Temporal TBL simulation numerical instability

[filippo-moroni]

Good morning,
I am trying to simulate a temporal TBL as Kozul et al. (2016) with a tanh initial velocity profile and white noise near the bottom wall.
The bottom wall translates with imposed velocity. Periodicity is imposed in x and z directions.

I attach here the link to the BC file for the temporal TBL case (adapted from the channel flow BC file):
https://github.com/filippo-moroni/Incompact3d/blob/modified/2-solver/src/BC-Temporal-TBL.f90
and the input file:
https://github.com/filippo-moroni/Incompact3d/blob/modified/2-solver/examples/Temporal-TBL/input.i3d

I am experiencing numerical instability when the stretching of the grid is used in combination with the white noise.

These are the 3 initial time steps:

Simulations are stable when stretching or noise are removed (so individually they do not represent a problem).

Stability is recovered in both cases of uniform mesh I tried (istret = 0 and istret = 3 with large beta).

I tried different spatial and time schemes (2nd order/6th order in space, explicit and semi-implicit time integration) and also different approaches for the white noise (very low intensity, different locations, filtering). I also tried different BC for the top wall (no-slip and free-slip).

How can I solve in order to trigger turbulence but also limit the number of points in y-direction (so using istret = 3)?

Thank you,
Filippo

[filippo-moroni]

I solved using v4.1 instead of v4.0 (that had the bug for istret = 3)