Py fai matchup sept2026

readme.md

@@ -6,4 +6,44 @@ They work by fitting the diffraction geometry for a series of images of powder p
 You will need to edit/modify them for any other detector geometry. 
 If you can log in at ESRF you can find the data files referenced here, otherwise at Zenodo : [![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.6344000.svg)](https://doi.org/10.5281/zenodo.6344000). 
 
-Uploaded to github on 6 Dec 2021 by Jon Wright.
+Uploaded to github on 6 Dec 2021 by Jon Wright.
+
+## Addendum, September 2026: `fitpksEiger_pyFAI_2026.9.0.dev0+1e06043e4.ipynb`
+
+This notebook is a copy of `fitpksEiger.ipynb` (the ImageD11-based module-geometry fit is
+unchanged, cell for cell, apart from one deliberate fix noted inline with `# FIXED !!!!`:
+the pyFAI spatial-distortion file now uses pyFAI's own 0-to-1 pixel-corner convention
+instead of the original -0.5/+0.5 one). It appends a full cross-check of the result against
+pyFAI:
+
+- converts the converged ImageD11 geometry to a pyFAI PONI file, using pyFAI's own
+  `pyFAI.geometry.imaged11` ImageD11<->pyFAI conversion helpers, referencing the corrected
+  spatial file;
+- compares peak positions computed by ImageD11 and by pyFAI's `Geometry.tth`/`chi` for the
+  same peaks, converged geometry, and reference CeO2 rings (should - and does - agree to a
+  few 0.1 mdeg, since this checks the conversion, not two independently-fitted geometries);
+- as a third, independent check that doesn't touch ImageD11's peaksearch at all: runs
+  `integrate2d` directly on the raw calibration image and computes a background-subtracted,
+  per-azimuth-bin peak centroid for every visible CeO2 ring, with contaminated bins (those
+  whose region-of-interest touches a masked/gap pixel or the detector edge, grown by 1
+  pixel) detected and excluded rather than averaged over.
+
+**To reproduce**, in addition to what the original notebooks above already need:
+
+- **pyFAI from source, not a release.** The ImageD11<->pyFAI conversion
+  (`pyFAI.geometry.imaged11`) lives only on an as-yet-unmerged branch:
+  [`kif/doc_calib_ID11`](https://github.com/kif/pyFAI/tree/doc_calib_ID11) off
+  [silx-kit/pyFAI](https://github.com/silx-kit/pyFAI), specifically commit `1e06043e4`
+  ("upgrade tutorial"; `pyFAI.version` reports `2026.9.0-dev0`). Build/install pyFAI from
+  that checkout (`pip install -e .` or a meson-python editable install) rather than from
+  PyPI/conda.
+- **The raw calibration image**, in addition to the `eigerSpatial/*.flt` peak files the
+  original notebooks already regenerate: the notebook's final integrate2d cross-check reads
+  frame index 1 of `scan0001/eiger_0000.h5` directly (same beamtime/dataset as everything
+  else here - ESRF path `/data/id11/nanoscope/blc12841/id11/CeO2/CeO2_Spatial_distortion_apr2021/`,
+  or the corresponding file in the Zenodo record linked above if running outside ESRF), plus
+  `hdf5plugin` to read it (bitshuffle/LZ4-compressed Eiger data).
+
+Everything else the notebook produces along the way (`newspatial_pyFAI.h5`, `1.poni`,
+`fitted_eiger_pars2.dat`, etc.) is generated by the notebook itself; nothing else needs to
+be supplied by hand.