Things to Watch Out For
Last Updated Aug 1, 2008
This page contains the brief summaries of important issues that all Suzaku users should be aware of. It will be updated frequently to reflect our latest understanding of real and potential issues in Suzaku data analysis.
Most V1.x to V2.x transition issues are believed to have been solved.
- All Suzaku observations have been (re)processed with V2.x.
- HXD/PIN background files have been released (updated 2008 Jun 17).
- HXD/GSO background files have been released.
- XIS night earth background database and tool have been released.
- There is a new version of the ABC guide that corresponds to V2.x data anlysis, although it does not yet include the description of GSO data analysis.
The remainder of this page documents the known calibration and software issues.
1. Calibration Issues
ISAS maintains the calibration status and caveats pages.
In addition, we note the following items:
- HXD/PIN - XIS Cross Normalization (Updated 2008 Jul 31)
- With the 2008-07-09 release of CALDB used with Version 2 data,
the normalization of PIN data relative to XIS0 data is 1.16 for
observations of the Crab at the XIS nominal position, and 1.18
for those at the HXD nominal position (uncertainty is ~0.015). See
Suzaku
Memo 2008-06 for details.
With older calibrations, the normalization of PIN data relative to XIS0 data is 1.06-1.09 for observations of the Crab at the XIS nominal position, and 1.11-1.13 for those at the HXD nominal position. The rangees are for use of different energy bands for the PIN data. See Suzaku Memo 2007-11 for details.
- With the 2008-07-09 release of CALDB used with Version 2 data,
the normalization of PIN data relative to XIS0 data is 1.16 for
observations of the Crab at the XIS nominal position, and 1.18
for those at the HXD nominal position (uncertainty is ~0.015). See
Suzaku
Memo 2008-06 for details.
- PIN - GSO Cross Normalization (Posted on 2008 Jun 27)
- With the current calibration, in the 2007 March 20 observation
of the Crab at the HXD nominal position, there is a cross-normalization
problem at the 20% level (1.0:0.80, PIN:GSO) and ~10% residuals in
the GSO data.
The HXD team is working on resolving this issue. In the mean time, empirical arfs can be used to reduce both problems significantly.
- With the current calibration, in the 2007 March 20 observation
of the Crab at the HXD nominal position, there is a cross-normalization
problem at the 20% level (1.0:0.80, PIN:GSO) and ~10% residuals in
the GSO data.
- XIS: Energy scale and resolution calibration as verified using the 55Fe calibration soruce
- SCI data processed using V2.1.6.15 and earlier versions: After 2006 September, the gain at Mn K alpha line decreased at a rate of 30 eV/year in the FI chips and 50 eV/year in the BI chip. See notes on V2.0/2.1 processed XIS data for more. Also read about updated calibration.
- Non-SCI data: The energy scale is well calibrated untill July 2006 in FI units, but systematically higher by 10 eV in XIS1. After this time, gain gradually increased in XIS3, rapidly decreased in XIS0 and XIS2, and gradually decreased in XIS1.
- 2x2 mode (non-SCI): With the current calibration, there is an offset in the 2x2 mode energy scale relative to the 5x5 mode. This offset is sensor, time, and energy dependent.
- For full details, see XRT/XIS Energy scale page at ISAS.
- XIS: Flux calibration for small extraction regions (users of data with
1/4 Window or 1/8 Window options are likely to be affected).
- XIS flux calibration is most accurate for large, circular
extraction regions centered on the source. For example, a radius
of 250 pixels (260 asrcsec) ensures that 99% of the flux of a point
source is included in the extraction region. Users should use a
large extraction region if at all possible.
The current calibration is noticeably less accurate for small extraction regions. This is not a major issue for r>150 arcsec (agreement among sensors to better than a few %), but becomes serious for r<100 arcsec. For 50 arcsec region, XIS1 in particular will measure fluxes that are 20-30% lower than it should (XIS0 appears most reliable in this respect, followed by XIS3 and XIS2).
The XRT and XIS teams are working on refinement of the PSF and the effective area calibrations that will improve the situation. However, because of the interaction of two instruments and two calibration items, this is proving to be a time-consuming process.
- XIS flux calibration is most accurate for large, circular
extraction regions centered on the source. For example, a radius
of 250 pixels (260 asrcsec) ensures that 99% of the flux of a point
source is included in the extraction region. Users should use a
large extraction region if at all possible.
- XIS: Contaminant thickness for XIS0 (Updated 2008 Jun 18)
- The thickness of contamination for XIS0 is increasing more
rapidly than the model in the calibration file in CALDB releases
2007-12-04 (XIS Calibration Files: 20071122) and older.
Spectral fits to recent XIS0 data using response files created
using these contamination calibration files will therefore
"detect" excess NH.
Updated calibration has been released in the 2008-01-13 release of CALDB (XIS Files 20080113), and a further update for XIS0 is in the 2008-06-02 release of CALDB (this is unaffected by other problems). Read details.
- The thickness of contamination for XIS0 is increasing more
rapidly than the model in the calibration file in CALDB releases
2007-12-04 (XIS Calibration Files: 20071122) and older.
Spectral fits to recent XIS0 data using response files created
using these contamination calibration files will therefore
"detect" excess NH.
- PIN responses by epoch:
- With V2.x processing, data from all PIN units should be analyzed together. However, due to the changes in bias voltage and the software threshold, response matrices appropriate for the epoch of the observation should be used in spectral fitting.
- V2 PIN background files selected periods (Revised
2008 Jun 17)
- There are now two types of PIN non X-ray backgroun (NXB) files, "tuned" and "quick." The tuned model has an estimated systematic uncertainty of 1.3%, which is significantly better than for the quick model. Therefore, the tuned NXB files should be used whenever available.
- However, production of the tuned NXB files takes 1-2 months after the processing of observations. The quick NXB files can be produced with a much shorter delay of 1-2 weeks. Therefore, the HXD team will continue to produce NXB files using both methods.
- Dead-time correction is not necessary for the tuned or the quick NXB files. In this respect, the tuned NXB files differ from the V1.2 bgd_d NXB files, even though both are based on the same family of models.
- V2 GSO gain correction for selected period (Added 2008 Jan 16)
- The HXD team has discovered that the GSO gain correction using the current version of hxdpi shows a small discontinuity around 2006 March 3 - 2006 May 9; the energy scale of reprocessed data becomes different by several % from that of response and background. HXD team is now developing the new version hxdpi.
- Reproducibility of HXD non X-ray background (NXB) (Updated 2007 Dec 27)
- Current reproducibility of the PIN NXB is +/-3.2% (1 sigma) in (15-40 keV) (see Suzaku Memo 2007-09). The effect of this should be taken into account in the data analysis. The Technical Description for proposers contains a recipe for how one might do so.
- As to the GSO background, a similar description will be prepared after the release of the V2 NXB files.
- Cosmic X-ray background (CXB) in HXD data
- The background files provided by the team only accounts for the NXB component. Whereas the cosmic background makes a very small contribution to the GSO data, it is not negligible for PIN data. We have provided a recipe for including the CXB in the data analysis.
2. Software Issues
- HXD: hxdtime and hxdpi on Fedora (Linux) machines (Updated Aug 1, 2008)
- When compiled from source using GNU 4.1.2 suite of compilers on Fedora 7 machines, hxdtime and hxdpi may fail with a segmentation fault at the end of the run. On the other hand, this problem is not seen Scientific Linux machines. It now appears likely that this problem is common to Fedora 7, 8, and 9 machines using GNU compilers v4.X; it can be avoided by compiling HEAsoft using GNU 3.x compilers. For more, consult the HEAsoft bug list page.
- XIS: xissim and xissimarfgen (Added Jul 31, 2008)
- The 2008-07-09 release of CALDB requires these two tools, both of which rely on the raytracing library, to be updated to those found in Version 8 or higher of Suzaku FTOOLS. The new software is backward compatible with older calibration files. However, older versions of software will not run with the latest CALDB.
- XIS: xispi with remote CALDB access (Updated Jul 31, 2008)
- The bug in xispi reported below has been fixed with the
Version 9 release of Suzaku FTOOLS.
The older version of xispi has a bug in the code that judges whether PARARELL_CTI parameters should be re-read from the calibration file, for observations with SCI. The bug makes xispi read the calibration file many times unnecessarily. While the results of the calculations are correct, there is a high overhead particularly with remote CALDB access.
- The bug in xispi reported below has been fixed with the
Version 9 release of Suzaku FTOOLS.
- XIS: arf generation
- The software to generate ancilary response files (ARFs) for XIS is based on the ray-tracing software. We provide hints on how to speed up ARF generation.
- Updated 2008 Jul 31: in the simple case (point source centered on the extraction region), it is simplest to generate the response when extracting the spectrum within xselect. Simply use the resp=yes option with the extract spectrum command. This will leave a combined .rsp file, having run both xisrmfgen and xissimarfgen.
- XIS: rmf generation
- While the process of generating XIS rmf is fast, there are tricks one can use in this stage, or immediately thereafter, to speed up the spectral fit.
If you have any questions concerning Suzaku, visit the Feedback form.