Hinode/XRT stray light correction

SUMMARY of the XRT Stray Light Study -- Measurement and Correction.

Link to the 2015 phase 1 summary page.
Link to the 2015 phase 2 summary.
Link to the 2016 summary.

(20th August, 2014 by Aki Takeda)
(28-Jan-2015: correction factor part added.)
(3-Mar-2015: correction factor polynomial fitting added.)
(22-Sep-2015: plot and fitting of the factor corrected.)

The visible stray light component contaminated in the X-ray filter images were measured during the eclipse season of 2014.
Introduction to the stray light in XRT

1. Dates of the measurements and purposes.

timeline No.	date & time	link to the page	purposes
1	6-May-2014, 17:50UT	to the 1st test page	collects light curve, Ti-poly (250msec) at Dick Center
2	15-May-2014, 18:20UT	to the 2nd test page	light curve, Ti-poly (1.4 sec) & Al-mesh (707msec) at DC
3	27-May-2014, 20:30UT	to the 3rd test page	light curve, Al-poly (1.0 sec) & C-poly (1.41 sec) at DC
4	3-Jul-2014, 18:30UT	to the 4th test page	light curve, Ti-poly (1.4 sec) & Thin-Be (2.83 sec) at DC
5	30-Jul-2014, 18:30UT	5th and 6th test page	light curve, Med-Al (32 sec) at DC
6	31-Jul-2014, 15:49UT		light curve, Med-Be (24 sec) at DC
	31-Jul-2014, 17:27UT		light curve, Thick-Al (64 sec) at DC
	1-Aug-2014, 16:26UT		light curve, Thick-Be (64 sec) at DC
7	8-Aug-2014, 18:56UT	to the 7th test page	light curve, Ti-poly (4.0 sec) & C-poly (4.0ec) at N-pole
	8-Aug-2014, 20:40UT		light curve, Ti-poly (1.4 sec) & C-poly (1.4 sec) at E-limb
	8-Aug-2014, 22:30UT		light curve, Ti-poly (4.0 sec) & C-poly (4.0 sec) at S-pole

2. Main results.

Out of 9 X-ray analysis filters, significant stray-light components were detected only with the Ti-poly and C-poly filters.
Stray-light images consist of the following 3 components:
- solar disk component (brightest).
- structures particular to each filter (Ti-poly: stable, C-poly: pointing dependant).
- dark component, i.e., fine stripes similar to those found in dark frames. (time dependant, weakest contribution).
At the polar- and limb-pointings, the stray-light images are different most significantly in the solar disk component.

3. Stray-light image gallery.

Ti-poly series (click to enlarge):

Disk Center pointing	North Pole pointing	East Limb pointing	South Pole pointing

XRT20140515_182503.4.fits	XRT20140808_191124.0.fits	XRT20140808_205020.5.fits	XRT20140808_222905.4.fits

Movie of all stray-light images in Ti-poly

C-poly series (click to enlarge):

Disk Center pointing	North Pole pointing	East Limb pointing	South Pole pointing

XRT20140527_204601.0.fits	XRT20140808_191137.1.fits	XRT20140808_205031.0.fits	XRT20140808_222918.5.fits

Movie of all stray-light images in C-poly

How to correct an image (example):

IDL> dir_l0='/disk/data/HINODE/xrt/level0/'
IDL> file=dir_l0+'2014/05/15/H1800/XRT20140515_182503.4.fits' ; T-poly stray-light image at DC pointing
IDL> mreadfits,file,index,data
IDL> xrt_prep,index,data,i_stray,d_stray,/norm,/float,/desp

; img_before : an image to be corrected, 1024x1024, Ti-poly, at DC poining, PREPed, intensity normalized.  
; img_after  : an image after correction.
; d_stray    : the stray-light image prepared in the above.

IDL> img_after = img_before - d_stray

4. Corrected image gallery.

Ti-poly at disk center pointing

BEFORE [synoptic (triple) composite image]	AFTER [stray light image (prep-ed) subtracted]

C-poly at disk center pointing

BEFORE [synoptic (double) composite image]	AFTER [stray light image (prep-ed) subtracted]

Ti-poly at N-pole pointing

BEFORE [synoptic (double) composite image]	AFTER [stray light image (prep-ed) subtracted]

C-poly at N-pole pointing

BEFORE [(double) composite image before twilight]	AFTER [stray light image (prep-ed) subtracted]

Ti-poly at E-limb pointing

BEFORE [(double) composite image before twilight]	AFTER [stray light image (prep-ed) subtracted]

C-poly at E-limb pointing

BEFORE [(double) composite image before twilight]	AFTER [stray light image (prep-ed) subtracted]

Ti-poly at S-pole pointing

BEFORE [(double) composite image before twilight]	AFTER [stray light image (prep-ed) subtracted]

"Ghost" from 2007 solar eclilpse image (from Kano-san's report 21-Feb-2007 to XRT_COs).
Thin-Be CME watch image (and its enhancement obtained on 1-Dec-2015, processed by Trish Jibben, private communication on 4-Dec-2015).

C-poly at S-pole pointing

BEFORE [(double) composite image]	AFTER [stray light image (prep-ed) subtracted]

5. Cosmetic correction of past/future browse images (PNG).

Correction of May-2012 images by 2014 stray-light patern.

synoptic (triple) composite image	stray-light image subtracted	stray-light image * 0.5 subtracted

The stray-light is increasing: the intensity on 9-May-2012 is about half of that on 15-May-2014.
In order to correct the images in the past (or outside eclipse season), we need to know the temporal variation of the stray light intensity (from 9-May-2012 to present).

5.1 Determination of the factor for the correction.

Let's assume that the images are corrected in the following way:

[image_corrected] = [Image_uncorrected] - factor(k) * [stray-light_2014]

The first rough estimation of the correction factor applied to the initial PNG products was as follows (shown in the green line in the plot below) :

k = 0.5 (9-May-2012 18:00 through 14-Aug-2012 23:59)
k = 0.7 (15-Aug-2012 00:00 through 31-Dec-2012 23:59)
k = 1.0 (1-Jan-2013 00:00 and onward)

More reasonable value of the factor is determined by using intensity correlation between Al-mesh and Ti-poly images. Note that Al-mesh images are not affected by the stray-light.

[step 1] Find a pair of Al-mesh and Ti-poly images obtained in the same synoptic program (taken almost simultaneously).
[step 2] Applying the mask on an Al-mesh image, extract only pixels whose intensity is in the range, [0,500] in DN (see the right image).
[step 3] Apply the same mask on the corresponding Ti-poly image.
[step 4] Correct the Ti-poly image with a factor, k (later try variations from 0.45 to 1.10).
[step 5] Take cross-correlation between the masked Al-mesh and corrected Ti-poly images (see scatter plots below).
[step 6] With different values of k, determin the one that minimizes the square sum of the residuals of the data points of the corrected image (red dots) to the line fitted for 1-9 May, 2014 images (blue line in the above plot).


White: Pixels from 1-May-2014 6UT to 9-May-2014 6UT, masked synoptic images. Blue line: linear fit (by LADFIT, `yy=2.451+0.341xx`). Red (left): 9-May-2014 18UT image BEFORE* correction, Red (right): AFTER correction with the correction factor of 0.5.

The temporal variation of the factor k, determined by the above steps 1-6 are shown with '+' signs in the plot below (see the next subsection for the red and blue lines).

The stray light intensity increases from the beginning(9-May-2012) through around Jan-2013.
Short term increase (with ~3 weeks period) are seen, probably due to the increasing amount of contamination on the CCD, and clearance by the CCD bakeout.

5.2 Fitting the variation of the factor, k.

[NOTE]
This sub-section(5.2) has a new version in which the fitting was re-done using the whole data in phase 1 (9-May-2012 through 14-Jun-2015).
[NOTE]

The above data points are determined for the Al-mesh and Ti-poly pairs, each of the pair image are 'intact' (i.e., with no saturation nor telemetry drop). However, some of the intact Ti-poly images that do not have the Al-mesh counterpart are not included.

Applying the polynomial fit (7th order) to the data points between 9-May-2012 and 31-Dec-2014 (red line in the above plot).

k0 = pl7coef(0)           $    ; pl7coef(0)=   2.1924060 
       +tt*pl7coef(1)     $    ; pl7coef(1)=  -0.040220413 
       +tt^2*pl7coef(2)   $    ; pl7coef(2)=   0.00035595453
       +tt^3*pl7coef(3)   $    ; pl7coef(3)=  -1.4460938e-06
       +tt^4*pl7coef(4)   $    ; pl7coef(4)=   3.1434595e-09
       +tt^5*pl7coef(5)   $    ; pl7coef(5)=  -3.7773749e-12
       +tt^6*pl7coef(6)   $    ; pl7coef(6)=   2.3664218e-15
       +tt^7*pl7coef(7)        ; pl7coef(7)=  -6.0336921e-19

To avoid negative values of the faint structures due to over-correction, the offset value of -0.05(shifting downward) was applyied as the final value of the correction factor (blue line).

k = k0 - 0.05

I made two IDL routines:
1) GET_SLCORFACT.PRO : Calculate the correction factor(k) at an arbitrary time (returns blue line points).

IDL> print,get_slcorfact('9-may-2012 18:00')
      0.50596564
IDL> print,get_slcorfact('15-may-2014 18:00')
      0.98621502
IDL> print,get_slcorfact('1-may-2012 00:00') 
*** the input time before [9-may-2012] set to 0.
2012/05/01 00:00:00.000
      0.00000
IDL> print,get_slcorfact('1-mar-2015 00:00')
*** the input time after [31-dec-2013] set to 1.
2015/03/01 00:00:00.000
      1.00000

2) RD_SLCORFACT_RAW.PRO : Provide raw data points (returns '+' sign points).

IDL> rd_slcorfact_raw, index,factor
% Compiled module: RD_SLCORFACT_RAW.
INDEX           STRUCT    = -> MS_163834261002 Array[1938]
FACTOR          FLOAT     = Array[1938]
IDL> utplot, index.date_obs,factor,psym=1  ; to give the above plot

5.3 The correction applied to the current SCIA PNG products.

The PNG products for May-2012 through Dec-2014 were re-processed using the k value calculated by the GET_SLCORFACT.PRO. (2 March, 2015).

The full-res images (2048x2048) included in the above period were also corrected by sabtracting the half-res stray light image(1024x1024) re-bined to 2048x2048.

The PNG products after Jan-2015 are tentatively corrected with the value of k=1.0. The k value will be updated in the future by a new fitting applied to the new data.