[Issues] Poor astrometry on SUBARU test field

[John P. McFarland]

Hi Philippe,

Thank you for your feedback.

The RMS of 0.8-1.1 arcsec is indeed high, particularly for this data that 
_appears_ fine.  There is a large number of solutions for WFI observations 
near the celestial equator (primarily photometric standard fields) with RMS 
of between 0.4 and 0.8 arcsec.  This data does not appear to fall in the 
same category, but it needs to be confirmed.  These circumstances are why I 
asked Tim to post this message so that Erik and others could give feedback 
on this unusual occurrence.

GAstrom will improve on the relative astrometry, but it is uncertain how 
much the absolute astrometry will be affected in this case as the reference 
to source pairings for a given frame will be unchanged.  It is my experience 
that the overall RMS is somewhat homogenized in these situations, but it is 
highly data dependent.

As for the 64 SourceList limit, this is a technical limit related to 
accounting in AssociateList and its associated SourceLists.  It does not 
appear to affect the global astrometric solution which uses AssociateList, 
but Kor was investigating this prior to his forced absence.  As a 
work-around, this limit can be ignored in one's own checkout, but be aware 
that it is not fully tested, and some unusual results are always possible, 
however unlikely.  A permanent solution will be put into place when it has 
been determined safe, which will hopefully be soon.

Cheers,


-=John


On Thu, 21 Aug 2008, Philippe Heraudeau wrote:

> Dear Tim,
>
> You should certainly run "Gastrom" at some point to derive the global
> astrometric solution on your observations if they
> have enough overlapping.
> Astrom gives you a first solution whose accuracy should be about the one
> of the reference catalogue i.e. about 0.3" for USNO.
> Thus, 1" rms seems already too large for Astrom and I leave Erik and
> John comment on it but it should be improved by
> Gastrom which will give you the final astrometric solution.
>
> Note however than GAstrom cannot handle more than 64 sourcelists (or
> chips) at the moment as indicated in
> astro/main/AssociateList.py:            raise Exception,
> '__append_sourcelist: Cannot handle more than 64 SourceLists at the
> moment!'
>
> As I can see, your data set has 100 chips, so you should make a
> selection of max. 64 chips for Gastrom...
>
> Maybe John or Erik can tell us when this limit will be removed since it
> represents only 2 omegacam pointings!
>
> All the best,
> Philippe
>
>
>
> Tim Schrabback wrote:
>> Dear John and Erik,
>>
>> during the Leiden tutorial I randomly picked a multi-color SUBARU data set
>> to work with. The results for the astrometry (Astrom) were poor, with
>> an RMS exceeding 1" for some chips/exposures. Looking at the frames they
>> visually look fine. I think it would be very valuable to understand the
>> cause for such poor results in order to further improve the Astro-WISE
>> system.
>>
>> If you want to check this yourself, this is how I ran the astrometry
>> task:
>>
>> pars=Pars(AstrometricParameters)
>> pars.AstrometricParameters.process_params.MAX_RMS=1.1
>> pars.AstrometricParameters.preastromconf.RMS_TOL=5.0
>> dpu.run('Astrometry', i='SUP', o='deep16_0', C=1,pars=pars.get())
>>
>> We are currently performing extensive astrometry tests (also using
>> different reference catalogs) using CFHTLS-Wide MEGACAM data outside of
>> Astro-WISE. It would be good to get these data fully supported in
>> Astro-WISE at some point, so we can do cross-checks.
>>
>> Thanks and best regards,
>>
>> Tim
>> _______________________________________________
>> Issues mailing list
>> Issues at astro-wise.org
>> http://listman.astro-wise.org/mailman/listinfo/issues
>>
>
>
>