Assuming you're working in solution, in principle simple zg pulse program is all that you need. There is no standard parameter set for 33S, however, because the sensitivity is very low and you will have very significant quadrupolar broadening, so there has been extremely limited use of 33S nmr. The quadrupole moment is higher than that of 14N, which is itself rather difficult to observe in all but very small or symmetric molecules. The sensitivity at natural abundance is at least 10x worse than carbon, even without allowing for fact that the broad lines will reduce sensitivity even more.

For setup, (NH4)2SO4 or 2M Cs2SO4 in D2O come at about 0ppm, whilst CS2 comes at -333ppm. You want to use a short acquisition time and short D1, perhaps 0.05s for the standard and even less for real compounds, so at least you can collect a large number of scans relatively easily. There are a couple of example spectra of the standards at http://chem.ch.huji.ac.il/nmr/techniques/1d/row3/s.html -real samples will be somewhat broader even than CS2, and may be too broad to see. Elevated temperature and low viscosity solvents may help for the real samples.

If you are working in the solid state, Cement and Concrete Research 36 (2006) 1781 – 1783, has some interesting results, but at much higher field (900MHz for proton), which makes things vastly easier in solids.

What sort of samples are you looking at?