I have some code that uses fits_read_img to read in a fits image, which works fine. The problem is when I use ds9 to convert one slice of a binary data cube into a fits image and it cannot read it in correctly any longer. IDL can read the new image in correctly, but fits_read_img no longer can. I don't get an error, in fact it looks almost identical, but in the third or fourth decimal place it has the wrong value. Has anyone seen this before? do you know a workaround?