Complex-Valued 2D-CNN Equalization for OFDM Signals in a Photonics-Aided MMW Communication System at the D-Band

Orthogonal frequency division multiplexing (OFDM) is a promising solution in photonics-aided millimeter-wave (MMW) communication systems, which combine the features of fiber and wireless links. In such a fiber-wireless system, the broadband OFDM signals could be affected by the multipath effect, chromatic dispersion, frequency synchronization error, and other complex impairments. Therefore, both inter-symbol interference (ISI) and inter-carrier interference (ICI) will cause performance degradation. In this experiment, a D-band photonics-aided MMW communication system is demonstrated. To compensate for the linear and nonlinear distortions during the fiber and wireless transmission, we proposed a novel time-frequency domain equalizer based on a two-dimensional convolutional neural network (2D-CNN) with a complex-valued structure. Moreover, a residual branch is adopted. We investigate the bit error rate (BER) performance of the OFDM signals transmission with different modulation formats. With the aid of complex-valued 2D-CNN based equalization, 40 Gbit/s 16QAM and 55 Gbit/s PS-64QAM signals can be successfully transmitted over 200 m wireless distance at the D-band while satisfying the HD-FEC (3.8 × 10
⁻³
) and SD-FEC (2.0 × 10
⁻²
) threshold, respectively. Moreover, the complex-valued structure can bring significant sensitivity improvements compared to the real-valued 2D-CNN equalization.