Versions Compared

Old Version 11

changes.mady.by.user Mikhail Dubrovin

Saved on

compared with

New Version Current

changes.mady.by.user Mikhail Dubrovin

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Content

Table of Contents

Mask

To create mask use Mask Editor command med or launch it through the calibman.

Geometry center

ix_cent, iy_cent = det.point_indexes(runnum)

879, 871 (then x and y are swapped to Cartesian system).

Equatorial mask

Mask parameters

Code Block
Wedge  871  879  454  387  148  196 1 k False False
Wedge  871  879  454  387  -27   21 1 k False False

...

Code Block
t   = pars[0]              # figure type, ex. 'Wedge'
x   = float(pars[1])       # x coordinate of the wedge center
y   = float(pars[2])       # y coordinate of the wedge center
r   = float(pars[3])       # 1-st radius of the wedge 
w   = float(pars[4])       # radial width of the wedge 
t1  = float(pars[5])       # 1-st angle
t2  = float(pars[6])       # 2-nd angle
lw  = int(pars[7])         # line width
col = str(pars[8])         # color
s   = self.dicBool[pars[9].lower()]   # isSelected boolean parameter
rem = self.dicBool[pars[10].lower()]  # isRemoved - helper parameter

Image ModifiedImage Modified

Involved sensors in adopted cxif5315 geometry: 0, 1,    9,15,    16,17,   25,31

...

Code Block
---------
| 1 | 0 |
----+----
| 2 | 3 |
---------

Arc mask

Code Block
Wedge  871  879  454   26  -174  186 1 k False False

...

Involved sensors in the top part of the image in adopted cxif5315 geometry: 0, (1), 7,   8, (9), 15

Equatorial and arc combined mask

 

Radial background subtraction

For some reason polarization correction does not work well in this experiment for entire image.

Comparison of the 2-d interpolated radial background subtraction 

  • nbins (rad:500, phi:1)

Image RemovedImage Removed

  • nbins (rad:500, phi:32)

Image RemovedImage Removed

SIngle angular bin still works fine in our ROI defined by both masks.

Comparison of the pfv2 with Cheetah list of peaks

Code for comparison: cxif5315/peak-list-comp-cheetah.py

Peak lists for comparison:

  • peakfinder: pfv2-cxif5315-r0169-2016-03-28T15:02:47.txt
  • cheetah: r0169-cheetah-peaks.txt - with unknown origin...

Comparison algorithm and conditions

Use pyimgalgos.TDFileContainer for both peakfinder and cheetah

Code Block
    fc_ch = TDFileContainer(fncheet, indhdr='frameNumber', objtype=TDCheetahPeakRecord) # , pbits=256)
    fc_pf = TDFileContainer(fnpeaks, indhdr='Evnum', objtype=TDPeakRecord) #, pbits=256)

...

Masks for segments

Code below shows how to generate mask n-d arrays for particular set of segments (2x1s)

Code Block
shape_cspad = (32,185,388)
seg1 = np.ones((185,388))
mask_winds_all = np.zeros(shape_cspad, dtype=np.int16)
mask_winds_equ = np.zeros(shape_cspad, dtype=np.int16)
mask_winds_arc = np.zeros(shape_cspad, dtype=np.int16)

mask_winds_all[(0,1,7,8,9,15,16,17,23,24,25,31),:,:] = seg1
mask_winds_equ[(0,1,9,15,16,17,25,31),:,:] = seg1
mask_winds_arc[(0,7,8,15),:,:] = seg1

Only listed segments are highlighted on plots:

 

Image AddedImage AddedImage Added

Background

Averaged background shape

 

Background shape evaluated in exp=cxif5315:run=162 with common mode correction - central segments got offset to higher intensity

 

Image Added

Background shape evaluated evaluated in exp=cxif5315:run=162 without common mode correction - common mode offsets averaged out to zero and image represents true background shape.

Command to average data with options for processing algorithms:

Code Block
det_ndarr_average -d exp=cxif5315:run=162 -s CxiDs2.0:Cspad.0 -n 10000 -m 5000 -a 3 -f nda-bkgd -p -v

skips 5000 and averages next 10000 events.

Deployed as: /reg/d/psdm/cxi/cxif5315/calib/CsPad::CalibV1/CxiDs2.0:Cspad.0/pixel_bkgd/95-end.data

 

Image AddedImage Added

 

Radial background subtraction

polarization correction factor map orientation should be consistent with geometry file - in cxif5315 it should be rotated by 90°.

Code Block
from pyimgalgos.RadialBkgd import RadialBkgd, polarization_factor

    X, Y, Z = geo.get_pixel_coords()
    rb = RadialBkgd(X, Y, mask, nradbins=500, nphibins=1) 
    pf = polarization_factor(rb.pixel_rad(), rb.pixel_phi()+90, 91.33e3)

# in event loop:
    nda, title = rb.subtract_bkgd_interpol(arr * pf) * mask

For example we apply this algorithm to the water background averaged sample from exp=cxif5315:run=162:

Image AddedImage Added

Radial background fails to work in the region of shadow, where image miss symmetry. This region can be removed by constraining the range of angles phiedges=(40, 325).

In the next plot the radial range is also constrained as radedges=(5000,100000) for the purpose of example:

Code Block
# the same as in previous case, but
rb = RadialBkgd(X, Y, mask, nradbins=500, nphibins=1, phiedges=(40, 325), radedges=(5000,100000))

Image AddedImage Added

SIngle angular bin still works fine in our ROI defined by both masks.

Background subtraction examples

Subtract background shape evaluated in run 192

Code Block
    from pyimgalgos.GlobalUtils import subtract_bkgd
		# once per run:
        nda_peds  = det.pedestals(runnum)
        nda_bkgd  = det.bkgd(runnum)      # get n-d array with averaged background from calib/.../pixel_bkgd
        nda_smask = det.mask(evt, calib=False, status=True, edges=True, central=True, unbond=True, unbondnbrs=True)

        # windows for background normalization
        winds_bkgd = [(s, 10, 100, 270, 370) for s in (4,12,20,28)] # use part of segments 4,12,20,28 to subtr bkg

        # in the event loop
        nda_raw = det.raw(evt

...

Statistics for number of peaks:

  • total in Cheetah: 14002 
  • matched: 10593
  • not-matched: 3409
  • fraction of matched: 0.757

Plots for number of peaks in peak-finder, Cheetah, and number of non-matched:matched Cheetah peaks.

Image RemovedImage RemovedImage Removed

Peak-finder number of peaks accounts for arc region.

Peak finding

Try to work with  versions > v1 of peak-finders:

Parameters:

 

 

 

Code Block
alg_arc = PyAlgos(windows=winds_arc, mask=mask_arc, pbits=2)
alg_arc.set_peak_selection_pars(npix_min=4, npix_max=500, amax_thr=0, atot_thr=600, son_min=5)

alg_equ = # all parameters are the same as for alg_arc

        #peaks_arc = alg_arc.peak_finder_v2(nda, thr=10, r0=6, dr=0.5)
        #peaks_equif nda_raw is not None :
            nda = alg_equ.peak_finder_v3 np.array(nda_raw, rankdtype=5np.float32, r0=6, dr=0.5) ??????????? finds too many peakscopy=True)
            nda -= nda_peds
 
            # Subtract background shape averaged for pure water
        peaks_arc    nda = alg_arc.peak_finder_v4subtract_bkgd(nda, thrnda_low=10bkgd, thr_high=150mask=nda_smask, rank=5, r0=6, dr=0.5)

 

 

 

...

winds=winds_bkgd)

Image AddedImage Added

Radial background subtraction

Image corrected up to 80mm, rings span for entire phi

Image AddedImage Added

 

Image corrected up to 80mm, 40<phi< 320:

Image AddedImage Added

OLD IMAGE PROCESSING

Subtract background shape evaluated in run 192. 

Note

Background shape was evaluated WITH common mode correction; central 2x1s got offset due to non-uniform water background shape.

Averaged Fraser- transformed image using angles from fit to 2 arc peaks

Image AddedImage Added


Spectrum of intensities


Spectrum of intensities of all quad 2x1-segment 0 (close to beam) and 4 (water ring region)

before background subtraction

Image AddedImage Added

after background subtraction

Image AddedImage Added

  • masked pixels contributes to peak at 0
  • 1-, 2- and 3- photon peaks are seen
  • common mode correction before background subtraction does not work well in this data sample due to significant fraction of 1-photon peak next to noise peak, which makes an offset due to illumination.
  • common mode correction after background subtraction does not work - it moves noise peak to 0 and destroys background subtraction results.

Note

Potentially any non-dark data spectra can be used to calibrate pixel gain.

References