This chapter describes some aspects of the GIS performance which users should be aware of when reducing and analyzing GIS data.
For the latest information about the GIS, please consult the Web pages, `GIS News' at the URL
and `Calibration Uncertainties' (§1.7)
Although the GIS has four modes, in practice the pulse height (PH) mode has been used for virtually all scientific observations. The multi-channel pulse count (MPC) mode was used occasionally at the beginning of the mission. The FITS keywords which pertain directly to the GIS modes are described in §3.4.
In this mode, the on-board CPU calculates the event position and discards background events using a specially designed event selection algorithm. Only accepted X-ray events are sent to the ground. With the standard bit assignments, Pulse-height spectra consist of 1024 channels for each detector.
PH mode allows some flexibility in how the telemetry is allocated. There are 31 bits per event to describe the PHA value, X and Y position, rise time, spread and timing. Note that changing the bit assignments in PH mode only changes the digital resolution, i.e., the binning, of the data. For example, the standard 1024-channel PH mode spectral responses can be used for 256-channel PH mode spectra, provided the response has been rebinned (using the FTOOL rbnrmf) 3.1.
In the following table, the various alternative choices are given in parentheses, while the standard values are not:
|=0 for nominal setting.|
The standard temporal resolution (timing bit=0) is 62.5 msec (in HIGH bit rate) and 500 msec in MEDIUM bit rate: using bits for timing makes the temporal resolution times better. It is often useful to increase the number of timing bits to achieve higher time resolution. For example, for several observations of fast X-ray pulsars, the following bit assignments were used and found effective:
|8(PH)-6(X)-6(Y)-5(RT)- 6(Timing)||in high bit rate||(1.0 ms resolution)|
|8(PH)-6(X)-6(Y)-0(RT)-10(Timing)||in medium bit rate||(0.5 ms resolution)|
The bit assignment in a PH mode events file is described by the header keyword, `TIMEDEL'.
Telemetry capacity limits in the PH mode are 128, 16, and 4 cts/s/sensor for the HIGH, MEDIUM, and LOW telemetry rates respectively.
This mode is optimized for bright sources, trading temporal resolution and telemetry capacity for positional information: only pulse-height information is recorded - in the form of histograms, similar to proportional counter data. This is also a back-up mode against CPU failure, since the data are processed without the intervention of the CPU. Pulse-height spectra consisting of 256 channels for each time bin are produced. Alternatively, each time bin can contain a set of N spectra of 256/N channels for N equal portions of the time bin.
You can extract light curves and spectra, but not images, for MPC mode data using XSELECT. The lack of positional information (for the spatial gain calibration) makes analyzing MPC data problematic. Spectral response matrices do not exist for MPC mode, so accurate spectral analysis is not possible for the general user.
A single GIS event invokes electric pulse-heights on the 16 16 anode wires of the Imaging Photo Multiplier Tube. The GIS calculates the event position as the centroid of these 16 16 pulse-heights. As a consequence of the position determination algorithm, the GIS has a limited spatial resolution. The point spread function (PSF) of the GIS alone is a Gaussian with a FWHM of arcmin, where is the energy in keV. Note that this is smaller than the 3-arcminute half power diameter of the XRT PSF, but at soft X-ray energies is larger than its sharp core (the FWHM of the XRT PSF is 50 arcsec). As a result, the convolved PSF of the XRT+GIS becomes much broader than that of XRT alone, and the original detailed structures in the XRT image are smeared in the GIS image. Studying small structures on a scale of an arcminute, which may be carried out with the SIS, is very difficult with the GIS.
Although the distributed GIS data contain all the events from all over the detector (50 arcmin diameter), the GIS usable Field of View (FOV) is limited by several factors. Firstly, most non-X-ray background events in the GIS occur close to the walls of the detector, i.e., at the edge of the FOV. Secondly, there is an internal calibration source which emits monochromatic X-rays and which appears at the edge of the GIS FOV. Third, not only is background high, but the gain is also less accurately known at the edge of the FOV. Indeed, because of the high background and uncertain gain, only events within the central 44 arcmin diameter of the GIS FOV are aspected (assigned RA and DEC). It is a standard practice to exclude data taken outside of a central diameter of 40 arcmin. For more details, see §5.5.2.
In xselect, once the datamode has been set, the various mode-dependent quantities, such as the number of energy channels, are set automatically. It is, however, useful to know what the corresponding keywords are in the GIS data file.
PHA_BINS= 1024 / number of bins for PHA RISEBINS= 32 / number of bins for RISE_TIME TIMEBINS= 1 / number of bins for TIME SP_BINS = 1 / number of bins for SPREAD RAWXBINS= 256 / number of bins for RAWX RAWYBINS= 256 / number of bins for RAWY
Note that the values in this example correspond to the standard bit assignment. Furthermore, when RISEBINS=1, the file contains no rise time information; consequently, background rejection based on the RTI, or invariant rise-time, using gisclean, cannot be performed.
HV_RED = 'OFF ' / HV-reduction on/off HVH_LVL = 3 / HV-High level (0 - 7) HVL_LVL = 4 / HV-Low level (0 - 7)
The values given here correspond to valid data: files which have different values are garbage, and are recognized as such by XSELECT which does not, by default, include them in `obscats' (observation catalogs).
In the GIS calibrated event files, each event is given a PI value which is proportional to the input photon energy. In the process of calculating PI values from original PHA, instrumental gain is determined, which is a complicated function of time, detector position and temperature.
Details on the GIS gain determination method are explained at the following page:
In the standard GIS processing, nominal accuracy of the GIS gain determination is +/- 1 which is considered to be the GIS absolute energy accuracy.
So far, it has been confirmed that the GIS gain determined by extrapolation in the standard processing is consistent with that determined by interpolation. However, if users want to apply the latest gain calibration when it is made available, follow the following procedure (see also the URL address above):
The update_gis_gain.pl script creates a new gain history file using gis_temp2gain.fits by running the temp2gain FTOOL, followed by ascalin to recalculate the PI values of the GIS event files using the new gain history file.
It is possible for Guest Observers to manually adjust the gain and redetermine the PI values using ascalin.3.2
Here is an example to re-determine PI values that are 0.978 (=1/1.022) times the original values, for the events file ad43001000g300370h.evt:
ascalin datafile=ad43001000g300370h.evt calfile=caldb tempofile=\ ft950908_0323_0800.ghf attitude=none gainnorm=1.022
Here, ft950908_0323_0800.ghf is the gain history file which is in the aux directory of the data distribution package or the ASCA archives. The plausible gain normalization factor has to be determined independently by, for example, using the XSPEC gain command. Note that the new PI values will be the old PI values divided by the input gainnorm. This is made so that the output of the XSPEC gain command can be directly used here. Instrumental spectral features at the gold M-edge (2.2 keV) and Xenon L-edge (4.8 keV) may be used as fiducial marks for the gain adjustment.
Also, please consult the `GIS Gain Correction' Web page at the URL
for more details and examples.
Here are some other important points to bear in mind with the GIS.
The data taken with the standard bit-assignment have 128 spectral channels, as opposed to the ordinary 1024. Please consult the original announcement from the GIS team for details on this problem. This is available at
For spectral analysis of the data taken during this period, RMFs with 128 channels have to be used corresponding to the 128 channels of the PHA data. The 128 channel RMF is available in the ASCA calibration database at 3.4
When making an ARF with ascaarf, the 128 channel RMF should be used so that the ARF will have 128 channels.