2016-04-23 10:36:43 |
Alex Valavanis |
description |
The gde program currently uses a FTCS solution to the diffusion equation, which becomes unstable if large time-steps are used.
It would be better to implement a robust matrix-based Crank-Nicolson solver, although this would need some thought if we're wanting time-dependent diffusion coefficients.
I'm setting this as a "Wishlist" bug, because the existing solver includes stability checking, so it can't give stupid answers... it's just slow/annoying. |
The qwwad_diffuse program currently uses a FTCS solution to the diffusion equation, which becomes unstable if large time-steps are used.
It would be better to implement a robust matrix-based Crank-Nicolson solver, although this would need some thought if we're wanting time-dependent diffusion coefficients.
I'm setting this as a "Wishlist" bug, because the existing solver includes stability checking, so it can't give stupid answers... it's just slow/annoying. |
|