The optic elements of the spectrograph must be adapted to the telescope optics. The key parameter here is the equality of focal ratios of telescope and spectrograph. When the F# of the spectrograph is smaller than the F# of the telescope, only a partial area of the collimator (and the grating) is used. In opposite situation, a part of starlight is lost.
On the other hand is to adjust the pixel size of CCD on the resolution of the spectrograph. There is no advantage if the spectrograph resolve 1 angstrom, but records 3 angstroms on 1 pixel and integrates 3 resolution units into a 1 pixel value (undersampling). Or, conversely, the resolution element of 1 angstrom is spread over 5 pixel, that is all about the same pixel value measured (oversampling )? And captures additional noise of the to many participating pixels?
According to the Nyquist Theorem, the spectral resolution element (the most fine structure that is mapped in the spectrum) has to expose 2 pixels. Then the system shows the best statistical behavior.
Now we can distinguish different situations. .
- You own a telescope and would like to buy or build a suitable spectrograph. Then, the focal ratio is just fixed and you can safely calculate with SimSpec how the matching spectrographs optics should look. Respect certain restrictions, such as weight issues, desired resolution, etc. By the way, you can change focal ratios by Shapley- and Barlow- lenses (if you later change the telescope)!
- You build, buy, or possess a spectrograph and would like to buy the best matching telescope. Then the weight and attachment problems are to be solved. As a rule, probably a Schmidt-Cassegrin system (SC) will be a good choice because of a generous back focus range and weight tolerance on the focuser.
- The CCD camera should correspond approximately to the Nyquist theorem – with a littele tolerance to oversampling . Watch out! Too small pixels, and the widespreaded megapixel madness use in spectroscopy little! CCDs with 20 µm pixels are usually much more sensitive and lownoisy, such as those with 5 µm pixels.
If somebody was LHIRES III want to buy, then I recommend to the telescope, for which the LHIRES was designed and optimized: the C11. I had decided for the C14. But that was an (but not worse) error. Because of the nearly 4 m focal length of the C14 the diameter of the star image in focus counts approximately 60 to 80 µm (average seeing in Germany of about 2 to 4 „). So I have to adjust the slit to 40 µm to cut not too much of the star disk/image . Effectively I lose while about 1 / 2 to 2 / 3 of the starlight on the slit, so that brings the bigger aperture of the C14 compared to C11, ultimately nothing. With 40 µm slit width, but I lose resolution compared to the situation at C11, where would be because of the smaller focal length and thus smaller star picture in focus a slit width of 20 to 30 µm to be sufficient. For this reason, I use a Shapley lens in front of slit to reduce the focal length of C14.
For such decisions necessarily ask SimSpec and run through all the possible alternatives. Identify the main consequences for the selected output parameters (degrees of freedom) in SimSpec: minimum size for the optical elements and characteristic parameters such as resolution, dispersion, etc.