Accurate Calculation of Bit Error Ratios in Optical Fiber Communications Systems

University Of Maryland Baltimore County, Baltimore MD

Investigators

Abstract

0400535 Menyuk Since the invention of the erbium-doped fiber amplifier and the advent of long-haul optical fiber communications systems, it has not been possible in general to accurately calculate the bit error ratios (BERs). Nonlinearity in the optical fiber transmission line is the ultimate source of this difficulty. Nonlinear interactions between the signal and the noise can significantly affect the noise distribution prior to the receiver, invalidating simple analytical approaches. At the same time, the low BERs that one must calculate -on the order of 10 -15 in some cases - make standard Monte Carlo simulations impractical. Transmission nonlinearity also leads to pattern dependences and a spread in the received marks and spaces that is not due to noise and must be taken into account. In wavelength-division-multiplexed (WDM) systems, thousands of bits can interact, so that accurately accounting for the pattern dependences is difficult. Additionally, it is critically important to model the receiver accurately, including receiver- induced intersymbol interference, and one must be able to take into account the impact of forward error correction and signal processing. In the past few years, the PIs have developed a set of theoretical tools that are capable of addressing all these issues and have put the accurate calculation of BERs within reach. For each of these issues, they have developed at least two different methods, one deterministic and one statistical, that are in excellent agreement for the most part and are mutually self-validating. However, considerable further development is needed to test these methods in practical contexts, to validate reduced approaches that might be useful in certain contexts, and to validate results experimentally. The intellectual merit of the proposal resides in their taking sophisticated mathematical tools that have been applied primarily to idealized settings and developing them into effective computational methods that can be used to calculate BERs in real-world experimental systems with significant complexity. The broad impacts of the proposed research lie in its potential to substantially impact information technology and in its educational features. They expect that after further development, these tools will impact system design, lowering both development and systems costs and encouraging innovation. By careful comparison to experiments, they expect these tools to provide new insights into how these complex systems work. These methods that the PIs develop will be disseminated in courses taught at UMBC and short courses that they teach externally, as well as in publications. Interdisciplinary interactions between students in Mathematics and students in Engineering is a key feature of the proposal. Graduate research assistants will carry out most of the proposed work as part of their education, and outreach to under-represented groups is another key feature of the proposal.