Low-Resolution Joint Encoding-Modulation Optical Fiber Communication System

Low-resolution digital-to-analog converters (DACs) and analog-to-digital converters (ADCs) have low costs and low power consumption. Nevertheless, the introduced quantization noise results in significant performance degradation. To guarantee the performance of low-resolution optical transmission systems, we propose and experimentally demonstrate a low-resolution optical fiber communication (OFC) system based on joint encoding modulation (JEM). The proposed system utilizes a deep learning (DL)-enabled JEM network to extract image features for both image encoding and modulation. It employs a joint demodulation-decoding (JDD) network to reconstruct the transmitted image. We integrate the additive white Gaussian noise (AWGN) model into the JEM and JDD networks, training them jointly to enhance their adaptability to low-resolution transmission channels. Our simulation results demonstrate that JEM does not require switching between different modulation formats like traditional approaches to adapt to varying levels of noise. The experimental results demonstrate that the JEM-OFC is more robust to the impairments of low-resolution transmission than the PAM4-based traditional OFC, especially when the quantization bits are between 2 and 3. When using 4-bit DAC and ADC, our proposed approach significantly improves the compression capability of the system compared to the traditional OFC structure while maintaining the same transmission performance. Additionally, the JEM-OFC system exhibits more stable performance, especially in low-received optical power regime. The results show that it can improve 2 dB receiver sensitivity when the quantization bit is 4.