For an instant start-up in CCD spectrography, you will need an irisation filter for special photographic effects such as Cokin's #40 filter. Its cost of 300 F (50$) is the only one to incur in order to obtain the first spectrum using your CCD camera.
You will have to install the filter 2 or 3 centimeters in front of the CCD's surface. The filter holder's drawer will probably make this easy. If needed, the Cokin filter holder can easily be modified, since it is made of plastic.
Aim at a star field, use an exposure time of a few seconds, and you will obtain an image such as the one in figure 10.
All stars in the field will contribute a spectrum, sometimes too weak to be visible in this reproduction. Note the spectra are somewhat narrow, a sure sign that they are properly focused, but at the same time star images at order 0 are strongly out of focus, which betrays field curvature implied by this type of construction. Looking carefully at the spectra you will see a few spectral lines, and in particular a "red" line at 7609A from oxygen in the atmosphere.
Figure 11 shows how to extract a field star's spectral profile.
An important step in spectral lines processing involves computing the plate ratio P and creating a graduation along the horizontal axis. To this end one uses some easily identified spectral lines such as atmospheric O2 at 7609A (in fact more than one line, but close enough to be seen as a single line within the spectrograph's resolution). On the spectral profile, the distance between this line and the order 0 image is measured. Let us assume the value is 178 pixels. Since the pixel size is 9 microns, the distance e in millimeters between the line and order 0 is : 178 x 9.10 –3 = 1,602 mm. Since the grating has 100 grooves/mm one then computes the diffraction angle b for 7609 A, and finds b=4,364°. Then d, distance between the CCD and the grating, is expresssed as
Now on, in order to estimate wavelength l for a spectral line whose position in pixels x is know, the following formula should be applied (dispersion is supposed to be linear here)
Figure 12 shows an example of spectral profile for Vega obtained with a simple Cokin #40 filter.
The spectrum in figure 12 is spectrally limited by the CCD's responsivity. On the left the spectrum starts in the blue from about 3900 A. On the right, it disappears in the infrared at a wavelength of about 1 micron (in the figure the spectrum was voluntarily interrupted at about 8000 A as a consequence of orders superpositions). Between those limits, the quantum efficiency has a complicated evolution and numerous intensity variations in the spectrum take origin in the CCD detector itself. An important part of spectrum processing consists in suppressing this detector contribution and also that of the grating's efficiency which itself depends on the wavelength (this will be seen further on when analysing the spectrum for quasar 3C273).
Though perfectly usable for astronomy, the Cokin filter has a drawback in that it diffracts the light flux identically for orders of opposite signs; this is not in favor of detectivity whenever weak objects are observed. Fortunately there are better and cheaper gratings as is shown in figure 13.