Sound source separation in diesel engines can be implemented using a Wiener filter or spectrofilter that can extract the combustion contribution from the overall noise. This filter characterizes the transfer function between a cylinder pressure and a measurement point. Thus, a spectrofilter is computed for each cylinder, which induces that an engine is represented by several filters. Its computation depends on operating conditions, which makes the full characterization of the engine very time consuming. The goal of this study is to determine filter parameters modifying the perception of these sounds. An experiment has shown that the engine timbre is mainly dominated by the cylinder pressure. This allows the use of an averaged filter that can be applied to all operating points without lowering the accuracy of the source separation. To that purpose, modal parameters are identified for several spectrofilters to estimate them in the most efficient way, while allowing a good reconstruction. Perceptual experiments are conducted to analyze the effects of this estimation on the combustion noise.