From: Milton Aupperle <milton@outcastsoft.com>

Date: April 21, 2011 11:19:20 AM MDT

To: Astro_IIDC@yahoogroups.com

Subject: Re: [Astro_IIDC] Spectroscopy

Hi  Ralph, Darryl and silent others:

Thanks for the feedback.

Last night (while watching the hockey games) I had a look at the work flow that other people have posted on line  that shows what needs to be done to generate spectra.

First off, if your going to do Spectrum measuring, you need to have done flat frame and dark frame correction to your images or movies. To improve Signal To Noise, it's also a good idea to stack multiple frames too, which Astro IIDC can do for you. A major PITA is going to be that you likely will have to rotate the camera and slit / prism / grating so that the star spectra your measuring is not overlapped by other stars and that means you need to shoot new Flat frames after you do the rotation. So this rotation thing will be a major PITA unless your going to acquire a FlatMan or other light screen devices to do flat frames on the fly.

The initial profile generation (which Equinox Image already does) isn't terribly hard too do. You have to basically select two line points, one begin the zero order image of the star and then extending to the end of the first order spectra (which also indicates the spectra direction too low to high frequency). Then we generate a rotated box that you adjust to select the height of of the spectra to measure, including some empty background sky area.

After that, we need to extract and subtract the sky background spectra, which isn't terribly involved either. It will require the user to select a good area of background adjacent the spectra they are measuring, with no stars.

Determining the horizontal pixel to light frequency  conversion factor requires the user to identify two absorption or emission lines of known wave length (or one emission lines of known wave length and the zero order star image it was generated from)  and identifying them in the profile. This parameter changes with focal length, so with any scope where you have a movable mirror, it needs to be re-done if the mirror / focal length is moved.

However after that it gets "messy" and really involved.

Basically we need to calibrate the "Imaging System" (CCD, optics, grating / prism etc.) so that we can adjust the color spectra it produces. That involves imaging a star that has an existing Library Spectra curve for it. Then we smooth both curves out to remove any spikes / dips and generate a multiplier curve. The resulting curve will adjust our "Imaging System" image to the Standard curve, so they create the same spectral curves. This likely doesn't need to be done more than once, except if your imaging objects below about 50 degrees when atmospheric extinction raises it's ugly head and we get 5% or more differences in R G B color strength. If you change your optics, then this will affect the curve too, so you don't change anything once calibrated (that includes, CCD, gains, binning etc. etc.). Change any camera / optical parameter and then your back to generating new curves.

To implement the above, I need to get some of these "Library Spectra curve" for various bright stars, set them up in a format that I want to use and  embed them in Astro IIDC. After that, you have to create calibrated images of a matching star and then calibrate your system before you can generate calibrated stars. I will need to build all the code that handles this, including saving and loading calibration curves from previous tests. I'm also sure there will be some hidden "gotcha" things too that only long periods of testing will show up too.

So as you can see, this is not a trivial exercise at all. On the users part, there are a lot of things that need to be done before they get tot he point of analyzing the spectra and even more work afterwards. On my end, I have a tonne of work and testing that needs to be done to make this work as easily as possible, as well as document it.

The first three steps (i.e. raw spectra profile, sky spectra profile and converting horizontal pixels to wavelengths of light) are not too bad, but the rest gets really really involved in a hurry. So  I'm not sure how far I want to go with this, especially since only a hand full of people seem interested in it. If I just did the first three steps, the resulting data in a text file could be plotted or processed in a spreadsheet  for the rest of it.

Also, I'm not planning on exporting graph images at all and you'd generate your graphs in another package like "Plot" or Excel or Numbers or AppleWorks etc.

Thoughts anyone?

Milton Aupperle