Oscilloscopes step on the gas

Testers of high-speed systems such as 10 Gb/s telecom networks face two problems. First, testing at high data rates is more difficult than with slower systems. Second, detecting irregularities is more important because increased rates tend to enhance or accentuate irregularities in the signal stream. One of the more important measurements at these rates is jitter testing. While such tests are difficult at 10 Gb/s, they are much more difficult on the even faster systems (40 Gb/s and higher) now under development.

Traditionally, high-speed oscilloscopes have been used to conduct analysis of optical transmission signals. The oscilloscopes evaluate the signal integrity, also known as signal quality. Because these sampling oscilloscopes typically had bandwidths of only up to 2 GHz and sampling rates of less than 5 Gsamples/s, engineers conducted analysis by showing the data averaged over a long period using multiple acquisitions at the lower sampling rate. In addition, many high-speed oscilloscopes could only locate a signal problem, not trace it back to the cause, which often forced engineers to guess. Unfortunately, most of these constraints were created by hardware and chip limitations, so oscilloscope manufacturers could do little to correct the deficiencies.

Conducting analysis over long periods was an acceptable practice—albeit not a perfect one—when transmission rates were slower. Experienced engineers learned to narrow the possible cause of signal degradation based upon the actual glitch or error. Neither scenario, however, is acceptable given today's complex signals.

Real-time signal analysis becomes increasingly important when data is sent in packets rather than in a continual stream. With the use of data packets at higher transmission speeds, many signal anomalies can occur within a short period. To address this development, single-shot measurement capability is now common on most high-speed oscilloscopes so that analysis can be conducted on much shorter record lengths.

The ability to conduct real-time signal analysis is a feature of new oscilloscope design. Use of silicon germanium (SiGe) chips—substantially faster than conventional silicon chips—has improved oscilloscope performance. Other factors include CMOS memory chips that can accept 10 Gb/s data rates for up to nearly 50 million acquisition points, and streaming architectures that allow complex signal analysis to be performed 10 to 100 times faster than before.

These technological advances greatly improve the capabilities of oscilloscopes. The fastest oscilloscopes now have sampling rates of 20 Gsample/s and can capture signals with bandwidths of up to 6 GHz. These performance improvements offer great promise for high-speed signal analysis, especially in the area of jitter testing. For transmissions at slower rates, such as 100 Mb/s, 1 ps of jitter was not a concern, but at rates of 10 Gb/s and above, the margin for error is much smaller. One picosecond can significantly degrade 10 Gb/s signal performance, so improved sampling rate and bandwidth are imperative to conducting analysis of high speeds signals.

There are other benefits to the improvements in high-speed oscilloscopes. Engineers no longer have to guess the cause of a signal anomaly. High-speed oscilloscopes can now provide eye patterns that not only locate the problem but provide the necessary signal analysis, allowing engineers to track it back to its origin (see oemagazine, February 2001, page 48). Complementing the advances in oscilloscope design are improvements in optical-to-electrical (O/E) and electrical-to-optical (E/O) converters, as well as software.

Current O/E converters provide improvements in both bandwidth and sensitivity. Modern optical communications systems employ a broad range of optical wavelengths covering the spectrum from 700 nm through 1600 nm. Advances in semiconductor technology have led to O/E converters with responsivities in excess of 1 V/mW, covering a wavelength range from 950 nm to 1630 nm and with bandwidths of more than 3.5 GHz. These O/E converters can be used in conjunction with a real-time oscilloscope to test optical communications signals at data rates of 2.5 Gb/s.

A primary measurement of the performance of an optical transmitter is its extinction ratio (ER). The extinction ratio is defined as the ratio of the optical power of a laser's "on" state (P₁) to its "off" state (P₀). This relationship is expressed as the ratio:

Figure 1. Eye pattern from a 2.5 Gb/s transmitter shows all of the optical levels that the signal takes on as it varies between its '0" and '1' states. Standards specify regions in both amplitude and time where a compliant signal cannot deviate (dark blue areas). This signal deviates into the mask significantly during the transitions between on and off (1 and 0).

Figure 2. A filter that approximates the frequency response of an ideal optical receiver provides a more useful measure of a system performance than the transmitter pattern alone. This image is a filtered version of the eye pattern in figure 1.

Additionally, the transmitter output should be compared to a mask template that represents the compliance of the transmitter to a given standard, such as ITU-T G.958. Optical mask, or "eye mask," standards specify amplitude and time regions into which a signal cannot deviate and remain compliant (see figure 1).

The example measurement shows that the signal being tested deviates into the mask significantly during the transitions between 0 and 1 and there is a large amount of "ringing." The ringing is characteristic of the relaxation oscillations that are common in directly modulated lasers. This oscillation is normally above the frequency response range of the optical detectors used in the communications link and so is not seen by the threshold detector within the receiver. In order to make the eye pattern and extinction ratio tests more meaningful, the user can apply a filter to the output of the O/E converter in the test set to approximate the frequency response of an ideal optical receiver (see figure 2). The combination of O/E converter and filter constitutes a reference receiver and is the primary component of all time-domain measurement equipment used for evaluating optical communications transmitters.

Figure 3. Normalized attenuation of an optical reference receiver as a function of frequency normalized to bit rate (e.g. the horizontal value of 1corresponds to 2.5 Gb/s). The dotted lines represent the tight tolerances for frequency response.

Building an accurate optical reference receiver is critical because the shape of the data eye, and hence its relationship to the mask, is directly related to the accuracy of the reference receiver response (see figure 3). Because an inaccurate response can cause a good device to fail the mask test or a bad device to pass, the O/E converter, filter, and oscilloscope front end must be precisely matched. Test equipment manufacturers normally match the components during production. The oscilloscope, O/E converter, and filter are delivered to the user as a matched set whose calibration is only valid for the oscilloscope and channel on which the calibration was originally performed. To further complicate the issue, repairs or calibrations to any part of the reference receiver chain must be done together.

There is a better way. The advent of oscilloscopes with high bandwidths and sampling rates has enabled the development of digital reference receivers, which use digital signal processing to implement the reference receiver filter. The filter coefficients are computed using the known response of the O/E converter (stored in a memory chip within its housing) and the response of the oscilloscope channel, along with the desired response of the reference receiver. The filter coefficients are determined automatically when the O/E converter is connected to the oscilloscope, eliminating the need for factory calibrations. The O/E converter can be moved from channel to channel or from oscilloscope to oscilloscope and the calibration is maintained. This represents a great advantage for manufacturing up-time since repairs and calibrations can be performed separately on individual components.

An added advantage to the digital reference receivers is their ability to support an arbitrary data rate. For example, the standard specifies that the 3-dB point of the receiver must be 0.75 times the data rate. Knowing this, forming a proper filter is simply a matter of setting this data point into the filter's coefficient search routine.

Another important measurement for today's high-speed oscilloscopes is timing jitter. This is a critical measurement in optical communications systems because the ability of a receiver to track jitter in the incoming data stream is directly related to the error performance of the system. Jitter in optical communications systems is defined in terms of the time interval error (TIE). TIE is the deviation in the timing of the transitions of the digital data from their ideal (jitter-free) locations (see figure 4). The data transitions occur nominally at multiples of one unit interval (UI), which is equal to the inverse of the bit rate. The deviations from this value represent the TIE and are expressed in UI.

Figure 4. Jitter track (upper trace) and data pattern (lower trace) vs. time of an optical data transmission system. The upper trace is scaled in ps and shows the variation of the time interval error (TIE) for each bit. Note the oscillations in the unfiltered optical signal. (LECROY).

The traditional means of measuring jitter on an oscilloscope involves measuring the statistics of the eye pattern data at the zero crossing points. The peak-to-peak and RMS values of the distribution of the time instances during which the optical signal is present at the zero crossing point indicate the total jitter. This measurement, however, is only a partial indication of the performance of the optical link because it does not take into account the rate at which the TIE varies. This rate is very important since optical receivers employ phase-lock loops (PLLs), which are circuits that extract the data timing from the incoming data stream so that the bits can be decoded. The PLL can track low rates of TIE, and so its level at these rates is less critical. Jitter analyzers and time-interval analyzers have been developed specifically to measure jitter in such cases.

The high sampling rate and phase stability of today's digital oscilloscopes allow engineers to measure jitter using the same techniques employed in jitter analyzers. In a long record, which contains enough data transitions so the lowest desired jitter rate can be measured, an oscilloscope acquires the serial data signal. For example, a minimum jitter rate of 1 kHz requires a data record at least 1-ms long—10 million samples at 10 Gsamples/s. The scope determines the nominal locations of the data transitions by processing the data record through a "golden PLL." This is a mathematical computation of the average time interval between data transitions. The PLL operates by allowing the mean bit period to vary over long time intervals while remaining fixed over shorter time intervals. The operation of the golden PLL is defined by its transfer function, which has a high-pass characteristic whose corner frequency is defined by the data rate. In practice such a PLL can be implemented by a weighted, moving average of the TIE versus time, which is the equivalent of a digital low-pass filter.

The TIE is computed relative to this recovered clock so the resulting error values only contain contributions from rapidly varying time interval errors. The statistics of these filtered TIE values are analyzed to determine the total peak-to-peak and RMS jitter. The jitter can be (and usually must be for various measurement standards) further broken down into random and deterministic components by various techniques. Random jitter is noise-like and often very low. Its sources are usually thermal in nature. Deterministic jitter is related to the data pattern itself or some other systematic source such as the power supply of other digital signals.

High data rates bring with them more stringent testing parameters that have required high-speed digital oscilloscopes to undergo significant changes. The new oscilloscopes provide strong signal-analysis capabilities for faster signals due to increased bandwidth capability and higher sampling rates. Complementary advances in O/E converters and the development of software packages have transformed these traditional general-purpose electronic instruments into more customizable communications analyzers that can address the needs of the optical market.

Similar improvements and advances will continue in order to meet the faster transmission speeds currently under development. Although these higher data rates remain a few years away, the need for accurate high-speed digital oscilloscopes is ever-present. oe