Incorrect sensitivity curve for spectroscopic data
Kimmo Lehtinen wrote on May 06, 2011
Hi
I was analyzing our long-slit data from ESO VLT/FORS instrument and noticed that the sensitivity curve I derived was different from what it used to be. I am sure that the sensitivity curve I derived earlier is OK.
I found out the reason to be the following: in the file 'std', which is the output flux file from the standard-task (and used by SENSFUNC), the observed spectrum of the standard star is shifted in wavelength.
That is, the observed spectrum of the standard star in the file 'std' is the same as previously, but it is shifted by about 250 Angstrom into longer wavelengths. That will make the derived sensitivity curve different. This was easy to note because the sensitivity curve drops
rapidly below about 4000 Angstrom.
I am using IRAF V2.15.1a on a 64-bit Linux/Centos machine. I installed IRAF at 4th of May.
I am following the manual 'A User's Guide to Reducing Slit Spectra with IRAF'.
I am using an arc-lamp image to get the wavelength solution for the standard.
I also noticed that in the first plot of the identify-task, showing the spectrum of the arc-lamp, the number of pixels is over 4000, although my spectrum has only 2048 pixels.
However, later within the identify-task when I use the 'h' graph key to plot wavelength versus pixel coordinate, the pixel values
seem to be correct.
The following files can be found at the page www.helsinki.fi/~kklehtin/IRAF
FORS2.ms.fits Spectrum of the standard star from apall-task
cFORS2.fits A trace from the arc-lamp image. Output from apall-task, and input for the identify task
idcFORS2 Output from the identify-task. In this file the wavelengths and pixel coordinates seem to be correct.
dFORS2.ms.fits spectrum of the standard star, after using the dispcor-task. I used 'linearize=no'. This file is used as an
input to the standard-task.
Thanks for your help, Kimmo Lehtinen
University of Helsinki, Finland
I was analyzing our long-slit data from ESO VLT/FORS instrument and noticed that the sensitivity curve I derived was different from what it used to be. I am sure that the sensitivity curve I derived earlier is OK.
I found out the reason to be the following: in the file 'std', which is the output flux file from the standard-task (and used by SENSFUNC), the observed spectrum of the standard star is shifted in wavelength.
That is, the observed spectrum of the standard star in the file 'std' is the same as previously, but it is shifted by about 250 Angstrom into longer wavelengths. That will make the derived sensitivity curve different. This was easy to note because the sensitivity curve drops
rapidly below about 4000 Angstrom.
I am using IRAF V2.15.1a on a 64-bit Linux/Centos machine. I installed IRAF at 4th of May.
I am following the manual 'A User's Guide to Reducing Slit Spectra with IRAF'.
I am using an arc-lamp image to get the wavelength solution for the standard.
I also noticed that in the first plot of the identify-task, showing the spectrum of the arc-lamp, the number of pixels is over 4000, although my spectrum has only 2048 pixels.
However, later within the identify-task when I use the 'h' graph key to plot wavelength versus pixel coordinate, the pixel values
seem to be correct.
The following files can be found at the page www.helsinki.fi/~kklehtin/IRAF
FORS2.ms.fits Spectrum of the standard star from apall-task
cFORS2.fits A trace from the arc-lamp image. Output from apall-task, and input for the identify task
idcFORS2 Output from the identify-task. In this file the wavelengths and pixel coordinates seem to be correct.
dFORS2.ms.fits spectrum of the standard star, after using the dispcor-task. I used 'linearize=no'. This file is used as an
input to the standard-task.
Thanks for your help, Kimmo Lehtinen
University of Helsinki, Finland
Francisco Valdes wrote on May 06, 2011
For some reason the value of CD1_1 in the arc extraction is 2. This is understood by IDENTIFY. I'm not sure how this affects the downstream work. Try using hedit to reset the value to 1 and redo the reductions. If this is still a problem include the std file produced by standard that you noted seemed to be shifted.
Frank Valdes
Frank Valdes
Kimmo Lehtinen wrote on May 06, 2011
Hi Frank
I changed the value of CD1_1 to 1.
The first plot of the identify-task, showing the spectrum of the arc-lamp, gives now the correct number of pixels, about 2000, instead of about 4000. This error of a factor of two must have been caused by the incorrect value CD1_1=2.
However, the final sensitivity curve is still incorrect.
At the page www.helsinki.fi/~kklehtin/IRAF you can find the text-files std_old and std_new which are the old and new output files from the standard-task.
cheers, Kimmo
I changed the value of CD1_1 to 1.
The first plot of the identify-task, showing the spectrum of the arc-lamp, gives now the correct number of pixels, about 2000, instead of about 4000. This error of a factor of two must have been caused by the incorrect value CD1_1=2.
However, the final sensitivity curve is still incorrect.
At the page www.helsinki.fi/~kklehtin/IRAF you can find the text-files std_old and std_new which are the old and new output files from the standard-task.
cheers, Kimmo
Francisco Valdes wrote on May 06, 2011
Hi Kimmo,
Thanks for the extra information.
I now need to know something specific about what you think is wrong. You say things are shifted but is there some quantitative feature that tells you this.
My concern is the following. Your arc fit looks quite reasonable but there are no constraining lines near the end of the spectrum, particularly in the blue. So the wavelength you derive for the first pixel can be off significantly from what you might thing it should be and will change with the fitting order you use. So leaving off the blue end, is there evidence for a shift? After you dispersion calibrate the standard star you could check the strong absorption lines and see if they are at the right wavelength, say using SPLOT. If the wavelengths appear correct in the dispersion calibrated standard star then there doesn't seem to be anything that can go wrong with sensfunc.
So we need a specific quantitative feature which is wrong to figure out the source of a problem.
Yours,
Frank
Thanks for the extra information.
I now need to know something specific about what you think is wrong. You say things are shifted but is there some quantitative feature that tells you this.
My concern is the following. Your arc fit looks quite reasonable but there are no constraining lines near the end of the spectrum, particularly in the blue. So the wavelength you derive for the first pixel can be off significantly from what you might thing it should be and will change with the fitting order you use. So leaving off the blue end, is there evidence for a shift? After you dispersion calibrate the standard star you could check the strong absorption lines and see if they are at the right wavelength, say using SPLOT. If the wavelengths appear correct in the dispersion calibrated standard star then there doesn't seem to be anything that can go wrong with sensfunc.
So we need a specific quantitative feature which is wrong to figure out the source of a problem.
Yours,
Frank
Last post on May 06, 2011