Asynchronous nonlinear optical sampling with parallelized signal reconstruction

We present a scalable software framework for synchronization and waveform reconstruction in asynchronous nonlinear optical sampling, designed to enable efficient processing of long, bandwidth-constrained measurement records. High-speed optical signals exceeding electronic baseband bandwidth are sampled using a nonlinear optical sampling gate based on four-wave mixing and a low-bandwidth photodetector, and are resynchronized entirely in software by estimating the sampling time step through spectral peak refinement using zero-padded Fourier transforms. Because the proposed framework relies primarily on Fourier analysis, peak estimation, and time-axis remapping, it avoids iterative optimization and is inherently well suited to parallel execution. We experimentally validate the framework using 10-GHz mode-locked laser pulses and 10-Gb/s pseudo-random bit sequence (PRBS) signals, and quantitatively assess reconstruction quality in terms of extinction ratio, Q-factor, pulse width, and timing jitter. A composite figure of merit is introduced to guide parameter selection under trade-offs between amplitude and timing fidelity. GPU-based parallelization of the proposed software framework yields an approximately tenfold reduction in processing time compared with a CPU implementation, demonstrating scalability to long datasets. The reconstructed waveforms with picosecond pulse width indicate picosecond-scale temporal resolution, highlighting the applicability of the framework to waveform-level diagnostics in future ultra-high-speed communication systems.