20 replies to this topic

[revans]

I notice that when I compare flux calibrated stellar spectra with reference spectra there usually is no issue with the absorption/emission features being present and at the correct wavelengths but there is usually some difference in slope between the continuum in my spectra vs the reference spectra.  I'm doing pre-calibration and background subtraction and my instrument response curve looks normal.

 

What are the likely causes of this difference in continuum slopes and does it have any significance?   It seems to occur whether I use a largely automated process like Demetra or a more manual process like Rspec.

 

I couldn't find a reference library spectrum for an a1m iv star and so used an a0 iv star but I assumed that they should be similar at low resolution.  I was interested in differences in the continuum using different processing software tools vs a reference spectrum.

 

 

On the other hand, Vega worked out more precisely with Demetra processing compared to aov reference spectrum:

 

 

So, this is more the kind of match I'd like to see all the time.... and am not sure what is preventing it... the most obvious answer I guess is the instrument response curve but the same one can give a good continuum match to reference for one star and but not another.

 

 

Rick

[bkc]

I have not done any stellar spectroscopy but I have done some spectroscopy in the distant past. The thing that stands out to me is the absorption dips in the reference spectrum (green) and Rspec (red) spectrum are moving with respect to each other. The Rspec dips are at longer wavelengths when looking at the short wavelength end of the spectrum and it is reversed as the long wavelength end of the spectrum. For now, I am ignoring the absorption line near 660 nm.

 

Combine that observation with the decrease in overall intensity of the Rspec spectrum at long and short wavelengths. It could be explained by a misalignment in the spectrum. When I looked at the Rspec website they show the user aligning the spectrum. I wonder if the spectrum was not quite aligned with respect to the horizontal box the demo showed. This would compress the position of the absorption lines relative to their true locations if the Rspec algorithm did not quite scale the spectrum correctly.

 

The decrease in overall intensity at short and long wavelengths in the Rspec spectrum could then be explained if part of the spectrum at both ends of the rainbow were not is the analyzed rectangular box.

 

I am assuming there are no absorption/transmission issues since your Vega specturm looks good. I am also ignoring the Demetra spectrum, but I think it could be explained by a similar argument. I would have expected both software packages would be smart enough to allow for some rotation, but maybe that cannot be tolerated when trying to avoid contamination with other star spectra.

 

Well, it is just an idea.

 

bkc

[revans]

I have not done any stellar spectroscopy but I have done some spectroscopy in the distant past. The thing that stands out to me is the absorption dips in the reference spectrum (green) and Rspec (red) spectrum are moving with respect to each other. The Rspec dips are at longer wavelengths when looking at the short wavelength end of the spectrum and it is reversed as the long wavelength end of the spectrum. For now, I am ignoring the absorption line near 660 nm.

 

Combine that observation with the decrease in overall intensity of the Rspec spectrum at long and short wavelengths. It could be explained by a misalignment in the spectrum. When I looked at the Rspec website they show the user aligning the spectrum. I wonder if the spectrum was not quite aligned with respect to the horizontal box the demo showed. This would compress the position of the absorption lines relative to their true locations if the Rspec algorithm did not quite scale the spectrum correctly.

 

The decrease in overall intensity at short and long wavelengths in the Rspec spectrum could then be explained if part of the spectrum at both ends of the rainbow were not is the analyzed rectangular box.

 

I am assuming there are no absorption/transmission issues since your Vega specturm looks good. I am also ignoring the Demetra spectrum, but I think it could be explained by a similar argument. I would have expected both software packages would be smart enough to allow for some rotation, but maybe that cannot be tolerated when trying to avoid contamination with other star spectra.

 

Well, it is just an idea.

 

bkc

 

What you describe could likely be due to small errors in the calculation of slant and/or small errors in the correction of smile/rotation.  Using RSpec with a slit spectrograph is more manual than the automated processing Demetra can do.  It is easier to get a consistent result using Demetra with my data, but you can do pretty well with RSpec with care and patience.  

 

I was more interested in problems with the continuum slope than wavelength calibration errors.  There are many ways to address wavelength calibration errors but I fundamentally only know one way to affect the continuum at my beginners level of understanding.... and that is by discrepancies in producing the very best instrument response curve possible.  However, what I can't understand is why the same instrument response curve produced more continuum error in Beta Aurigae than in Vega when the spectra were taken with the same rig within a few days of each other with the spectrograph and telescope remaining intact as a unit in the meanwhile.  

 

I suspect that there must be other factors besides the instrument response curve that affect the slope of the continuum but don't have a great effect on wavelength calibration of absorption features.  I just don't know what they are and it has more or less left me at a plateau in my learning curve.  I just don't seem to be able to consistently get a perfect continuum result.  I suspect that background subtraction might play a pivotal role here, but I have been doing background subtraction so there is nothing to correct in this that I can think of.

 

Rick

[bkc]

Did you try rotating the same spectrum several time to see how the results varied? I still think the differences are explained by mechanical alignment issues. More correctly, the software inability to handle some alignment error cues and let you know. Can you send me a file with intensity vs wavelength? Just pm it to me if you can. It looks like an interesting problem.

 

If you have something producing absorption in the wings of the spectra, it should be there all of the time. And that still would not explain the differences between the Rspec and the Demetra. The same applies if you have a source producing energy in the middle of the spectra.

 

Brian

[robin_astro]

How are you deriving the response for these two spectra? What are you using as your reference star in each case ?

[revans]

How are you deriving the response for these two spectra? What are you using as your reference star in each case ?

For the ALPY calibration I created an instrument response curve generated in RSpec using Vega.  By the time I learned how to use Demetra, Vega had westered a bit and I created an instrument response curve in Demetra using Menkalinen (Beta Aurigae). 

 

Rick

 

 

[robin_astro]

There are far too many unknowns and possible causes to troubleshoot efficiently there. 

 

Do this type of exercise using a set of stars with actual known spectra (not just a published spectral classification) taken close to the same air mass

http://www.threehill...ts_20161105.pdf

following precisely in detail the procedure described here

http://www.threehill...on_20221222.pdf

[robin_astro]

If you are using  the response to self calibrate the same star this should always work to produce a match to your reference spectrum and if it does not you have a problem with how you are calculating the response. I would not recommend using the response generated in one program with another program (some software for example treat how the flat correction is done differently) or using the same response between observing sessions.

[revans]

Did you try rotating the same spectrum several time to see how the results varied? I still think the differences are explained by mechanical alignment issues. More correctly, the software inability to handle some alignment error cues and let you know. Can you send me a file with intensity vs wavelength? Just pm it to me if you can. It looks like an interesting problem.

 

If you have something producing absorption in the wings of the spectra, it should be there all of the time. And that still would not explain the differences between the Rspec and the Demetra. The same applies if you have a source producing energy in the middle of the spectra.

 

Brian

Here is a link to the wavelength calibrated spectra of Vega done 12/23/23 with the ALPY on my C5 f6.3 with a 533MC camera:

 

https://www.dropbox....k1pdqxk4kz&dl=0

 

Rick

[revans]

If you are using  the response to self calibrate the same star this should always work to produce a match to your reference spectrum and if it does not you have a problem with how you are calculating the response. I would not recommend using the response generated in one program with another program (some software for example treat how the flat correction is done differently) or using the same response between observing sessions.

Yes, you remind me that the reference spectrum I used to get the instrument response curve in Demetra matched the spectral class of Beta Aurigae from the Demetra library, whereas when I plotted the three curves comparing reference, RSpec and Demetra curves in thread #6 I used only a similar class star that I found in the RSpec library (a0iv).  So that may be why the Demetra curve doesn't match the reference precisely as far as the continuum goes.

 

When I have time tomorrow, I'll dig the correct reference star out of the Demetra library and see if it matches my result for Beta Aurigae obtained in Demetra.  It should.

 

Rick

[bkc]

Have you made any progress with your spectra?

Brian

[revans]

Have you made any progress with your spectra?

Brian

I need some clear and not too cold weather to get some more spectra with the DADOS.  I'd like to see better consistency with my continuum shape in the flux calibrated result whether I use Demetra or the more manual RSpec.  I think it will be a slow process but eventually I'm sure I'll get things right.  Meanwhile in another current thread it looks like the ALPY 200 is going to become a possibility soon and I'm interested in seeing how that compares as well.

 

Thanks for your suggestions.  

 

Rick

[bkc]

I will be interested in seeing more results. I played with the data you sent me and found that only a very slight rotation of the spectrum relative to the horizontal could explain the slight shift in peaks relative to the reference spectrum. Demetra talks about this and a smile distortion.

 

The shape of the background has more going on. I would like to see how much work is required to obtain good spectra as I might consider a little spectroscopy in the future.

Brian

 

Correction: Data you made available in one post.

Edited by bkc, 31 January 2024 - 10:23 PM.

[robin_astro]

  Meanwhile in another current thread it looks like the ALPY 200 is going to become a possibility soon and I'm interested in seeing how that compares as well.

 

 

I recommend you fully explore the performance of your current spectrographs  first. The ALPY200 is a special option for the  ALPY  for a very specific use (Very faint targets with strong broad features at the limit of detection using  very low resolution  ) This is primarily for experienced spectroscopists who  have already developed the skills to measure faint objects but want to explore this demanding region which is currently inaccessible using existing spectrographs. If you use it on bright targets you will be disappointed with the results compared with the standard ALPY600. 

 

Robin

Edited by robin_astro, 01 February 2024 - 09:18 AM.

[mborland]

Does anyone know of a more complete set of reference star spectra than the one available for ISIS? I frequently find that I can't find anything close to my target.

 

I often end up using a nearby A0V star and the Pickles file (p_a0v.dat), but I wouldn't think that is very reliable because  the distance to the A0V star surely matters in the spectrum. (I.e., a more distant A0V would be reddened.)

 

--Michael

Edited by mborland, 28 February 2024 - 12:11 AM.

[robin_astro]

 

I often end up using a nearby A0V star and the Pickles file (p_a0v.dat), but I wouldn't think that is very reliable because  the distance to the A0V star surely matters in the spectrum. (I.e., a more distant A0V would be reddened.)

 

 

Hi Michael,

 

Francois Teyssier's reference star finder on this page

http://www.astronomi...tsSpectro0.html

direct link

http://www.astronomi...rFinder_V3.xlsm

lists suitable hot main sequence stars with low reddening ie with low E(B-V) for use with the Pickles reference which is generally good enough.

 

MILES stars also need to be used with caution as many of them are potentially variable particularly if they are not main sequence (They were not really designed to be standard references). 

 

If you need a very precise calibration then better are the CALSPEC spectra which are standards calibrated in absolute flux also in ISIS. There are not many of them though so you generally need to take a couple of them at different air mass and correct for the extinction at the target air mass, which needs stable conditions. (This is how the professionals do it)

 

Cheers

Robin

 

EDIT:- corrected direct link  

Edited by robin_astro, 28 February 2024 - 07:19 AM.

[robin_astro]

 the distance to the A0V star surely matters in the spectrum. (I.e., a more distant A0V would be reddened.)

 

 

Interstellar extinction is distance dependent of course but  also very dependent on the direction. For example extinction looking out of the galactic plane is generally low compared with into the galaxy. There is a calculator in NED which gives the total galactic extinction in a particular direction, though this is of course an upper value and nearby stars within the galaxy will be much lower.

https://ned.ipac.cal...tion_calculator

 

Cheers

Robin

[robin_astro]

 suitable hot main sequence stars with low reddening ie with low E(B-V)

E(B-V) is the difference between the measured (B-V) brightness and the expected (B-V) based on the spectral class. If the spectral class is correct and there is no extinction (interstellar or circumstellar) then E(B-V) should be zero.

 

Here are  tables for (B-V) for all spectral classes

https://ui.adsabs.ha.....234F/abstract

Edited by robin_astro, 28 February 2024 - 07:38 AM.

[robin_astro]

there is also MELCHIORS, a new catalogue of flux calibrated stars which looks promising but the files are  difficult to read.

https://ui.adsabs.ha...A.107R/abstract

I believe Christian buil is developing a tool to help with this

[Olivier_Garde]

there is also MELCHIORS, a new catalogue of flux calibrated stars which looks promising but the files are  difficult to read.

https://ui.adsabs.ha...A.107R/abstract

I believe Christian buil is developing a tool to help with this

Yes and we will present a new web interface to use this MELCHIORS database of reference spectra and we have new online tools for spectgroscopy. To learn more, we have an online workshop the 12th march at 20UTC, to subscribe :

https://us06web.zoom...F75vvdYOwuFA_ 3
 

This workshop is in french but you can ask your question in english.

[robin_astro]

there is also MELCHIORS, a new catalogue of flux calibrated stars which looks promising but the files are  difficult to read.

https://ui.adsabs.ha...A.107R/abstract

I believe Christian buil is developing a tool to help with this

I have just checked and SpecInti already has a conversion tool. Item 20 in the "toolbox"

http://www.astrosurf...ecinti1_en.html

so it should be possible to use it to batch convert all the MELCHIORS spectra to a more easily readable format.

The list of hot main sequence MELCHIORS spectra could then be added to Francois Teysser's existing spreadsheet (similar to what is already done with the MILES spectra) which would then identify the most suitable nearby reference stars.

 

Robin