Bin x 2 - problem with ISIS

[Peter Velez]

Thanks for your thoughts David - there is much to digest here.

Unfortunately, I don't have luxury of refocussing the LISA each night - my set up is remote. Its a 6+ hour drive away.

What I can certainly do is refocus on a cold day or early morning or dusk when the temperature is low in preparation for the later temperatures. I had suspected that the focus undertaken during the day was fine but its doubtless been quite soft by night.

As to the analysis you suggest, I will work through that

Pete

[Peter Velez]

Hi,
Could you join a FITS file of the 2D spectrum of your calibration lamp in bin 2x2?
Be aware not to over expose the red lines.
Is your calibartion lamp a Relco?

Benji

Benji

I'm unable to attach a .fit file here for some reason. Is there an alternate way of providing it?

Pete

[Benjamin Mauclaire]

http://dl.free.fr

[Robin Leadbeater]

I'm unable to attach a .fit file here for some reason. Is there an alternate way of providing it?

Pete

You can add it as a zip file, as here for example
http://www.spectro-aras.com/forum/viewt ... 457#p13667

Cheers
Robin

[Peter Velez]

OK

here are 2 spectra I have processed using ISIS after having successfully calibrated a B9V star. Each is taken from 3 x 10 second images of the Ar Ne calibration lamp taken with the LISA calibration unit. The bulb is the standard Shelyak supplied bulb. One set was taken during the day with an ambient temperature of 16C, the other at night with an ambient temp of around 9C.

I plotted each in Plot Spectra. Its clear that the 16C set are not as well aligned with the expected emission lines. For example, the 9C spectra has a clear peak at 5944 while the 16C has a corresponding (and higher) peak at around 5948. I suspect focus and temperature is the issue here.

With the 9C Ar Ne frames I can calibrate with ISIS using an A reference star to under 0.5 RMS - with a bit a tweaking. Interestingly, I find that I need to adjust the X-pixel for the 5944 line slightly to achieve a match. Visually, it looks like x-pixel 517 or thereabouts. No match with the calibration assistant. Drop that to say 514 and I have a good match.

I suspect that the LISA was focussed on a cool day, but not at an ambient temp that matches the temperature when I am imaging - so trying to use calibration frames taken the day after imaging is not a sound strategy. I expect I also need to wait for a very cool day or around dusk after the scope has cooled down to refocus.

Pete

[Peter Velez]

I've been messing about with ISIS again this morning. Here is what I have determined. I am sure this is all obvious to everyone so I should perhaps change the heading to Obvious Things About Calibration (which I should have known):

1. Ar Ne frames should be taken at the same time (and therefore the same temperature) as reference star images - with my set up I have found that the position of emission lines may shift by as much as 4 A for images with a delta temp of 7 degrees C. So an attempt to use the Calibration Assistant function to match a calibration frame and the Ha line for a reference star will fail if there is a significant difference between the 2 images.

2. The Calibration Assistant for me fails in 2 ways - firstly, ISIS can't generate a match based on the nominated x-pixel for the reference wavelength (for me 5944) and so selects an alternate line which is way out. This has been fairly common for me. Secondly, ISIS accepts the nominated x-pixel for the reference wavelength but automatically determines an incorrect Slant which prevents suitable matching. I find that short exposures (10 seconds) are more likely to generate an accurate Slant than longer frames. This seems counterintuitive to me - at 30 seconds I am not quite saturating the Ar Ne frames and I would have thought that it would have been easier for ISIS to determine the slant. But I suspect that the internal focus on my LISA is a bit soft which smears the lines used to determine slant resulting in poor outcomes.

3. The internal focus of the LISA is temperature dependant. I suspect that I have focussed my instrument during the day and I have not replicated observing conditions. So my focus looks good and sharp when its set up but produces poor results even when I take Ar Ne frames at the same time as my reference star images.

4. This is odd to me - I still struggle to use a .lst file to calibrate images binned x2. Today, I used the Calibration Assistant to calibrate a data set with RMS of about 0.33. I then exported the emission lines used in that polynomial using the Dispersion function and set up a new .lst file. I then determined the dispersion for the Ha line using the Dispersion (no parameter) command line and added that to the .lst. Selecting the same x-pixel for the reference line at 5944 using the .lst, I should in theory have generated the same calibration. Instead it generated an RMS of 18.32. Clearly with my data ISIS needs the second reference given by the Ha co-ordinate too.

I'll keep playing with the data to see if I can refine my technique.

Pete

[Peter Velez]

OK, I will offer my thoughts.

... I usually focus on the 5852 line in the middle of the spectrum and aim for a FWHM of about 2.5 pixels. This is with the 23 micron slit and binning x1. I have the same CCD in my camera.

David

David, a question - you bin x2 using the 23 micron slit. I had thought that the 35 micron slit was a better match for our camera. The theoretical FWHM achievable by the LISA for that slit size is around 1.7 and 2.6 for the 35 micron option. Have you determined that the narrower slit is better after taking account of your seeing?

I played around with SimSpec but couldn't get a sensible result - doubtless a function of the GIGO principle - Garbage In, Garbage Out - as I haven't included the appropriate parameters. At SSO seeing is in the range of 1.5 - 2.5 arcseconds per pixel.

If resolution is important and I have good quality skies, a narrower slit is a better option so perhaps I am better served joining you with the 23 micron slit.

What do you think?

Pete

[Peter Velez]

Hi David,
The method you described to compute RP (resolution power) will not describe LISA caracterictics, ie non linear dispersion and focus along wavelength band width covered.
RP depends on wavelength as you said, as RP=Lambda/Delta_lambda where Delta_lambda=FWHM(of line at Lambda), it is needed to use the FWHM associated to the calibration line.
Therefore RP of blue calibration lines will be lower and then describes the unfocussed of this wavelength zone.
Cheers,
Benji

Thanks Benji - I understand (I think).

The Dispersion command in ISIS generates a list of the dispersion for specified wavelengths. I generally crop my spectra at 3900 and 7500. The range of dispersions at these wavelengths for my last properly calibrated spectrum was 3.6596 and 3.6590 respectively with a minimum dispersion of 3.6122 at 5031 A. This confirms that the LISA dispersion is not linear.

Pete

[Peter Velez]

You can investigate how your resolution varies across the spectrum by measuring the FWHM of the calibration lines in pixels using the FWHM function in ISIS. Multiply these FWHM values by the dispersion in A/pixel from the A1 value of the polynomial fit to get the FWHM in A. Then divide the wavelength of each line by its FWHM to get the resolution you are achieving at each line. This is a very educational graph! You can then experiment with the focus to see how the shape of that graph changes as you vary the focus. What you find is that there is a compromise between balancing the resolutions at the blue and red ends of the spectrum. I can get a resolution of 1200 or better in the centre of the spectrum but it drops to around 700 at the ends. This reflects the compromises necessarily made in designing the LISA at an affordable price.

David,

I had a crack at this yesterday as well. The limits of my Excel skills were exposed by the graphing! What was clear is that the FWHM function generated inconsistent results for me when measuring the calibrated spectrum. I will try again with a raw image and see how that goes.

Pete

[David Boyd]

Pete,

I almost always use the 23 micron slit to get good resolution on spectral lines. This matches my seeing which averages around 3 arcsec with a C11 plus focal reducer. I bin x1 with my camera to get a sampling factor of just over 3. Binning x2 would be undersampling. In my experience theoretical resolutions predicted by designs are often not achieved in practice. With seeing as good as you mention, why use a wide slit? Your resolution is going to be determined by the star image size. At 35 micron you are almost running slitless (slight exaggeration ).

I am puzzled why the ISIS slant function does not give you consistent results. Provided you are measuring a well exposed calibration line, it should be accurate. One thing I would suggest if you have doubts about the stability of your LISA over an evening is to take calibration frames before and after taking spectra of each star. You can then average these in ISIS. This may slightly worsen resolution but should give a more accurate wavelength calibration.

Inconsistent results for the FWHM function may point to an underexposed calibration frame or a misaligned camera. It is worth checking that the dispersion of the spectrum is well aligned along the X axis so your calibration lines are aligned along the Y axis.

I am still puzzled why the .lst file method does not work well. This may be because of your focus and calibration issues. Once you have refocused the LISA and generated some good well-exposed calibration frames you should be able to use the .lst method successfully. This is the .lst file I use:
4
1.8
7383.980
7272.936
7067.218
6965.431
6506.528
6402.248
6143.063
5944.834
5852.488
5400.562
4965.080
4879.864
4764.865
4545.052
4158.590

One thing to note is that your spectrum will not be accurately wavelength calibrated beyond the region covered by these lines because extrapolating the 4th order polynomial fit is unconstrained.

David