By default the task will reduce the effective area produced in the ARF
if the source area contains bad pixels/columns or contains area which lies
outside the CCD boundaries. For an extended source this reduction is
simply the ratio of the dead area within the source region to the good area.
For point sources a fine grid is used to determine the contribution to
the enclosed energy of each dead point, i.e. the dead regions are
normalised by the fraction of the point spread function lying within that
region before being subtracted from
the total effective area. The task uses the parameter extendedsource
to choose between these options. Although the correction for point sources
is quite good the extended source correction assumes a uniform surface
brightness of the source. If this condition is not met then quite large
errors can be introduced. A way to avoid this potential source of error is
to apply a detector mask in the accumulation of extended source spectra.
The detector mask is a binary (1/0) detector image, where all CCD "dead"
areas are set to 0. The masked out pixels are not used in the calculation
of the effective area (see
section 5.2.1).
Hence, the integrated effective area over the extended source surface brightness takes into account only the fraction of the source which has been
effectively imaged.
A detector mask is generated with emask
starting from an exposure map
as generated by eexpmap, which in turn requires an image generated
by evselect
or xmmselect. In command-line terms:
eexpmap imageset=field_of_view.img attitudeset=0060_0122700101_AttHk.ds
eventset=0060_0122700101_EPN_S001_ImagingEvts.ds
expimageset=exposure_map.fits pimin=1000 pimax=10000
emask expimageset=exposure_mask.fits detmaskset=detector_mask.fits
If the fraction of counts, lost to bad pixels and chip gaps is large, then it is recommended to use a finer resolution when determining this contribution. A warning is issued by the program if it sees that the resolution used is too coarse.
In a similar way the task corrects for area of the source region which is outside the FOV (selectable with the parameter ignoreoutoffov), outside the observing window or on a CCD explicitly excluded by a CCD selection within the data subspace.
This behaviour can be turned off by setting withbadpixcorr=false. CCD gaps and area outside the FOV are automatically corrected for if withbadpixcorr=true. To correct for bad pixels, a file containing bad pixel extensions must be specified on the command line by badpixlocation=file. The bad pixel information is usually stored in the event file from which the spectrum was constructed. Bad columns and bad rows (MOS only) are also handled if they are specified in the OFFSETS column of the badpixlocation file. Pixels adjacent to chip gaps, bad pixels and bad columns are also corrected for if the FLAG selection in the datasubspace has been set accordingly. The code checks each bit mask in the event flag to see if pixels next to a particular type of bad pixel or bad column should be excluded. In this way spectra created with, for example, the #XMMEA_EM and #XMMEA_SM flagging can produce different effective areas for certain selection regions.
We recommend users to utilize #XMMEA_EM for accumulating EPIC-MOS spectra. Spectra extracted using this selection criteria have been demonstrated to be in agreement with the EPIC calibration status document (http://xmm2.esac.esa.int/docs/documents/CAL-TN-0018.pdf).
The task creates temporary files called BADPIXnn.ds, where nn is the
CCD number, in the current directory while calculating the bad pixel
correction. To make these files permanent specify
withfilteredset 
yes
on the command line.