8.8 Gbps PAM-4 visible light communication link using an external modulator and a neural network equalizer

Area of Science:

• Optical Communications
• Signal Processing
• Machine Learning

Background:

• Visible light communication (VLC) systems offer high bandwidth potential.
• External modulators, while enabling high frequencies, can introduce low-frequency fluctuations and inter-symbol interference (ISI).
• Existing methods struggle to mitigate these distortions effectively in high-speed VLC.

Purpose of the Study:

• To experimentally demonstrate a high-speed VLC system using a LiNbO3 external modulator.
• To address and mitigate low-frequency fluctuations and ISI inherent in such systems.
• To achieve a record data rate for externally modulated VLC.

Main Methods:

• Utilized a LiNbO3 external modulator for pulse amplitude modulation (PAM)-4 signal transmission.
• Proposed and implemented a novel neural network, termed Low-Frequency Neural Network (LFNN), incorporating low-frequency signals as a second label.
• Employed 7% hard-decision forward error correction (HD-FEC) for performance evaluation.

Main Results:

• Achieved a data rate of 8.8 Gbps using PAM-4.
• Successfully mitigated low-frequency fluctuations and ISI using the LFNN.
• Met the bit-error-ratio (BER) limit of 3.8 × 10^-3 under HD-FEC.

Conclusions:

• The proposed LFNN effectively combats distortions in externally modulated VLC systems.
• This work establishes a new benchmark for data rates in visible light communication using external modulation.
• The demonstrated system shows significant promise for future high-speed wireless optical communication.