FITS Format
Contact John Barentine (primary) or your Observing Specialists (secondary) for additional information.
Contact Jack Dembicky for web page errors, corrections, ommisions, or additions.
This page last updated: July 21, 2004 - JMD
This page last checked: July 21, 2004 - JMD
GRIM FITS format storage change
The MC has been modified to change the way GRIM images are stored to disk when FITS format is selected. IRAF format storage of grim images and storage of DIS images in either IRAF or FITS format has not changed.
Storage of GRIM images has been somewhat problematical because GRIM does an internal 'bias subtraction' which leaves one with images that are not pure 16-bit unsigned numbers (like they are off of a 'normal ccd') instead, the 16-bit numbers are differences of numbers, and can be negative. In fact the average counts per pixel in a blank grim image is about -2000. At the same time, one wishes to preserve as much as possible the full dynamic range of the A/D converters of the instrument and data, so that using signed 16-bit numbers can also cause one to lose dynamic range for values which go beyond 32767, which occurs for long grim integrations on bright sources.
The solution the Grim team has come up with is to internally add a constant of 10000 to every pixel of the 'bias subtracted' grim image which makes it conform to the unsigned 16-bit number standard, in that all pixel values after this addition of 10000 are in the range 0-65535. When you save a GRIM image to disk in IRAF .hhh .hhd format, you get this +10000 'ushort' image with an apparent bias level of around 8000.
The Grim team. however has requested that the original (without the +10000 added to every pixel) pixel values be available to users who take images in FITS format. To facilitate this they provide the level of the constant (10000) that was added to every pixel in the form of BSCALE and BZERO keywords which the put out on status with every image taken. By using BSCALE and BZERO within a FITS header, one can recover the full 16-bit dynamic range of the data, even though some is negative and some is positive above 32767. We have now implemented this for storage of GRIM files in FITS format. When one reads these FITS files into IRAF using rfits or strfits, they no longer read in to a 16-bit ushort image, but rather into a (32-bit) real with 'full dynamic range' further more, the bias level of these 32-bit images is at -2000 (not at +8000).
I hope this doesn't cause confusion. Basically observers who use Grim should be consistent in taking all their data as
filetype iraf
or
filetype fits
and not switch between them, though one can convert if one wants to by adding/subtracting 10000. The difference is a constant of 10000 between the two and double the storage for the fits image after it is read into iraf, (the fits files are still 16-bit).
Please let the Observing Specialists know if this causes trouble.
