Coherent optical OFDM modem employing artificial neural networks for dispersion and nonlinearity compensation in a long-haul transmission system

• Main Author: Jarajreh, Mutsam Abdel-karim
• Other Authors: Ghassemlooy, Zabih
• Published: Northumbria University 2012
• Subjects: 621.382; H600 Electronic and Electrical Engineering
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.558071

In order to satisfy the ever increasing demand for the bandwidth requirement in broadband services the optical orthogonal frequency division multiplexing (OOFDM) scheme is being considered as a promising technique for future high-capacity optical networks. The aim of this thesis is to investigate, theoretically, the feasibility of implementing the coherent optical OFDM (CO-OOFDM) technique in long haul transmission networks. For CO-OOFDM and Fast-OFDM systems a set of modulation formats dependent analogue to digital converter (ADC) clipping ratio and the quantization bit have been identified, moreover, CO-OOFDM is more resilient to the chromatic dispersion (CD) when compared to the bandwidth efficient Fast-OFDM scheme. For CO-OOFDM systems numerical simulations are undertaken to investigate the effect of the number of sub-carriers, the cyclic prefix (CP), and ADC associated parameters such as the sampling speed, the clipping ratio, and the quantisation bit on the system performance over single mode fibre (SMF) links for data rates up to 80 Gb/s. The use of a large number of sub-carriers is more effective in combating the fibre CD compared to employing a long CP. Moreover, in the presence of fibre non-linearities identifying the optimum number of sub-carriers is a crucial factor in determining the modem performance. For a range of signal data rates up to 40 Gb/s, a set of data rate and transmission distance-dependent optimum ADC parameters are identified in this work. These parameters give rise to a negligible clipping and quantisation noise, moreover, ADC sampling speed can increase the dispersion tolerance while transmitting over SMF links. In addition, simulation results show that the use of adaptive modulation schemes improves the spectrum usage efficiency, thus resulting in higher tolerance to the CD when compared to the case where identical modulation formats are adopted across all sub-carriers. For a given transmission distance utilizing an artificial neural networks (ANN) equalizer improves the system bit error rate (BER) performance by a factor of 50% and 70%, respectively when considering SMF firstly CD and secondly nonlinear effects with CD. Moreover, for a fixed BER of 10-3 utilizing ANN increases the transmission distance by 1.87 times and 2 times, respectively while considering SMF CD and nonlinear effects. The proposed ANN equalizer performs more efficiently in combating SMF non-linearities than the previously published Kerr nonlinearity electrical compensation technique by a factor of 7.