IEEE task force members look to settle 10-Gigabit Ethernet PMD debate

Sept. 1, 2000



The IEEE P802.3ae Task Force, charged with establishing the specifications for the upcoming 10-Gigabit Ethernet standard, will meet this month to work out an issue that some feel has become more of a business problem than a technical matter-the physical-media-dependent (PMD) interface for multimode optical fiber.

The task force has five PMD objectives as part of the 10-Gigabit Ethernet effort, including determining PMDs to accommodate singlemode-fiber applications at distances of 40, 10, and 2 km, as well as 300 m of multimode fiber and 100 m of what is termed "installed multimode fiber," which is commonly understood to mean 62.5-micron/160-MHz-km fiber. Task force members have entertained more than 20 proposed PMDs for the various requirements, according to the chair of the IEEE P802.3ae Task Force, Jonathan Thatcher of WorldWide Packets (Spokane, WA). By its plenary meeting this past July, that list had been whittled to five. Two of these were approved at that meeting: a 1,550-nm serial interface for the singlemode 40-km specification and a 1,310-nm serial interface for the singlemode 10-km and 2-m benchmarks.

Neither of these PMDs is considered desirable for multimode-fiber applications, according to task force member Bill Wiedemann, vice president of business development at Blaze Network Products (Dublin, CA). Thus, at least one other PMD from among the remaining three-a 1,300-nm wide WDM (WWDM) proposal, an 850-nm coarse WDM (CWDM) offering, and an 850-nm serial option-must be selected. However, agreement on one or more of these PMDs proved elusive, say both Thatcher and Wiedemann, as the conflicting goals of members employed by systems suppliers, transceiver manufacturers, and fiber providers could not be reconciled.

From a technical standpoint, all three proposals meet the 300-m-over-multimode-fiber objective and both WWDM proposals meet the 100-m-over-installed-multimode-fiber objective. The 850-nm serial PMD requires next-generation high-bandwidth multimode fiber to meet the 300-m specification. In addition, the 1,300-nm WWDM option will accommodate a distance 300 m over installed multimode fiber as well as some singlemode applications. However, the task force seeks PMDs that are not only "feasible" but "optimal"-and optimal, like beauty, is in the eye of the beholder.

For example, system vendors are looking for as small a set of PMDs as possible to decrease the number of transceivers in their inventories and simplify product designs. Thatcher points out that the 10-Gigabit Ethernet standard will have two PHYs, one called WAN and the other LAN (depending on whether they naively attach to the SONET OC-192 infrastructure or are a simple "speed-up" of Ethernet, not on the transmission distances they can support). PMDs are paired with PHYs to create "ports"; with two PMDs already approved, system vendors are on the hook for four port types. Each multimode PMD approved means two more ports the systems houses may be forced to support-and no systems house wants to support 10 ports, observers agree. In voting last July, task force members from these companies showed strong support for the 1,300-nm WWDM option in the belief that it would meet the majority of the requirements their customers will face.

Transceiver manufacturers, meanwhile, are naturally interested in bolstering their product lines, so the greater the number of PMDs, the more transceivers they can sell. At the same time, different vendors have aligned themselves with different PMDs-Agilent Technologies is behind the 1,300-nm WWDM option, Blaze backs the 850-nm CWDM offering, and companies such as Picolight and Cielo are said to back the 850-nm serial proposal-and no company wants to see the PMD behind its leading product shunted aside in favor of a PMD that supports a competitor.

The third camp comprises the manufacturers of the upcoming next-generation multimode fiber, which approve of the 850-nm serial PMD and probably wouldn't mind the 1,300-nm WWDM proposal, as well.

The conflicting interests of these parties ensured that no option, or combination of options, received the 75% approval of the committee members necessary for adoption. The 850-nm CWDM option came closest, with a nearly 64% approval rating.

Neither Thatcher nor Wiedemann were willing to predict whether the task force members will be able to put aside their differences and push one or more of the PMDs through approval this month. Thatcher says that technical enhancements to one or more of the proposals, possibly in the form of the addition of equalization, might make certain PMDs more appealing. However, Thatcher says he is no more than "hopeful" that the voters will reach agreement.

"Of the three proposals that we were unable to vote in, I think the committee at large believes they are all technically feasible, that they're all reasonable technical solutions," Thatcher concludes. "So the decisions are basically decisions that are not based on 'this one is higher risk,' etc. They're very much business-related decisions compare to technical decisions."