ess.sans.beam_center_finder.beam_center
ess.sans.beam_center_finder.beam_center#
- ess.sans.beam_center_finder.beam_center(data, data_monitors, direct_monitors, wavelength_bins, q_bins, gravity=False, minimizer='Nelder-Mead', tolerance=0.1)#
Find the beam center of a SANS scattering pattern. Description of the procedure:
obtain an initial guess by computing the center-of-mass of the pixels, weighted by the counts on each pixel
from that initial guess, divide the panel into 4 quadrants
compute \(I(Q)\) inside each quadrant and compute the residual difference between all 4 quadrants
iteratively move the centre position and repeat 2. and 3. until all 4 \(I(Q)\) curves lie on top of each other
- Parameters
data (
DataArray
) – The DataArray containing the detector data.data_monitors (
Dict
[str
,DataArray
]) – The data arrays for the incident and transmission monitors for the measurement run.direct_monitors (
Dict
[str
,DataArray
]) – The data arrays for the incident and transmission monitors for the direct run.wavelength_bins (
Variable
) – The binning in the wavelength dimension to be used.q_bins (
Union
[int
,Variable
]) – The binning in the Q dimension to be used.gravity (
bool
, default:False
) – Include the effects of gravity when computing the scattering angle ifTrue
.minimizer (
str
, default:'Nelder-Mead'
) – The Scipy minimizer method to use (see the Scipy docs for details).tolerance (
float
, default:0.1
) –Tolerance for termination (see the Scipy docs for details).
- Returns
Tuple
[Variable
,Variable
] – The beam center position as a vector.
Notes
We record here the thought process we went through during the writing of this algorithm. This information is important for understanding why the beam center finding is implemented the way it is, and should be considered carefully before making changes to the logic of the algorithm.
Use a + cut, not an X cut
The first idea for implementing the beam center finder was to cut the detector panel into 4 wedges using a cross (X) shape. This is what Mantid does, and seemed natural, because the offsets when searching for the beam center would be applied along the horizontal and vertical directions. This worked well on square detector panels (like the SANS2D detector), but on rectangular detectors, the north and south wedges ended up holding many less pixels than the east and west panels. More pixels means more contributions to a particular \(Q\) bin, and comparing the \(I(Q)\) curves in the 4 wedges was thus not possible. We therefore divided the detector panel into 4 quadrants using a
+
cut instead. Note that since we are looking at an isotropic scattering pattern, the shape of the cut (and the number of quadrants) should not matter for the resulting shapes of the \(I(Q)\) curves.Normalization inside the 4 quadrants
The first attempt at implementing the beam center finder was to only compute the raw counts as a function of $Q$ for the sample run, and not compute any normalization term. The idea was that even though this would change the shape of the \(I(Q)\) curve, because we were looking at isotropic scattering, it would change the shape of the curve isotropically, thus still allowing us to find the center when the curves in all 4 quadrants overlap. The motivation for this was to save computational cost.
After discovering the issue that using a
X
shaped cut for dividing the detector panel would yield different contributions to \(I(Q)\) in the different wedges, we concluded that some normalization was necessary. The first version was to simply sum the counts in each quadrant and use this to normalize the counts for each intensity curve.This was, however, not sufficient in cases where masks are applied to the detector pixels. It is indeed very common to mask broken pixels, as well as the region of the detector where the sample holder is casting a shadow. Such a sample holder will not appear in all 4 quadrants, and because it spans a range of scattering (\(2{\theta}\)) angles, it spans a range of \(Q\) bins.
All this means that we in fact need to perform a reduction as close as possible to the full \(I(Q)\) reduction in each of the 4 quadrants to achieve a reliable result. We write ‘as close as possible’ because In the full \(I(Q)\) reduction, there is a term \(D({\lambda})\) in the normalization called the ‘direct beam’ which gives the efficiency of the detectors as a function of wavelength. Because finding the beam center is required to compute the direct beam in the first place, we do not include this term in the computation of \(I(Q)\) for finding the beam center. This changes the shape of the \(I(Q)\) curve, but since it changes it in the same manner for all \({\phi}\) angles, this does not affect the results for finding the beam center.
This is what is now implemented in this version of the algorithm.