IBM and Corning are pursuing a vertically integrated program to develop a scalable all-optical network infrastructure encompassing local-, metropolitan-, and wide-area networks by exploiting the bandwidth and transparency of optical fiber resources. IBM is also working on related, more advanced work on multi-wavelength devices under the same program.
The program is partially sponsored by ARPA under contract MDA-972-95-C-0001.
The ICON program has the following goals: (1) to develop a scalable architecture based on wavelength division multiplexing (WDM) and wavelength routing, by which available wavelengths are reused many times in the network, with explicit support for asynchronous transfer mode (ATM), and (2) to develop the component technology required for the network, including transmitters, receivers, wavelength routers, and optical amplifiers capable of amplifying a large number of wavelengths at reasonable cost.
Successful achievement of these program goals should lead naturally to the rapid commercialization of all-optical networking technologies. The parties to this proposal bring the market access and commercial infrastructure required to commercialize the components, software, hardware, systems and services to be developed under this program.
At the lowest level in the protocol hierarchy, the architecture provides a large number of transparent ``pipes'' or {\em lightpaths} that may be reconfigured if required. These lightpaths are then used to interconnect electronic ATM packet switches and all-optical broadcast star local networks, and are made available to other applications as well. We are developing and using a complete architecture consisting of algorithms for network design, routing, wavelength assignment, and reconfiguring the lightpath topology, and a framework for controlling the network.
The transmitters, receivers and routers required for the network are being realized using a planar optical grating chip, with suitable additions such as arrayed laser driver chips, arrayed photodetector chips, or planar optical switches, all of which can be manufactured at low cost. We are calling this unified approach our optoelectronic chipset.
Our system experience has indicated that it is most desirable that the receive versions of the chipset modules have sensitivities at least as good as commercial APD receivers, and that the transmit modules have several dB better output power into the fiber than do commercial DFB laser diodes. Therefore, we are investigating low-cost wavelength-flat erbium doped fiber amplifiers (EDFAs) to be packaged within the same overall module as the chipset recevier or transmitter technology.
Our three-year goal is to realize an architecture and the components capable of supporting 32 wavelengths, each at 2.4 Gb/s, in the wide-area, and 128 wavelengths, each at 2.4 Gb/s, in the local area.
We expect to commercialize components (arrayed WDM and tunable transmitters, arrayed WDM and tunable receivers, routers and amplifiers) as well as software (protocol implementations) and other hardware (ATM workstation adaptors, ATM interfaces etc.), while simultaneously taking the results of our architectural work to standards bodies.
While the main thrust of this proposal is an integrated program to build a WDM network using modifications of available devices, the proposal also contains an effort to develop a fundamentally new set of devices based on a simple monolithic semiconductor structure that could serve as WDM transmitters, receivers, or amplifiers. Such a device, if perfected, could truly revolutionize WDM. The low cost and simple structure would allow its use even in short length computer links.
The device, whether configured as a laser, amplifier, or detector, uses a tapered waveguide structure as the frequency selective element. It requires only rudimentary growth and processing and does not contain diffraction gratings. Its one dimensional structure uses less real-estate and should lower cost by orders of magnitude. Configured as an amplifier, the input wavelengths are transferred from an input waveguide through an active cavity to the output waveguide, with each wavelength traversing the cavity at a different position. Control currents applied to different active regions control the gain of the corresponding wavelength. Feedback causes the amplifier to lase, while reverse biasing the p-n junction of the active cavity allows its use as a demultiplexer, with the photocurrent of each wavelength appearing at the corresponding contact pad. Tone low cost monolithic structure would replace the source, amplifier, and detector in a WDM system, substantially simplifying and lowering the cost of such a network.
Though the work described in the last two paragraphs is technically more risky than the further development of available devices, the payoff can be enormous. Our balanced approach of simultaneously pursuing both paths toward low-cost optoelectronic chipsets assures success with the highest payoff possible.