Wander and jitter defects roam in broadband networks
In fiber-optic networks, data edge phase variations in transmitted digital signals can cause clock and synchronization problems that degrade voice, data and video services
Dan wolaver and dana a. cooperson
tektronix Inc., Microwave logic Products
In broadband networks using synchronous digital hierarchy (SDH) or synchronous optical network standards, both jitter and wander characteristics must be monitored for deleterious effects. As in traditional asynchronous networks, synchronous network elements have clock-recovery phase-locked loop circuits that are sensitive to jitter. Unlike earlier networks, however, synchronous network elements rely on low phase drift between timing signals
for proper operation. If the drift exceeds certain limits, large phase steps occur in the payloads and degrade signal transmissions.
Standards bodies, such as Bell Communications Research and American National Standards Institute (ANSI) for North American countries and the International Telecommunications Union-Telecommunications (ITU-T) Standards Sector for overseas countries, classify digital phase variation as either jitter or wander, depending on frequency. Specifically, phase variation greater than 10 hert¥is defined as jitter, whereas phase variation less than 10 H¥is defined as wander.
To control problems related to jitter and wander, ANSI has published specifications that limit Sonet interface jitter and wander; Bellcore standards define the jitter and wander performance required of Sonet equipment; and the ITU-T sets interface and equipment jitter and wander specifications for SDH systems internationally.
Because of the ubiquity of asynchronous networks, service providers primarily focus on payload jitter as a major cause of service problems. Wander is as yet misunderstood and unappreciated as a source of errors in services carried over emerging synchronous fiber-optic cable broadband networks. As the size of synchronous networks increases, the role of wander as a source of payload errors is expected to become more evident, and related problems are anticipated to worsen.
Sources of error
Fortunately, specialized test tools and instruments are available for design conformance and network deployment measurements to proactively locate the sources of wander-induced errors before they cause service problems. Once identified, these sources can be replaced or removed, thereby improving network reliability and service quality.
Because wander is a relatively new measurement parameter specified for fiber-optic networks, its sources, impact on the payloads carried by broadband networks, and generation and measurement mandate investigation.
Jitter, wander and phase transients degrade a digital signal caused by clocks that generate these digital signals and by the multiplexed systems that transport digital signals. Jitter is the short-term variation from an ideal position in time of important events (for example, rising edges or level crossings) of a digital signal. Wander is the long-term variation from an ideal position in time of the important events of a digital signal.
Short-term implies phase oscillations with spectral components greater than or equal to 10 Hz; likewise, long-term implies variations with spectral components less than 10 Hz. A special class of wander events known as phase transients embodies large excursions in phase of limited duration--that is, less than 10 seconds.
Jitter is controlled in a synchronization-distribution network by requiring stratum clocks to accept inputs with high levels of jitter and to generate outputs with low levels of jitter. The same is true of controlling jitter in telecommunications transport networks. Multiplexers and repeaters must meet jitter transfer, generation and tolerance specifications. Sonet jitter specifications are detailed in Bellcore GR-253-CORE and ANSI T1.105.03-1994 standards.
Sonet and SDH systems use data pointers to keep track of the location of each payload being transported. If the phase between two optical line signals drifts slowly, the network element`s first-in/first-out synchronization buffer must eventually make up the difference with a pointer adjustment. This pointer adjustment does not change the phase of optical signals; rather, it shifts the payload within the outgoing optical signal.
When the payload is desynchronized as it exits the system, any pointer adjustments will result in payload jitter. If this jitter is excessive, it causes a first- in/first-out "spill down the line," which results in severely erred seconds, unhappy customers and loss of revenue for the service provider.
Phase drift can also occur between synchronized clocks. These clocks use a control mechanism to maintain lock with their reference source. However, the control mechanism is prone to cause wander over time.
There are three principal sources of wander in synchronous broadband systems:
Drift in synchronization circuits
Reconfiguration or improper configuration of the synchronization path
Change in delay due to cable expansion with temperature.
The first two sources typically overwhelm and therefore minimize the impact of the third source. Although wander affects synchronous communications networks that use fiber-optic cables for transport, the fiber itself is not a significant cause of wander.
Over time, the clock circuits used to supply network synchronization can experience offset voltage drift. This voltage drift will cause phase drift in phase-locked-loop circuits, which are responsible for locking one clock in a system to another. This phase drift, one of the main causes of wander, results from the imperfect functions of electronic circuits.
Another cause of wander in networks can be inadvertently introduced by a network provider. In Sonet and SDH systems, the synchronization source for each network element is typically provisioned through a local craft access device or remote maintenance system. Often, wander problems can be traced back to improper provisioning of the network elements by the network provider. For example, timing loops or breakage of the synchronization path--which can result from improper provisioning--can lead to excessive wander.
Even with proper configuration of the synchronization path, reconfiguration can occur because of automatic or manual timing reference switching, entry into self-timing operation when the external reference is lost, recovery from self-timing operation, automatic clock diagnostics, hardware protection switching or pointer adjustments for payload signals carried over Sonet or SDH networks. Each reconfiguration causes a phase step; multiple reconfigurations result in random phase walk, or wander.
In asynchronous networks, payloads with reported errors are routinely tested for jitter, and appropriate corrective action is taken. In synchronous networks, however, the cause of this jitter--wander between the Sonet or SDH line and the synchronization reference--can go undetected. Thus, the symptom rather than the cause is treated. In addition, service-affecting wander problems that are not evident today will arise as synchronous networks grow in size.
Control of first-in/first-out slips and Sonet pointer adjustments requires that all network clocks operate at the same frequency within an achievable limit. This control is accomplished by building a synchronization network that provides synchronization references traceable to highly accurate clocks known as primary reference sources.
While monitoring the results of synchronization problems (such as pointer activity and first-in/first-out slips) provides useful information, practical experience dictates a more proactive approach to performance monitoring to find synchronization problems before a service impairment occurs. Wander and frequency accuracy indicate a signal`s synchronization performance.
The following issues relate to measuring wander and frequency accuracy:
A known good clock reference for comparison is required. This clock can be the primary reference source from which the network gets its timing, or a cesium clock that provides a portable reference. (These clocks are accurate to within one part in 1011, or to 9 microseconds per day.)
Three parameters are used to characterize wander--time interval error, maximum time interval error and time deviation.
The frequency accuracy of a clock source is specified as x parts in 10y. This parameter is typically calculated from the time interval error data by calculating the slope of its plot.
Wander measurements are generally performed at network elements and with specialized performance monitoring test set equipment.
Bellcore, ANSI and the ITU-T have set wander limits--in terms of maximum time interval error and time deviation--that Sonet and SDH equipment must not exceed.
If a time deviation or maximum time interval error limit has been exceeded by equipment in the field, additional information might be required to troubleshoot a problem. Field experience shows that a time interval error plot that shows the phase of a signal over time is important. This data, along with an accurate time stamp, should be retained for several days.
To ensure that payload jitter stays within acceptable limits in large Sonet networks, Bellcore (GR-253-CORE) and ANSI (T1.101-1994) have specified limits on the wander generated by network elements and on the maximum wander amplitude that may appear on output signals at network interface points, respectively. These specifications cover five areas:
Wander generation. This specification limits the maximum time interval error and deviation on the output line signals generated by a network element that is timed by a wander-free reference that possesses a specified level of noise jitter in the 10- to 150-H¥jitter frequency band.
Wander transfer. This specification limits the time deviation on the output line signals generated by a network element that is timed by a wandered line signal with a specified time deviation versus integration time mask.
Phase transients. This specification limits the maximum time interval error on the output line signals generated by a network element during synchronization rearrangements, such as an internal switchover from one reference signal to another.
Derived DS-1 wander. This specification limits the maximum time interval error and time deviation on the digital signal, level one output timing signals generated by a network element that is timed to an optical carrier, level n line signal.
Interface wander. This specification limits the maximum time interval error and time deviation on all output signals appearing at Sonet optical and electrical signal interface points, for example, at OC-n or STSX-n crossconnects.
The ITU-T has set similar limits on these parameters for SDH networks.
A Sonet or SDH wander measurement instrument can check the time interval error as a function of time (where the phase has been filtered by a 10-H¥low-pass filter) between an optical signal and a DS-1, E1 or other clock reference signal, or between two clock reference signals. In addition, some test instruments can generate wander on a DS-1, E1 or other reference signal for stress testing. These instruments can test all the network interface and network element wander requirements found in current Sonet/SDH standards.
Generally, wander and phase transient generation for stressing is done in design or evaluation laboratories as part of Sonet or SDH specification conformance testing. On the other hand, wander and phase transient measurements are done in laboratories, as well as in installed networks, as part of performance monitoring and troubleshooting.
Typically, a test instrument uploads the time interval error data and then calculates and plots the maximum time interval error and time deviation information. These plots are compared against standard limit plots, known as masks, to determine whether the device or network under test passes or fails.
Network element clocks
Current Sonet and SDH standards place requirements on the wander generation of the internal clocks used in network elements. Wander generation is measured by synchronizing a network element to a "wander-free" reference signal that contains a specified level of jitter and then measuring the resulting wander on an optical output line.
A special test instrument can measure wander generation on the output of a network element by using its transmitter to add wander to a specific type of synchronization reference, such as a building integrated timing supply clock signal, applied to the timing input of the network element under test. The instrument`s receiver measures the resulting wander (time interval error) on the OC-n output of the network element. This data is then uploaded to a personal computer to calculate time deviation and maximum time interval error information. Lastly, the results are plotted and compared with the standard masks.
Verification that a network element meets the wander generation and other specifications helps ensure that fiber-based Sonet and SDH networks will operate error-free. Without a specialized test instrument capable of time interval and maximum time interval error measurements and time deviation calculation and plotting, verification of a network element`s adherence to wander specifications becomes difficult and time-consuming. u
Dan Wolaver is principal research engineer, and Dana A. Cooperson is marketing manager, at Tektronix Inc., Microwave Logic Products, Chelmsford, MA.
Because wander data occurs at low frequencies, it can consist of hours of recorded phase information. To condense this information into a concise measurement of synchronization quality, key wander parameters have been defined.
Time interval error (or wander in nanoseconds). A wander measurement requires that a "wanderless" reference be designated. This reference is usually a cesium-based or global positioning satellite-based clock. The wander, or time interval error, at the beginning of a measurement is defined to be zero. An error plot typically shows the phase changes recorded since the measurements began. Consider, for example, that the initial phase difference between the reference clock and the measured signal is 3 ns. Then, 100 seconds later, it is 22 ns. Consequently, the error value at 100 seconds is 19 ns (22-3=19). The example plot depicts measurements in a synchronous optical network on an OC-12 (622-megabit-per-second) optical line.
Maximum time interval error (related to peak-to-peak wander, in nanoseconds). The maximum time interval error is used to bound peak-to-peak phase movements and frequency offsets at network interfaces. This error, which is calculated from the time interval error, is a measure of wander that characterizes frequency offsets and phase transients. Another value--maximum time interval error (S)--is defined as the largest peak-to-peak wander in any window of length S. Note that high synchronization quality is determined by a low maximum time interval error value. The example plot is calculated from the data above, plus the requirement mask that sets the maximum time interval error limits for wander generation on the OC-n line, where n = 1, 3, 12, etc., according to the synchronous optical network hierarchy. The error is used to characterize phase transients because it captures worst-case information.
Time deviation (related to root-mean-square wander, in nanoseconds). The time deviation is used to define the amount of phase noise at the network interfaces. It is a measure of wander that characterizes its spectral content, that is, the frequency(ies) where wander is concentrated. The definition of time deviation (t) is the root-mean-square of the filtered time interval error, where the bandpass filter is centered on a frequency of 0.42/t. High synchronization quality is determined by a low deviation value. The example plot shows data calculated from time interval error data, and the requirement mask that sets time deviation limits for wander generation on the OC-n lines. The deviation is considered a superior statistic for steady-state performance because it is not dominated by one or two worst-case events, as is maximum time interval error.