Research And Design Of Electronic Equalization Based On Most Likelihood Sequence Estimation

Fiber communication system has become a most important part of information transmission networks for its massive channel capacity and high transmission rate. As the transmission rate increasing, the dispersion problem, which is not very serious when the transmission rate is low, is no longer neglectable. Due to the widely adoption of 10G/s fiber and its systems, total substitution is not economical. To increase facility utility, much attention has been drawn on equalization technology. Comparing to optical signal equalization, electronic equalization methods based on electronic signal processing is much more economical and feasible, with higher automation accessibility. Among the many electronic equalization methods, the digital signal processing algorithm—Most Likelihood Sequence Estimation, is considered as the best solution theoretically.Most Likelihood Sequence Estimation(MLSE)consists of two modules:the channel estimation module and the Viterbi module. The channel estimation module aims to establish mathematic model for the transmission channel and stimulate the distortion process of the signal; the Viterbi module, on the other hand, is to do the MLSE of the distorted signal using Viterbi algorithm, based on the channel model. In the thesis, two kinds of channel model have been researched—the most likelihood probability (ML) model and the finite impulse response (FIR) filter model. For the first model, a nonparametric method is adopted in this thesis; for the FIR filter model, a simplified Fast Fourier Transformation (FFT) channel estimation method is proposed. A transition based Viterbi algorithm is also investigated and implemented in the research, combined with the two channel model mentioned above, to form two sorts of MLSE equalizer. Using different stimulation software tools, a co_stimulation system is built for analysis and comparison of the two equalizers.Co_stimulation result shows that, the proposed simplified FFT channel estimation method performs better than the ML model. Moreover, the operation time of simplified FFT method is an order of magnitude faster than the ML model.