There are other architectures; however, they usually are some combination of the ones just described.

Architectural Choices

Conventional hierarchy. The conventional architecture is commonly used by local exchange carriers (LECs) and is described in current industry standards for synchronization and timing. The conventional synchronization plan starts at a primary node with a triple-redundant set of Cesium clocks. Cesium clocks provide reference that is standardized as Stratum 1--which is less than 1x10^-11 error in frequency at any time. While Cesium clocks meet this standard, GPS-referenced rubidium clocks have been shown to achieve better than 1x10^-12 in frequency at any time.

The primary node is where all network elements are timed from the Cesium clock. Lower level clocks are used for holdover, in case the terrestrial transmission links to the primary nodes fail. At the first level below the Cesium clocks are the Stratum 2 BITS clock, having a 1x10^-10 per day holdover stability. These clocks are used at toll centers. Below the Stratum 2 BITS clocks are the Stratum 3 BITS clocks, which have less than 255 slips in 24 hours and are used in Class 5 central offices (COs). Further out--at the customer's site--are Stratum 4 clocks, which are integrated into channel banks or customer premises equipment (CPE). Stratum 4s provide no holdover; rather, they smooth out jitter and wander, providing a clean clock source for transmitted data flowing from the customer site into the network.

The conventional system works only with asynchronous transmission equipment, not with Sonet. The system provides a grade of service and availability that is consistent with the prevailing voice-traffic services it supports. Even Stratum 1 timing sources have routine 648 nanosecond (ns) hits associated with reference switching--this is not allowed under the new clock requirements for Sonet. BITS clocks are low quality, and have transients that exceed today's Sonet requirements (Figure 1).

Sonet distribution with conventional hierarchy. When Sonet was introduced, the ability to transport timing using the optical line overhead initially was seen as a great benefit. Unfortunately, ring topologies complicate matters. In addition, the low quality of Stratum 3 holdover prevents it from being used with Sonet systems. This has prompted the industry to develop a new Stratum 3E standard with less than a 1x10^-8 per day error and a well-defined, short-term, phase noise output specification.

A Sonet transmission system distributes synchronization with the head-end of a Sonet chain, or the "top" of a Sonet ring timed from a higher quality Stratum clock. Subsequent nodes use line timing to drive local lower-quality Stratum clocks. Stratum quality is best at "1" (the lowest number) and worst at "4" (the highest number).

Hierarchical BITS are used at each node; they take the line-timing from the Sonet multiplexer, smooth it out and return it to the mux. The multiplexer then can be used for line timing that continues on downstream or around the ring. Special care must be taken to create a ring that cannot be configured to take timing from itself, which creates a timing loop (Figure 2).

Sonet distribution with synchronization messaging. In the event of a fiber failure, a ring may rearrange itself to transport traffic; however, it may be unable to find a good timing source without creating timing loops.

While the ring reconfigures, it may not be able to deliver service. One way out of this dilemma is to use a BITS clock at each Sonet node. Another way is to use synchronization status messages (SSMs) which allow the ring to reconfigure its timing.

BITS clocks still are used frequently to bridge the time it takes a ring to find a good source of timing. In such intervals, the Sonet node operates from its internal Sonet maintenance mode clock, if no other clock is used. The resulting quality level, while above Stratum 4, is below Stratum 3 and does not guarantee traffic transport (Figure 3).

Sonet timed from conventional hierarchy. Faced with such problems, many networks use the existing asynchronous network to transport timing signals to BITS clocks at the Sonet nodes. This is an overlay of two technologies, and an interim topology. It's problematic, because the BITS clocks that are used are frequently of the older variety, and may not be stable enough to prevent excessive pointer adjustments by the Sonet nodes (Figure 4).

GPS-derived timing and synchronization. The above situations faced by carriers changed as soon as products became available using information continuously transmitted by the GPS satellite constellation to discipline clocks. No terrestrial distribution would be required; every node could derive timing independently from a GPS reference.

Holdovers of the Stratum 3E or Stratum 2 quality levels are available on many of these clock systems, as they can use external T1 references should an antenna or GPS receiver fail. In most failure scenarios, internal line timing can be used, thus eliminating the cost of redundant GPS. (Figure 5).

DIRECT COSTS

The conventional approach is probably the least expensive in terms of direct equipment costs. Stratum 3E clocks are not required to keep any part of a network operational. In addition, use of asynchronous transmission facilities is a well-known, well-understood methodology.

Sonet-based systems add immediate direct costs, because Stratum 3E clocks or better BITS clocks are required at every node. In some cases, direct costs increase when existing asynchronous transport is used for timing. The GPS timing model increases costs, because a GPS BITS clock is placed at every node, and associated cabling to the GPS antenna outside must be installed.

Early GPS systems needed to view at least four satellites to operate properly, and antennas required an unobstructed view of the sky. However, newer GPS systems need to view only one satellite, and even can operate through a window on the side of a building--suitable for urban applications.

INDIRECT COSTS

The conventional model has one hidden cost: the loss of revenue caused by the use of transport that otherwise can be made available for network traffic. Sonet models have few hidden costs. Only the use of existing asynchronous links results in lost revenues. Existing derived DS1s may be used for timing distribution; however, pointer adjustments and other impairments make this a rarely used option.

Another hidden and potentially high cost for Sonet is the need to have a network management system for locating problems. In fact, the maintenance and upkeep of such management systems may be more costly in some instances than the Sonet network itself.

The GPS model has no indirect costs; the satellite system is federally maintained and reception is free.

OPERATION AND MAINTENANCE

The conventional system, because of its hierarchy, can be operated and maintained simply. Troubleshooting is performed with conventional problem-locating tools on the transport part of the network. In such cases, technicians carry travel cases with portable Cesium or rubidium oscillators, used to identify synchronization and timing troubles. Although the architecture is labor-intensive when identifying trouble, it is simple to understand. Maintaining proper records is key in identifying facilities used for timing distribution.

When a Sonet system fails, it still may be difficult to find qualified personnel to locate and identify the problem. Sonet architectures typically have prolonged troubleshooting intervals; in addition, transport services may be disrupted during an outage. Timing systems with built-in test and measurement, such as synchronization performance monitoring using automatic threshold crossing alarms, work with network management systems to identify trouble-spots. Two factors make finding and tracing circuits in a Sonet network difficult: the need for automatic record-keeping; and problems associated with Sonet add/drop and insert capabilities.

To ensure a high level of network availability, many Sonet systems use a large number of Stratum 2 BITS clocks to enable repairs during working hours. Unfortunately, Stratum 2 BITS clocks typically use rubidium-based oscillators for long holdover stability (more than a month to the first slip). Rubidium oscillators exhibit wear-out symptoms in less than 15 years. Some timing systems have rubidium devices that require calibration or repair every seven to 10 years.

GPS timing systems are independent. Each system operates until it enters a holdover or failure mode. A far simpler network management system is needed to find nodes not in GPS mode, and only those nodes need to be configured to take other sources of timing (such as Sonet line timing) until the fault is fixed. If the GPS timing equipment can automatically fall back to a DS1 reference or has suitably high holdover stability, then faults will not cause service outages. There is time to repair the fault during normal working hours, not during weekends or evenings.

All architectures use BITS clocks to some extent. Not all systems need large or complex network management systems; but all can derive benefits from network management. Some networks are excessively complex and fragile, and do not deliver the expected grade of service. Under failure conditions, some networks are more robust than others, but generate hidden costs associated with antennas, network management systems or perhaps BITS clocks at every Sonet node.

Austin Lesea is Vice President of Advanced Product Development and a Director of Larus Corporation, San Jose, California

Reprinted with permission from AMERICA'S NETWORK, February 1, 1998
AN ADVANSTAR PUBLICATION Printed in the U.S.A.