SAS Watchout

Can I use a box region to extract EPIC-pn Timing or Burst mode spectra?

EPIC-pn fast mode data is produced with a RAW-X and a RAW-Y axis and the SAS allows spectra to be produced from these spatial axes using any of the shapes (circle, ellipse, box etc.) that may be used with imaging mode data. However, for the EPIC-pn fast modes, SAS versions 15 and earlier only support the calculation of an effective area file (ARF) from a very limited range of regions as explained below.

Timing mode

Source and background spectra should be extracted using a RAW-X range only.

e.g. for a single contiguous region

     RAWX in [28:32]

or for a region which excludes the central columns (to avoid pile-up or to extract the background)

     RAWX in [20:27] || RAWX in [33:40]

See the thread:

     http://www.cosmos.esa.int/web/xmm-newton/sas-thread-pn-spectrum-timing

for more details.

Do not use a RAW-Y range or extract the data with a box region such as:

     (RAWX,RAWY) IN box(36.5,89,2.5,83,0)

as this is not necessary and may result in the production of an incorrect ARF.

Burst mode

The top 20-40 RAW-Y rows of Burst mode data are contaminated by direct emisison from the point source (see Kirsch M., et al., 2006, A&A, 453, 173) and should be removed.

The spectrum may be extracted using:

     RAWX in [33:41] && RAWY<=140

Currently this is not fully supported by the arfgen / backscale programs and will produce ARF and BACKSCAL values which are 180/140 times too large. The BACKSCAL error is not important if a region with the same RAW-Y range as the source region is used to extract background data, e.g.

     RAWX in [2:8] && RAWY<=140
 

In this way the scaling factor used to subtract the background in a spectral fitting program will be correct.

The error in the ARF will produce fluxes which are 180/140 times smaller than they should be. This error is energy independent.

Alternatively spectra may be extracted using a box:

     (RAWX,RAWY) IN box(37,70.5,2.5,70..,0)

This will produce BACKSCAL values which are 200/140 times too large and a correct ARF. Again, as long as the background is extracted from a box with the same RAWY dimensions as the source region, then this will not be a problem.

However, great care must be taken with the box dimensions as arfgen will use the box width naively to calculate the extraction region. e.g. for a box

     (RAWX,RAWY) IN box(37.5,70.,2.5,70.,0)

it will assume a RAWX width of 5 columns (2.5 is the half-width in the box expression). In reality, the spectrum will have been made from columns 35,36,37,38,39 and 40, i.e. 6 columns. This will result in not only the wrong BACKSCAL value but also the wrong effective area.

Similarly

     (RAWX,RAWY) IN box(37,70.,2,70.,0)

will give the effective area for 4 RAWX columns whereas the data has been extracted from 5 (35-39 inclusive). For this reason boxes should always be defined using an integer centre and a fractional width or vice-versa, e.g. supported boxes are:

     (RAWX,RAWY) IN box(37.5,70.5,2,70.,0) or
     (RAWX,RAWY) IN box(37,70.,2.5,69.5,0)

In the first case this will correctly use the four columns, 36,37,38 and 39 and in the second case it will correctly assume a region with the five columns, 35,36,37,38,39.

Note that excluding central columns using two boxes by e.g.,

     (RAWX,RAWY) IN box(30,70.5,2.5,70.,0) || (RAWX,RAWY) IN box(38.5,70.5,2,70.,0)

is not supported and will produce the wrong result.