256 GBaud C-band RRM-based OOK Transmission Using CNN-GRU and Bi-LSTM Equalization

Ring resonator-based modulators are essential in high-speed intra-data center communication links, providing efficient modulation with high bandwidth and low power consumption. In this work, we demonstrate the performance of two neural network (NN)-based equalizers in a ring modulator-based C-band 256 GBaud on-off keying amplification-free transmission, which currently represents state-of-the-art symbol rate transmissions. We evaluate two recurrent NN-based models: a convolutional neural network combined with a gated recurrent unit (CNN-GRU) and bidirectional long short-term memory. Owing to their ability to operate with both long and short-term memory, combined with advanced recurrent gating architectures, both models outperform the conventional decision feedback equalizer in both back-to-back and 100 m single-mode fiber transmission scenarios. Unlike a DFE with 99 taps, the proposed equalizers achieve a bit-error-ratio below the 6.25% overhead hard-decision forward error correction threshold.