Experimental realization of convolution processing in photonic synthetic frequency dimensions

Area of Science:

• Photonics
• Optical Computing
• Signal Processing

Background:

• Convolution is a key operation in neural networks and signal processing, demanding significant computational resources.
• Photonic convolution offers a promising alternative to electronic methods, potentially overcoming computational bottlenecks.
• The synthetic frequency dimension allows for compact device designs by utilizing light's spectral properties.

Purpose of the Study:

• To experimentally demonstrate convolution operations within the synthetic frequency dimension.
• To showcase the synthesis of arbitrary convolution kernels using modulated ring resonators.
• To explore methods for enhancing kernel implementation capabilities.

Main Methods:

• Utilizing a modulated ring resonator to synthesize convolution kernels.
• Employing predetermined modulation waveforms for accurate kernel synthesis.
• Performing convolution computations with input frequency combs and synthesized kernels.
• Introducing an additive offset to expand achievable kernel types under limited modulation strength.

Main Results:

• Successful experimental realization of convolution in the synthetic frequency dimension.
• Accurate synthesis of arbitrary convolution kernels.
• Demonstration of convolution computation between frequency combs and kernels.
• Validation of the additive offset technique for broader kernel implementation.

Conclusions:

• The synthetic frequency dimension is an effective approach for data encoding and computation in photonics.
• This method leads to the development of compact and scalable photonic computation architectures.
• The demonstrated techniques pave the way for advanced optical signal and image processing.