Waveguide configuration analysis for optical joint transform correlator: optimizing accuracy in photonic convolutional accelerators

We present a detailed analysis of waveguide configuration strategies for optical Joint Transform Correlator (JTC) implementations in photonic convolutional accelerators. Using experimental measurements from a fabricated chip with 1D lens-based Fourier transform capabilities, we develop PyTorch-based simulations that incorporate measured component non-idealities to investigate how different zero-padding arrangements and waveguide configurations affect convolution accuracy. Our PyTorch-based simulations reveal optimal design parameters for waveguide spacing and padding strategies, providing practical guidelines for future photonic accelerator implementations. The results demonstrate how strategic waveguide configuration can enhance convolution accuracy while maintaining fabrication feasibility.