Owner's review of the rotation figure:
- The in-group merge names the correction surfaces it fits (decay, absorption,
modulation; RotationScaleMerge) instead of the word "corrections".
- New "autodiff" badge, on the fits whose derivatives are automatic: Refine and
the rotation indexer's lattice refinement (XtalOptimizer/XtalRefine, forward-
mode Dual numbers), Post-refine (ceres::AutoDiffCostFunction in PostRefine.cpp)
and the stills geometry bundle adjustment (GeometryRefiner). Not on the rigid
body (analytic plus forward differences), the stills partiality refinement
(central differences), the beam-centre fit or the stills indexer's candidate
refinement. The Ceres text on Refine is gone.
- Text back to about the owner's budget and box sizes: the meaning moved to
arrows ("pass 2, or pass 1 if it merges better" into the P1 merge, "beam
centre" from the pre-scan to Index) and badges; "rival ..." phrases removed.
New: the stills figure (rugnux_pipeline_stills.svg) from the same script, same
style and legend. Stills = no goniometer axis or a stationary one
(RugnuxRotationData). Geometry bundle adjustment first, only with a known cell
(dashed); pre-scan masks the beam stop and bad pixels; per image: decode,
profile, spots, index (1D FFT, or ffbidx with a cell; min-pix chosen per image
by what indexes: tested), refine beam/orientation/cell, excitation-error
prediction, profile fit; whole dataset: consensus cell and per-image scale (CPU),
symmetry, merge with partiality post-refinement and error model, structure, report.
Both figures are on RUGNUX_OVERVIEW.md, the stills one under "Stills instead".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PXc1zD1yxHqaR1DGdCTHL4
leonarski_fandClaude Opus 5.5
f9fa106688
