Transponder manufacturers debate electronics

By STEPHEN HARDY

While the XFP 10-Gbit Small Form Factor Pluggable proponents celebrate the release of electronics from 10-Gbit/sec optical transponder/ transceiver modules (see page 1), several transponder vendors say systems designers want modules to incorporate more electronic functions, not fewer. XENPAK manufacturers aren't about to scrap their designs, either. So are electronics moving out of 10-Gbit/sec optical modules or migrating in? The consensus answer among vendors polled at the Optical Fiber Communications (OFC) Conference in March is "both."

Optical modules come in two forms: transceivers and transponders. Traditionally, transceivers do little more than convert optical signals to electrical and vice versa at a certain line rate. If the optical transmission rate is greater than what the line-card electronics can support, multiplexing and demultiplexing functions (and potentially other physical layer signal processing) must be added to the board to accommodate this mismatch. Transponders gained popularity because they incorporated these physical layer (PHY) functions into the module, which enabled systems designers to add an optical interface to their products without having to worry about including the translation functions in their line-card designs. Several transceiver designers reacted to this trend by incorporating PHY electronics into their devices as well; however, most still called their offerings transceivers, which can cause confusion when comparing optical modules.

Regardless of what they call their devices, several manufacturers of transceivers and transponders report that the desire to avoid putting PHY functions on the board remains strong. Gadi Lenz, co-founder and chief operating officer of Kodeos Communications (South Plainfield, NJ), and Brian Schrader, the company's vice president of marketing, say that metro and long-haul systems designers looking at their new Optical Modem have expressed a strong interest in adding forward error correction (FEC) to the current roster of in-module electronic functions. Any exceptions to this trend generally come from systems houses that have a proprietary FEC approach, Lenz reports. Amit Jain, vice president and general manager of Vitesse Semiconductor Corp.'s Optical Systems Div. (San Jose, CA), suggests that developers of ultra-long-haul and submarine systems frequently wish to do a lot of the electronics on the board.

Dennis Chu, director of marketing at Oki Semiconductor's Optical Components Div. (Sunnyvale, CA), agrees that developers targeting telecommunications applications, particularly those who need control signaling to support voice communications, are likely to want to keep the electronics in the module. Both Fred Leonberger, chief technology officer at JDS Uniphase (San Jose), and K.C. Wang, vice president of advanced products at OpNext Inc. (Thousand Oaks, CA), suggest that advances in lower-cost PHY electronics would also support the retention of electronics in optical modules.

However, these sources and others say that onboard electronics approaches such as outlined in the XFP multisource agreement (MSA) may prove popular for LAN applications, with Vitesse's Jain adding very-short-reach installations to this category. (Jain suggests that transponders smaller than typical 300-pin MSA devices, such as Vitesse's, may compete in LAN applications as well.) The attraction of XFP will depend on how robust the optical signal needs to be and how far the transmission needs to travel. The shorter the distance, the less the requirement for FEC and other types of signal conditioning and control, and the more stringent the cost sensitivity, the more likely an approach such as XFP will find favor, say the transponder sources.

But in the longer term, several observers suggest that XFP represents the future of 10-Gbit/sec optical modules. For example, Steve Perna, vice president and general manager at PMC-Sierra's Optical Networking Div. (Burnaby, British Columbia), points out that a similar migration occurred at 2.5 Gbits/sec as CMOS processes advanced enough to accommodate the line rate. Warner Andrews, vice president of marketing at Picolight Inc. (Boulder, CO), agrees. However, he points out that XENPAK has an advantage in the near term, even in LAN applications, because its 10-Gigabit Attachment Unit Interface (XAUI) is an established part of the 10-Gigabit Ethernet (10-GbE) standard and chips supporting this interface are available from multiple sources.

Shuo Zhang, Ethernet optics product marketing manager at Agilent Technologies Inc.'s Networking Solutions Div. (Santa Clara, CA) also believes that the momentum behind XENPAK will carry a lot of weight with many 10-GbE systems designers. However, storage-area networking equipment is another matter.

"Storage is just beginning the first phase of 10-Gbit link design," Zhang reports. "The XFP MSA and serial electrical interfaces are of particular interest to this segment. Several of the key OEMs believe that serial electrical interfaces are necessary to meet their system needs." Some designers are looking at XAUI as a proven interface that is easy to implement, he adds.

"The nature of this market is that early system design slots are important in the long term," he concludes.