A Large Eddy Simulation of a subsonic operating point (test case number two from Bake et al. [9]) of the EWG is presented in this paper. This study has three goals. The first is to provide a larger database on the generation of indirect noise than the one provided by RANS simulations and analytical models. The second is to determine the dominant noise in the subsonic case (direct or indirect). The third is to understand the reason of the overestimation of the pressure peak noise when the nozzle Mach number increases (i.e. Figure 1). As an explanation to the third phenomenon, Howe [20] suggested firstly that in cases with flow separation in the nozzle diffuser, vortex sound is strongly correlated with entropy noise and can dramatically reduce the overall sound level. Secondly, the strong deformation of the hot slug in the nozzle throat reduces the entropy gradients within the front and rear interfaces of the slug generating a decrease in acoustic pressure.