Non linear calibration a bit hit & miss!

[Gasman]

Hi chaps

Like the title says my non linear calibration accuracy seems a bit hit and miss sometimes!.

I`ve been waving a Neon nite light over the end of the scope and then maybe a stack of 10x 1 second exposures to get something like this raw image -

I then do a slight rotation of 0.6 degs and after a smile correction and vertical crop to give 60 pixel high spectrum which I process then identify about 6 lines to calibrate it using Christian`s Neon map off his site to get something like this -

I then use this reference spectrum to calibrate a spectrum from say beta CrB  which is a G0V type and overlay a GOV reference spectrum but some thing still not quite right. No cropping done so all spectra the same size. Maybe I should try another light source for the calibration?. I tried the fluorescent starters which had a lot wider spectrum than the Neon but quite dim maybe a combination of the two needed?

cheers

Steve

[Merlin66]

Yeah,

The neon works best to calibrate the red wavelengths...

What lines did you select? Obviously the spread should be as wide as possible...did you end up with a quadratic calibration or just a best fit linear?

I've been using the Argon Filly lamp for calibration on my Hb to Na spectra and the neon for only the Ha.

The Relco starter lamp will give you better success in the shorter wavelengths....

The other way is to cheat!

If you know the features on your spectrum - CaK, Hb, Na, Ha, Telluric etc then use them to generate a non-linear calibration....this should remain reasonable constant unless you change your set-up.

Edit:

It's surprising the neon doesn't result in at least a good calibration around Ha.....

It could be you didn't make your tilt/smile corrections properly.

Firstly, the tilt should be corrected using the Y axis reading of the center of the target spectrum on your CCD (say Y=300).

Then crop both the target/ reference to a strip say Y1= 280, Y2= 320 - 40 pixel high strip centred on the target.

Correct this strip for tilt and apply the same correction to the target. If you correct for smile/ slant then also make the same correction to the target. For completeness in BASS I select the binning area (height)position the same for both.....

This should at least Show the Ha at 656.3nm - If not then there's something else wrong.

[Merlin66]

The attached jpg shows the dramatic differences you can get across the height of a reference lamp (with a 6mm slit height).

The slant if not corrected for the target position can give significant error.

[NickK]

Have you switched the gain/amp off to get maximum linearity? Or corrected for it?

[Merlin66]

Nick,

the non-linearity is caused by the varying dispersion (prism element) rather than the camera.....

[robin_astro]

Hi Guys,

A few of pointers here.

It is vital to treat the lamp and star spectra identically and keep them in sync (ie same geometric corrections, slant, tilt, smile) and then measure the lamp lines at the same Y position as the star spectrum otherwise you will get offsets.   (Vspec is excellent here as it stores star and lamp binned spectra in the same file so transfering lamp calibration to star happens automatically  

Make sure you are using unblended lines, many are not at ALPY resolution

To get a good non linear calibration fit it is important that your lamp lines span the measured region well. Extrapolating beyond the measured lines can give big errors particularly if you use a high order fit (I never go beyond a 3rd order which is fine for every spectrograph I have and often 2nd order will do)  This means unless you delve into very faint neon lines at t he blue end, a neon lamp alone is not enough. The Ne/Ar starter is one approach but again watch out for blended lines. Easier though is to use the same trick as with the Star Analyser and use Balmer lines in the reference star (which you always take anyway, right?) or even a combination of the two/three.  Christian Buil covers all the options here

http://www.astrosurf.com/buil/isis/guide_alpy/resume_calibration.htm

Cheers

Robin

[robin_astro]

The Ne/Ar starter is one approach but again watch out for blended lines. Easier though is to use the same trick as with the Star Analyser and use Balmer lines in the reference star (which you always take anyway, right?) or even a combination of the two/three.  Christian Buil covers all the options here

http://www.astrosurf.com/buil/isis/guide_alpy/resume_calibration.htm

Fortunately with the high stability of the ALPY, unless you are looking for super accuracy, this is pretty much a one off operation unless you dismantle it. subsequent spectra can be calibrated using the same equation, perhaps with an offset check using one line (Balmer in the reference star or lamp) - the equivalent of one point calibration using the zero order with the Star Analyser. dont try using this technique with less stable spectrographs with flexure though. There the lamp and star spectra need to be taken with the telescope pointing in the same direction.

Cheers

Robin

[jnp]

Hi Steve,

The Neon waving has to be even (e.g. circular motion) or the spectra may get slightly shifted to one side.

Double check you have identified the correct reference lines during the calibration. It would be useful to see a shot of the data in the data grid in the calibration Lines screen, The error column normally shows if there is an issue with a calibration point.

As Ken suggests, you need a non-linear 'calibration fit' due to the prism within the Alpy

As Robin + Ken suggest, the same geometric corrections must be applied.  Load the reference + target spectra in a new project. Tick the "Apply to all images" checkbox when carrying out smile, tilt corrections on the reference spectra. The "Apply to all images" checkbox applies the same changes to the reference + target images.

If you must crop, the Bulk Image processing screen is useful to ensure the reference and target spectra have the same cropping.

Hope this helps

John

[Gasman]

Hi all

Apologies for delay replying, just helping to increase UK tax revenues

You all mention a few very helpful points which has made me stop and think!.

Ken - the target spectra, I tend to just use the raw fits after a bit of stretch is applied to give an image like this -

Admittedly this isn`t the best shot and no smile correction was done as the spectrum is only a few pixels high. In my early days with using a Star Analyser I used to crop the spectrum to 1 or 2 pixels and then stretch it to about 30 pixels ,maybe I should do that again now with the Alpy slit type?. So the lack of treating the target spectrum the same as the reference is obviously a contributing factor to the dodgy accuracy. I also alternate between Rspec and BASS which might also not help?.

This is a shot of the calibrated Neon reference data at 3rd order with lines I`m pretty sure correspond with Christian`s, I even compare/count the relevant peaks beforehand!

Robin - I think maybe I was placing too much emphasis on just using the Neon spectrum which just occupies half of the image. You never seem to stop learning at this game do you ?(or maybe its just me ).

John - Never gave the waving a thought aaaaagh! I just held it over the end!

cheers again

Steve

[Merlin66]

Steve, Don't stretch the spectral image out to 30 pixel etc.!!!!

All the data is in that narrow band.

Just watch the selection of the "binning zone" to catch the extents of the spectrum and use the "remove sky background" above and below.

[Gasman]

Hi Ken

I thought that stretching the image might show the smile more as its a bit hard to see at that size?.  I do mean `stretch` as in physically making it bigger and not `stretch` as in processing  to bring out the detail!

Steve

[Merlin66]

"stretch" as in making it physically bigger....

Try an experiment....make a profile from the original "narrow" spectral image....then do a stretch to make it say 30 pixel high and then make a second profile - compare the two....

I'd be very surprised if the larger stretched version shows any more detail - in fact it may just create more noise!

[Gasman]

Hi Ken

My confusion is that any smile is easily seen in the raw Neon spectrum as its as tall as the slit will allow but any smile on the target spectrum is hardly noticeable or do we just assume that because there is smile on the Neon then there must be the same smile on subsequent spectra taken with the same setup and thus require the same correction, does that sound clear or jibberish?   

Steve

[Merlin66]

I generally ignore smile correction to the narrow spectrum. (The answer to the question is yes, any distortions showing in the reference will also be in the spectrum)

With a hi-res littrow, slant correction, for me, is more important.

Just watch the Y co-ordinate for the reference/ image........

[robin_astro]

Hi All,

It is important not to ignore the slant (or smile) in the narrow star spectrum just because you cannot see it. Correcting these in the lamp or star spectra  will potentially cause the lines to move and so you should make the same adjustment to both lamp and star.  You  can measure the slant (and smile) in the  lamp spectrum lines and the tilt in the star spectrum and then apply all corrections to both spectra.   Correcting the slant and smile in the star spectrum is also important for background subtraction.  The features in the sky spectrum will not be subtracted correctly if there is any remaining geometric distortion. 

It is possible to use a static lamp over the telescope aperture provided you mount it in line with the slit. If you shine the light obliquely at the slit this is when you can get small shifts.

Cheers

Robin

[Merlin66]

Agreed!