SAS watchout page

My BACKSCAL (or effective area) has crazy values: help!

When backscale or arfgen look at the spatial region used to extract a particular spectrum, they expect to find a region defined in POS(X,Y), DET(DETX,DETY) or EQPOS(RA,DEC) coordinates. However in some circumstances evselect or xmmselect produce a region with the units AXIS2D1(X,Y) or AXIS2D1(DETX,DETY). In general this axis is misinterpreted by the tasks, and causes them to calculate wrong values (often wildly wrong).

If bad pixel correction is switched on (withbadpixcorr=yes) this can sometimes be detected by an unfeasibly high bad pixel correction factor in the task output (when SAS_VERBOSITY>=4), e.g.:

backscale::arfgen:- Bad pixel & chip gap fraction: ~43.23 %

This problem has so far been detected in two cases:

1. A spatial region was extracted with the axes reversed, i.e.what should have been a region of type POS(Y,X) became AXIS2D1(Y,X).
2. A spectrum was extracted from a filtered event list where the filtering was performed with a region file, e.g.:
   evselect expression="region(bkg_region.fits)&&(PATTERN == 0)&&(PI in [200:10000])"

   Nonetheless, spectra extracted from a normal unfiltered event file using a region file don't seem to suffer from this problem.

Its probable that more lines of processing will run into this error. Please check the DSTYPEn keywords in the header of the spectrum if you suspect that your BACKSCAL value of effective areas are wrong, e.g. with the ftools command:

fkeyprint spectrum.ds DSTYP

The workaround for this problem is straightforward. It is sufficient to use a FITS editor (such as fv) to change the AXIS2D1 string to the right value, e.g. change DSTYP3 = AXIS2D1(Y,X) to DSTYP3 = POS(Y,X), or DSTYP5 = AXIS2D1(DETX,DETY) to DSTYP5 = DET(DETX,DETY), and then run backscale and/or arfgen again.

For complicated spatial regions the identifier may be AXIS2D2 or AXIS2D3, but the same principle applies.