Telecommunications Channel Equalization In Optical Burst Receiver

Dispersion is one of the limited factors in the quality of optical fiber communication system. The broadened optical pluses coming from dispersion interfere with each other, which will cause inter-symbol-interfere (ISI) and finally will cause b error. The industrial prospect of traditional optical dispersion compensation methods is limited due to high cost and complex technology. Electronic dispersion compensation (EDC), which uses electronic channel equalization techniques, such as decision feedback equalization (DFE) can restore the distorted signals with low cost, low power, and simple technology. The electronic channel equalization techniques have strong practical value in optical fiber communication.Recently, IP business is growing explosively. In the optical networks which can carry IP business, and in the future all-optical network, data are transferred as burst packets (or optical packets). Different from traditional continuous data, burst packets come from different transmitters. The receiver which uses DFE doesn't know where the received burst packets come from. So DFE can't adjust its tap coefficients with the unknown channel transmission characteristics. Through adding training sequence into burst packets, DFE can initialize its tap coefficients with the training sequence. In addition, the lengths of burst packets are much less than traditional continuous data. The proportions of the training sequences in burst packets can influence the transmission efficiency directly.The related properties about using DFE at the environment of optical burst are studied in this paper. Simulation model is founded to simulate the actual fiber communication system with MATLAB. The key factors, such as the length of training sequence, the iterative step of the minimum mean square error (LMS) algorithm used for initializing DFE equalizer's tap coefficients, and the transmission distance of optical fiber communication system, which effect equalization performance are analyzed and optimized. Through optimizing these factors, the length of training sequence is reduced to 2Kb in the premise of ensuring communication quality, which makes the transmission efficiency improve as far as possible.Two improved methods are proposed from two aspects: 1. for the selection of initial tap coefficients, a more effective method is found within a module simulated metropolitan area ring network (MARN). 2. Optimized iterative steps are different for the burst packets which come from different transmission distances. However, the receiver can't select and optimize the iterative step. In order to effectively solve this problem, an automatic method of adjusting the iterative step is presented.