I've placed two test files on the web, both 4x4x4 pixel data cubes. They
can be found at:
http://gravlens.physics.ucdavis.edu/~auger/work/gaiatest.fits
http://gravlens.physics.ucdavis.edu/~auger/work/gaiacrash.fits
The first file gives an error message when opened and the second causes a
segfault on our 32bit linux systems. On an a 64bit AMD system using the
September release of starlink, gaiacrash.fits still causes a segfault, but
the gaiatest.fits file does not cause any problems.
The gaiatest.fits header is:
SIMPLE = T / conforms to FITS standard
BITPIX = -64 / array data type
NAXIS = 3 / number of array dimensions
NAXIS1 = 4
NAXIS2 = 4
NAXIS3 = 4
EXTEND = T
CTYPE1 = 'LINEAR '
CRPIX1 = 1
CRVAL1 = 1
CD1_1 = 1
CRPIX2 = 1
CRVAL2 = 1
CD2_2 = 1
CRPIX3 = 1
END
and the gaiacrash.fits header is the same except no CRPIX3 keyword.
Regarding NaNs and blanks: when I have NaNs in my images, they are
displayed in gaia as NaN in my image and are therefore unaffected by
changes of the 'blank' pixel color. I suppose what I'm wondering is if it
makes sense for gaia to assume a BLANK = '[representation of NaN]' FITS
header card even in the absense of such a card to that NaNs are
automatically displayed as 'blank'.
I hope this helps!
Matt
On Fri, 2 Mar 2007, Tim Jenness wrote:
>
> Can you point us to your example file? I'm sure it's a FITS header problem.
>
> I think Floating point fits defines BLANK to be NaN.
>
> Tim
>
> On Fri, 2 Mar 2007, Matthew Auger wrote:
>
>> Hi...I just built (and tried using the pre-compiled binaries for) the
>> newest starlink release and get an error message when I try to open a
>> datacube with gaia. The error is "wrong number of columns" followed by some
>> sort of description of the file. This happens when I use the following
>> header keywords:
>> CRVAL1
>> CRPIX1
>> CD1_1
>> CTYPE1
>> CRVAL2
>> CRPIX2
>> CD2_2
>> CRPIX3
>>
>> If I also include CRVAL3 and CD3_3 the problem persists, and if I don't
>> include CRPIX3 the program crashes complaining (in a gdb backtrace) about a
>> line in gaiaUtils.c (which investigation shows results from ncoords=3 and
>> ncoords_in=2 so there is no condition to instantiate the coords_in array).
>> Any suggestions? Other than the message, everything seems to work OK.
>> Oh...except there is one odd behavior: datacubes display nan's as blank
>> pixels (which is good!) because they seem to add the 'BLANK' header
>> keyword, whereas 2d images don't. Is there a way to automatically convert
>> nan's to 'blank' or do I have to add the header keyword myself? Also, the
>> 'blank' color encoding remains set (ie the radial button is still selected)
>> when I change colormap/stretch/intensity map, but the blank coloring is not
>> active and I have to re-select it. Is this expected?
>>
>> By the way, gaia is a great product--thank you very much for providing it
>> to us!
>>
>> Best regards,
>>
>> Matt
>>
>
>
|