Electronic developers have long designed very accurate systems with relative time clocks. However, time stamps derived by separate systems, using relative timing with no synchronizing connection, can't be related to each other. This shortcoming is overcome by synchronizing each separate system to an absolute time standard. Time stamps produced by systems all synchronized to the same absolute time standard can be related to each other within the precision of the traceability process.

For example, the interval between zero crossings of a power cycle between two power grids can be determined, making it possible to phase lock the two grids. Data packet transition times across a continent can be measured, making it possible to provide critical data bandwidth measurements. Also, telephone call billing records can be accurately time annotated. However, absolute timing systems must answer design questions not faced in the construction of independent systems: What is the applicable legal time standard? How will time be transferred from the time standard's master clock to the embedded application? What is a realistic performance expectation? Does the resulting system offer adequate traceability?

Time standards

The second, a measure of time interval, is one of the International System of Units (SI) base units, determined and controlled by the Bureau International des Poids et Mesures (BIPM) in Paris. In 1960, the second was defined in terms of the resonant frequencies of the Cesium 133 atom-thus atomic clocks became our official sources of time. In the United States, the National Institute of Standards and Technology (NIST) operates a very accurate Cesium clock, NIST7, whose output is coordinated with other high performance Cesium clocks-operated by national timing laboratories in Germany, Canada, and Japan-to produce a very accurate estimate of the second.

Formally created in 1971, International Atomic Time (TAI) is the accumulated number of seconds since January 1, 1958, expressed as Modified Julian Day, Hours, Minutes, Seconds. By design, TAI is perfectly linear; however, the rotation of the earth is not. Thus, there is a need to have a time scale that does not gain or lose time relative to the rotation of the earth.

The result was a hybrid time standard, Universal Time Coordinated (UTC). The UTC time scale is stepwise linear, perfectly matching the linearity of TAI except when an extra second is inserted to effectively keep mean high noon within 0.9 seconds of 12:00 TAI. Reported observations from many optical and radio observatories are 'coordinated' by the International Earth Rotation Service (IERS), also in Paris, France. The result is a periodic announcement by IERS to adjust the UTC time scale by one leap second. These adjustments, if called for, occur on either 31 December or 30 June when time advances from 23:59:59 to 23:60:00 rather than reverting immediately to 00:00:00. As of November, 2000 there are 32 accumulated leap seconds.

Universal time

UTC is a computed time standard provided by BIPM 30 days after the fact, so there is no actual UTC clock. However, most countries maintain physical UTC clocks that are their official time standard. The naming convention for national clocks follows the form: UTC (organization acronym here). Although it is not a real clock, the source is sometimes termed UTC (BIPM). The United States operates UTC (USNO, U.S. Naval Observatory) and UTC (NIST). By an Act of Congress, these clocks are official United States time-of-day standards for military and commercial time respectively. As a consequence, systems intended to operate within the United States should use UTC (USNO) or UTC (NIST) as the root time. Similarly, systems intended to operate under the laws or rules of one country should use that country's UTC time standard.

Because there is no single official international clock, the various national clocks can differ by more than one microsecond, and intercontinental time-transfer errors can be substantial; designing systems that must operate internationally with high precision poses a special challenge. Enter the GPS. By default, GPS-because of its worldwide, 24-hour availability and great accuracy-has become the de facto time standard used to support international synchronization applications.

GPS is operated by the United States Air Force and consists of ground stations and 24 satellites orbiting the earth every 12 hours. The orbital inclination and spacing is such that four or more satellites are 'in view' at any given time, necessary for a complete position and time solution. GPS primarily provides highly accurate position data, but precise time is also an essential part of its operation and transmissions. GPS satellites carry several atomic clocks, cesium and rubidium, that are steered to UTC (USNO). These satellites broadcast their own 'GPS Time', which, like TAI, is a constant rate time scale, not synchronized to the earth's rotation.

GPS is widely used as a relay agent for UTC (USNO). The USNO continually monitors each satellite's clock, computes an offset relative to UTC (USNO), and uplinks this offset. The satellites broadcast this offset along with GPS time. GPS timing receivers account for this offset, outputting UTC traceable to USNO. For distinction, we will term this UTC (GPS)-using italics as a reminder that this standard has no official international status.

Given these highly accurate national master clocks, the challenge shifts to transferring their time to a target sub-system. Time transfer involves a Reference Clock Source (RCS) time-transfer method, and the target sub-system. Various types of RCSs are available depending on their intended time-transfer method: satellite, GPS; telephone dialup, automated computer time service (ACTS); Internet public time servers, Intranet dedicated time servers; NTP; radio broadcasts like WWVB, WWV. Essentially the RCS is hardwired into the master clock. As a result, the time output from each primary RCS has zero error at its point of transmission, as relative to the collocated master clock.

GPS-TFMs are the most effective when timing accuracy of less than one microsecond and a high accuracy or stability frequency reference is needed. In addition, GPS-based systems operate anywhere in the world, 24 hours per day. However, GPS systems require installation of an antenna. Although quite small, the antennas do require a clear view of the sky and are generally mounted on the roof. The antenna and its attendant cable installation can rule out GPS if there is no roof access (in the case of downtown, high-rise office buildings) or if the system has to be moved around within a building. Module outputs can be serial time data, one pulse per second (PPS), or frequency reference outputs.

Following protocol

NTP is a client/server application protocol designed to work over TCP/IP data networks. Sub-systems generally operate as a client, obtaining their time from servers located either within the enterprise firewall or outside from primary public timeservers. If the target sub-system is using a high performance microprocessor and a TCP/IP stack, the NTP client may be able to use these resources and operate without any additional hardware. However, if this places a burden on the host processor or if there is no TCP/IP stack, it is possible to integrate an NTP-TFM. These modules typically employ a fast processor, real-time OS, and a TCP/IP stack.

NIST in Boulder, Colorado offers an ACTS. The service is free, but does require a long distance call to Boulder. The calls are short, less than 1 minute, and depending on accuracy requirements can be scheduled anywhere from hourly to monthly. The protocol operates in two modes: one way is where the client simply receives a time message with no allowance for line delay; the other mode takes place when, on receipt of the initial time broadcast, the client immediately responds with a synchronization request. The server then resends the time compensated for its estimate of line delay. This protocol results in 1 to 5 milliseconds client offset error.

The ACTS dialup system is especially easy to implement on general-purpose workstations and third-party software is available free or at nominal cost to synchronize Windows and Macintosh platforms. ACTS is an excellent service that, in today's TCP/IP networking world, is often overlooked. The dialup connection is quite deterministic and secure and systems based on ACTS tend to be very reliable and easy to troubleshoot.

WWVB (low frequency), WWV and WWVH (high frequency) are time broadcast services operated by NIST. Systems based on these radio services can be very economical to implement as demonstrated by the commercial home and travel clocks offered with 'atomic clock' synchronization. However, these systems all suffer from variable signal reception, which can differ from one location to another and at different times of the day. From a product design standpoint, there is no guarantee that systems based on these radio broadcasts will work reliably at the user's location. These systems are mainly used for portable data acquisition systems or where a direct NIST connection is mandated.

Traceability

Laws, rules, or bureaucratic policy decisions, rather than the science of measurement or ISO standards, may often dictate traceability in a particular application. According to ISO's International Vocabulary of Basic and General Terms in Metrology (VIM), traceability is defined as:

The property of a result of a measurement or the value of a standard whereby it can be related to stated references, usually national or international standards, through an unbroken chain of comparisons all having stated uncertainties.

Traceability of all national clocks, including UTC (NIST) and UTC (USNO), relative to UTC (BIPM), is available in the BIPM's publication Circular T. NIST and USNO continually measure and publish offsets of their master clocks relative to GPS. By arrangement, these agencies mail traceability reports to any interested organization and make them available online. Thus, there is complete interlocking traceability between all of these time standards.

Most often, traceability means that the system design demonstrates an unbroken chain of comparisons all having stated uncertainties between the target sub-system and the cognizant national time laboratory's master clock. As a practical matter, how much uncertainty is acceptable? Whether a given system can be traceable falls under the one percent rule: If the time-transfer error between a master clock and the designed sub-system is less than one percent of the required timing accuracy, then the system time can be considered 'traceable,' provided the system meets reasonable standards of reliability, deterministic behavior, availability, and non-interference.

The need for traceable UTC time finds its way into countless applications. Those discussed above, highlight cases where GPS, ACTS, NTP, and WWVB time and frequency modules offer a pre-designed solution. In all of these cases, the designer can demonstrate high timing accuracy by tying the target sub-system's time stamps to the official national UTC clock and the international time standard system.

Design Examples

• Security Trade Time Stamping: The National Association of Securities Dealers (NASD) has a requirement that member-trading firms time stamp transactions with three-second accuracy traceable to NIST. Many of these securities dealers are located in downtown, high-rise office buildings, which rule out GPS and radio solutions requiring an outside antenna. Both ACTS dialup and NTP would meet the requirement technically. The dialup solution would be easier to embed in a trade ticket time-stamp machine. Either ACTS or NTP could supply time for computer workstations. Since these computer systems would already have network access, NTP would be the best choice technically. However, internal control and policy is necessary to assure that operators do not tamper with the time stamps.

• Cellular Base Station Frequency Reference: Code Division Multiple Access (CDMA) cell systems require both precise time and frequency standards in the base station. The time accuracy requirement is 7 microseconds. Embedded GPS TFM's are the only practical solution. Fortunately, the cell systems, by their very nature, have an antenna and the addition of a GPS antenna is not a major concern.

• Traffic Signals: City traffic light systems need to be synchronized to permit constant traffic flow and alter their on/off patterns. Only a few seconds accuracy is required, but cost is a major factor. Without telephone or network connections, WWVB radio receivers or GPS TFMs are ideal solutions.

• Data Center Recovery: One of the primary uses for accurate network synchronization (although traceability is not usually an issue) is network event logging. Data centers are usually under stringent guarantees of quality of service and minimum down times. In the event of network downtime, an exact reconstruction of the sequence of preceding events is essential for a quick and correct diagnosis; accurate time tags are vital to the process. Furthermore, in the case of wide area networks (WANs), precise time stamps are the only way to correlate events between remote locations.