Provisional method to analyze the XIS data with SCI
2007-04-10 by the XIS team
In the current pipeline products (Ver. 1.x), pulse invariant (PI)
values of the data with the Spaced-row Charge Injection (SCI) are not
correct. Here we introduce a provisional method to calculate the PI
values with reasonable accuracy for the SCI data.
The PI values are calculated in the current processing using the
Charge Transfer Inefficiency (CTI) appropriate for the no-SCI data
regardless of the SCI usage. This introduces a systematic error
in the PI values
of the SCI data. The XIS team is now preparing tools to process
the SCI
data correctly, which will be used from the Ver. 2.0 processing.
However, some observers may want to analyze the SCI data as soon as
possible. Thus we introduce here a provisional method to analyze the
SCI data in addition to the calculation of PIs with reasonable
accuracy. However, this method is restricted to the FI data
(XIS0,2,3) in normal mode without any options. We would like to
ask the users to use this method carefully with an understanding
of the
systematics associated with it.
By analyzing some data taken for the SCI calibration, we have
found that
the CTI is almost zero when the SCI is applied. In fact, a PI
calculation assuming that the CTI is zero gives an energy scale with
systematic errors of < 30 eV in the iron band, except for the BI CCD
(XIS1). Hence, if you want to analyze the SCI data with a reasonable
energy scale, please reprocess them using the zero-CTI CALDB
files, as
explained below.
Note:
The principle of the SCI is described in the Suzaku Technical
Description (http://www.astro.isas.jaxa.jp/suzaku/doc/suzaku_td/).
In addition, the availability and the restrictions of the SCI are
summarized in
http://www.astro.isas.jaxa.jp/suzaku/planning/xismode/
This document consists of three subsections below. First, we
summarize systematic errors of the SCI data with the zero-CTI
correction. The energy scale and resolution are described in detail.
After the explanation of reprocessing the SCI data, we show how to
create the XIS Redistribution Matrix File (RMF) and Auxiliary
Response
File (ARF) files for the spectral analysis.
1. Systematic errors of the SCI data
with the zero-CTI correction
Systematic errors of the SCI data with the zero-CTI correction are
summarized below.
| Energy scale |
< 30 eV at 6 keV.
|
| Energy resolution |
The width of a narrow line can be reproduced relatively well, if
you use
an RMF created with the method in
Section 3. The excess
width would be much smaller than 20 eV at 6 keV.
|
| Effective area |
Additional systematic errors due to the SCI are considered to be
negligible, if the SCI and adjacent rows are removed
(Section 2) and
corresponding ARF created with "xissimarfgen"
(Section 3) is used.
|
| Quantum efficiency |
Because the zero-CTI is just an approximation, it may affect the
quantum detection efficiency of the data through the grade
classification. However, the systematic error is considered to be
negligibly small unless the data statistics are extremely good.
|
Energy scale and resolution
Figure 1 shows a time history (2006 August - 2007 February) of the
apparent center energy of Mn Kalpha line from the 55Fe calibration
source, which illuminate corners of the XIS0 and XIS1, segment A.
The
data we used here are publicly available from DARTS/Suzaku (http://
www.darts.isas.jaxa.jp/astro/suzaku/).
However, the PIs are recalculated with the provisional method
described in Section 2 below. Each mark in the plots corresponds
to an
observation sequence. While the XIS1 shows a systematically lower
energy scale of ~30 eV, the line center energies of the XIS0 are
consistent with 5.895 keV (dashed-lines in the figures) within ~10
eV.
Note that the line center energy becomes lower with time even if the
SCI is applied to observations, although the energy decrease is
slower
than that without the SCI. The time dependence of the energy
scale with
the SCI and the cause of the lower energy scale of the XIS1 are now
under investigation. Full treatment of the SCI data including these
issues will be available in the Ver. 2.0 processing. New ftools and
CALDB files for the Ver. 2.0 processing will be released in June.
Once the Ver. 2.0 processing is started, whole the Suzaku archive
will
be reprocessed with appropriate CTI parameters.
 |
 |
|
Figure 1: Time history of the apparent center energy of Mn Kalpha
line. The PIs are recalculated with the zero-CTI CALDB files.
|
|---|
Figure 2 shows apparent line widths of Mn Kalpha obtained with the
RMF
that was determined by the ground calibration. Excess widths (sigma)
of ~20 eV for the XIS0 and ~35 eV for the XIS1 are clear from the
figure. Because the RMF has a resolution (FWHM) of ~135 eV at this
band, the energy resolutions of the XIS0 and XIS1 with the SCI are
estimated to be ~140 eV and ~155 eV, respectively. For the FI
sensors
(XIS0,2,3), the excess widths would be 0 eV at 6 keV, if you use
an RMF
with the resolution on 2005-08-13 (XIS door was opened). A method to
create an RMF is explained in Section 3.
 |
 |
|
Figure 2: Time history of the apparent line width of Mn Kalpha line.
The line widths are obtained with the RMF based on the ground
calibration. The excess widths would be much smaller than 20 eV at 6
keV, if the RMF created with the method in Section 3 is used.
|
|---|
Since charge loss due to the CTI depends on the distance from the
readout nodes, i.e., the number of charge transfers, we investigated
the ACTY dependence of the energy scale in the iron band. Figure
3 shows
the center energies of Fe XXV Kalpha emission from the Perseus
cluster
observed at the end of August, 2006. We divided the Segment B of
XIS0
into six regions and derived the center energy at each region. It
can
be seen that the ACTY dependence of the CTI with SCI is smaller than
that without SCI. For Segment B and C of all sensors, we examined
such ACTY dependences and fit the data by linear functions. The
slopes are summarized in the Table 1. Combined with the systematic
error of ~10 eV derived from the calibration sources, we now conclude
that a systematic error of the energy scale due to the zero-CTI
correction is probably lower than 30 eV at the center of the XIS FOV.
|
| Figure 3: ACTY dependence of the center energy
of Fe XXV Kalpha emission |
|---|
Table 1: The ACTY dependences of the energy scale due
to the CTI
| | Slope [eV/pix] |
| Segment \ Detector | XIS0 |
XIS1 |
XIS2 | XIS3 |
| Segment B | -(26+/-15)E-3 | -(2+/-18)E-3 |
-(11+/-11)E-3 | -(34+/-15)E-3 |
| Segment C | -(17+/-7)E-3 | -(21+/-23)E-3 |
-(31+/-17)E-3 | -(13+/-9)E-3 |
From the preliminary analysis of E0102.2-7219, the improvement
of the
energy resolution is also verified at low energies. However, further
calibration at this band including the energy scale is still in
progress.
2. How to reprocess the SCI data with the
zero-CTI CALDB files
The PI values can be calculated with the tool "xispi" using the
zero-CTI CALDB files. The recalculation method is summarized below.
Recent versions (HEASOFT v6.2 or later) of "xispi" can be applied
directly to the cleaned event files. However, older versions can be
applied only to the unfiltered event files. Thus, we also describe
how to apply it to the unfiltered data and to screen them to create
the cleaned event files; this is basically the same procedure used in
the pipeline processing.
This is an example of the OBS_ID=501008010 observation.
The files required in the reprocessing are as follows;
501008010/auxil/ae501008010.ehk.gz (Extended House Keeping file)
501008010/auxil/ae501008010.hk.gz (Common House Keeping file)
501008010/xis/event_cl/ae501008010xi[0-3]_0_[35]x[35]n066a_cl.evt.gz
(Cleaned event files)
501008010/xis/event_uf/ae501008010xi[0-3]_?_[35]x[35]3n066z_uf.evt.gz
(Unfiltered event files)
501008010/xis/hk/ae501008010xi[0-3]_0.hk.gz
(XIS House Keeping files)
In the explanation below, these files are assumed to be linked
appropriately to your working directory.
A. Correction to the cleaned event files:
(Note: this is possible with the HEASOFT v6.2 or later versions of
"xispi".)
1. Make the cleaned event file only with EVENTS and GTI extensions.
unix% xselect
xsel:SUZAKU > set datadir .
xsel:SUZAKU > read event ae501008010xi0_0_3x3n066a_cl.evt.gz
xsel:SUZAKU-XIS0-STANDARD > extract event
xsel:SUZAKU-XIS0-STANDARD > save event ae501008010xi0_0_3x3n066a_cl2.evt
>Use filtered events as input data file ? >[no] no
xsel:SUZAKU-XIS0-STANDARD > exit
>Save this session? >[no]
2. Recalculate the grade and the PI values by explicitly specifying
the CALDB files in which the parallel CTI is zero. The CALDB
files
are labelled as ae_xi[0-3]_makepi_20060522_nocti.fits. You can
download these files in the following URL:
http://www.astro.isas.jaxa.jp/suzaku/analysis/xis/sci/zerocti/
unix% xispi infile=ae501008010xi0_0_3x3n066a_cl2.evt \
outfile=ae501008010xi0_0_3x3n066a_nocti.evt \
hkfile=ae501008010xi0_0.hk.gz \
makepifile=ae_xi0_makepi_20060522_nocti.fits
The house keeping (HK) files are in the "xis/hk" subdirectory.
3. Next, filter out the SCI and its previous/next rows in the cleaned
event file to remove fake events created by the SCI. We prepare a
convenient shell script, "cut3rows.csh", instead of a direct input
of selection criteria in xselect. The script can be downloaded at
the following URL:
http://www.astro.isas.jaxa.jp/suzaku/analysis/xis/sci/zerocti/
unix% ./cut3rows.csh \
ae501008010xi0_0_3x3n066a_nocti.evt \
ae501008010xi0_0_3x3n066a_cut3row_nocti.evt
The SCI rows are currently defined as follows; ACTY = 52, 106, 160,
214, 268, 322, 376, 430, 484, 538, 592, 646, 700, 754, 808, 862, 916,
970, 1022. Then you can filter out these rows by directly specifying
the ACTY columns.
unix% xselect
xsel:SUZAKU > set datadir .
xsel:SUZAKU > read event ae501008010xi0_0_3x3n066a_nocti.evt
xsel:SUZAKU-XIS0-STANDARD > filter column "ACTY=0:50 54:104 108:158 162:212 216:266"
xsel:SUZAKU-XIS0-STANDARD > extract event
xsel:SUZAKU-XIS0-STANDARD > save event ae501008010xi0_0_3x3n066a_nocti.evt_cuttmp1
>Use filtered events as input data file ? >[no] yes
xsel:SUZAKU-XIS0-STANDARD > filter column "ACTY=270:320 324:374 378:428 432:482 486:536"
xsel:SUZAKU-XIS0-STANDARD > extract event
xsel:SUZAKU-XIS0-STANDARD > save event ae501008010xi0_0_3x3n066a_nocti.evt_cuttmp2
>Use filtered events as input data file ? >[no] yes
xsel:SUZAKU-XIS0-STANDARD > filter column "ACTY=540:590 594:644 648:698 702:752 756:806"
xsel:SUZAKU-XIS0-STANDARD > extract event
xsel:SUZAKU-XIS0-STANDARD > save event ae501008010xi0_0_3x3n066a_nocti.evt_cuttmp3
>Use filtered events as input data file ? >[no] yes
xsel:SUZAKU-XIS0-STANDARD > filter column "ACTY=810:860 864:914 918:968 972:1020"
xsel:SUZAKU-XIS0-STANDARD > extract event
xsel:SUZAKU-XIS0-STANDARD > save event ae501008010xi0_0_3x3n066a_cut3row_nocti.evt
>Use filtered events as input data file ? >[no] no
xsel:SUZAKU-XIS0-STANDARD > exit
>Save this session? >[no]
B. Correction to the unscreened event files:
(Note: this is a bit cumbersome. We recommend the update to the
latest version of "xispi")
1. Recalculate the grade and PI values with the zero-CTI CALDB files.
unix% xispi infile=ae501008010xi0_1_3x3n066z_uf.evt.gz \
trcor_caldbfile=ae_xi0_makepi_20060522_nocti.fits \
cticor_caldbfile=ae_xi0_makepi_20060522_nocti.fits \
grade_caldbfile=ae_xi0_makepi_20060522_nocti.fits \
pha2pi_hkfile=ae501008010xi0_0.hk.gz \
pha2pi_caldbfile=ae_xi0_makepi_20060522_nocti.fits \
outfile=ae501008010xi0_1_3x3n066z_uf_nocti.evt
The unfiltered event files are in the "xis/event_uf" subdirectory.
2. Screen with xselect (filter column), using GRADE and STATUS
information.
unix% xselect
xsel:SUZAKU > set datadir .
xsel:SUZAKU > read event ae501008010xi0_1_3x3n066z_uf_nocti.evt
xsel:SUZAKU-XIS0-STANDARD > filter column "GRADE=0:0 2:4 6:6 STATUS=0:65535"
Calibration source areas are not removed in the pipeline
processing. You are requested to do this yourself if needed. If
you are
using xselect, substitute "GRADE=0:0 2:4 6:6 STATUS=0:65535"
for the
above criterion.
3. Screen with xselect (select hk), using common HK, to discard the
data during maneuver.
xsel:SUZAKU-XIS0-STANDARD > read hk hkfiles=ae501008010.hk.gz expand=yes
>HK file directory >[./]
xsel:SUZAKU-XIS0-STANDARD > select hk "AOCU_HK_CNT3_NML_P==1"
xsel:SUZAKU-XIS0-STANDARD > extract event
xsel:SUZAKU-XIS0-STANDARD > save event ae501008010xi0_1_3x3n066z_uf_nocti.evt_gsmnvcut
>Use filtered events as input data file ? >[no] yes
The common HK file is in "auxil" subdirectory.
4. Screen with xselect (select hk), using XIS HK, to discard the data
during data rate low. The criteria are given by "Sn_DTRATE < 3",
where n=0, 1, 2, 3, depending on the sensor ID.
xsel:SUZAKU-XIS0-STANDARD > read hk hkfiles=ae501008010xi0_0.hk.gz expand=yes
>HK file directory >[./]
xsel:SUZAKU-XIS0-STANDARD > select hk "S0_DTRATE<3"
xsel:SUZAKU-XIS0-STANDARD > extract event
xsel:SUZAKU-XIS0-STANDARD > save event ae501008010xi0_1_3x3n066z_uf_nocti.evt_dtratecut
>Use filtered events as input data file ? >[no] yes
5. Screen with xselect (select hk), using extended house keeping
(EHK)
file, to discard the data during SAA.
xsel:SUZAKU-XIS0-STANDARD > read hk hkfiles=ae501008010.ehk.gz expand=yes
>HK file directory >[./]
xsel:SUZAKU-XIS0-STANDARD > select hk "SAA_HXD==0 && T_SAA_HXD>436 && ANG_DIST<1.5"
xsel:SUZAKU-XIS0-STANDARD > extract event
xsel:SUZAKU-XIS0-STANDARD > save event ae501008010xi0_1_3x3n066z_uf_nocti.evt_saacut
>Use filtered events as input data file ? >[no] yes
The EHK file is in "auxil" subdirectory.
6. Cleaned events are then selected with the following criteria:
xsel:SUZAKU-XIS0-STANDARD > read hk hkfiles=ae501008010.ehk.gz expand=yes
>HK file directory >[./]
xsel:SUZAKU-XIS0-STANDARD > select hk "ELV>5 && DYE_ELV>20"
xsel:SUZAKU-XIS0-STANDARD > extract event
xsel:SUZAKU-XIS0-STANDARD > save event ae501008010xi0_1_3x3n066z_uf_nocti.evt_elvcut
>Use filtered events as input data file ? >[no] no
xsel:SUZAKU-XIS0-STANDARD > exit
>Save this session? >[no]
8. Remove hot/flickering pixels by applying "cleansis".
unix% cleansis datafile=ae501008010xi0_1_3x3n066z_uf_nocti.evt_elvcut \
outfile=ae501008010xi0_1_3x3n066z_nocti.evt \
cellsize=5 logprob=-.53E+01 bthresh=3 \
phamin=0 phamax=4095 chipcol=SEGMENT
9. Finally, as explained in the step (A-3), filter out the SCI row
and
its previous/next rows in the cleaned event file.
3. How to create the RMF and ARF files
The energy resolution is improved from ~200-220 eV to ~140-150 eV at
the iron band with the application of the SCI. The improved
resolution is almost the same as that just after the XIS door opened.
Therefore, in order to create the RMF for the SCI data by
"xisrmfgen",
the RMF generator in HEASOFT v6.2, set the observation date on
2005-08-13T00:00:00 as follows;
unix% xisrmfgen \
phafile=NONE \
instrume=XIS0 \
clk_mode=normal \
date_obs=2005-08-13T00:00:00 \
ebinfile=none \
quantefffile=ae_xi0_quanteff_20060825.fits \
rmfparamfile=ae_xi0_rmfparam_20061024.fits \
makepifile=ae_xi0_makepi_20060522_nocti.fits \
outfile=xis0_with_SCI.rmf
As explained in the step (A-2), the SCI and its adjacent rows are
removed as dead lines. The dead lines correspond to 5.6% of the
image. This ratio is changed by the source-extraction region.
Therefore, if you want to estimate an X-ray flux of a source more
precisely, you need to exclude these lines in the ARF
calculation. It
is easy to prepare an image FITS file of the source-extraction
region,
in which the dead lines are already removed, and to give it to an
input parameter "regfileN". Note that the image must be created in
DETX/Y coordinates since the SCI rows are parallel to RAWX. The ARF
for the SCI data is created as below. Users should refer to the
paper
by Ishisaki et al. (PASJ, 59, SP1, S113 (2007), astro-ph/0610118) to
understand the proper usage of this tool.
1. Create image FITS files of a region from which a source spectrum
is extracted.
unix% xselect
xsel:SUZAKU > set datadir .
xsel:SUZAKU > read event ae501008010xi0_0_3x3n066a_cut3row_nocti.evt
xsel:SUZAKU-XIS0-STANDARD > set xyname DETX DETY
xsel:SUZAKU-XIS0-STANDARD > set xybinsize 1
xsel:SUZAKU-XIS0-STANDARD > filter region xis0_source_detxy.reg
xsel:SUZAKU-XIS0-STANDARD > extract image
xsel:SUZAKU-XIS0-STANDARD > save image xis0_source_detxy.img
xsel:SUZAKU-XIS0-STANDARD > set xyname X Y
xsel:SUZAKU-XIS0-STANDARD > set xybinsize 1
xsel:SUZAKU-XIS0-STANDARD > filter region xis0_source_skyxy.reg
xsel:SUZAKU-XIS0-STANDARD > extract image
xsel:SUZAKU-XIS0-STANDARD > save image xis0_source_skyxy.img
xsel:SUZAKU-XIS0-STANDARD > exit
>Save this session? >[no]
2. Create mask images by "ftimgcalc", an image manipulator in HEASOFT
v6.2. When creating a mask image of the detector, you need to
prepare a region file excluding the SCI and its adjacent rows.
Attach
a corresponding region file of the dead lines to your source region
file. The region files can be downloaded at the following URL:
http://www.astro.isas.jaxa.jp/suzaku/analysis/xis/sci/zerocti/
Detector
unix% cat xis0_source_detxy.reg xis03_scirows_detxy.reg \
>xis0_source_detxy_cut3row.reg
unix% ftimgcalc \
outfile=xis0_source_cut3row.img \
expr='regfilter("xis0_source_detxy_cut3row.reg", A.P1, A.P2) ? (1) : (0)' \
a=xis0_source_detxy.img
Sky (Uniform sky is assumed here)
unix% ftimgcalc \
outfile=xis0_source_uniform_sky.img \
expr='regfilter("xis0_source_skyxy.reg", A.P1, A.P2) ? (1) : (0)' \
a=xis0_source_skyxy.img
3. Run "xissimarfgen" to create the ARF.
unix% xissimarfgen \
instrume=XIS0 \
source_mode=SKYFITS \
source_image=xis0_source_uniform_sky.img \
num_region=1 \
region_mode=DETFITS \
regfile1=xis0_source_detxy_cut3row.img \
arffile1=xis0_with_SCI.arf \
limit_mode=NUM_PHOTON num_photon=400000 \
accuracy=0.005 \
phafile=none \
detmask=none \
gtifile=none \
date_obs=2006-09-26T14:18:16 \
attitude=none \
ea1=266.4995 ea2=119.1678 ea3=184.9994 \
rmffile=xis0_with_SCI.rmf \
estepfile=default
ISAS/JAXA Department of High Energy Astrophysics
Last Modified: Wednesday, 04-Jul-2007 12:46:32 JST
