Spectroscopy isn't supposed to be about "pretty pictures" in fact the first thing you do to a coloured spectrum is convert it to black and white. It's all about the scientific knowlege rather than the aesthetics. However, I do find the specra very beautiful and the idea of "spreading out" the light from a star into its constituent colours and discovering secrets hidden within it is wonderful.
To make a pretty spectrum takes quite a bit of processing.
There are 6 stages:
1) Produce your raw image.
2) Rotate and crop the First order spectrum:
You can already see 2 of the hydrogen absorption lines in the blue part of the spectrum. These are created when the electrons in the hydrogen absorb specific wavelengths of the spectrum.
3) Use software to produce a profile.
I'm a total convert to RSpec. It's a wonderfully intuitive program which turns a complicated and potentially frustrating process into an absolute pleasure.
As you can see under the profile, we already have a spectrum, but there are 2 problems with it:
Because it's much brighter in the centre, you lose the detail at the sides. Sirius has very strong hydrogen lines but these are hardly visible here. Also there are dark lines caused by the non linear nature of the camera's CCD (see previous post)
In particular, notice the big dip in the centre of the graph which produces a wide dark line in the specrum. This is an artifact of the Nikon D60 sensor, and nothing to do with any real features of Sirius. Also the "noise" in the graph is creating spurious lines.
So the next stage is to produce a spectrum which has been corrected to take into account the variations and limitations of the CCD.
4) As in the previous post, we do this by subtracting the raw profile from a reference profile and use this to calibrate the image. Now we have a spectrum which is a fairly accurate representation of the actual enegy of the various wavelengths in the light from Sirius. But...
... we have a new problem now. The left side is bright and the right side is dark.
5) So we need to normalise the curve using another great feature in RSpec called Spline Smoothing.
Having taken some time to get my head around spline smoothing, I'd love to say more about it, but I came to the conclusion that people would naturally fall into one of 2 camps; those who already know what it was, and those who feel quite strongly that if they've got by in life thus far without knowledge of curve nominalisation using spline smoothing, they are happy to continue without going into it in too much detail now...
Anyway, you get the idea from the blue line. The software replots the graph based on how far the points on the profile are above and below the line.
And this produces a normalised graph, which in turn creates a more uniform spectrum:
Almost there. We now have a "balanced" spectrum and the hydrogen lines are becoming clearer, but the process has also emphasised the noise. To get rid of this we simply...
6) Perform a new spline smoothing operation to iron out the artifacts.
And now the final result. But first, here's my best effort from my early heady days as a spectroscopist (2 weeks ago to be precise)
I made this by simply stretching the original image in Photoshop, before discovering the joys of RSec.
And this is made using the same raw image.
Ta dah!
The 3 main hydrogen absorption lines in the visual spectrum of Sirius are now lovely and clear, and I'm looking forward to producing a set of spectra covering different classes of star.
Recent Comments