XFP not a sure thing, yet

Nov. 1, 2003
The market for 10-Gbit/sec transponder/transceiver modules is surprisingly active, given their lack of actual deployment in today's networks. Four competing multisource agreements (MSA) may account for some of this action (five, if you count the incumbent 300-pin configuration). Everyone seems to have an opinion about which form factor and laser technology is appropriate for what application. While it may be a bit early to declare a market winner, many vendors are backing the XFP—and Lightwave Europe's online pundits seem to agree. When asked what the future holds for XENPAK, XPAK, and X2 transponders in light of the emerging XFP form factor, the largest percentage of respondents indicate that those transponders are moving toward obsolescence. This transition isn't likely to occur overnight, however.

Established in 2002 by the likes of Finisar and JDS Uniphase, among others, the XFP MSA outlines an ultra-small form factor. Higher-density platforms and chassis will be needed in the future to drive per-port costs down, and the XFP transceiver may well emerge as the best option. Unlike the other modules, the XFP transceiver does not include an internal multiplexing/demultiplexing function. All multiplexing/demultiplexing is done on the host board, which affords system designers more flexibility.

Early XFP adopters include SONET system vendors replacing their 200- and 300-pin transponders. Many of these slots should be replaced—albeit gradually—by XFP modules, claims Rami Kanama, product manager of transceivers at Infineon Technologies AG (Munich).

That said, any widespread transition from legacy transponders to XFP transceivers will be delayed until volumes increase, which will in turn lower costs enough to justify network deployment. In the interim, look for XENPAK and its smaller XPAK/X2 cousins to gain some degree of industry acceptance.

A key advantage of the XENPAK module is its support of "every flavour of optics you could dream of for 10 Gbits/sec," says Warner Andrews, vice president of product marketing at Picolight (Boulder, CO). This, along with XENPAK's support of existing chassis, may keep it a viable option for OEMs at least in the near-term. Demand for 10-Gbit/sec-capable systems isn't exactly soaring, and it's expensive to requalify existing systems for the new transceivers, notes Andrews.

XPAK/X2 transponders may also find applications in the network. Developed to give OEMs a wider range of options for system design, these form factors are similar to the XENPAK, only smaller. Infineon's Kanama believes the XPAK with integrated physical layer will emerge as the module of choice for Fibre Channel and Ethernet applications. Early Ethernet system vendors adopted XENPAK modules, but most new designs for Ethernet should use XPAK, he says.

While each form factor seems to have its inherent advantages and disadvantages, they are all trending in the same direction: Everyone wants a smaller, more intelligent, lower-cost module. Today's devices are designed to deliver increased quality of service to the end user.

Unlike 300-pin transponders, the four MSA devices are now mostly hot-pluggable; the modules can be plugged in or replaced without powering down the system. New-generation devices—particularly XFPs—also feature digital diagnostics and optical monitoring, enabling OEMs to perform preventive maintenance on both the transceiver and optical connector. This function allows the host board to query each transceiver for its current operational parameters, including transmitted and received optical powers, internal module temperature, laser bias current, and supply voltage. Both Finisar and Ignis Optics, for example, recently unveiled XFP transceivers that provide not only dynamic diagnostics, but also static, customer-specific data, including vendor/module identification and link type.

Emerging 10-Gbit/sec modules also consume less power than their 200- and 300-pin MSA predecessors. The first transponders on the market consumed about 15 W of power. Today's modules need just 2 W of power, thanks to improvements in material processes. First-generation transponders utilised gallium arsenide, then silicon germanium. Now manufacturers are tapping into 0.18-µm and more recently 0.13-µm CMOS processes.

And that seems to be the tip of the iceberg. As volume ramps and technology improves, transceivers will be made even smaller, more intelligent, and more efficient. "We think at 10 Gbits/sec, there's likely to be some activities to lower the cost even further, which will be enabled by advancements in silicon," surmises

Picolight's Andrews. "What we need to see now is a volume market appear for one of these form factors, and then we can all put our heads together—the silicon vendors and the optics vendors—and come up with even lower-cost silicon form factors."