Jitter and wander measurements keep SDH/SONET in synch

Data integrity in today

Wolfgang Miller, Acterna, Germany -- Test & Measurement World, 6/1/2001

New and revised standards for the jitter and wander properties of SDH/SONET system components lead to some new challenges when it comes to measuring these properties. Further developments of SDH/SONET and development of the new optical transport network (OTN) will make even greater demands on jitter and wander test equipment in the future. Major challenges include higher bit rates such as STM-256 at 40 Gbit/s and measurement of clock quality during transmission of user signals via the OTN.

Jitter and wander refers to short- and long-term movements in signal rates across the network. SDH/SONET are, by definition, synchronous systems. Therefore, phase stability of clock and data signals throughout the network is fundamental. In practice, network elements experience interfering factors that can affect synchronisation. The results are bit errors, slips, data loss, and frequency interference that have the potential to impair transmission quality. Jitter and wander measurements quantify these errors and enable network operators to maintain synchronism within acceptable limits.

To achieve international compatibility at network interfaces, the ITU-T and other standardisation bodies stipulate maximum permissible limits for jitter and wander. These limits apply to network interfaces when you incorporate system components into a complete network. Important examples include ITU-T Recommendations G.823, G.824 and G.825 (see Table 1).

Developers of system components have particular interest in how a network element will behave as an individual component under ideal conditions. Suitable requirements have been derived from the above standards and incorporated into the standards for system components — e.g., ITU-T Recommendations G.783 and G.781.

The requirements break down into three categories:

• Jitter and wander at output interfaces,
• Jitter and wander tolerance of input interfaces, and
• Jitter and wander transfer functions.

Output interfaces may not exceed certain limits for jitter and wander within defined frequency ranges. Input interfaces should also tolerate these permissible amplitudes with no problems. The third item — transfer functions — requires that the relationship between the input and output of a network element meets certain criteria.

Measuring instruments to analyse these properties will provide comparable results only if they use comparable measurement techniques with the same degree of accuracy. This is why the ITU-T created Recommendations O.171 and O.172 for jitter and wander measuring instruments (see Table 2). Based on experience gained from Recommendation O.171 for PDH applications, a new Recommendation known as O.172 was published in 1999 for measurements on SDH equipment. O.172 takes into account the special requirements of this newer technology.

The Box “Jitter Jargon” explains some of the terms you’ll frequently come across when making jitter and wander measurements.

Measuring output jitter

Jitter exists at the output of a network element even if its input signal — used for synchronisation — is ideal. This “intrinsic jitter” arises in the network element’s own circuitry. Possible sources include:

• Phase noise in clock oscillators,
• Spurious response of crystals used in clock oscillators,
• Crosstalk to the clock supply from other system components,
• Pattern-dependent delay in scramblers and encoders, and
• Insufficiently sharp edges in the digital signal.

When you develop network elements, look at the output jitter early on to ensure that it doesn’t violate maximum permissible values. You should measure the values within specified jitter bandwidths. Recommendations generally define two jitter limits: one value for high-frequency jitter and one value for broadband jitter.

Figure 1. Jitter measuring instruments compare transmitted and received signals.

Figure 2. Synchronising the measuring instrument and the network element with a reference avoids jitter components due to uncontrolled pointer activity when measuring pointer jitter.

Figure 3. The reference clock simultaneously synchronises the network element and measuring instrument to enable relative wander measurements.

Connect the signal under test to the input of the jitter measuring instrument (see Fig. 1). If necessary, you can use the output signal from the test instrument to provide a valid signal to the input of the device under test. Recommendations don’t usually specify a measurement time, but between 1 and 15 minutes prove useful in practice. Most important for proper evaluation is the maximum peak-to-peak jitter unit interval, UI_pp, within the measurement interval.

Pointers pose problems

“Pointer jitter” is a particularly critical parameter that results from pointer activity in an SDH signal. Pointer adaptations cause phase hits of 8 or 24 bits in the transported signal (e.g., for a 2 Mbit/s data stream). Clock processing in the demultiplexer smoothes out these phase hits in the outgoing signal. However, some phase modulation remains, and low-frequency components predominate. The prescribed jitter measurement (high-pass) filter highly attenuates these jitter amplitudes. However, the large amplitudes that arise (several UI) have a decisive effect on the measurement result.

Recommendation O.172 describes the special properties required by instruments in order to deliver reliable and repeatable measurements. This includes compliance with the transfer function specified for the measurement filter as well as “drivable” input circuits in the jitter measuring instrument.

“Mapping jitter” has its origin in the stuffing processes used to map a PDH signal, and both jitter types are present when measuring pointer jitter. This is known as “combined jitter”, and the standards specify relevant maximum values for this parameter.

You can make this measurement by using a test generator to feed a valid SDH signal with the proper signal structure to the device under test. To avoid jitter components due to uncontrolled pointer activity, you should synchronise the measuring instrument and the device under test using the same reference clock (see Fig. 2).

For repeatable results, you need to employ standardised pointer sequences (see ITU-T Recommendations G.783 and O.172).

The test procedure must have three phases:

• An initialisation phase ensures that the buffer memory of the pointer process is in a defined starting condition. A sufficient number of pointer movements are transmitted in the same direction as the test sequence.
• A cool-down phase allows the desynchroniser’s clock processing circuitry time to settle.
• A measurement phase, during which the pointer test sequence is initiated and the jitter is measured until at least one complete sequence has elapsed. The result is the maximum peak-to-peak jitter during the measurement phase. The measurement must generally be performed with two filter combinations (f1 to f4 and f3 to f4) to meet both limits.

Further variants of this procedure use the reverse pointer direction and various offsets of the traffic bit rate to assess the worst case. These measurements can be very time-consuming unless you use test methods that automatically execute multiple test passes and generate test reports.

Pointer wander

In the measurement just described, the phase disruptions that result from pointer events also have a low-frequency characteristic, known as wander. The next obvious step is to record, besides the jitter, this wander in the wander-jitter transition zone. The newly revised draft version of G.783 also stipulates the relevant maximum time interval error (MTIE) limit curves for PDH bit rates in the 1.5 Mbit/s hierarchy. The basic test setup is the same as for measuring pointer jitter but uses a different evaluation. Now, you must determine the MTIE as a function of an observation interval between a few milliseconds to 100 seconds. A 100-Hz low-pass filter limits the measurement frequency range and your test needs to use a sampling frequency of at least 300 Hz.

SDH signals simultaneously transport timing information for synchronising network elements. Unlike jitter, though, there is no way to fully suppress wander in network elements. Wander generated in any network element thus contributes to the wander accumulation in the whole network. To minimise the effect, G.783 and G.813 give useful specifications for the clock quality of SDH network elements.

What we measure are relative wander measurements in which the network element is synchronised with a reference clock. This clock simultaneously acts as the reference for the wander measurement (see Fig. 3).

Figure 4. Wander measuring instruments show permissible wander amplitudes as limit curves for mean time interval error (MTIE) and time deviation (TDEV). Measurements usually take between 15 mins to 2 hours during which the instrument computes the MTIE and TDEV as a function of the observation interval.

Specifications predominately show permissible wander amplitudes as limit curves for MTIE and time deviation (TDEV). Your measuring instrument records the phase variation with respect to the phase value at the start of a measurement, usually at 30 samples/s with a 10 Hz low-pass cutoff. The measurement usually lasts 1000 to 10,000 seconds (approximately 15 minutes to 2 hours). Then, your measuring instrument computes the MTIE and TDEV as a function of the observation interval. Figure 4 gives a sample analysis result together with MTIE and TDEV tolerance masks.

Other wander tests measure the phase jumps when switching the incoming reference signal and for switch-over to holdover mode. Again, limit masks exist for this measurement. For millisecond phase jumps, you need to use higher sampling rates (e.g., 300 samples/s) with a low-pass cutoff at 100 Hz.

Jitter & wander tolerance

The capability to process certain jitter and wander amplitudes at signal and synchronisation inputs without errors is necessary to reliably operate an error-tolerant transport network. Jitter and wander are predominantly stochastic phenomena, making it difficult to generate such signals with precisely defined properties. Accordingly, the standard measurement uses sinusoidal jitter or wander test signals to obtain meaningful results. The specification is in the form of a minimum-tolerance curve for different jitter and wander frequencies as per Figure 5.

Figure 5. Because jitter and wander are stochastic, measurements use sinusoidal jitter or wander test signals to obtain meaningful results. Standard G.825 provides a specification as a minimum-tolerance curve for different jitter and wander frequencies.

The test increases jitter amplitude at the different jitter frequencies until bit errors (exceeding a defined threshold) occur at the output of the network element. The jitter amplitude just before the output crosses this error threshold is the maximum tolerable jitter of the input under test.

Automated test procedures for determining the jitter and wander tolerance over the entire frequency range are very useful, particularly if the results can be output as tables and graphs. Also useful are algorithms that quickly locate limit amplitudes, as are automatic test procedures for measuring the wander tolerance due to very low measurement frequencies down to 10 mHz (or a 24-hour period).

Figure 6. You can test the jitter tolerance of a network element, for example an add/drop multiplexer (ADM), by feeding in a test signal with sinusoidal jitter or wander modulation. The test increases jitter amplitude at the different jitter frequencies until bit errors occur.

 Due to their simplified clock recovery circuitry, SDH regenerators have no jitter suppression below certain frequency limits. When a network chains together large numbers of regenerators on a long communications path, the jitter amplitude grows with the sum of the jitter amplitudes of the individual regenerators. To limit this effect, individual regenerators must not amplify input jitter. Tolerance masks permit only minimal “jitter gain” that amounts to 0.1 dB in the passband.

During a test, the measuring instrument feeds a signal modulated with sinusoidal jitter to the input of the network element under test. The test uses the largest possible jitter amplitude, but the element should easily tolerate the input. The high amplitude improves the signal-to-noise (S/N) ratio and thus the measurement accuracy. However, in case of non-linear behavior of a device under test, different amplitudes can result in different frequency responses. In this situation, you should perform the measurement with a small, constant amplitude.

By measuring jitter amplitude at the output of the network element you can calculate the transfer function in dB. You should repeat the measurement at a number of jitter frequencies in the passband and stopband. Spurious jitter away from the test frequency, particularly at lower amplitudes, can impair measurement accuracy. However, you can obtain accurate results by narrowband selection of the test signals, which limits spurious effects.

The maximum permissible gain of 0.1 dB (about 1%) demands extremely high precision, making it very important to take the intrinsic frequency response of the test setup into account. You can compensate for intrinsic errors by performing a reference measurement without the DUT. Today’s test sets have this capability built-in, together with automated test procedures

For more information

Acterna, “Leading the Way with Innovative Jitter & Wander Test Solutions”, Application Note 71, Wavetek Wandel Goltermann, 1999.

Wolfgang Miller is a project manager with Acterna, Eningen, Germany. He also has responsibilities for international standardisation in the area of communications technology.

1

2