XFP-E compromises size for pluggability, tunability
At the OFC/NFOEC Conference this past March, Opnext, Avanex, and Mitsubishi Electric announced a multisource agreement (MSA) for an XFP-E (extended-size) transceiver designed for 80-km and tunable DWDM applications. However, the system vendor community seems to be of two minds about the new form factor. There are those who favor the XFP-E as an interim step to achieve longer distances and tunability (goals the XFP currently is unlikely to support), and there are those who will eschew the XFP-E in favor of alternative options-or who will simply wait for the equivalent functionality in the standard-sized XFP package.
There are two limiting factors for the standard XFP: its size and its thermal constraints. The XFP-E is twice the size of the standard XFP; it is, essentially, two XFPs placed side by side but integrated into a single module. The larger size is critical for accommodating longer-reach optics-which presently tend to be too bulky for the XFP-and, ultimately, incorporating tunability.
The XFP’s mechanical constraints are directly related to its thermal capacity; longer-reach optics draw more power. Some tunable laser modules, for example, draw more than 8 W, which is beyond the XFP’s 3.5-W power budget. (In general, the smaller the form factor, the lower the power budget.) The existing larger form factors-XPAK, X2, and XENPAK-allow much higher power consumption. The XENPAK, for example, features a 9-W power budget, while the X2 can accommodate 4 W. By contrast, transceiver manufacturers are attempting to achieve a 1-W power budget for the emerging SFP+ form factor.
Long-reach SONET applications normally consume at least 3 to 4 W, notes Shing-Shwang Yao, senior engineering manager in Mitsubishi Electric’s USA Semiconductor Division (http://global.mitsubishielectric.com). With that kind of power consumption, the temperature of the package or case will increase. The cooling capacity of the standard XFP is limited to 30° to 40°C. “If a customer uses several packages in a series from upstream to downstream, the last one would be hotter because all the cool air is becoming warmer and warmer,” explains Yao. “The temperature rise would be very high at that last one, the downstream one. We need to have a larger case area in order to dissipate more heat easily.”
Thanks to its increased surface area, the XFP-E features a power budget of around 6 W and an extended temperature range of -40° to +85°C, enabling its use in longer-reach DWDM applications. “It is important to keep the laser diode chip at a certain temperature range of 25°C or 30°C to keep the wavelength drift smaller,” says Yao. “Because the wavelength demand is high, [the laser] cannot be allowed to shift more than 0.4 nm.”
Of course, if size is the XFP-E’s chief differentiator, one wonders why system vendors do not simply implement the existing XPAK, X2, and XENPAK form factors. According to Ed Cornejo, director of product marketing at Opnext (www.opnext.com), it makes sense to implement an XFP-E instead because it is electrically and optically compatible with the standard XFP transceiver. “The electrical interface doesn’t change, which means you’re still supporting a serial electrical or ‘Ziffy’ [XFI] interface,” he explains. “X2 or XENPAK would force you to use XAUI, and [system engineers] don’t want to do that. They prefer to just extend the mechanical package a little bit but keep the electrical the same.”
One potential reason why a system vendor might not embrace the doublewide XFP-E is its port-density limitations. Because the XFP-E consumes more physical space, its port density tops out at around eight, says Cornejo. If the system vendors “are trying to pull more than eight 10-gig ports out of the same panel, then [the XFP-E] is a disadvantage,” he notes. “It really comes down to your port-density or bandwidth plan. How much bandwidth are you looking to get through your line card? That becomes an issue.”
Interestingly, all three founding members of the XFP-E MSA report that they also are working on a tunable standard-size XFP. “We’re still not there, technology-wise,” admits Cornejo, “but we certainly plan to get there.” As such, he views the XFP-E as “an interim step.”
The folks at Avanex agree. “Our belief is that there is room for the XFP-E as the first step in the introduction of tunability into pluggables,” says Yves LeMaitre, vice president and general manager of Avanex’s Optical Components Group (www.avanex.com). “It’s kind of the Holy Grail right now in transponders/transceivers, to have a tunable/pluggable or pluggable/tunable.” LeMaitre does not see anyone overcoming the mechanical and thermal challenges of the standard XFP within the next two years. XFP-E is, therefore, a “good compromise” to give system designers a faster transition to pluggability and tunability, he says.
In fact, pluggability was one of the main goals of XFP-E MSA. “We said, ‘Let’s try to move everything pluggable as fast as possible,’” recalls LeMaitre. He cites the trend toward more modular architectures-a result of outsourcing engineering and design activities-as a key driver of pluggability. Instead of redesigning or respinning their line cards with each new technology implementation, pluggability enables OEMs to reuse the same equipment.
“Most of the architecture should stay the same as you switch from 10 gig to super 10 gig or from super 10 gig to a tunable 10 gig or from a tunable 10 gig to a 40 gig,” contends LeMaitre. “You are really trying to get these basic building blocks that OEMs can reuse and reuse and reuse, which helps them reduce the cost of introducing new technologies. We saw that there was a strategic advantage to move to pluggables for everyone in the food chain of this industry,” he adds.
Not all system vendors have jumped on the tunable/pluggable XFP-E bandwagon, however. Cisco Systems (www.cisco.com), for one, does not see the value of the XFP-E, at least for its near-term product roadmap. “The real question here-which is an interesting observation on the whole pluggable concept-is, ‘If I go tunable, do I still need pluggability to the extent that it makes sense today?’” muses Thomas Scheibe, manager of Cisco’s Transceiver Market Group. “You could argue that once I tune everything across, say, all 32 or more channels, the advantages of having a pluggable transceiver are diminished. On pluggables, it’s a fixed wavelength; you have to have the right wavelength to plug in and out, whereas with tunables, you can just tune to whatever wavelength you want,” he notes.
For Lawrence Gasman, president of CIR (www.cir-inc.com), the value of tunable/pluggables or pluggable/tunables may be contingent on their price points. Pluggability and tunability “both increase the efficiency of operations but in different ways,” he explains. “If you could bring them together, then you’d make swapping out more efficient and carrying inventory more cost-efficient. Doing both of these things is quite attractive, as long as you’re not paying huge premiums.”
For its part, Cisco is shipping “the alternative solution,” as Scheibe calls it. “Instead of putting [the XFP-E] on a pluggable form factor, just put it on the module that sits straight on the line card,” he says. “If you don’t package it as a pluggable transceiver, you have more options for cooling. We do ship cards with tunable transceivers on the line card today.”
Beyond the pluggable/tunable question, Scheibe wonders if the industry needs yet another form factor for 10 Gbits/sec. There are five form factors to date: XENPAK, X2, XPAK, XFP, and the recently emerged SFP+. XFP-E would be the sixth. Scheibe recalls Cisco’s rationale for developing a pluggable DWDM XENPAK back in 2003. “We looked at it and said, ‘Should we do a new form factor or should we just stay with the existing one?’ We decided it’s great to do DWDM in pluggable, but you should stay with an existing form factor so you can at least take advantage of what’s out there, given that the market is really not that big.”
Gasman has had similar thoughts. Volumes for 10-Gbit/sec devices are getting bigger, and CIR is predicting 10-Gbit Ethernet shipments in the millions “a few years out,” he says. “But you don’t need a lot of different form factors for that-just the low-cost, short-reach ones that make the most sense. My guess is that quite a number of these form factors will sort of drop by the wayside over a period, or they will be out there but they won’t be widely supported. It’s difficult to imagine there could be more than four or five down the road,” he admits.
LeMaitre acknowledges that there is a contingent of system vendors who are still using and designing around the 300-pin SFF to provide advanced functionality for WDM networks. Those vendors are likely to wait for the standard-size XFP to provide the same level of functionality. Nevertheless, he says, “We feel very comfortable that there is a nice market for this type of product, where customers require longer distance, low dispersion, and probably tunability in a pluggable format. The rest is going to depend on how fast the technology is integrated at the chip level, probably for the laser on the modulator, to enable the same level of functionality in the XFP.”
In the interim, Avanex has prototype XFP-Es available and already is shipping to customers for evaluation. Opnext plans to have its version by the third quarter of this year. Mitsubishi Electric, meanwhile, was first to market with an XFP-E, announcing a series of transceivers back in September of 2005.
Meghan Fuller is senior editor at Lightwave.