Stressed receiver sensitivity testing has evolved into a critical procedure in the singlemode realm--and multimode testing is expected to follow suit.
By JOHN FRENCH
Circadiant Systems Inc.
Though new and relatively complex, singlemode 10-Gigabit Ethernet (10-GbE) stressed receiver sensitivity (SRS) testing has evolved into a critical requirement for ensuring compliance and interoperability of equipment and devices within a 10-GbE network. Similarly, as the market for lower-cost transceivers based on vertical-cavity surface-emitting laser (VCSEL) technology increases, multimode SRS testing is gaining widespread acceptance.
Until recently, automated multimode SRS test equipment has been unavailable, making it difficult to produce truly compliant multimode components and systems. The multimode market has continued to evolve and is expected to grow significantly in the next few years. With that growth comes the challenge for test equipment providers to provide the capability for automated R&D and manufacturing testing of 850-nm multimode components and systems to meet established standards for 10-GbE (IEEE 802.3ae).
Why SRS conformance for 10-GbE?
In SONET/SDH networks, every link is engineered; to bring a link up, engineers are assigned to architect and monitor the link to assure appropriate levels of optical amplification or attenuation, dispersion compensation, polarization compensation, and other critical network characteristics necessary to achieve requisite performance levels.
However, this is not the case with a 10-GbE network link in a traditional data-communications market. The IEEE standards body has determined that Ethernet links are non-engineered; plug-and-play performance is required. Yet many of these 10-GbE links provide similar distances as SONET/SDH links and approximately the same bit rate. In fact, 10-GbE LAN links are marginally higher, operating at 10.3 Gbits/sec as opposed to 9.9 Gbits/sec on nominally 10-Gbit/sec SONET/SDH systems.
With this decision by the standards bodies, a shift in responsibility (and expense) was subtly thrust upon component manufacturers and system vendors to provide users with plug-and-play performance and vendor-blind interoperability--freeing the users from the time-consuming and expensive link engineering typically conducted by carriers. In the SONET/SDH world, these interoperability tests were generally performed using near-perfect test signals with additional margin added to the testing pass levels to compensate for predicted field conditions.
Under today's 10-GbE standards, there is no engineer to ensure the link. Testing performed by network equipment and system vendors therefore must be more rigorous to guarantee plug-and-play performance and interoperability in the field. As a result, the requirement shifts from performing tests with perfect signals to performing tests with impaired signals that mimic the worst-case allowable conditions that might actually exist in the real world.
An SRS test consists of four basic impairments to an otherwise perfect optical signal, all applied to the signal simultaneously. The four impairments are 1) a poor extinction ratio; 2) horizontal timing jitter; 3) vertical (or amplitude) jitter; and 4) slowed rise and fall times.
With these impairments placed on the optical signal, system performance is then judged for acceptable bit-error rate (BER) performance at low light levels combined with low optical modulation amplitude. By testing each component and system under the same conditions that could exist in the field, the risk of actually deploying a faulty or poorly-performing network element is greatly reduced, plug-and-play performance is evident, and the probability of achieving vendor-blind interoperability is enhanced.
Although the standard was defined for both singlemode and multimode devices, there hasn't been much activity in the multimode market until recently. There is a growing trend toward more activity for multimode components and systems in the market--necessitating a closer look at the special characteristics of multimode components and the availability of multimode SRS test equipment.
Sending very high-speed signals through a multimode fiber is more problematic at 10 Gbits/sec than at 1 Gbit/sec. At 10 Gbits/sec, each bit is afforded just one-tenth of the time, and if signals walk off by as much as a tenth of a nanosecond, problems are more likely. As a result, the multimode fiber and multimode set-ups for 10 Gbits/sec are much more sensitive than either singlemode at 10 Gbits/sec or multimode at 1 Gbit/sec.
Although the requirements for the four signal impairments are the same for multimode as singlemode, the values or amount of impairment in each case--singlemode vs. multimode--are different because multimode systems behave differently in the field. Multimode SRS test sets are actually very similar to singlemode sets, but the impairment-setting "knobs" are adjusted differently. Additionally, all components--including attenuators, fiber, connectors, couplers, switches, and transmitters--must be multimode.
An SRS test set includes jitter generation and measurement equipment to calibrate the amount of jitter placed on the signal. A bit-error rate test (BERT) is required to send data. There must be a source of vertical interference, typically another synthesizer and a mixer. In addition, a set of Bessel-Thompson filters is necessary to slow the rise and fall time of the data stream, and an instrument-grade transmitter is needed to adjust for a poor extinction ratio.
Although there are several solutions available to create all these signal impairments, these solutions typically require a separate box or boxes for generating individual impairments. This configuration requires a full rack of equipment tied together with another box for sending the output through an attenuator into a scope to view and adjust until the requisite impaired signal is produced. The signal is then applied to the device under test (DUT) and the results are put into a BERT receiver to count errors.
Single-box test set
Less than two years ago, the first SRS test set emerged for singlemode SRS testing, incorporating the means to apply all four signal impairments to a DUT using a single piece of test equipment. In anticipation of the need for similar multimode test sets, a multimode SRS test set recently debuted at the Optical Fiber Conference in March.
At the conference, a private area was set up for testing different multimode devices from several manufacturers. Since most of the devices had already undergone testing with perfect signals, the results were interesting. Potential problems were found in several devices that did not show up during initial in-house testing with a "perfect eye."
If perfect signals are not likely under real-world conditions, why would anyone test equipment using perfect signals? To be fair, the 10-GbE standard is relatively new. Also, most of the 10-GbE people are coming from an IP world and simply have a different mindset--mainly that everything should just plug in and work. They see no need to spend extra money on tests.
But now that the standard has been ratified, SRS testing has become critical in singlemode environments, and it's quickly evolving to the multimode realm. SRS testing is the key indicator to ensure plug-and-play repeatability. Even though many other tests check for various equipment operating characteristics, the SRS test is the crucial test from a figure-of-merit standpoint.
Today's 10-GbE market is poised for rapid growth, and an increased recognition of the importance of SRS testing is growing alongside it. SRS testing is no longer simply a well-recognized procedure; it is a critical step to ensuring interoperability for both singlemode and multimode components and systems. With test equipment currently available, vendors and manufacturers are obligated to ensure their products meet the criteria set forth by the IEEE 802.3ae standard--and the interoperability and plug-and-play requirements demanded by their customers.
John French is executive vice president for Circadiant Systems. He may be contacted at firstname.lastname@example.org.