Re: automatic processing of data

[Tom Droege <tdroege@veriomail.com>]

Michael,

Why so many dark field frames?  3 does a good job and gets rid of cosmic 
rays with high probability.  5 seems overkill.  OK, I guess if you use a 
lot then the dark will have less noise.  Seems to me that this is a pretty 
useless  low noise as it should be much less than the noise in the data 
frames.  The data frame noise will always be much larger, and when added in 
quadrature, the dark frame noise will just disappear.

Using a lot of dark frames is not a "free" improvement.  It takes more time 
to compute.  It also makes it harder to track temperature changes.  One 
might argue that a sliding 3 or 5 frame dark might give the best result.

My observation is that there is little to be gained by fussing with the 
dark.  To convince me to use 20 or 50 dark frames I would want to compare 
the result with using a smaller number.

Tom Droege

At 03:45 PM 12/20/00 -0500, you wrote:

>   In Tech Note 62, I scaled the darks before combining them because
>the temperature of (at least one of) the chips wasn't controlled
>very well during the night.  I should have simply averaged the
>dark frames together.  For some reason (and probably incorrectly),
>I decided to rescale them to a common mean before taking the median.
>
>   Perhaps my reasoning was "I must use a median, not an average,
>when I combine dark frames; if I use an average, then cosmic rays
>will contaminate the output pixel values.  But wait -- I can't
>simply combine all these dark frames with a median, because their
>overall levels are changing due to temperature changes.  Rats.
>Well, if I rescale them all to have the same mean value, then I
>can use a median filter on them."
>
>   The proper thing to do in this case, with a bunch of dark frames
>taken at different temperatures, is ... throw them all out and
>take a new set of dark frames.
>
> > Any thoughts on minimum number of frames for a good dark/flats?  Or trying
> > to quantatize how you reject images for flat creation?
>
>   The more darks and/or flats, the better.  Thirty to fifty dark frames
>would be nice.  It's harder to get good flatfield frames, but a similar
>number would be good.
>
>
>                                            Michael Richmond