Echo State Network Based Nonlinear Equalization for 4.6 km 135 GHz D-Band Wireless Transmission

Reservoir computing (RC) is a novel computational framework derived from recurrent neural networks (RNN). It can reduce the training complexity of RNN and is suitable for time-series learning tasks. The echo state network (ESN) is one of the most popular forms of RC. In this paper, an ESN-based equalizer is applied to perform signal equalization in a wireless D-band communication system to compensate for the nonlinear distortion. Based on the photonics-based technology and multiple amplifiers, a long-range wireless transmission system is successfully established at D-band. We experimentally demonstrated that our proposed wireless transmission link can realize up to 4.6-km wireless delivery of over 8 Gbit/s quadrature phase shift keying (QPSK) millimeter-wave signal at 135 GHz with a bit-error-rate (BER) less than the hard decision forward error correction (HD-FEC) threshold of 3.8 × 10
⁻³
. Compared with the traditional CMA equalizer and the Volterra equalizer, the experimental results show that the ESN-based equalizer can achieve a better balance between the BER performance and the computational complexity.