Hi, I have done some work in the past looking at flow past a circular cylinder. We userd a comercial code, RANS turbulence models and then some rough DES models. My question is not about grid resolution but more model resolution in the spanwise direction. Does annyone have a feel for how many crossectional dimensions the model should extend int he spanwise direction, particularly when looking to undertake DES/LES simulations. I have herd that there should be a minimum of 10 diameters, however has annyone ever looked at this ? What are the potential problems ? Bob

10 is way to much if you are using periodic boundaries. Spalart published a paper a few years ago about DES calculations around a cylinder, they probably did spanwise domain sensitivity studies.