Tunable optical matrix convolution of 20-Gbit/s QPSK 2-D data with a kernel using optical wave mixing.

Area of Science:

• Photonics
• Optical Signal Processing
• Nonlinear Optics

Background:

• Optical matrix convolution offers advantages over electronic methods, particularly for high-rate, phase-encoded data, by eliminating costly optical-to-electronic-to-optical (OEO) conversions.
• Phase-encoded data formats like quadrature phase-shift keying (QPSK) are crucial for high-capacity optical communication systems.

Purpose of the Study:

• To experimentally demonstrate a reconfigurable optical matrix convolution system for processing QPSK-encoded data.
• To showcase the system's ability to handle variable kernel sizes and high data rates.

Main Methods:

• Serialization of two-dimensional (2-D) input data and generation of time-shifted data replicas (TSDRs).
• Convolution of 2-D data with a 1-D kernel using nonlinear wave mixing in a periodically poled lithium niobate (PPLN) waveguide.
• Adjustment of kernel coefficients by tuning the relative phase and amplitude of kernel pumps.

Main Results:

• Successful demonstration of reconfigurable matrix convolution with error-free outputs for both 2x1 and 3x1 kernels.
• Achieved 16-quadrature amplitude modulation (QAM) output with an error vector magnitude (EVM) of ~5.1-8.5% for a 2x1 kernel.
• Achieved 64-QAM output with an EVM of ~4.9-5.5% for a 3x1 kernel, processing data at rates from 6-20 Gbit/s.

Conclusions:

• The proposed optical matrix convolution technique is a tunable and reconfigurable approach for high-speed signal processing.
• This method effectively avoids OEO conversions and achieves high-order modulation formats with low EVM, paving the way for advanced optical communication systems.