This work describes a novel approach for obtaining digital samples of porous media based on a statistical knowledge of the microstructure of interest. The present formulation introduces a contact law based on bubble physics that is capable of handling interferences among spherical primitives of different diameter placed in a representative elemental volume, while the volume is compressed to yield a target porosity. The result is a statistically accurate mathematical model of a permeable, spherical-void-phase porous material that has the added feature of being spatially periodic in all principal directions. To validate the approach, digital samples of spherical-void-phase carbon foam were generated and discretized for use in hydraulic and thermal Computational Fluid Dynamics simulations. Relevant transport properties were computed from the simulation results, and compared to similar data found in the literature. (C) 2014 Elsevier Ltd. All rights reserved.