The original ASCA data collected on-board include `bad' data which have to be rejected in the course of data analysis. In this context, `bad' means that such events are inappropriate for scientific analysis, for a number of different reasons which will be explained in this chapter. Standard data screening (see §5.5) is carried out during processing at GSFC, and the cleaned events lists are found in the `screened' directory on an ASCA distribution data set or in the ASCA Archive (they have names of the form *.evt). The standard data screening criteria have been found by the ASCA GOF to provide a good balance between rejecting bad events without overly compromising signal-to-noise, and convenience (providing a manageable set of criteria).
Although one can start from these pre-screened events lists to extract images, light-curves and spectra, you may want to apply your own data screening criteria, which may be looser or tighter than the standard criteria. Unscreened data are also distributed, the files having names of the form *.unf and can be found in the unscreened directory (distributions earlier than REV1 also had `RAW' data files in the raw directory).
Even if you are not going to apply your own data selection criteria, it will be helpful to understand the kind of data screening that should normally be applied to your data before extracting scientific products. Here we describe the individual criteria and how they are used to remove bad events. We also suggest how to tailor the criteria to match individual analysis priorities. Firstly, screening criteria common to both GIS and SIS are described. Then, criteria unique to the GIS and SIS are discussed separately.
There are three methods available to perform your own customized screening. These are implemented with the following programs:
xsel:ASCA-GIS2-PH > select mkf "ACS==0 && COR>4 && ELV>5" > Enter the filter file directory >[.]
The meaning of these parameters will be explained below. Screening criteria may be written in Fortran style (e.g., ELV.gt.5) or in C style (ELV5). The select mkf command prompts the user to enter the directory where the mkf file or files is/are located.
More technical details of screening will be given in chapter 6. In this chapter we concentrate on the screening criteria themselves.
Recently, the GIS team conducted a systematic study of the characteristics of the GIS internal background. This has lead to a better understanding of screening criteria that should be used, in the sense of increasing the source signal-to-noise by avoiding the rejection of too many good events. A summary of this study and the resultant new screening criteria is given in section 5.3.1.
In this section, we explain data screening criteria common to both the GIS and SIS.
From the satellite, the angle between the Earth's limb and the pointing direction is known as the elevation angle. At low elevation angles the target is viewed through the Earth's outer atmosphere which absorbs and/or scatters X-rays and hence distorts the spectrum. Negative elevation angles in the mkf file identify Earth occultations. We have found that data quality degrades with elevation angles less than 5 degrees. At angles greater than 10 degrees, the data quality no longer depends on elevation angle. This means that users can experiment with setting the minimum acceptable elevation angle at 5 degrees (lax) or 10 degrees (strict), depending on whether they want to boost signal-to-noise at the expense of possibly compromising data quality.
Note that sensitivity to elevation angle for a given observation depends on target position. Targets close to the ecliptic poles are the most susceptible, as the elevation angle can remain modest throughout the observation.
The elevation angle is recorded in the mkf file in a
column named ELV.
(Note that for REV0 distributions ELV was known
Hence the standard elevation angle criterion you might use can be written as
ELV 5 .
ELV 5 .
If the pointing of the satellite fluctuates too much, then photons from the same part of the sky are not incident on the same part of the detector. Since the response depends on detector position, this could lead to inaccurate results. To avoid this happening, only data taken during stable pointing should be used. Also, it is possible to select events based on the deviation from the mean pointing. The mkf parameter ACS should be zero when the satellite is in pointing mode. The mkf column ANG_DIST contains the root mean square deviation of the instantaneous pointing from the mean pointing. Setting ANG_DIST to be less than 0.01 degrees (0.2 times the HPD of the XRT) is a sensible default.
In such cases that there is significant deviation from the mean pointing, it is usually mostly limited to the first 1000-2000s of the observation, before the star tracker fine-tunes the pointing. Deviations after this are usually negligible. For some short observations, where the mean pointing is strongly affected by this initial segment, a larger value of the ANG_DIST threshold may be advisable. (Please note that the ANG_DIST column is not present in REV0 mkf files).
The ANG_DIST values should be positive by definition as long as the
attitude is correctly determined.
Hence, in order to select stable pointing data,
the selection criteria you might use
can be written as,
ACS==0 && ANG_DIST 0 && ANG_DIST 0.01.
ACS==0 && ANG_DIST 0 && ANG_DIST 0.01.
The South Atlantic Anomaly (SAA) is a `hole' in the geomagnetic field which
allows cosmic rays to penetrate further than usual. The particle background
in the SAA is extremely high. In fact, the high voltage of the GIS is
usually turned off during passage through the SAA,
but in the SIS the rate is not so high that the SIS is turned off.
SAA passages can be explicitly excluded by selecting on
the mkf column SAA:
requiring that the value be zero excludes SAA passages.
Hence, the mkf expression to choose data outside the
SAA is written as,
Note that this will have practically no effect for GIS data, since the GIS should always be turned off during SAA passages.
Cut-off rigidity (COR) is a local measure of the ability of the geomagnetic field to repel cosmic rays. Specifically, it is the minimum momentum (in units of GeV/c) with which a cosmic-ray particle can penetrate as far as the satellite orbit. Since cosmic rays induce instrumental background, low values of COR identify those parts of the orbit which have high background. For the SIS, a value of 6 GeV/c yields plenty of good events without seriously compromising data quality. For weak sources, less than cts/s the user may want to experiment with higher values of COR to get the best compromise between signal-to-noise and low background at high energies. For the GIS, it has been found that the COR threshold can safely go down to 4 without being affected seriously by background events, IF this is combined with other appropriate selection criteria (see section 5.3.1).
Note that since background is roughly proportional to COR, the actual spectrum used for background subtraction should have the same distribution of COR as the on-source spectrum. This will automatically be the case if the background spectrum is extracted from the same screened events list as the on-source spectrum. If, on the other hand, blank-sky background is used, the COR distribution should be forced to match.
A decrease in the COR
threshold from 6 to 4 GeV/c will typically result in a 20%
increase in the net exposure time,
while the increase in background may not be a
serious problem, depending on the source intensity and the type of analysis
Thus the lower threshold that is
ultimately adopted is
up to you, but the selection
criterion may be written as
The cut-off rigidity is recorded in the mkf files in a column named COR. (Note that in REV0 mkf files, the corresponding column name is COR_MIN).
Before explaining screening criteria unique to the GIS, it will be helpful to become familiar with GIS internal background characteristics. The description consists of excerpts from the Web page at the URL
and from the article `Reproducibility of GIS non-X-ray background' in the ASCANews #5.
The GIS intrinsic (particle) background is known to vary with the orbital position of the satellite. Also, there are transitory particle events which should be excluded in data analysis. Three useful quantities which can be used to get a handle on the GIS intrinsic background and to select good time intervals with are:
The Radiation Belt Monitor (RBM) is a small PIN diode detector attached to the bottom of GIS2. RBM is sensitive to particle background and used to automatically reduce the high-voltage of the GIS sensors on-board.
GIS monitor counts are house keeping parameters to count all the events (X-rays or particles) which hit the GIS pulse-height lower discriminator. Details can be found at the Web page with URL
Among the six GIS monitor count-rates for each sensor, H0 and H2 are considered to include very few X-rays and hence are good particle monitors. X-rays are included in the L1 monitor counts, so L1 must be positive as long as the observation is carried out correctly. Below, the sum of the G2_H0, G2_H2, G3_H0 and G3_H2 monitor count rates shall be referred to simply as `H02'.
Data taken from 59 blank-field observations between June 1993 and August 1994 were studied. The distribution of the RBM and H02 counting rates during ASCA orbits, along with contours of COR are shown in
which are also linked from the document
It is found that there is a good correlation between COR and H02.
Figure 5.1 shows this correlation between COR and H02 for these blank sky data. Despite the good anti-correlation, there are events having anomalously high H02 values at around COR . These are transitory particle events which should be rejected in data analysis. These flare-like events turn out to have hard energy spectra, so we call them `hard flares'. The hard-flare events, as well as the high background regions with high H02 and low COR values, can be easily rejected by throwing away data above the two following two broken lines in the diagram: H02=45.0 and H02= 0.45 COR -13 COR +125.
In figure 5.1, the bifurcate below COR=8 may be due to inaccuracies of the COR map we are using in the ASCA mission. On the lower branch, which corresponds to the low COR region around the east of Cuba and Florida, the H02 values are not as high as on the upper branch, in spite of the fact that COR 6.
In addition, another kind of transitory particle event is found, which we will call a `soft flare', The soft-flares have soft energy spectra and accompany increases in the RBM count-rate, but not H02. Hence they can be rejected using RBM criteria. It is found that there are two orbital regions where the soft-flares occur preferentially (regions surrounded by broken-dot lines in the figures
The RBM criteria should be stricter in these regions. Suggested, conservative criteria to reject `soft-flare' events are RBM 18.75 cts/s everywhere and RBM 6.25 cts/s in these two regions.
show the RBM and H02 count distribution after all the above three screening criteria (H02 45 cts/s, hard-flare cut and soft-flare cut) are applied. This can be achieved with the `strict' criterion in ascascreen (see §5.3.3).
Most non-X-ray background events in the GIS occur close to the walls of the detector, i.e., at the edge of the Field of View (FOV). Not only is the background high at the edge of the FOV, but the gain is also less accurately known. Indeed, because of the uncertain gain, only events within the central 44 arcmin diameter of the GIS FOV are aspected (assigned RA and DEC). Also, at the edge of the FOV in both detectors, internal Fe55 calibration sources are attached. These high-background rings and calibration sources have to be excluded when analyzing GIS data. This can be easily done by applying a spatial region filter in DETX and DETY coordinates.
Bona fide X-ray events occupy a tightly defined region in the plane formed by the gain-corrected - i.e., invariant - rise time (RTI) and the pulse height (PHA). The XSELECT command gisclean will reject events outside this region. Note that RTI is not always available. There are cases when the number of timing bits is increased to achieve high time resolution, and as a trade-off the rise-time bits are lost. In this case, gisclean cannot be applied, and the data will have more internal background than otherwise. It is important to note that the standard GIS response matrices assume that the data has had the RTI filter applied.
Firstly, the following screening should be applied to the GIS data. These are defaults in the GSFC REV2 processing and ascascreen (FTOOLS v3.7 and later). Standard REV1 screening is described in §5.5.
GSFC REV2 processing and ascascreen assumes the following region selections (in the case of 256256 spatial resolution). In terms of SAOimage region files, these spatial regions are described below. These are modified and improved comapred to REV1 (see §5.5.2).
In some instances you may want to modify these region selections. To do so, you can either edit the above region files manually or you can make a region file within SAOimage. This new region file can then be used with the XSELECT command filter region as in the following example:
xsel:ASCA-GIS3-PH > filter region g3_cal.reg
where g2_cal.reg is the name of the region file. Note that you specify the region in DETX and DETY coordinates (this is the default).
This is carried out by the XSELECT command gisclean without arguments:
xsel:ASCA-GIS2-PH > gisclean
This is performed as long as the GIS events files have the RTI column correctly populated. For high time resolution data which do not have rise-time information, this step is skipped.
After the above screening has been done, you can specify your own further screening criteria using mkf parameter expressions. The version of ascascreen later than that distributed with FTOOLS v3.6 should have three default GIS data screening criteria: `loose', `standard REV2' and `strict'. The `standard REV2' selection is used in the REV2 processing and is different to earlier processing versions (see §5.5 for standard REV1 selection criteria). Here, we explain these three screening criteria in ascascreen. Users can choose one of them, or specify their own data selection criteria.
Besides the screening criteria explained in section 5.2, no selection is made to reduce the GIS background. This will make almost all the GIS data available, and it is recommended that this option be used only for bright sources, for which the particle background is negligible compared to the number of source photons.
GIS X-ray events are included in the monitor counts G2_L1 and G3_L1, so these quantities must be positive as long as observation is going normally.
In addition to the loose screening above, the following conditions are further imposed for this option:
SAA==0&&COR>4 &&G2_H0+G2_H2+G3_H0+G3_H2<45 && G2_H0+G2_H2+G3_H0+G3_H2<0.45*COR**2-13*COR+125&&RBM_CONT<100
The first two items exclude regions with high particle background during the orbit. The third and fourth items reject the high background data and the hard flares respectively. The last, RBM_CONT, criterion excludes the strong soft flares. See section 5.3.1 for an explanation of the hard and soft flares.
Instead of the RBM_CONT100 criterion in the standard screening, we use stricter soft-flare rejection criteria as follows:
(RBM_CONT < 6.25 || (RBM_CONT <18.75 && !((SAT_LON > 200 && SAT_LAT < -16) || (SAT_LON > 255 && SAT_LAT < 10 && (SAT_LAT < 0.36*SAT_LON -97))|| (SAT_LON <250 && SAT_LON > 160 && SAT_LAT > 16)|| (SAT_LAT > 8 && SAT_LON<250 && SAT_LAT > -0.53*SAT_LON+123))))
This means that RBM_CONT should be less than 18.75 outside of the two regions where soft flares occur preferentially and less than 6.25 inside the two regions.
Hot and flickering pixels, which appear as false events, are a manifestation of radiation damage to the CCDs (see §4.4). Although they are unavoidably included in telemetry, they can be straightforwardly removed on the ground. The sisclean algorithm rejects those pixels which register an event more often than expected from Poisson statistics. Since they contain no astrophysical information and are not accounted for in the available response matrices, hot and flickering pixels should always be removed (see §4.4.1).
The number of hot and flickering pixels depends on epoch (the problem has worsened since launch) and on clocking mode (more of a problem in 4-CCD mode than in 1-CCD mode).
Grade is a one-dimensional description of the shape of the charge cloud created by an event in a CCD. Certain grades have better spectral resolution than others (grade 0 has the best resolution) or are more likely to be bona fide X-ray events (grades 1, 5 and 7 contain mostly non X-ray events). A combination of grades 0, 2, 3 and 4 are the best-calibrated, and this combination provides a good compromise in resolution and signal-to-noise. However, there are two important cases where users might want a different selection. First, FAST mode data have different grade assignments: only grade 0 events should be selected. Second, since analyzing a light curve does not require a spectral response matrix, users can increase the signal-to-noise of SIS light curves by including grade 6 events.
In addition, please note that:
As explained in section 5.2.1, ELV is considered an appropriate criterion to reject data contaminated as a result of being taken near the Earth rim. In the case of the SIS, which is also sensitive to optical and UV radiation, data quality is further impaired when the FOV is close to the illuminated face of the Earth, by light scattered within ASCA XRT. This necessitates two elevation angle criteria: one for when the Earth's limb is dark, and a second, stricter one for when the limb is bright.
The Bright Earth elevation angle is contained in a column named BR_EARTH in the mkf file. For the bright limb, the threshold for the parameter BR_EARTH is clocking-mode dependent. For 4-CCD mode data, the elevation angle can be set in the range 15-40 degrees for SIS0, with 20 degrees being a sensible default. For SIS1, the recommended range is 15-20 degrees. For 1 and 2-CCD mode data, lower values of the threshold (down to 10-15 deg) appears to result in no noticeable degradation of data quality. However, it is best to experiment for each observation.
To avoid telemetry saturation and photon pile-up, the SIS is not used to observe very bright objects. This means that when high count rates do occur in the SIS, it is often the result of occurrences such as passing through the SAA or observing the bright limb of the Earth's atmosphere. The problems are accentuated by the existence of light leaks in the SIS. The mkf column Sn_PIXLm records the number of pixels per readout (REV0) or per second (REV1 and above) which exceed the event threshold in SISn/chipm. By imposing an upper limit on Sn_PIXLm, these non X-ray peaks can be rejected. In practice, however, the same default settings are used for all eight chips and depend only on the readout time (i.e., on clocking mode). In the standard GSFC processing the threshold values used for 1, 2 and 4-CCD modes are 100, 75 and 50 respectively.
There are several effects to watch out for:
Strictly speaking, if one needs to obtain the most accurate fluxes, telemetry-saturated frames should not be used since the count-rates will be incorrect. Even if the source is relatively weak, hot and flickering pixels may still cause telemetry saturation in the SIS. If the priority is to obtain a high signal-to-noise spectrum and one is not too concerned about accurate fluxes then telemetry-saturated data can be used. In any case, it is a good idea to compare results with and without telemetry saturation.
The mkf files for REV1 and thereafter contain a column called Sn_SATFm where n refers to the SIS (0 or 1) and m is the chip number. This parameter is zero during periods of NO telemetry saturation. Therefore, unsaturated data is selected by specifying Sn_SATFm==0.
To derive the true pulse-height in each pixel, the dark-frame level (see §4.6.2) is subtracted from the total pulse-height. The dark frame level varies across each chip and depends on the radiation environment of the instrument. Because of these dependencies, the on-board computer calculates a 16 x 16 map of the dark frame level for each chip and updates the map periodically. However, when the satellite clears the SAA, radioactive decays persist for a few seconds, raising the dark frame current. This elevation of the dark frame level occurs so rapidly that the dark frame map is not accurately calculated until about 4 readout times have elapsed (i.e., 16, 32 and 64 seconds for 1-CCD, 2-CCD and 4-CCD mode, respectively). The mkf column T_SAA gives the time in seconds after a passage through the SAA. Note that before the first SAA passage of an observation, T_SAA is negative. This means that for a 1-CCD mode observation, for example, the appropriate setting is that T_SAA be less than zero and greater than 16. Users should also be aware that mkf files have 32-s bins.
Since an incorrect dark-frame level affects light curves less than spectra, users extracting light curves only might want to omit the T_SAA criterion (as well as the T_DY_NT criterion below).
To derive the true PHA in each pixel, the dark frame level (see §4.6.2) is subtracted from the total PHA. Dark frame varies across each chip and depends on the radiation environment of the instrument. Because of these dependencies, the on-board computer calculates a 16 x 16 map of the dark frame for each chip and updates the map periodically.
The dark-frame map described above is also inaccurate when the the satellite crosses the Day-Night boundary, again because the dark-frame level changes so rapidly. The mkf column T_DY_NT gives the time in seconds after a passage through the Day-Night or Night-Day transition. Note that before the first Day-Night Transition in an observation, T_DY_NT is negative. This means that for a 1-CCD mode observation, for example, the appropriate setting is that T_DY_NT be less than zero and greater than 16. Users should again be aware that mkf files have 32-s bins. As with T_SAA, light curves are less affected than spectra, if the dark-frame error is not corrected with T_DY_NT.
The version of ascascreen in FTOOLS v3.6 does not apply the correct value of T_SAA and T_DY_NT. Instead of converting 4 readout times into the appropriate number of seconds, ascascreen always sets T_SAA and T_DY_NT to be greater than 4 seconds, regardless of clocking mode. This is not a serious bug because applying the T_SAA and T_DY_NT criteria, even correctly, does not screen many events. Since ascascreen does not prompt the user to set T_SAA or T_DY_NT, the way around the problem is to edit the _mkf.sel file. Specifically, you should first run ascascreen as usual, but type ascascreen -q instead of ascascreen. This will cause ascascreen to stop before running XSELECT. Second, type ls to identify two files which ascascreen has just produced: the _mkf.sel file, which contains the screening criteria, and the .xco file, which contains the XSELECT commands. Third, use an editor to overwrite the correct values of T_SAA and T_DY_NT in the _mkf.sel file (16, 32 or 64 seconds for 1-CCD, 2-CCD and 4-CCD modes, respectively). Finally, run XSELECT; then after choosing a session name, type @ followed (without a space) by the name of the .xco file. By following these steps you are effectively running ascascreen as usual but with an interruption in the middle to correct the bug. The bug should be fixed in FTOOLS v3.7 and later.
Following is a summary of typical screening criteria to be applied to SIS data.
For convenience we summarize the standard screening criteria used in GSFC REV1 processing.
SAA==0 && ELV>10 && COR>6 && ACS==0 && ANG_DIST>0.01
Standard RTI-rejection applied, as implemented in gisclean. GIS outer-background ring and calibration source removed with the following SAOimage-style region files:
Remove hot and flickering pixels. Apply the following further screening criteria:
FOV>0 && BR_EARTH>20 && Sn_SATFm==0 && T_SAA>16M && T_DY_NT>16M && Sn_PIXLm<Twhere M is equal to 1, 2 and 4 and T is equal to 100, 75 and 50 for 1-, 2- and 4-CCD modes respectively.