I'm using 6th order tri-diagonal differencing scheme as described in Lele for a turbulence DNS. I want to find out the smallest possible turbulent structure that can be run in the DNS. My questions are.
- Does eddy size matter wrt the differencing scheme?
- If yes, how do I find out the limit.
The obvious idea would be to simulate a turbulent field and increase the no of points until mesh independence is achieved, however in the paper of Lele, there are Fourier error analyses of these different schemes.
Would it be correct if I used the resolving efficiency and divide my wavenumber space in the DNS into resolved and unresolved?