On 18 October 2011 13:53, Jean-Baptiste Marquette <[log in to unmask]> wrote:
> On http://fits.gsfc.nasa.gov/registry/sip.html one read:
> "The SIP convention is supported by WCSTOOLS (written by D. Mink), the
> Starlink AST library, and the IDL ASTROLIB library."

> How AUTOASTROM does deal with SIP convention ?

AUTOASTROM does not deal with the SIP convention. The SIP convention
and the ""QV" convention used by AUTOASTROM are alternatives.

> Another question: how to use a local reference catalog instead of a network
> query ? I have sequences of hundreds of images on the same field to
> calibrate, so a single query (for example 2mass@cds) should be enough to
> define a suitable reference catalog.

I'll leave Peter to answer that.

David



> JB
>
> Le 18 oct. 2011 à 12:05, David Berry a écrit :
>
> On 17 October 2011 13:31, Jean-Baptiste Marquette <[log in to unmask]> wrote:
>
> Hi David & Dustin,
>
> I come back on this issue concerning astrometric calibration on EROS-2
>
> images with distortion.
>
> You will find on anonymous ftp.iap.fr, directory
>
> pub/from_users/marquett/Astrom an archive containing:
>
> the image lm00417tbr8k0657.fits calibrated using Starlink's AUTOASTROM
>
> the image lm00417tbr8k0657.new.fits calibrated using default parameters
>
> of solve-field from astrometry.net
>
> various files yielded by this process
>
> The original image was linearly calibrated using Emmanuel Bertin's Scamp.
>
> In Gaia-Starlink version 4.4.1 on Mac, it appears that:
>
> calibration by AUTOASTROM is in agreement with 2MASS catalog over plot.
>
> calibration by solve-field shows an offset compared to 2MASS, more
>
> pronounced in RA
>
> In ds9 6.1.2 on Mac, both images show an offset, not the same way, compared
>
> to 2MASS catalog.
>
> My guesses are:
>
> The QV keywords from AUTOASTROM are correct in lm00417tbr8k0657.fits.
>
> However, they are not properly managed in ds9.
>
> The SIP keywords in lm00417tbr8k0657.new.fits are wrong, for unknown
>
> reasons. But I noted that the CRPIX values are close to the image center.
>
> This seems to me not correct, as the image is part of a CCD mosaic, thus the
>
> optical axis is far away from the image.
>
> It would be nice to have your opinion on this issue. This is critical to
>
> make the choice how to calibrate the 1.55 million EROS-2 images.
>
> Thanks for your help and expertise,
>
> Cheers
>
> Jean-Baptiste
>
> Further to this, I suspect we should be thinking to change AUTOASTROM
> so that it produces header with the new TPV projection plus PV
> keywords, rather than TAN projection plus QV keywords. They are
> exactly equivalent systems, but TPV+PV will presumably become more
> recognised by a wider body of software now.
>
> David
>
>
>
>
>
>
> Le 4 oct. 2011 à 09:48, David Berry a écrit :
>
> On 3 October 2011 20:50, Jean-Baptiste Marquette <[log in to unmask]> wrote:
>
> Le 3 oct. 2011 à 12:30, David Berry a écrit :
>
> On 30 September 2011 09:10, Jean-Baptiste Marquette <[log in to unmask]> wrote:
>
> Dear Starlink gurus,
>
> This may be of interest for you, I would be glad to have your expertise.
>
> Cheers
>
> Jean-Baptiste
>
> Début du message réexpédié :
>
> De : Jean-Baptiste Marquette <[log in to unmask]>
>
> Objet : Rép : astrometry.net & EROS-2
>
> Date : 30 septembre 2011 09:57:54 HAEC
>
> À : Dustin Lang <[log in to unmask]>
>
>
> Le 29 sept. 2011 à 14:51, Dustin Lang a écrit :
>
> My purpose is to write a python pipeline to test parallel calibrations on my
>
> 4-core Mac Pro, and to run it on a future (at least) 12-core one. So I'm not
>
> sure that web queries will be as efficient as local indexes.
>
> Ok, good.
>
> Try these files:
>
> http://astrometry.net/downloads/indices/
>
> index-203-##.tar.bz2
>
> index-204-##.tar.bz2
>
> The ## says what part of the sky it covers: see the map here:
>
> http://trac.astrometry.net/browser/trunk/src/astrometry/util/hp.png
>
> For example, the LMC is in healpix 08.
>
> cheers,
>
> dustin
>
> Hi Dustin,
>
> I have the results of a preliminary test, and they are quite unexpected.
>
> The screenshots below show a zoom of the outermost corner of an external CCD
>
> of the EROS-2 "blue" camera, where 2MASS objects are over-plotted as green
>
> circles. The "failed" one is from a solve-field calibration, the correct one
>
> is from a Starlink's autoastrom one. I also join the FITS header of both
>
> images. These images are displayed with Gaia-Skycat version 4.4.1 on my Mac
>
> Pro under Mac OS X Lion (10.7.1). I used solve-field with default
>
> parameters, and 2MASS at CDS as reference catalog for autoastrom.
>
> Note that both images show such a positional offset in ds9.
>
> My guess is that Gaia copes correctly with autoastrom's QV keywords, maybe
>
> not with SIP convention.
>
> Hi Jean-Baptiste,
>
> Thanks for the information, and sorry for the delay in replying.
>
> GAIA uses the AST library to do handle the WCS information. I've just
>
> tried using AST to transform some coordinates from the SIP header
>
> (astrometry.head) you supplied. For instance, transforming pixel
>
> coords (1,1). AST gives:
>
> 87.77907666 -70.45308005
>
> For comparison, I did the same thing using two other independent
>
> implementations of the SIP distortion:
>
> astrolib (i.e. pywcs): 87.78922947 -70.45330567
>
> wcstools                 : 87.77907666 -70.45308005
>
> So AST agrees (excellently!) with wcstools, but is significantly
>
> different to pywcs. So unless  AST and WCSTOOLS are both wrong,
>
> interpretation of SIP headers in gaia is probably correct.
>
> Previous versions of DS9 have used wcstools to do WCS conversions and
>
> used AST for plotting coordinate grids, so since AST and wcstools
>
> agree, I would also expect GAIA and DS9 to agree. [The next version of
>
> DS9 (version 7.0 - currently available for beta testing) has swapped
>
> over to using AST instead of wcstools for all its WCS conversions.]
>
> David
>
> PS In case you are interested, a Python interface to AST is available
>
> and will be described in a poster at ADASS next month.
>
> Hi David,
>
> Thank you for your answer and analysis.
>
> Gaia is effectively announced as compatible with SIP by GSFC on
>
> URL http://fits.gsfc.nasa.gov/registry/sip.html
>
> Thus I wonder why there is a significant difference when I over-plot 2MASS
>
> objects on both images in gaia (one with AST header, the other with SIP
>
> header). They should be in accordance.
>
> Dis you see the images ?
>
> Do the offsets shown by ds9 and gaia for the SIP image look the same?
>
> If so, then there is presumably a real error in the sip headers,
>
> introduced by the software that generated the headers.
>
> The reason that ds9 shows offsets for the autoastrom solution is that
>
> ds9 prior to version 7 ignores the QV keywords that autoastrom uses to
>
> encode the distortion. So the autoastrom image will be displayed by
>
> ds9 without any distortion. What would be interesting is if the
>
> offfsets shown by ds9 are the same for both QV and SIP distortion.
>
> David
>
>
> Cheers
>
> JB
>
> ---- Starlink User Support list For list configuration, including
>
> subscribing to and unsubscribing from the list, see
>
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=STARLINK
>
> ----
>
> Starlink User Support list
>
> For list configuration, including subscribing to and unsubscribing from the
>
> list, see
>
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=STARLINK
>
> ---- Starlink User Support list For list configuration, including
>
> subscribing to and unsubscribing from the list, see
>
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=STARLINK
>
> ----
> Starlink User Support list
> For list configuration, including subscribing to and unsubscribing from the
> list, see
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=STARLINK
>
> ---- Starlink User Support list For list configuration, including
> subscribing to and unsubscribing from the list, see
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=STARLINK

----
Starlink User Support list
For list configuration, including subscribing to and unsubscribing from the list, see
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=STARLINK