ISIS 5.2.1 version (for Alpy 600 and more)

[Christian Buil]

Hello,

La dernière version de ISIS est juste disponible :

http://www.astrosurf.com/buil/isis/isis.htm

Pour les utilisateurs de Alpy 600 questions fonctions supplémentaires
et aussi une aide spécifique :

http://www.astrosurf.com/buil/isis/guide_alpy/tuto.htm

Enfin une guide sur l’évaluation de la réponse instrumentale et
de la transmission atmosphérique :

http://www.astrosurf.com/buil/isis/guid ... method.htm

-------------------------------------------------------------------------------------------------------------------------
Hi all,

The latest version of ISIS just available:

http://www.astrosurf.com/buil/isis/isis_en.htm

For Alpy 600 users, new functions and a specific tutorial:

http://www.astrosurf.com/buil/isis/guid ... uto_en.htm

Also a guide about instrumental response evaluation (in french
sorry, but many illustrations).

http://www.astrosurf.com/buil/isis/guid ... method.htm

Christian

[Andrew Smith]

Christian - Thanks as always for a new version of ISIS and for the tutorial on atmospheric extinction and instrument response.

On the instrument response I am not sure I fully agree with your analysis. If you are using a flat field, as in the examples, to correct the target and reference image does this not cancel out the" true" instrument response (i.e. as you define it in your paper) and leave, for want of a better term, a "flat field response"? I am assuming that the flat field is made in such a way that it correctly illuminates the telescope entrance pupil and so experiences the same instrument response as the target and reference images and is not a synthetic flat field or corrected in some way as can be done in ISIS.

If a calibration unit is used between the telescope and the spectrograph you will get a mixed response curve from the telescope & flat field response.

In both cases the stability of the "instrument response" curve will depend on the stability of the flat field light source.

I may well have misunderstood the paper or how ISIS does the processing.

I would welcome your views on this.
Regards Andrew

EDIT - I had missed the last part of you tutorial (the paper ran out!) where you comment on the need for a stable lamp for the photometric method but I stll think my points stand for both methods.

[Robin Leadbeater]

Hi Andrew,

Yes the "true instrument response" for a slit spectrograph has very little to do with the instrument and is in effect the flat lamp spectrum, particularly if the flat lamp is in front of the telescope. Personally I am very wary of mixing flats from different sessions, but...

For the differential method the reference and target spectra would be taken during the same observing run therefore I would normally expect to use the same flat for all spectra. In this case there is no problem.

For the absolute method, the "true instrument response" is indeed only valid as long as everything which affects the instrument response remains unchanged. This includes the flat lamp and any changes in the system not included in the flat measurement (eg in the telescope optics if an internal flat lamp is used).
Even if the "true instrument response" is carried over between observing sessions. I would be inclined to take at least one reference star and use that to check all is ok. This (rather like a check star in photometry) would show up any potential problems with instrument response or extinction but would not distinguish between the causes. An examination of the two flats (the one used for the "true instrument response" and the one used to correct the target) could help identify any problems.

Cheers
Robin

[Christian Buil]

Hi Andrew and all,

For the differential méthod, the applied equation is:


The color effect of the flat-field is identical to the numerator and denominator in the second term, which has no influance. (this is the same situation for the optical spectral transmission or quantum efficiency...).

The situation is different for the photometric method. Here the basic equation which the parameters that affect the actual signal after ISIS processing:



- Topt is the optical specral transmission (telescope + spectrograph);

- Qccd is the spectral quantum efficiency of the detector;

- F is the spectrum of the calibration flat-field white light (almost a black body);

- Planck (2750 K) is a synthetic Planck curve (calculated) for a 2750 K black body (fixed internal parameter of ISIS).

The calculed Planck curve is for compensate the strong blue color deficit in flat-field typical "white" lamps.

The term [...] is the "true" instrumental "response". The value of Planck equation(2750 K) is to get a response curve approximately flat.
Easier to check.

The temperature of 2750 K is the typical value of a halogen lamp.

The presence of the term "F" explains the dependence to the color temperature of the lamp calibration in the compututed "response".

I have completed the web page these equations.

---------------------------------------------------------------------------------

Un peu plus d'explication dans la page consacré à l'évaluation de la réponse instrumentale,
en précisément mieux l'importance de la variation de la couleur de la lampe d'étalonage flat-field.

Christian

[Martin Dubs]

Hi everybody,
there seems to be some confusion about the instrument response.

The term [...] is the "true" instrumental "response".

If the F in the denominator would be the true flux of the near blackbody lamp then I would agree to call this instrument response. It would include the optical transmission and CCD efficiency, with some correction for the deviation of the flat from a true blackbody of 2750 K.
But in ISIS the calculation is done differently, with F is denoted the MEASURED signal of the flat, which also contains optical transmision of the spectrograph and CCD efficiency. Therefore in the term in brackets CCD efficiency and spectrograph transmission cancel out and a rather flat response curve results after dividing the spectrum by a reference. Basically it contains only the difference of the lamp spectrum from a blackbody and differences in the lamp optical path and star optical path through the instrument. This makes it very easy to approximate this response with a smooth curve. But it should be called, as Christian does in part of his description, PSEUDO response or calculated response. Therefore one also determines absolute flux before dividing the measured binned signal in ADU by the flat signal.
By the way the same procedure is followed in the treatment of echelle spectra, where the blaze function of the grating and detector sensitivity are eliminated in the same way (in Audela and ISIS).
I do not know if this is any clearer now, but I tried to see it from my point of view.

Regards, Martin

[Andrew Smith]

Thanks for your responses Robin & Christian.

I will restrict my comments to the differential case as I need to study the photometric method more before I fully understand it.

In the differential case I fully accept your equation but I think there is a subtle point that is not explicit in the equation. When you divide the flat into the target and reference star images this is done pixel by pixel and so you get an exact cancellation of the pixel by pixel instrument response. This leaves the imprint of the flat spectral response in each image. However, when the spectra is extracted unless the target and reference are recorded in on exactly the same pixels or the flat field is perfectly flat then they will not divide out exactly and leave an error.

With my LISA the calibration unit flat field is not perfect and shows a gradient perpendiculary to the dispersion. I know ISIS can flatten this but then dividing out by the flat will not fully cancel the instrument response.

I suspect the impact of this is quite small and only of concern to a nerd like me, but I will do some experiments to investigate as the long hours of daylight restrict real observing!

Regards Andrew

[Robin Leadbeater]

Basically it contains only the difference of the lamp spectrum from a blackbody and differences in the lamp optical path and star optical path through the instrument. This makes it very easy to approximate this response with a smooth curve. But it should be called, as Christian does in part of his description, PSEUDO response or calculated response.

Exactly. The term "instrument response" or even "response" seems completely misleading in this context as it has very little to do with the instrument or the way it responds to signals and in this case also appears to be software specific as it has been normalised using a hypothetical 2750K black body curve. I think we need a new name for it. flat lamp correction factor perhaps ?

Robin

[Robin Leadbeater]

In the differential case I fully accept your equation but I think there is a subtle point that is not explicit in the equation. When you divide the flat into the target and reference star images this is done pixel by pixel and so you get an exact cancellation of the pixel by pixel instrument response. This leaves the imprint of the flat spectral response in each image. However, when the spectra is extracted unless the target and reference are recorded in on exactly the same pixels or the flat field is perfectly flat then they will not divide out exactly and leave an error.

Hi Andrew,

As far as I am aware, In all cases the spectrum images are still divided by the flat pixel by pixel. This is key to the use of flats. The discussions on instrument response only relate to the way the effect of the flat lamp spectrum, which is imprinted on the profile as a consequence of the flat correction, is handled

Cheers
Robin

[Andrew Smith]

I agree Robin but that was not the point I was trying to make! So when words fails let’s try some equations. If the measured flux is m and the true flux is M with instrument response I and atmospheric extinction A then using t, r and f to label the target, reference and flat field then:

1) mt = Mt.I.At for the target
2) mr =Mr.I.Ar for the reference
3) mf = Mf.I for the flat field

The assumption under differential correction is that At=Ar =A.

So if the target and reference are flat fielded using mf then:

4) mt’ = Mt.I.A/Mf.I = Mt.A/Mf
5) mr’ = Mr.I.A/Mf.I = Mr.A/Mf

Where I have used the ‘ to label the flat fielded corrected flux and I have ignored the fact that in flat fielding the result is normally multiplied by some average of the flat filed. As we agree the division is done pixel by pixel.

Now the target and reference spectra are extracted into a 1D profile we apply the equation

Target True profile = Reference True Profile.Target Observed Profile/Reference Observed Profile

So if we take Mr from the Miles data base (suitably filtered to match the observed resolution) say Mrdb

Then we get:

Mt = Mrdb.mt’/mr’ =( Mt.A/Mf)/Mr.A/Mf) = Mt as long as Mrdb = Mr which we assume.

But this is wrong!

For while the atmospheric extinction A in mt’ and mr’ was measured at the same pixel (x,y) for all (x,y) that is not true of Mf which were measured at (x,yt) and (x,yr) where yt is the row the target spectra is recorded on and yr the row the reference spectra is measured on. (This is complicated by the extraction of the optimal profile across row but the point stands.)

So unless Mf(x,yr) = Mf(x,yt) i.e. the flat is truly flat then the cancellation is not perfect unless the target and reference are recorded on the same rows.

A similar analysis can be done for the flat taken at the spectrograph and then in addition the telescope response will not fully cancel out.

As I said before my LISA has a strong gradient in flats made using the calibration unit so I need to see if this is a real or theoretical issue?

Regards Andrew

[Christian Buil]

The equations given are fisrt approximation of course !
Pseudo noise are always presents (physical measure is a complex task )

I agree, the "F" term is actually the flat-field measured signal (some problems with English).
A more accurate equation can be for example:



with Tspe, the spectral transmission of spectrograph part.

If the calibration module is in front the entrance slit (situation for LISA/Alpy with optional module),
we have:



with P, now is the emitted spectrum by the "white" flat-field calibration lamp,

and, in first approximation:



In fisrt approximation because, I agree, the optical path can be not
the same for the stellar and flat-field flux for example (i.e. gradiant effect).

Note also, the order of operations is important : the flat-field correction
is made at the 2D level image (not at the profile level). So the CCD pixel to pixel
non uniformty response is well erased (and the vertical position of spectrum trace
is less important in the image, PRNU it is the main justification of flat-field correction).

Christian