Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://xmm.vilspa.esa.es/docs/documents/CAL-SRN-0177-1-2.ps.gz
Äàòà èçìåíåíèÿ: Wed Sep 15 12:35:01 2004
Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 22:30:08 2012
Êîäèðîâêà:
Ïîèñêîâûå ñëîâà: http astrokuban.info astrokuban
XMM-Newton CCF Release Note
XMM-CCF-REL-177
Low energy noise rejection re nement for pn.
M. Kirsch
September 15, 2004
1 CCF components
Name of CCF VALDATE List of Blocks
changed
CAL VERSION XSCS ag
EPN REJECT 0002 2000-01-01T00:00:00' CORRECTION VALUES 3.152 NO
2 Changes
The CCF contains the information how incorrect o set shifts for speci c pixels can be reconstructed
from the brightness of these pixels at 20 adu. This information was derived from correlating the
20 adu brightness with the corresponding value in the residual o set map. As the 20 adu brightness
is also in uenced by other e ects, an image containing the median of several 20 adu images was
subtracted before computing the correlation. This was the method used for deriving the values in
EPN REJECT 0001.CCF.
In the meantime it turned out that also the values in the residual o set maps are in uenced by
other e ects, similar to the situation with the 20 adu images. Subtracting the median of several
residual o set maps (which contain the temporally stable features) before correlating them with
the residual 20 adu images reduced the scatter and led to an improved correlation. This was the
motivation for the update of the CCF:
EPN REJECT 0001.CCF:
derived from correlation of (20 adu brightness - median 20 adu brightness) with residual o set
EPN REJECT 0002.CCF:
derived from correlation of (20 adu brightness - median 20 adu brightness) with (residual o set -
median residual o set)
1

XMM-Newton CCF Release XMM-CCF-REL-177 Page: 2
Figure 1 shows the original correlation, used for deriving the values in EPN REJECT 0001.CCF
kod ­ t.ps : t.ps ( ), 30­Apr­04 / 14:46:52, P 1
0 1 2 3 4
­4
­2
0
2
4
6
52
388
1407
5433
53208
134367
54125
3348
96
Q 0
energy
shift
to
be
applied
[adu]
excess counts at 20 adu per 1 ks per pixel
kod ­ t.ps : t.ps ( ), 30­Apr­04 / 14:47:13, P 1
0 1 2 3 4
­4
­2
0
2
4
6
12
71
491
3948
36048
87543
37304
2732
108
Q 1
(without CCD 6)
energy
shift
to
be
applied
[adu]
excess counts at 20 adu per 1 ks per pixel
kod ­ t.ps : t.ps ( ), 30­Apr­04 / 14:47:34, P 1
0 1 2 3 4
­4
­2
0
2
4
6
295
1177
2934
7625
51301
126241
53863
4694
528
Q 3
energy
shift
to
be
applied
[adu]
excess counts at 20 adu per 1 ks per pixel
kod ­ t.ps : t.ps ( ), 30­Apr­04 / 14:47:24, P 1
0 1 2 3 4
­4
­2
0
2
4
6
104
598
2027
5896
51820
131585
55034
3969
207
Q 2
energy
shift
to
be
applied
[adu]
excess counts at 20 adu per 1 ks per pixel
Figure 1: Pixel by pixel correlation between the excess counts at 20 adu, normalized to 1 ks, and the
values in the (uncleaned) o set map, for all quadrants (arranged according to their position on the detector).
While the energy shifts to be applied can only occur as integer multiples of 1 adu ( 5 eV), the excess counts
exhibit a much smoother distribution due to their normalization to 1 ks. Their distribution is illustrated
by the thickness of \histogram tubes", which were all expanded to have the same maximum thickness. The
number of pixels which they contain are given at right. This correlation was derived by combining the
observations of the Lockman hole in XMM revolutions 522, 523, 525, 526, 527, 528, and 544 ( 636 ks total
exposure time), all taken in fullframe mode. For Q 1, exclusion of CCD 6 was necessary in order to
get a better correlation. For Q 3, four noisy columns were excluded. A vertical red line indicates the
median value of the distributions. Shaded areas show the ranges of excess counts for which a certain energy
shift should be applied. Note how the brightness of pixels at 20 adu responds di erently in the four quadrants
to energy shifts. No such variation was found within the CCDs of the same quadrant. K. Dennerl (MPE)
Figure 2 shows the improved correlation, used for deriving the values in EPN REJECT 0002.CCF.

XMM-Newton CCF Release XMM-CCF-REL-177 Page: 3
kod ­ t.ps : t.ps ( ), 15­Sep­04 / 09:23:40, P 1
0 1 2 3 4 5 6 7 8
­4
­2
0
2
4
6
34
247
949
3612
31630
139379
65781
10454
517
Q 0
energy
shift
to
be
applied
[adu]
excess counts at 20 adu per 1 ks per pixel
kod ­ t.ps : t.ps ( ), 15­Sep­04 / 09:24:02, P 1
0 1 2 3 4 5 6 7 8
­4
­2
0
2
4
6
15
87
615
4227
31722
133847
68355
12899
801
Q 1
energy
shift
to
be
applied
[adu]
excess counts at 20 adu per 1 ks per pixel
kod ­ t.ps : t.ps ( ), 15­Sep­04 / 09:24:56, P 1
0 1 2 3 4 5 6 7 8
­4
­2
0
2
4
6
100
527
1635
4614
30064
132497
65376
11695
733
Q 3
energy
shift
to
be
applied
[adu]
excess counts at 20 adu per 1 ks per pixel
kod ­ t.ps : t.ps ( ), 15­Sep­04 / 09:24:19, P 1
0 1 2 3 4 5 6 7 8
­4
­2
0
2
4
6
52
374
1358
4211
30754
137467
66631
11044
600
Q 2
energy
shift
to
be
applied
[adu]
excess counts at 20 adu per 1 ks per pixel
Figure 2: Pixel by pixel correlation between the excess counts at 20 adu, normalized to 1 ks, and the values
in the cleaned o set map, for all quadrants (arranged according to their position on the detector). While the
energy shifts to be applied can only occur as integer multiples of 1 adu ( 5 eV), the excess counts exhibit
a much smoother distribution due to their normalization to 1 ks. Their distribution is illustrated by the
thickness of \histogram tubes", which were all expanded to have the same maximum thickness. The number
of pixels which they contain are given at right. This correlation was derived by combining the observations
of the Lockman hole in XMM revolutions 522, 523, 525, 526, 527, 528, and 544 ( 636 ks total exposure
time), all taken in fullframe mode. For Q 3, four noisy columns were excluded. No exclusion of CCD 6
was necessary. A vertical red line indicates the median value of the distributions. Shaded areas show the
ranges of excess counts for which a certain energy shift should be applied. Note how the brightness of pixels
at 20 adu responds di erently in the four quadrants to energy shifts. No such variation was found within the
CCDs of the same quadrant. From Dennerl et al., Proc. SPIE 2004 (Glasgow)
More information can be found in Dennerl et al., Proc. SPIE 2004, which describes speci cally
how EPN REJECT 0002.CCF was obtained.
3 Scienti c Impact of this Update
The di erences to the results obtained with EPN REJECT 0001.CCF are very subtle, however treat
the behavior of the variation of the residual o set maps in the correct way.

XMM-Newton CCF Release XMM-CCF-REL-177 Page: 4
4 Estimated Scienti c Quality
The o set correction recovers errors of up to several 10 eV for point sources that happen to fall on
a bright patch caused by a wrong o set value.
The reduction of the detector noise allows to extend the useful energy range down to an instru-
mental energy of 120 eV. Note that Figure 3 is not showing the di erence between images processed
with EPN REJECT 0001.CCF and EPN REJECT 0002.CCF, but the di erences between the original im-
age and the image after correcting the o set shifts with EPN REJECT 0002.CCF. A comparison of
EPN REJECT 0001.CCF and EPN REJECT 0002.CCF in images will not show any obvious di erences.
kod ­ ... : check0.ps ( ), 19­Aug­04 / 10:30:26, P 1
­3 ­2 ­1 0 1 2
cm
­3
­2
­1
0
1
2
cm
­15' ­10' ­5' 0' 5' 10'
­10'
­5'
0'
5'
10'
1 1
64
2 1
64
1
64
Q
0
0
199
4
64 5
64 6
64 1
Q
1
0
199
7
1 64 8
1 64 9
1 64
Q
2
0
199
10 64
11 64
12 64
1
Q
3
0
199
0 5 10 15 20 25 30 35 40 45 50
­5000
0
5000
10000
15000
XMM / EPIC pn pn0367ff_med_*_map.fits
kod ­ ... : check0.ps ( ), 19­Aug­04 / 10:12:36, P 1
­3 ­2 ­1 0 1 2
cm
­3
­2
­1
0
1
2
cm
­15' ­10' ­5' 0' 5' 10'
­10'
­5'
0'
5'
10'
1 1
64
2
64
3 1
64
Q
0
0
199
64 1 5
64 1 6
64
Q
1
0
199
1 64 8 64 9
1 64
Q
2
0
199
10 64
1
11 64
1
12 64
Q
3
0
199
0 5 10 15 20 25 30 35 40 45 50
­5000
0
5000
10000 15000
XMM / EPIC pn pn0367ff_med_*_map.fits
Figure 3: Images of all events with 20 adu in the data set 0367 0137550901. Left: original image, right: image
after correcting the o set shifts with EPN REJECT 0002.CCF. Black pixels are caused by the fact that the o sets applied
on{board were higher than 20 adu, so that there are no 20 adu data available. This problem diappears above  22 adu.
5 Test procedures & results
Tests were performed with the data set 0367 0137550901, a 30 ks observation of the Vela SNR in
FF mode. In general, the di erences to the results obtained with EPN REJECT 0001.CCF are very
subtle.

XMM-Newton CCF Release XMM-CCF-REL-177 Page: 5
6 Expected Updates
The rst part of this task (the energy shift correction) is particularly important for observations
where the o set map was not transmitted. For more recent observations, where the corresponding
o set map is available, the o set shifts can be directly derived from the o set map. That is not yet
implemented.