There are two quite different needs for spectra to assist in the analysis and modelling of an eclipsing binary.
The first need is for a classificatory spectrum, to determine the system’s Harvard spectral type and - though harder - its Morgan-Keenan luminosity class. By knowing these you can place the binary in the H-R diagram, and so make assumptions about its temperature, stellar masses, and (to some extent) evolutionary history. This in turn strongly constrains the range of models that can be fitted to your light curve. Sometimes the spectral type is available in the literature, but it’s often far from reliable so an independently obtained spectrum can be valuable.A classificatory spectrum can be obtained with modest equipment. Experience is needed to interpret the result. Here is a spectrum of YY Gru (mag. V=12.75-13.65) obtained by Terry Bohlsen (Armidale, NSW) using a LISA classical spectrograph on a 250mm telescope. The camera is an Atik 314L+ using a 23 um slit, with 40 minutes integration. This gives a resolution of about R=1400. The spectrum is late G or maybe K0.
Radial velocity spectra
The second needis for a radial velocity determination (not recessional, but orbital). For this the difference in doppler displacement of the spectrum at maximum approach in the orbit, and maximum recession, is measured.From photometric analysis the orbital inclination to the line of sight can be determined, and combining that with the radial velocity determination gives the actual orbital velocity. Since the orbital period is known from photometric analysis, the actual size of the orbit follows trivially. Then by Newton’s version of Kepler’s third law, the actual masses of the stars are derived. But if the radial velocity is unknown, orbital size and stellar mass and much else can only be estimated indirectly.The sad story is that to measure these doppler displacements takes the equipment out of the hands of the amateur. To obtain the doppler displacement of a magnitude 7 to 8 star requires about a 1-metre telescope fitted with a very high resolution spectrgraph.If you haven’t got a radial velocity determination how can you make up for it? You have to rely on what you can infer from the spectral classification, as described above. That can be tricky since you’re dealing with two stars together, not one; and moreover interactions between the two can make the brighter one overluminous for its spectral class. But many papers on eclipsing binaries have to rely on such estimates, and any modelling carried out is explicitly based on those assumptions. That’s not merely better than nothing; it’s well worth doing as giving one informed account of nature of the binary, available as the basis for later research - which may well correct some of the assumptions.
Reporting your classificatory spectrum
You will probably working with a team who need your spectrum and its classification.
The uses of both types of spectral data, aside from their intrinsic interest, are as inputs to system modelling.