Jerome Kieffer

Hi Jesse, I know integrating detectors are striking back ... we have Jungfrau detectors here and I need to implement the error propagation for those detectors as well. My point...

Apparently, when data are actually synthetic, the two approaches look equivalent ... 

So Jesse, you are right, we need to keep the option error_model="azimuthal", and implement them all (ain't gonna be funny). This makes me wonder how to deal with heavily tilted...

I am trying to find a configuration where a detector is tilted by ~45° ... this will ensure the solid angle within a ring varies enough to have significant impact...

To investigate, the geometry of fit2d helps. One can set the tilt to ~70°, the beam being centered on the detector and used this as a starting point.

This is apparently UN-avoidable with Bragg-peaks: ``` import numpy ensemble = numpy.ones(1> 977.5615234375 31234.706249089617 sigma = numpy.sqrt(ensemble.sum()/len(ensemble)) print(sigma) >>> 31.26598028908577 (1e6-mean)/sigma, (1-mean)/sigma >>>(31952.3785673625, -31.23399664453812) ```

We can expect that the default tolerance from all_close are good.

I suspect this is the same bug: ``` > gonioref.sload(json_file)#, gonioref > > =============================================================== > > ---------------------------------------------------------------------------RuntimeError > Traceback (most recent call > last) in 1 > print(open(json_file).read()) 2 ---->...

Please refer to this tutorial ... https://github.com/silx-kit/pyFAI/blob/master/doc/source/usage/tutorial/Goniometer/MX-calibrate/MX-calibrate.ipynb

not closed but low priority for now.