LHires III shifts with long exposures - how to handle?

[Tonis Eenmae]

By practice, we use tellurics only for checking spectra, mostly in case of high SNR + smooth continuum (cooler stars have spikes everywhere, rendering it unuseable). They are too low by SNR for a general use, and dependent on sky quality. However, DIB and tellurics certainly useable at low resolution (e.g. < R~500).
The background sky I usually get, is useable only at certain conditions (lowest resolution e.g. R~500 with full spectrum - bright enoguh pollution with proper lamps, e.g. not LEDs), otherwise they are mostly absent or weak in case of my typical use (R~3000 - 18000, in a narrow band of LHires).

To check wavelength scale using tellurics, correlation works rather well. Definitely upwards from R~5000. I have to admit that I have done very little spectroscopy on cool stars and I haven't investigated if telluric lines are totally useless there.

What I meant by using DIB-s is to shift spectra together (after heliocentric correction of course), not determining wavelength - because they are mostly not exactly symmetric ones, with few exceptions. In the case of low resolution, of course that asymmetry becomes negligible and they can be suitable for wavelength scale BUT their wavelenghts are "fixed" after heliocentric correction (that can be also very small in low res spectra).

I wonder how weak sky spectrum you get in e.g. 10 minute exposure? But when averaging (individually) on both sides of stellar spectrum? In the case of my "little baby" (a 1.5-meter telescope), sky emission is always visible when observing with R~1000 and R~2500 when exposures are ~10 minutes. I sometimes use them around Halpha to construct dispersion curve, most often I do that for R~5000 observations.

Wishing happy new year to all of you!
Tõnis

[Peter Somogyi]

Tonis,

I do around R~3000 and R~17000, with only a 0.3m scope.
Tellurics are either blended (R~3000) or very noisy, and only a few of them (R~17000: only 100A coverage so have very few tellurics - sometimes in middle of emission -, e.g. 6510-6605A in case of H-alpha). Toward H-beta, tellurics just nonexistent. I also can't find any useable tellurics near HeI 6678, despite high-SNR continuum necessary for that project.
Sky emissions I have is only around 5800A, which are rarely present in the typical wavelegth regions I observe.
Most projects I do is lower-SNR (continuum used to be roughly 500-2000 ADU at 330 ADU bias level, that's a few -e.g. 9 - pixel wide, for noise see ATIK 428 EXm specs).
Tellurics around H-alpha definitely used to be in the noise range (something like a 50 ADU signal amplitude), whilst normal calibration lines I have is about 3000-10000 ADU (2-3 pixels of 9 micron width). I admit for Be star projects this could be sometimes useable, but these projects tend to have higher standards for RV accuracy, too (very hard to check if it fulfils, finally the professional will do the drop if doesn't like a spectrum seriewise, with or without feedback).
Since these methods are varying by quality (continuum SNR, observed range), it would be overwhelming to check whether it suits the given requirements (that's many times unspecified till processing spectra and determining average quality of gathered serie by various observers).

Jack,

I've read the BaSS manual and installed the latest beta now, and found facility to stack 1D profiles (Operation -> "Stack and Align profiles"), but unfortunately always getting an error message "All the profiles must have the same width" even when I manually crop them either in ISIS or the BaSS X-Axis (input: N times 1D spetrca processed with ISIS). I stop my effort in this direction for now, maybe I continue later.

Robin,

I agree that ISIS would be the best place for adding this feature, however you know free software always driven by authors personal interests, and since 2012 nothing happened to support this well known problem. In case ISIS would be dedicated for professional use, this feature would either already exist or treated as a bug (Spectroscope model: "LHires III" dedicatedly supported, now it looks it doesn't).

Now I'm going to try SpectroCalc (will take time).

Peter

[James Foster]

To: Peter & maybe Robin,

I frequently do exposures with the LhiresIII with 45min (2700sec) subs; Nova Per 2020 for instance in H-Alpha, H-Beta, and CaK near 3900A. Isis V5.93 has been my default software for processing LhiresIII, LISA-IR, and Alpy600 spectra. I don't notice too much resolution drop when my local seeing is good (i.e. <2.5 arc sec) when shooting H-Alpha with 2400 l/mm grating with LhiresIII; typically R=13.5K. I probably should "tweak" my LhiresIII more to get what I first achieved 3 years ago; R=16.3K. However with sub-nominal seeing, R seems to be proportionate to my local conditions, and when its bad (3-5 arc seconds or worse) my R drops to 10K or below.

"....My main problem left here, is only the amount of processing work of this method with ISIS - too much manpower with many mistake possibilities (fits header). This is also why I don't encourage others to tread this path...."

When I process my Ha for BeSS database V2 submission, after processing in IsV5.93, I use the H2O tab to see how much my telluric lines are shifted, say best shift is -0.03A. I hit cancel and manually (in IsV5.93) shift the spectrum and save. I for into the file retrieve folding, while still in the profile tab, and use WIn10 open with to bring the 1D spectrum up in Bass. I call up the Atmo-Ha wavelength file under Measurements and Elements to check and see it the telluric adsorptions coincide with the 18 Atmo-Ha wavelegth refernces. Once this is confirmed, I call up the profile of the fits file and manually link the object with Simbad's catalog name. I also manually adjust the exposure time since the duraction is keyed to the start and end of the .fits subs taken. In the comments, I usually input the spectra shift I used. After the 1D-spectra is saved, in a subfolder called "ALL Processed Spectra_BeSS Check" (whole directory tree example: "F:\SPECTRA2018\__Home16Dec18(-5)CC13-LISA IR-23micronSLIT\ALL Processed Spectra_BeSS Check"). I go into BeSS to upload for quality check.....these days 95% of my spectra are accepted without comment. The use of the Bass program takes up no more than 10% of my spectra processing time and assures good quality of my spectra's H2O telluric line alignment.

BTW, here is my longest spectrum, 6 X 2700 sec.'s subs (4.5 Hrs) on Nova Per 2020 (16Dec20) in the Cak region with the CDK17+LhiresIII+Atik460ex CCD at -9 cooling temp (R=4675):


James

p.s. I'm going to be paying more attention to the mechanical aspects of spectra process as I shift towards using the eShel.....just ordered my ASI6200mm Cmos Camera!

[Peter Somogyi]

James,
Thank you for sharing your method, good to see you also care with the shifts - at least the in case of high-res H-alpha region!
If I'm right, you don't have any tellurics for the CaII region (not even for H-beta), so you can't check for any shift.
In case with fiber fed eShels, you won't meet this problem (although I saw professionals still taking bracketed spectra with fiber-eShels, but only for an error check).
I'm still continuing to search for the best possible processing method, such that can be used in general.

All:
I hae briefly tried SpectroCalc now (2D wizard), however I greatly miss the 2D preprocess GUI of ISIS (3. Calibration tab) that allows a visual feedback what is going to happen in 2D exactly.
Clearly something automatic happening that I don't understand.
For the first try with my spectra, it couldn't find my spectrum line on a noisy image, and thrown various errors finally aborting with an exception. I stop trying it for now.
A very basic thing I don't understand, why it requests only a dark, and no cosmetic+bias - how can it calculate what to substract then?

Trying SpectroCalc/1D Wizard (for fits header generation purpose), it also complained of different wavelength regions coming from ISIS as an input, similarily to BaSS.

For this particular problem, found a workaround in ISIS: subtracting zero fits (generated from 1st spectrum by subtracting itself - that preserves its header very well) from each spectra as a serie (1 mouse click). The output 1D sequence can be fed now for both SpectroCalc and BaSS.
However, its sum via SpectroCalc doesn't contain any essential header data (e.g. DATE-OBS).

Trying again the already mentioned Bass/Operation->Stack and/or align profiles (now it works!), the generated sum only partially preserves the following fields:

BaSS_stack_header.png (36.48 KiB) Viewed 4166 times

I believe this is not enough.

The next approach I continue with, is to remain within ISIS, and process both the original way (with a dummy calibration - good header) and with per-exposure calibration + sum in 1D, and finally copy the relevant header fields from one to the other.
I will definitely build a script for that copy (either astropy or IRAF, I will see).
I have a feeling, that this could be much easier to be implemented in ISIS.

Peter

[Benjamin Mauclaire]

Hello Peter,

I had also observed this phenomenon in the firsts years I worked with Lhires3 2400mm grating in 2005.
Yes Lhires isn't mechanical stable.
It can be improved by adding screws for camera-spectrograph and spectrograph-telescope interfaces and increase the tightening of all the screws of the spectrograph. Despite this, long run can still sometimes produce an horizontal shifting (along wavelength axis).

This is why, I added among others an horizontal registration early in the Spcaudace's pipelines, assuming taking a calibration lamp spectrum at the begining and at the end of the serie numbered as you do with target spectra:
calibration_nova_per_2020-1.fit and calibration_nova_per_2020-2.fit
They aren't stacked as they are used for measuring horizontal shifting.
As Spcaudace works with fits files during the whole processing, all files header are conserved.
Consequently, all functions deal with fits files such as stack, geometrics corrections or calibration.
As an example, you can stack a series with many ways:
- Do a simple addition of files and manage fits header: spc_spsomme nova_per_2020- addi
- Synthax is:
spc_spsomme generic_1d_spectra_name sum_method(addi/moy/sigmakappa/med)
moy=mean, med=median

Nethertheless, to get all the corrections done, you should use pipelines with the files given from night observation (target raw spectra, darks, flats, calibration spectra). No special settings is needed (it manages all spectral wavelength domains obtained with 2400 or 600 mm grating for example), just fill the pipeline form. See the following tutorials and video (do a google translate if needed):

- Computing the instrumental response [IR] (pipeline 1):
http://spcaudace.free.fr/docs/index.php ... line1.html
- Process night spectra (pipeline 2a=stellar targets, pipeline 2b=nebular stargets):
http://spcaudace.free.fr/docs/index.php ... ine2a.html

Standard processing steps are done with automated detections (angles, 2.5 sigma binning, etc ; although it can be adjusted):
1. preprocessing: making master dark and flat, dark and flat corrections
2. geometric corrections: vertical and horizontal registrations, smile on X (tilt) and Y (slant) axis corrections
3. staking or not (time series like used for pulsating phenomenon)
4. optimized binning
5. wavelength calibration: 2 lines lead with 1 order polynom and 4 or more lines woth 3rd order polynom
6. dividing by IR spectrum obtained in pipeline 1
7. excluding spectrum edges if needed
8. continuum rescaling or normalization (optionnal)
9. Bess format export (optionnal)
10. PNG format export (optionnal)

All these processing steps are done in one go without any intereaction needed.
My 15 years experience in spectroscopy helped a lot and leads to a scientific quality which allowed publications in peer-reviewed journals.
But quality results can be checked at the end of the pipeline in a synthetical display in Audela's Console. All these information is stored in fits header keywords, thus processing history (operations: binning width, binning height, geometrics corrections parameters, etc.) and quality checking (RMS on calibration lines, RMS on telluric lines, resolution power measured, etc.). See tutorials for exemple.
In fact Spcaudace has been designed to achieve all these boring but tricky tasks for a scientific exploitation afterwards that requires time to be freed up.

International hot line (not too hot, it's winter here): +33 701 702 703

Benji

[Peter Somogyi]

Hello Benji,

Many thanks for the description, it is good to know you address the same problem!
I have started learning spcAudace now.

Peter

[Benjamin Mauclaire]

Hello Peter,

To get started quickly, you should do a training with the archive available in each pipeline tutorial and follow the steps described in it.
Then you process your own spectra.

Note that individual long exposure, ie >1200s, can't be corrected from mechanical flexures in it. The consequence is the broadening of lines and RP will decrease. You can see it in telluric lines which become broader. Only shifting between spectra can be corrected.
I noticed that this phenomenon depends on the sky zone pointed by the telescope (I have an SCT, so no internal flexures, only the spectro).

Benji

[Peter Somogyi]

Benji,

If I'm reading the calibration method correctly, there are only 2 calibration frames requested by Pipeline 2a for 1 target (via google translated link):

" 4. Pipeline Calculation Options
...
Calibration with several lamp spectra:
To even better manage the mechanical deformation of the spectrograph, it is advisable to take a calibration lamp spectrum at the start and at the end of the series. Here, the spectra ne_zet_tau-1.fit and ne_zet_tau-2.fit were recorded. In this case, leave the option "o" (yes) checked by default."

However, I am taking calibrations not only at begin & end of the serie (for 1 target), but also between the serie elements (that is clearly the better way).
Am I looking at the right pipeline?

Thank you,
Peter

[Benjamin Mauclaire]

Peter,

Here you have to use only the first and last calibration spectra associated to a serie named such as starname_calib-1.fit and starname_calib-2.fit.
Remove other intermediate calibration spectra from the working directory.
With that, Spcaudace will compute a linear law for applying the right horizontal shift to each studied target spectra of the serie, assument the first one won't be shifted.

Cheers,

Benji

[Peter Somogyi]

Benji,

That method is still not the best for precision, because there are times when the shifts are nonlinear - which Robin noted here in this thread:

http://www.spectro-aras.com/forum/viewt ... 697#p15228

e.g. Gam Cas, HeI 6678, R~17000: I typically do these exposures (6 x 10 minute):

cal-1 (starting cal.)
GamCas-1
GamCas-1-cal
GamCas-2
GamCas-2-cal
...
GamCas-6
GamCas-6-cal (end cal.)

Within 1 hour, there can be sudden jumps (e.g. sudden temperature cooldown is a typical effect in my area), and no useable tellurics exist for any error check in this region. Moreover, flat errors (which is a 2nd problem of LHires - can't be fully corrected) with smooth continuum causing this type of target unusable for cross correlating individual exposures.

Hence both my existing method (2.c.ii) and SpectroCalc (2.c.iii) are still addressing this problem definitely better.

Cheers,
Peter