Hinode/XRT stray light correction 2017

XRT Stray Light Study 2017 -- Phase 3 (after 27-May-2017).

Link to the 2014 summary page.
Link to the 2015 phase 1 summary.
Link to the 2015 phase 2 summary.
Link to the 2016 (phase 2) summary.
Link to the 2017 phase 2 (before 3rd rupture).
Link to the 2018 phase 3 (before 4th rupture).

(26 June, 2017 by Aki Takeda)
(11 Sept, 2017 added results of later measurements.)
(22 Sept, 2017 added section of image correction.)
(10 Oct, 2020 corrected wrong label of "after PREP comparison" plot for Al-mesh.)

0. Motivation.

On 27-May-2017, around 11UT, a saturation of G-band images was observed (reported by DV [xrt_co:31950]). This strongly suggests another rupture of the XRT pre-filter. The flare patrol images (taken with Ti-poly) also show the increase of stray light component (see images below). The new rupture seems to have occurred during the period XRT paused observation due to SAA passage and S/C night (i.e., no observation between the two images below).

1. Purposes of the project 2017, phase 3.

Since 27-May-2017, XRT's stray light status has entered a new stage, so we call it "phase 3". (Note that we were in phase 2 since the 2nd pre-filter rupture event on 14-Jun-2015).

Measure the current level of the stray light of Ti-poly.
Also measure other thin filters (Al-mesh, Al-poly, and thin-Be) if they are affected by the stray light increase of this time.
Measure non disk-center stray light images for Al-mesh and Al-poly.

2. Summary of Results.

Ti-poly stray light intensities after 27-May-2017 (phase 3) are about 2.4 times larger than before (09-May-2017, phase 2). Now the maximum intensity of the stray light component exceeded 3500 DN/sec (around DC, 2x2 binning). So half-res observations with 1-sec exposure are not able to quantitatively detect X-ray signals.
Al-mesh stray light intensities after 27-May-2017 are ~10 DN/sec (around DC area, in 2x2 binning), and increased by about 2.3 times relative to the amount measured on 9-May-2017 (phase 2, 3-4 DN/sec).
Al-poly stray light intensities after 27-May-2017 are about 2.5 DN (around DC area, in 2x2 binning), while it was negligible at the measurement in phase 2 (< 1 DN/sec on 9-May-2017).
The stray light component in the thin-Be, med-Be filters are still negligible.
The stray light component in med-Al is still negligible, but a slight increase (5-10 DN per 32sec) from the measurement at phase 2 (11-May-2017) is seen in the plot. This needs to be confirmed by the measurement next eclipse season.
G-band VLS=open and close images have nearly equal intensities in the phase 3. Sunspots are barely seen by subtraction, but their intensity contrast is weak.

3. Measurements 2017 (phase 3) at a glance.

ID No.	date & time (UT)	XOB	purposes (link: obtained plots & images)	memo
5	13-Jun-2017, 17:31:30	1B98	light curve, Ti-poly (1.0s), and Al-mesh(2.83s) at DC.	prev. bakeout + 6 days. Ti-poly > 2500 DN. G-band monitor images saturated.
6	14-Jun-2017, 18:07:00	1B99	light curve, Al-poly (4.0s), Thin-Be (11.3s) at DC	prev. bakeout + 7 days. G-band monitor images saturated. Sunset during thin-Be stray light exposure.
7	15-Jun-2017, 18:42:30	1B9A	light curve, med-Al (32.0s) at DC	prev. bakeout + 8 days. G-band monitor images saturated. Sunset during stray light exposure.
8	16-Jun-2017, 17:39:30	1B9B	light curve, med-Be (22.6s) at DC	prev. bakeout + 9 days. G-band monitor images saturated. Sunset during stray light exposure.
9	18-Jun-2017, 17:12:00	1B9E	light curve, Ti-poly (500ms), Al-mesh (2.83s) at DC	prev. bakeout + 11 days. Monitor imgs to 384x384 no binning.
10	19-Jun-2017, 17:47:30	1B9F	light curve, Al-poly (4.0s), Thin-Be (11.3s) at DC	prev. bakeout + 12 days. Monitor imgs to 384x384 no binning. Sunset during thin-Be stray light exposure.
11	25-Jul-2017, 17:45:30	1BAC	light curve, Al-mesh(1.4s) and Al-poly(2.0s) at E-limb	prev. bakeout + 7 days. Monitor imgs to Al-mesh 384x384 full-res.
12	26-Jul-2017, 18:21:00	1BAC	light curve, Al-mesh(1.4s) and Al-poly(2.0s) at N-limb	prev. bakeout + 8 days. Monitor imgs to Al-mesh 384x384 full-res.
13	27-Jul-2017, 17:18:00	1BAC	light curve, Al-mesh(1.4s) and Al-poly(2.0s) at W-limb	prev. bakeout + 9 days. Monitor imgs to Al-mesh 384x384 full-res. Al_poly image looks strange. Need to confirm.
14	28-Jul-2017, 17:53:00	1BAC	light curve, Al-mesh(1.4s) and Al-poly(2.0s) at S-limb	prev. bakeout + 10 days. Monitor imgs to Al-mesh 384x384 full-res.
15	8-Aug-2017, 17:51:30	1BAE	light curve, Al-mesh(1.4s) and Ti-poly(500ms) at DC,	prev. bakeout + 1 days. Monitor imgs to Al-mesh 384x384 full-res. No NGT event, but X-ray well absorbed.
16	9-Aug-2017, 18:27:30	1BAF	light curve, Al-poly(2.8s) and Thin-Be(11.3s) at DC,	prev. bakeout + 2 days. Monitor imgs to Al-mesh 384x384 full-res. No NGT event, but X-ray well absorbed.

4. Stray light images 2017, at a glance.

Ti-poly filter

Tp 20150718 (1.4s) - phase 2 - [bo +1d]	Tp 20170509 (1.0s) - phase 2 - [bo +14d]	Tp 20170613 (1.0s) - phase 3 - [bo +6d]	Tp 20170618 (500ms) - phase 3 - [bo +11d]	Tp 20170808 (500ms) - phase 3 - [bo +1d]

After PREP comparison

20170618 v.s. 20170509 (bo+11d v.s. bo+14d)	20170808 v.s. 20150718 (bo+1d v.s. bo+1d)

Al-mesh filter

Am 20150718 (1.0s) - phase 2 - [bo +1d]	Am 20170509 (2.9s) - phase 2 - [bo +14d]	Am 20170613 (2.9s) - phase 3 - [bo +6d]	Am 20170618 (2.9s) - phase 3 - [bo +11d]	Am 20170808 (1.4s) - phase 3 - [bo +1d]

After PREP comparison

20170808 v.s. 20150718 (bo+1d v.s. bo+1d)	20170808 v.s. 20170509 (bo+1d v.s. bo+14d)

Am 20170725 (1.4s) - E-limb (ph3) -	Am 20170726 (1.4s) - N-limb (ph3) -	Am 20170727 (1.4s) - W-limb (ph3) -	Am 20170728 (1.4s) - S-limb (ph3) -

Al-poly filter

Ap 20160511 (4.1s) - phase 2 - [bo+9d]	Ap 20170510 (4.1s) - phase 2 - [bo+15d]	Ap 20170614 (4.1s) - phase 3 - [bo+7d]	Ap 20170619 (4.1s) - phase 3 - [bd+12d]	Ap 20170809 (2.8s) - phase 3 - [bd+2d]

After PREP comparison

20170619 v.s. 20160511
(bo+12d v.s. bo+9d)

Ap 20170725 (2.0s) - E-limb (ph3) -	Ap 20170726 (2.0s) - N-limb (ph3) -	Ap 20170727 (2.0s) - W-limb (ph3) -	Ap 20170728 (2.0s) - S-limb (ph3) -

thin-Be filter

tnBe 20160511 (5.8s) - phase 2 -	tnBe 20170510 (11s) - phase 2 -	tnBe 20170614 (11s) - phase 3 -	tnBe 20170619 (11s) - phase 3 -	tnBe 20170809 (11s) - phase 3 -

med-Be filter

med-Be,20170512(32s) - phase 2 -	med-Be,20170616(23s) - phase 3 -

med-Al filter

med-Al,20170511(32s) - phase 2 -	med-Al,20170615(32s) - phase 3 -

5. Image correction.

Correction of Al_mesh composite image.
- Composite image taken 2017/08/08 17:33:48
- Stray-light image taken 2017/08/08 18:01:26

[Click image to enlarge]

Before subtracting stray light img.	After subtracting stray light img.	Intensity profile, image(*,512)

Correction of Al_poly composite image.
- Composite image taken 2017/08/09 18:14:00
- Stray-light image taken 2017/08/09 18:38:21

[Click image to enlarge]

Before subtracting stray light img.	After subtracting stray light img.	Intensity profile, image(*,512)

6. Attempts to derive the "fudge" factor (k-factor).

[review of the Ti_poly image correction during phase 1]
Based on our previous study (on the stray light intensity during phase 1), the Ti_poly stray light intensity varies with CCD contamination layer thickness. We corrected the Ti_poly images taken anytime during phase 1, using a single stray light image taken during the eclipse season in 2014 by assuming the following simple relation.

 I_{_corrected} = I_{_uncorrected} - I_{_straylight} * k

The time variation of the factor k was determined by using the intensity correlation between Ti_poly and Al_mesh images taken closely in time so that the correlation after subtracting stray light component becomes closest to the correlation of these two filters before we have the stray light problem. The factor k was well determined with this method, and it turned out that the resulting k values predominantly reflect the variation of CCD contamination thickness obtained from G-band images.

[Attempt 1]
So, we first tried the same method as the Ti_poly image correction during the phase 1. This time, we use Al_mesh and Al_poly intensities, and compare their correlation in phase 3 with that in phase 1 (during which the stray light component of Al_mesh or Al_poly filters were both negligible). However, this method didn't work: We could not determine the value of k in a descent value range (0.5 to 2.0) by comparison with the phase 1. We also tried to find the k that maximize the correlation coefficient after the stray light correction, but without success. The failure of this method is due to the fact that the stray light subtraction does not make substantial change in distributions of data points of cross-correlation (Fig. 6.2):

In phase 3, both Al_mesh and Al_poly filters are affected by the stray light (we have no other choice, because thin-Be and thicker filters are low S/N at quiet region).
The stray light intensities of both filters are still small compared with that of Ti_poly filter in phase 1. (Ti_poly stray light intensity during phase 1 was 7 to 8 times stronger than the current Al_mesh stray light).

[click image to enlarge]

Fig.6.1 : Al_mesh vs Al_poly intensity correlation in May 2015 (phase 1).	Fig.6.2 : Al_mesh vs Al_poly intensity correlation on 14-Jun-2017 (phase 3).

[Attempt 2]
As the next step, we tried to use the G-band stray light (VLS=close) variation to estimate the variations of Al_mesh and Al_poly stray light, by assuming their behavior being the same as G-band's.

The G-band stray light intensities are measured regularly in the course of creating CCD contamination layer thickness DB by K. Yoshimura. To compare with these data, the stray light intensities of Al_mesh and Al_poly were measured for the same location as G-band's (i.e., 256x256 area around disk center for half-res images), then compare with the G-band stray light variation scaled to match the level of stray light intensities observed with Al_mesh and Al_poly filters.

However, it turned out the behavior of Al_mesh and Al_poly stray light is different from G-band's (Fig. 6.3 and 6.4). This suggests that there is other factor (than CCD contamination) that affects Al_mesh and Al_poly stray light intensity. One candidate of such factor is the difference in pointing. All the stray light image used for these plots were taken at the disk center for Hinode satellite, but there is an offset between Hinode's and XRT's pointing which varies with season. The time variation of the offset is generally large in eclipse seasons, according to K. Yoshimura who prepares the XRT co-alignment DB. While the stray light level is lower, the effect of other factors becomes larger.

So, it is therefore not appropriate to use the G-band stray light variation to simulate the time variation of Al_mesh and Al_poly stray light intensities.

[click image to enlarge]

Fig.6.3 : Al_mesh and G-band stray light variation.	Fig.6.4 : Al_poly and G-band stray light variation.

[Conclusion and my suggestion]
Our two attempts to determine the correction factor (the k-factor, or fudge factor) both failed. We should not provide the correction factor for the Al_mesh and Al_poly stray light correction. Rather, we should simply use a stray light image taken at the closest XRT pointing to data, then allow about 10% error for this correction.

7. G-band images analysis.

G-band VLS=open and close images have nearly equal intensity in the phase 3. Although sunspots (AR12662 and 12663) are barely seen by subtraction, but their intensity contrast is worse than that of contamination spots.

(click image to enlarge)

Comparison with the Phase 2 (Jan-2016).

(click image to enlarge)