Fitcoords ‘Fitted coordinates at the corners of the images’ error
tfish13 wrote on May 09, 2014
Hello! I'm using trying to wavelength calibrate some 2D spectra and having trouble somewhere in the Identify/Reidentify/Fitcoords process. I am running through the reduction process for two sets of data (red and blue spectra), taken simultaneously. The red wavelength calibration has worked fine, but the blue calibration is giving me crazy coordinates:
Red spectra:
Map User coordinates for axis 1 using image features:
Number of feature coordnates = 3426
Mapping function = chebyshev
X order = 6
Y order = 6
Fitted coordinates at the corners of the images:
(1, 1) = 9839.911 (2048, 1) = 5204.399
(1, 1028) = 9847.215 (2048, 1028) = 5099.2
Blue spectra:
Map User coordinates for axis 1 using image features:
Number of feature coordnates = 360
Mapping function = chebyshev
X order = 6
Y order = 6
Fitted coordinates at the corners of the images:
(1, 1) = 3344.803 (2048, 1) = 17397.47
(1, 1028) = 107993.6 (2048, 1028) = -32470.48
The resulting coordinates for the blue spectra using fitcoords are supposed to range between 3000 - 5500 Angstroms, so something is very off. The reduction process has been identical for both sets of data, with the only exception being that the blue comp lamps only have about 4-6 lines available for calibrating. Is this what is giving fitcoords fits?
Cheers,
Travis
Francisco Valdes wrote on May 12, 2014
Hello Travis,
If you would like to send me your fitcoords database file for this problem spectrum I can take a look. Off hand I don't have an answer. Please also include a header listing of the blue spectrum file.
Frank Valdes
Francisco Valdes wrote on May 12, 2014
Please use valdes@noao.edu for the email. Frank
Francisco Valdes wrote on May 13, 2014
Hi Travis,
Thanks for the information. What I see is that there are only 5 lines. Using the default order of 6 is too high and becomes unconstrained, particularly towards the edges where the corner values are computed. So you need to adjust the order, particularly the x order. Below I show the result of using a 3rd order and the corners seem more reasonable. I did not play around too much with the fitting but with a little tuning of the function and orders you should get a good result. If there are more spectral lines you could try and trace them but I know this is often not possible in the blue.
Regards,
Frank
NOAO/IRAF V2.16 valdes@tiamat.tuc.noao.edu Tue 08:38:07 13-May-2014
Longslit coordinate fit name is tpcomp.0005b.
Longslit database is database.
Features from images:
tpcomp.0005b
Map User coordinates for axis 1 using image features:
Number of feature coordnates = 360
Mapping function = chebyshev
X order = 3
Y order = 6
Fitted coordinates at the corners of the images:
(1, 1) = 2545.09 (2048, 1) = 6283.646
(1, 1028) = 2553.052 (2048, 1028) = 6274.229
tfish13 wrote on May 13, 2014
Hooray! This worked great. Thanks for the help, Frank.
Last post on May 13, 2014