iraf.net

I have a very weird problem with XDIMSUM. It seems to work, but sometimes one of the sky-subtracted imagesis screwed up. What happens is that the left ~200 columns in the image are shifted last to the right of the image. The rest of the images is shifted to the left. I repeated the reduction few times and it seems to happen randomly and on random images. Any idea why this happens? I have not had this problem before.Thanks.

What you describe sounds like the boundary extension behavior (i.e. what a task does near the edges of the image). This is only set to "reflect" in the XZAP task but you didn't say which task you were using introduced the swap.I did notice that in XZAP the script incorrectly sets a 'nearest' boundary extension (i.e. sets the parameter without quoting the value), but I would expect this to throw a syntax error. It's possible you could be using a parameter setting from your uparm directory so one thing to try would be to unlearn all the tasks in the package, or better yet, create a new iraf login directory with MKIRAF so you don't have any saved params and try reducing again. This won't explain the random changes however. Frank will see this and may have other comments.-Mike

Hi Mike,The problem happens in XMOSAIC during the master pass. The problem should not be in XZAP because I switched it off. Moreover, 200 px is a bit too much to be edge effect. It's like the left part of the sky subtracted image is shifted to the right (all objects are visible). Well, one picture = 1000 words, so here is an example pair of images. The stars in the circles are the same objects. The second circle in the sky-subtracted image marks the fixed bad pixels, which in the input image are in the upper right and left corners. Hope this images will help figure out what is the problem
[img:bce9cedced]http://centra.ist.utl.pt/~vall/ds9.jpg[/img:bce9cedced]

The bad pixel correction comes after the sky subtraction, and after that no further image manipulation is done (just checked xmaskpass.cl). It seems that the mess up happens after the bad pixel correction, i.e. the bad pixels are corrected right but somehow the image gets screwed up after that.

I turned the bad pixel correction off and the problem is gone.

The shifting behavior is usually the result of a corruption of the FITS file where the header block(s) are not correct. The result is either an "premature EOF" or the wraparound depending on whether a header block is lost or added. In the past there were problems with the FITS handler that could result in this. These have been gone for some time AFAIK but maybe you have an older version of IRAF. It could also be a bug in one of the image tasks in XDIMSUM such as the bad pixel fixing.Your workarounds are a) what you did by empirically turning steps off or b) use imh format instead of FITS format. If this is happening with the latest version of IRAF and xdimsum then it is problem that should be debugged. If you have a reproducible case you could provide the recipe and images (along with information about the OS type) and we could debug the problem (hopefully).Yours,
Frank Valdes