We introduce a computer-assisted procedure for folding large RNA chains into three-dimensional conformations consistent with their secondary structure and other known experimental constraints. The RNA chain is modeled using pseudoatoms at different levels of detail--from a single pseudoatom per helix to a single pseudoatom for each nucleotide. A stepwise procedure is used, starting with a simple representation of the macromolecule that is refined and then extrapolated into higher resolution for further refinement. The procedure is capable of folding different random-walk chains by using energy minimization, allowing generation of a range of conformations consistent with given experimental data. We use this procedure to generate several possible conformations of the 16S RNA in the 30S ribosomal subunit of Escherichia coli by using secondary structure and the neutron-scattering map of the 21 proteins in the small subunit. The RNA chain is modeled using a single pseudoatom per helix. RNA-RNA and RNA-protein crosslinks, reported in current literature, are included in our model. Footprinting data for different ribosomal proteins in the 16S RNA are also used. Several conformations of the 16S RNA are generated and compared to predict gross structural features of the 30S subunit as well as to identify regions of the 16S RNA that cannot be well-defined with current experimental data.