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

在不久的将来，对电信网络容量的高需求可能导致混合电光信号处理向全光处理的转换，从而利用光域中可用的最大带宽。满足这一需求的一种方法是在时域和波长域中进行复用。使用皮秒持续时间的光脉冲，其可以是网络的类似孤子的过部分，可以首先在时域中执行复用（即，时分复用或TDM）用于本地到城域网应用，然后在波长域（波分复用或WDM）中用于广域覆盖。这种情况使人们得出这样的结论：要解决的关键问题是如何利用纯光学领域强大的数字处理技术，在非常基础的层面上最大限度地减少噪声的有害影响。其思想是，对于数字编码的数据[1（0）由脉冲的存在（不存在）表示]，代替使用以模拟方式作用于信号并且不可避免地引入3dB噪声的线性放大器，可以使用数字开关放大器或光再生器。与此同时，纯光数字交换可能比电光交换更可靠和更快。此外，还需要光交换来实现其他联网功能，例如解复用，以便以非常高的速度处理用于寻址网络上不同用户的数据分组的报头。 我们的初步实验表明，光纤的参数非线性可以被利用来执行分组交换全光网络中所需的功能，如光纤高速缓存缓冲器、皮秒脉冲全光再生器、全光限幅器和可调谐时钟再加工模块。这些器件利用了玻璃光纤的超快参量非线性，因此能够以超过100 Gb/s的速度工作。此外，它们对于部署分组交换、超高速时分和波分复用全光网络至关重要。 到目前为止，在我们所有的实验中，标准色散位移光纤（DSF）已被使用。为了实现数据处理功能，需要使用几百米量级的光纤长度，脉冲宽度为ps，峰值功率为几瓦。在这里，我们建议探索使用高非线性光纤，如微结构光纤（MF，这是现在才开始商业化），在高速全光处理中执行essentialfunctions。由于它们的强引导行为，MF可以缠绕成非常紧密的环，这表明它们可能适合于紧凑的模块化开关封装。具体而言，我们建议利用高非线性微结构光纤开发全光数据处理模块。这些包括基于参数放大的高速缓存存储缓冲器，其将能够在数十Gb/s范围内操作。通过使用高非线性光纤，可以用市售的瓦特级光放大器来满足平均泵浦功率要求。我们将进行实验来探索阅读、写和擦除存储数据模式的各种方法，我们的工作表明，超快参量非线性可以被利用来提供宽带可调增益或动态增益调制用于时钟恢复。我们建议将两者联合收割机结合起来，以证明光学锁相环，这在原则上可以非常快，因为它们依赖于克尔非线性的相位鉴别。在进行上述实验研究的同时，我们还将建立各种光学系统的数值模型.这将提供一个设计工具来确定参数值，从而使实验装置能够最有效地运行。我们以前已经证明了亚皮秒脉冲在光纤线路中稳定传播的可能性，其中共轭增益被用来补偿线性损耗。我们建议组装一个再循环回路实验，其中线性损耗将由一对非简并参量（共轭）放大器补偿。这两个放大器的位置将根据进一步的理论/数值结果来选择。我们将通过实验和理论研究亚皮秒脉冲的稳定性，并将实验结果与数值模拟结果进行直接比较。

The high demand likely to be placed on the capacity of telecommunication networks in the near future urgesconversion 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 aswell as wavelength domains. Using picosecond-duration optical pulses, which could be soliton-like overportions 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 exploitedto perform functionalities that will be needed in packet-switched all-optical networks, such as fiber-opticcache-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 andhence are capable of operating at speeds in excess of 100 Gb/s. Moreover, they will be essential for deployingpacket-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 lengthson the order of 100's of meters are required for used with ps-duration pulses of a few watts peak power toachieve the data processing functions. Here we propose to explore the use of high-nonlinearity fiber, suchas microstructure fiber (MF, which is only now becoming commercially available), to perform essentialfunctions in high-speed all-optical processing. Because of their strongly guiding behavior, the MFs canbe wound into very tight loops, suggesting that they could potentially fit into a compact modular switchingpackage. 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 broadbandtunable gain or dynamic gain modulation for clock-recovery. We propose to combine the two to demonstrateoptical phase-lock loops, which in principle can be extremely fast as they rely on the Kerr nonlinearityfor envelope-phase discrimination. Simultaneous to the above experimental studies we will also developnumerical models of the various optical systems. This will provide a design tool to determine the parametervalues allowing the most efficient operation of the experimental setups. We have previously demonstratedthe possibility of stably propagating sub-picosecond pulses in fiber lines in which conjugating gain is usedto compensate the linear loss. We propose to assemble a re-circulating loop experiment in which linear losswill be compensated by a pair of non-degenerate parametric (conjugating) amplifiers. The location of thetwo amplifiers will be chosen based upon further theoretical/numerical results. We will experimentally andtheoretically study the stability properties of the sub-picosecond pulses by making various signal and noisemeasurements, and will compare the experimental results directly with numerical simulations.