They're not as flashy as 100 Gbit/s. In fact, they're running at 10 Gbit/s.

But tunable XFPs -- optical modules that are driven by tunable lasers and pluggable into XFP-sized slots -- are going to be one of the hot topics at OFC/NFOEC next week, and probably for the next year or more.

Components vendors have talked about tunable XFPs for a few years, but 2011 is going to be the year that substantial shipments start rolling.

"It's going to be the really big optical technology this year," says Andrew Schmitt, an analyst with Infonetics Research Inc. "Twelve months ago, everybody already loved it. Cisco loved it. But there was only one supplier. When you have one supplier, it's like having no suppliers."

JDSU and Emcore Corp. (Nasdaq: EMKR) were the first to announce devices, in 2009, and JDSU was the first to ship. (See JDSU, Emcore Shrink Tunables.)

Most major module vendors now have tunable XFPs that they've announced or will announce at next week's OFC/NFOEC. The list includes Emcore, Finisar Corp. (Nasdaq: FNSR), Fujitsu Optical Components Ltd. , Menara Networks , Oclaro Inc. (Nasdaq: OCLR) and Opnext Inc. (Nasdaq: OPXT).

"This is going to create the environment that will make it a viable commercial technology," Schmitt says.

If JDSU's experience is any guide, the market is ready. In the fourth quarter of 2010, tunable XFPs accounted for more than 10 percent of JDSU's communications optics revenues of $168 million. Schmitt expects the tunable XFP market to reach $100 million in 2012, compared with $30 million in 2010.

Read on to find out more about these little beasties.

— Craig Matsumoto, West Coast Editor, Light Reading

Performance
Tunable XFPs are arriving in two batches. Early designs target metro/regional distances of about 80km, replacing fixed-wavelength XFPs. This is the type that JDSU first shipped.

Other modules target long-haul applications, aiming for performance good enough to replace 300-pin transponders. This case requires higher performance -- specifically, better optical signal-to-noise ratio. These devices haven't begun shipping yet.

Vendors are tending to use the same design for both types of parts. They're differentiated by chirp: the metro/regional modules typically use negative chirp in the optical signal, while the long-haul parts have zero chirp, because it's assumed the signal will be going through an amplifier at some point.

Usage cases
In metro/regional networks, tunable XFPs will simply replace fixed-wavelength XFPs. One big advantage here is the simplification of inventory -- the same case that made tunable lasers so exciting when they first came out. Instead of stocking one module for each wavelength, which meant having to keep 40 or 80 piles of the things, carriers can now just stock a tunable.

Tunable XFPs also fit in with the idea of the flexible optical network, where wavelength assignments can change quickly and can easily be rerouted through Reconfigurable Optical Add/Drop Multiplexer (ROADM)s. Operators can now consider having ports change colors, so to speak.

On the long-haul side, the tunable XFP can make a dramatic difference, because it's replacing the larger 300-pin transponder. A line card previously holding one transponder can be designed to fit four tunable XFPs -- or eight, as some JDSU literature cites -- so the potential is there to at least quadruple system densities.

"There hasn't been a lot of refresh in 300-pin designs in the last few years. They've really squashed all out of it that they can," Schmitt says.

The big arguments will likely be around 300-pin replacement parts. "There is considerable discourse amongst vendors about who can and who cannot meet the specs of 300 pin transponders," wrote analyst Ed Zabitsky of ACI Research , in a report on JDSU's fourth-quarter earnings.

Integration
Primary pieces of a tunable XFP include a tunable laser (duh), a modulator, a wavelength blocker, and a Semiconductor Optical Amplifiers (SOAs).

JDSU and Oclaro pride themselves on monolithically integrated parts -- that is, they've got these elements designed onto one indium phosphide chip. One advantage to this is that there's no need to align the parts so that the light passes through correctly. Integration removes that assembly step.

Other vendors, including Emcore and Opnext, counter that assertion by pointing out that integrated devices tend to have lower yields. In the end, that results in higher costs, they say. They also note that an integrated part isn't as flexible, which can be a factor considering carriers' requirements can differ slightly. As you might have deduced, Emcore and Opnext are using hybrid integration.

Page 3: Vendors

The table below indicates the time frame for tunable XFPs to arrive on the market. Not all information was available at press time; specifically, Fujitsu and Menara hadn't been polled about their statuses yet. Light Reading intends to update the table during OFC/NFOEC.

Table 1: Tunable XFP Shipment Timetables

Back to Page One: Introduction

— Craig Matsumoto, West Coast Editor, Light Reading