Speaker
Description
As the SLAC high-level applications environment begins migrating towards Python-based frameworks, rigorous validation against legacy tools is critical for operational continuity. We present a comprehensive benchmarking study comparing a new Python wire scanner analysis framework (slac-tools) against legacy MATLAB software at LCLS using 379 historical datasets from routine operations since 2022.
MATLAB .mat files were converted to HDF5 format, preserving wire encoder positions, PMT traces, and original beam size measurements. The Python pipeline—including automated profile windowing, Gaussian fitting, and beam size extraction—was applied identically to converted datasets, treating historical data as live acquisitions.
Comparison of 1062 measurements showed 62% yielded physically valid results after two-stage filtering: excluding non-physical disagreements (≥100% difference) and statistical outliers (1.5×IQR). Valid measurements demonstrated excellent agreement with median difference of 0.31% and mean of 1.38%—well below operational uncertainty. Beam sizes ranged from 20-1000 microns across multiple sectors and modes.
The Python framework is in the process of deployment for live acquisition with comparable performance. This work establishes a systematic methodology for validating modern analysis tools using real operational data and demonstrates HDF5 as a durable format for diagnostic archiving.
| I have read and accept the Privacy Policy Statement | Yes |
|---|