How to remove discrepencies in a 2D pattern produced from the integration of an image calculated with calcfrom2d ?

[loichuder]

I want to write a test that checks that the integration 2D works as intended. For this, I am generating arrays for radial, azimuthal and the resulting integrated patterns:

tth = numpy.linspace(0, 30, 200)
chi = numpy.linspace(-180, 180, 180)
intensity = numpy.full((chi.size, tth.size), 10)

I am then using ai.calcfrom2d to create an image whose 2D integration should give the intensity array:

# ai is the azimuthal integrator
calc_image = ai.calcfrom2d(intensity, tth, chi, mask=ai.mask)

But the problem is that when I do the integration on this calc_image, I do get "holes" in the integrated pattern:

result = ai.integrate2d(calc_image, tth.size, chi.size)
integrated_intensity = result.intensity

I attached an HDF5 file that contain all the data for easier browsing: test_pyFAI_2d.zip

I guess some bins are empty but I do not really understand why ? How could I avoid this ? Should I create a more sparse intensity array ?

Full Python script

import pyFAI
import pyFAI.detectors
from pyFAI.azimuthalIntegrator import AzimuthalIntegrator
import numpy
import matplotlib.pyplot as plt
geometry = {

"dist": 10e-2,  # 10 cm

"poni1": 10e-2,  # 10 cm

"poni2": 10e-2,  # 10 cm

"rot1": numpy.radians(10),  # 10 deg

"rot2": 0,  # 0 deg

"rot3": 0,  # 0 deg

}

ai = AzimuthalIntegrator(

detector=pyFAI.detectors.detector_factory("Pilatus1M"), wavelength=1e-10, **geometry

)

tth = numpy.linspace(0, 30, 200)

chi = numpy.linspace(-180, 180, 180)

intensity = numpy.full((chi.size, tth.size), 10)
calc_image = ai.calcfrom2d(intensity, tth, chi, mask=ai.mask)
result = ai.integrate2d(calc_image, tth.size, chi.size)

integrated_intensity = result.intensity
f, ax = plt.subplots(ncols=2)

ax[0].pcolormesh(intensity)

ax[0].set_title("Original")

ax[1].pcolormesh(integrated_intensity)

ax[1].set_title("Integrated")
plt.show()

[kif]

The holes you see are consequences of the gaps in the detector "Pilatus 1M" which are hard-coded in the pyFAI.detectors class.
https://github.com/silx-kit/pyFAI/blob/main/src/pyFAI/detectors/_dectris.py#L69
You can always re-define the detector's mask like:

detector = pyFAI.detector_factory("Pilatus1M")
detector.mask = numpy.zeros(detector.shape, "int8")

Note that the calcfrom1/2d does interpolation while integrate performs histogramming, the difference is subtle but you won't get back what you input initially.

[loichuder]

The holes you see are consequences of the gaps in the detector "Pilatus 1M"

Of course, that makes sense. I wonder why I didn't think about that 🤦

Thanks!