Because the smallest size of very fine particles that can be separated in an impactor is proportional to the ratio Re/M2, there is an interest in studying these instruments at moderate Reynolds numbers, Re, and at Mach numbers, M, up to sonic conditions. Here an inertial impactor of fixed geometry is tested at Reynolds numbers between 40 and 840, and in the transonic flow regime, at downstream to upstream pressure ratios between 0.42 and 0.79. Earlier theoretical and experimental work has shown that conventional impactors, with nozzles having relatively large length-to-diameter ratios, loose their resolution at Re smaller than several hundreds due to boundary layer growth at the nozzle walls. This effect is minimized here by means of nozzles having rapidly converging walls down to their exit, embodied by a thin-plate orifice with thickness to diameter ratio of 0.014. The nozzle to plate distance is equal to the orifice diameter. Test dioctyl sebacate (DOS) aerosols, with diameters ranging from 0.05 to 0.24 μm, were generated using a Differential Mobility Analyzer (DMA). Collection efficiencies were measured on-line at constant Reynolds and variable Mach numbers using two electrometers; one to monitor the current of charged particles impacting on the collecting plate, the other for the current of uncollected particles. A new method based on sweeping over the Mach number for a given test aerosol, rather than on varying the aerosol size at a fixed Mach number, provides an inversion technique which eliminates completely the effects of multiply charged particles from the DMA. Measurements show excellent resolution even at Re = 100, as well as through the transonic region, so that neither compressibility nor viscous phenomena limit the size of the smallest particle that can be inertially separated with high resolution. The limiting factor is Brownian diffusion. Its negative effects on particle collection efficiency curves are measured here for the first time using NaCl particles with diameters between 0.008 and 0.016 μm.