Fine vacuum gridding leads to numerical instability

[matt-pharr]

Vacuum does not automatically adjust the number of grid points it counts as "near the singularity" for the purpose of separating the singularity into a non-singular and an analytic term. This means that when the number of poloidal grid points gets sufficiently high, we start getting points close enough to the singularity that they start blowing up and the accuracy starts to suffer. This was originally noticed by Stuart Benjamin, I believe. A solution to this may be to adjust the number of designated singular points based on the number of poloidal grid points.

[jhalpern30]

@matt-pharr is there a known equilibrium/resolution where this bug appears and/or a script that reproduces and plots it? i.e. "Solovev ideal example when mtheta > 2048". Ok if not, but don't want to repeat work tracking one down

[matt-pharr]

@jhalpern30 not off the top of my head but you could ask Stuart Benjamin as I think he tracked this down at some point.

[logan-nc]

@jhalpern30 it seems like things are working robustly now... without a concrete example of something going bad, I'd argue for closing this rather than keep it around as a ghost. Do you agree or is there a concrete example of something going wrong to address?

[jhalpern30]

Agreed - I thought this was a bigger issue than it appears that it is, I took some test cases and ran up to like mtheta = 8096 and never observed this, so I think its an edge case. If we end up finding a legitimate reason I can add in a fix, but not worth the effort rn