28 January 2017 at 3:38 pm #573680Kevin GurneyParticipant
Here is another recent observation. I had selected it based on it being generically ‘B’ but was pleasantly surprised to see it was in fact Be… [Its still early enough in my spectroscopic career for me to be excited by the simple fact you can see all this stuff from your own back yard! 🙂 ]
So as I guess we expect, there is strong Ha emission, but the Hb region (panel at bottom left) looks interesting… It appears to show superposition of an absorption and emission line; presumably from the star itself and its surrounding material?
Note that the panels at the bottom are taken from the rectified spectrum (shown in the middle). Also note there was no observation for a reference star here – incoming clouds prevented. I therefore just used a spectrally close reference star to get at least something ball-park (?).
Incidentally, if you are interested only in specific features of a spectrum, what is the wisdom about using such rectified spectra as their basis?
Attachments:29 January 2017 at 10:10 am #577859
I’ve been doing spectroscopy for a couple of years now and I still get excited with each new spectrum. You never know what you will see and this is very true of Be stars. The hydrogen lines can be a complex mix of absorption and emission from the star and disc configurations which change over time.
I found 48 Per in the BeSS database. I’ve put a link below, though it does not look as though the link preserves all of the criteria I entered. You can select the H-beta region by entering 4861 in the wavelength. This shows pervious spectra with various mixes of absorption and emission of this line.
I’m not sure of the effect of rectification on the science value of a spectrum, so I avoid it. The BeSS standard uses the words ‘clearly advised against’ where it calls it ‘continuum normalisation’. For personal interest it is of course fine, and others with more knowledge may be able to advise on its effect. The bit where I am particularly unsure is its effect on line strength measurements, like equivalent width. This may not be big since the effects to the continuum on either side will be scaled by a similar amount, but I’ve never done an experimentation to discover the effect.
Similarly, for personal interest it is fine to use alternative methods to correct the response, but as soon as you want to do any science or submit to a database then a proper response correction is needed.
Andy29 January 2017 at 11:07 am #577860
As I understand it (and I may well be wrong) in stellar classification the spectra are rectified and this is built into the definition of equivalent width (see The Classification of Stars Jaschek and Jaschek).
The issue of what processing spectra should have depends I think on what scientific purpose the spectra is to be put and to a degree what type of instrument it was captured. BeSS had a particular goal and hence it’s rules.
As an instrumental example short exposures (<10min) I can see no statistical difference between a bias frame and a dark from for my SXVR H694 at -10deg so why use a dark frame and increase the noise why not use just a (constant) bias correction? Similarly for a fiber fed spectrograph where the spectra is always on the same pixels why flat field if you use a standard star for instrumental correction?
Just some random thoughts.
Regards Andrew29 January 2017 at 2:19 pm #577865
You raise some important and interesting points, possibly worth their own forum threads.
I agree the processing of a spectrum should depend on the scientific use to which it is intended to be used. When submitting to databases though, it is my opinion that it is best not to rectify the spectrum unless necessary. The reason for this is that a researcher can always rectify a spectrum should they wish to do so, though they cannot unrectify it. Note that when a rectified spectrum is submitted to the BAA Spectroscopy Database then the BSS_NORM fits header should be populated.
I have no strong opinion on whether or not to rectify a spectrum when no response correction has been applied.
On the dark vs bias front, I would expect to see differences between dark and bias frames with amateur CCDs. I also use an SXVR-H694 and I see a big difference between my bias and dark frames.
For the following stats I had the CCD set to -20C with 600 second darks, with statistics computed by MaxIm DL from single raw bias and dark frames.
Mininum 882 1006
Maximum 1315 56522 (due to hot pixels not cosmic ray hits)
Average 1084.580 1133.068
Standard Deviation 15.960 211.396
I also advise to flat field as know of a couple of reasons why the response correction from a reference star would not correct for all variations in the CCD.
First, when using a reference star to create an instrumental and atmospheric response curve, then there is a degree of smoothing to remove small scale variations in the curve. This is to give the bulk wavelength dependency. So by its very nature it does not remove the individual pixel to pixel variations.
Second, the target star and reference star on the CCD need to be exactly aligned so that the light path through the spectrograph illuminates the CCD pixels in exactly the same places. This is nearly impossible with a slit spectrograph, but I am not sure if a fibre fed system would get around this.
I am happy to be proven wrong if my understanding is at fault. Also please excuse this lengthy post but I thought it worthwhile for these important topics.
Andy29 January 2017 at 4:28 pm #577866
Andy, I am not trying to make a big issue of this. I agree any spectra going into a general database needs to meet it’s standards and the BeSS standards is fine.
When I did my dark v bias comparison I used a region without hot pixels as they would be rejected in a standard statistical test when calculation the mean and variance of a sample.
The examples I gave were just to point out different style of instruments have different characteristics and the need to think about what processing is relevant. (Another example is to flat field a slitless spectra or not.)
Regards Andrew29 January 2017 at 6:26 pm #577868
I don’t know 48 Per specifically but I think what you see is typical in some Be stars. The Be star spectrum is a combination of the normal main sequence B star (dominated by broad winged Balmer lines with He also) plus a Balmer line emission component from the disc, H alpha being the strongest but also sometimes in H beta if the emission is strong. If the ssytem is seen pole on as would seem to be the case here, then there is no doppler component from the rotation of the disc so the emission is a single narrow line which is narrower than the photospheric absorption line (because of the lower gas density in the disc). When the two are superimposed, the H alpha emission can dominate, hiding the whole absorption line including the wings, but where the emission is less as here in H beta we still see the photospheric absorption line with a small emission in the central core. (The same effect is also sometimes seen at H alpha in cases where the emission is particularly weak.)
Robin29 January 2017 at 6:51 pm #577869
Concerning rectification, I think it depends to what use you are going to put the spectrum. For your own personal use or in a specific study, you just need to do what is required for that application (Which as an extreme example could be all about as accurate as possible wavelength calibration and nothing needed concerning flux calibration if all you wanted to know was the RV) .
The problem comes where you might be uploading data to a general database where you do not know to what use the data is to be put. The rule I apply (which is the philosophy behind the BeSS database) is that you should do all reduction that the user cannot do (eg wavelength calibration, flux calibration by correcting for instrument response) but not steps that the user could do for themselves (eg heliocentric correction, rectification, telluric removal). making sure that the appropriate flags are set in the header.
In this specific case however, If I was uploading this to a database, I would rectify it. To upload it as is including the continuum would be misleading. The reason is that the continuum shape is completely artificial having been made to fit that of a typical star of similar spectral type. For all we know this might bear no resemblance to the actual spectrum. By rectifying it, we are not losing any data, since the continuum shape is not derived from the spectrum and the user should they wish, could add back the continuum shape using their own typical B star spectrum.
Concerning measurements on rectified spectra, EW measurement of absorption lines is no problem as they are measured relative to the continuum. Measurement of emission line strengths is problematic though as the actual flux in the line is usually the important parameter rather than the EW (as in general emission and continuum flux are not related) and measuring the relative strengths of these at different wavelengths (even in relative flux) requires a non rectified accurately flux calibrated spectrum.
Robin29 January 2017 at 7:06 pm #577870
Re darks:- With my extremely faint supernova spectra, I no longer do a traditional dark correction though i still take them to produce a defect map and use this with the bias frame. I have not taken that step with other spectra yet though. I use cloudy nights to take darks so I have plenty of opportunity to take very large stacks !
re flats:- I always take and use them with slit spectra as I find it is the only way to accurately remove some fine scale QE variations (QE ripples at low resolution and interference type effects at high resolution). Getting a big enough stack to avoid introducing noise without getting saturation is a problem with a halogen flat source and wide spectral range though. (eg ALPY). Again cloudy nights come to the rescue here.
For slitless work I have come to the conclusion that flat correction is effectively impossible so just take a flat so I can avoid any obvious defects and position the star zero order (target and reference) in the same place in the field.
Robin29 January 2017 at 7:10 pm #577871
Please excuse my long posts 🙂
It is good to have these discussions as I am sure that others have similar questions.
Andy30 January 2017 at 3:46 pm #577874
I would agree with all Robin has said above. I would,however, caution that you need to check what your processing is doing what you expect and not adding any anomalies.
While I have not done this yet I am considering taking the mean of my bias frames and subtracting it, rather than the bias frame, from the other frames (targets, flats, darks). The logic is that for CCD (but not CMOS) the bias or offset is not a function of the pixels but of the output electronics. It should, at least in the short term be constant, which is estimated by reading out an unexposed frame. Some amateur cameras and professional ones have an over-scan region for this purpose as it avoids any residual dark current light leaks etc.
If this is a valid approach, and I think it is, it avoids adding in noise during the bias removal. Thoughts welcome.
Regards Andrew30 January 2017 at 8:14 pm #577877Kevin GurneyParticipant
Thanks for all your comments. I’ve learned a lot..
I am now reworking some spectra in ISIS and am using PlotSpectra (by Tim Lester) for post-processing, measurement and plotting.
I think Robin drew attention to this as I recall?
Its well made and easy to use – albeit with a slightly opaque (at least for me!) measurement dialogue box – but I am persisting..
Kevin31 January 2017 at 12:33 pm #577881Tony RoddaParticipant
Some great discussion on frame calibration.
Does a Dark not contain Bias, by definition?
As in, we already use flat-darks in ISIS Masters and indeed have that option for photometry in AIP4WIN?
The latter allows a fixed offset/Bias value.
- You must be logged in to reply to this topic.