Re: APASS quality checking
Patrick Wils
The astrometry doesn't seem to be very good. Matching with other catalogs reveals many stars 10 arc seconds or more from their nominal position. Luckily there seems to be some systematic effect, needing (additional?) non-linear terms in the plate solving. I matched one APASS field (30172) with CMC14 up to 20 arc seconds and removed those stars which had abs(V-r') > 0.5 to avoid false identifications. I then plotted all stars whose position differed by 3 or more arc seconds in 30172.gif. The error bars are an indication of the size of the difference in RA and Dec, the colors give an indication of the direction (negative or positive difference in RA or Dec). The small dots are stars which differ less than 3 arc seconds in position. The donut shapes reveal that there are some systematic errors which need to be corrected before the data set can be studied thoroughly.
Plotting Verr against V reveals a "cloud" of stars with Verr > 0.4, clearly separated from the main group of stars (see V-Verr.gif). I can only think this is some artificial effect created by the pipeline. Only about 0.2% of the stars are affected, but I think it needs to be addressed.
Further on the photometry, I matched stars with Loneos.dat <http://vizier.u-strasbg.fr/viz-bin/Cat?II/277> and plotted the differences in Loneos-V.gif (173 matches within 3 arc seconds). Large differences only occur for stars with a large uncertainty Verr, except for three stars. Other catalogs however indicate that the APASS value is more likely to be correct in these three cases. There might be some systematic trend in the difference V_APASS - V_Loneos, with APASS a few centimags brighter on the bright end and fainter on the faint end, but this needs to be investigated further once the two issues above are resolved.
Patrick
Doug Welch
Hi Patrick,
Beautiful work!!! Arne - how is the world coordinate system for each
frame determined at present?
Cheers,
Doug
> ------------------------------------------------------------------------
>
>
> ------------------------------------------------------------------------
>
>
> ------------------------------------------------------------------------
>
Arne Henden
just a linear transformation;
I knew that some distortions were present (it is tough to deliver a
3-degree flat field without
distorting it), but I was hoping that they were less than what is
shown. Oh well; more software!
While CCD observing is nice in that you save the images and can
reprocess, I also save
sufficient information with each star extraction that I can work with
the starlists for this
correction. While I could work with the PinPoint WCS parameters in
the fits header, I'd
rather do it in the pipeline for a number of reasons.
The cloud of objects with Verr>0.4 are because we are doing
center-to-corner overlap on the
fields (to remove any systematic vignetting signature), and in a few
cases, the overlap field
was taken on a nonphotometric night and was not removed in this
initial processing. Those
are the three fields that I mentioned in my first posting.
Patrick's distortion map is very helpful; it shows that the primary
distortions are radially
symmetric, so that corrections can be modeled easier. However, there
is an insidious problem
with geometric distortion: it also affects the photometry. That is
much harder to model, so
it may be a later pass through the catalog. The intent is to complete
the observations,
and then go back and do a global solution to clean up remaining systematics.
What is also nice is that the photometric errors show that we saturate
at V=10 and
have a faint limit around V=17, exactly as predicted and what I was
hoping for. The
bright end systematic error when comparing with the Loneos calibrators
are because
those stars are saturated. On the faint end, I bet that the APASS
magnitudes will
end up being slightly brighter than other photometry as the aperture photometry
will include blended objects. We can remove some of that in the final
processing
if we decide to do psf-fitting.
Deep-sky imagers don't care about many of these aspects, as they image
with a fixed
field center and distortions aren't obvious in pretty pictures.
Researchers get upset. :-)
And Doug, thanks for the 3-D field map!
Arne
> --
> You received this message because you are subscribed to the Google Groups
> "aavsodatasection" group.
> To post to this group, send email to aavsodat...@googlegroups.com.
> To unsubscribe from this group, send email to
> aavsodatasecti...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/aavsodatasection?hl=en.
>
>
Stupendous_Man
> Patrick's distortion map is very helpful; it shows that the primary
> distortions are radially
> symmetric, so that corrections can be modeled easier. However, there
> is an insidious problem
> with geometric distortion: it also affects the photometry. That is
> much harder to model, so
> it may be a later pass through the catalog. The intent is to complete
> the observations,
> and then go back and do a global solution to clean up remaining systematics.
The geometric distortion can affect the photometry,
certainly, but vignetting and flatfield errors will
also cause photometric errors with a radial
pattern. I would guess that those two sources
might yield larger errors than the geometric
distortion.
Do you have a plan for isolating each of these
sources of error and correcting for them separately,
or do you plan to perform a single radial correction
to remove all the effects in a single step?
If you need to add a correction for geometric
distortion into the pipeline somewhere,
this might be of some use:
Michael
Doug Welch
Folks,
Patrick Wils has another nice data-mining paper in the IBVS:
"New Double-Mode and Other RR Lyrae Stars from WASP Data"
Wils, P.
http://www.konkoly.hu/cgi-bin/IBVS?5955
The interesting selection criterion for follow-up resulted in a larger
fraction of RRc stars than have historically been discovered.
Cheers,
Doug