main
Staged pipeline (coarse -> level0 -> level1 -> tilt -> camera_length) for detector geometry refinement using CrystFEL's Millepede-based align_detector, replacing the old crystfel/0.10.2 detector-shift workflow. Each stage iterates indexamajig + align_detector to convergence before advancing, and the run produces four diagnostic plots (per-stage convergence, per-panel refinement support, indexed pattern count, unit-cell before/after). The coarse stage falls back to crystfel/0.10.2's predict_refine/det_shift (dead in 0.12.0, always reports 0.000/0.000) to bring badly-off starting geometries within Millepede's ~100-300um working range before handing off, since this crystfel/0.12.0 build has no --big-errors flag to relax its default reject threshold. Validated end-to-end on p23523's previously-failing geometry (99.4% reject rate without coarse pre-alignment). crystfel_split_mille.py and millepede-refine are the supporting indexing/ alignment helpers the pipeline reuses; replot_pipeline_v2.py regenerates the four plots from an already-completed run's on-disk data, for recovery if a run crashes after the expensive cluster work is done. SESSION_SUMMARY.md documents the investigation findings behind this work. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
swissmx_tools
Tools and utils for swissMX such as CTA interaction, data recovery and additonal jfjoch scripts
Description
Tools and utils for swissMX such as CTA interaction, data recovery and additonal jfjoch scripts
88 KiB
Languages
Python
97.2%
Shell
2.8%