We experimentally demonstrate high-speed metro-scale optical transmission of a single sideband (SSB) 4-ary pulse amplitude modulation (PAM-4) signal based on a silicon photonic dual-drive Mach–Zehnder modulator (MZM). We propose a novel, to the best of our knowledge, artificial neural network (ANN) structure of soft combined ANN (SC-ANN) to compensate for both linear and nonlinear impairments of the signal. SC-ANN obtains the enhanced performance by averaging the outputs of conventional ANN with different sizes. With the help of the SC-ANN, we achieve a 320 km standard single-mode fiber (SSMF) transmission of 184 Gb/s (92Gbaud) PAM-4 with a bit-error rate (BER) below the 20% soft-decision forward error-correction (SD-FEC) threshold of ${2.4} \times {{10}^{ - 2}}$, and the optical signal-to-noise ratio (OSNR) penalty is only 0.3 dB compared to the back-to-back (BTB) results.

中文翻译：

---

带有硅MZM和基于软组合人工神经网络的均衡的高速SSB传输

我们实验证明了基于硅光子双驱动马赫曾德尔调制器（MZM）的单边带（SSB）四进制脉冲幅度调制（PAM-4）信号的高速城域规模光传输。据我们所知，我们提出了一种新颖的软组合人工神经网络（SC-ANN）的人工神经网络（ANN）结构，以补偿信号的线性和非线性损伤。SC-ANN通过平均具有不同大小的常规ANN的输出来获得增强的性能。借助SC-ANN，我们实现了184 Gb / s（92Gbaud）PAM-4的320 km标准单模光纤（SSMF）传输，且误码率（BER）低于20％的软判决前向纠错（SD-FEC）阈值$ {2.4} \ times {{10} ^ {-2}} $，与背对背（BTB）结果相比，光信噪比（OSNR）损失仅为0.3 dB。