Spectral shift during sequence acquisition

[Christian Buil]

The mechanical flexure of Lhires III during a long time observation sequence is a classical and well know problem.

Here a typical example concerning a recent observation of faint star AZ Cas (20 x 600 s sequence on a C9 and
a Lhires III 2400 l/mm, 19-microns slit).

Most important: I have taken one neon lamp spectrum just before starting of the sequence and
one neon lamp spectrum just at the end of the sequence for control (non intermediate neon lamp for the occasion).

Only the individual first neon spectrum is used for spectral calibration under ISIS. ISIS use this unique
spectrum for compute the FWHM (mean FWHM of three available lines). Also, the software use this unique starting
neon lamp spectrum for compute spectral dispersion equation.

Here a typical comparison on the fisrt AZ Cas spectrum and the latest of the sequence (remeber,
all calibrated by using the initial neon spectrum):



The spectral shift (or drift) is evident (here 2.34 pixel of 9.08 size).

Now I activate the spectral shift registration of ISIS (General tab):



The result is better. Also the absolute spectral error is probably minimized
because the reference spectrum (the first of the sequence) is taken
nearly the same time of neon lamp spectrum.

I recommand the procedure for long duration sequence on moderate brigthness
object with sufficient constrasted and numerous stellar spectral line (for an efficient
registration) i.e. F,G,K, ... type stars. Ideal for AZ Cas campaign.

More about ISIS processing of Lhires spectra:

http://www.astrosurf.com/buil/isis/guid ... to1_en.htm

Christian B

[Robin Leadbeater]

Hello Christian,

What algorithm are you using to register the spectra?
In general I think it would be useful to have more documentation with all programs to explain exactly what algorithms are being used in each function. (I do not like using programs without knowing exactly what they are doing)

Cheers
Robin

[Christian Buil]

Robin, I use a nearly standard cross-correlation method for measure spectral shift between spectra (Crsoss Corelation Function - CCF). See another application:
http://www.astrosurf.com/buil/tauboo/exoplanet.htm

The detailled description of all the algorithm is a considerable job. The format is equivalent to a book.... for the future...

Chrisitan

[Christian Buil]

Robin says:
===================================================================


Hi Keith

I have not used the new ISIS tool yet but have resorted to doing this sort of thing manually where I have had significant shifts during runs and have not had sufficient neons to calibrate out the shifts. (My preference is to take frequent neons, preferably after each 20min exposure. With my remote controlled setup, taking neons only takes typically 30sec so little observing time is lost even if done for every exposure. By averaging before and after neons for each exposure, the need to re-register is avoided and the absolute calibration is even better.)

I have though had experience of using cross correlation techniques in areas other than spectroscopy. The effectiveness of cross correlation algorithms depends on the prominence of features in the signal so personally I would check how it performs on a case by case basis, at least initially (Many strong narrow features will work best and it would probably be advantageous to bandpass filter the spectrum first before applying cross correlation to avoid the algorithm attempting to lock onto gross broad shapes in the profile or pixel to pixel noise. It is not clear if this is being done here.)

The calibration is still based on the neon lines taken at the start of the series. These should be closest to correct but I personally would always do a final check (and tweak if needed) of the spectrum based on tellurics if they are available and clear enough as this will always give the most accurate absolute calibration.

Note that if there are significant shifts during a set of exposures then separately calibrating them (either using before and after exposure neons or by re-aligning them using the cross correlation algorithm) will ensure that the potential resolution loss due to bluring caused by the drift is minimised compared with say combining all the exposures and using a start and end of run neon average for calibration.

Cheers
Robin
=====================================================

I full agree with Robin's comments.

Keith note two points:

- first, ISIS register option is not a standard procedure for me. The normal utilisation is for very long time sequence and for spectra spectra showing numerous contrasted lines. From my proper experience, the AZ Cas observation is a rare exemple of register automatic procedure usage.
- second, with a Lhires III the systematic spectral calibration quality check by using telluric lines is well recommanted (is telluric lines can be detected of course). From my proper experience, the control is systematic for moderate magnitude Be stars program for example (if the star spectrum is at the detection limit, telluric lines are not visible, but a possible small spectral calibration error is acceptable for this type of documents).

Christian B

[Andrew Smith]

I wonder if a modelling approach to correcting mechanical (and possible thermal) shits in wave length might be possible. What I have in mind is a similar approach to that pioneered by Patrick Wallace for telescope pointing and implemented on both professional and amateur telescopes. With my Paramount ME it reduces pointing errors to a PSD of about 14 arcsecs and a sky RMS of about 10 arcsecs reducing the errors by at least 100 times.
I found that with my LISA and a heavy FLI ML3800 camera I got repeatable shifts in the position of the internal Neon lines of a few pixels when flipping the telescope from horizontal to horizontal through 180 degrees. Fitting a model to the shift v altitude and azimuth should be possible especially if a physical model were developed rather than fitting arbitrary functions. However, “Syncing” into the model might still be required each night.
Any views on if this would be worth trying?

Andrew

[Robin Leadbeater]

Hi Andrew,

I understand this is done on professional telescope mounted spectrographs. It would certainly be interesting to try. I did a bit of very preliminary work on this for my setup a few years ago but I did it during the day and ISTR was plagued by the effects of temperature as the sun fell on the spectrograph at different angles!
As a simple intermediate step, perhaps predicting shifts based on neons at the start and end of a run and linearly interpolated based on the exposure times might help, though I am not certain that the movement is well behaved and repeatable in my case at least. I sometimes see little shift between exposures followed by larger shifts as though there is some stiction and hysteresis. Although I have not rigourously confirmed it, I have a sense that the size of the effect seems to be different for the two grating holders I have which gives some clue as to where at least some of the movement is coming from.

Cheers
Robin

[Keith Graham]

Hi Robin and Christian,

I began this discussion before ISIS V5.0 was announced. When working with Be stars, I always take neon spectra before and after a sequence run, combine the 2 neon spectra in Maxim DL, and use the center point of the 6506A line of that combined neon spectrum image in ISIS for calibration. I then tweak the calibration using telluric lines. My problem is that if there is a shift in the spectrum, the 6506 line in the combined neon image is broader than in each of the individual neon spectra. So my concern is that I believe ISIS determines resolution based on the neon spectrum. So I am wondering if this is the true resolution of the Be star spectrum. I have experimented by using either the first or last neon spectrum (which ever was the sharper) for calibration. This resulted in an increase in the resolution shown in the ISIS log. I did not see a problem with using only one neon spectrum as long as I did a final tweak with telluric lines.

I am currently attempting spectra in the AZ Cas campaign. The problem here is that there are no telluric lines for tweaking. We are totally dependant on the neon spectra for calibration. So we are left with calibrating using the two combined neon spectra that contains those “bloated” lines with the possibility that resolution will “suffer”

Question #1
Is there any weakness or problem with calibrating using only one neon image as long as I do the final tweak with the tellurics?

Question#2
If ISIS uses the neon image to determine resolution, would not the resolution based on that image be different from the actual resolution of the stellar spectrum?

About the same time I was questioning these calibration/resolution issues, Christian announced ISIS V 5.0, which contains the new wavelength registration algorithm. In reading the new manual and playing with this new feature, I somehow got the idea that I needed to register the first and last “pro” file to get an accurate calibration. To me, this was a clever idea because I could actually see for myself when proper registration was achieved – just as I could see when accurate registration was achieved using telluric lines. But I soon realized that this did not include all of the intermediate images, and this prompted my latest questions to Chrstian.

If I understand Christian’ explanation properly, the Wavelength Registration feature is applicable for only very long runs where there are no telluric lines. Robin, I still have all of the emails we exchanged almost 2 years ago when you were so patient with me and extremely helpful in getting me jump started into spectroscopy. A point you make here is one that you made two years ago and has stuck with me. You always want to know the process software is taking to achieve what it is trying to do. The beauty of using telluric lines is we can actually see what is happening to get accurate registration. When using the new wavelength registration feature in ISIS, we cannot “see” what the software is doing to calibrate the spectrum, so we rely on the expertise of the author that the result will be accurate. This is why I wanted to be certain I was using this feature correctly before submitting any AZ Cas spectra to ARAS.

Christian, thank you for your clarification on using the wavelength registration feature only for long sequences with many lines as with AZ Cas. It does make perfect sense to use tellurics when possible. But I am still a little unclear as to just how this feature should be used, so could you please tell me if I have it correct?

Question#3
For my AZ Cas spectra, if I enter the sequence and first neon spectrum into ISIS as usual and check the Wavelength Registration box, the final resulting profile will be fully calibrated and acceptable for submission. Is this correct? If not, what am I missing?

Thank you and

Cheers,

Keith

[Christian Buil]

A quick answer for your latest question Keith:

Question#3
For my AZ Cas spectra, if I enter the sequence and first neon spectrum into ISIS as usual and check the Wavelength Registration box, the final resulting profile will be fully calibrated and acceptable for submission. Is this correct? If not, what am I missing?

Yes, is you check Wavelength Registration box, ISIS automatically register individual spectra and the final profile is OK for publishing (remember, ISIS produce first the intermediate files @proxxx, then, near the final stage of processing, add the @proxxx files for generate the final FITS profile - a good idea is to examine some @proxxx file for verify if registration processing is ok along the sequence - use for example comparison tool).

Christian

[Martin Dubs]

Hello Keith,

before modeling the flexure of the LHIRES I would suggest you try to reduce it. In the beginning of spectroscopy, with long focal length prism spectrographs flexing was minimized by using counterweights (usually one on each side) with a lever system which acted against the weight of the spectrograph. A similar system could be used to balance the camera weight on the LHIRES. I think you can find some examples in Hearnshaw's book: Astronomical Spectrographs and their History. Your Paramount ME should carry the additional weight. Of course some thought has to be given to the off-axis position of the camera, which makes the task more interesting.

Regards, Martin

[Terry Bohlsen]

Is this mostly a problem with the LHIRES?
With my LISA and the ST8 camera I can slew all over the sky and the neon lines don't move. I have tried taking many sets of exposures throughout the night and they are all identical. This makes it much easier.
Cheers

Terry