Discussion Forum

In using the .mf interface of the AC/DC Module, a very useful numerical trick is to set the conductivity of the air around a conductor to have a small (but nonzero) value. Magnetic field effects in the conductor are essentially unaffected, and avoiding an exact 0 facilitates numerical convergence.

This trick has been serving me really well, but now I have run into a problem with it. I would like the domain of interest in which the magnetic effects occur to have a quite small conductivity, even down to 10^-3 or 10^-6 S/m. The problem is, I have been using =1 S/m or .1 S/m for the conductivity of the air domain around this domain of interest, which would no longer be a good approximation.

The obvious solution of simply lowering the air conductivity further than .1 S/m causes a failure to converge. I have been able to get convergence for air conductivities as low as .01 S/m, but not much past that. So is there a way to set up Comsol to take extra care with this value, like a custom precision or error tolerance? Or, is there a way around using a very small air conductivity value and still getting accurate results?