[News] Potential Data Problems
Koen Kuijken
kuijken@strw.leidenuniv.nl
Thu, 5 Jun 2003 09:44:52 +0200 (MEST)
Dear Mark,
excellent initiative. I have added some comments below.
Greetings
Konrad
> 1/ Science Flats:
>
> When creating flat field frames from the science images (master science
> flats) there must be a mechanism for eliminating unsuitable frames (a quality
> determination). Of importance here, is the elimination of all frames containing
> large objects (bright, saturated stars, nearby galaxies, etc) whose extent exceeds
> the dither pattern size. Otherwise, there will be sufficient overlap between
> successive frames that object residuals will be in the master flat and resulting
> in "holes" in the flatfielded images.
> This is relatively easy to test for since we have a number of SExtractor output
> parameters that we can use. For example, in a high S/N SExtractor pass remove
> all frames from the master science flat creation with:
> A_IMAGE >= dither/jitter [in pixels] [A_IMAGE is the 2nd order moment along
> the image major axis at the chosen
> detection threshold]
> (Other relevant SEx parameters could be ISOAREA_IMAGE, FWHM_IMAGE,
> XMIN_IMAGE, YMIN_IMAGE, XMAX_IMAGE, YMAX_IMAGE . . . Emmanuel suggestions?)
>
In case of jitter data, this will remove all frames - we expect several
m=8 stars per field. So this will only work if
- this procedure is applied per CCD, not per field
- many OBs at different pointings are incorporated. In that case a more
traditional masking of bright outliers may be sufficient.
>
> 2/ Masking of Spurious/Unwanted Features via:
>
> (i) Automatic detection and masking of prominent features such as:
>
> (a) Stellar spikes (NOTE: not entirely trivial as stellar spikes
> can vary in length/profile/relative_position as
> a function of seeing, and with camera rotation
> with respect to telescope spider.
> Under conditions of very good seeing, one can
> get spikes that go across the entire CCD).
>
> (b) Saturated stars (NOTE: the masking of saturated stars needs
> to be sufficiently conservative to also mask
> potential offset halos due to reflections in the
> optics around very bright stars).
>
> (c) Satellite trails (NOTE: these are not always uniform. Satellite
> tumbling may result in large brightness variations
> along the track. How well does the Terapix wavelet
> transform software work at detecting these?).
>
> (d) Reflections/ghosts/halos
>
> (ii) Manual masking (is this going to be a Panorapix utility?)
>
(b,d) we think we know where there reflections will lie, from raytracing
> 3/ Fringe Pattern Correction:
>
> This can be relatively straight-forward when the observing conditions are
> stable and photometric. It can be a nightmare when they are not.
> For light of a given wavelength, the observed fringe pattern can be very stable,
> being determined by the thickness of the chip at each pixel. The effect is not
> very strong for continuum sources, since the fringe pattern is smeared out.
> However, the fringe pattern seen in flattened images is the result of a few strong
> atmospheric emission lines (mostly OH and O). Since it is caused by the sky, and
> not inherent in the sensitivity of the detector to the stellar flux, the fringe
> pattern should be treated as an additive term remaining after the flat-fielding.
> The problem is that under variable sky conditions the OH and O emission varies in
> intensity apparently independently from one another. Since they emit at different
> wavelengths, the strength of the fringe pattern will vary substantially relative to
> the sky brightness from exposure to exposure. I have heard the claim that this
> can be modelled and corrected (WFI users/Meisenheimer?) but I have not seen the
> results and am not familiar with the exact technique.
>
The claims I have heard concern INT WFC data (similar detectors). Mike
Irwin has developed a technique which robustly fits the amplitude of the
fringe pattern and subtracts it out. He claims that for i' and bluer, a
single fringe pattern works very well (down to better than 1% of the
original amplitude), but that in z' (or I_c) there are indeed different
families of lines tat vary independently. He also developed a technique to
fit these simultaneously, but it was not as successful. We should try this
out on WFI data, though the lack of an i' filter is a pain (WFI I filters
extend much further into the nasty near IR sky lines).
> 4/ Illumination Correction:
>
> Undoubtably, OmegaCam will show significant large-scale spatial gradients in
> the photometry across its field of view and across its individual CCD's. The
> primary reason for this is the non-uniform illumination of the field. These
> systematic variations can be removed using superflats (Clowe & Schneider 2001;
> Alcala et al. 2002), or, perhaps better, by observing densely distributed
> secondary standard star fields (globular clusters . . .). This was done for
> WFI by Koch et al. (see http://www.mpia-hd.mpg.de/SDSS/data) who found strong
> spatial gradients of about +/- 0.05 magnitudes across the field of view. By
> doing photometry on a globular cluster with SDSS calibration, they fit the
> residuals, and were able to subtract these from the measured instrumental
> aperture magnitudes to remove the spatial dependence of the photometry. As
> a side effect, observations of these dense star fields produce a good 1st
> order astrometric correction.
>
The non-uniform illumination is corrected with superflats, indeed, but the
WFI problem is something else - sky concentration, which is a defocussed
reflection of the whole field that is present in every exposure. It has
the effect of adding a smooth bump to the flat fields as well as to the
sky, so that dividing one by the other results in a nice flat background
but with suppressed source fluxes in the middle of the field. This
photometric corerction is distinct from an illumination correction, and
can only be measured through systematic study photometric standard fields,
or by taking many large dither steps on the same field and so
reconstructing the gradient of the pattern.
The photometric standard fields being established with the INT should be a
good reference for calibrating out this effect.
> 5/ Oblique Scattering:
>
> Since the VST is a fast telescope (f/5.5) and we will be obsering quite
> a bit in grey/bright time, we will undoubtably see the effects of
> oblique scattering. Here, dust and optical defects get illuminated
> obliquely by a gradient in sky brightness (eg. moon). This will not be
> a major issue, but it can produce strange scatter patterns, reflections,
> ghosts, etc. Though definitely not repairable, perhaps these effects can
> be detectable.
>
> 6/ Radial Wavelength Dependence:
>
> Another small effect will occur in the Ha, Stromgren v, and night sky gap filters.
> The narrow-band interference filters (eg. night sky gap filters: delta(lambda)~100A,
> or even the Stromgren v: delta(lambda)~250A) will essentially act as a Fabry-Perot
> between the front and back strata of their interference coatings. In other words,
> there will be a wavelength dependence with radius. We will measure this when
> Sagem delivers the filters. They have promised a 0.2% relative wavelength accuracy
> across the field. So, for one of the night sky gap filters centered, say, at
> 8500A, we can expect a wavelength offset of about 20A from center to edge. (This
> is relevant to line-emission object searches etc., and needs to be considered by
> these projects, but it could be argued whether this needs to be a pipeline issue).
>
Certainly an effect - in my view this is not a primary pipeline issue
though. It is very hard to calibrate on-sky except through rigorous
spectral followups. My guess is that narrow-band searches can get a long
way using a theoretical prediction of the effect based on the lab
measurements we will make.
>
>
> Hope this is a start,
>
> Cheers,
>
> Mark
>
>
>
> =====================================================
> Dr. Mark J. Neeser
>
> Institute for Astronomy and Astrophysics, _/ _/ _/ _/_/ _/
> Ludwig-Maximilians-Universitaet Muenchen _/ _/ _/_/ _/ _/ _/ _/
> _/ _/ _/ _/ _/ _/ _/_/
> Scheinerstrasse 1, _/ _/ _/_/ _/ _/
> D-81679 Muenchen, Germany
>
> Tel: +49-89-2180-5994
> Fax: +49-89-2180-6003
>
> email: neeser@usm.uni-muenchen.de
>
> _______________________________________________
> News mailing list
> News@astro-wise.org
> http://listman.astro-wise.org/mailman/listinfo/news
>