[Aavso-photometry] Flat frame quality and itseffectonphotometric accuracy

[Michael Newberry]

Comments below...

----- Original Message ----- 
From: "Lionel Catalan" <lcatalan at lakeheadu.ca>
To: "'Michael Newberry'" <mnewberry at mirametrics.com>
Sent: Tuesday, August 12, 2008 11:50 AM
Subject: RE: [Aavso-photometry] Flat frame quality and 
itseffectonphotometric accuracy


> Hi Michael,
>
> Here are my answers to your questions.
>
> 1. How much vignetting do you have from center to corner? (as in signal at
> center vs signal at corner)
>
> The amount of vignetting is about 5%. (The brightest spot near the centre 
> of
> the image is about 26600 ADU and at the darkest corner the signal is 25200
> ADU). Most of the time, my target and comparison stars are all fairly 
> close
> to the centre of the image, so they wouldn't be subject to that much
> vignetting.


    This would mean that 5% systematic error is the most you could get from 
center to corner if you did no flatfield correction at all before doing 
aperture photometry.


>
> 2. What is the signal level at the center of the flat?
>
> Approximately 26,000 ADU
>
> 3. So you apply dark and/ir bias corrections to your raw flats?
>
> I apply both bias and dark corrections to the raw flat frames with Maxim 
> DL.


    Are the darks scaled to the exposure time of each flat and data frame? 
Or do you take darks of the same exposure time as each flat and data frame?


>
> 4. What is the average signal of a flat field frame?
>
> Approx. 25,000 ADU. This is close to 50% of the CCD linear range.
>
> 5. Are the flats filtered, and do you use twilight, dark sky, or a screen?
>
> Flats are filtered and I use the sky at dawn or at dusk.
>
> 6. What is your degree of artificial light pollution?
>
> Light pollution is moderate. I live at the edge of a relatively small town
> in the boreal forest.
>
> 7. Are your master flats normalized to common signal level during or 
> before
> median combining?
>
> I don't understand what you mean by normalizing the flats to common signal
> level , but the signal level is about the same for all flats since the
> software adjusts the exposure time to keep the signal within a narrow 
> range.
>


    Aha! This could be a problem but perhaps not the entire explanation for 
your systematic errors. No kind of rejection combining (i.e., median, rank, 
clipping) will work properly if the images are not scaled to a common signal 
level (or offset, rather than scaled if you are combining bias frames). This 
is called "normalizing". In other words, draw a large statistics box (say, 
100x100 or so) at the center of the flats. Is the median value the same? To 
get the values to be the same, the images need to be scaled by applying to 
all pixels the same numerical factor. Normalization produces images for 
which the pixel values differ by statistical differences (noise) rather than 
by a difference in the overall intensity of each image. Then and only then 
will your median combining give you the master flat that you expect. Without 
normalizing, you will get 1) a master flat that is just the image of middle 
intensity or 2) you may get some kind of hybrid image that does not 
accurately portray your flat field variations (i.e., "garbage"). When you 
aplied such a master flat, the first case would add noise to your data 
frames and the second case would cause systematic errors in the photometry.

You should look at the flats you combine to see that they and the master 
frame make sense. After normalizing, you should plot the intensity through 
exactly the same pixel row or column through the middle of every flat. Make 
an overplot (on one graph) so you can see all the curves together. If you 
can, also plot the exact same pixels across the master flat. All of these 
curves should (or "must") overlap---plus or minus only statistical 
noise---across the width of the frame. That way, you at least know that your 
flats and master flat make sense.

Michael

>
> Thanks,
>
> Lionel
>
> -----Original Message-----
> From: Michael Newberry [mailto:mnewberry at mirametrics.com]
> Sent: Tuesday, August 12, 2008 2:43 PM
> To: Lionel Catalan; aavso-photometry at mira.aavso.org
> Subject: Re: [Aavso-photometry] Flat frame quality and
> itseffectonphotometric accuracy
>
> So, can you tell me a little about your processing? See the questions I
> asked in the previous Email.
>
> Michael
>
 [...]