Observed occasional numerical stability problems in some cases:
long range and complicated bathymetry
complicated sound speed profile with :smooth interpolation
In most cases these are resolved by using a lower tolerance or a stiffer solver, but not always. For now :smooth interpolation on SSP should be considered "broken".
The problem with :smooth SSP seems to do with q becoming negative. Perhaps the curvature in SSP is leading to caustics that are not dealt with correctly. Related to org-arl/AcousticRayTracers.jl#6 ?
Observed occasional numerical stability problems in some cases:
:smooth
interpolationIn most cases these are resolved by using a lower tolerance or a stiffer solver, but not always. For now
:smooth
interpolation on SSP should be considered "broken".