[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Fits Header Fix Progress



I have been working on the fits header.

At 01:58 PM 4/7/01 -0400, you (Michael Richmond) wrote:

>   I'm trying the reduce the images on Disk 17.  Here's what I've found
>concerning the FITS headers:
>
>      1. some keywords have changed (relative to old Mark IV keywords)
>
>             CRVAL1   is now   RA
>             CRVAL2   is now   DEC
>             UT       is now   UTSTART


The plan is to put the same thing I am putting in RA into CRVAL1.  Ditto 
goes for the other items, just duplicate them so programs find things 
wherever they look.


>   Warnings for others who are using Disk 17:
>
>      a. the RA value is in sexigesmial notation, formatted as a string:
>               '13:03:43.4'
>
>      b. the Dec value is NOT in sexigesimal notation, but is formatted
>         as a string:
>               '1        '

This has been fixed.


>      c. the EXPTIME values have a significant scatter.  What I mean
>         is that the values in the FITS headers are (for example)
>
>                  101.8, 101.4, 100.7, 101.6, 99.4, 100.2, etc.
>
>         when one might have expected identical values of "100" in
>         each case.  This causes problems for me when I try to
>         create master dark frames.  I'd like to combine dark images
>         with the same exposure times ... but no two images have
>         (exactly) the same exposure time.  Moreover, I'd like to
>         use the 100-second darks to subtract from the 100-second
>         object frames ... but again, the exposure times aren't
>         all 100 seconds.
>
>         I'm dealing with this by binning all exposure times to
>         the nearest 10 seconds before I try to group frames together.
>         It's a minor pain.

OK, I have rewritten the program to use the ON TIMER(t) feature of 
QBasic.  Once turned on it jumps to a subroutine every t 
seconds.  Unfortunately it can only be set to a minimum of one 
second.  Every second, I check to see how long the shutters have been 
opened, and compare this to the desired value.  This somewhat reduces the 
uncertainty.  It is not great.

1       69.98
2       69.47
3       69.53
4       69.53
5       69.64
6       70.40
7       70.08
8       71.46

The commanded exposure was 70 seconds.  Note that #8 was a dark frame which 
has slightly different code.   This may or may not be significant.  In any 
case there is a minor improvement.


>         Do we know if the variations are real?  That is, was the
>         exposure time for a frame which says "100.4" in the FITS
>         header really 100.4 seconds?

We still have no way of knowing the real exposure.  I think that what you 
command is just as good a thing to use as what the program puts in the header.



>      d. the IMAGETYP values are often incorrect.  I found that
>         the dark frames are misidentified:
>
>            frame            has IMAGETYP         but is really
>          -------------------------------------------------------
>            2747                dark                 object
>            2749                object               dark
>
>            2763                dark                 object
>            2765                object               dark
>
>            2779                dark                 object
>            2781                object               dark
>
>            2795                dark                 object
>            2797                object               dark
>
>            2810                dark                 object
>            2812                object               dark
>
>          Apparently, there's an "off-by-two" error in the assignment
>          of IMAGETYP keyword.


This is now fixed.  An initialization problem.

Tom Droege


>   On another note, the V-band master flatfield (made from 34 object
>frames) shows small-scale features: blurry spots (perhaps due to
>spots on the lens?) and vague dark shapes near two of the edges
>(tree branches?).
>
>   I'll continue to work on the data over the weekend.
>
>                                                      Michael Richmond
>
>
>
>
>