next up previous contents
Next: 7. HXD Data Analysis Up: Suzaku ABC Guide Previous: 5. ``README FIRST" for   Contents


6. XIS Data Analysis

6.1 Introduction

The XIS consists of four CCD detectors, three of which are ``front-illuminated'' (FI) and one ``back-illuminated'' (BI). The BI chip has an increased quantum efficiency at low ($< 1$keV) energies with a small decrease at higher energies. Although the detectors have seen significant improvements since the ASCA SIS, the data reduction is quite similar to that of ASCA SIS and Chandra ACIS.

Users should familiarize themselves with the current issues with the XIS and XIS analysis (the loss of XIS2, the XIS0 anomaly, SCI, energy scale of non-SCI data, energy scale of SCI data, contamination, pile-up, timing mode, and others) in chapter 5.

6.2 Cleaned Event File Content

XIS data begin as part of the RPT telemetry downloaded from Suzaku, and are converted into a collection of FITS files by the mk1stfits routine at ISAS. mk1stfits does not reject any events or apply any calibration to the data but merely converts RPT into FITS files. Once the files have been processed through the pipeline, they are included in the standard data download in the directory xis/event_uf.

Table 6.1: XIS calibration steps.
Calibration Item Tool Comments
Assign Arrival Times xistime Not for P-sum mode
Calibrate Sky Coordinates xiscoord Attitude wobble is now modeled
Assigning Pixel Quality xisputpixelquality  
Compute PI xispi Works for data with and without SCI
Compute GTIs xisgtigen Can be used to monitor telemery saturation
Add Microcode xisucode Updates header information about the mode

The XIS mk2ndfits pipeline task is then run on the mk1stfits output to create filtered, calibrated output event files, which are placed in the event_cl subdirectory. The calibration steps are summarized in Table 6.1. If updates of any of these tools became available since the initial processing of a dataset, the user can reprocess the data, see section 6.3 and 6.4 for details. A recalculation that is necessary in most cases in order to obtain the best available calibration, is the update of the PI energy scale with the newest calibration using xispi (next section).

In addition to the above calibration steps, the event files in the event_cl have been screened using two broad classes of screening, event by event and by good-time intervals (GTI). The former includes event grade, which encodes the pattern of charge distribution among neighboring pixels and can be used to distinguish between X-ray and charged particle events. The GTI screening is used to select time intervals where the instrument is stably pointed at the source without being blocked by the Earth, and to exclude high background intervals. Both classes of screening are applied by the processing pipeline.

The Version 2 pipeline screening criteria are summarized in Table 6.2. These are also the screening criteria applied by the tool xisrepro, see next section.

Table 6.2: XIS screening criteria.
Type Criterion Comments
Event by event GRADE=0:0 2:4 6:6 ASCA grades indicating X-ray events
  STATUS=0:524287 Bad columns, charge injection rows removed
  cleansis Flickering pixels are removed
GTI AOCU_HK_CNT3_NML_P==1 Attitude control in pointing mode
  ANG_DIST$<$1.5 Instantaneous pointing within 1.5 arcmin of mean.
  Sn_DRATE$<$3 Telemetry rate SuperHigh, High, or Medium
  SAA_HXD==0 Satellite is outside SAA
  T_SAA_HXD$>$436 Time since SAA passage $>$436 sec
  ELV$>$5 Pointing direction $>$5 deg above Earth
  DYE_ELV$>$20 Pointing direction $>$20 deg above sunlit limb of Earth

6.3 Energy Scale Reprocessing (Xispi) and Screening using Xisrepro

The XIS team regularly updates the CALDB files (e.g., ae_xiN_makepi_20090615.fits), which are used by xispi which calculates the PI values. The files include time-dependent CTI parameters and thus enable us to correct the decrease in the gain after 2006 September.

The processing pipeline V2.1.6.16 is the first version to use these revised makepi files. Users should check the pipeline version used for processing their data by reading the PROCVER keyword in their data files. See chapter 5 for a history of makepi file releases. Significant makepi updates will continue to be annouced on the Suzaku GOF web pages.

For window mode data only the Suzaku FTOOLS Version 12 xispi or later should be used to perform the reprocessing, see for details.

The Version 2 processing pipeline used older versions of the makepi files up to V2.1.5.15. For SCI-on XIS data taken 2006 September, this resulted in the gain of Mn K alpha calibration line decreasing at a rate of about 30 eV/year in the FI chips -- see

Users can reprocess their own SCI-on data (processing Version 2.x) as follows.

First, run xispi to recalculate the PI values. Note that the XIS HK files are in the xis/hk subdirectory.

> xispi infile=ae101005070xi0_0_3x3n066z_uf.evt.gz \
        outfile=ae101005070xi0_0_3x3n066z_uf_new.evt \
The hidden paramter makepifile should be set to CALDB if accessing the latest CALDB, or explicitly specify the latest ``makepi'' file.

Since the grade determination is based on the CTI-corrected pulse height values in the PHA column, users should reprocess starting with the unfiltered event files. Once all unfiltered event files are reprocessed with xispi, they must be screened. For convenience, we provide an xselect script, which references the event selection criterion file and the standard good-time interval selection file Users should download these three files into the current working directory, make sure the filter file (.mkf in ../../auxil directory) is uncompressed if using Suzaku FTOOLS Version 13 or earlier, read in the updated unscreened event file(s) into xselect, and then type

xsel > @xisrepro
This will cause xselect to apply the event selection, remove flickering pixels (using cleansis), and apply the standard GTI selection. If desired, users can edit xis_mkf.sel to adjust the screening criteria. Xselect will pause and ask the users to give the output (screened) event file name.

If there are more than one unfiltered event file, it is recommended to reprocess them all individually and then combine the reprocessed files in the screening step.

6.4 Full XIS Reprocessing and Screening using Aepipeline

The powerful FTOOL aepipeline (available since Suzaku FTOOLS version 14, i.e., HEAsoft v.6.8) duplicates the Suzaku processing pipeline. It allows the user to run all or part of the pipeline for XIS (detector selection available) and/or HXD (detector selection available) and to vary the calibration files used. A number of other pipeline processing parameters can also be changed. Please, refer to the help text available, e.g., by typing ``fhelp aepipeline''.

The pipeline performs calibration - for the XIS this includes running xisgtigen, xistime, xiscoord, xisputpixelquality, and xispi - as well as data screening - with defaults as given in Table 6.2 for the XIS. The screening criteria can be modified by changing the parameters xis0/1/2/3_expr from their default values.

The GTI file for telemetry un-saturated times created by hxdgtigen (aeNNNNNNNNNxis_0_tlm.gti) is applied by aepipeline.

The following is an example of a simple call of aepipeline, performing recalibration (stage 1) and rescreening (stage 2) of the XIS data for a given observation or sequence number (here: 403049010) applying the default calibration and the default screening criteria.

> aepipeline indir=/scratch/403049010 \
outdir=/scratch/reprocessed/xis \
steminputs=ae403049010 entry_stage=1 \
exit_stage=2 clobber=yes instrument=XIS


(1) The outdir directory may not be a subdirectory of the indir directory. If it is, wrong results can be produced (double counting of events), especially in the case of multiple reprocessing runs. While the reporting will soon be improved, this currently happens without warning or error messages.

(2) The header keyword in the event files indicating the version of the processing pipeline used for reprocessing ``PROCVER'' is not updated by aepipeline. Check the log file produced by aepipeline for pipeline version information.

6.5 Extracting Products

The primary tool for extracting data products (spectra, lightcurves, exposure maps) from XIS data is xselect, which is part of the general HEAsoft distribution. xselect can apply filters which select user-defined times, sky regions, or particular event flags. It then uses the filtered events to create a (binned) spectrum (as well as generating the necessary calibration files), a lightcurve, or an exposure map. Basic parameters commonly used for common data screening are in the filter, or mkf, file. See section 3.7 for a list of xselect default settings for Suzaku.

6.5.1 Combining 3x3 and 5x5 Editing Mode Data

The editing mode (3x3, 5x5, or 2x2) is chosen by the operations team for every interval of an observation, based on the available telemetry (see section 3.6.3). The observed data are sorted by editing mode and stored in separate files, with the mode indicated in the file names. The event files from all available modes need to be combined in order to obtain the full exposure. For the 3x3 and 5x5 modes this can be done in a straight forward way by reading the event files from both modes into xselect together and then proceeding with the filtering and extraction of products (images, lightcurves, spectra) in the same way as for a single event file. In the standard product extraction, only the central 3x3 pixel information is used even for the 5x5 event data, so there is no qualitative difference between the 3x3 and 5x5 event data. Some observations of bright sources contain 2x2 event data, however. Products from those should be extracted and analyzed separately, at least for the spectral analysis, since these event data do not contain a part of the pixel information useful for refining the photon energy determination.

6.5.2 Additional Screening

Additional filtering can be applied to the screened data at this stage using the ``select mkf'' command. Xselect assumes that the filter file is located in ../../auxil relative to the current working directory, with the file name *.mkf. Users who prefer to work under a different directory can use the set mkfdir command to change the location of the filter file. Note that if aepipeline is used to reprocess the data (section 6.4), a copy of the .mkf file is placed in the output directory of that tool. With the default set-up it has a file name ending with .mkf and for Suzaku FTOOLS Version 13 and earlier it is necessary to uncompress it.

Additional filtering could include applying a more strict version of the screening already applied in pipeline processing. For example, some observations may be more sensitive to the effects of solar X-rays scattered from the sunlit Earth. In this case, users may want to experiment applying DYE_ELV$>$25 to the data and see if it makes a difference.

Another item that affects the particle background rate is geomagnetic cut-off rigidity (COR, which is in the mkf file; a slightly updated version, COR2, is currently available only in the ehk file). For example, applying the criterion ``COR$>$6'' can reduce the effective exposure time somewhat but may improve the signal-to-noise ratio.

One final type of screening concerns telemetry saturation. It is not expected that this is a major issue in the majority of observations, as long as data obtained with a low telemetry rate setting are excluded, hence the pipeline does not apply GTIs based on non-saturation of telemetry. However, GTI files for intervals of unsaturated telemetry are available in the xis/hk subdirectory with filename ending in _tel_uf.gti, and these can be applied by using the ``select time file'' command in xselect.

In general, users are encouraged to explore the effects of different values for all the cuts and selections described above on their own dataset by making lightcurves of mkf parameters.

6.5.3 Relaxing Screening Criteria

In case the user wants to relax some of the screening criteria, this can be done by starting from the unfiltered data and applying a different set of screening criteria than the standard one, e.g., using the xisrepro tool, see section 6.3.

6.5.4 Region Selection

For a point source, circular extraction regions centered on the source should be used to extract source spectra and light curves. Generally an extraction radius of 260 arcsec (250 pixels), which encircles 99% of the point source flux, is recommended. Recent software and calibration releases, however, allow radii as small as 60 arcsec (Suzaku FTOOLS verion 8 and CALDB release 2008-07-09, or later). As the vignetting is relatively small, a large fraction of the remainder of the XIS chip is in principle available for background subtraction. Sky Regions

Users generally want to extract light curves or energy spectra from specific regions on the sky. Such region selections can be done in the ``SKY'' image displayed by ds9/saoimage. Select a region and create a region file. This file can then be applied by the xselect ``filter region'' command. Note that sky coordinates are the default image coordinates in xselect. After using other coordinates enter ``set image sky'' to go back to sky coordinates. Detector Regions

Particular regions within a single detector can also be selected using detector coordinates. Since the default image coordinates in xselect are sky coordinates, enter ``set image det'' to switch to detector coordinates before extracting images. While detector coordinates are defined so that all the XIS images have the same direction (§3.3), the four XIS sensors on the baseplate are rotated by $90^{\circ}$ or $180^{\circ}$ relative to each other. The ACT coordinates are the actual location on the CCD chip, which may be useful when investigating instrumental characteristics at particular chip positions (such as extracting the calibration source spectra). Entering ``set image raw'' followed by ``extract image'' will extract XIS ACT images. XIS performance is dependent on segments, and particular segments may be selected with the ``select event'' command. Events on segment A, B, C, and D have ``SEGMENT'' column values 0, 1, 2, 3, and 4, respectively.

6.5.5 Calibration Source Locations

The two far-end corners from the read-out node (i.e., those with large ACTY coordinates) of each sensor are illuminated by the $^{55}$Fe calibration sources. The illuminated areas are roughly sketched on the two-page XIS instrument summary that can be found at tsujimot/pg_xis.pdf. In addition, XIS images for events with PI values between 1500 and 1800 clearly show the illuminated areas.


The area of the extraction region, as a fraction of total area of the coordinate space (sky or detector), is recorded in the extracted spectra in the BACKSCAL keyword. Since the total area of the coordinate space is different between sky and detector coordinates for Suzaku same size regions will have different BACKSCAL values depending on the coordinate system. XSPEC automatically scales the background using the ratio of the BACKSCAL keywords before subtracting it from the source spectrum. For timing analysis, users must manually check the BACKSCAL keywords and subtract the correctly scaled version of the background lightcurve from the source lightcurve.

6.6 Generating RMF and ARF Files

6.6.1 Rebinning XIS Spectra $+$ Generating Responses using Xisresp

Xselect gives the user the option to rebin/group the spectra, when saving them. If the user answers ``yes'', the result will be physically rebinned from 4096 channels to 2048, with the following variable binning scheme:

0 699 1
700 2695 2
2696 4095 4

Xselect writes the above 3 lines into a file, chanfile.txt, then runs ``rbnpha binfile=chanfile.txt'' - users can do the same outside xselect for any XIS spectra that have been saved with "no" as the answer to the "rebin/group" question. Building Response Files: The simple Way (1)

If the object in question is a point source, and the extraction region is centered on that source, user can use the ``response=yes'' option when saving the spectrum:

xsel> save spec resp=yes

For an extended source the user can use the ``response=extend'' option:

xsel> save spec resp=extend

Xselect will run xisrmfgen and xissimarfgen to build the response file. In the case of an extended source, the WMAP image in the spectral file is used as the input image. If, in addition, the user opts to rebin the spectrum, xselect will also rebin the response files. This will leave a .rsp file, combining the .rmf and .arf files.

Note, however, that the .rmf and .arf generation with xselect, and therefore xisresp (next section), is not recommended for complicated cases, especially complex extended sources. For such cases, please run xisrmfgen (section 6.6.3) and xissimarfgen section 6.6.4 individually. Building Response Files: The Simple Way (2)

The script used by xselect to build the response is also available as a stand-alone script, xisresp. Usage:

xisresp filename <fast|medium|slow> regionfile extend? echo?

Xisresp is available at:

The fast/medium/slow parameter controls the binning of both the spectrum and the response: slow corresponds to no rebinning, medium to the default channel binning of xselect (and a similar binning in photon energy space), and fast a father linear factor 2 binning.

Note that xisresp does not rebin the background spectrum, when ``medium'' or ``fast'' is specified. That must be done by the user using the ``rbnpha binfile=chanfile.txt'' command described above.

Also note that xisresp does not check if the input spectrum has already been rebinned. Thus, using xisresp with the fast or the slow option on a spectrum saved rebinned in xselect will result in an error. (Using xisresp with the medium option on a spectrum that was rebinned and saved will produce an error message from rbnpha -- however, in this case, the error message can be ignored, since the script tries to rebin the spectrum as the last step, which fails but was unnecessary to begin with.)

6.6.2 Combining Spectra and Responses using addascaspec

Addascaspec is available as an ASCA FTOOL which can be used to combine the spectra and responses of Suzaku XIS (FI chip) data. It requires a 4-line Ascii file, listing the source spectral files, background spectral files, ARF files, and RMF files. It should have two or three columns depending on the number of active FI XIS units. For example, create the following file

x0.pha x2.pha x3.pha
x0b.pha x2b.pha x3b.pha
x0.arf x2.arf x3.arf
x0.rmf x2.rmf x3.rmf

and call it fi.add (this assumes a specific but obvious file naming convention). Then,

> addascaspec fi.add fi.pha fi.rsp fi_b.pha

will run several FTOOLS to create a combined source spectral file (fi.pha), a combined background spectral file (fi_b.pha), and a combined (RMF x ARF) response file (fi.rsp). Note that the operation to multiply and add the individual response files may be extremely memory-intensive, depending on the quality and the size of the original response files.

Note the update of the default and handling of the error statistic (errmeth parameter) in addascaspec with HEAsoft Version 6.7 and a bug fix with HEAsoft Version 6.8, see

6.6.3 Generating RMF Files using Xisrmfgen

The XIS response generator, xisrmfgen, takes into account the time variation of the energy response, appropriate for XIS data obtained with or without spaced-row charge injection (SCI). It is relatively straightforward to use, see example below.

> xisrmfgen
xisrmfgen version 2006-11-26
Name of input PI or IMAGE file or NONE[xis0-5b5w.pi]
Name of output RMF[xis0-5b5w.rmf]

The information concerning the instrument, clock mode, and observation date is taken directly from the header of the spectral file given6.1.

Note that xisrmfgen requires the spectral file to have a WMAP (weighted map) in detector coordinates. This is the default in recent (HEAsoft 6.1.2 or later) releases of xselect, although older versions defaulted to sky coordinates.

The following warning message:

xisrmfgen: WARNING: Weighted map or image is not in DET coordinate.
xisrmfgen: WARNING: Use constant weight on whole CCD.

indicates that the WMAP is in SKY coordinates. In this case, xisrmfgen generates a response file assuming a constant WMAP over the whole CCD. The current xisrmfgen does not consider spatial variation of spectral response on the CCD chip, which is negligible for the current data. Therefore, the practical effect of this is negligible. Nevertheless, it is advisable to generate spectral files with the WMAP in DET coordinates. To do so with older versions of xselect, issue the command:

xsel:SUZAKU-XIS1-STANDARD  set wmapname detx dety

6.6.4 Generating ARF Files using Xissimarfgen (or Xisarfgen)

Xissimarfgen is a ray-tracing based generator of ancillary response files (ARFs) for the Suzaku XIS. It is a powerful tool, which has far more parameters and modes of usage than the typical guest observer needs (or probably wants to know about). Since xissimarfgen calculates ARFs through Monte-Carlo simulations (it ray-traces X-ray photons through the Suzaku XRT and XIS and counts the number of events detected in the user-defined extraction regions), users need to simulate a sufficient number of photons to limit the statistical errors to an acceptable level.

For further details, users should refer to the paper by Ishisaki et al. in the Publication of the Astronomical Society of Japan (Ishisaki et al. 2007, PASJ, 59, 113;

For point sources very similar ARFs to the ones produced by Xissimarfgen can alternatively be produced by Xisarfgen, an FTOOL available since 2011. It uses a faster though potentially less accurate approach based on pre-calculated CALDB files (table interpolation) instead of on full ray-tracing. Further details are given in section 5.3.5 and in the Suzaku memo ``The fast ARF generator xisarfgen''
( Example: ARF of a Point Source

Here we give an example of generating an ARF file for a point source observed on-axis, using the data set ID 100012010. The following is an XIS1 image of the observation, in which one can see a bright point source at the center (Fig. 6.1). We assume that a spectrum from the white encircled region in the image has been extracted, and show how to generate a corresponding ARF file.

Figure 6.1: Example for an extraction region for a point source.
Image etacar_srcreg

Region files in ds9 format and in physical coordinates can be fed into xissimarfgen; when using this combination, the binning used to extract the image does not matter. To save a source region, one specifies the coordinate system ``Physical'' in the ``File Coordinate System'' row in the ``Region'' menu on ds9. Here is the example etacar_phys.reg:

# Region file format: DS9 version 3.0
# Filename: ae100012010xi1_1_5x5n001_cl.evt.gz[EVENTS]
global color=green font="helvetica 10 normal" select=1 edit=1 \
move=1 delete=1 include=1 fixed=0 source

corresponding to the white encircled region above in the physical coordinates. Then, run the following

xissimarfgen clobber=yes \
instrume=XIS1 \
pointing=AUTO \
source_mode=J2000 \
source_ra=161.264962 \
source_dec=-59.684517 \
num_region=1 \
region_mode=SKYREG \
regfile1=etacar_phys.reg \
arffile1=xis1_etacar.arf \
limit_mode=NUM_PHOTON \
num_photon=400000 \
phafile=etacar.pi \
detmask=none \
gtifile=100012010/xis/event_cl/ae100012010xi1_1_3x3n001_cl.evt \
attitude=100012010/auxil/ae100012010.att \
rmffile=ae_xi1_20060213.rmf \

Some options specify calibration files or Monte-Carlo simulation parameters that can be adjusted each time xissimarfgen is run. These are:

Other options fixed by the observation or by the upstream analysis. These are:

The ``pixq_[min,max,and,eql]'' parameters are not specified on the command line since we use the default setting (bad columns, pixels, and charge injection rows are excluded; the calibration source region is not subtracted).

Here is an example, in which the source position is specified in SKYXY coordinates.

xissimarfgen clobber=yes \
instrume=XIS1 \
pointing=AUTO \
source_mode=SKYXY \
source_x=784.5 \
source_y=786.5 \
num_region=1 \
region_mode=SKYREG \
regfile1=etacar_phys.reg \
arffile1=xis1_etacar.arf \
limit_mode=NUM_PHOTON \
num_photon=400000 \
phafile=none \
detmask=none \
gtifile=100012010/xis/event_cl/ae100012010xi1_1_3x3n001_cl.evt \
attitude=100012010/auxil/ae100012010.att \
rmffile=ae_xi1_20060213.rmf \
estepfile=default Example: ARF of a Uniformly Extended Source

Here we show an example of generating an ARF file for a uniformly extended source, using observation 102002010. The following is an XIS0 image of the observation, in which the strong emission from SNR E0102.2$-$7219 is evident. The goal is to search for possible extended emission from the surrounding areas. The calibration sources have to be cut out in order to avoid degradation of the data quality. This is done by typing

XSEL> select events "(STATUS<524287)&&(STATUS%(2**17)<2**16)"

in xselect.

Figure 6.2: Example for extraction regions for an extended source.
Image xis0img

As can be seen in the image (Fig. 6.2), events at the two corners where the calibration sources are located have been removed. We extract two spectra from the top-left half and bottom-right half of this image, using the region files e0102_tophalf_phys.reg

# Region file format: DS9 version 3.0
# Filename: xsel_image.xsl
global color=green font="helvetica 10 normal" select=1 edit=1 \
move=1 delete=1 include=1 fixed=0 source
and e0102_bottomhalf_phys.reg
# Region file format: DS9 version 3.0
# Filename: xsel_image.xsl
global color=green font="helvetica 10 normal" select=1 edit=1 \
move=1 delete=1 include=1 fixed=0 source

Then, the appropriate ARFs can be generated using the following command:

xissimarfgen clobber=yes \
instrume=XIS0 \
pointing=AUTO \
source_mode=UNIFORM \
source_rmin=0 \
source_rmax=20 \
num_region=2 \
region_mode=SKYREG \
regfile1=e0102_tophalf_phys.reg \
regfile2=e0102_bottomhalf_phys.reg \
arffile1=e0102_tophalf.arf \
arffile2=e0102_bottomhalf.arf \
limit_mode=MIXED \
num_photon=2000000 \
accuracy=0.005 \
phafile=e0102_tophalf.pi \
detmask=none \
gtifile=../xis/ae102020010xi0_cl.evt \
attitude=../auxil/ae102020010.att \
rmffile=e0102_tophalf.rmf \

As in the point source example, certain parameters specify calibration files or Monte-Carlo simulation parameters that can be adjusted for each run of xissimarfgen:

Other parameters depend on the data or on the upstream analysis:

To double-check that the intended source region was used by xissimarfgen, it can be displayed using ds9. For example, type

> ds9 e0102_tophalf.arf

while the selected STATUS bits can be confirmed in the standard output from xissimarfgen.

Note that the ARF files generated using the above command are normalized to the sizes of defined emitting regions. In the above example, the xspec output (e.g., the normalization parameter, the flux) assumes emission from an encircled region with 20arcmin radius. Generating ARFs in Sky Coordinates: Caveats

Users may need to specify the sky reference position when generating ARFs in SKY coordinates. Please refer to Appendix 2.3 of Ishisaki et al. (2006) for more details.

When users choose Reducing Run Time for xissimargfgen

  1. Reducing the number of energy bins of the ARF:

    Computation time of an arf table is proportional to the number of energy steps used in the calculation, where the energy steps are defined by the ``estepfile'' parameter:

    estepfile [filename]
       Energy step file or built-in steps. The built-in energy steps are:
       "full" : calculate effective area for each RMF energy bin. Very slow.
       "dense" or "default" : dense sampling (2303 steps). Slow.
       "medium" : medium sampling (157 steps). Moderate.
       "sparse" : sparse sampling (55 steps). Fast.

    In many cases it will be sufficient to use the estepfile=medium option: while a broad band fit will not yield significantly different results or improvements compared to the estepfile=default case, the arf computation time will go down by a factor of $\sim15$ (2303/157).

  2. Optimizing the ``num_photon'' and ``accuracy'' parameters:

    When limit_mode=NUM_PHOTON, computation time is also proportional to the number of faked photons defined by the ``num_photon'' parameter.

    For a point source ARF limit_mode=NUM_PHOTON and num_photon=400000 is recommended, but for faint sources limit_mode=MIXED, num_photon=200000, and accuracy=0.005 is acceptable.

    For a uniform sky ARF limit_mode=MIXED, num_photon=2000000, and accuracy=0.005 is recommended.

    For ARFs of extended sources (including uniform sky ARFs), visual inspection of the accuracy of ARFs, by plotting the effective area in xspec, is highly recommended.

  3. Running xissimarfgen with a fast CPU using a fast code:

    The Monte-Carlo ray-tracing simulation runs numerous floating point calculations, and so Athlon64s usually run xissimarfgen faster than Pentium4s. If available, 64-bit codes compiled on 64-bit Linux run about 1.5 times faster than 32-bit codes on the same PC.

  4. Running ARF caluculations for different sensors (XIS0/1/2/3) on different machines.

  5. Generating ARFs of multiple accumulation regions within the same observation simultaneously.

  6. Defining the smallest emission region possible when generating diffuse source ARFs.

    If an emission region defined by the ``source_image'' parameter is too large compared with the event extraction region, computation time gets slower without improving the quality of the simulation. For generating a uniform sky ARF, we recommend source_mode=UNIFORM, source_rmin=0, and source_rmax=20.

    Similarly, if the spacecraft attitude is not stable after a maneuver and the emission region moves out from the event accumulation region, the ARF calculation becomes slow.

6.6.5 Faster Fitting of XIS Spectra

  1. Rebinning the RMF:

    Because the standard RMF of the Suzaku XIS has 7900 energy bins (2eV steps, 0.2-16keV) times 4096 PI bins, xspec needs a lot of memory to read the RMF and time to fit a spectral model. This fine-step matrix is usually over-sampled for moderate flux sources with featureless X-ray spectra (e.g., AGNs).

    Users can rebin the RMF in both channel- and energy-spaces with rbnrmf. Note that the spectral file also needs to be rebinned, when the RMF is rebinned in channel-space. Users can also specify the channel-space rebinning factor using the ``rebin'' parameter of the xisrmfgen.

    The RMF energy bins are determined with the default set of parameters
    ebin_lowermost=0.20, ebin_uppermost=16.0, and ebin_width=2.0. Users who are only interested in the soft band spectrum can reduce the RMF size by 25% with ebin_uppermost=12.0. When the spectral model to fit is featureless (no strong emission lines), ebin_width=4.0 or ebin_width=8.0 will give almost the same fit result. Older versions of the RMF, e.g., ae_xi0_20050916.rmf, in the CALDB has non-equal-width energy bins with 4096 steps in 0.2-12.0keV. Users can copy these energy steps, by specifying

    ebin_mode=1 ebinfile=ae_xi0_20050916.rmf

    ARFs must be re-created when the RMF energy bins are changed.

  2. Combining XIS0, (XIS2,) and XIS3:

    The XIS team recommends adding the spectra and response for the units with the frontside illuminated (FI) chips. XIS1 spectra, however, must be fitted separately since its (backside illuminated, or BI chip) response is distinctly different from those of the FI chips.

6.7 Pile-Up

Figure 6.3: Incident versus observed count rates of a point source for the FI sensor. The thick colored lines show the range that can be observed without strong pile-up for a given window option (defined by the figure legend) and burst option (defined by the time values $t_\textrm {b}$ indicated in the figure). Note that the window frame times $t_\textrm {w}$ are 8, 2, and 1s for the full, 1/4, and 1/8 window options, respectively, with exposures per frame of $t_\textrm {b}$ for a given burst option.

Image pile-up

The XIS is a position-sensitive integrating instrument, with a nominal interval of 8s between readouts. If during the integration time more than one photon strikes the same CCD pixel, or one of its immediate neighbors, these cannot be correctly detected as independent photons; this is the phenomenon of photon pile-up. Here, the modest angular resolution of the Suzaku XRT is an advantage. However, photon pile-up can be a problem when observing bright sources. Note that the attitude fluctuation of the satellite does not mitigate the photon pile-up, because the attitude drift is negligible within the exposure of 8s. Fig. 6.3 shows the range of the incident counting rates that can be observed without the pile-up.

When pile-up occurs, both image and spectral data are distorted. The energy spectrum tends to become harder and the total photon flux tends to decrease. In extreme cases, all events may be discarded as non-X-ray events at the image center, and the local photon flux becomes effectively zero. This means that a point source image would have a hole at the center. This is the simplest method to detect photon pile-up.

The fraction of ASCA grade 1 events shows a strong correlation with the pile-up fraction, i.e., guest observers can check the significance of the pile-up in their data by studying the fraction of ASCA grade 1 events. The simplest method to correct for pile-up is to remove the image center from the event extraction region. This means that the events are extracted from a ring shaped region. In this case, a corresponding ARF file needs to be calculated.

See also section 5.3.3 for (a) links to a 2011 XIS pile-up study with case examples and (b) introduction of two routines to determine pile-up (aeattcor2, pileest).

6.8 XIS Non X-Ray Background

Screened XIS event data still include particle background events (Non X-ray Background: NXB) and X-ray background events. These contributions can best be estimated from off-source area of the same XIS CCD chip, but this is not always possible for extended sources. Alternatively, we can estimate the particle background during an observation from data taken when the satellite sees the night side of the Earth (night Earth), i.e., when Suzaku sees no X-ray emission through the telescopes. Night Earth data are collected by the XIS team and stored in the CALDB. Related files are:

ae_xi?_nxbsciof_yyyymmdd.fits: SCI-OFF event file
ae_xi?_nxbscion_yyyymmdd.fits: SCI-ON event file
ae_xi?_nxbvdchk_yyyymmdd.fits: HK file with the detector temperature
ae_xis_nxbcorhk_yyyymmdd.fits: HK file of the cut-off rigidity
ae_xis_nxborbit_yyyymmdd.fits: orbit file

Here, yyyymmdd is the release date of these CALDB files. If a source observation in question was taken after the time period in the ``DATE-END'' column of a set of those files - normally ending several days to a month before the release date - newer background files have to be downloaded from the CALDB page.

The XIS team has developed a tool that collects the appropriate NXB data from CALDB and automatically generates NXB images and spectra. The tool xisnxbgen is available in the HEAsoft version 6.4 or later.

6.8.1 Xisnxbgen

Tawa et al. (2008, PASJ, 60, 11) have shown that the XIS NXB only varies with the cut-off-ridigity (COR) value at the satellite location. Based on this result, xisnxbgen sorts NXB data by COR values, generates an NXB spectrum and image for each COR range (defined by the ``sortstep'' option), and combines them weighted by the exposure time ratio of each COR range included in the GTIs of the supplied spectral source file. The default NXB indicator is COR2 (revised cut-off rigidity). The other indicators such as the obsolete cut-off rigidity COR and the PINUD rate, which can be calculated with aemkpinudhk, can also be used via the hidden option ``sortkey''.

When users choose non-standard event by event selection criteria, this has to be defined using hidden parameters of xisnxbgen (``grades'', ``pi_min'', ``pi_max'', etc.). Xisnxbgen takes care of non-standard gti filters (for example, screening with satellite elevations from the bright/night earth) when it calculates the exposure time ratio of each COR2 range from the GTIs of the supplied spectral source file.

Here, we show an example of generating an NXB spectrum and image. First, we need to create a spectral FITS file of the source, which we call etacar_nebula_x0.pi in this example. We name the output NXB spectral file etacar_nebula_nxb.pi. We also need to input the source region file (etacar_nebula_x0.reg), from which we created etacar_nebula_x0.pi, and the coordinate system (SKYREG), on which the region file is described. The attitude and orbital files for the data are ae402039010.orb and ae402039010.att, respectively, which are found in the auxil directory in the data distribution. With this, we can run:

> xisnxbgen etacar_nebula_nxb.pi etacar_nebula_x0.pi \
SKYREG etacar_nebula_x0.reg \
ae402039010.orb ae402039010.att

Xisnxbgen displays all the options that were selected after the text ANL: *** xisnxbgen show parameter ***. We recommend that the user confirms that the options are specified as intended. The ouput file has an NXB spectrum in the 1st extension, an NXB image in detector coordinates in the 2nd extension, and an NXB image in sky coordinates in the 3rd extension. The sky coordinates map in the 3rd extension ignores the region file selection, i.e., the user will get a sky NXB background image of an entire CCD chip (to display the detector/sky background images with ds9, type ds9 etacar_nebula_nxb.pi[2] or ds9 etacar_nebula_nxb.pi[3] on the command line). These results can be used in xspec as background, or for NXB subtraction in the sky image.

To produce an NXB image within a certain energy range, the lower and upper boundary PI values (3.65eV/channel) can be specified using the pi_min and pi_max parameters. For example:

> xisnxbgen etacar_nebula_nxb.pi etacar_nebula_x0.pi \
SKYREG etacar_nebula_x0.reg \
pi_min=274 pi_max=548 \
ae402039010.orb ae402039010.att

The NXB spectral file is not affected by these options, that is, it also has values below pi_min and above pi_max.

Things to be considerd when using xisnxbgen:

COR2          :  EXPOSURE (s)  FRACTION (%)
  0.0 -   4.0 :       1184.0         2.161
  4.0 -   5.0 :       3560.0         6.498
  5.0 -   6.0 :       3224.0         5.885
  6.0 -   7.0 :       3672.0         6.703
  7.0 -   8.0 :       3408.0         6.221
  8.0 -   9.0 :       3808.0         6.951
  9.0 -  10.0 :       4288.0         7.827
 10.0 -  11.0 :       5096.0         9.302
 11.0 -  12.0 :       7368.0        13.449
 12.0 -  13.0 :       6440.0        11.755
 13.0 -  99.0 :      12736.0        23.248
          SUM :      54784.0       100.000
        TOTAL :      54784.0       100.000
COR2          :  EXPOSURE (s)  FRACTION (%)  SPEC (cts) IMAGE (cts)
  0.0 -   4.0 :       4584.0         1.254        37.2       878.0
  4.0 -   5.0 :      19136.0         5.235       129.3      3077.0
  5.0 -   6.0 :      18816.0         5.148       122.7      2547.0
  6.0 -   7.0 :      18640.0         5.099       104.1      2275.0
  7.0 -   8.0 :      20520.0         5.614       104.4      2308.0
  8.0 -   9.0 :      23792.0         6.509       112.0      2491.0
  9.0 -  10.0 :      43136.0        11.801       162.7      4004.0
 10.0 -  11.0 :      45496.0        12.446       179.9      4098.0
 11.0 -  12.0 :      42992.0        11.761       174.1      3710.0
 12.0 -  13.0 :      44456.0        12.162       171.2      3698.0
 13.0 -  99.0 :      83968.0        22.971       310.0      7219.0
          SUM :     365536.0       100.000      1607.6     36305.0
        TOTAL :     365536.0       100.000     36305.0     36305.0
    EFFECTIVE :     370262.3       101.293      1679.1     37775.9

6.9 Creating an Exposure Map

For the study of extended sources with the XIS, it is necessary to know the exposure times as well as vignetting at various sky locations within the XIS image.

6.9.1 Types of Exposure Maps

One type of exposure map can be created by simply considering the detector field of view and the spacecraft attitude, the result being the actual exposure time per sky pixel. Such exposure maps can be created by using xisexpmapgen, which allows users to exclude unused pixels such as bad columns, hot/flickering pixels, SCI rows, and the $^{55}$Fe calibration source area. See section below as well as the help file of xisexpmapgen for further details.

In the other type, the effective exposure times per sky pixel are calculated, taking into account the vignetting of the XRT. Below, we describe how to use xissim to simulate a ``flat field'' image for this purpose.

6.9.2 Running Xissim

As an example, we show how to simulate an XIS0 flat field image at 2.45keV of the observation sequence 102002010. The attitude wobbles during this observation are included in the simulation by supplying the attitude file and a GTI table.

> xissim instrume=XIS0 enable_photongen=yes photon_flux=1 flux_emin=1.0 \
flux_emax=10.0 spec_mode=1 image_mode=2 time_mode=0 limit_mode=1 energy=2.45 \
ra=16.0083 dec=-72.0313 sky_r_min=0 sky_r_max=20 exposure=15825.09 \
pointing=AUTO gtifile=cleaned.evt\[GTI\] attitude=ae102002010.att \
ea1=16.007012398071 ea2=162.031577674707 ea3=29.330729822566 \
xis_rmffile=/FTP/caldb/data/suzaku/xis/cpf/ae_xi0_20060213.rmf \
outfile=sim_x0.fits phafile=allarea.pi


Note that the output file has only $\sim $10% of the seed photons. This is because most of the photons are absorbed or blocked by mirrors or instruments.

6.9.3 Extracting a Flat Field Image

The simulated events created by xissim have STATUS information, which describes the quality of each simulated photon. Thus the simulated event files should be screened using the same STATUS criteria as for the observed events (see Table 6.2).

Then the flat field image can be extracted in xselect, making sure that the same XY binning as the observed image is used.

6.9.4 Smoothing the Flat Field Image

It is difficult to avoid statistical fluctuation in a simulated flat field map, so it is often desirable to smooth the map using, e.g., ximage or ds9. We assume that the flat field map has been smoothed, with the file name flatfield_smo.img.

6.9.5 Trimming the Flat Field Image

A smoothed map generally has rough edges, so it is useful to trim such a map with a masking image, which can be done using xisexpmapgen.

> xisexpmapgen expmap.img cleaned.evt ae102002010.att

Here ae102002010.att is the attitude file, and cleaned.evt is used as the value of the ``phafile'' parameter to supply the XIS mode (such as the window option).

The output file (expmap.img) contains two maps; a mask image in detector coordinates in the 1st extension and an exposure map in sky coordinates in the 2nd extension. Here, we need a masking image in sky coordinates, and so use the image in the 1st extension.

By displaying the 1st extension, one can empirically determine a good threshold for masking. For a threshold of 5000s, for example, use:

> fimgtrim infile=expmap.img\[1\] threshlo=5000 threshup=5000 \
const_lo=0 const_up=1 outfile=skymaskmap.img

This produces a masking image, called skymaskmap.img. This may have to be rebinned to match the binning of the exposure map (by default, xselect bins Suzaku images by a factor of 8), before multiplying with the smoothed flatfield image.

> fimgbin skymaskmap.img skymaskmap_8bin.img 8
> farith flatfield_smo.img skymaskmap_8bin.img flatfield_smo_trim_8bin.img "*"

6.9.6 Applying the Flat Field Image

> farith source_raw.img\[0\] flatfield_smo_trim_8bin.img input_vigcor.img "/"

The above applies the smoothed, trimmed, binned flatfield image flatfield_smo_trim_8bin.img to the primary extension of the observed image source_raw.img by dividing the latter by the former. A vignetting corrected image, here input_vigcor.img, is produced. The flat field image can be scaled to make it a true effective exposure time map, although the normalization depends on the purpose of such an operation.

Depending on the scientific objectives, it may well be desirable to subtract particle, cosmic X-ray, or Galactic X-ray background from the observed image before dividing by the flat field.

6.10 Initial Processing: Details

The remainder of this chapter describes the details of the initial processing for the XIS. These steps, already performed in the processing pipeline (Table 6.1), can be repeated by users if necessary.

6.10.1 Assigning Arrival Times

xistime assigns the corrected arrival time to the XIS events, and performs the fine time measurement for the Burst mode and Window option. Here is an example for its application:

xistime infile=filename_uf.evt.gz \
              outfile=xistime_outfile.fits \

infile is the the name of the XIS event fits input file,
outfile is the name of the output event fits file created,
timfile is the name of the input fits file with timing information, ae<obsid>.tim, which can be found in the Suzaku data distribution package of each observation.

6.10.2 Calculating Sky Coordinates

xiscoord combines the position of the observed counts on the XIS detector with the orbit and attitude information to calculate the ACT, DEC, FOC and sky X/Y values for XIS event files. xiscoord uses either the attitude file assigned on the basis of the event input file name (the default), or fixed Euler angles if the parameter attitude is set to EULER. The RA and DEC used by the program can either be read from the header of the input event file or set manually. In the former case the command is:

xiscoord infile=xistime_outfile.fits outfile=xiscoord_outfile.fits \
attitude=DEFAULT pointing=KEY

infile is the name of the XIS event fits input file,
outfile is the name of the output file created - see caveat below,
attitude indicates where to get the attitude information from,
pointing indicates where to read the RA and Dec - pointing set to KEY reads them from the header of the input event file
Users should be aware of the following:
1) When the attitude parameter is set to ``Default", the code searches for a file named ***.att in the same directory as the input file. This can be bypassed by specifying the full path to the file on the command line.
2) We have found that xiscoord does not produce output files on several unsupported platforms (Mandrake 10,..). Users are advised to check the supported platforms (see and run only on a supported platform.

6.10.3 Put Pixel Quality

xisputpixelquality runs on the output of xiscoord. The command is:

xisputpixelquality xiscoord_outfile.fits xisputpixelquality_outfile.fits
infile is the name of the XIS event fits input file (output from xiscoord),
outfile is the name of the output file created.

The hidden parameters, badcolumfile and calmaskfile, should point to CALDB. Users may want to examine the differences (if any) between the input and the output files of xisputpixelquality.

6.10.4 Computing PI Energy Channels

As its name indicates, the xispi routine calculates the XIS PI and grades values from the PHAs, see also section 6.3. In addition to the input event file, the routine needs the CALDB files ae_xi[0-3]_makepi_[date].fits and the housekeeping file associated with the input event file. If the CALDB option is not set properly and the file has to be input manually, users should check that the latest ``makepi" file is used. The command to run xispi is:

xispi infile=xisputpixelquality_outfile.fits outfile=xispi_outfile.fits \
 hkfile=HKFILE.fits makepifille=CALDB
infile is the name of the XIS event fits input file,
outfile is the name of the output file,
hkfile is the House Keeping file located in the xis/hk directory. This is not the hk file from the auxil directory,
makepifile is a hidden parameter, set to ``CALDB'' by default.

6.11 Standard Screening

Both bad pixel filtering and grade selections are done by the processing pipeline and implemented in the cleaned files distributed to the users. Users can find a complete example of filtering at: In addition, we provide an xselect command file and files containing event and mkf selection expressions via: We explain the steps below, but see also section 6.3 for the availablity of a screening script.

6.11.1 Bad Pixel Filtering

The cleaning of hot and flickering pixels is done using cleansis, available as a standalone script at the GOF website

cleansis was originally written for analysis of the ASCA SIS data and removes hot and flickering pixels based on a Poissonian analysis. It has since been adapted for work on Swift and Suzaku data. This generalized version is available in all releases after 6.0.6 of HEAsoft. Users should make sure that their version of HEAsoft is current.

To run cleansis on Suzaku XIS event files type cleansis chipcol=SEGMENT on the command line, give the input and output filenames, and use the default values of the remaining parameters.

6.11.2 Grade Filters

The GRADE column shows the event grade, which is determined from the distribution of pulse heights among the 5x5 (or 3x3 or 2x2) pixels. The standard spectral responses provided by the XIS team assume GRADE 0,2,3,4, and 6. Only events with these grades should be selected (within the xselect task):

select event "GRADE==0||GRADE==2||GRADE==3||GRADE==4||GRADE==6"

Or, equivalently:

filter grade "0,2-4,6"

The Suzaku instrument teams recommend the following cuts to be applied within xselect:

select mkf  "SAA_HXD==0 && T_SAA_HXD>436 && ELV> 5 && DYE_ELV>20" \ 
mkf_name=MKF_filename  mkf_dir=/path-to-the-MKF-file/

1) mkf_name and mkf_dir should be set automatically by xselect by ``read events'',
2) The ``select mkf'' command creates a time filter of good times. To actually filter the events, users must then issue the command ``extract events".

Satellites launched into low-Earth orbit, such as Suzaku, pass through the South Atlantic Anomaly (SAA). During a passage, the high particle flux makes the instruments unusable. The mkf column SAA_HXD has a value of 0 when the satellite is not in the SAA and so the selection condition is SAA_HXD==0 (this is based on the current extent of the SAA as determined empirically using the HXD data). Even when the satellite emerges from the SAA, the background is still high, the mkf column T_SAA_HXD indicates the amount of time since an SAA passage. For the XIS, T_SAA_HXD can be as low as 60 seconds. However, the HXD background stays high for much longer. The instrument teams have recommended adopting the same condition for both instruments, hence the cut of T_SAA_HXD$>$436 is imposed on the XIS data.

The two remaining criteria are recommended by the instrument teams to reduce the contamination from Earth's atmosphere. The first is applied to the elevation angle, mkf column ELV, the angle between the target and the Earth's limb. Only data with an elevation angle larger than 5 degrees should be considered. The second concerns the elevation angle from the day Earth rim and helps reduce contamination in the Nitrogen and Oxygen lines from X-rays scattered on the Earth's atmosphere. Users who can ignore the low energy part of their spectrum (below 0.6keV) may want to explore the possibility of relaxing the cut on DYE_ELV.


... given6.1
It is also possible to run xisrmfgen without specifying a spectral file; see the help file for details
... criteria6.2

next up previous contents
Next: 7. HXD Data Analysis Up: Suzaku ABC Guide Previous: 5. ``README FIRST" for   Contents
Katja Pottschmidt 2013-09-04