Advantages of timing networks with cesium atomic clocks

A good example of synchronization is a network element receiving information on an ingress DS1 that must be transmitted within a (different) egress DS1. If there is a frequency offset between the ingress and egress links, the network element will experience either a surplus or deficit of information bits. This results in information units being deleted, or repeated, respectively.

The term slip generally refers to the deletion/repetition of a 125-ms unit of information. In the case of a 64 kbps voice channel created by sampling at 8 kHz with one octet per sample, a slip results in the deletion/repletion of one octet.

For a voice call, the impact of an occasional slip on service quality may be minimal to the human ear; for a call between fax machines the effect can be more dramatic. A slip is different from a bit-error where a "1" is incorrectly decoded as a “0”. Forward error correction schemes can correct for bit errors but not for slips.

Considering that every link between network elements is a potential weak link, the ITU-T has developed standards for acceptable slip rates as well as how to budget this requirement between different segments of a hypothetical reference connection (ITU-T Recommendation G.822, Controlled slip rate objectives on an international digital connection, is available at www.itut.int).

Ideally, there should be zero frequency offset between any two interconnected network elements. As a practical matter of maintaining a high quality of transmission network-wide, however, every node in a communication network requires access to a primary reference source (PRS).

The PRS acts as the heartbeat of the network. It provides the precise and accurate synchronization and timing essential for the reliable and efficient transmission of voice, video, and data throughout every type of digital network. Precision synchronization is essential for maintaining network reliability, efficiency, and performance.

Relevant network and telecom standards such as ITU-T G.811 and ANSI T1.101 define a primary reference source as one with a frequency offset of less than 1x10^-11.

(ITU-T Recommendation G.811, Timing requirements at the outputs of primary reference clocks suitable for plesiochronous operation of international digital links, is available at www.itut.int. ANSI TI1.101 Synchronization Interface Standard is available at www.atis.org.)

A PRS can be provided directly with a stand-alone clock or indirectly. In the indirect method, a radio receiver receives timing frequency signals that are used to extract a frequency reference traceable back to a primary atomic standard.

GPS and GLONASS, for example, provide such a frequency reference utilizing a satellite network and LORAN does so through a terrestrial system. In all these systems, the cesium atom provides the basis of timing since the original timing reference is predicated on cesium atomic clocks. In fact, the International Standards' (SI) definition of the second is based on the CS-133 isotope of cesium. A typical cesium clock is shown in Figure 1.

The operation of a cesium-beam atomic clock is based on the transition between two states in the cesium atom. Simply put, the atom has two ground states. The transition between the states is associated with a photon of frequency 9,192,631,770 Hz, precisely.

By bathing a stream of cesium atoms in an RF field and counting the number of atoms that change state, the frequency of the RF field can be calculated. By embedding this "physics package" in a closed loop with a (high-quality) quartz oscillator, the oscillator can be steered to be on-frequency.

In a simplistic sense, the physics package emulates a tank circuit with very high Q—on the order of several million. The accuracy of the cesium clock is dependent on the number of cesium atoms "interrogated" and improves as the number of atoms in the beam increases. Since the cesium atoms are being "consumed", there is a trade-off between accuracy and tube life.

Cesium atomic clocks play a critical role in the establishment and maintenance of time scales such as UTC. With accuracies of better than one part in 10¹³, cesium atomic clocks are used in laboratories such as the National Institute of Standards and Technology (NIST) to establish and maintain national time-scales.

Devices used in metrology applications trade-off life-time and cost for accuracy. With the availability of low-cost, long-life, cesium-beam-tube atomic clocks, providing accuracies of one part in 10¹³ and better, direct cesium-based synchronization is cost-effective for telecommunications and networking applications.

Managing time distribution
An accurate clock is the foundation of a stable network. Timing offsets between nodes result in lost data and retransmissions.

For non-real-time services such as web surfing, packet retransmissions waste bandwidth. For real-time services, such as IPTV or online gaming, data loss results in degraded service quality. This is why standards call for stratum-1 accuracy, which means that clock signals used for synchronization must provide accuracy better than 1x10^-11.

It is worth noting that when used in a telecommunications context, the term "stratum-1" refers to frequency accuracy. In the NTP (Network Time Protocol) context, the term "stratum-k" applied to a server indicates how far down the chain the server is. For example, a stratum-k server gets its time-reference from a stratum-(k-1) server. A stratum-0 timeserver gets its reference from external (non-NTP) means and most stratum-0 timeservers deployed today utilize GPS as a time-reference.

All the network elements in a central office derive their time-base from a reference provided by a central clock, called a Building Integrated Timing Supply (BITS), which distributes accurate timing throughout the office. This architecture is cost-effective because it places the highest-cost component, namely the holdover oscillator, in a single device rather than in every network element.

The BITS derives its timing signal reference either from a local PRS or, more commonly, via a network feed, typically a DS1. That is, traditionally, network timing was distributed over the network itself in a hierarchical fashion as shown in Figure 2.

The BITS is a slave clock (ITU-T Recommendation G.812, Timing requirements of slave clocks suitable for use as node clocks in synchronization networks, available at www.itut.int.)

The holdover oscillator quality determines its stratum level. However, as networks migrate from copper to fiber, network timing distribution suffers. For instance, a DS1 carried over a T1 facility can be used as a synchronization reference.

SONET, however, introduces jitter and wander through pointer adjustments that render a bearer (i.e. traffic bearing) DS1 inadequate for use as a synchronization reference. In these situations the recovered SONET (optical) line clock is used to create a “derived DS1” for use as a BITS reference.

Needless to say, the more stages that the timing signal passes through, the less stable the signal becomes. In this case, clock accuracy has nothing to do with long-term accuracy but rather with short-term stability. Furthermore, if any clock in the synchronization chain loses its reference signal and goes into a holdover mode, all the downstream clocks are affected.

A popular method of distributing PRS clock signals is based upon GPS. GPS provides a timing signal traceable back to UTC. This method avoids the jitter and wander degradation experienced when the timing signal is distributed over the network. In general, GPS-based synchronization is highly accurate and cost-effective but it is not always the most appropriate approach. A GPS-based PRS is constructed utilizing a GPS receiver in conjunction with a high-quality oscillator such as an OCXO (double oven quartz) or rubidium secondary atomic standard.

The upfront cost of a cesium atomic clock is greater than that of a GPS-based PRS. But when installation and maintenance costs are taken into account there are many applications where a direct cesium atomic clock implementation is more cost-effective.

A GPS-based PRS incurs cable and antenna costs, and running cable from the basement of a building located in an urban canyon to a roof position that has satellite line-of-sight can quickly add up.

Many metropolitan environments also have limited roof access and availability, increasing the cost of installing an antenna. While a long cable will not necessarily degrade the incoming RF signal from a timing perspective, planning the cable drop, acquiring the necessary space, and laying the cable could eliminate any cost advantage of a GPS-based PRS over a stand-alone cesium clock.

Additionally, the multiple elements of a GPS-based PRS introduce multiple points of failure into the synchronization subsystem and make troubleshooting and maintenance a far more complex affair than monitoring a green light/red light on a cesium atomic clock.

There are also environmental factors to take into consideration. GPS signal loss is subject to the weather. Lightning protection mechanisms can fail and cables can be accidentally cut. The antenna and its external structural and protection systems require periodic inspection and testing to ensure that the total antenna system continues to meet the original engineered specifications. Commonly, when a GPS-based timing system falters, the antenna is the culprit. Anecdotal evidence suggests that visible antennas make for great target practice.

For many applications, a GPS-based PRS is simply not an option. Equipment collocated in a telecom closet with other equipment, for example, may be limited to rack space only, without access to the roof or the ability to penetrate a wall to run a cable to an antenna.

The alternative is to purchase a reference clock from other companies collocated in the closet. Given that a network will require at least two clock feeds for redundancy and that renting a clock can run several hundred dollars a month, rental costs can easily exceed the cost of owning a cesium clock. Additionally, direct ownership of a clock source provides additional reliability, generates less paperwork, and minimizes human error issues that can result in "finger-pointing."

Direct clock ownership
Bringing a direct PRS into a network eliminates many frustrating timing distribution problems. In principle, network elements can be chained, deriving a reference from an upstream network element in order to discipline its internal time-base.

This may preserve the long-term frequency accuracy but additive clock noise in the network element clock will degrade the reference provided downstream. Additive wander will have a direct effect on how far you can distribute a synchronization signal.

ITU-T Recommendation G.803 (Architecture of transport networks based on the synchronous digital hierarchy (SDH), available at www.itut.int) specifies that a transport network element should have a clock traceable back to a primary reference source. It also provides guidelines as to how many slave clocks, or BITS, can be present in a chain before the quality of synchronization becomes sub-par. The fewer slave clocks in the chain the better.

Networks must be engineered to meet the most stringent timing requirements. Service providers cannot, in general, control the mix of traffic so meeting stringent requirements means they can ensure that all customers consistently get the best possible service experience.

For a large network, it may be prudent to provide several timing frequency sources through multiple cesium atomic clocks that can provide redundancy, mitigating the impacts of failures or fiber cuts. Multiple PRSs also simplify administrative procedures and reduce maintenance costs, particularly when troubleshooting activities must be invoked.

Cesium atomic clocks also provide significant benefits as networks move from ring to mesh topologies. Because it is mandatory that timing loops be avoided, the origin of a frequency reference is critical for maintaining synchronization accuracy. With a ring topology, it is clear that the clock signal is either coming from one direction or another.

With a mesh topology, however, the addition of new elements to the mesh can introduce provisioning errors that disrupt the traceability and introduce timing loops. A direct clock source reduces the impact of such errors on system stability by shortening and simplifying the distribution chain. This is especially important since maintaining traceability across even a simple network has become increasingly complex. The Table illustrates some of the common issues that complicate network synchronization. The availability of a direct PRS, such as a cesium atomic clock, alleviates many of these negative symptoms.

Minimizing complexity
Cesium atomic clocks have a built-in reference based on fundamental properties of matter. This characteristic can reduce complexity considerably. Although the internal is surrounded by electronics, it derives its accuracy based on the quantum-mechanical behavior of atoms.

While cesium atomic clocks are one of the most complex clock technologies available, they are packaged make this complexity transparent to the user and are often referred to as red light/green light devices. As long as the green light is lit, the clock is providing an accurate synchronization frequency. With such simple external packaging, a cesium clock can be installed in less than 30 minutes and maintenance is a matter of confirming that the green light is on.

Cesium atomic clocks is quite complex and cannot be repaired in the field. When a clock experiences a failure, it illuminates the red light and triggers a network alarm. Typically, this is not a catastrophic event. The BITS that is fed by the primary reference clock has a holdover clock that will continue to supply a stratum-1-level signal to the network while a technician replaces the failed clock with a spare clock and then ships the failed unit off for repair.

Given their self-contained packaging, cesium atomic clocks are extremely robust and come with warranties on the order of 12 years. As a consequence, cesium atomic clocks require little to no maintenance during their lifetime, typically 15 to 18 years. Just for reference, with an accuracy of 1x10¹², the error accumulation over 12 years is less than 400 ms.

Frequency versus time
It is important to be aware of the difference between synchronization based on frequency and synchronization based on time traceable to UTC. A traceable time source provides an absolute time, which is important for applications such as time stamping or where time-of-occurrence needs to be measured in two different locations. A wireless network is a good example of a network that utilizes traceable time to ensure that signal handoffs between two cells are synchronized, minimizing dropped calls for mobile users.

Most networks and applications—including self-contained and private networks—do not require traceable time. The stable timing frequency provided by a stand-alone cesium atomic clock yields a relative time. Since all nodes communicating with the network share the same relative time, this is sufficient for reliable synchronization.

If a network requires time traceable to UTC, then the appropriate technology is a distributed clock over GPS with a cesium atomic clock acting as a local source and holdover clock. If the network does not require time traceable to UTC, a cesium atomic clock efficiently provides the necessary PRS.

Because synchronization is critical for efficient network operation, having a direct PRS within every office is recommended. This is because the widespread deployment of PRSs greatly reduce the recurring operational expenses associated with maintaining accurate and reliable synchronization distribution throughout a network. For example, after extensive study Telcordia has prescribed several methods of PRS deployment, with a minimum of one local PRS for every local serving area (GR-436-CORE, ,Digital Network Synchronization Plan, available from Telcordia Technologies www.telcordia.com

While the needs of each network are unique, using cesium atomic clocks as part of a PRS strategy will ensure that the most reliable time and frequency signal is distributed throughout the network. Cesium atomic clocks are immune to external failure, do not require an antenna or complex installation, are simple and cost-effective to maintain, and for many applications have a lower total cost of ownership while providing higher reliability and accuracy than GPS-based systems. Direct ownership of the cesium frequency enables service providers to deliver to their customers the most consistent service with the best performance available.

About the authors
Jim Olsen is Symmetricom's business development manager for North America. Formerly the company's director of advanced technologies, he has authored numerous white papers and has made significant contributions to others. His articles on synchronization have appeared in industry books and trade publications. He can be reached at jolsen@symmetricom.com. Kishan Shenoi is chief technologist at Symmetricom. Prior to joining Symmetricom, he worked for DSC Communications Corporation and ITT Advanced Technology Center. Dr. Shenoi holds more than 20 patents as well as a PhD in Electrical Engineering from Stanford University, an MS from Columbia University, and a BTech from the Indian Institute of Technology. He can be reached at kshenoi@symmetricom.com.