I have a question about LES.Is the 'filter' only used for theoritical analysis or it has to be implemented in the code? Nuemrical method such as FVM is a 'implicit filter', isn't it? So if the grid is fine enough, and sub-grid turbulence model is used, then it is LES, no explicit filter is needed, is that true?

Hi,

Almost all LES are carried out without an explicit filtering of the N-S equations because the mesh and the numerical method act as implicit filters. However, some LES purists recommend to use an explicit filtering in order to reduce error in subgrid modeling.

Moreover, filters are required for some subgrid models, such as dynamic Smagorinsky : they're used to obtain a test field with higher cutoff length to take advantage of Germano's identity.

Hope this helps

Best regards.

If you explicitly compute the Leonard terms, you need to specify a filter.

Unless otherwise one uses dynamic family of models people donot use explicit filters. ALthough some use.