The measurement of earthquake source parameters is affected by large uncertainties, and different approaches lead to large variability in results. One crucial aspect is the trade-off between attenuation (Q) and corner frequency (fc) in spectral fitting: The source corner frequency, inversely proportional to the fault size, can be severely masked by attenuation and site effects. In this article, we describe a method to solve the trade-off based on the fit of displacement spectra to find the source characteristics (corner frequency, fc, and the signal moment, Ω0) and the single-station attenuation operator (t*), in addition to the site response. We follow a parametric approach based on the use of 3D Q seismic tomography and a bootstrap-based method for selecting the best spectra fit. The correction of attenuation with synthetic values derived by 3D attenuation tomography efficiently deals with the trade-off between source and path terms, leading to small uncertainties in the determination of source unknowns (fc and signal moment, Ω0), thus yielding constrained estimates of source parameters for low- to medium-magnitude earthquakes. We show an application to the Emilia 2012 seismic sequence, for which we computed the source parameters for 1240 aftershocks (from an initial dataset of 1748) with local magnitude ranging from 2.0 to 4.7 using the spectral fit from P and S waves. About 80% of stress-drop estimations are characterized by relatively low uncertainties (within 20% of the estimated values), with maximum values of about 40% for the remaining 20%. The attenuation correction is effective to determine source parameters for small-magnitude earthquakes; hence, we obtain reliable estimates of source parameters for the entire aftershock sequence. This approach gives the opportunity to infer the mechanical state of a complete fault system by taking advantage of the larger number of low-magnitude events (with respect to the largest ones) that always follow a major earthquake.