Tim,
> SIMPLE = T /
> BITPIX = -32 /
> NAXIS = 2 /
> NAXIS1 = 299 /
> NAXIS2 = 299 /
> CRPIX1 = 212464.628 /
> CRVAL1 = 0.000000000 /
> CDELT1 = -0.001666666707 /
> CTYPE1 = 'GLON-CAR' /
> CRPIX2 = 10.394 /
> CRVAL2 = 0.000000000 /
> CDELT2 = 0.001666666707 /
> CTYPE2 = 'GLAT-CAR' /
> CROTA2 = 0.000000000 /
> LONPOLE = 0.000000000 /
> WAVELENG= 8.28000e-06 /
> BUNIT = 'W/m^2-sr' /
> TELESCOP= 'MSX ' /
> INSTRUME= 'SPIRITIII' /
> ORIGIN = 'AFRL-VSBC' /
>
I *HATE* CAR projections! They give more trouble than all other
projections put together. The problem here is the value of the CRPIX1
keyword....
CAR projection is defined as valid for 180 degrees longitude on either
side of the reference longitude (i.e. from CRVAL1-180 to CRVAL1+180).
In your case a CDELT1 of 0.001666666707 results in 180 degrees
being equivalent to 108000 pixels, so the projection is valid from
CRPIX1-108000 to CRPIX1+108000, which is from pixel 104464.63 to
pixel 320464.63. So the valid region does not overlap your image region
(which ends at pixel=299).
The solution... modify the CRPIX1 value by subtracting the number of
pixels in 360 degrees of longitude. So in your case change CRPIX1
to
CRPIX1 = 212464.628 - 360/CDELT1
= -3535.37
Changing CRPIX1 by 360/cdelt1 effectively changes changes the
corresponding sky longitude by 360 degrees which has no effect on the
displayed RA but results in the native spherical coords being in the right
region [-180,+180].
And Yes, FitsChan should probably do this automatically when it comes
across this situation, but currently it doesn't.
David
|