Is there any reason I couldn't use a very coarse mesh (fine enough for RANS) also for LES/DES?

I am slightly confused about the practical difference between LES and URANS. If I refine the grid, I will resolve appropriately smaller eddies with URANS, much like with LES. So on the same grid, what do I gain by switching to LES?

Why not using SAS instead?

Scale Adaptive Simulations?

I don't have an exact problem to solve, just trying to understand the difference in real life

LES models are typically dynamic

The coefficients are dynamically determined by the simulation itself

For RANS the model coefficients are fixed

but wait

if your temporal resolution is more refined you're breaking the rules

what do you mean by that?

I mean that could break the assumption of URANS/RANS itself

like running a pimpleFoam with a very small timestep?

Yes

As an example take a look at this thread

very nice, thank you very much

🙂

Or this one:

Implicit filtering is not due niether to LES nor related to the OpenFOAM code. It's a fact of low-order FVM, since the basic principle of FVM is the MEAN VALUE THEOREM (MVT) which, implicitely, is a median filter.

i meant this: https://www.cfd-online.com/Wiki/Large_eddy_simulation_(LES) However, most practical (and commercial) implementations of LES use the grid itself as the filter (the box filter) and perform no explicit filtering.

It's difficult to assess the practical difference, if the theoretical reasons are not clear. Not going to make an exhaustive list, but the theoretical difference between RANS and LES, is what kind of kernel you use to filter. Very, extremely, roughly speaking in RANS you filter in time and in LES you filter in space...

And that I answered

that's the answer as well, thank you very much

Yes, the reason is called Kolmogorov scales, or scales where eddies dissipate to viscosity.

The MVT in FMV will guarantee your simulations will not crash, but you'll get spurious numerical viscosity.

This is not entirely true. And this is not the difference between LES & RANS.

could I estimate required grid spacing from one of Kolmogorov scales (integral? energy transfer?)