To accommodate the exponential growth of network services in the five-generation (5G) and beyond wireless system, 50 Gb/s/λ passive optical network (PON) is developed for mobile xhaul applications. Intensity modulation and direct detection (IMDD) technology together with digital signal procession (DSP) is being considered as the promising solution for 50 Gb/s/λ PON due to its low cost, low power consumption, and compact footprint. Different DSP algorithms with varied structures are proposed for linear and nonlinear impairments compensation in the high-speed PON, while the performance and complexity analysis of these algorithms is still missing. To find the most efficient equalizers, in this paper, four conventional equalizers, including feed-forward equalizer, decision feedback equalizer (DFE), Volterra equalizer (Vol) and Volterra DFE equalizer (Vol-DFE), together with two deep learning equalizers namely fully-connected neural network, and long short-term memory equalizer are experimentally compared in a 10G optics based 50G-PON system in terms of the equalization performance, computation complexity, optimization difficulty, and generalization ability. After the evaluation of our proposed fair comparison algorithm, we consider Vol-DFE is the most efficient one considering both performance and complexity. Attributes to the strong and efficient equalization capability of Vol-DFE, C-band 50 Gb/s PAM-4 signal transmission can be supported in a 10G optics based IMDD system with a 3 dB bandwidth of 6.11 GHz, and a power budget up to 38 dB can be achieved.

Increasing the serial bitrate to 25-Gb/s per channel as well as coexisting with the same optical distribution network of the 10G-ethernet passive optical network (EPON) is challenging. The bandwidth limitation and the reduced dispersion tolerance at high date rate make it much more difficult to achieve higher loss budget. In this paper, we have first shown a demonstration of a symmetric 4 × 25-Gb/s time-wavelength-division-multiplexed passive optical network system based on nonreturn-to-zero on-off-keying format supporting 0-20 km reach in O-band. 10G-class directly modulated lasers (DMLs) are employed to serve as both upstream and downstream transmitters owing to the low cost and the frequency chirp induced by DML can be used for bandwidth equalization. Instead of using digital signal processing, the chirp management and frequency equalization are simultaneously realized in optical domain through our proposed dispersion supported equalization technique. By employing a spool of dispersion-shifted fiber in optical line terminal (OLT), a single device can be used to simultaneously equalize the frequency response of multiple downstream and upstream channels. Then, we further use a semiconductor optical amplifier as a booster and preamplifier at the OLT for downstream and upstream, respectively. A power budget of 26 and 32 dB with 0-20 km reach of standard single mode fiber for downstream and upstream transmission at 1310 nm is obtained. The experimental results shown in this paper reveal that our proposed cost-effective scheme would be a promising candidate for the next generation 100G-EPON.

Volterra nonlinear equalizer (VNLE) is widely investigated for linear and nonlinear distortions compensation in optical communication systems. Despite the powerful equalization ability, the required high computation complexity limits its real-time implementation on the hardware. To simplify the complexity, the structure of the VNLE needs to be well optimized. Traditionally, manual optimization of the VNLE structure is blind, and traversal search is inefficient and may result in multiple structures. Therefore, the optimal VNLE structure with the best equalization performance and the lowest equalization complexity for the same performance are still unclear. In this paper, a novel search method named AutoVolterra is proposed to efficiently find the optimal structure of VNLE by using deep reinforcement learning to learn the functional relationship between structure parameters and optimization objectives. Experimental demonstration of 50-Gb/s PAM-4 signal transmission in a bandwidth-limited intensity modulation and direct detection system is performed to verify the effectiveness of AutoVolterra. Both AutoVolterra and greedy search are applied to Volterra feedforward equalizer, Volterra decision feedback equalizer and Volterra-Pruning. The bit error ratio and equalizer's complexity results suggest that the structure searched by AutoVolterra can optimize the equalization performance, and also significantly reduce the complexity, and improve the pruning quality of VNLE to assist real-time implementation on the hardware. Besides, AutoVolterra can also provide a reliable equalization performance benchmark and complexity benchmark to help us verify the actual capabilities of other algorithms.

Directly-modulated laser (DML) is widely employed in intensity modulation and direct detection (IMDD) system due to its low cost and high output power. However, the corresponding frequency chirp is regarded as one of the main disadvantages for its application in passive optical networks (PONs). In this paper, we theoretically analyze the frequency response evolution of DML based system under different chirp and dispersion conditions, proving that the system bandwidth can be improved by interactions between negative dispersion and DML chirp. Based on this concept, we experimentally demonstrated downstream 50 Gb/s PAM4 signal transmission over 20 km single-mode fiber (SMF) access based on the 10 Gb/s DML operating at 1310 nm and avalanche photodiode (APD). A dispersion-shifted fiber (DSF) providing -150 ps/nm dispersion at 1310 nm in the optical line terminal (OLT) is used to pre-equalize the frequency response of bandwidth-limited directly modulated signals in the optical domain. Thanks to our proposed dispersion-supported equalization (DSE) technique, the system bandwidth can be improved by 5 GHz. Feed-forward equalization (FFE), decision feedback equalization (DFE) and Volterra filter are employed to evaluate the signal performance improvement, respectively. By evaluating the receiver sensitivity, the DSE combined with FFE scheme shows 2 dB improvement than the complex Volterra algorithm, indicating its potential to reduce the complexity of digital signal processing (DSP) and therefore a lower cost and power consumption in PON.

For the first time, an unsupervised learning training method was proposed for neural network (NN)-based blind equalization in intensity modulation and direct detection (IMDD) system. The whole scheme can directly train an NN-based equalizer with only the received signal rather than original symbols as reference. Besides, the unsupervised learning can also help a well-trained NN to keep its performance in face of varying system status, such as wavelength shift and bias fluctuation, in practical applications. We evaluate the performance of the proposed scheme in a 50 Gb/s IMDD system. The experimental results confirm that the proposed unsupervised learning can train fully-connected NN-based equalizer as well as convolutional neural network (CNN)-based equalizer to reach the same performance as the one trained by supervised learning. Besides, in face of signals with different bias current of the directly-modulated laser (DML), the unsupervised learning method can train the NN to keep the best performance. It can be proved that the scheme can help maintain the performance of NN-based equalizer against the continuous system parameter variations.