Ultra-High-Capacity Optical Communications and Networking: Data processing modules using high-nonlinearity fiber for advanced optical networking

Northwestern University, Evanston IL

Investigators

Abstract

The high demand likely to be placed on the capacity of telecommunication networks in the near future urges conversion of hybrid electro-optical signal processing to all-optical processing, exploiting the largest bandwidth available in the optical domain. One way of catering to this demand is by multiplexing in time as well as wavelength domains. Using picosecond-duration optical pulses, which could be soliton-like over portions of the network, one can first perform multiplexing in the time domain (i.e., time-division multiplexing, or TDM) for local to metropolitan-area network applications and then in the wavelength domain (wavelength-division multiplexing, or WDM) for wide area coverage. This scenario leads one to conclude that the key issue to be addressed is how to take advantage of the powerful digital-processing techniques in the pure-optical domain, that minimize the detrimental effects of noise at a very fundamental level. The idea is that, for digitally encoded data [1's (0's) represented by the presence (absence) of pulses], instead of using linear amplifiers which act on signals in an analog fashion and inevitably introduce 3 dB of noise one can employ digital-switching amplifiers or optical regenerators. At the same time, pure-optical digital switching is potentially much more reliable and faster than electro-optical switching. Furthermore, optical switching will also be needed to implement other networking functions, such as demultiplexing to process at very high speed the header of a data packet used for addressing to different users on the network. Our preliminary experiments show that the parametric nonlinearity of optical fibers can be exploited to perform functionalities that will be needed in packet-switched all-optical networks, such as fiber-optic cache-memory buffers, picosecond-pulse all-optical regenerators, all-optical limiters, and tunable clock re-covery modules. These devices take advantage of the ultrafast parametric nonlinearity of glass fiber and hence are capable of operating at speeds in excess of 100 Gb/s. Moreover, they will be essential for deploying packet-switched, ultrahigh-speed time-division and wavelength-division multiplexed all-optical networks. In all of our experiments thus far, standard dispersion-shifted fiber (DSF) has been used. Fiber lengths on the order of 100's of meters are required for used with ps-duration pulses of a few watts peak power to achieve the data processing functions. Here we propose to explore the use of high-nonlinearity fiber, such as microstructure fiber (MF, which is only now becoming commercially available), to perform essential functions in high-speed all-optical processing. Because of their strongly guiding behavior, the MFs can be wound into very tight loops, suggesting that they could potentially fit into a compact modular switching package. Specifically, we propose to utilize the high-nonlinearity microstructure fibers to develop all-optical data processing modules. These include a cache storage buffer based upon parametric amplification that will be capable of operating in the 10's of Gb/s range. With use of the high-nonlinearity fiber, the average pump power requirement can be met with commercially-available watt-class optical amplifiers. We will carry out experiments to explore various ways of reading, writing, and erasing the stored data patterns. Our work has shown that the ultrafast parametric nonlinearity can be exploited either to provide broadband tunable gain or dynamic gain modulation for clock-recovery. We propose to combine the two to demonstrate optical phase-lock loops, which in principle can be extremely fast as they rely on the Kerr nonlinearity for envelope-phase discrimination. Simultaneous to the above experimental studies we will also develop numerical models of the various optical systems. This will provide a design tool to determine the parameter values allowing the most efficient operation of the experimental setups. We have previously demonstrated the possibility of stably propagating sub-picosecond pulses in fiber lines in which conjugating gain is used to compensate the linear loss. We propose to assemble a re-circulating loop experiment in which linear loss will be compensated by a pair of non-degenerate parametric (conjugating) amplifiers. The location of the two amplifiers will be chosen based upon further theoretical/numerical results. We will experimentally and theoretically study the stability properties of the sub-picosecond pulses by making various signal and noise measurements, and will compare the experimental results directly with numerical simulations.