next up previous contents
Next: 6. XIS Data Analysis Up: Suzaku ABC Guide Previous: 4. Suzaku Data Analysis   Contents


5. ``README FIRST" for the Suzaku Data Analysis

5.1 Introduction

This chapter contains the details of the current status of the data analysis. Users should check this chapter to answer the following questions:

Users should also consult the following web pages:
which serve a similar purpose to this chapter but may be updated more frequently; and
for information regarding the processing pipeline; and
which contains the calibration uncertainties.
Furthermore, users are encouraged to contact us via the comment webpage at

5.2 On-board Changes

5.2.1 XIS2 Loss

One of the XIS units with an FI chip, XIS2, suffered catastrophic damage on 2006 November 9. Since then, no astronomically useful data have been obtained with XIS2, although some diagnostic mode data are taken. Users should therefore not expect any cleaned event files for XIS2 observations taken after 2006 November 9.

The default telemetry allocations for the different XIS units and their changes over time are described in the Technical Description.

5.2.2 XIS0 Anomaly

An anomaly occurred in XIS0 on 2009 June 23. In this case the leak charge is confined within a portion of the CCD resulting in a dead area in segment A (ACTX$\sim $80 to $\sim $140). Therefore, an area discriminator (ACTX=70$-$150) is applied, in order to discard pseudo-events due to leak charge. How the leak charge affects the calibration of not only segment A but also the other segments is currently under investigation. Preliminary analysis indicates a very small but a detectable change in the gain of XIS0. Detailed information can be found in the Suzaku memo available at

5.2.3 XIS1 Anomaly

The XIS1 showed an anomaly, starting 2009 December 18, some time between 12:50 UT and 14:10 UT. A bright and persistent spot ($\sim $90 pixels in FWHM) suddenly appeared at the end of the segment C in all images taken during day-earth observations, while none was found during night-earth observations. The XIS team believes that the anomaly stems from optical light leaked from a hole in the optical blocking filter created by a micrometeorite hit. The XIS team continues to operate the XIS1 in the same way as before the anomaly. Science observations of soft diffuse emission with low surface brightness might be impacted by data degradation. Possible effects under investigation are: increased inaccuracy of the calibration (flux calibration), increased noise due to optical and UV light, and a decreased rate of effective data recording (due to an increase in spurious events). Detailed information can be found in the Suzaku memo available at

5.2.4 Additional Optical Blocking Filter Holes in XIS1 and XIS3

A total of 10 additional bright spots were found in the frame dump images of XIS1 and XIS3 in 2013, likely caused by small holes in the optical blocking filters. No effects on the X-ray data are expected in most cases. See Suzaku memo at for further details.

5.2.5 XIS Spaced-Row Charge Injection (SCI)

The cumulative effect of in-orbit radiation damage creates charge traps in CCDs, leading to an ever increasing charge transfer inefficiency (CTI). This changes the PHA-Energy conversion factor as a function of the number of transfers before the charge can be read out, hence on the position on the CCD. This also degrades the spectral resolution.

The ability to inject charge into the CCD chips has been designed into Suzaku XIS. This can be used to fill the charge traps and therefore ameliorate the effects of the CTI. This operation is called spaced-row charge injection, or SCI: In 2006 September, SCI has been used for selected observations to test its effectiveness. Once this has been confirmed, it was offered as an option to all users in 2006 November. Since 2007 April (Cycle 2 observations), the use of SCI is the default.

With Version 2 processing, this is purely a data quality issue. Users need not take action, since information about the SCI operation is encoded in the CI keyword, which is read by Suzaku FTOOLS as necessary and the appropriate calibration is used. However, users may wish to check the value of this keyword. The value of the CI keyword should be 0 for observations without SCI, and 2 for observations with SCI.

5.2.6 XIS1 Change to CI$=$6keV in 2010/11

From the start of the SCI operation, a charge equal to the one produced by a 6keV X-ray photon (``6keV equivalent'') was injected for the FI devices while a smaller amount (``2keV equivalent'') was used for the BI device, XIS1. By 2010 it had been shown that with an increased SCI amount of 6keV equivalent for XIS1 the high-energy response improves significantly with no measurable degradation of the low-energy response. Therefore, the amount of charge injection for XIS1 was increased from 2 to 6keV equivalent for better high energy response in 2010 and 2011. The changes were made gradually for different clocking modes and retrospectively for some calibration observations. More information can be found on the Suzaku GOF web pages at,
in the Suzaku memo available at,
and in the CALDB release note The ``makepi'' and ```rmfparam'' file versions listed in the release note (ae_xi1_makpi_20111018.fits, ae_xi1_rmfparam_20111020.fits) or later are required to analyze these data. Xisrmfgen, xissimarfgen, and xisarfgen automatically choose the appropriate CALDB files based on header information of the spectral input file (specified using the phafile parameter).

A general rule is that most (including all GO) observations were made with CI$=$2keV before the change and with CI$=$6keV after the change at the dates shown below in Table 5.1. Some exceptions can be found in XIS calibration observations made in 2010-2011; some calibration observations were made with CI$=$6keV before the changes and with CI$=$2keV after the changes in order to track the long-term gain change for both CIs.

Table 5.1: Dates for the XIS1 CI$=$2keV to CI$=$6keV change for different modes.
Date Unit Mode
2011-06-01 XIS1 Injection charge increased to 6 keV for Normal (no option).
2011-08-22 XIS1 Injection charge increased to 6 keV for Normal (1/4 window).
2011-09-01 XIS1 Injection charge increased to 6 keV for Normal (0.1s burst).
2011-10-06 XIS1 Injection charge increased to 6 keV for Normal (1/4 win$+$1.0s burst, 1/8 win).
2011-10-11 XIS1 Injection charge increased to 6 keV for Normal (1/4 win$+$0.1s, 0.3s, & 0.5s burst).
2011-10-25 XIS1 Injection charge increased to 6 keV for Normal (0.5s, & 0.62s burst). Check CI Setting Using XIS1 Event File

Before 2011-04-01:

Check the ``submode ID'' which is part of the XIS1 event file name. If the file name includes ``u'' (e.g., ae105027010xi1_0_3x3n069*u*_cl.evt.gz), it was taken with CI$=$6keV. If it includes ``b'' (e.g., ae105027010xi1_0_3x3n069*b*_cl.evt.gz), it was taken with CI$=$2keV.

After 2011-04-01:

Check the ``microcode ID'' in the XIS1 event file.

fkeyprint ae103001020xi1_0_3x3n066c_cl.evt.gz  CODE_ID

If the ID is one of the ones listed in Table 5.2, the observation was made with CI$=$6keV.

Table 5.2: Microcode IDs for XIS1 with CI$=$6keV.
Code_ID Comment
129 BI Normal Mode Periodic CI 1CI/54rows speed4x IA TrailMask (Obsolete)
130 BI Normal Mode Periodic CI 1CI/54rows speed4x IA TrailMask
131 BI Normal 1/4 Win at the XIS pos. with SCI and TrailMask
132 BI Burst 0.100sec SCI trail mask for CI=6keV
133 BI Burst 1/4Win 0.994sec SCItrailmask CI=6keV XIS1 XIS-nominal
134 BI Normal 1/8Win SCItrailmask CI=6keV XIS1 XIS-nominal
135 BI Burst 1/4Win 0.130sec SCItrailmask CI=6keV XIS1 XIS-nominal
136 BI Burst 1/4Win 0.299sec SCItrailmask CI=6keV XIS1 XIS-nominal
137 BI Burst 1/4Win 0.495sec SCItrailmask CI=6keV XIS1 XIS-nominal
138 BI Burst 0.500sec SCItrailmask CI=6keV
139 BI Burst 0.620sec SCItrailmask CI=6keV XIS1 NXB Increase with CI$=$6keV

The increased charge injection amount for XIS1 led to an increase in its NXB level: In the SCI operation, artificial charges are injected in every 54th row. Some fraction of the charges is carried over to the adjacent rows. Events in the charge injected rows (SCI_ROW) and in those next to them (SCI_TRAILING_ROW) are removed onboard, but a small fraction of the charges is carried over to the overnext rows (SCI_2ND_TRAILING_ROW). Due to the increase in the total amount of injected charges, the remaining charges in the second trailing rows increased, resulting in the increased NXB level. See section 5.3.7 for a description of how to estimate and/or mitigate the increased XIS1 NXB with CI$=$6keV.

5.2.7 PIN Epochs: Low Energy Threshold & Responses

The bias voltage on-board and low energy threshold in the ground processing of various subsets of PIN units have been adjusted since launch to reduce noise events. This changes the characteristics of these PIN units in several discrete steps.

With Version 2 processing, data from all PIN units should be analyzed together. However, due to the changes in the bias voltage and the software threshold, response matrices appropriate for the epoch of the observation (see Table 7.3 in chapter 7) should be used in spectral fitting.

5.2.8 Additional Attitude Wobble

Since launch, the 3-axis attitude control using the Inertial Reference Unit (IRU) had been done with the combination IRU-X/Z/S1. On December 18th, 2009, a switch from IRU-S1 to S2 was performed. The IRU-S2 or S1 controls the attitude in the DET-X direction while the IRU-X controls the attitude in the DET-Y direction. In this configuration the accuracy of the attitude control along the DET-X axis declined by roughly a factor of two. On June 15th, 2010, another gyro, IRU-Y, was activated. The IRU-Y is more sensitive to the attitude of the DET-X direction than is the IRU-S2. IRU-S2 was used for attitude control, and the IRU-Y for attitude determination. Since June 25th, 2012, IRU-Y has also been used for attitude control which is good to $\sim1$arcmin or better, with the attitude determination being good to $\sim20$arcsec or better, i.e., back to the original situation. Table 5.3 summarizes these changes. The actual behavior differs from observation to observation. Some possible effects are listed in the following.

Table 5.3: History of IRU operations.
Time Period Attitude Control Attitude Determination
  Configuration, Accuracy Configuration, Accuracy
2006/07/10 - 2009/12/18 IRU-X/Z/S1, $\lesssim1$arcmin IRU-X/Z/S1, $\lesssim20$arcsec
2009/12/18 - 2010/06/15 IRU-X/Z/S2, $\lesssim2$arcmin IRU-X/Z/S2, $\lesssim2$arcmin
2010/06/15 - 2012/06/25 IRU-X/Z/S2, $\lesssim2$arcmin IRU-X/Z/Y, $\lesssim20$arcsec
2012/06/25 - present IRU-X/Z/Y, $\lesssim1$arcmin IRU-X/Z/Y, $\lesssim20$arcsec

Possible pointing determination errors between December 18th, 2009, and June 15th, 2010: XIS images may be elongated along the direction of the DET-X axis by 1arcmin. The recommended treatment is to:

Possible flux variations of the XIS data taken between Dec 18th, 2009, and June 25th, 2012: The pointing direction is corrected on the ground to within the calibration uncertainties. The calibration error of the pointing direction correction is larger for the data taken between December 18th, 2009 and June 15th, 2010 (see above). The image is reconstructed before the data are distributed to the users. But even in the attitude corrected data the count rate of the XIS detectors might vary by up to a few tens of percent. The variability is larger for off-axis sources because the variability results from a change of the effective area, where the XRT vignetting function is smooth on-axis while sharp off-axis (suzakumemo 2008-05). The count rate varies with the 96min satellite orbit. The recommended treatment is to:

Photometry using XIS1 with a window option: The arteficial flux variation described above is most pronounced for the XIS1, especially when operated with a narrow window option such as 1/4 or 1/8 and/or at the HXD aimpoint. The inaccuracy in the attitude control along the DET-X axis is larger than along the DET-Y direction. Unfortunately, the narrower boundary of the window option is set on the DET-X axis for XIS1 (and the inoperable XIS2) while it is on the DET-Y axis for the others (XIS0 and 3).

Detailed information and example images can be found in the Suzaku memos available at (2010-04, -05, -06).

5.3 XIS Software Tools, Recipes

5.3.1 Reprocessing and Screening: Aepipeline

Since 2010 the FTOOL aepipeline is available. It can be used to apply the newest energy calibration, as well as to perform data screeening for both XIS and HXD data. See section 6.4 for more information on using aepipeline for XIS data.


(1) Aepipeline has an input directory parameter and an output directory parameter, among others. The output directory may not be a subdirectory of the input 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.

5.3.2 Cleansis Runaway Problem

The FTOOL ``cleansis'' is applied in Suzaku data processing, when going from the unscreened (in the event_uf subdirectory) to the cleaned (event_cl) event files. When users reprocess data (e.g., to apply a new gain calibration), it is generally recommended to start with unscreened data, and the use of cleansis is one of the recommended screening steps. This tool was originally written for the ASCA SIS. Its algorithm is inherently statistical: it examines the distribution of the number of counts per raw detector pixels, and rejects those that have far more counts than likely. This step is usually repeated to eliminate lower level flickering pixels.

The versions of cleansis in HEAsoft v6.7 and earlier occasionally suffer from a runaway problem, creating a hole in the XIS data near the peak of the image. Users who encounter such a hole should first check if one is present in the unscreened event data. If this is the case, this is most likely due to a severe case of photon pile-up and is unrelated to cleansis. If, on the other hand, the hole is seen only in the cleaned data, cleansis is the likely cause.

The version of cleansis available since HEAsoft 6.8 incorporates an improvement in the algorithm which greatly reduces, if not completely eliminates, instances of such runaways. Alternatively, there are two run-time parameters that can be adjusted to reduce or eliminate this problem:

Both these methods can prevent the runaway, at the possible cost of retaining some low-level flickering pixels in the screened data.

5.3.3 Improved Attitude Correction and Pile-Up Estimation

Section 6.7 provides a general description of the pile-up effect and gives typical count rate ranges that can be observed with minimal pile-up (Fig. 6.3).

In addition, a detailed pile-up study has been performed in 2011, see$\sim $yamada/soft/XISPileupDoc_20120221/
and Yamada et al., 2012, PASJ, 64, 53. This study includes case examples that can be used as analysis recipes (web page: ``Super Bright Source'', ``Relatively Bright Source'', ``Dim Source''; paper: Table 2 - exclusion regions for a sample of sources).

Two FTOOLs routines are available, aeattcor2 and pileest, which were especially designed to aid in the analysis of Suzaku data that might be afflicted by photon pile-up.The tool aeattcor2 further improves the attitude correction for the slow wobbling of the optical axis for bright sources by using their detected image to create a new attitude file. Note that this tool cannot be used if a clear double image of the source is visible in the image. pileest is designed to be run after aeattcor2. It will create an image that will show an estimated, minimum pile-up fraction for user specified levels. This allows the user to create a region file that excludes the most piled areas, and then to estimate the effective pile-up fraction of the remaining events. The two FTOOLs were created based on the externally contributed routines, and, written in S-lang and are designed to be run on the command line using ISIS, the Interactive Spectral Analysis System, as a driver. Both utilize a number of S-Lang Modules Packages. For all intents and purposes, the scripts behave similar to typical Unix command line tools. They are available from

5.3.4 Exposure Maps, Vignetting Correction

For the study of extended sources with the Suzaku XIS, it is necessary to know the exposure times as well as vignetting at various sky locations within the XIS image. One type of exposure maps 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. For the other type, the effective exposure times per sky pixel are calculated, taking into account the vignetting of the XRT. A detailed description can be found in section 6.9 of the XIS chapter and at It consists of the following steps:

5.3.5 A faster ARF Generator than Xissimarfgen: Xisarfgen

The FTOOL xisarfgen was designed to calculate ARFs for point-like sources using pre-calculated (via ray-tracing) files in the CALDB (ae_xrt*_effarea_XXXXXXXX.fits and ae_xrt*_psf_XXXXXXXX.fits). This is usually much faster than using xissimarfgen (which always carries out ray-tracing), especially when the Euler angles are given as an input instead of the attitude file (see examples below). The different methods of calculation result in only small differences between the effective areas in the ARFs produced by xissimarfgen and xisarfgen. This is illustrated in the Suzaku memo ``The fast ARF generator xisarfgen''

The usage of xisarfgen is similar to that of xissimarfgen, e.g.:

xisarfgen phafile=filename rmffile=filename \
source_mode=J2000 source_ra=value source_dec=value \
region_mode=SKYREG num_region=1 regfile1=filename arffile1=filename \


xisarfgen phafile=filename rmffile=filename \
source_mode=J2000 source_ra=value source_dec=value \
region_mode=SKYREG num_region=1 regfile1=filename arffile1=filename \
attitude=none ea1=value ea2=value ea3=value

5.3.6 XIS Non X-Ray Background Estimation

Before Suzaku FTOOLS Version 7, the tool and the data set of night Earth observations released from the XIS team were used to estimate the non X-ray background (NXB).

With Suzaku FTOOLS Version 7, which was a part of HEAsoft v6.4 released in December, 2007, the tool was upgraded to xisnxbgen and was released as one of the official FTOOLS. The NXB database was released as a part of the official Suzaku CALDB.

Now users should use xisnxbgen with the Suzaku CALDB instead of to estimate the NXB. The tool is now obsolete. The FTOOLS help document and section 6.8.1 give detailed information about xisnxbgen.

5.3.7 Estimation & Mitigation of XIS1 NXB Increase with CI$=$6keV

The increased charge injection amount for XIS1 led to an increase in its NXB level: In the SCI operation, artificial charges are injected in every 54th row. Some fraction of the charges is carried over to the adjacent rows. Events in the charge injected rows (SCI_ROW) and in those next to them (SCI_TRAILING_ROW) are removed onboard, but a small fraction of the charges is carried over to the overnext rows (SCI_2ND_TRAILING_ROW). Due to the increase in the total amount of injected charges, the remaining charges in the second trailing rows increased, resulting in the increased NXB level. When the events in the second trailing rows are removed for CI$=$6keV, the level is consistent with that for CI$=$2keV. Estimate XIS1 NXB

The tool xisnxbgen, when used with the appropriate nxbevent file in the CALDB, will estimate the NXB level appropriate for the amount of charges injected. Due to the current structure of the CALDB, xisnxbgen is unable to pick the correct file automatically for XIS1 data taken with CI$=$6keV. For such data, please explicitly specify the nxbevent calibration file:

xisnxbgen nxbevent=ae_xi1_nxbsci6_XXXXXXXX.fits

If you have a copy of the Suzaku CALDB locally installed, this file can be found in the CALDB data/suzaku/xis/bcf directory. If you are using the HEASARC version of the CALDB remotely, it is best to download this specific file from:

Note that the NXB calibration files ae_xi*_nxb*_20121201.fits are wrong and should not be used (section 5.5.8). Mitigate XIS1 NXB Increase with CI$=$6keV

In the pipeline processing, events in the second trailing rows are NOT removed. Users should make their own choice of whether they remove second trailing rows for a lower NXB level with a smaller effective area. The recipe for doing so is the following:

5.3.8 XIS Burst Option Exposure Times

The XIS team has released a version of xistime allowing to apply the option ``bstgti=yes'' as part of the Suzaku FTOOLS V14 in HEAsoft v6.8. This produces detailed good time intervals (GTIs). Previously, the GTIs of the XIS event files did not contain the detailed GTI information for burst mode data (e.g., one line for each 2 second exposure separated by 8 seconds). This led to inaccuracies in the exposure times after data filtering.

Note that the GTI extension may then contain a large number of rows, potentially causing problems in down-stream software. Also note that this option works even when the data were taken without burst mode, but in this case its application is not recommended.

5.3.9 Timing Mode

From AO-5 on, the Timing mode is used for a restricted numbers of observations. However, the calibration accuracy of the Timing mode remains worse compared to the normal mode. A recipe for reducing Timing mode data can be found at

Due to comparably large numbers of hot/flickering pixels in the BI CCD and due to increased charge leakage in XIS0 in the Timing mode since the 2009 anomaly, the Timing mode is available only for XIS3. Since only one dimensional information is available in the Timing mode, the distinction between X-ray and non-X-ray events becomes inaccurate. This means that the Timing mode has a significantly higher non-X-ray background than the Normal mode. Preliminary analysis showed that the non-X-ray background in the Timing mode is one or two orders of magnitude larger than in the Normal mode. Because the available data are limited, its detailed behavior, e.g., dependence on the cut-off rigidity, is not known.

The in-flight calibration of the Timing mode is on-going and is expected to improve in the near future. However, it will not reach that of the normal mode, because (1) the Spaced-row Charge Injection (SCI) is not available on board and (2) the CTI correction is difficult in the ground processing. The updated calibration information of the Timing mode will be released as part of the CALDB. Unavailability of the SCI or the CTI correction means that the energy scale and resolution are significantly different from those of the normal mode. Preliminary analysis showed that the gain was lower than the nominal value by $\sim $10% and the energy resolution had degraded to $\sim $5% at 5.9 keV in early 2009. The former may be corrected if appropriate calibration information becomes available, but the latter not.

In addition, the effective area of the CCDs may change in the Timing mode: even a small number of hot/flickering pixels produces a relatively large dead area in the CCD, which reduces the effective area. This effect is time dependent. It should be noted that the accuracy of the ARF file may also be degraded, because we need to use a rectangular extraction region for the Timing mode.

The nominal time resolution in the Timing mode is 7.8ms, but this depends on the image size of the X-ray source. If the X-ray image is much larger than 128 pixels ($\sim $2arcmin) along the Y-address, the time resolution may become worse than 7.8ms. Because the Y-address represents the photon arrival time in the Timing mode, the image extension (in the Y-direction) works as a low-pass filter. This means that, even for a point source, some signal is leaked to the adjacent time bins due to the tail of the XRT PSF. Therefore, the frequency response is somewhat reduced near the Nyquist frequency of the nominal time resolution of 7.8ms.

5.4 HXD Software Tools, Recipes

5.4.1 Reprocessing and Screening: Aepipeline

Since 2010 the FTOOL aepipeline is available. It can be used to apply the newest energy calibration, as well as to perform data screeening for both HXD and XIS data. See section 7.4 and 7.6.3 for more information on using aepipeline for HXD data.


(1) Aepipeline has an input directory parameter and an output directory parameter, among others. The output directory may not be a subdirectory of the input 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.

5.4.2 HXD Spectra and Lightcurves: Extraction Tools

Since 2010 a set of FTOOLS is available that can be used to extract PIN (hxdpinxbpi, section 7.5.6) and GSO (hxdgsoxbpi, section 7.7.2) spectra, as well as PIN (hxdpinxblc, section 7.9.1) and GSO (hxdgsoxblc, section 7.10.1) lightcurves. These tools also perform additional tasks like background extraction or deadtime correction where necessary. See fhelp texts for these tools or sections indicated in the previous sentence for more detailed information.

5.5 XIS Calibration Updates

5.5.1 Contamination Model

Figure 5.1: Data points indicate the C column density of the contaminant derived from observations. The solid lines indicate the best fit empirical model to the temporal evolution of the contamination for each sensor, as available, e.g. for xissimarfgen, through the CALDB files released in 2013 September (ae_xiN_contami_20130813.fits). The dashed lines indicate the previous version of the calibration files (ae_xiN_contami_20120719.fits).
Image abc_xis_rates_byelem_c

Figure 5.2: Same as Fig. 5.1 but for the O column density.
Image abc_xis_rates_byelem_o

Figure 5.3: Same as Fig. 5.1 but for the N column density.
Image abc_xis_rates_byelem_n

In late November 2005, contamination in the optical path of each sensor became apparent. Spectra of celestial sources show that the contaminant is predominantly carbon. Monitoring of 1E0102.2$-$7219 and RXJ1856.5$-$3754 showed that the contamination is changing at a different rate for each sensor leading to an equivalent additional column density of C of $\sim 2-7 \times 10^{18}$cm$^{-2}$ (as of mid 2013; see Fig. 5.1). Observations of the bright earth show that the contaminant is twice as thick at the center of the field of view than at the edge, a pattern that tracks the temperature distribution on the optical blocking filter (OBF). This suggests that the contaminant is on the spacecraft side of the OBF, rather than on the CCD detector surfaces.

The ARF generator xissimarfgen takes the contamination effect into account. That is, the ARF calculated with xissimarfgen by default includes transmission through the contaminant, taking observed data up to a certain date for each sensor into account. Those dates are defiend by the CALDB file used. Transmission values beyond those dates are extrapolated. The FTOOL xiscontamicalc is handy for determining contaminant transmission values at specified detector coordinates and dates, and for manually changing transmission information in an ARF file. Contamination Model $\leq $ Suzaku FTOOLS Version 17 - released before 2011 June

Intitial studies suggested that the contaminant is mainly DEHP (C$_{\rm 24}$H$_{\rm 38}$O$_{\rm 4}$, or C/O=6 by number). The C/O = 6 description has been used for calibration up to Suzaku FTOOLS Version 17. This version corresponds to CALDB file tags 20081023 [XIS1, 2, 3] and 20090128 [XIS0] taking data up to March 15, 2008, and December 14, 2008, into account, respectively. Suzaku FTOOLS Version 17 and below are not compatible with newer CALDB contamination files than these. Contamination Model $\geq $ Suzaku FTOOLS Version 18 - released in 2011 June

With Suzaku FTOOLS Version 18 an improved contamination model has been released. While the spatial dependence is the same as in the previous version, the elemental composition (C, O, H) of the contamination is now variable allowing for more accurate spectral modeling, especially in the 0.4-0.5keV energy range. This version corresponds to CALDB file tag 20091201 [XIS0-3] taking data up to June 26, 2009, into account. Suzaku FTOOLS Version 17 and below are not compatible with these new CALDB contamination files, while Suzaku FTOOLS Version 18 can handle the new as well as the old CALDB files. Contamination CALDB Files further improved - released in 2012 September

Figure 5.4: 2011 April 26 observation of PKS2155$-$304 fitted using (a) old and (b) new models of the contaminant.
Image pks2155comp

Figure 5.5: 2012 June 11 observation of the Cygnus Loop fitted using (a) old and (b) new models of the contaminant.
Image cygloop-comp

The 20120902 version of the Suzaku XIS CALDB includes updated calibration files for the XIS contaminant build-up (ae_xiN_contami_20120719.fits, for use with Suzaku FTOOLS Version 18 or higher). These represent a significant improvement in the calibration of the low energy response of the XIS. The XIS team has revised the spectral models for 1E0102.2$-$7219 and RXJ1856.5$-$3754, included PKS2155$-$304 observations, and analyzed newer calibration observations obtained since the previous release.

For the center of the XIS FOV, the time dependence of the chemical composition of the contaminant was revised and for XIS1 is now including nitrogen. The model of the spatial distribution of contaminant was also updated. In addition to the use of the new model for the contaminant build-up at the center of the instrument, and the use of the recent calibration observations, the parameter for the spatial distribution is now allowed to vary over time. The CALDB update information at gives the equation used for the spatial dependence. See Fig. 5.4 and Fig. 5.5 for examples of the improvement of spectral residuals with the new model. Contamination CALDB Files further improved - released in 2013 September

Building on the 2012 version the contamination files have been further improved. The calibration according to the further updated files (ae_xiN_contami_20130813.fits) is shown as black solid lines for the different XIS modules and for different elements in Fig. 5.1 to Fig. 5.3. The dashed lines represent the 2012 calibration (ae_xiN_contami_20120719.fits). While there is generally very little difference between the two for data taken before 2012, the improvement can be considerable for newer data.

5.5.2 XIS Reprocessing History

Table 5.4 gives a list of revisions of the makepi CALDB files that are used in the processing pipeline since V2.1.6.13:

Table 5.4: Revisions of the makepi file.
CALDB Update File Name Comments
2007-06-27 ae_xiN_makepi_20070611.fits  
2007-07-31 ae_xiN_makepi_20070730.fits  
2007-11-01 ae_xiN_makepi_20071031.fits Important for SCI-on$^a$
2008-02-01 ae_xiN_makepi_20080131.fits  
2008-09-05 ae_xiN_makepi_20080825.fits Important for SCI-off $2\times2^b$
2009-08-13 ae_xiN_makepi_20090615.fits  
2010-01-23 ae_xiN_makepi_20091202.fits  
2010-12-06 ae_xiN_makepi_20100929.fits  
2011-06-30 ae_xiN_makepi_20110621.fits  
2011-10-10 ae_xiN_makepi_20110907.fits  
2011-11-09 ae_xiN_makepi_20111018.fits  
2012-02-09 ae_xiN_makepi_20111227.fits Do not use, see section 5.5.7
2012-07-03 ae_xiN_makepi_20120527.fits  
2012-11-06 ae_xiN_makepi_20121009.fits  



The energy scale calibration for both SCI and non-SCI data is included in Version 2 processing, and in general achieves an accuracy of 0.2% at 6keV.

Significant makepi updates will continue to be announced 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.

5.5.3 Energy Scale for SCI Data

The XIS team constantly updates CTI and gain calibration of XIS data taken with SCI (Table 5.4). Note that cleaned data from processing Version or earlier suffer from time and energy dependent effects in energy scale calibration, and should therefore be reprocessed. This can be done by running xispi on unscreened files (running it on screened files will lead to inaccurate results, since new calibration changes the event grades, which are used for screening). The updated files must then be screened to produce new cleaned files. See: for more detailed instructions. Also see section 5.3.1 above for the introduction of a high-level script for both, reprocessing and screening.

5.5.4 Energy Scale for SCI-Off Data of 2006

For some non-SCI data taken in 2006, the energy scale calibration is noticeably worse than stated above. Discrepancies of up to 40eV have been noticed among different XIS units.

5.5.5 Energy Scale for Window Option Data

A discrepancy in the energy scale of window option data compared to the full window data has been resolved with an xispi update in Version 12 of the Suzaku FTOOLS by incorporating a CTI correction formula that takes the two types of charge transfer into account. With this, the energy scale of the FI and BI data in 1/4 window mode with SCI-off are in agreement to within $\pm20$eV at 6keV with that of the full window mode (before 40eV for the FI CCD and 60eV for the BI CCD). For the SCI-on data good agreement between the full window and the 1/4 window modes has subsequently been shown as well.

5.5.6 Energy Scale for $2\times 2$ Editing Mode Data

For SCI-off data there was an offset in the $2\times 2$ mode energy scale relative to the $3\times3$/$5\times5$ modes. This is believed to have been fixed with the release of the ae_xiN_makepi_20080825.fits files in 2008 September. Users should apply the new calibration using the same procedure as for SCI-on data processed with V2.1.6.15 or earlier.

The energy scale for $2\times 2$ mode with charge injection is expected to be similar to those of the $3\times3$/$5\times5$ modes. However, this expectation has not yet been verified with actual calibration. Users of $2\times 2$ mode data with SCI should carefully check for gain discrepancies before combining $2\times 2$ events with $3\times3$/$5\times5$ events. Note that $3\times3$ and $5\times5$ events can generally be combined.

5.5.7 Energy Scale Calibration Files Ae_xi0,3_makepi_20111227.fits wrong

A bug was found in the gain calibration files ae_xi0,3_makepi_20111227.fits in the CALDB. These files are used by xispi to calculate the PI and grade values from the PHA values. For window option data, an additional gain correction is usually made to match with the gain for the full window in the pipeline processing. The additional correction was mistakingly NOT made for these files, i.e., for processing the XIS0 and XIS3 data taken with a window option in the period between 2011-12-29 and 2012-07. This leads to line energies that are wrong by up to $\Delta E
\sim 20-30$eV at $6-7$keV. At lower energies, the effect is smaller. The affected ObsIDs are listed in Table 5.5.

Table 5.5: Observations processed with wrong gain files for XIS0 and XIS3.
Object Observation Date Sequence Number
PERSEUS_1_4_WIN 2012-02-08 106007020
E0102$-$72_1_4_WIN 2012-04-23 107003010
PKS2155$-$304 2012-04-27 107010010
GX304$-$1 2012-01-31 406060010
4U1705$-$44 2012-03-27 406076010
AXJ1846.8$-$0240 2012-04-01 407019010
4U1630$-$47 2012-02-13 906008010
EXO2030$+$375 2012-05-23 407089010
3C273 2012-07-16 107013010

Data processed with the CALDB version 20120703 (ae_xi0,3_makepi_20120527.fits) or newer are corrected for this issue. The affected data listed in Table 5.5 have to be reprocessed by the user using, e.g., aepipeline (see sections 5.3.1 and 6.4) or xispi:

xispi \
infile=ae${seqnum}xi${sensor}_*_uf.evt \
outfile=ae${seqnum}xi${sensor}_*_new.evt \
hkfile=ae${seqnum}xi${sensor}_*.hk.gz \

See also ISAS page and CALDB release note 20120530_MakepiUpdate_win14GainTable.pdf under

5.5.8 XIS NXB Calibration Files Ae_xi*_nxb*_20121201.fits wrong

There was a problem in the 2013-01-10 release of the XIS CALDB regarding the NXB database (files ae_xi*_nxb*_20121201.fits), such that xisnxbgen will produce a background map for the 1/4 window region only, regardless of the option used for your observation.The XIS CALDB has since been updated. The 2013-03-05 release (files ae_xi*_nxb*_20130228.fits) corrects the above problem. See following ISAS page for more details:

5.5.9 Calibration Uncertainties near 2keV

Figure: Example for typical residuals around 2keV due to instrumental effects (blue: BI - XIS1, red, combined FI - XIS0+3), with respect to a smooth continuum fit (top) and modeled with two Gaussian lines (bottom), see text for line energies (Suchy et al., 2011, ApJ, 733, 15). The 2012 improvements (see text) are not taken into account. While these improvements reduce the residuals their typical shape is still apparent in some observations.
Image si_au_residuals

Spectral modeling residuals have often been seen in the 1.6keV to 2.3keV region, especially for bright sources. They are due to imperfect calibration of the instrumental Si K edge at $\sim $1.8keV (and maybe some influence of imperfect calibration of the instrumental Si K-alpha fluorescence line at $\sim $1.7keV as well) and the Au M edge at $\sim $2.2keV. An example - not yet taking the 2012 improvements described below into account - is shown in Fig. 5.6, where residuals of a spectral fit to data from a 500mCrab bright outburst of the accreting pulsar 1A1118$-$61 are displayed (Suchy et al., 2011, ApJ, 733, 15), for a smooth continuum model (top) and a model with two additional Gaussian lines (bottom) at 1.82$\pm$0.01keV (BI, ``emission'') or 1.89$\pm$0.01keV (FI, ``absorption'') as well as 2.21$\pm$0.01keV (BI and FI). Improved Si Edge QE Calibration, mainly for XIS03 - released 2012 June

The XIS team has released a set of new quantum efficiency files (ae_xiN_quanteff_20120428.fits). This update or later versions modify the XAFS structures around the Si K edge ($\sim $1.8keV) and the Al K edge ($\sim $1.6keV) by adopting new absorption coefficients. This leads to the quantum efficiency for the BI CCD XIS1 changing only a little - not solving the 2keV residual problem - while for the FI CCDs XIS03 it changes significantly - improving the residuals near 2keV. The X-ray transmission of the optical blocking filter around the Al K edge has also been updated. To use the new calibration, install/use the latest CALDB and rerun xisrmfgen; there is no need to reprocess the event files or regenerate the ARF files. See also: Improved Si Edge Response Calibration for XIS1 - released 2012 December

Suzaku FTOOLS Version 20 was released in 2012 December as part of HEAsoft version 6.13. Xisrmfgen has two new hidden parameters in this release, bi_si_edge_mode and fi_si_edge_mode. The former is set to 1 by default, which instructs xisrmfgen to use a new and improved response model around the Si K edge for XIS1. Setting it to 0 reproduces the old response. The latter is reserved for future use. It is currently set to 0 by default and should not be changed from the default for the time being.

5.5.10 Calibration Uncertainties at low Energies for highly absorbed Sources

Figure: Spectra and best fit models for an observation of GX301$-$2 including data from the three individual XIS instruments and the two HXD instruments. At lower energies, one clearly sees the constant level in the modeled flux discussed in the text (Suchy et al., 2012, ApJ, 745, 124).
Image apj411544f2_lr

The response function of the XIS detectors includes a constant component seen below the main peak (Koyama et al. 2007, PASJ, 59, S23). Although the origin of the component is not fully understood, a part of the component is due to X-ray photons absorbed at a place close to the boundary between the depletion layer and the insensitive layer, in which case only a small fraction of electrons can be collected as a signal (Matsumoto et al., 2006, SPIE Proc., 6266, 626641).

In highly absorbed sources, like the accreting pulsar GX301$-$2, this component can become dominant at low energies (pulse height channels). It is difficult to calibrate in space and therefore has a comparatively large uncertainty in the current response function. For sources that are weak at soft energies it might be necessary to ignore those energies if they are showing residuals associated with this component. In GX 301-2 the residuals seem to be stronger for the BI spectrum than for the FI ones (see Fig. 5.7 and Suchy et al., 2012, ApJ, 745, 124). In this case the $<2$keV data were not included in the final fit.

5.5.11 XRT/XIS Effective Area

The 2008-07-09 release of the Suzaku CALDB contains an updated calibration of the XRTs and the corresponding updates to the XIS quantum efficiency files, while the Suzaku FTOOLS release Version 8 contains updates of the raytracing library used by the xissim and xissimarfgen response generator tools (older versions of these tools do not work with the new calibration files, while the new version of the raytracing library is backward compatible with the older calibration files). The new calibration can be applied without reprocessing the event data. The only requirement is to re-generate the response files using xisrmfgen and xissimarfgen.

As part of this release the vignetting and imaging calibrations have been updated. The geometry parameters for each quadrant of each XRT have now been individually adjusted. As a result, the calibration of the point-spread function (PSF)/encircled energy function (EEF) has improved, particularly beyond 2arcmin from the image center. The simulated EEF now coincides with the observed values within 4% from 1 to 6arcmin radius.

The combination of the XRT and XIS calibration changes regarding the effective area is such that, for observations at the XIS nominal position, analyzed using a large (e.g., 250 pixel radius) extraction region, changes are relatively small. Note, however, that the relative normalization for the PIN data is now 1.16 and 1.18 for the Crab nebula observed at the XIS and HXD nominal positions, respectively (with estimated errors of 0.015). Note also that the changes in the EEF and in the vignetting calibration affect the effective area calibration as a function of source location and extraction region radius. Be advised that effective area calibration may have changed significantly for some extraction regions.

5.6 HXD Calibration Updates

5.6.1 PIN Noise & Energy Range

Figure 5.8: Example of an overall PIN spectrum with noise contamination.
Image pinnoise_contami

The leakage current of the PIN sensors has been gradually increasing due to radiation damage. Therefore, the HXD team has updated the PIN threshold level file (ae_hxd_pinthr_NNNNNNNN.fits) several times. This file is used to remove the noise events so that the clean event file does not contain them. However, after releasing an updated threshold level file, the PIN noise level further increases, and noise events sometimes leak into the clean event files before the release of new threshold files, especially for observations where the HXD temperature was high. (The temperature can be checked at .) In that case, PIN noise events appear at 10-15keV as a steep increase to the lower energy as shown in Fig. 5.8. With the current PIN response matrices of all epochs, systematic uncertainties of typically 5% and 3% remain in the $12-15$keV and the $>15$keV range, respectively. The HXD instrument team generally recommends to ignore data below 15keV.

Figure 5.9: Example of 64 individual PIN spectra with noise contamination.
Image pinnoise64pin

Looking at the individual 64 PIN spectra can already be used now to identify noise contamination. The noise level is different between PINs, unlike the cosmic signals and background signals. In the XSELECT, the following command selects the events of an individual PIN diode:

filter column "UNITID=2:2 PIN_ID=3:3"

In this example the events of PIN no. 3 of the Well UNIT no. 2 are extracted. UNITID is in the range of 0 to 15 and PIN_ID is in the range of 0 to 3. Fig. 5.9 shows an example of the individual 64 PIN spectra for data with noise contamination. An example of a C-shell script for creating the 64 PIN spectra and plotting them into postscript files using XSELECT and XSPEC can be found at:

5.6.2 GSO Gain Calibration GSO Gain Calibration $\geq $ Suzaku FTOOLS Version 16 - released in 2010 March

Figure 5.10: HXD-PIN (black) and HXD-GSO (red) spectra and fit residuals for the Crab as observed on 2005 Sep. 15 using the nominal HXD pointing position. The adopted model is wabs$\times $bknpowerlaw, with $N_{\rm{H}}$=3.8$\times $10$^{21}$cm$^{-2}$, photon indices of 2.09 and 2.27, a break energy of 103keV, and a nomalization of 10.9photos keV$^{-1}$ cm$^{-2}$ s$^{-1}$ (at 1keV). To illustrate the effect of the correction arf file, fit residuals without applying the file are displayed as well (blue). The 20% difference of the 70-400keV flux between the data and the model without the arf and the bump-like residual around 50-70keV due to calibration uncertainties around the Gd-K edge are reduced by introducing the correction arf file, resulting in better agreement with the HXD-PIN spectrum.
Image fitbknpower_webpage_hxdnomi_2005_webfinal

On 2010, March 29, the HXD team has released updated software (as part of Suzaku FTOOLS version 16 within HEAsoft v6.9) as well as a new CALDB file (ae_hxd_gsogpt_20100323.fits), implementing a revised gain calibration for the HXD-GSO data. This extends the low energy limit of the usable GSO data from 70keV down to 50keV. Note that a new type of CALDB gain history calibration file, new non X-ray background files, and new response $+$ correction arf files have to be applied (Fig. 5.10). One of the most important differences in the pipeline processing is that instead of the old type of GSO gain calibration file, ae_hxd_gsoght_YYYYMMDD.fits, the new type, ae_hxd_gsogpt_YYYYMMDD.fits, has to be used with hxdpi. Details are described by Yamada et al., 2011, PASJ 63, No. 5, and at and
Support for the old software/calibration/backgrounds for new observations has stopped for observations from 2010 spring onwards. However, older observations can be reduced with both, the new and the old set up, the latter by using hxdpi_old in Suzaku FTOOLS Version 16+, instead of hxdpi, for reprocessing.

For an explanation on how to determine which calibration setup was used for the pipeline processing (``cleaned events'') of a specific GSO dataset and of when and how to perform reprocessing for a specific GSO dataset see section 7.3, sections, and paragraphs on the $\geq $ Suzaku FTOOLS Version 16 background model events below in this section. GSO Gain Calibration GSOGPT CALDB File further improved - released in 2011 October

Since the initial release of the new GSO gain calibration setup described above an improved version of the CALDB file has been released (ae_hxd_gsogpt_20110819.fits). It corrects more recent GSO data in order to keep gain fluctuation below 1%. For more details see CALDB release note at This calibration file has been used in the pipeline processing (``cleaned events'') since 2011, November 3. See paragraphs on the $\geq $ Suzaku FTOOLS Version 16 background model events below in this section for a statement on when data processed with the previous GSOGPT file (ae_hxd_gsogpt_20100323.fits) should be reprocessed. GSO Gain Calibration $\geq $ Suzaku FTOOLS Version 16 - new GSO Non X-ray Background Files

Table 5.6: $\geq $ Suzaku FTOOLS Version 16 GSO NXB model event file properties.
Availability Dates Directory METHOD$^a$ METHODV$^b$ GSOGPT File
2005-08-17-2011-11-03 gsonxb_ver2.5 LCFIT(bgd_d) 2.5ver0912-64 20100323
2011-03-01- gsonxb_ver2.6 LCFIT(bgd_d) 2.6ver1110-64 20110819
$^a$ FITS header keyword to distinguish background modeling methods.
$^b$ FITS header keyword to distinguish revisions.

Two new sets of HXD/GSO Non X-ray Background (NXB) event files corresponding to the version of hxdpi released with Suzaku FTOOLS Version 16 (HEAsoft v6.9) are available since early 2012. Table 5.6 shows for which dates they are available. They are called ver2.5 or ver2.6, with GSO energy calibration file ae_hxd_gsogpt_20100323.fits and ae_hxd_gsogpt_20110819.fits used in their creation, respectively. The NXB files can be downloaded from and
New ones become available around one month after the observation date.

The FTOOLS Version 16 and later pipeline processing (``cleaned events'') uses
ae_hxd_gsogpt_20100323.fits for observations performed before 2011, November 3, and ae_hxd_gsogpt_20110819.fits for observations performed after 2011, November 3, as indicated by the header keyword GSOGPT_F. Users should choose the background model accordingly. In most cases it is not necessary to reprocess $\geq $ FTOOLS Version 16 GSO data (for reprocessing of earlier FTOOLS version data see section 7.3 and sections taken before 2011, November 3, since the difference between the analysis results for the two gsogpt files is not large. Only for bright sources (above 100mCrab in the GSO band), it is recommended to reprocess the data from 2010, March 1, to 2011, November 3. Once the data have been reprocessed with ae_hxd_gsogpt_20110819.fits, background ver2.6 should be used. Aepipeline automatically chooses the newest calibration file.

5.7 Cross Calibration

5.7.1 Flux Cross Calibration: XIS0,1,3

Figure 5.11: Relative flux normalization between different XIS sensors. The ratio of the best-fit normalization values are compared between two sensors at the XIS and the HXD nominal positions. Observations of a power-law source in the Normal clocking mode were used. The best-fit constant values are indicated by solid lines.
Image norm_pos_normal

Table 5.7: Relative flux normalization between different sensors.
Position Ratio Mean Standard deviation
XIS-norm XIS1/XIS0 1.026 0.016
  XIS3/XIS0 1.014 0.017
HXD-norm XIS1/XIS0 1.004 0.019
  XIS3/XIS0 0.980 0.022

The stability of the relative flux normalization between the three XIS sensors is shown in Fig. 5.11. No significant time dependence is found. The relative normalization remains constant. The mean and standard deviation are summarized in Table 5.7, separately for the XIS and HXD nominal positions.

5.7.2 Flux Cross Calibration: XIS and HXD

Observations of the Crab have been used to study the cross-normalization of XIS and HXD. With Version 2 processed data, the normalization of PIN data relative to XIS0 data is currently 1.16 for observations at the XIS nominal position, and 1.18 for those at the HXD nominal position (with estimated errors of 0.015).

The normalization of the GSO relative to the PIN was found to be $\sim0.8$. This discrepancy can be considerably reduced with the temporary workaround of applying the GSO ``correction arf'' in addition to the standard arf, see

Cross normalization factors should be taken into account in joint spectral fits of XIS and HXD data.

5.7.3 Flux and Spectral Shape Cross Calibration: Multi-Mission

See, e.g., Tsujimoto et al., A&A, 525, 25 (2011).

next up previous contents
Next: 6. XIS Data Analysis Up: Suzaku ABC Guide Previous: 4. Suzaku Data Analysis   Contents
Katja Pottschmidt 2013-09-04