In practice, arfgen does not use the input detector map 'as is'; instead it uses a filtered form of the detector map in which the points that lie outside the regions used to accumulate the input spectrum are excluded.
A consequence of this is that the user must ensure that the source regions lie within the bounds of the input detector map, otherwise the output ARF could be grossly inaccurate. The task performs a crude check to see if regions are compatible with the detector map, and issues a warning if this is not the case. The user is welcome to be conservative and create a detector map over a very large window, enclosing the regions plus a lot more. In this case the number of image map pixels should be correspondingly greater to ensure a good coverage over the region of interest, which will result in a performance overhead during the initial filtering process.
The filtered map is available as a by-product,
by setting withfilteredset 
true
and specifying a name for the output dataset
via the parameter filteredset. The map
is provided in pixel list format.
The filtering operation can be switched off,
if desired, by setting filterdss false.
In this case, the regional information is ignored,
and all pixels in the input detector map
are written to the 'filtered map', which
in turn means that all pixels are
considered during the ARF computation.
arfgen supports region information specified in terms of DET or POS (sky) coordinates. In the former case, the region specification can be used to filter the DET-based detector map directly. In the latter case, an additional step is necessary to transform the region specifications to the DET system. The transformation is performed by making a call to the task attcalc.
