Bhanu Gupta

Regarding capture Frequency(f) <5 you would not require refining the mesh but having longer signal length which means more flow time for recording the data(Atleast 0.2sec matured signal for f=5 Hz resolution). LES would definately require resolving BL and hence can be expensive, I would suggest using Stress Blended Eddy Simulation(SBES) turbulence model. SBES is scale resolving turbulence model resolving core flow using LES and near wall modeling using SST KW. This is one of the recommended turbulence model for Acoustics without resolving the boundary layer.