Performances of Neural Networks and LDPC Decoders for OFDM High Speed Transmission in Optical Fiber

This paper presents a comparison of the performances of coded OFDM (Orthogonal Frequency Division Multiplexing) for optical broadband transmission using Random Neural Network (RNN), a variant of RNN, Soft Decision Decoding (SDD) and LDPC channel decoding techniques. The performance evaluation is based firstly on the Error Vector Magnitude (EVM) to assess the effects of imperfections in the optical channel, and secondly on the estimated Bit Error Rate (BER) based on OSNR. The simulations are performed at a rate of 10Gb/s over 1000 km using VPI software cosimulation environment. The results show that the error correcting codes, particularly LDPC codes, are well suited and efficient for broadband.These codes provide satisfactory solutions for OFDM, reducing the effects of chromatic dispersion (CD), polarization mode dispersion (PMD), Intersymbol Interference(ISI) and non-linearities.