Most of the information we get from celestial objects is deduced from their spectra. High-resolution spectroscopy for stellar work typically means a spectral resolving power of 50000 or more. This type of work can be done in three ways, 1) with a high order echelle spectrograph, using a cross disperser to separate orders into a two dimensional format suitable for CCD detectors, or 2) with a moderate order grating spectrograph at a coude focus, preferable used in only one order at a time, or 3) with a Michelson interferometer having a large path length. The first arrangement gets large wavelength coverage by sacrificing accuracy. The second arrangement allows high accuracy but limited wavelength coverage. The interferometer is used mainly in the infrared and only for special projects because it is very slow, equivalent to a single-channel spectrometer scanning the spectrum, but it can achieve high accuracy if carefully done.
My equipment here at Western Ontario is at the coude focus of our 1.2-meter telescope. The resolving power is about 100000. The visible window (3800 to 10000 Angstroms) is covered in the 7th through the 17th orders of a 316 l/mm grating, but only a few tens of Angstroms is recorded in each exposure. The narrow wavelength range reduces scattered light to a minimum thereby avoiding one of the major faults of spectrographs. The spectrograph is the room seen on the right half of the observatory ground floor in the picture.
The entrance slit of the spectrograph is replaced by a Richardson image slicer. This is a marvelous device that brings more light into the spectrograph, especially in our climate where the seeing disk (image size) of the star can be several seconds of arc across. Here is what our 'red' slicer looks like:
Additional information can be found in these references: