Page 1 of 1

dispersion different in ISIS

Posted: Wed Aug 22, 2018 6:57 pm
by Erik Bryssinck
Hello,

Benji Mauclaire pointed me to a problem with my HR spectra of X Cyg regarding to the dispersion. The dispersion in my spectra are to low. If i check this out in the FITS header, CDELT= 0.08 A/px for spectra processed with ISIS 5.9.3 , i think this is the dispersion ? I processed the same data-files of X Cyg but now in BASS 1.9.7, and i get a dispersion of 0.16 A/px ??? Maybe is this a problem in the software or what is the cause of this ? This is really a problem if you compare spectra 1D-FIT files which are processed with different software.

Maybe is the cause the binning of the CCD camera ? I operate the ccd-camera in 2x2binning. How deal with this or how can this be solved ?

thanks for your time,

Erik Bryssinck

Re: dispersion different in ISIS

Posted: Wed Aug 22, 2018 9:26 pm
by Robin Leadbeater
Hello Eric,

CDELT is the dispersion in A/bin in the calibrated spectrum which is not the same as A/pix. A/pix varies along the spectrum. When you wavelength calibrate the spectrum using a non linear calibration, the wavelength difference between the measurement at each pixel varies along the spectrum. To produce a spectrum with equal wavelength intervals, the software then interpolates the measured values. (Some professional software eg IRAF can leave the spectrum with unequal bins and includes the wavelength fit equation in the header but this is not so easy to plot or work with)

Some software eg Visual Spec, BASS just interpolate using a bin width equal to the average interval. (Average A/pix) This can produce artifacts though (aliasing). To reduce this effect, I understand ISIS resamples the spectrum at a smaller interval (by interpolating) so the CDelt is n * smaller than the average A/pix (In your case the average A/pix is ~0.16 but the A/bin in the final spectrum from ISIS is 0.08)

This does not matter though, the resolution is the same which is the important factor. If you are interested in the A/pix, this is approximately equal to the linear coefficient in the wavelength fit recorded in the log (A1)

Cheers
Robin

Re: dispersion different in ISIS

Posted: Thu Aug 23, 2018 3:11 am
by Christian Buil
This is the exact reason pointed by Robin. I divide the original pixel sampling by two for reduce risk of digital noise in the actual and future processing.

Christian

Re: dispersion different in ISIS

Posted: Thu Aug 23, 2018 1:48 pm
by Benjamin Mauclaire
Hi Christian,

Erik's setup is: C11_L200_1800_TriusSX814
The problem was that in the fits file header of Erik were: NAXIS1=1409 and CDELT1=0.08 with CRVAL1=6500.
This can't be correct with his setup.
In his last processed spectrum, NAXIS1=3298 and CDELT1=0.08 but he doesn't how this happened as he only swichted off heliocentric correction.

Unfortunatly, correcting digital noise risk add a new risk: this make confusing on knowing the real setup caracteristics and add an other operation on data, then it is moving away from pure initial data. Moreover all sensors don't need such precaution.

Benji

Re: dispersion different in ISIS

Posted: Fri Aug 24, 2018 7:28 am
by Erik Bryssinck
Benjamin Mauclaire wrote:Hi Christian,

The problem was that in the fits file header of Erik were: NAXIS1=1409 and CDELT1=0.08 with CRVAL1=6500.
This can't be correct with his setup.
In his last processed spectrum, NAXIS1=3298 and CDELT1=0.08 but he doesn't how this happened as he only swichted off heliocentric correction.
I checked several spectra from my archive and i see that the NAXIS1 is different in allmost all the spectra. The NAXIS1 is the amount of pixels of the 1D file and depended on the cropping which i use in ISIS.

best regards
Erik

Re: dispersion different in ISIS

Posted: Fri Aug 24, 2018 1:31 pm
by Robin Leadbeater
Hi Eric,

fits files are just arrays of numbers. NAXIS1 just means the number of rows in the fits file. (length of data axis 1)

In the original 2D image fits file this the width of the image in pixels. In the calibrated 1D spectrum it is the number of bins in the wavelength axis. These do not have to be the same.

There is no direct connection between NAXIS1 in the calibrated spectrum and NAXIS1 in the original image. For example in an extreme case of a non linear prism spectrograph the dispersion A/pixel can be many times larger at one end of the spectrum compared to the other but the A/bin will be the same at all wavelengths in a calibrated and linearised spectrum.

Even if you do not crop, NAXIS1 in the calibrated spectrum will not necessarily correspond to the number of pixels in the original image. NAXIS1 is the number of bins (along the wavelength axis) after upsampling in ISIS, wavelength calibration, linearisation and interpolation.

Robin

Re: dispersion different in ISIS

Posted: Fri Aug 24, 2018 1:37 pm
by Robin Leadbeater
Benji is correct in that to preserve the original data you should ideally use unequal bin sizes in the calibrated spectrum, in which case there will be a one to one correspondence between pixels and bins in the wavelength axis. In practise though most published spectra have been linearised with equal bin sizes.

Robin