spectro-aras.com

Ici :

http://www.astrosurf.com/buil/instrument_response/

un article qui décrit en détails ma manière de calculer la réponse instrumentale, un sujet très recurant dans les discussions (!).

J'ai essayé de rationaliser un peu la chose en allant vers le principal.

An article about the computation of instrument response based on artificial and natural light sources (in French but illustrated).

Christian

Merci pour cet article.

Thanks Christian for this interesting and important new approach to flux calibration. There are already some discussions on this new method elsewhere so I thought I would bring them together here and add some more comments for discussion

http://www.spectro-aras.com/forum/viewt ... 311#p12621
http://www.spectro-aras.com/forum/viewt ... =10#p12661
http://www.spectro-aras.com/forum/viewt ... 315#p12657

(Note that ISIS has already been modified to allow the new technique to be used.This means any instrument responses produced using 5.9.3 or earlier will not work correctly with versions later than this)

The proposed method is very interesting and has a number of advantages over the method of using reference stars for each observation measured close in time and position to the target. (Less noise at the uv end from the flat correction. No need to do flats or find reference stars for every observation)

It is similar in a lot of ways to how professionals observe. They typically have very stable instruments and predictable atmospheric conditions compared with my setup though so I need to be cautious before moving to this method.

For the method using a specific reference star for each target, only two conditions need be met to give an accurate result. We need a star with a reliable reference spectrum close in time and position and nothing in the instrument or atmosphere should change between measuring the target and reference star.

With the new method which does not use a conventional flat correction and uses a master instrument response and an atmospheric model, several different conditions need to be met

1. The flat field must be free of local defects such as dust etc. Or alternatively separate flat correction containing this information needs to be made.

2. The instrument response needs to be stable long term (weeks/months?) between measuring the master instrument response and using it on the target. I am not sure if I can meet this conditon because:-

a I remove and replace my spectrographs and cameras often. Also the LHIRES for example is known to be not very stable and I change wavelengths. The new technique expresses the details in the instrument response (eg ripples) by wavelength instead of correcting by pixel postion using a flat. This should help with stability problems but I need to check if the intrument response is actually repeatable long term.

b My telescope (C11 +reducer) suffers from chromatic aberration which changes the instrument response depending on focus. The focus would need to be the same for the master instrument response and the target which would be difficult to guarantee. (There is also the issue of atmospheric dispersion as I do not observe at the parallactic angle). Observing a specific reference star close to the target helps manage this problem

3. The atmospheric conditions need to be constant long term between measuring the master instrument response and making the observation, or at least for some hours if corrected for using a reference star measurement on the night. I think this might be difficult to achieve at my location or at least would limit the number of nights I could observe.

Fortunately it should be quite easy to test if these conditions can be met using some existing observations. If I measure a master instrument response as suggested and then use this to re-reduce some past reference star measurements I should be able to see how accurate this new method is. I would recommend anyone considering this new technique to try this.


Cheers
Robin

An additional comment on Robin's point 2b. Slit losses at slit orientations other than the parallactic angle can have a significant airmass- and wavelength-dependent effect on the spectral profile depending on how well seeing is matched to slit width.
http://www.caha.es/pedraz/RS/refract_slit.html
Using a reference star spectrum close to the target star largely mitigates this effect as both spectra suffer the same slit losses. Without using a reference star, this becomes a much more significant issue in cases where is it not possible for whatever reason to adopt the parallactic angle.

David

David, yes, atmospheric refraction is very important topic.

In my paper, I well separate this problem from the calculation of the instrument response.

Resolving the errors induced by refraction by observing a reference star close to the target is certainly a possibility, but ... (1) the effect of diffraction is quite variable depending on the way to guide the stars, also the precision is not always at the rendez-vous (2) this is a cumbersome procedure if it must be carried out systematically.

A modern approach for amateur spectrography is to correct the diffraction at the acquisition time. Our future is to adopt an effective AOD (costly !?) or adapt the observation strategy. For example, here by intentionally orient the slit perpendicular to the horizon line:



This can be optimized (motorization, computerization) in the future... See also this topic:

http://www.spectro-aras.com/forum/viewt ... f=8&t=2109

Christian

An English version of my text concerning instrument response is now available :

http://www.astrosurf.com/buil/instrument_response_us/

Some infos are added.

Une mise à jour (en français) ici : http://www.astrosurf.com/buil/instrument_response/

Christian

Hello Christian

Robin Leadbeater wrote: 1. The flat field must be free of local defects such as dust etc. Or alternatively separate flat correction containing this information needs to be made.

Instead of using a PRNU as you suggested, would it be better to make this correction using a "residual flat image" made by dividing the original flat by the column averaged flat? This would give a cleaner separation of spatial and spectral components in the flat. The residual flat image containing the spatial component would be used to correct the spectrum image and the column averaged flat containing the spectral component would be used to generate the instrument response as in your method

Cheers
Robin

Bonjour Christian.

Merci pour ton article très instructif.
J'ai utilisé ta méthode pour corriger un spectre de v694 Mon que j'ai acquis en Mars 2019, avec un LHIRES3, un réseau à 150tr/mm et une ASI183MM Pro.

Pour la réduction du spectre, j'avais initialement utilisé une étoile de référence proche et acquise immédiatement après v694 Mon. Mais la correction reste difficile dans le bleue, probablement dû à la faible hauteur de l'étoile sur l'horizon.

Avec la méthode que tu proposes dans ton article, j'obtiens une meilleure correction dans le bleu. J'ai essayé cette méthode sur un spectre de la même étoile obtenu en Février (30 jours avant), mais cette fois, la correction reste mauvaise dans le bleu.
Peut-être est-ce dû cette fois au paramètre AOD et autres paramètres atmosphériques (que j'ignore) qui ont changé fondamentalement entre Février et Mars?

Dans tous les cas, la méthode donne la possibilité de rattraper des acquisitions qui seraient entachées de défaut.

Ibrahima.

Merci pour les tests Ibrahima,

Le chromatisme atmosphérique demeure un problème important, trop souvent ignoré. En fait, pour aller au bout de la démarche. soit on observe proche du méridien ou pas trop bas sur l'horizon, ce qui est bien sur contraignant, soit on dispose d'une méthode pour corriger efficacement le chromatisme. Quelque part, viser une étoile de référence proche est un moyen pour essayer de corriger le chromatisme... mais sans mesurer celui-ci directement. Mais la technique est consolatrice de temps, avec aussi un risque de décalage temporel induisant une erreur (variation significative de l'angle paralactique dans le temps et en fonction de l'edroit visée).

Ce fameux angle, entre l'axe long de la fente et la direction verticale, devrait être un paramètre plus systématiquement rappelé. Par exemple, dans les exemples exposés, compte tenu de la date d'observation, est-il possible d'avoir la valeur de cet angle (outils "Vitesse héliocentrique").

Mon avis est qu'il faut développer des méthodes matérielles et/ou algorithmiques pour corriger ce biais de chromatisme (ici les télescopes AltAz sont potentiellement avantagés, mais rares, à moins d'utiliser un Dobson).

Je vois pas mal d'observateurs orienter la fente (axe long) suivant l'axe AD, ce qui ne me semble pas une excellente chose car on observe souvent assez proche du méridien local, et alors le défaut de chromatisme est alors maximal.

Ma réflexion actuelle est de modifier dynamiquement l'orientation de la fente (et donc du spectro) en utiilisant un niveau à bulle électrique ou un inclinomètre...

Christian

.

... ou d'utiliser un trou à la place de la fente.

Benjamin