Instrument Model Bugs: general

Bugs

1. If the NAXIS keywords in the PSF image file are odd, Sherpa does not convolve the PSF with the model correctly.

2. In order to replace the instrument model with a different ARF on multiple loops of a script, it is necessary to call "erase data" inbetween resetting the instruments.

3. Cases where source and instrument model reinitialization does not occur.

Bugs

1. If the NAXIS keywords in the PSF image file are odd, Sherpa does not convolve the PSF with the model correctly.

   For example, if you have an image that is 55x55 pixels. To check these keywords, you must examine the raw file header:

   unix% dmlist odd_psf.fits header,raw | grep AXIS
   Key    3: I             *NAXIS        = 2                    / number of data axes
   Key    4: I             *NAXIS1       = 55                   / length of data axis
   Key    5: I             *NAXIS2       = 55                   / length of data axis
   

   Workaround:

   if the PSF image is binned to an even number, e.g. 56x56 pixels instead of 55x55, it will have even NAXIS keywords and no model distortion will occur.

2. In order to replace the instrument model with a different ARF on multiple loops of a script, it is necessary to call "erase data" inbetween resetting the instruments.

   Otherwise you will see an error of the form:

   Error: illegal instrument definition.
          Mapping from ARF to PHA bin unknown - provide RMF.
          Instrument definition deleted.
   

3. Cases where source and instrument model reinitialization does not occur.

   The current version of Sherpa is designed such that source and instrument models are not "reinitialized" with every fit, goodness, or visualization command; reinitialization only takes place if a change has occurred (e.g., one has changed the filter) such that reinitialization is required. However, there are some cases where reinitialization does not occur, but should:

   • If one changes the normalization within a FARF or FARF2D (FEXPMAP) model, that change is not noticed in subsequent fits.

   • If one subtracts/unsubtracts, sometimes the change is not noticed, e.g.,

     sherpa> print(get_dcounts_sum(foo).value)
     264.124
     sherpa> unsubtract
     sherpa> print(get_dcounts_sum(foo).value)
     264.124
     

   Problems such as these are being fixed as they are discovered.