Ethernet’s Impact on Optical Component Testing

Rick King

The optical networking industry is redirecting its focus. Unlike the new-product development feeding frenzy of the 1990s, carriers and service providers now want to leverage their infrastructures and concentrate on cost-effective ways to deliver services profitably to end users -- the commercial and residential customers at the network edge, where bandwidth demand is insatiable because of Internet use and multimedia data-hungry applications.

Redirection will require low-cost technologies, modular plug-and-play components, compatible interfaces and system interoperability on both the manufacturing and test sides to stretch network viability to maximum levels. Equipment manufacturers have conformed accordingly by taking a 25-year-old technology that has low implementation costs, a reputation for solid reliability and a track record for relatively simple installation and maintenance, and developing it to the next level.

Figure 1. Because component suppliers and systems manufacturers are pushing the limits as never before in the 10 Gigabit Ethernet space, component testing is the key to helping determine where those limits end.

The Gigabit Ethernet standard, which is a staple in corporate and public data networks, fueled the move of Ethernet from local area networks into metropolitan area networks. With the advent of the 10 Gigabit Ethernet standard, this indomitable technology is primed to tackle the metropolitan area network head-on as the low-cost alternative to SONET and its international counterpart, synchronous digital hierarchy (SDH). It will bring seamless compatibility with time division multiplexing and dense wavelength division multiplexing optical networks, and establish a common technology at SONET OC-192 rates that unites the local, metropolitan and wide area networks for end-to-end optical Ethernet networking.

As a result, suppliers of optical test equipment have their work cut out for them. They must maintain a balance between providing instrumentation for legacy local and metropolitan area network technologies while developing solutions for a continually evolving Ethernet technology with new data rates and with higher speeds. They must do so while remaining mindful of interoperability and modularity, cost savings and fast network deployment demanded of their customers' products, as well as the need to comply with new standards and specifications.

The 10 Gigabit Ethernet standard (IEEE 802.3ae), for example, includes a specification for degraded signal testing to speed up deployment of the technology using a worst-case method. The stressed receiver sensitivity test defines a new method for receiver performance measurement using a precisely degraded stimulus known as a stressed eye. To address this whirlwind of standards and customer demands, modular plug-and-play test equipment may be the easiest answer.

Ethernet vs. SONET

Designed as a technology for carrying voice traffic, SONET's bailiwick has been the network core, where the highest speeds are required. When the first 10 Gigabit SONET implementation was deployed a few years ago, the main criterion was performance, with cost a much lower priority (Figure 2). This environment used externally modulated lasers to emit cleaner, higher-power signals transmittable over long distances. With fewer interconnects in the core, the ability to control which receiver would receive and tune the transmission was much greater. The core, which has attracted a smaller number of players, was left relatively "pure" and uncluttered. The transmitters and receivers used were proprietary devices generally from the same manufacturer. Interoperability standards existed but took a back seat. The emphasis was on technology innovation.

Figure 2. The first 10 Gigabit SONET implementation used externally modulated lasers to emit clean, high-power signals transmittable over long distances. Note this signal's high power level, good signal-to-noise ratio and high fidelity.

Now cost has assumed center stage, thanks to a demand for low-cost Internet protocol services, plus surging data traffic fueled by projected growth in e-business commerce, Internet protocol virtual private networks, hosting services and voice-over packets. Service providers are demanding higher-capacity solutions that simplify and reduce overall network connectivity costs while enabling service differentiation and high reliability.

Higher bit rates and the extension of reach bring the Ethernet standard to a new level but maintain its traditional benefits over SONET/SDH of cost, convenience and service. Gigabit Ethernet has lower per-port costs through more cost-effective equipment and packet overhead savings. Contributing to this is its flexibility and the advantage of dynamic bandwidth. With SONET, on the other hand, as a customer's bandwidth demands increase, additional bandwidth must be purchased in large increments at fixed prices, whether or not the bandwidth is fully employed.

Ethernet providers also can supply bandwidth in a matter of days, compared with alternate options, and often in a matter of hours. In addition, end users have access to a suite of services outside the traditional data transport services, such as regional Ethernet-to-Ethernet connectivity, and to multimedia applications such as real-time video and voice over IP.

Overall, 10 Gigabit Ethernet promises to promote efficient high-speed networks at the edge with its easy, uncomplicated migration to higher-performance levels without disruption; its low cost of ownership, acquisition and support; its reduced learning curve with familiar management tools and a common skills base; its support of new applications and data types; its flexible network design; and its use of low-cost, modular, plug-and-play components from multiple equipment manufacturers that enable interoperability between assemblies.

With the evolution to 10 Gigabit Ethernet, end users should expect the technology to compete head-tohead with SONET when it comes to speed and distance. With the latest specification, Ethernet can run on a fiber optic line up to 40 km long. The specification also allows equipment manufacturers to add logic to the physical layer to convert traffic to SONET/SDH OC-192 rates. The physical layer will clock traffic into the local area network at 10.3 Gb/s but will match SONET/SDH at 9.95 Gb/s. The relatively simple logic involved should not significantly affect the cost of Ethernet in the metropolitan area network, and some equipment vendors have even estimated that Ethernet in this network may cost 80 percent less than SONET/SDH.

One avenue that IEEE addressed to help reduce costs in the network environment was to make the 10 Gigabit Ethernet standard compatible with directly modulated lasers, but this meant considering compromises. The optical signals of directly modulated lasers possess much more overshoot and ringing than their externally modulated counterparts. And lower power output means that the signal-to-noise ratio is not as good, making the resulting signal more challenging for the receiver to analyze and translate back into data (Figure 3).

Figure 3. Directly modulated lasers can drive down network costs, but network designers must make some trade-offs because their signal-to-noise ratio is not as good with externally modulated lasers. The signal shown is typical of a lower power signal used for data transfer.

Interoperability and modularity also are key to the cost equation, in light of the objective of making Ethernet components commodity items. The issue is to enable the migration to low-cost technologies that meet acceptable requirements for the applications for which they are intended and to promote the plug-and-play concept. This would enable local and metropolitan area network operators to purchase 10 Gigabit Ethernet products from an assortment of vendors and to plug them into their systems for instantaneous operation with other Ethernet technology.

IEEE also has spent a considerable amount of time defining form-factor specifications that focus on interoperability. For example, the Xenpak form factor allows optical transceivers from various vendors to be used interchangeably, representing a major step in adopting 10 Gigabit Ethernet products. In addition, it will permit a single product -- a blade for a modular chassis-based switch or an uplink module for a stackable or fixed-configuration switch -- to use the various 10 Gigabit Ethernet media options and consequently support different fiber types and distances.

The impact on testing

With component suppliers and systems manufacturers pushing the limits in the 10 Gigabit Ethernet space to accommodate the low-cost demands of service providers, testing has never been more critical to help them judge where those limits end. The IEEE 802.3ae standard requires that the test signal be purposefully degraded for receivers and transmitters carrying 10-Gb local or wide area network traffic. This means that manufacturers who want their equipment included in 10 Gigabit Ethernet systems deployed in the network must use worst-case testing techniques. The device under test should function within allowable parameters and interoperate on the day it is shipped, without the need for tuning.

One example is stressed eye testing, which entails testing the device's eye pattern and sinusoidal jitter tolerances under specific conditions. With the new testing, the designer can take a pristine signal source and add noise and jitter to see how much signal fidelity can be degraded.

In the stressed eye test for receivers, the eye is partially closed vertically and horizontally, while carrying a test pattern generated from a laser with a poor extinction ratio. This Layer 1 test simulates a worst-case transmitter and link feeding a receiver. The standard requires that the minimum bit error rate be met with the degraded signal feeding the receiver. Because all 10 Gigabit Ethernet manufacturers must meet the specification with these signals, interoperability is assured.

Besides this advantage, the approach should reduce costs and shorten the trial-and-error feedback process among component suppliers, systems manufacturers and service providers, thereby speeding the deployment of 10 Gigabit Ethernet systems into the network. Several commercial stressed-eye-testing products are already on the market.

Interoperability and modularity are top priorities with test equipment manufacturers, who are walking a delicate tightrope. They're trying to be backward-compatible with the various Gigabit Ethernet data rates -- the SONET-friendly 9.9-Gb version and the 10.3 version -- while also moving to accommodate the constantly proliferating new data rates such as the 11.1-Gb, forward-error-correction version, which enables transmission over a longer distance and correction of more errors. Clearly, the need for modular solutions to test multiple rates is an emerging issue on the test equipment side.

Meet the Author

Rick King is vice president of the optical product line at Tektronix Inc. in Beaverton, Oregon.