Things to Watch Out for When Doing ASCA Analysis (Updated 2001 January 5)
Users who have some familiarity with ASCA data analysis often ask about the key things to watch out for, especially since the ASCA calibration is time-dependent. This page is not meant to replace the "ASCA Data Analysis Guide" but rather serves as a quick reference which points out the most serious pitfalls and will be updated regularly. Much of the following is based on FAQs to the ASCAHELP line. This page will also be useful for people with no familiarty with ASCA who don't necessarily want to analyse ASCA data but would like a quick summary of "caveats" for ASCA data analysis. An example of such a person would be someone who is refereeing an ASCA paper but has no experience with ASCA. For more techincal details on some of these problems please consult the ASCA calibration uncertainties page for details.
WARNING: SIS data below 1 keV, taken after late 1994 should now be interpreted with caution.None of the current software corrects for a decrease in efficiency of SIS0 and SIS1 below 1 keV (up to about 20%), so errors in continuum parameters and/or column density will result. This also creates problems for background subtraction. See item  below for details.
Bright or Weak source?First decide whether you should be worrying about calibration uncertainties or background subtraction systematics or both. Although the answer depends on the spectrum of your source, very generally speaking, if your source is less than ~0.1 ct/s then you likely do not need to worry about calibration uncertainties. Your results will be dominated by background-subraction systematics. For very weak sources (~0.01 ct/s or less) do not be surprised if you get significantly different spectral results for SIS and GIS. This kind of count rate is at the limit for spectroscopy with the GIS. This is because the background becomes comparable to the source and also source photons may fall on the GIS support grid (an energy-dependent effect). Neither of these two effects can be accurately accounted for.
Most of all, use common sense!
Things to be Concerned About
Be aware of the following:
1] AscaarfThe default for ASCAARF is to apply both XRT empirical effective area factors, which reduce residuals to Crab spectral fits to 3% or less (i.e. ARFFIL=yes and FUDGE=yes); see ASCA calibration uncertainties page for details.
2] Increasing Divergence of SIS and GIS spectraBelow approximately 1 keV SIS0 and SIS1 spectra since approximately late 1994 have been showing an increasing divergence from each other, and from the GIS data, towards lower energies. This is in the sense that both SIS0 and SIS1 efficiencies below 1 keV have been steadily decreasing over time. This loss in efficiency is currently NOT corrected by any of the software. In particular, CORRECTRDD does NOT solve this problem yet. At 0.6 keV the discrepancy between SIS1 and GIS2, GIS3 can be as much as 40% for data taken in January 2000. It is important to realise that this problem is present 1-CCD mode (and worse in 2 and 4-CCD modes). It is believed that the problem is due to increased residual dark current levels in the SIS and a decrease in Charge Transfer Efficieny (CTE) and an effort to model the effects are currently underway. Meanwhile, one can estimate the magnitude of the effect and apply a crude, empirical "correction" in terms of an excess "effective absorption" column density. Full details can be found in this report.
A related problem is of course, that the released blank-sky data cannot be used for background-subtraction for later SIS data since that blank-sky data was taken early on in the ASCA mission. We are working on putting together some blank-sky data suitable for use with data from all epochs.
Other SIS/GIS DiscrepanciesMany people report discrepencies between SIS and GIS even for "old" data which did not suffer from the problem above. Most commonly, significant differences between the SIS and GIS slopes and/or column densities. Before reporting any such discrepancies to us, please bear in mind the following. You cannot compare SIS and GIS column densities of the order of 10^21 cm^-2 or less because the GIS is not sensitive to such small column densities. Check the background subtraction- certain mistakes in the analysis (e.g. defining an extraction region which goes outside detector boundaries) can lead to incorrect BACKSCALE factors. Also, subtelties with fitting complex spectra separately to SIS and GIS may result in *apparent* differences because of the important differences in the responses of the two types of instrument. Also, for very weak sources, of the order of ~0.01 ct/s spectral differences can be quite large, as this is at the limit of spectroscopy for the GIS due to background and other systematics.
3] SIS PH to PIThe calibration file to convert SIS PH values to PI is based on measurements from Cas A; for GO observations made after the date of the last Cas A observation used in the file, the relevant values used for PH->PI conversion are based on linear exptrapolation. The file currently used is sisph2pi_110397.fits which means it was made in March 1997 (the actual Cas A measurements were made in July 1996). Therefore PH to PI conversions for all observations after July 1996 are based on linear extrapolation. It now appears that the conversion factors are actually evolving non-linearly with time, resulting in a significant gain miscalibration for recent data; the SIS team has released a workaround.
Note that non-standard chips are less well calibrated than the nominal ones (i.e. SIS0/chip 1 and SIS1/chip 3). There is also a dependence of the PH/PI relation on CCD mode, which is currently being investigated.
4] Low-energy SIS featureLine-like residuals have often been reported at ~0.5-0.6 keV in the SIS. Since this is right at the bottom end of the SIS response and the feature seems to be transient, it has not so far been possible to quantitatively study this effect. Many people now simply ignore the data below 0.6 keV. However, the data at the loweset energies carry little statistical weight so it may be more informative to include these data, or at least extrapolate the model down to 0.5 keV and show all the data in your paper as it may someday assist in quantifying the problem.
5] RDD effect in SISThe CORRECTRDD tool is experimental and, moreover, does not currently correct for a loss in the low-energy efficiency in the SIS. Click here for further details. Note that in 1-CCD mode the RDD effect (from Monte-Carlo simulations) is supposed to be negligible (see report by Dotani-san for details). Therefore, the *observed* loss in SIS efficiency (which has been getting worse in time since 1994) must be due to another cause (see item  ).
6] Calibration of 2-pixel EventsIt is known that the SIS response generator (SISRMG) is not accurate for grades 2,3,4 when the event threshold is high. The observed effect is that the grade 2,3 and 4 data lie below the grade 0 data at low energies - click here for details.
7] Lower Level DiscriminatorCheck whether the SIS lower-level discriminator was not changed during the observation. If it was, you need to make separate response matrices for the different settings. You can check the values of the lower-level discriminators in each event file by examining the value of the keyword S0_LVDL1 and S1_LVDL1 for SIS0 and SIS1 respectively.
8] Background SubtractionWhen using the released, blank-field backgrounds (or indeed background from some other blank field, from a non-detected source observation, say) bear in mind that the contribution from the Galactic background is different in different parts of the sky so this can affect your low-energy results.
Extended sources will particularly be subject to background systematics and caution should be excercised.
Also bear in mind the point made in item : released blank-sky data used for background subtraction on data taken after late 1994 (approximately) will give OVER-SUBTRACTION below 1 keV because the SIS low-energy efficiency has been decreasing with time and is not corrected by the software at present.
If you have any questions concerning ASCA, visit our Feedback form.