Experimental demonstration of wavelength shift keying in optical WDM networks
Young, James F.
Doctor of Philosophy
Fiber optic networking has developed incredibly over the past a few years. It offers the promise to support large numbers of users simultaneously, to provide users with integrated services at high speed, and to increase the length of transmission without regeneration of the signals. To achieve yet even bigger capacity and longer transmission spans, the physical properties of the link, such as nonlinear processes in the optical fiber have been extensively studied. This thesis reports a study of one of the principal fiber nonlinear processes---Four-Wave Mixing (FWM), which results in interchannel crosstalk in wavelength division multiplexing (WDM) networks. We introduce a coding and detection scheme---Wavelength Shift Keying (WSK)---to reduce this nonlinear noise. Wavelength shift keying using symmetrical wavelength assignment and balanced detection can easily be implemented on top of the standard WDM setup, with only moderate changes at the end equipment. Its capability to depress four-wave mixing crosstalk to first order will greatly improve performance in long-haul transport systems. An analytical study of the four-wave mixing noise characteristics is conducted in WDM networks using low-dispersion fibers. The results suggest a wavelength shift keying scheme that utilizes these properties to suppress four-wave mixing noise to first order. Simulations on system performance for various network scenarios are carried out in both standard WDM and WSK, and the results compared, using both Gaussian approximation and exhaustive calculations. A proof-of-principle experiment is then implemented on a testbed that emulates these systems to show the operability of the proposed scheme. Optical and electrical properties of the testbed such as power spectrum, transmission efficiency, and noise figure are also studied so that irrelevant factors are controlled. Bit-error rate (BER) measurements are conducted in the first order approximation, and the results compare well with the calculations. In the laboratory system, an erbium-doped superfluorescent fiber source is employed to generate the four-wave mixing noise spectrum, since normal laboratory-scale systems do not produce realistic FWM noise. The theory of importance sampling is considered for the second order behavior, and group measurements are conducted within the range of system limitations. Data analysis shows the validity of the first order comparison done previously, and provides a new approach to variance performance evaluation.
Electronics; Electrical engineering