A synchronization plan is the determination of the distribution of synchronization in a network. It involves the selection and location of a master clock or clocks, the distribution of primary and secondary timing, and the selection of the clocks and reference facilities. To achieve the best performance and most robustness from a synchronization network, several rules and procedures must be followed when developing a synchronization plan. Some of the most important rules are avoiding timing loops, maintaining a hierarchy, following the BITS concept, using the best facilities for synchronization reference trans-port, and min:unizing the cascading of the timing reference.

Timing Loops

Timing loops occur when a clock uses a timing reference that is traceable to itself. When such loops occur, the reference frequency becomes unstable. The clocks in a timing loop will swiftly begin to operate at the accuracy of the clock's pull-in range. This will result in the clock exhibiting performance many times worse than it does in free-run or holdover mode. Therefore, it is important that the flow of timing references in a network be designed such that timing loops cannot form under any circumstance. No combination of primary and/or secondary references should result in a timing loop. Timing loops can always be avoided in a properly planned network.

Maintaining a Hierarchy

Maintaining a hierarchy is important to achieve the best possible performance in a network. Under ideal or stress conditions, passing timing from better to worse clocks will maximize performance. Synchronization will still be maintained in normal operation if timing is passed from a worse clock to a better clock. Only performance may suffer slightly, since a better clock is more immune to short-term network impairments and will accumulate less timing error. It is only in the case where an upstream clock enters holdover or free run that a non-hierarchy causes major problems. In this case, the poorer performing upstream clock in holdover may have a frequency accuracy worse than the downstream clock can lock to. The downstream clock will not remain locked and will also go into holdover. This results in multiple clocks being in holdover and excessive slips in the network.

Building Integrated Timing Supply (BITS) Concept

Most administrations follow the Building Integrated Timing Supply (BITS) concept for synchronization distribution. In the BITS method, the best clock in an office is designated to receive timing from references outside the office. All other clocks in the office are timed from this clock. In many cases the BITS clock is a timing signal generator, whose sole purpose is for synchronization. Other administrations rely on clocks in switches or cross-connect systems for the BITS. The BITS clock should be the clock that is best performing in stress and holdover and is the most robust. By use of the BITS concept, the performance of the office will be dictated by the BITS clock, since only the BITS clock is subjected to stress on its timing reference.

Best Facilities for Synchronization Reference

Using the best facilities to transport a synchronization reference is required to minimize slips. The best facility may be defined as the reference with the fewest impairments. This refers to a reference that has the least average number of severely errored seconds (SES) and that is free from excessive timing instabilities (jitter and wander). References that are payloads on SDIVSONET should not be used for tinning, since they are subjected to pointer processing, which adds excessive wander and jitter to the reference. Similarly, references that are transmitted by ATM Constant Bit Rate services will exhibit large amounts of wander and should not be used for timing.

Cascading of Timing References

Cascading of timing references through a network should be minimized. Timing performance will always degrade as timing is passed from clock to clock. The more clocks and facilities in a synchronization chain, the greater the accumulated degradation will become, and the larger the frequency offset grows. Each facility will add impairments to which the clocks in the chain must react. Therefore, for best performance, synchronization chains should be kept short.

The synchronization plan for a network must meet several requirements, ranging from meeting performance objectives and satisfying service needs to being easy to maintain and administer.

Several performance objectives have been established by ITU, ETSI, and ANSI to control slip rates in carrier networks. ITU-T, ETSI, and ANSI allow the long-term frequency inaccuracy at the output of a digital system clock to be no worse than 1 x 10-2, 1 x 10-3 and 1 x 10-9. Short-term requirements of 1 to 10 microseconds of time error in a day must also be met at the output of each network clock.

In private networks, ANSI requires that the first customer premises equipment (CPE) in the synchronization chain have no more than 4.8 milliseconds of time error per day. This corresponds to 40 slips per day, but it is only an interim requirement. In the next few years, this requirement is expected to change of 18 microseconds of daily time error.

Service requirements for synchronization depend on the application. ITU has established a slip rate threshold of one slip in every five hours for an "acceptable" end-to-end connection that will support the service needs of digital data, video, facsimile, and most encrypted services. In ,some networks, a tighter requirement of one slip per day is needed to support encryption. Signalling System 7 (SS7) networks also typically require no more than 1 slip per day. All these service requirements will be achieved if the 1 x 10-11 long term frequency accuracy and 1 to 10 microsecond daily timing error requirements are met.

SDH and SONET services require good short-term stability. ANSI requires that the band-limited short-term noise at the output of any clock not exceed 100 nanoseconds. In addition, the band-limited short-term noise from a BITS clock should not exceed 17 nanoseconds. This implies that the network operate at a stratum 3E or local clock performance levels.

Carrier networks require stringent synchronization planning if performance goals arc to be achieved under normal operating conditions. A receiver clock normally operates with a frequency offset compared to the source clock to which it is locked. This frequency offset accumulates in a chain of clocks. It is a goal of carrier synchronization planning to ensure that these accumulated frequency offsets do not exceed the required 1 x 10-11 frequency accuracy.

Since synchronization performance is dominated by receiver clock and facility performance, the major focus of synchronization planning for carrier networks lies in the determination of tinting distribution and the selection of clocks and facilities used to time the network. The selection and determination of the number and locations of master clock or clocks for the network are largely based on performance goals and maintenance issues.

Carrier Network Synchronization Performance

The synchronization performance in a carrier network is characterized by three components: the accuracy of the master clock, the performance of the facilities distributing the reference, and the performance of the receiver clocks obtaining a reference over the facility.

The inaccuracies of the master clock usually contribute a small portion of the timing inaccuracies in a synchronization network. Primary reference sources (PRS) typically have long-term accuracies on the order of a few parts in 10-13 to a few parts in 10-12. This performance is usually 10-100 times better than the performance of receiver clocks locked to the PRS.

Synchronization performance is dominated by a combination of the facility and receiver clock performance. In normal operation, a receiver clock must lock to a source clock by extracting timing from a facility that has short disruptions or error bursts. The number of error burst events can range from an average of 1 to 100 events per day, depending on facility type, mileage, and other factors.

These constant degradations will adversely affect the distribution of the timing reference. The receiving clock will react to each error. It is allowed to move up to 1 microsecond in response to each error on its timing reference. Accumulation of facility errors and the resulting phase error in the receiving clock can result in the receiver clock operating with a frequency inaccuracy of a few parts in 10-12 to a few parts in 10-10. Therefore, receiver clocks contribute a much larger portion of timing errors and slips in a network than the PRS will.

Primary Reference Sources

Almost all carriers with digital networks rely on primary reference sources (PRS) for timing since nu recommends that networks operate with a long-term accuracy of 1 x 10 -11. A PRS is placed in a network. The PRS times all the equipment in the location in which it resides. This equipment, in turn, will time the rest of the network.

The slip rate contribution of a PRS is usually negligible. A network which derives timing from two PRS clocks will experience at most five slips per year, caused by the inaccuracy of the two clocks. This is negligible compared to the performance of receiver clocks. Receiver clocks typically operate with a daily performance that is 10 to 100 times worse than the PRS to which they are slaved. Therefore, it has been the trend of telecommunication network operators to rely more heavily on PRS clocks and to use multiple PRS clocks to time their network. This reduces the cascading of timing in the synchronization network.

The choice of PRS technology used by an administration depends largely on reliability, maintenance, and control. Performance is usually a secondary concern. LORAN-C or GPS technologies have been used extensively in North America. However, since GPS is controlled by the U.S. Department of Defense and LORAN-C is not global, many other administrations rely on Cesium clocks.

The locations in which the PRSs are used are determined by network topology. PRSs are usually placed in locations that will minimize the cascading of timing in the network. In this manner, the best performance can be achieved in the network. Additional sites that may require the use of PRSs are international switching locations. It is at these locations that one administration interfaces with another and all signals are transferred plesiochronously. It is important, therefore, to guarantee that these locations operate with the 1 x 10-11 frequency accuracy necessary for plesiochronous operation.

Interoffice Timing Distribution

The determination of the distribution of primary and secondary timing through a network depends on the network topology and performance requirements. Interoffice timing is configured such that timing loops are avoided, the hierarchy is maintained, cascading is minimized, and performance goals are met. Most administrations use traffic-bearing lines between offices to pass timing. Some rely on dedicated timing-only facilities between offices for improved maintenance, reliability, and performance.

Facilities with the best error performance, in terms of lowest number of daily severely errored seconds (SES) and least downtime, should be chosen to carry synchronization. Error rates typically correlate with the length of the facility. Therefore, short facilities are preferred. Satellite facilities are not used for timing distribution due to diurnal wander [lb References that are carried as traffic in Constant Bit Rate ATM, SDH, or SONET are also not used to carry the reference due to excessive wander.

Most DS1/E1 references passed between locations are carried on facilities that use asynchronous multiplexing (e.g. DS3 multiplexing). These multiplexers utilize pulse stuffing methods to transmit the reference without retiming or changing the long-term frequency of the signal. The inclusion of asynchronous multiplexing in the reference signal will not significantly affect synchronization performance or planning as long as jitter levels caused by the multiplexing remain controlled.

The amount of reference cascading that can be allowed from the PRS to all other clocks in the network depends on the facility and BITS clock performances and the performance objectives of the network. The design objective for most networks is 1 x 10-11 long term frequency accuracy or 1 microsecond per day of timing inaccuracies. To achieve this level, BITS clocks must have rearrangement maximum time interval error (MTIE) performance of less than 1 microsecond, facilities can only have a few errors per week, and timing cannot be cascaded to more than two to four offices. If very good BITS clocks and facilities are used, more cascading can be allowed and a fewer number of PRSs are required in the network.

Intraoffice Timing Distribution

The BITS configuration is the recommended method for intra-office timing. By using a BITS, the best clock controls the office, cascading is minimized, and administration is easier. In the BITS method, all equipment in the location receives two diverse references from the BITS in a star arrangement.

The stratum level required in a BITS clock is largely determined by performance requirements and the size of office. Stratum 2 and transit clocks are typically used in large toll offices, whereas stratum 3 or local clocks are used in smaller locations. If there is SONET or SDH equipment in the location, it is recommended that the BITS clock be at least a stratum 3E or local clock, respectively. This is to minimize the number of pointer adjustment events occurring on traffic through the SDH/SONET system.
Determination of which clock is best for a given stratum level is based on performance and maintenance issues. The most important performance parameter distinguishing clocks at the same stratum level is the MTIE during rearrangement. This performance parameter will have the greatest impact on daily performance of any function of the BITS clock.

Timing the BITS Clock

The BITS clock receives two or more references from other locations. If dedicated facilities are used to time the BITS clock, the facilities will terminate directly on the BITS. When traffic-bearing facilities are used, the reference facility that is chosen typically does not terminate on the BITS clock. In these cases, a device called a bridging repeater is used. A bridging repeater is an in-line device that taps onto any DS1/E1 signal and provides a copy of that signal. In this manner, any traffic-bearing line coming into an office can be used as a timing reference for the BITS clock.

In order to receive timing not terminated on the clock, the BITS clock also requires external timing capabilities. External timing is the feature of having ports on the device dedicated for synchronization only. Some equipment, however, can only extract timing from traffic that terminates on the equipment. If such equipment is chosen for the BITS and the reference line does not terminate on it, the carrier must dedicate two (or more) traffic ports to be used for synchronization only.

An important factor in supplying timing to the BITS clock must be that diversity is maintained among the references. Diversity must be achieved at all levels. The timing source should be diverse, corning from a different location and different equipment, over diverse facilities. In addition, power to all devices in the synchronization path, including facility multiplexers and bridging repeaters, should be diverse.

Timing Distribution from the BITS Clock

The BITS clock is used to provide synchronization to all equipment in its location. There are several methods used for intra-office timing distribution: dedicated DS1 and El synchronization lines, synchronous clock insertion units, composite clock lines, and square and sine wave signals.

DS1 and El signals are typically used to time digital switches and cross-connect systems in an office. If a timing signal generator (TSG) is used as the BITS, the TSG will supply these signals directly. If a digital switch or cross-connect system is used as the BITS clock, the device will typically only transmit traffic and not synchronization signals. Therefore, with a switch or cross-connect BITS, a bridging repeater or a clock distribution unit is used to supply DS1 and/or El timing signals.

Bridging repeaters are tapped into any outgoing DS1/E1 traffic-bearing line from the switch or cross-connect BITS clock. These bridging repeat¬ers can now supply timing (traceable to the BITS clock) to other clocks in the office.

Clock distribution units (CDU) are devices that accept a DS1/E1 reference and supply multiple references. They essentially act like a multiplicative device giving multiple references out for one incoming reference. A clock distribution unit is distinguished from a TSG in that the CDU does not include a clock and does not retime the line. It has no reference switching or holdover functions that are normally associated with a clock. If the input line to a CDU is lost, the CDU will cut off signals to all outgoing lines.

The DS 1/E1 BITS reference from the TSG, bridging repeater, or CDU is connected to the external timing input of all equipment in the location. The equipment which does not have external timing capability can receive timing directly from the BITS clock by either dedicating traffic ports for synchronization use only or by using a synchronous clock insertion unit (SCIU). An SCIU is a device whose only function is to reclock a DS1/E1 line that is passed through it. Other than reclocking, the SCIU does not affect the line being sent through it. With the use of the SCIU, BITS clocking can be injected on traffic-bearing lines and sent to equipment that does not have external timing. The advantage of the SCIU is that a star arrangement can be maintained without requiring the use of extra DS1/E1 ports in the receiving equipment.

In some equipment, the external timing input does not accept an external DS1 or El signal, but rather a 2.048 MHz, 1.544 MHz, or 8 kHz square or sine wave timing signal. For this equipment, TSGs and CDUs will often have square and sine wave timing signal options. These can be used to directly time the equipment. Many administrations, however, do not use this option and will either time such equipment by an SCIU, or looped-time it from other equipment in the location.

A composite clock is another means commonly used to time equipment in a location. A composite clock signal is used to synchronize equipment that has DSO interfaces, such as channel banks with DSO dataports and DDS equipment. The composite clock signal is a 64 KHz bipolar square wave with an 8 KHz signal riding on it in the form of a bipolar violation every 8 bits. Most TSG and CDUs will provide composite clock signals. Receiving equipment that uses composite clock signals requires only one timing reference. It should be noted that not all networks require composite clock signals.

Monitoring and Verification

A recent trend in telecommunication synchronization is the use of monitoring systems to verify synchronization performance. The timing performance of an office can be verified by comparing the timing of a DS1 or E1 line coming from that office with the timing from a known, highly-accurate reference. Monitoring can be done real-time and continuously by using a bridging repeater to tap into the DS1 line to be monitored and continuously measuring the line relative to a primary reference source (PRS).

Monitoring System

Verification is needed to detect and resolve timing degradations before they impact service. There are numerous sources of timing degradations, such as maintenance activities, circuit re-provisioning activity, clock diagnostics, and excessive facility errors. These degradations are not easy to detect or diagnose using standard alarm mechanisms. Monitoring allows for instant detection and simplified diagnostics of synchronization problems.

SONET/SDH

SDH and SONET use a pointer mechanism to identify the beginning of frames in their payload. If the payload timing differs from the SDH/SONET network element, or if the timing of two SDH/SONET network elements differs, adjustments are made in the pointers. Pointer adjustments cause significant amounts of jitter and wander. Therefore, ANSI recommends that timing is never passed as payload through an SDH/SONET facility. It recommends instead that timing is extracted from the optical signal, In this case, the tinting is derived from the previous SDH/SONET network element (NE) in the facility.

SDH and SONET also place further restrictions on the synchronization plan since BITS clocks in offices with SDH or SONET equipment should have at least a stratum 3E or local clock. This is to minimize the number of pointer adjustment events occurring on traffic through the SDH/SONET system.

Synchronization planning for digital private networks is extremely difficult since the private network topology is typically complex and unconstrained. The widespread use of stratum 4 clocks in these networks can lead to severe performance degradations and force excessive cascading of the timing reference. Private networks are also in a multivendor, multicarrier environment which makes synchronization planning complex. Proper planning requires the knowledge of the entire digital private network, All digital equipment and digital services must be included in the plan.

An additional complexity with private networks synchronization planning versus carrier network planning is that most reprovisioning in the private network affects the synchronization plan. This is because of the limited connectivity in a private network. A new plan is needed if any digital service, facility, or equipment is added, deleted, or changed in a private network. This is usually not the case in carrier networks.

Private Network Synchronization Performance

The majority of private networks rely on stratum 4 CPE clocks for their timing. Since stratum 4 CPE clocks perform 100-1000 times worse than stratum 4E or better clocks, private network synchronization performance is dominated by the performance of stratum 4 clocks.

Under stressed conditions, stratum 4 CPE clocks perform very differently than other network clocks. When a stratum 4 clock experiences a short interruption, it will declare the reference unusable and will switch its reference to a backup timing source, either another timing reference or its internal oscillator. During the reference switch, the clock will typically produce a large, fast phase hit of 10 to 1000 microseconds. This phase hit is often large enough to cause multiple slips and occurs on all outgoing lines of the CPE.

Downstream clocks are unable to remain locked to a reference with such a phase hit. To the downstream device, the phase hit is indistinguishable from a facility error. As a result, the downstream clock will switch its reference, cause another phase hit, and the error event propagates. Therefore, one error on a facility at the top of the synchronization chain can cause all lines and nodes in the synchronization chain to have errors.

Since transmission facilities carrying the timing reference can have from 1 to 100 error bursts per day, the performance of private networks using stratum 4 clocks is typically poor. It can be 1,000 times worse in performance than is seen in public networks, operating at an effective long-term frequency accuracy of 1 x 10-9 to 1 x 10-7. Slip performance of dozens of slips per day per CPE is not unusual.

Synchronization Sources

Private networks typically do not own and operate their own PRSs, but rather rely on the service-provider for timing. Many services provided by the carrier has timing that is traceable to a PRS. The private network can receive timing from these services.

It must be noted that not all services provide timing. The private network operator should check with the service provider to see if timing is provided. In general, in North America, switched services are timed, whereas point-to-point services may or may not be timed.

In Europe, both switched and point-to-point services usually have timing. To optimize synchronization, the private network should receive timing from as many PRS-traceable sources as possible, even if the sources come from different carriers. This will provide the best performance in the private network. In normal operation, the frequency accuracy of carrier network equipment is typically 100-1000 times better than that of a CPE. Therefore, the synchronization performance of a CPE will improve if it receives timing directly from a carrier source, as opposed to cascaded timing from another CPE. The use of multiple PRS-trace¬able timing sources improves performance, minimizes cascading of timing, and increases the diversity of timing sources in the network.

If the private network does not have any service that provides PRS traceable timing from the carrier, the private network must designate a master clock for the network. In this case, the private network is a digital island and does not interface digitally with any PRS traceable equipment. Therefore, the private network does not need to be timed from a PRS. The designated master clock is the clock with the best free-rain performance. The clock should also be centrally located in the network so that cascading of timing can be minimized. This master clock, usually a digital Private Branch Exchange (PBX) or multiplexer, is set to free-run. All equipment in the network obtains timing traceable to the master clock.

The hazard of operating a private network traceable to a clock that is not a PRS is when new services are added to the network. A new digital, PRS-traceable service can be added to the network later. The synchronization distribution must be changed so that the private network receives timing from the new service. If a new synchronization plan is not developed and master timing for the network remains as a free-running CPE, excessive slips will appear and make the new service unusable.

Interoffice Timing Distribution

Timing is passes between private network locations when a location is connected to the private network by point-to-point services that do not have carrier timing. This is typical in data networks. The rules for determining how timing is passed between private network locations is similar to carrier network rules: use the best (typically, shorter) facilities, minimize cascading, and avoid timing loops. Minimization of cascading is especially important in private networks since CPE cascade phase hits throughout a network. The shorter the synchronization chains are, the less stratum 4 CPE equipment is impacted.

The three major difficulties in interoffice timing in private networks are lack of secondaries, non-hierarchical situations, and diversity. Private networks typically have limited connectivity. Each location is connected to only a few other locations. Sometimes, it is impossible to provide a location with a secondary without forming a timing loop. In this situation, it is better to configure the network without a secondary than it is to run the risk of forming a timing loop. The performance of a CPE in a timing loop is typically 10-100 times worse than it is in free-run or holdover.

Non-hierarchical situations also arise due to lack of connectivity. Often, they are unavoidable. As discussed in Section III, the net-work will operate with non-hierarchical situations. These situations can be avoided by introducing a TSG in the timing source location.

Ideally, primary and secondary sources should be diverse, especially since most CPEs do not incorporate holdover in case of loss of all timing references. However, in private network diverse references may not always be possible. A location may be connected to only one other location. In many cases, access of private network service providers is not diverse. Also, in North America, having service from multiple carriers does not ensure diversity, since one carrier may lease lines from another.

Intraoffice Timing Distribution

In private networks, when a location relies on a CPE for timing, as opposed to a TSG, the use of the BITS concept is often infeasible or undesirable. Most CPEs do not have external timing options. Therefore, in order to use the CPE as a BITS, two dedicated DS1 ports (for synchronization uses only) and two bridging repeaters are required. This is usually more expensive than the cost of a TSG. In addition, if the CPE is not major equipment (such as a PBX or a multiplexer), the CPE is typically reprovisioned often and may not make a reliable timing source.

In situations where a BITS clock is feasible, the private network should use a BITS that has a stratum 4E or better clock, if available. If the private network is to meet the future ANSI requirement of no more than 18 microseconds of daily timing error, a stratum 4E or better BITS clock is required. Stratum 4 BITS clocks typically have daily timing errors that range from hundreds of microseconds to tens of milliseconds.

Dynamically Changing Synchronization Plans

Several CPE manufacturers incorporate a proprietary, dynamically-changing synchronization plan feature in their equipment. With this feature, the synchronization distribution continually changes among a network of equipment made by the manufacturer. One or more network timing sources are designated for the network. From these sources, the CPE automatically determines how to pass timing through the network, with the goal of minimizing the cascading of timing. If a CPE loses its timing reference, the entire synchronization distribution in the network will change to resupply all CPEs with a timing source with minimum cascading from the network reference.

In most cases, it is preferable for the private network to avoid the use of dynamic synchronization plans. Such a plan will lead to an excessive number of reference switching events. Since most CPEs with this feature have stratum 4 clocks, excessive reference switching causes large numbers of error bursts and very poor performance. In addition, the use of a dynamic plan does not allow the private network to follow the BITS clock concept and makes the network more difficult to administer.

SDH and SONET

As carriers introduce SDH and SONET into their networks, many private network services will be provisioned as payload on SDH or SONET systems. In these situations, all services for a private network location may be subjected to SDH/SONET pointer adjustments and the associated jitter and wander. Typically, the carrier's SDH or SONET equipment will not be on the private network operator's premises and, therefore, that operator will not have access to the SDH/SONET timing signal that is derived from the optical carrier. The private network operator must rely on the SDH/SONET traffic signal for timing. References passed as payload through SDH/SONET can have significant amounts of wander. A DS-1 or E1 signal, mapped and transported through SDH/SONET, can experience tens of microseconds of wander per day. This situation is currently under study in standards bodies and affected private network operators are handled on a case-by-case basis by the service provider.