[Aavso-photometry] Flat frame quality and its effectonphotometric accuracy

[arne]

Lionel Catalan wrote:
> I would like to start by thanking everybody who has been participating in
> this discussion about flats. I'll try to answer questions as best as I can.
> 
>  
> 
> Arne: You mentioned that the problem is likely due to scattered light. I am
> surprised because I thought that scattered light would be corrected by the
> process of aperture photometry since the background level is subtracted from
> the star signal. It seemed to me that as long as the aperture and annulus
> for a given star are similarly affected by scattered light, then the net
> star signal would be corrected. Am I wrong about this? My images usually
> have some gradient in the background caused by city light pollution or
> neighbor's lights, etc. 
>
The problem with scattered light is that it appears in your flats.  They
get divided into your raw science frames, and so any scattered light
systematically affects the photometry.  You get really uniform-looking
sky background, but the star flux is wrong. It is a confusing issue.  For
example, you state that you have 5 percent vignetting - you might actually
have *no* vignetting, and all of the pattern is due to scattered light.
You need to test to see for sure.

>  
> 
> The test proposed by Arne in which a star is raster scanned across the CCD
> would be difficult to carry out at my location (northwestern Ontario)
> because I almost never get perfect skies. Therefore, it would be difficult
> to distinguish the effects of change in transparency from those of poor flat
> fielding. Instead, I propose to do the following test. 1. Image an open
> cluster both before and after meridian flip making sure that the same star
> with good signal-to-noise ratio stays at the centre of the CCD. 2. Calibrate
> the images. 3. Do differential photometry on stars distributed across the
> CCD using the central star as reference. 4. calculate the difference in
> calculated magnitudes of the same stars before and after meridian flip. Plot
> the differences as a function of star position on the CCD image and look for
> trends. Would this be a good test?
> 
Your test would check for any non-radial pattern; anything that is
concentric would still pass through.  Flipping and recentering means
that a star that is 5arcmin from the center of the image on one side
of the pier is 5arcmin from the center of the image on the other
side of the pier.  So if you had scattered light coming in from one
side of the field, then you would see a systematic shift, but if it
mimics vignetting (which often happens since the light passes down
cylinders to get to the camera), you wouldn't see any effect.  The other
way is to image a known cluster and measure each star and compare with
the standard magnitude for that star - do this as a function of position.
The problem here is that stars have different color, so you really need
to transform your data before running this test to correct systematic
color errors first.  Another method is to raster-scan the cluster, even
on a non-photometric night, and use least-squares (basically use the
central stars to set the frame zeropoint) to adjust each frame before
performing photometry.  That is a difficult exercise.
>  
> 
> Regarding dust on the filters and the repeatability of the filter wheel, I
> checked that the dust donuts are at different positions using different
> filters (B,V, and I), so it is clear that most of the donuts are caused by
> dust on the filters. Moreover, when I divide two flat frames taken minutes
> apart with the same filter (and without turning the filter wheel in
> between), I get a perfectly flat image with only random noise. On the other
> hand, dividing flats taken with the same filter on separate sessions after
> having turned the filter wheel gives an image that shows the outlines of the
> dust donuts. So, I think that the problem is in a large part due to poor
> repeatability of the filter wheel. If I remember correctly, my SBIG filter
> wheel (CFW10) always turns in the same direction to change from one filter
> to another. So the repeatability of the filter wheel remains a problem even
> if the filters are always approached from the same direction. Therefore,
> removing the dust from the filters may be the only possible solution here.
> By the way, can somebody recommend a procedure for cleaning filters in order
> to remove as much dust as possible?
> 
Usually canned air works, as long as you keep the can vertical.
For more stubborn dust, you can use alcohol with care, being sure to
dab and not wipe.  Once clean, SCTs plus cameras tend to stay pretty
dust-free since there is little exposure to the air.

You are pretty close to a final solution, if you only have 5% vignetting
and only 1.5% shift upon GEM flip.  I'm sure you will get the rest of
the way soon.  Michael Newberry has given good advice for generic
flatfielding that will help.
Arne