Modern computing power has made noise simulation models, i.e., prediction of actual time histories of sound levels, practical. Switzerland uses its FLULA model for airport noise contours. NASA’s new RNM rotorcraft model uses simulation to address the complexities of rotorcraft noise. The Wyle/USAF NMSIM fixed‐wing model has been used to address significant effects associated with complex terrain. A requirement for this type of model is the three‐dimensional noise source properties of the aircraft being analyzed. This information is generally extracted from fly‐over noise tests. Because noise is needed on a complete sphere (or at least a hemisphere), parts of the noise sphere are associated with long propagation distances and/or grazing incidence to measurement positions. This paper reviews the practical problems associated with turning fly‐over measurements into source models. Examples from experience with RNM and NMSIM are presented.