17 May 2018 at 11:19 am #574037
I bought an Alpy 600 with guiding and calibration modules about three weeks ago. Raw spectra of astronomical targets look good to me but am climbing the learning curve in a few processing aspects. One of these is wavelength calibration.
Before starting to get to grips with this, I thought the Alpy calibration unit would be a doddle to use, but I soon ran into difficulties correlating my calibration image with the spectrum provided by Shelyak, especially at the blue end. Here the lines are weak and rather broad. I subsequently took spectra of several objects including Denebola with an unambiguous (I think) Balmer series. Using seven lines from this series and a cubic interpolation in BASSProject, I get a maximum residual of -0.175nm and 0.088RMS. (Going to fourth order polynomial didn’t seem to be justified as these quality indicators were only marginally improved and, as I later discovered, fourth order thoroughly screwed up extrapolation at wavelengths longer than Ha).
Using this calibration I tried to identify the lines on the (ArNe) calibration spectrum. This plot shows the results of my efforts:
(The spectrum of Eta Aur is included as a cross-check as it was taken during the same session as the ArNe calibration session. The Denebola spectrum was taken some days earlier and I was concerned about possible shift.)
Even with the help of the Denebola calibration I don’t feel confident about identifying the blue Ar lines and simply couldn’t make out the lines at 451.073nm and 476.487nm at all. Judging by the sharpness of the Balmer lines, I don’t think I have a resolution problem. It occurred to me that I should reduce the active binning region to reduce possible smearing caused by line “smile” but this didn’t have a visible effect on the profile.
The cubic fit does not result in a good fit at long wavelengths – the error rises to 4nm at the end of the plot which I consider unacceptable. I find, not unsurprisingly, that this becomes much worse with a fourth order polynomial fit. What is needed, of course, is a calibration point or points at very long wavelengths and here the Ar spectrum would seem to be ideal in augmenting the Balmer line calibration points which do a good job at the blue end. (I’m sure it wouldn’t hurt to include the Ne 585.429nm line too.) It doesn’t seem possible in BASSProject, to make a calibration from two separate images – have I missed something? Nor does it seem possible to combine the two spectra to produce a composite which shows all the lines. With an “add constant” function, I feel that it would be possible, but this function seems to be missing.
I am in the very early stages of getting to grips with ISIS and can see that this might have the functionality to cope with at least some of the above – should I make a more concerted effort to move to ISIS (or anything else)?
I’d be grateful for comments to the above.
Attachments:17 May 2018 at 12:23 pm #579496Mr Andrew Jonathan WilsonKeymaster
First of all welcome to the spectroscopy. There can be a steep learning curve at the beginning, but you are on the track and it does get easier. Once you get the hang of it then it is very rewarding, adding a whole new level to the hobby.
I wrote a BAA tutorial on using BASS with the Alpy, which can be accessed here:
In section 5 on page 23 I go through calibration, covering off many of the things you have found, like a cubic fit being a good compromise. I think the strength of lines can vary from calibration lamp to calibration lamp. In the tutorial I include plots/images of calibration lamps with the lines identified from 3 different sources, which may help in the identification. I found the following 6 lines worked well, though you may find a different set works best for you.
BASS is very user friendly and easy to get going with, while ISIS is great at processing spectra quickly with less intervention but it has a steeper learning curve and you have to get everything setup just right. There is also DEMETRA from Shelyak, which I have not used but I understand it was created for the Alpy and should also make processing spectra more straight forward.
Andy17 May 2018 at 5:50 pm #579499
Many thanks for the tips. I’ve downloaded your pdf tutorial – it looks to be excellent – just what I need and I look forward to going through it in detail, probably starting with section 5.
John17 May 2018 at 6:27 pm #579500
I dont use BASS but below are the lines used by ISIS from a typical run. (Given the pixel size and the postion of one strong line it finds them automatically and produces a 4th order fit with RMS 0.25A typically.) The fit deteriorates slightly beyond 3900A but ISIS can also include Balmer lines in the violet to improve the fit there
If your lines are blurred at the violet end it could be worthwhile trying a tweak of the focus. The ALPY optics are designed to give excellent performance into the UV but with some compromise at the IR end beyond ~7000A
If you have not come across this already Richard Walker’s list is also a good one for identifying lines to use in the lamp spectrum (not all lines may be present or may differ in intensity though from lamp to lamp, for example H leaks out through the glass over time. Also watch out for blends at lower resolutions)
Wavelength fit deviation
point #1 x = 338.470 lambda = 3946.250 dlambda = -0.150
point #2 x = 423.993 lambda = 4158.408 dlambda = 0.182
point #3 x = 565.202 lambda = 4510.823 dlambda = -0.093
point #4 x = 578.791 lambda = 4544.813 dlambda = 0.237
point #5 x = 623.978 lambda = 4657.878 dlambda = 0.022
point #6 x = 666.777 lambda = 4764.983 dlambda = -0.113
point #7 x = 746.795 lambda = 4965.095 dlambda = -0.015
point #8 x = 921.725 lambda = 5400.811 dlambda = -0.251
point #9 x = 1104.901 lambda = 5852.418 dlambda = 0.072
point #10 x = 1275.212 lambda = 6266.399 dlambda = 0.091
point #11 x = 1375.164 lambda = 6506.326 dlambda = 0.204
point #12 x = 1647.149 lambda = 7147.469 dlambda = -0.429
point #13 x = 1749.146 lambda = 7383.738 dlambda = 0.242
Coefficient a4 : 3.302033E-11
Coefficient a3 : -1.853850E-07
Coefficient a2 : 2.729276E-04
Coefficient a1 : 2.34638
Coefficient a0 : 3127.559
RMS : 0.248411
————————————————————————–17 May 2018 at 6:46 pm #579501
Attached is the wavelength calibrated spectrum of my ALPY lamp used to produce the calibration fit above (1D fits file, you may need to change the extension to fit. for some reason the forum allows .fits file extension but not .fit)
Robin17 May 2018 at 7:11 pm #579502
4200.67A looks to be a blend in my lamp spectrum and H alpha has completely disappeared from my lamp now
Robin25 May 2018 at 9:37 am #579530
Thanks for your comments, tips and wealth of information. And sorry for being a bit tardy with an acknowledgement.
I have been busy developing a utility to blend images of stellar and calibration lamp spectra. Using this I found that I couldn’t get consistent polynomial fits simultaneously to both stellar absorption lines and calibration emission lines, especially if the ‘scope had been moved in between. This pointed to a flexure problem. I was dismayed to find I hadn’t tightened all six of the Alpy thumbscrews properly. I should have checked this much earlier!
Now to look again at focusing for the blue end of the calibration line spectrum.
John25 May 2018 at 12:52 pm #579531
Glad you found the source of your flexure. If screwed down tight I have found the ALPY to be remarkably stable. I could not detect any flexure on mine in tests moving at around the sky.
If you have not come across it before, Christian Buil has some information on the effect of combining lamp and stellar lines which you might find interesting.
Though not necessary for most applications, if you are looking for the highest accuracy using stellar spectra dont forget to take account of the earth’s motion and the radial velocity of the star which can add up to a few tens of km/s. The effects are small but can be an Angstrom or two, detectable even with the ALPY.
Cheers31 May 2018 at 10:41 pm #579562
News of this nova reached me only a day after I had obtained my first stellar spectra with my new Alpy 600. As this object might show interesting developments as time went on, I thought I would obtain spectra on every observable night I obtained spectra on 14 nights (“evenings” more like) from 2018-05-02 to 2018-05-27.
I have been collecting data at a much greater rate than I could process it – as I was climbing up the processing curve as well as dealing with some acquisition issues, guiding especially. Thanks to Andy Wilson for his tutorial on BASS which I have used for this processing and also to Robin for tips and information.
But now I have processed spectra for three of the 14 dates (May-05, -13 and -27) and I am pleased with them. Here are the profiles, with Robin Leadbeater’s for comparison – and 1D fits files are attached too.
I am not sure I have the instrument response correction right, but I am pleased with the sharpness of the lines and to see that most of the lines are triplicated i.e with most of the prominent lines having two “side lobes”. I hope I have identified the He lines correctly – I am suspicious about the Fe line (which BASS reports).
I have read practically nothing about the spectra of novae as yet, but I see these as being blue and red-shifted versions of the central line, which conjures up the image of an accretion disk with material on either side approaching us and receding. If this is the case and if I have my sums correctly, the velocities are some 1,000-2,000 km/s, with the approaching material tending to be slower. Can this be correct? Also, it seems to me that these have developed with time – the side lobes aren’t prominent in Robin’s very early spectrum, nor of mine, 3 days later (2018-05-02). By 2018-05-13 they are well developed (easily visible on the raw images too).
These are hot off the press, and I would welcome feedback on the results – e.g. should I be concerned about the instrument response correction? I will be tackling the other 11 night’s spectra in the coming days. Eventually I will be subjecting the whole lot to a more detailed analysis and, before getting too far advanced, I’d appreciate hearing opinions about the accretion disk (novae are supposed to have them, I read) hypothesis and whether radial velocities really can be as high as I suggest. In fact I’d be grateful for a pointer to a primer on the spectra of novae and any early papers/comments on this one.
- You must be logged in to reply to this topic.