LES unsteady : Turbulent Mixing Layer

Flav

I'm studying the well-known Goebel and Dutton supersonic Mixing Layer under Fluent 5. My problem is 2D. The smallest element of the mesh is about 1e-5 m long. The domain is 0.5*0.048 m*m. The purpose of my study is to modelise the eddies produced by the turbulent mixing layer. I've already carried out a calculation with the k-eps model on a steady state, and it's allright. Then I set the solver to Viscous / LES. I use the "ideal gas" option, and the viscosity is set to "Suntherland".

Now, which solver do I have to choose ? Explicit or implicit ? Coupled or segregated ? Which time-step ? How many iterations per time step ?

I thought that the implicit solver was inconditionally stable, but it doesn't seem to be so. I can't find the correct settings to iterate towards a converged solution . (I also set the Courant Number to 0.5)

Concerning the inlets, I have input a profile for U along y. Do I have to add some kind of "noise" to U in order to trigger off eddies production ?

As you can see, I have many questions to ask. The fact is that the LES unsteady solver is not well explained in the Fluent Help/Tutorial files ...