Just as note, if the Re=4700 is the bulk-based, then you get a quite low Re_tau for which also a DNS is possible. How did you estimate 200 milion of nodes?

Dear Paolo
I am pleased to hear from you again and I appreciate that.
Referring to your fifth statement:
There is central differencing option for momentum equation (please see attached). Is that what you mean? If yes, I have grasped it.
Again, should I choose central differencing for the energy equation as well?
I have similar question regarding the transient formulation! which one is the preferable, bounded second order implicit or second order implicit? should I avoid bounded second order implicit for transient formulation too?

If your physics is just heat transfer, i would choose:

PRESTO! for pressure
Central differencing for momentum
Bounded Central differencing for temperature
Second order implicit for time
Non iterative time advancement with Fractional Step


Ideally, using Central differencing also for Temperature would be better, but i have no specific experience on this. I always used bounded for temperature, to avoid any possible violation of minima/maxima. You can give it a try and see what happens.

Dr. Lampitella, but according to professor Pruett; you need to apply a windowing procedure because the Navier Stoke equation is quasi-periodic in time. Therefore, to follow approach (3) then you need to apply Parzen windowing.

Dear Julio (let's stay on the personal level if you don't mind, Paolo or sbaffini is more than fine),

you are absolutely right. Still, for the purpose of computing an FFT with, say, thousands of samples, assuming the existence of periodicity is not, in my experience, really altering the underlying physics. You can do this check in a channel flow, comparing the spatial spectra with the temporal ones taken at some locations along the same line for the proper amount of time (up to the same amount of samples). Actually, i think that windowing here is much more invasive.

On a more general level, what i'm saying is just that he can have a set of signals at given locations, just as would happen in a laboratory experiment (because that's the hope for LES). He can then just use them as he would do for the experimental signals. In that context, windowing or not, FFT is very common.

Yes I agree with you. Could you please provide me with the paper of that channel flow where comparison with spatial and temporal spectral are done?
I have always had the same impression from the temporal spectral computation, following the idea of the experiment. But, if I am not mistaken ( I need to double check) I read a paper where the authors tried to get the energy spectrum by using the two point correlation instead of using just one probe. That paper was about turbulence in ocean, very different to a channel flow though...!!!

Well, i don't have a paper, it's just a check you can do by yourself if you have a channel flow data set.

I still suggest some administrators to make an extract of this discussion in a new post...