AG Peg and UV Aur

[Francisco Campos]

Hello,

here I report observations of two symbiotic stars. After a lot of problems with ISIS, finally I had to reduce them with VSpec.

The main problems with ISIS are that, because I work with a DADOS spectrograph and there is not an internal calibration lamp, I had to use a current energy-saving lamp and calibrate the spectra using only 5 lines (those of mercury at 4046, 4358, 5461 and 5770 A, and one of argon at 6308 A). Very soon I will built a better calibration lamp based on a RELCO starter.

Due to this issue, I have failed trying to calibrate the spectra with ISIS. So I proceded in a similar way with VSpec:

Firstly, I took single spectra of the variable star (8-10 single spectra) and the comparison stars (HD 208565 for AG Peg and HD 39357 for UV Aur). Without moving the scope, I took the calibration image with the lamp and a paper covering the scope objetive. Then, I took bias, darks and flats.

Then, with MaximDl, I rotate and calibrate the images with bias and darks only, because if I calibrate with flats, I don't know why, the spectra took a strange appearance and the intensities of the emission lines became completely altered and unreal What are the Maxim doing bad?

So, with VSpec. I calibrate all the images in lambda. I do binning of the star and background. I substract the background. Repeat the same with the comparison star (I think it is a good way to remove light pollution).

Then, I use the comparison star to determine the instrumental response, using the VSpec library.

Finally, I divide the profile of the variable star with the instrumental response.

Problem: I can't save the spectra in BeSS format: VSpec saves a fit image but if I try to load it, VSpec returns the error: "file fit with more than two plans: unreadable".

Well, I hope to do all this work in the future with ISIS. But I got stuck trying to calibrate in lambda only with the calibration lamp.

And here is a spectra of UV Aur. Please, if it is something bad, tell me how to correct it

_uv aur_20151213_914-franciscocampos.jpg (84.27 KiB) Viewed 8565 times

Greetings,
Fran

[Robin Leadbeater]

Hi Fran,

Problem: I can't save the spectra in BeSS format: VSpec saves a fit image but if I try to load it, VSpec returns the error: "file fit with more than two plans: unreadable".

If you save the spectrum profile as a fits file (for example using "file" "export fit for BeSS" it is saved as a 1D profile, not a 2D image. You should be able to reload it (or any other spectrum profile saved as a fits file) back into Vspec using the "file" "open profile" command (not "file" "open image" which expects a 2D image)

Hope this helps
Robin

[Robin Leadbeater]

The main problems with ISIS are that, because I work with a DADOS spectrograph and there is not an internal calibration lamp, I had to use a current energy-saving lamp and calibrate the spectra using only 5 lines (those of mercury at 4046, 4358, 5461 and 5770 A, and one of argon at 6308 A). Very soon I will built a better calibration lamp based on a RELCO starter.

Due to this issue, I have failed trying to calibrate the spectra with ISIS. So I proceded in a similar way with VSpec:

You can use ISIS to calibrate a spectrum from any spectrograph using any lines, similar to VSpec.

Display the profile you want to calibrate
Click the "dispersion" button, fill in the wavelengths of the lines you want to use and measure their positions in the spectrum profile (double click either side of each line)
click "compute polynomial"

You can then use this polynomial plus the position of one line to calibrate any spectrum using the "predefined dispersion equation" mode in the "general" tab

You can even generate a special file with the lines you want to use and the approximate dispersion and ISIS will identify the lines automatically for any spectrograph and calibration lamp in the same way as it can for the Shelyak spectrographs. This is used with the "file mode" option in the "general" tab.

There is more information on these methods somewhere in the ISIS tutorials, for example here
http://www.astrosurf.com/buil/isis/tuto ... ion_us.htm
though this is for an earlier version of ISIS so the screens are slightly different. (There are probably more up to date examples elsewhere)

Robin

[Francisco Campos]

Thank you very much for your help, Robin,

I didn't see the filter "format" and that's the reason I couldn't open the generated BeSS files. Problem solved!

About calibrating with ISIS... well, I will try again with these user guide. But I have to admit that ISIS is confusing for me: it is much more powerful and confortable treating spectra than VSpec, but I find VSpec more intuitive and easy to understand.

Another question I am considering is: are the flats really necessary? I don't know why, but if I process a spectra with flats and MaximDl (taken with an halogen lamp and exposing up to 80% of the full histogram), the results are strange, very strange: all the profile seems very different compared with the original, the intensity of the band changes al lot, the background is not uniform but very strong at the edges of the image, etc...

I have a lot to learn...

Greetings
Fran

[Robin Leadbeater]

Hi Fran,

I think taking flats is important otherwise you cannot be sure that any features you see on the spectrum are real or flat defects. (There are some people who advocate not using flats at all but of course they are just burying their heads in the sand !)

Spectroscopic flats are rather complex compared with conventional image flats though as they contain not only the usual spatial effects (vignetting, dust, pixel/pixel variations etc) but also wavelength dependent effects (The spectrum of the flat lamp and the response of the instrument etc.)

If you are using a slit spectrograph you can (and should) use flats. (using something with a smooth spectrum,normally a Halogen lamp - uniformity is less important than with imaging flats as we are only interested in the light that comes through the narrow slit) As well as getting rid of the usual flat defects it also removes the instrument response, replacing it with the smooth (black body) flat lamp spectrum, which is generally easier to deal with than the often more lumpy CCD response.
This is the reason the shape of the spectrum changes when you apply a flat. Do not worry, the effect cancels provided you apply the same flat correction to your reference star when you make the instrument response correction. (Note that some programs like ISIS assume that the flat lamp is a black body of a certain temperature and takes this into account. As a result the flat corrected spectrum produced by ISIS appears less distorted by the flat lamp spectrum)
One problem when using Halogen flats with full range spectra though like those produced by the ALPY/LISA is that the intensity of the flat lamp is much lower at the blue end. It is important to produce a flat which has enough counts at the blue end (so that the flat correction does not introduce additional noise), while not saturating at other wavelengths. To do this, take many well exposed flat images and sum them (I typically take ~30 flats)

In general the subject of spectroscopic flats is a fascinating one and I believe the story is not complete yet as far as amateurs are concerned and more research is needed. They are very much a necessary evil and I from what I have experienced I suspect there is no such thing as the perfect flat.

There are different sources and types of flat defects and ways of correcting them. It depends on what defects you have in your flat. To solve all flat problems, you may even need to use a combination of different types of flats (common with pro setups). From my experience and others I suspect only a light source ahead of the scope can work for all circumstances and even that is difficult to get perfect. For example there is an interesting recent post by Christian on here concerning the new internal LHIRES flat lamp (in French)
https://groups.yahoo.com/neo/groups/spe ... ages/16454
and some work done by Lothar Schanne investigating the use of EL panels (almost ideal though they have a limited wavelength range)
http://www.astrospectroscopy.eu/Einstei ... lats_e.htm
If you have fine scale interference ripples and a spectrograph with flexure you may even need to take the flat with the scope in the same orientation as the target
http://www.astrosurf.com/buil/infrared/obs.htm

Personally I use a halogen lamp and T shirt for the LHIRES but I keep my CCD clean and do not have intereference ripples.
For the ALPY I use the internal halogen lamp which is very easy to use and seems to work ok comparing standard star spectra but I have not done any comparisons with T shirt flats.
I plan to try the LHIRES upgrade flat lamp but given what Christian has said, I suspect it may need more comparisons done to understand it better.
For the Star Analyser, I take flats but do not use them conventionally (Slitless flats are very complex and it is currently beyond the scope of amateurs to use them to correct the spectrum image directly) I check the flat for localised defects and place the star to avoid them. The target and reference stars are placed in the same position in the field so any defects cancel as far as possible.

Robin

[Francisco Campos]

Hi Robin,

thank you very much for your help. I take usually my flats covering the telescope with a diffuse plastic and illuminating it with a 150W halogen lamp, I take 10 flats exposing 3/4 of the full histogram and I average them to obtain a master flat. Next time I will sum all the flats instead of average them.

Yes, in extreme blue (3600-3800 A) the flat is so faint that the noise is very strong. And yesterday, while I was treating a spectra of Z And and his calibration star, I noticed that the calibration star seemed saturated in the blue zone (the raw images had up to 26.000 ADU's only): VSpec measures intensities of 32767 ADU's in the blue zone, so I am worried about doing something bad.

Well, I have reanalyzed an observation of Z And with all the images calibrated with bias, darks and flats. The light pollution has been removed by substracting the background of the same spectra. The instrumental response has been calibrated using the star HD 222439 (A0 V). In the next figure you can see both spectra: in blue the original spectra obtained from images without flat treatement; in pink, the spectra calibrated with flats. The difference are quite relevant. Do you think the spectra is ok?

Greetings!
Fran

[Robin Leadbeater]

Hi Fran,

I think summing or (floating point) averaging the flats should give the same result.

Yes the difference between the flat corrected and uncorrected spectra does look strange. Any flat field corrections are normally at small scales, not broad differences in scale at the blue end like this. (Any broad shape effects should be removed by the instrument response). I can suggest a few things to check.

Have you corrected your flats with flat bias and darks ?

Are you bias and dark correcting your spectrum images before flat correcting ?

VSpec has a limited dynamic range compared with more modern image processing packages so check that the preprocessed spectrum images are not saturating in VSpec

If you process the reference star star spectra with and without flat correction using the instrument responses you have calculated do they both look the same and agree with the library version ? (This will check that your flat corrected instrument response is working ok)

Robin

[Robin Leadbeater]

And yesterday, while I was treating a spectra of Z And and his calibration star, I noticed that the calibration star seemed saturated in the blue zone (the raw images had up to 26.000 ADU's only): VSpec measures intensities of 32767 ADU's in the blue zone, so I am worried about doing something bad.

Yes I think that could be the problem. If the flat corrected image of the reference star is saturated in the blue in VSpec, the calculated instrument response will be too low at the bue end which will make the flat corrected Z And spectrum too high. You could try scaling the reference star flat corrected image (divide by 2 say) before loading it in VSpec to see if this corrects the problem

Cheers
Robin

[Francisco Campos]

And yesterday, while I was treating a spectra of Z And and his calibration star, I noticed that the calibration star seemed saturated in the blue zone (the raw images had up to 26.000 ADU's only): VSpec measures intensities of 32767 ADU's in the blue zone, so I am worried about doing something bad.

Yes I think that could be the problem. If the flat corrected image of the reference star is saturated in the blue in VSpec, the calculated instrument response will be too low at the bue end which will make the flat corrected Z And spectrum too high. You could try scaling the reference star flat corrected image (divide by 2 say) before loading it in VSpec to see if this corrects the problem

Cheers
Robin

Ok, I reanalyzed the spectra of Z And and checked how MaximDl does the pretreatment: flats have been treated (substracted) by bias and darks. To avoid saturation problems, I have stretched the histogram (linealy) with MaximDl in order to avoid saturation in VSpec. The final spectra is quite different and closer to the raw one, but I still see differences, specially if I compare the Balmer bands intensities. And this one concerns me specially, because all the data that I could extract from the spectra won't be valid.

In this image, the raw spectra is the pink one, and the new corrected one, in blue.

z and tratada y strech con el flat.jpg (58.32 KiB) Viewed 8380 times

What it is clear is that correcting images with bias and darks, the resulting images change very little (removes hot pixels and very few more). But the real problem comes with flats. When I process pictures or photometric images with flats, the image changes litte (basically, vignetting dissapears) but now, I don't understand what's happening.

Here's an image of the Z And spectra with only darks anb bias:

espectre cru.PNG (233.29 KiB) Viewed 8380 times

and here's the same image with darks, bias and flats. My spectrograph has 3 slits and, because I fill 3/4 of the full histogram and the slit I use is the narrower one, the two others are fully saturated (I don't use it, of course).

Espectre tractat.PNG (153.95 KiB) Viewed 8380 times

I hope to resolve the mistery asap and reduce other spectra...

Greetings
Fran

[Robin Leadbeater]

Hello Fran,

Your flat corrected spectrum image looks ok but you might need to average more flats to reduce the noise at the extreme blue end.

I will try to explain more clearly why the flat corrected raw spectrum looks different to one which has not been corrected

Consider a perfect spectrograph with no flat defects and ignoring atmospheric absorption, then :-

uncorrected measured spectrum = true spectrum * instrument response ------ 1
and the flat = flat lamp spectrum * instrument response

so when you do a flat correction:-

flat corrected measured spectrum = uncorrected measured spectrum / flat

so

flat corrected measured spectrum = (true spectrum * instrument response) / (flat lamp spectrum * instrument response)

so

flat corrected measured spectrum = true spectrum / flat lamp spectrum ------- 2

Note how ---1 is different from ---2 This is why you see a big difference in raw spectrum shape between the flat corrected and uncorrected raw spectrum

It also explains why the intensity of the flat corrected spectrum is much higher at the blue end (the halogen flat lamp spectrum is very low here)

The effect of the flat lamp spectrum will be the same for both target and reference stars so the effect of the flat lamp spectrum will cancel when you
correct for "instrument response"

Actually, as you can see, if you do a flat correction, you are in fact using the reference star to remove the flat lamp spectrum (and corrrect for atmospheric absorption), not to correct for the instrument response but most people still describe it as an instrument response correction

Robin