Information Rates of Approximate Message Passing for Bandlimited Direct-Detection Channels

2025-08-02

AI Summary

Achieving capacity逼近 in bandwidth-limited direct-detection fiber-optic channels remains challenging due to receiver nonlinearity. Method: This paper proposes a novel receiver architecture integrating generalized vector approximate message passing (GVAMP), multilevel coding, and successive interference cancellation. It is the first work to successfully apply GVAMP to noncoherent optical channels. Contribution/Results: The scheme achieves 5 bpcu spectral efficiency with 80 Mop/b information-bit complexity after 38 iterations—within 0.3 bpcu of the real-valued coherent capacity for amplified links, or equivalently enabling a 3 dB power gain for unamplified links—reducing the theoretical capacity gap by 0.7 bpcu. Bipolar modulation improves power efficiency by 6 dB over unipolar modulation and exhibits strong robustness to parameter variations, significantly outperforming state-of-the-art neural-network-based receivers.

Technology Category

• Search and Optimization: Non-convex Optimization
• Computer Vision: Learning & Optimization for CV
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

The capacity of bandlimited direct-detection channels is difficult to compute or approach because of the receiver nonlinearity. A generalized vector approximate message passing (GVAMP) detector is designed to achieve high rates with reasonable complexity. The rates increase by using multi-level coding and successive interference cancellation. The methods are applied to optical fiber channels with long intersymbol interference, as encountered in practice. Bipolar modulation operates within 0.3 bits per channel use (bpcu) of the real-alphabet coherent capacity for optically-amplified links, improving the best existing gap of 1 bpcu based on theory. Remarkably, bipolar modulation gains 6 decibels (dB) in power efficiency over unipolar modulation, and 3 dB for unamplified links. The detector is robust to changes in channel parameters such as the fiber length. The GVAMP complexity, measured in multiplications per information bit (mpib), is proportional to the number of iterations and the logarithm of the block length, and is substantially less than state-of-the-art neural networks. The receiver requires approximately 38 iterations to achieve a rate of 5 bpcu with 80 mpib.

Problem

• Computing capacity for nonlinear bandlimited direct-detection channels
• Achieving high rates with low-complexity GVAMP detector
• Improving power efficiency and robustness in optical fiber channels

Innovation

• GVAMP detector achieves high rates with reasonable complexity
• Multi-level coding and interference cancellation increase rates
• Bipolar modulation improves power efficiency and capacity gap