and frequency offset relative to the primary AFS. This information is utilized by the microprocessor to correct the VCXO output phase and frequency when a primary AFS failure is detected. Thus, switchover to the hot spare is accomplished with a measured transient phase glitch of only 50 ps (0.1 s time constant) and no observable permanent phase or frequency offset (well below the AFS noise level).

  Next, the paper describes a simulation of the difference between seamless and non-seamless switchover in the GPS system with Kalman filtering. Noise models that accurately reproduce actual GPS AFS Allan variances are utilized in the simulation. The simulation demonstrates the dramatic improvement in predictability after an AFS failure when seamless switchover is utilized. For the non-seamless case, the simulation shows that no meaningful navigation message can be formed for a 10-hour period after a switchover and errors do not return to pre-failure levels until after 24 hours. For the seamless switchover, no detectable performance degradation is observed during and after switchover.

  The mature satellite-based navigation systems that are now available (e.g.,GPS) have provided adequate positioning capability to users. However, their very success has been a driving force to increase accuracy, availability, reliability and integrity requirements. As a result, several satellite augmentation systems have been or are in the process of being designed, developed and/or tested in order to meet the ever-demanding requirements. This paper will provide a summary of the current possibilities to improve GPS performance, namely the impact of the GPS modernization program itself, augmentation with the satellite-based GLONASS, WAAS, MSAS and EGNOS systems, and augmentation with on-board aiding; e.g., barometers and clocks. Performances are discussed as a function of user mask angle. The impact of combined GPS/GALILEO is briefly addressed.

  Global Positioning System (GPS) navigation signals alone are not adequate to support aviation navigation. GPS accuracy is acceptable for all but precision approach applications, but integrity, continuity, and availability are lacking. The Wide Area Augmentation System (WAAS) hardware has been installed and is used for testing since last summer. Wide Area Reference Stations receive GPS signals and forward them to Wide Area Master Stations for corrections processing. Augmentation messages are sent to the Ground Uplink Stations for transmission to Geosynchronous (GEO) satellites that retransmit the messages to avionics receivers in aircraft. WAAS is a safety critical system, meaning that time to alarm is critical and transmission of hazardously misleading information is not allowed. Phase 1 WAAS will be commissioned in September of next year and will provide en route through non-precision approach services throughout the U.S. Flight Information Region. Precision approach services will be provided over roughly 50% of the continental U.S. END-state WAAS will provide CAT-1 precision approach navigation service throughout the continental U.S.

  The key to all WAAS operations is time. The GPS satellites, the GEO satellites, and all WAAS hardware suites must agree upon time. Precision time has been identified as one of the major untapped byproducts of WAAS.

  Raytheon Systems Company is currently designing the algorithms of the FAA’s Wide Area Augmentation System (WAAS). This paper will briefly overview the WAAS Network Time (WNT) algorithms and their integration into the overall system.

  The WAAS system is required to provide users with orbit, clock, and ionosphere corrections for single-frequency measurements of the GPS signal. The clock corrections will be made in the WNT reference scale. This reference scale is required to track the GPS time scale, while at the same time providing the users with the translation to UTC. Other internal requirements are placed on the WNT bias and frequency jumps from the accuracy demanded in the WAAS fast corrections.

  The WAAS architecture requires that WNT be computed at multiple WAAS Master Stations (WMS) using potentially differing sets of measurements from potentially differing sets of receivers and clocks. This design must provide the user a seamless transition when switching from one WMS's WNT realization to another.

  This paper expands on two previous papers, one presented at ION GPS-97 using simulated clock data as the input, and the other presented at the 1999 Technical Meeting using receiver data from the National Satellite Test Bed (NSTB) and preliminary WAAS data. We included some introductory sections from previous papers here for completeness. Here, we will present the performance results of the WNT algorithm using GPS clock and receiver data from WAAS Reference Station (WRS) using the standalone WNT algorithm with the WAAS prototype software. We will also show some recent WNT algorithm results from the operational WAAS software, indicating that the WNT design should meet the aforementioned requirements, taking into account possible WAAS WMS switches.

  Two tests are performed. Using two different subsets (taken from a single WRS thread to simulate two WMS networks) of the WRS sites, we demonstrate that coordinated WNT as computed at the two different networks closely agree (on WNT), thereby ensuring a seamless transition from one WMS network to another. We will also show how coordinated WNT behaves in the operational system with two actual WMSs. Next, we test how well WNT, for a single WMS network, agrees with estimated GPS time in the operational as well as in the standalone/prototype WAAS software.

  Raytheon Systems Company is currently designing the algorithms of the FAA's Wide Area Augmentation System (WAAS). This paper will discuss the GUS clock steering algorithms, performance, and validation and test results.

  The WAAS Wide Area Master Station (WMS) calculates WAAS Network Time (WNT) and clock parameters (offset and drift) for each satellite. The GUS clock is an independent free running clock. However, the GUS clock must track WNT (GPS time) to enable accurate ranging off the GEO signal-in-space. Therefore, a clock steering algorithm is necessary.

  The GUS clock steering algorithms reside in the WAAS Message Processor (WMP). The WAAS Type 9 messages (GEO Navigation Message), which are provided by the WMS, are used as inputs to the GUS WMP.

  The GUS WMP calculates clock adjustments. Based upon these clock adjustments, the frequency standard can be made to speed up or slow down the GUS clock. The GUS cesium frequency standard is controlled through very small frequency control signals so that the normal operation of the code and frequency control loops of the downlink signals will not be disturbed.

  Simulation and field results for the primary GUS will be discussed in the paper. The preliminary field results from AOR are that the systems settled to ± 100 nanoseconds with corrections occurring every 15 minutes after 70 hours shown.

  Accurate timing sources are becoming a very important issue in the development of networked telecommunication systems. Since the early 1990's, GPS has been exploited for this purpose. The common-view GPS time transfer technique is mostly used to minimize the timing error caused by satellite clock and ephemeris errors, and Selective Availability. This technique provides a typical timing accuracy of 50 nanoseconds (1 sigma).

  Currently, a new Wide Area Augmentation System (WAAS) is being developed under the authority of the Federal Aviation Agency (FAA). This system is a Satellite Based Augmentation System (SBAS) which will be used to enhance signal continuity, availability and integrity to GPS receivers. WAAS is scheduled to be officially commissioned in the summer of 2000.

  This paper describes the features and performance of a GPS/WAAS Timing Engine developed by Marconi Canada. The paper will discuss the features of the WAAS system and how it can be used to dramatically decrease the timing errors of a GPS engine below 10 nanoseconds. Results obtained using a GPS simulator and live signals will be analyzed. Comparative results between a GPS only and a GPS/WAAS Timing Engine will be presented. Finally, additional features of the GPS/WAAS Timing Engine, such as Time Receiver Autonomous Integrity Monitoring (TRAIM), will be discussed.

  The FAA is currently developing a Wide Area Augmentation System (WAAS), primarily for en route navigation. The system consists of a multitude of ground reference GPS systems and several geostationary communication satellites. The geostationary satellites will be broadcasting a "GPS"-like signal as well as a GPS health status and correction data. This new system offers a significant tool to the timing community.

  There are several unique advantages of using this system for frequency and time transfer:

• The geostationary satellites are always in view. They never set like the GPS and GLONASS satellites. This offers the ability for multiple sights to be permanently "phase locked" together by locking themselves to the common WAAS signal.
• They do not have any intentional distortions like the GPS SA dither.
• All of the dual frequency reference GPS receivers in the WAAS network are cesium controlled.
• The geo-broadcast signals are cesium based, generated and controlled by the WAAS network of cesium based GPS receivers.
• The WAAS network computes an ionosphere distortion model for the coverage area. This model is broadcast over the WAAS signal so that the users can make local corrections.

  This paper looks at a time transfer attempt using the AOR-W WAAS signal between the master clock at USNO Washington and the master clock at NRC Ottawa. A 30-day data set is examined from dual frequency GPS/WAAS NovAtel receivers that are set up at both institutions. In addition, an attempt is made at comparing GPS and AOR-W time through a standalone solution based on post-mission precise GPS orbits and clock corrections. The results will be presented.

  With the advance in accuracy and stability of modern atomic clocks, the need for a very precise method of detecting instabilities in their signals has arisen. For example, with two active hydrogen masers with a 1 x 10^-15 frequency difference, the drift rate of phase due to the frequency difference will be 1 fs per second or 3.6 ps per hour. No ordinary universal counter would be capable of resolving the difference between the two signals.

  In response to this need, a state of the art, computer based very high-resolution frequency and time interval measurement system (A7) has been developed. The measurement system integrates the most advanced phase Comparators with modern PC time interval-counting techniques.  The software consists of Stable32TM routines capable of measuring both first and second difference variances.  Internally the system also has a high stability rubidium oscillator and a high isolation 4 output distribution amplifier.

  With a hydrogen maser reference in a temperature controlled room, the A7 specifications state a short-term stability (Allan Variance) of 1.5 x 10^-13/gate time, resulting in 1.5 x 10^-13, 1.5 x 10^-14 and 1.5 x 10^-15 for 1s, 10s and 100s gate times (t). Initial results suggest even better performance than this with Allan variances of 5 x 10^-14, 8 x 10^-15, 9 x 10^-16 and 3.5 x 10^-16 for 1s, 10s, 100s and 1000s gate times (t). The Noise Floor is reached after about 1000s.

  Constant ambient temperature results in a typical drift of <2 ps/hr. A 1°C temperature change adds less than an additional 10 ps. A single shot rms resolution of 0.3 ps was also measured, enabling the A7 to easily resolve the 3.6ps/hr drift rate between two hydrogen masers.       

  The primary benefits of the A7 are improved accuracy and reduced measurement time. Fast measurements with high accuracy permit greater knowledge of the stability of the signal.

  The applications for a measurement instrument capable of such resolution are anticipated to be numerous, ranging from national standards and calibration laboratories, through Cesium, Rubidium and Quartz production to time transfer measurements.

Further work on reducing the size and improving the resolution is being carried out.

This paper shows how to distribute GPS-time with microsecond accuracy and below even in Ethernet-based distributed systems. Our approach rests upon a simple network controller-level hardware for time-stamping data packets as they leave and arrive at a node, which comes in two types: (1) the widely applicable memory-based time-stamping method exploited by our Network Time Interface (NTI) M-Module time-stamps data packets as the network controller accesses them in memory, and (2) for 10 Mb/s Ethernet, our experimental evaluation revealed a time distribution accuracy down to the microsecond range. Still, memory-based time-stamping can only be employed in conjunction with network controllers that do not store entire packets on-chip, and the available configuration parameters must be carefully chosen in order to cope with the various hidden sources of timing uncertainty. To escape from those limitations, we recently developed a novel MII-based time-stamping method that can be used in conjunction with almost any modern 10/100 Mb/s Ethernet chipset. The time-stamping hardware resides at the standard Media-Independent Interface (MII) between media access and physical layer here, which eventually leads to a time distribution accuracy well below the microsecond range.

  This paper introduces the NAV2300, a 12-channel, single frequency, C/A code receiver based on Analog Devices fixed point family of Digital Signal Processors (DSP). The receiver is capable of generating accurate timing outputs synchronized to either GPS or UTC to support a host of applications that require time synchronization.

  The NAV2300 receiver hardware comprises two principal blocks. A standard RF down converter and a GPS Digital Signal Processor (DPS). The DPS is equipped with programmable input/output flags that have been employed for generating the timing pulses.

  The NAV2300 GPS Receiver is capable of emitting a timing pulse with the rising edge of the pulse synchronized to either GMT or UTC. The timing pulses are generated within a minute of power up of the receiver even without the presence of the primary estimates such as almanac, position, time and date. With the presence of the primary estimates and valid ephemerides, the timing pulse is output within 20 seconds of the receiver power up. The accuracy obtained is comparable with the specifications quoted for a typical C/A code receiver despite a low cost crystal as the reference. The performance of the receiver can be enhanced by using a higher stability reference clock such as an OCXO.

  All figures quoted in the document have been obtained against rubidium and cesium sources. Tests have also been performed against P code receivers from Allen-Osborne. The timing accuracy obtained from the NAV2300 is a new benchmark in C/A code receiver technology.

  The NAV2300 is being targeted at the automobile and hand-held segment as a reliable and accurate time source.

  Military Satellite Communications (milsatcom) systems require precise timekeeping in order to take advantage of spread spectrum communication techniques. Though the level of precise timekeeping in milsatcom is typically not as stringent as that in satellite navigation, milsatcom nevertheless poses its own unique timekeeping problems. Specifically, milsatcom timekeeping must be robust, with the ability to autonomously detect and correct timekeeping problems during protracted periods when the ground control station is either not available or burdened with other pressing tasks. Here, we consider the ability of three different space-segment timekeeping subsystems to autonomously detect the failure of a single satellite clock and to then take appropriate action to remedy the situation. These systems include a Master/Slave system (similar to present day Milstar), an ensembling system (based on NIST’s AT1 algorithm) and a Kalman-Filter system (similar to GPS when it goes to cross-link ranging). Employing Monte Carlo simulation, we consider four types of "soft" clock failure: a time-jump failure, a frequency-jump failure, a failure arising from a sudden change in the clock’s frequency aging rate, and a failure arising from an abrupt increase in the clock’s random-walk frequency noise. Our results demonstrate that the three space-segment timekeeping subsystems require the addition of general "clock failure rules" to the basic algorithms that are associated with each system. Once in place, these rules provide for robust, autonomous space-segment timekeeping even in the presence of satellite clock failures.

  In the 1997 PTTI paper titled "Maintenance of HP 5071A Primary Frequency Standards at USNO," Chadsey and Kubik presented findings, to that date, on USNO's efforts to evaluate the devices' performance according to their internal operational parameters. In this paper, the author will present methods used to evaluate those parameters and automatically detect abnormal operation. Some of the developmental difficulties will be discussed, as well as some general guidelines on the tolerances USNO uses to detect problems with the HP 5071A Frequency Standards.

  National Standard Time and Frequency Laboratory of Telecommunication Laboratories (TL) is entrusted by the Bureau of Standards Metrology and Inspection (MOEA) to establish, maintain and disseminate the national standard time and frequency in Taiwan.

  At present, IRIG-B signals are widely utilized in service at TL, such as Network Time Protocol (NTP), Taiwan's Computer Time Service (TCTS), Audio Clocks, and so forth. These services process about two million requests each day. It is concern that the IRIG-B signals come from one generator.

  A Multiple Time Source Compare System (MTCS) is provided to enhance the reliability of the time source of our time dissemination service. The time sources of MTCS come from cesium clocks in two shielded rooms, and provide several kinds of different services on both UTC and local IRIG-B signals.

  In this paper, we will discuss some topics about MTCS; such as system architecture, hardware configuration, control procedures, and software programming.

  A new time interval counter has been developed for PC-based time and frequency measurement applications. The counter incorporates a multi-channel design, allowing the measurement of up to 11 input signals. Each input signal can be compared to an internal reference or to the integrated GPS receiver. Other features include 24-bit programmable dividers that allow the measurement of any frequency from 1 Hz to 100 MHz, single shot resolution of <30 ps and frequency stability of ~1 x 10^-15 at 10⁴s. A discussion of the counter design, the theory of operation, specifications and measurement performance and potential applications in time and frequency metrology will be included.

  An automated computer time distribution system has been set up at the National Measurement Centre (NMC) of the Singapore Productivity and Standards Board (PSB). The system broadcasts standard time to users through a computer and modem via telephone network.

  The server of the system measures and compensates for the transmission delays by measuring the round trip delay. It then corrects for one-way delays through time code advancement. The time code transmitted in the system is compatible with that from the service provided by the National Institute of Standards and Technology (NIST).

  For users who need traceable time, it is of particular interest to know the uncertainty of the time information received. Many factors contribute to transmission delays in telephone lines. The delay of various connections has been studied. Short and long period measurements of these different connections have been made. The influence of telephone traffic on delays has been analyzed. Variations in delay time under different conditions are considered. The results show that within Singapore, the uncertainties of a few milliseconds are achievable with this technique.

  Operation of the National Ignition Facility (NIF) laser requires both periodic and on-demand triggering of over two thousand client devices dispersed about a physically large facility. One element of the NIF facility timing system is a single-width 6U VME module, which triggers local client devices.

  The module accepts the system master 155.52 MHz optical time signal data stream and fires eight individual channels based on VME-programmed timing pattern matches. Each channel includes pattern match logic, a precision digital delay generator, and modular VCSEL laser or transformer-isolated electrical pulse outputs. Typical performance includes resolution of 1 ps, temperature influence of 0.5 ps/K, and 10-second jitter below 2 ps RMS.

  Architecture of the module, implementation of critical circuits, and performance limitations are discussed.

  Time transfer solutions produced routinely by many IGS Analysis Centers show nanosecond level discontinuities between daily solutions, whereas the discontinuities should be consistent with the formal errors of the clock estimates, normally about 0.2 ns, for realistic error estimates. We examine the relationship of clock jump discontinuities at these processing boundaries with measures of multipath delay on pseudorange data.

  Though the level of precise timekeeping for military satellite communications (milsatcom) applications may not be as stringent as that required for satellite navigation, milsatcom poses its own unique timekeeping problems. For example, milsatcom timekeeping must be precise without putting an undue burden on a ground station's workload. Further, milsatcom timekeeping must be robust, with the ability to autonomously detect and correct timekeeping problems during protracted periods when the ground control station is either not available or burdened with other pressing tasks. Here, we discuss three different space-segment timekeeping systems that could be employed in milsatcom applications, and our numerical simulations investigating their various attributes. These systems include a Master/Slave system (similar to present day Milstar), an ensembling system (based on NIST's AT1 algorithm) and a Kalman-Filter system (similar to GPS when it goes to cross-link ranging). The timekeeping performance of these three systems is characterized by the median time interval between ground station updates given a somewhat arbitrary two microsecond requirement for space-segment timekeeping. Among other effects, our simulations include: satellite temperature variations, satellite clock random noise, satellite clock frequency aging, time-transfer noise between satellites as well as time-transfer noise between satellites and the ground station. As we will show, milsatcom timekeeping involves a complicated interplay between satellite timekeeping hardware and the space-segment timekeeping system. Judicious choice of the hardware and space-segment system can allow weeks between ground station updates of the constellation's timekeeping.

  The carrier phase GPS technique allows time transfer between remote sites with 5-minute precision approaching 20 ps, and results are routinely made available by the Earth Orientation Department of the U.S. Naval Observatory (USNO). Although important technical issues involving long-term accuracy and calibration are still being resolved, it is possible to generate sample time and frequency scales to anticipate the full power of the technique. Using minor extensions of the USNO's post-processed Percival algorithm (SuperP), carrier-phase based frequency scales will be presented and compared to local USNO timescales.

  The National Ignition Facility (NIF) currently under construction at the Lawrence Livermore National Laboratory will contain the world's most powerful laser. The NIF laser is a research tool allowing scientists a glimpse into plasma interactions that are equivalent to those found in the center of the sun. The NIF will focus 192 pulsed laser beams onto a pea-sized target creating a fusion reaction between two isotopes of hydrogen. Synchronizing the pulsed laser beams and diagnosing the fusion reaction requires generation of over 1,000 precisely timed triggers.

  In the NIF the Integrated Timing System (ITS) has the responsibility of providing these triggers to laser and diagnostic sub-systems. These triggers will be delivered at precisely programmed times during the interval from one second before to one second after the laser beams strike the target. Some timing system clients require long-term timing stability under 20 picoseconds.

  In addition to precision programmable timing, the ITS trigger system allows simultaneous, independent use by multiple timing system clients. The design allows partitioning of trigger generator channels into subsets that are controlled by way of "keys". The transmission and detection of over 16,000 possible trigger keys are controlled by subsystem clients through shared databases or Graphical User Interfaces (GUIs). These keys can be dynamically shared to create arbitrary grouping of synchronized triggers.

  An overview of the trigger system requirements and discussion of the implementation architecture is presented. Requirements for key components are outlined and some system performance data are presented. Two related papers are presented at this meeting. One will cover details of the system that transmits the timing messages, including the keys; the other will discuss the delay generators that receive and decode these messages, then generate delayed electrical and optical triggers.

This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under contract number W-405-Eng-48.

  According to general relativity, frequency gravitational shifts are the consequence of time retardation in the vicinity of massive bodies. Time retardation must cause the same relative shifts of frequencies for oscillators of all types.

  According to the quasi-Newtonian approach [1], frequency gravitational shifts are caused by changes in parameters of oscillators: near a massive body the effective mass of classical oscillators are increased, and the energy levels of quantum oscillators are lowered. Thus, gravitational shifts in the cases of classical and quantum oscillators have different natures, and the shift predicted in the classical case are half the shift in the quantum case, which in a linear approximation coincides with the prediction of general relativity.

  Note that both general relativity and the quasi-Newtonian approach [1] agree with the experiments performed so far: gravitational effects are tiny, and they have only been observed with the help of precise quantum oscillators. But recently an ultra-stable quartz; i.e., classical, oscillator became available.

  It would be of interest to compare a quartz oscillator with a quantum frequency standard onboard a plane, searching for a variation of their frequency difference which is correlated with a change in altitude. According to the general relativity, the difference in their gravitational shifts should be equal to zero. According to the quasi-Newtonian approach, a 20-km change in altitude should cause an effect on the order of 1.1 x 10^-12. It could be detected by an ultra-stable quartz oscillator and by a transportable H-maser for averaging times of about 10s with the sampling time; i.e., period of altitude change, is several minutes long.

  An absence of difference in gravitational relative shifts of frequencies of quartz and hydrogen standards could be treated as an additional argument in favor of time retardation in the vicinity of massive bodies.

1. Grishaev A.A. On the possible violation of the correspondence principle in gravitational shifts of frequency. Report for 10-th Russian Gravitational Conference, Vladimir, 20-27 June 1999. Abstracts, p.144 (in Russian).

  The IEN time and frequency laboratory is developing an automated TWSTFT station to join the experimenter’s network on the INTELSAT satellite, which is already operational. The hardware and software are still under test that was first checked in April 1999, when the IEN station participated in the international measuring sessions using the OCA code and time-slot. The chance of having three MITREX modems available at the laboratory allowed measurement of the transmitting and receiving delays by suitable cross-connections between the modems.

  In the measurement set-up, which will be described, a common reference frequency was applied to the modems and all the possible combinations of Tx + Rx time interval measurements were performed.

  Unfortunately, the measurement scheme does not allow the estimation of the separate transmitting and receiving delay values. Under suitable assumptions, it may be shown that the variances and covariances of the individual delays may be estimated from the measured quantities. This estimation requires a noise de-coupling technique similar to the known N cornered hat method used to evaluate the noises of the individual clocks from comparison measurements.

  This can give an insight of the modem channels instabilities, both in the short-term and long-term, as one of the instability contributions in the calculation of the TWSTFT uncertainty budget.

  The Real Istituto y Observatorio de la Armada (ROA) was founded in 1753 to increase the knowledge of the Spanish Navy's midshipmen in astronomy matters. Since then, the Observatory has been the astronomical center of the Spanish Navy and it is the oldest Spanish observatory. Jorge Juan, the Spanish Navy Officer, who founded the Observatory, participated in the expedition to Perú to measure the length of one degree of arc of a meridian circle organized by the Royal Acadamie des Sciences de Paris. Its activities have been historically related to astrometry, ephemerides, geophysics and time and frequency fields.

  The Astronomy department was charged with maintaining the astronomical and civil time scales. It was also responsible for synchronizing the master clocks of the Spanish War Ships before they left port. In 1972, an independent department was founded to maintain the time scales using physical standards. A few years later this Time Department (Sección de Hora) received its first atomic clocks. Sección de Hora has been contributing to BIPM for TAI since these former days.

  In this paper we will discuss the activities of Sección de Hora in the time and frequency fields, as depositary of the Spanish time and frequency standards. These activities include time scale algorithm studies; time transfer using single channel and multichannel GPS receivers, time transfer using geodetic GPS receivers and TWSTT.

  ROA is an active and permanent participant in international cooperation in time and frequency activities.

  The National Measurement Laboratory (NML) is responsible for the maintenance and high-level dissemination of Australia's standards for physical measurement.

  In addition to the usual time and frequency calibration and dissemination function, the NML Time and Frequency Group is active in the following areas:

• Development of a trapped ytterbium ion frequency standard, both laser and buffer gas-cooled.
• Two-Way Satellite Time Transfer with links to NIST (Fort Collins) and CRL (Tokyo).
• GPS Time transfer system development, with a growing network of time transfer systems, based on the Motorola Oncore GPS engine and the LINUX operating system located at laboratories within Australia and in the Asia-Pacific region.

  The talk will outline recent progress in these areas.

  The Time and Frequency Department of the Istituto Elettrotecnico Nazionale, as the national laboratory for time and frequency metrology in Italy, is charged with the realization and dissemination of the national time and frequency standards, and the development of new frequency standards and synchronization techniques.

  In the field of atomic frequency standards, two research projects are in progress; one regarding the development of a maser based on the coherent population trapping phenomenon and the other, on the realization of a cesium atomic fountain at IEN in cooperation with NIST and the Politecnico di Torino.

  Concerning the realization of the national time scale UTC(IEN) and its international traceability, a new time scale generator is under development with improved reliability. GPS and GLONASS multichannel receivers are also undergoing performance evaluation.

  The work for the INTELSAT approval and for reaching the operational status of a VSAT transceiver in the Ku-band is in progress. This VSAT will be used with a MITREX modem to join the international two-way synchronization network. The associated measurement system has been completed.

  In the frame of a new CCTF Working Group, studies on time scales algorithms are being carried out. In particular, the generation of time references in space for the new navigation systems are being studied in collaboration with the DLR in Germany.

  The mission and organization of the Time and Frequency Division of the National Institute of Standards and Technology (NIST) will be reviewed and a discussion will be presented of recent activities. Among the major recent milestones have been the development of cold-atom frequency standards, including the construction of a new cesium fountain frequency standard and the design of a frequency standard for operation on the International Space Station. In addition, the use of GPS carrier phase for frequency transfer at the 1 x 10^-15 level has been demonstrated, and the power radiated by WWVB has been increased to 30 kW. Other areas to be discussed include time scales, time transfer, optical frequency standards, and time and frequency dissemination services.

  Communications Research Laboratory (CRL) is a national institute, which is responsible for the national frequency standards in Japan. For this mission, CRL has been conducting many research and development projects in the field of time and frequency standards. To introduce these projects, we will present the following topics:

• Development of primary frequency standards, such as an optically pumped cesium beam frequency standard and a fountain type cesium frequency standard;
• Timekeeping at CRL;
• Millisecond pulsar timing observations for long-term stability of time scale;
• Construction of low frequency stations for the dissemination of the standard time and frequency;
• Precise time transfer using two-way satellite time transfer and GPS carrier phase method;
• Construction of the calibration system for traceability of frequency standards in Japan; and
• Basic research on the next generation global satellite navigation system, such as the spaceborne hydrogen maser, and the precise time transfer method between the satellite and ground station.

  The Time and Frequency section of the National Physical Laboratory (NPL) has an active program of work in time and frequency metrology. At the centre of NPL's time and frequency work is the maintenance and development of the UK's national time scale UTC(NPL). NPL maintains an ensemble of five atomic clocks, two Sigma Tau active hydrogen masers and three HP 5071A commercial cesium clocks. Work is well underway to write a clock algorithm to combine the outputs from all of the clocks, and possibly incorporate other atomic clocks located remotely within the UK. NPL is actively developing new clocks and frequency standards; in particular work on a new cesium fountain clock is well advanced.

  Over the last decade NPL has been steadily developing its time and frequency transfer capabilities. Since 1993 NPL has been routinely operating a Two-Way Satellite Time and Frequency Transfer (TWSTFT) earth station and is actively participating in the associated development work and CCTF working group. This year, a second TWSTFT earth station was purchased further enhancing NPL's capabilities in this area. Over the last two years NPL has been advancing a program of work on geodetic GPS time transfer. A new Ashtech Z-12T receiver has been purchased to complement NPL's existing TTR-4P receiver and an additional receiver is being ordered. NPL has participated in international campaigns organized by the University of Berne, along with the joint BIPM/IGS working group on geodetic GPS time transfer. Software has been developed at NPL for comparing geodetic GPS receivers over short baselines. This is now being upgraded to operate over extended baselines. A 3S Navigation combined Glonass/GPS receiver has been working routinely at NPL since September 1998. NPL has been contributing both to the BIPM Glonass measurements and the IGS IGEX campaign.

  Satellite based time and frequency dissemination is becoming more important within the UK. NPL has undertaken work to validate time and frequency dissemination using GPS Disciplined Oscillators, in particular their traceability to UTC(NPL). In addition NPL is setting up several time and frequency dissemination services, in particular a GPS common-view service based on the Motorola Oncore receivers.

  The U.S. Naval Observatory has purchased one additional maser and ordered six more cesium standards. Soon there will be 50 cesium standards and 15 masers in our main facility at Washington, D.C., and 10 cesiums plus 3 masers at the USNO Alternate Master Clock (AMC), at Schriever Air Force Base. The temperature and humidity environment of these clocks has been improved with better regulation, and the Master Clock DC power backup system rendered more robust through improved charging systems, batteries, and cabling.

  Each clock's phase is measured on at least two different measurement systems, and separate mean time scales are generated using every measurement system. Our Master Clock is steered using data from a time-interval counter and switch system, which is in the process of being modified so as to be quieter, more accurate, and more robust. The modifications involve reduction of components, improved switches, installation of phase-stable cabling and one PPS regenerators with faster rise times and less jitter, and elimination of BNC connectors and most ground loops.

  We are improving the timing ties between our two Washington, D.C. clock buildings and modifying many of our steering algorithms so as to better remain on time and frequency in the event of connectivity loss between buildings or with the USNO-AMC.

  We have improved our Master Clock steering through use of an intermediate mean, steered to predictions of UTC(BIPM), and based only upon masers and cesiums which are characterized using the Percival algorithm and a cesium-only mean. This system has allowed UTC(USNO) to stay within 8 ns of UTC(BIPM), and we are considering switching to steering using a "gentle mean," which is steered to UTC(BIPM) with the gentlest possible algorithm, and an average of masers that are characterized using an unsteered cesium-only mean.

  We have continued to monitor LORAN transmissions, calibrated several GPS receivers for military users, and extended the number of users benefiting from our calibrated TWSTT program and our NTP services on the NIPRnet and SIPRnet.

  We have completed a new laboratory for our atomic fountain project, and are constructing a new GPS antenna mount to reduce multipath.

  We are improving time transfer through the development of better GPS receivers and the improvement of carrier phase techniques.

  The Total variance approach involves periodically extending a data run beyond its normal measurement duration and in such a way that a particular statistic is expected to have the same value with extended data as without. We describe a procedure for improving the confidence of the estimation of the modified-Allan variance (Mvar) using this approach. We have found in simulation studies that if a reflection-only extension procedure is applied to Mvar's individual estimates, we obtain a new estimate of Mvar which has an overall improvement in confidence indicated by an increase in equivalent degrees of freedom for the five common integer power-law noises while minimizing bias. We investigate tradeoffs between confidence and bias of the estimate. The methods outlined in this paper can be applied to other classes of variances, including structure function and wavelet transform approaches and the Hadamard, Picinbono, and pulsar variances.

  All U.S. Naval Observatory (USNO) steering situations involve compromises to minimize the degradation of short-term stability of a steered clock while gaining maximal benefits from the long-term stability of the reference. In the case of steering UTC(USNO) to UTC, extra complications arise due to the 30-day data interval and the 15-day delay associated with the transfer of new information. A technique that minimizes the amount of control required to steer the USNO mean to UTC will be presented. Different strategies designed for optimal steering of UTC(USNO) and a backup master clock system located at the USNO will be described. Some of these strategies involve steering a maser to an intermediate mean that is steered to an extrapolation of UTC. Examples of optimal steering on real data will be reported.

  The estimation of individual instabilities of N clocks, when only differences of phase-time clock measurements are available, can be carried out even supposing cross-correlation among clocks.

  Based on the original analysis of this problem, developed by P. Tavella and A. Premoli, we describe in detail an analysis of the constraint function (inequality restriction imposed by the positive definiteness of the N x N covariance matrix R).

  Then, we will solve a constrained minimization problem, based on the Kuhn-Tucker theorem. We will propose our particular Kuhn-Tucker equations: the already well known constraint function and two new candidates objective functions.

  Advantages and disadvantages in the choice of objective functions will be considered. The first one is particularly interesting by being a second order function. It can be demonstrated that it is strictly convex, this fact, and the convex character of the feasible domain solution, reduces the problem to a strictly convex problem with a single global solution. The second function is not even a convex function, but it expresses properly the dependence among clocks. It is a function whose physical meaning adjusts very well to the property that we require to minimize.

  Finally, we will propose to solve two constrained minimization problems. The first one will let us reach the solution in a first approach. This solution will be used as an initial condition in the resolution of the second problem.

  An overview on the Optimization algorithm used: Sequential Quadratic Programming (QP) method followed, Quadratic Programming subproblem solved in each iteration, line search performed and active set strategy to solve the QP problem, will be carried out.

  Examples with some experimental data from the Real Observatorio de la Armada en San Fernando illustrate the capabilities of this proposal.

Paper 25

  The accuracy of International Atomic Time (TAI) is based on a small number of primary frequency standards (PFS) that aim at realizing the SI second, the unit of proper time. Following the 14^th meeting of the Consultative Committee for Time and Frequency (April 1999), the BIPM time section is reconsidering the way in which the PFS data are used for evaluating the duration of the scale unit of TAI and how they are reported in Circular T and other BIPM publications.

  The frequency comparison of a PFS with TAI is generally realized with the laboratory comparing the PFS to a clock participating to TAI over a given time interval (or the PFS is itself such a clock), and the BIPM computing the link to TAI. The uncertainty s in the frequency comparison is calculated as the quadratic sum of two components s A and s _{B. s B} is provided by the laboratory and is expected to represent the combined uncertainty in the frequency shifts originating from systematic effects. s A is estimated by the time section (with inputs by the laboratory) and is expected to represent a statistical uncertainty on the frequency comparison. In some cases a laboratory may report the direct frequency comparison between the PFS and TAI along with its uncertainty. In the new procedure, it is proposed to break down the estimation of s in a larger number of elementary components and to report all of them in the BIPM publications. This will make the uncertainty evaluation more transparent and traceable. All presently contributing PFS are reviewed under the proposed procedure.

  In addition, the BIPM regularly computes an estimation of the duration of the scale unit of TAI, d, using an accuracy algorithm [1], combining the data of all PFS. The parameters used in this algorithm are re-evaluated and a new estimation of d has been carried out over recent years. The standard uncertainty on the determination of d is now estimated to be 3 x 10^-15.

1. Azoubib J, Granveaud M., Guinot B., Estimation of the scale unit duration of time scales, Metrologia 13, 87, 1977.

  In navigation satellite systems, it is necessary to determine the difference between the on-board time and the reference time for each satellite. This offset can be estimated in real-time by filtering time measurements collected over a ground station network and has to be extrapolated when the satellite is out of visibility of this network. This analysis has been carried out at the CNES in the GNSS2 context and leads to specifications on adjustment and extrapolation errors of the on-board time.

  The purpose of this paper is to predict these two errors from the noise levels and drift coefficients of the on-board clock. The first step consists of estimating the noise levels over a one-day sequence of time error measurements. This is achieved by using an estimator which is insensitive to deterministic drifts (such as the "pulsar variance"). The second step consists in measuring the drift coefficients of the time error sequence with the least squares method. The third step is the extrapolation of the measured time drift over a 3.5 hour duration. In the last step, the extrapolation errors are estimated from both the noise levels and the adjustment uncertainties, which depend on the noise levels.

  After the description of the method, an example will be given using real data. The predicted extrapolation uncertainties will be compared to the real extrapolation errors for a parameter estimation period varying from 3 hours to 24 hours.

WITHDRAWN

  This paper discusses the new GPS receiver technology being introduced to all U.S. Military Services and U.S. allies, now using standard PPS receivers. The new receivers implement two very distinct, yet highly synergistic characteristics. The ability to acquire the P(Y) code direct, without the use of the C/A-code signal and the new tamper-resistant architecture Selective Availability Anti-Spoof Module (SAASM) security.

  In this paper, the focus is on the direct P(Y) for the crypto-keyed receiver and the precision time and frequency applications that can take advantage of this new capability.

  Also included in the discussion is the future of the Selective Availability (SA) signal degradation and the overall scenarios of civil and military GPS signal accessibility.

  A new Global Positioning System (GPS) timing receiver has been built for the U.S. Naval Observatory. The new receiver is a 12 channel, Precise Positioning Service (PPS), unit capable of tracking GPS Y-Code signals and internally removing the effects of Selective Availability. It is based on the receiver being used in the GPS Monitor Stations by the GPS Control Segment.

  Testing is being done at the Naval Research Laboratory (NRL) using the GPS signal simulator. The NRL simulator is capable of producing 10 simultaneous signals with Selective Availability and Y-Code. Initial test results show that prior problems with timing calibration due to uncertainties in the initial conditions in the receiver have been solved. Ionospheric delay measurements in the receiver have been shown to be stable to less than one nanosecond with a small bias. Additional data will be presented showing the measured effects of temperature on the receiver.

  The Naval Research Laboratory has measured the effects of several modulation schemes proposed for use as the new Global Positioning System (GPS) M-Code on the operation of existing GPS timing receivers. Three candidate codes were tested at varying power levels to quantify the effects of the new signal added to a traditional GPS signal generated by a 10 channel GPS signal simulator. The M-Codes were generated in hardware to make representative Binary Offset Carrier and Manchester waveforms. Because the new codes may be run at considerably higher power levels than the current C/A and Y code signals, the M-Code signals were added in at power levels up to 40 dB over the existing signals. The test results show that the presence of M-Code signals is detectable. The effects on noise in the pseudorange and timing outputs are small.

  GLONASS P-code has two main advantages for precision time synchronization. First, GLONASS P-code has a chip length that is 1/10^th of the GLONASS C/A-code chip length and about 1/5^th of the GPS C/A-code chip length. This allows GLONASS P-code pseudo-range measurements to be considerably more precise than comparable GPS or GLONASS C/A-code measurements. Second, GLONASS P-code is transmitted on both L1 and L2 frequencies, without Anti-Spoofing (AS) encryption. The absence of AS encryption allows GLONASS P-code measurements to be used for high-precision ionospheric measurements.

  GLONASS data are subject to a receiver bias, which may be different for each GLONASS frequency. The spread of these biases across satellites can reach 15 nanoseconds and therefore mask other noise sources. Based on the data available so far, GLONASS frequency biases appear to be a function of temperature and relate to specific receivers. Once calibrated with respect to a reference receiver, and provided that temperatures are maintained via laboratory air-conditioning together with a temperature stabilized antenna set-up, these values remain fairly constant and can therefore be compensated in the software. In this paper we describe a test of long distance time transfer using GLONASS P-code multichannel common-view measurements between some U.S. and some European time laboratories. At both sites temperature stabilized antennas were used, and GLONASS frequency biases were determined by means of a portable reference receiver. Use of IGEX GLONASS precise ephemerides provides the best conditions for this test.

  The Royal Observatory of Belgium operates simultaneously as a Time Laboratory participating in the realization of TAI and as a GPS station belonging to the IGS network. Our institute is also equipped with a combined GPS/GLONASS multichannel receiver from 3S-Navigation, involved in the IGEX campaign. We use the advantages of this collocation to study the efficiency of GPS and GLONASS measurements for time transfer.

  In this paper, time transfer using the GLONASS P-Code is investigated. The GLONASS P-code is not perturbed by Anti-Spoofing (A-S); it is transmitted on both the L1 and L2 carriers, allowing high precision ionospheric delay corrections. Furthermore, the wavelength of the GLONASS P-code is about five times shorter than the one of the GPS C/A-code, leading to a measurement noise approximately five times smaller than the corresponding noise for GPS C/A-code measurements. For all these reasons, the use of the GLONASS P-code for time transfer is very promising.

  The results we present here are based on data sets collected by combined GPS/GLONASS receivers driven by a hydrogen Maser, involved in the IGEX campaign, and collocated with time laboratories participating to TAI. First, we compare the results obtained with GLONASS P-code using the broadcast ephemerides and those based on the precise GLONASS ephemerides, computed within the frame of the IGEX campaign. The possible influence of temperature variations at the receiver site is also investigated. Finally, the use of the GLONASS P-code is compared with the frequency transfer using GPS code and carrier phase measurements.

  Two low-cost single-channel fast-sequencing GPS receivers (Type K+K GPS5 made by K+K Messtechnik GmbH, Braunschweig, Germany) tracking the GPS signals of all satellites in view and using carrier phase smoothing have been studied in time scale comparisons between the Physikalisch-Technische Bundesanstalt (PTB) and Technical University Graz (TUG). The receiver output signal is a 1 PPS pulse which is the physical representation of UTC(USNO) obtained as the average from the individual satellites (SATMIX). At the PTB and at the TUG the 1 PPS is measured with respect to UTC(PTB) and UTC(TUG), respectively. After exchange of data, the time scale differences UTC(PTB) - UTC(TUG) are calculated and compared with those obtained by the GPS Common-View (CV) and TWSTFT methods. These comparisons show that the daily mean values of UTC(PTB) - UTC(TUG) found with the SATMIX technique are only a factor of two more noisy (about four ns) than those obtained with the CV method (about two ns). Measurement results over about one year will be presented. It appears that the SATMIX technique is a cost-attractive alternative method for time comparisons.

  An advanced version (K+K GPS5-2k) of the receiver presented here, including a time interval counter, is under development. The K+K GPS5-2K can be operated as a stand-alone receiver like the K+K GPS5 but it can also be used together with a computer. On a data port the K+K GPS5-2K will provide all data which are necessary to convert the recorded data in data files according to the CGGTTS GPS data format or any other format developed in the future. The expected features of the advanced receiver are described briefly.

  We have recently established a stable time scale in the Standards Laboratory at the Santa Clara site of Agilent Technologies, to provide a reference for developing and testing GPS time-transfer products. The time scale consists of a high-performance Cesium frequency standard that is steered to follow UTC(USNO, MC) using the output of a commercial GPS-disciplined oscillator. The disciplined oscillator technique is subject to various errors, such as antenna position, broadcast ephemeris, propagation effects, local multipath, and inaccuracy in the GPS/UTC broadcast correction, as well as instrumental factors. The magnitude of these errors is often time-dependent, and not easy to predict. As a consequence of the long time-constant used in the steering loop, the intrinsic frequency noise of the Cesium standard is also significant. To provide an independent calibration, and to determine the overall accuracy of the time scale, we also carry out common-view time transfer with the USNO according to the BIPM schedule. It is well known that many time transfer errors are considerably reduced by using the common-view technique. The common-view receiver system is portable, and has been calibrated by short baseline, common-clock operation at NIST. In the paper we will discuss the effects of various factors on the accuracy, stability, and traceability of the time scale. Data extending over about 185 days comparing the common-view and one-way time transfer methods will be discussed and compared with data from an independent system operating at Agilent Laboratories. We estimate the stability of the time scale to be about 10 ns rms, and the traceability to UTC(USNO, MC) to be about 10ns at the one sigma level for measurements averaged over several days. Systematic differences between the one-way and common-view results will be discussed.

  Free electrons in the earth's ionosphere cause a frequency-dependent group delay in the bands used for the spread-spectrum GPS ranging signals. This delay constitutes a potential source of error in timing measurements. The ionospheric delay can be removed by dual-frequency ranging using the L1 and L2 signals. Two-frequency time receivers are currently expensive, and are not very reliable when tracking low elevation satellites when Anti-Spoofing is enabled, unless decryption can be used. Single-frequency L1 receivers can correct for the delay using a detailed model of the ionosphere scaled by data contained in the navigation message broadcast by the satellites. However, because of unpredictable variations of the ionosphere, this so-called single-frequency correction is only expected to absorb 50% of the effect. The uncorrected ionospheric delay can cause significant errors in L1 timing systems such as disciplined oscillators. This unpredictable source of error is of increasing importance with the approach of the solar activity maximum. Estimates of the ionospheric delay, calculated from two-frequency geodesic measurements, are now available from several sources, but these are not easy to apply in real time. In the future, WAAS and other system real-time ionosphere estimates for some geographical areas will be available with suitable receivers via geo-stationary satellites. We report attempts to model the zenith ionosphere correction from observations of the L1 code and carrier phase GPS observables made with a multichannel receiver module. Code and carrier phase is affected by dispersion with opposite signs, but carrier phase always contains an ambiguity. It will be shown that by measuring derivatives of the observables, the ionospheric delay can be estimated approximately by post-processing the output of a single-frequency receiver. The use of a single frequency avoids inaccuracy due to group delay differences in the receiver for L1 and L2 signals. Real-time estimation could be performed if the batch processing procedure were replaced by a suitable filter. Preliminary results will be presented, and compared with post-processed IGS estimates.

  Several International GPS Service (IGS) analysis centres now provide ionosphere products in the form of global ionosphere Total Electron Content (TEC) maps. Typically, these maps contain the zenithal TEC as a function of longitude and latitude in a 5° by 2.5° grid, renewed every two hours. Starting in August 1999 the ionosphere products of the IGS Centre for Orbit Determination in Europe (CODE) analysis centre are used for the ionospheric correction of several long and medium distance time links in the TAI network. We show the improvements in stability and accuracy resulting from the use of the IGS products rather than the standard STANdardisation AGreement (STANAG) model for the ionosphere. When compared to ionospheric corrections from on site dual-frequency measurements with non-calibrated receivers (the method previously used for two long distance links). In addition, we see roughly equivalent stability, but improved accuracy and reliability, due to the adjustment of satellite and receiver biases in the IGS solution and to the large number of receivers used by the IGS. We expect that in the near future several more time links will be corrected for ionosphere using the IGS products, which should prove advantageous especially in light of the up-coming strong ionospheric activity due to the solar maximum expected around 2001.

  One of the most efficient techniques for precision estimates of timing signals is based on multidimensional optimal Kalman filtering [1]. For synchronization needs, fast and accurate Kalman filters of time error, frequency offset, and aging are extremely important to realize frequency and time standards based on GPS timing signals. Under condition of very small signal-to-noise ratio that is a usual case for GPS-based measurements of precision time and frequency, approximately 24 hours are necessary to obtain the accuracy of 10^-12 if an integrating LP filter is used. For this reason, current efforts are to minimize the processing time with the same accuracy.

  We present the results of theoretical and experimental studies of different types of optimal and quasi optimal Kalman filters based on crystal and rubidium oscillators with the reference timing signals of the Motorola GPS UT+ Oncore Timing receiver. Pursuing this goal, we consider the filter equations with different definitions of their coefficients and investigate an output signal and its variance behavior in time under real conditions. It allowed us to compare various Kalman algorithms and correspondent optimal filter structures for crystal and rubidium oscillators.

  We mark that, in fact, optimal and quasi-optimal algorithms exhibit the same asymptotic level with under an arbitrary noise. If the frequency behaves as a stationary random function then it does not matter what type of Kalman filter is used. On the contrary, in the non-stationary case the most accurate and fast results may be obtained based on the optimal filtering only.

  There is considered three-dimensional Kalman filter intended for optimal estimation of time error, frequency offset, and frequency aging being based on the oscillator signal model. An application is for synchronization needs of digital communication networks and metrology. The results are given for the measured and estimated value comparisons with respect to the quartz crystal oscillator.

  Many practical results of the Kalman filters use are considered in application to the oven controlled quartz crystal oscillator (OCXO) with AT-cut resonator and rubidium frequency standard. While investigating relatively stationary signals, we used a rubidium source. However, simulation of the non-stationary ones was done based on an OCXO. Many illustrations of the original and filtered processes are discussed for different approaches to the Kalman filter coefficients definition. Estimates of the filtering errors and the processing speed are also given.

  GPS carrier phase time transfer depends on performance of antennas, receivers, cables, amplifiers, and other electronics on environmental and other parameters which is studied under controlled and repeatable conditions. Zero-baseline and short-baseline experiments are used to quantify the effects of temperature and humidity on simple instrumental delay, and to provide information about hysteresis and apparently un-modelable or stochastic effects.

  Time transfer solutions produced routinely by many IGS Analysis Centers show half-nanosecond level discontinuities between daily solutions. We present evidence that these discontinuities are related to the loss of phase-ambiguity information, and discuss methods of handling the problem. Using one method in particular, improved time transfer results from a variety of baselines are presented, and where possible are compared to Two-Way Satellite Time Transfer data.

  We have constructed a link between PTB in Braunschweig, Germany and NIST in Boulder, Colorado using carrier-phase GPS receivers. The link will provide a direct measurement of the frequency difference between UTC(PTB) and UTC(NIST), and can also be used to provide a direct comparison between the primary frequency standards at the two laboratories. Based on our previous work with this method, we expect to be able to realize the frequency comparisons with an uncertainty of about 2 x 10^-15 using one day of averaging. This uncertainty is smaller than the combined uncertainties of the primary frequency standards in both laboratories, and it will therefore support a near-real-time comparison of these primary frequency standards without degrading their capabilities with the noise of the transfer system.

  A dedicated time and frequency transfer experiment over the Atlantic by GPS Carrier Phase (GPS CP) has been operating for more than one year. For this experiment a Geodetic Time Transfer Terminal (GeTT) was installed at the PTB and another at the USNO. The data is processed in the framework of a small network of IGS stations, most of which are driven by H-masers.

  Frequent comparisons between GPS CP and TWSTFT throughout the campaign allowed a comparison of the long-term stability of the two entirely independent techniques. Small discrepancies between the time transfer by GPS CP and the time transfer by TWSTFT have been observed. Over almost one year the difference was of a sinusoidal nature with a period of one year, implying a seasonal effect. Possible causes for this variation are analyzed and discussed.

  As a second issue we discuss the results of other baselines in the network. We focus on the performance of the GPS CP technique as a function of the baseline length, as well as on the potential differences between receiver types.

  GPS codes and carrier phases measured by multichannel geodetic receivers can be used for accurate frequency transfer applications when using geodetic data analysis methods.

  We will demonstrate the main advantage of the use of the GPS carrier phases: high performance frequency stability over short measurement periods (typically Modified Allan deviations of 10^-16 to 10^-15, for averaging durations of less than 12 hours, can be obtained).

  At the Royal Observatory of Belgium, previous on-site studies have shown the sensitivity of the frequency transfer to the temperature variations around the hardware (GPS receiver, cable, and antenna). Now, some parts of the old set-up have been improved. In a first step, the GPS receiver was moved to an environmentally controlled chamber; temperature variations are kept smaller than 0.5°C. In a second step, the old antenna cable has been replaced by a new cable with a low electrical length change versus temperature. The GPS antenna, a Dorne Margolin T, was not replaced and still contributes to the error budget caused by the temperature variations. We will report on recently performed on-site tests using the new temperature-stabilized set-up.

  Characteristic for the geodetic analysis method is the simultaneous estimation of site positions, troposphere corrections and other parameters, based on daily data sets. This daily estimation process introduces jumps in the estimated clock differences, which increase to a few tens of nanoseconds. We will focus on the minimization of the jumps between successive days in order to improve the long-term stability of the frequency transfer using GPS carrier phases.

  Recent developments performed with SATRE two-way time transfer modems resulted in significant performance upgrades and operational improvements of the TWSTFT method.

  The presentation covers new operational modes including carrier phase measurements and advanced modulation techniques. An overview of recently discovered systematic error sources, which might apply to a pseudo-noise modulated system, in general, is presented. The effects of these errors on system calibration and system operation are discussed together with means to reduce them. The resulting improvements are presented and the significance to similar applications is demonstrated.

  Finally, the layout of a potential commercial system for wide-spread use of two-way time and frequency transfer via satellite is presented with the goal, to make available to any commercial user such highly precise methods, which have been verified and refined for more than 20 years by primary time laboratories.

  A precise time and frequency transfer between the satellite-onboard atomic clocks and the ground reference clocks has been planned for the Engineering Test Satellite-VIII (ETS-VIII) under the collaboration between the Communications Research Laboratory (CRL) and the National Space Development Agency (NASDA). We will show the plan and the status report of this experiment.

  ETS-VIII is a Japanese Geostationary Satellite, which will be launched in 2002. It has missions for mobile communication experiments and for precision timing experiments using atomic clocks in space. For the latter purpose, it will be equipped with two cesium beam frequency standards, which were developed for GPS satellites.

  CRL is conducting the precise time and frequency transfer experiment between the cesium atomic clock on ETS-VIII and the ground reference clock. For this experiment, a special time transfer system, which was proposed by CRL, is now under development. In this system, we adopt a two-way time transfer method with the use of the carrier phase information for the precise timing measurement. This system also provides a function of compensating the internal delay variations of the transmission pass and the receiving pass in both satellite and ground station for accurate time transfer. It will attain a precision better than 10ps on the measurement of the time difference between on-board standards and the ground reference clocks. We have completed the on-board engineering model equipment and are constructing the prototype flight model.

  In response to advancements in receivers and increased emphasis on traceability of frequency to the national standard, NIST time and frequency radio station WWVB recently underwent improvements. This includes a 7 dB boost in radiated power, resulting in significantly greater signal availability throughout North America. This paper describes these improvements, theoretical coverage, and consumer orientated receivers projected to number in the millions of units in the next few years.

  Two methods are widely being used to synchronize Time and Frequency of a communications system. One is using earth networks, the other using satellites. In the method using satellites, a transmitting station sends the high precise time information, satellite position data and error correction information, etc., to the satellites. To accomplish this, the satellite data manipulation tool of a transmitting station should perform the reliable and fast interaction with other tools when acquiring, transforming and sending data to the base-band. In this paper, we present the design and core algorithms of the satellite data manipulation tool in a transmitting station as an R & D part of the Time and Frequency transfer system using Koreasat of Korea Telecom.

  We are studying the performance capabilities of a microwave frequency reference based on Raman scattering in Rubidium vapor excited by a diode laser beam. Initial tests using commercially available components show promise for development of a compact transportable Rb clock that would be low-power, with low acceleration sensitivity, while maintaining a fractional frequency instability of ~ 10^-11. Several different configurations have been evaluated with a table-top experiment, using both ⁸⁵Rb and ⁸⁷Rb isotopes.

  The basic design uses a diode laser tuned near the Rb D1 atomic transition at 795 nm. The laser beam passes through a small, heated cell containing Rb atoms with a Ne buffer gas. Inside the atomic vapor, a second optical field is generated by stimulated Raman scattering. This field co-propagates with the incident field and is frequency-shifted by the ground state hyperfine splitting of the atom (3.036 GHz for ⁸⁵Rb and 6.835 GHz for ⁸⁷Rb). Both the transmitted and generated fields are detected by a fast photodiode. The resulting RF beat note is then used as the basis for the frequency reference. This simple design has no need for a quartz crystal, a microwave source or a microwave cavity. Disadvantages are that the linewidth of the signal is broadened (³  100 kHz) and frequency-shifted by the intense laser field that is required.

  To improve the frequency stability we have implemented different configurations, with some penalty of complexity. These include: closed-loop RF feedback, stabilizing of the laser frequency, direct modulation of the laser at the hyperfine frequency (or half of it) with an external oscillator, RF power detection and DC detection.

  At the present, we have observed linewidths of ~600 Hz and have measured fractional frequency instability of a few times 10^-11 at 100 s. The limiting factors at longer measurement times are thermally induced shifts, since the present system has no active temperature control on the vapor cell. At short measurement times we seem to be limited by laser noise or external oscillator noise depending on the configuration. The results obtained with these various configurations will be discussed at the conference.

  This paper presents progress in development of the passive hydrogen masers at IEM KVARZ. The passive hydrogen maser physics package and electronic systems are described and the results of measurements of the passive maser's frequency stability and its immunity to variations in external temperature and magnetic field are reported.

  The availability of Vertical Cavity Lasers (VCSELs) has allowed the development of new generations of optical systems, which can benefit from the small size, low power consumption and high reliability of these light sources. Among diode laser families, VCSELs are particularly attractive for a number of reasons: (1) relatively simple and robust fabrication; (2) very low threshold current and thus low power consumption; (3) surface emission, thus simplified packaging and two-dimensional array capabilities; (4) circular emission apertures and thus good beam shape in the near and far-field; and (5) single-mode emission with a high wavelength tuning rate and range.

  Rubidium or cesium-based atomic time standards could benefit greatly from the use of customized VCSELs as a pump for Rb or Cs. Required is a high spectral purity coupled with emission at precisely the pump wavelengths of interest; 780 or 852 nm. The Swiss Center for Electronics and Microtechnology (CSEM) has developed VCSELs for atomic time standard applications, which fulfill these requirements, and is preparing their industrialization. Optimized laser parameters include single-mode emission with a high side-mode suppression ratio, polarization stability, narrow linewidth, high tunability and low noise. In this paper, we will discuss the principles underlying VCSEL operation, the required fabrication technology and finally, the electro-optical characteristics as these relate to the use of VCSELs in atomic clocks.

  We will present device results on 780 nm VCSELs customized for operation at elevated temperatures for use in Rb time standards. An emphasis will be placed on spectral properties, tunability and stability. Based on these developments, it is expected that VCSELs may soon play a significant role in a new generation of compact atomic time standards.

  Reliable, continuously operating atomic frequency standards with excellent long-term stability are needed for both the generation and maintenance of time scales and for autonomous spaceflight applications. Practical linear ion trap based mercury ion standards have been developed for ground based use at the U.S. Naval Observatory and in the NASA Deep Space Network with no lasers, microwave cavities, or cryogenics. These standards operate continuously and short term stability as low as 2 x 10^-14/tau^1/2 has been previously demonstrated. The measured long-term flicker floor has typically been limited to about 7 x 10^-16 at 100,000 seconds with long term drifts of 1-2 x 10^-16/day. This stability limitation result from the relatively large second order Doppler shift and the instability of this offset over long averaging times.

  In this presentation, we report initial measurements showing a dramatic reduction of sensitivity to the second order Doppler shift using a 12-pole linear ion trap standard [1]. An offset reduction of greater than 20 times has been measured and is expected to directly enable a corresponding improvement in long term stability. A flicker floor of 1 x 10^-16 can be reasonably expected with exceedingly low long-term drift. This advance could have a significant impact in improving the stability of time scales in national timing laboratories and in onboard satellite clock applications, e.g., GPS satellite clocks or autonomous deep space missions. Recent measurements with this ion trap configuration in a frequency standard for NASA's Deep Space Network and in an upgrade of a previously developed JPL trapped ion standard for use at the U.S. Naval Observatory will be discussed.

  The possibility of using multilevel transitions, as in Coherent Population Trapping, in order to implement a high stability atomic frequency standard, has recently raised interest. The physical behavior of such an approach is different from that used in standards based on intensity optical pumping in two level transitions systems and provides some advantages.

  In this paper, two different approaches for the realization of a frequency standard based on the Coherent Population Trapping phenomena in alkali atoms such as ¹³³Cs and ⁸⁷Rb will be described. In both cases, the required multi-chromatic laser radiation is obtained by frequency (or phase) modulation of a single laser diode.

  In the first approach, use is made of the so called "dark line", an electromagnetic induced transparency effect that takes place, for example, when two laser radiation fields are used for the atomic excitation in a so-called L scheme. When the frequency difference of the two laser beams equals the hyperfine frequency splitting, a non-absorption dark line appears in the transmission profile.

  In the second approach, the magnetization generated in the Coherent Population Trapping process is used to implement a Maser without population inversion and without threshold.

  In order to realize such a Maser, a quartz cell containing the alkali atoms ensemble and a buffer gas is placed inside a microwave cavity. The magnetization generated in the L scheme creates a microwave field inside the cavity, which reacts back on the atoms extracting power from the atomic ensemble through stimulated emission. In this case, advantage is taken of the microwave coherent emission having a high signal-to-noise ratio, without the limitations introduced by the threshold characteristics of standard Rb and H-masers.

  Studies of the light-shift effect in these approaches and of the short-term stability of our prototypes will be presented in detail at the conference.

  Market demands compactness and important cost reduction on time and frequency references, able to reach the highest achievable stability (within temperature range, with time: short-term, mid-term, long-term, etc.).

  Main applications that require such improvements are orbitography, timekeeping or frequency reference equipment, positioning and localization systems.

  The new C-Mac miniature ultra-stable oscillator CFPO-1, as time and frequency reference, achieves the piezo state-of-the-art for overall high stability and gives a competitive alternative to atomic clocks.

  Industrialization of recent designs (directly derived from space references), and process control improvements have made possible cost effective solution, which reaches very high performances, using standard and industrial structures and current components.

  The main characteristics, reached at 5 and 10 MHz with the CFPO-1 H0 device, fitted with C-Mac SC cut quartz crystal resonators, are :

• Thermal stability better than 2.1 x 10^-11 from 20 to +60 C,
• Aging few 10 x 10^-13 per day,
• Short-term stability (Allan deviation) few 10 x 10^-14 on 10s,
• Package style 67 x 60 x 40 mm³ (about 0.15 litre leaded package),
• Compatibility with 12 V/15 V/24 V supply voltage,
• Timekeeping (mean square error of prediction): less than or equal to 2 ns over 3 hours; best 20 ns / 50 ns typical over 24 hours.

This paper describes results on this type of device.

  Atomic clocks are ideal frequency standards to fulfill navigation mission requirements. Rubidium Atomic Frequency Standards (RAFS) in particular have been widely used in the GPS system, showing very good performance and a constantly increasing reliability. In space applications, Rubidium clocks are preferred over other frequency standards such as Cesium, because of their lightweight and reduced volume, lower cost and longer lifetime.

  The European Satellite Navigation System, Galileo, calls for specifications similar to those of GPS. Accuracy requirements are even more stringent in Galileo compared to GPS as in the case of civil aeroplane navigation.

  Temex Neuchâtel Time (TNT) manufactures commercial Rubidium atomic frequency standards. The company is proprietary of the whole technology, from the Rubidium gas cell to the electronics assemblies and sub-assemblies. This is why ESA decided to start a development project, which has the goal to qualify a Rubidium for space applications, based upon the TNT's physics package technology.

  This article will provide you with an insight into Galileo's requirements for on-board clocks. Details will be provided on the design and performance of the clock that is presently under development. Follow on activities for a lifetime test on the unit will be described.