Re: [TASS] Flat fielding

[Andrew Bennett <andrew.bennett@NS.SYMPATICO.CA>]

On Thu, 16 Sep 1999 09:23:32 -0700, Chris Albertson
<chrisja@JPS.NET> wrote:

> ...  Would it be better for me to average four
>200 line images then use one 800 line image?
>
>So when we "divide by a vector derived from a single image"
>we are actually "dividing by a vector derived from 800
>scan lines"  Maybe 800 is the wrong number and 400 or
>3200 would be better?

As has been pointed out to me, I know nothing whatever
about this. But I never let ignorance stop me before!
Judging by the selection of flat vectors available in
the Technical Notes, e.g. orion 2-6 in TN0009, the large
scale structure can vary by at least 20% between two 800
line images. Root-n arguments would suggest averaging
400 of these to get 1%. Other sets (e.g. TN0025) show
much smaller variation. But even averaging huge numbers
won't get rid of systematic errors. My gut feeling (uninformed
by fact as it may be) is that the systematic errors
are likely to be large.

These flats were derived from the images themselves;
those derived from twilight images might or might not
be better. Judging from the MK IV twilight images, there's
not much in it

I am assuming that these large scale variations are
entirely due to background light variation and are
essentially unrelated to the real variation of
sensitivity (mostly vignetting) that we are trying to
correct.

"Dome flats" look like the way to go for the MK IV.

>  I can send you a CD with raw images
>on it. if you want to experiment.

An excellent idea. I will contact you off-list.

Andrew Bennett, Avondale Vineyard, Nova Scotia, Canada.