A recent development in photonics—the advent of the large-scale photonic integrated circuit (PIC)—will fundamentally alter the assumptions driving telecom system design, and will redefine the way carriers build optical networks.
In carrier networks, fiberoptic transmission systems are responsible for converting digital information to pulses of light, then transmitting those pulses across a carrier’s fiberoptic cable plant. The designers of these systems, since their advent in the early 1980s, have used a host of separately packaged discrete optical devices, such as lasers and detectors, in designing complete systems. Functional integration was done at the system level, by interconnecting these devices with optical fiber across multiple circuit packs.
As their networks grew, carriers found that discrete optical components accounted for a disproportionately large fraction of network cost. The primary culprit was one specific function in the optical network: the optical-electrical-optical (OEO) conversion for the electronic regeneration of an optical signal; to clean up the digital information along a fiberoptic link; to groom or switch the digital signal; or to add or drop customer traffic. To reduce OEO cost, systems designers introduced more discrete components intended to reduce the need for OEOs, such as optical amplifiers, dispersion compensators, Raman amplification, and optical filters.
Optical-systems designers were certainly successful at reducing OEO cost as evidenced by the remarkable decline in cost of bandwidth in multiwavelength networks. Unfortunately, they also had to accept a major design tradeoff. As the diversity and complexity of devices in the systems increased, so too did the complexity and rigidity of the network. As a result, many carriers are now asking not just for cost savings, but also for the system simplicity and network reconfigurability they had when OEOs were more common.
Photonic integration provides a solution. Earlier this year, Infinera announced the first large-scale PICs, which integrate not only multiple devices (for instance, lasers and modulators), but also multiple “copies” of those devices, each using a different wavelength of light, onto small chips of indium phosphide. We have actually developed two PICs, a 100-Gbit/s transmitter and a 100-Gbit/s receiver, both supporting 10 wavelengths, all modulated at 10 Gbit/s. These chips implement a full OEO, and thus consolidate a wavelength-division-multiplexing (WDM) system on a chip.
So, the fundamental assumption underlying optical telecom system design—that OEO conversion was so expensive it must be minimized wherever possible—is no longer valid, and the system designer can rethink optical system architecture and carrier network implications. Since Infinera is vertically integrated we’ve had the opportunity to do exactly that. We’ve designed our own PIC-based system, worked with carriers to model the use of that system in networks, and found that PIC technology takes the economic downside out of deploying OEOs wherever required to manipulate the underlying digital data traversing the network. While arguably possible using traditional, discretely packaged optical-component technology, such architectures would not be economically viable without the use of PICs.
A promising future
By minimizing the economic downside of OEO conversion, PIC technology obviously offers carriers a broad range of opportunities for exploiting the upside. An increased number of OEO conversions means greater access to digital subwavelength bandwidth-management capabilities to switch, groom, multiplex, and add/drop customer traffic. At the same time, more network locations become capable of supporting customer access, enabling carriers to expand market presence and reduced backhaul of customer traffic.
Also, by eliminating the need for numerous “signal conditioning” components, OEO conversions simplify the engineering, operations, and evolution of the network by reducing the need to consider “all-optical” issues such as chromatic dispersion, wavelength banding, wavelength planning, and end-to-end optical reach.
In addition, OEOs maximize network reconfigurability by allowing carriers to add or change customer traffic in the digital domain, independently of the optical-line system. This stands in sharp contrast to the “all-optical” approach, in which any service changes can only be performed on wavelengths, not on actual customer circuits.
Given the technical feasibility and clear economic benefits of PIC technology, we believe that all fiberoptic networks will soon take advantage of photonic integration. We also believe that the innovation curve for PICs will be steep, just as it has been (as evidenced by Moore’s Law) for electronic integrated circuits, providing currently unimagined benefits down the road. The coming years should be exciting times for designers of optical chips, systems, and networks.
DAVE WELCH is chief technology and strategy officer and cofounder of Infinera, 1322 Bordeaux Drive, Sunnyvale, CA 94089; e-mail: [email protected].