DNN-assisted phase distance tuned PSK modulation for PAM4-to-QPSK format conversion gateway node

Area of Science:

• Optical communications
• Signal processing
• Machine learning

Background:

• Heterogeneous networks require flexible signal format conversion.
• Traditional optical-electrical-optical conversion is inefficient.
• Direct intensity-to-phase mapping faces challenges like phase noise distortion.

Purpose of the Study:

• To propose an optical gateway for direct PAM to PSK conversion.
• To enable flexible intensity-to-phase mapping in optical networks.
• To mitigate phase noise distortion and maximize achievable information rates.

Main Methods:

• Development of an optical gateway for PAM to PSK conversion.
• Implementation of a proof-of-principle experiment.
• Utilizing deep learning-based decision algorithms at the receiver for signal recovery.
• Optimizing phase distance for PSK signals.

Main Results:

• Demonstrated successful optical conversion from PAM to PSK signals.
• Identified non-uniform irregular phase noise distortion in optically converted PSK signals.
• Achieved a performance improvement of approximately 4 dB gain in generalized mutual information.
• Validated the effectiveness of deep learning-based recovery for phase distance-tuned PSK signals.

Conclusions:

• The proposed optical gateway effectively converts PAM to PSK signals without O-E-O conversion.
• Optimized phase distance and deep learning recovery significantly enhance PSK signal performance.
• This approach offers a promising solution for efficient data transmission in future optical networks.