bad columns not masked properly?

[Stupendous Man <richmond@a188-l009.rit.edu>]

  Tom wrote:

> Looking at a possible problem with the I images.
> 
> There is a bad "column" on the chip.  This was noted by Andrew 
> Bennett.  This causes stars that traverse this column to dip in value on 
> the frame where the cross.  This gives plots for constant stars that dip as 
> they pass over this column.  In fact, they usually have a low point then go 
> to zero for several frame times, then pick up another low point rising off 
> the defect.
 
...

> 1) The patch table does not seem to match where the defect is seen with DS9 
> which sees it at x = 1959, y = 461
> 
> master_flat_I.msk
> 
> 0       6072    0       2023    0       2
> 1899    62      450     470     1948    1953
> 
> The I.msk seems to have values for entry 1899 that are close to the defect, 
> but they do not match the spot I see with DS9.

  In the documentation for the pipeline,

        http://spiff.rit.edu/tass/pipeline/pipeline.html

the values in the mask file are explained.

:: The columns are 
:: 
::   1.region ID. Because many regions are found, but fail to 
::     meet the criteria for acceptance, the surviving regions usually
::     have big gaps in ID values. 
::   2.number of connected bad pixels in the region 
::   3.minimum row of all bad pixels in the region 
::   4.maximum row of all bad pixels in the region 
::   5.minimum col of all bad pixels in the region 
::   6.maximum col of all bad pixels in the region 

  So the second line in the mask file describes a bad region 
which covers columns 1948 to 1953, and rows 450 to 470.

  The reason that this doesn't match the apparent place of the
bad column -- column 1959, according to Tom -- is that the mask
is applied to each image AFTER it has been "trimmed".  The edge
rows and columns are removed from each image early in the 
reduction process.  The "ccdproc.param" file explains how many
rows and cols are trimmed.

> 2)  The mask does not realize that the defect extends down the whole column.

  Yup.  The program which created the mask didn't get it right.
In a case like this, the user should edit the mask file manually,
so that its second line looks something like this:

> 1899    62      0      2028     1948    1953

  _Now_ the mask will cover the entire set of columns (assuming that
the trimmed images have 2029 rows).

  The pipeline isn't going to get this part right automatically.
The user should check images manually every now and then, and modify
the mask files as needed.


> 3)  I think there is one more problem in that the overscan and dark columns 
> are not correctly selected.  Looking with something like DS9 will show 
> where they really are.

  Absolutely right.

  
                                             Michael