Abstract
In this paper, we address the problem of modulation format identification (MFI) for few mode fiber (FMF) transmission in elastic optical networks (EONs). The MFI accuracy is studied under different FMF channel conditions including mode coupling (MC), optical signal-to-noise ratio (OSNR), and chromatic dispersion (CD). Artificial neural network, trained using features extracted from the asynchronous in-phase quadrature histogram (IQH), is proposed to investigate the identification accuracy. Extensive simulation results have been conducted to identify six modulation schemes widely used in polarization division multiplexing coherent optical networks. This includes PDM-BPSK, PDM-QPSK, PDM-8QAM, PDM-16QAM, PDM-32QAM, and PDM-64QAM transmitted at 10 Gbaud network transmission speed. The results show that the proposed MFI achieves a successful average identification accuracy exceeding 98% in the presence of low MC when the incoming signal OSNR is greater than 20 dB. However, the effect of high MC and CD = 1100 ps/nm reduces the average accuracy to 90%. Further, the MFI accuracy is investigated under different symbol rates such as 14 and 20 Gbaud.