In a longitudinally ventilated tunnel, a minimum airflow velocity is required to prevent the formation of a back flow of smoke. Theoretical and semi-empirical models are available to relate this “critical velocity” to the fire heat release rate. Experiments were carried out on a scale-down densimetric tunnel model in order to investigate how the buoyant source size may affect the critical velocity. A particular result is that, by increasing the source size and for a given convective heat release rate, the critical velocity reaches a minimum in case of pure plume, for which momentum and buoyancy effects are balanced.