The electronics industry has reaped the benefits of integration since the advent of the integrated circuit in 1958. The photonics industry, by contrast, had been slow to follow until Silicon Valley startup Infinera (www.infinera.com) burst on the scene in 2003 with the first commercially available photonic integrated circuit (PIC). The company recently debuted its 1.6-Tbit/sec PIC, which features a whopping 240 optical components on a single indium-phosphide (InP) substrate. Though only a lab demonstration, Infinera’s 1.6-Tbit/sec PIC is representative of the kind of Moore’s Law data rate increases the company has achieved over the last three years, and it has some of its competitors nervous.
Chip-level integration reduces cost, size, and power consumption, explains Rick Dodd, senior director of product marketing at Infinera, who notes that the company’s PICs are “essentially a WDM system-on-a-chip.” All the lasers, modulators, amplifiers, multiplexers/demultiplexers, and other components typically found in WDM systems are integrated onto one single chip, enabling “radical footprint savings,” says Dodd. The company’s 1.6-Tbit/sec PIC, for example, is the data-rate equivalent of 160 10-Gbit/sec line cards.
Perhaps the biggest advantage of a PIC-enabled system, like Infinera’s Digital Transport Network (DTN) platform, is its more cost-effective optical-electrical-optical (OEO) conversion capabilities. “We all keep thinking, ‘We need to move to an all-optical network’ because ultimately, you want to keep everything in optics, but that’s not necessarily the case,” notes Eve Griliches, research manager of telecom equipment at IDC (www.idc.com) and author of the recent report, “Leading the Charge in Photonic Integration, Infinera Gets Disruptive, Makes Waves.” Being able to do cost-effective OEO conversions when and where they want to gives carriers far more flexibility, she says.
All of which leads to the obvious question: If the benefits of photonic integration are so numerous, why is Infinera the only company with a commercially available system based on PICs? “It is a significant technical challenge,” admits Dodd, adding that Infinera required a huge capital investment — to date $205 million — to develop its components and system. Moreover, he says, it would be difficult for another vendor to justify spending that kind of cash just to be the second company with the technology.
Griliches believes that component vendors are still feeling the sting from the telecom crash; for the past 4 or 5 years, they have only manufactured products with a solid revenue source. Plus, she says, they would sell fewer components if they did develop PICs. “What incentive does a JDSU have to start on something like this?” she wonders.
That doesn’t mean that Infinera has cornered the market on photonic integration, however. Griliches says it’s possible that other optical component companies may begin looking at integration, though probably not on the scale of Infinera’s PICs. Moreover, she wonders if perhaps Infinera itself could be purchased by a larger company, say a Lucent-Alcatel or a Ciena. With more available cash, the company could branch out from its primarily long-haul focus into the metro and access space, bringing its economies of scale to those markets.
Despite the obvious benefits of photonic integration, there are challenges associated with integrating multiple components on a single substrate, contends Steven Robinson, vice president of sales with Meriton Networks (www.meriton.com), which plays primarily in the metro regional space. “You could have thermal issues, power dissipation and crosstalk, where you modulate one laser and the thermal issues affect the adjacent laser. Or you could have electrical coupling issues in which high-speed modulation on a receiver, for example, affects the adjacent receiver.”
Moreover, Robinson worries that the PIC represents a single point of failure. “If the chip goes down, you don’t lose just 1 or 2.5 gig of traffic, but you lose 50 or 100 gig of traffic,” he says. “There are concerns that you have to have the reliability to be rock solid before you can put everything through that.”
The folks at Infinera counter this claim with the argument that the fiber coupler or pigtail, used to connect lasers, modulators, and other components, is the weak link from a reliability point of view, typically accounting for more than half of system failures. The conventional 80-channel system features more than 560 fiber couplings, says Dodd. Infinera’s PICs, by contrast, eliminate fiber couplers by connecting components via InP-based waveguides, thereby preventing 97% of system failures, says Dodd. He reports that the company’s PICs now have surpassed 3.5 million hours in the field without a single failure, which he believes is a testament to the devices’ robustness.
The proverbial proof is in the pudding. Infinera has inked deals with seven publicly announced customers spanning three continents, including Level3, Global Crossing, FLAG Telecom, XO Communications, OnFiber, CityNet, and German ISP freenet. In 2005, the company’s first year of shipping the DTN system, Infinera ranked number one in 10-Gbit/sec long-haul DWDM ports shipped, according to the Dell’Oro Group. Of course, success among the Tier-2s and -3s is one thing; the vendor has yet to announce an incumbent customer. Part of the reluctance on the part of the RBOCs may be the lack of a second source for Infinera’s components, modules, and systems, but Griliches believes it has more to do with the fact that the long-haul portions of their networks are already built out. “I think there are definitely options for completing sales in the metro space for Infinera with the Tier-1s,” she notes.
As for photonic integration technology itself, Griliches believes that, with a little more investment, photonic integration could be the most disruptive technology to hit the optical market in a decade. PICs could enable the development of very inexpensive reconfigurable optical add/drop multiplexers (ROADMs), mini-ROADMs, and PON equipment. “They could literally revolutionize the optical networking market and turn over all the [existing] equipment to much more cost-effective equipment,” she notes. “Everyone would move to optics if that were the case.”
For now, Infinera’s success has not gone unnoticed. When asked if the company’s competitors were concerned about its recent customer traction, Griliches reports that they are — and for good reason. “When it really comes down to it, what’s the difference between Nortel, Lucent, Alcatel, Ciena, and Siemens? Okay, so maybe someone is using CoreOptics transponders and someone else is using JDSU transponders, but for the most part, they are using off-the-shelf, available transponders. There is no secret sauce in any of this stuff, except for how you tweak the individual components and put it all together,” she says.
The optical market has collapsed to the point where a handful of component vendors are supplying a handful of system vendors, she adds, leaving little room for differentiation. “I think [Infinera’s competitors] see that in three to five or six years, that could be a real problem.”
Meghan Fuller is senior editor at Lightwave.