[kaya]

I have a 3d Navier Stokes solver

x and y axes are Fourier treated and z(vertical) is handled with Pade-4 (4th order compact) scheme

Non-linear conective parts(which are skew-symmetric) on fourier modes are 3/2-padded.

I am using a fractional 1-step method to time step with AB2. Firts find an intermediate velocity without having pressure in the momentum equations, solve the poisson equation for pressure (which is staggered on on z axis), add the derivatives on intermediate velocities to find velocities at the new time step.

I run simple low reynolds boundary layer flows and I see some wiggles (2h waves mainly on the vertical axes) I don't know why exactly.

Does anyone have any idea?

[FMDenaro]

generally, that is a sign of a pressure decoupling ... have you checked if you fulfill the divergence-free constraint in each node?

[kaya]

Yes, I had checked that, continuity was satisfied.

after solving for pressure, unlike an iterative scheme I add pressure derivatives to intermediate velocities and don't iterate again, just go to the next time step. since its the poisson equation that I solve at the very last, its also expected to have divergence free condition. But can this solution strategy cause local wiggles when there is a lot of shear ( such as having two opposite sign vortices very close to each other )?

or would it be wrong to think that maybe the grid was not fine enough?

thanks

ps. I currently can't run the code now since I am developing a new version from scratch; yet i would like to hear your thoughts because to my experience, in my spectral simulations I always had some (either visible or not) 2h waves (wiggles) somewhere close to where I put say some strong body forces and never understood what was going on behind.

[FMDenaro]

Originally Posted by kaya

Yes, I had checked that, continuity was satisfied.

after solving for pressure, unlike an iterative scheme I add pressure derivatives to intermediate velocities and don't iterate again, just go to the next time step. since its the poisson equation that I solve at the very last, its also expected to have divergence free condition. But can this solution strategy cause local wiggles when there is a lot of shear ( such as having two opposite sign vortices very close to each other )?

or would it be wrong to think that maybe the grid was not fine enough?

thanks

ps. I currently can't run the code now since I am developing a new version from scratch; yet i would like to hear your thoughts because to my experience, in my spectral simulations I always had some (either visible or not) 2h waves (wiggles) somewhere close to where I put say some strong body forces and never understood what was going on behind.

spectral methods can produce Gibbs phoenomenon in case of shar gradients...that happens when viscosity is low and not causing difusion to dissipate spurious high wavenumbers. But you wrote that are yousing de-aliasing, therefore wiggles of 2*h wavelenght should not appear.

I would check still for pressure decoupling ...