Discussion Forum

I have gone a few cycles with Comsol support on this one without resolution, so hopefully you the user community can help.

Using Comsol 3.5a with AC/DC module.

I am trying to model a toroidal current ring within a toroidal conductor. The electromagnetic aspect of this model is symmetric about the Y axis, but other parts (to be added later) are periodic every 5 degrees of toroidal angle, so I am setting up the electromagnetic part as a 3-D, 5 degree slice of a torus.

How do I set up the boundary conditions at the toroidal boundaries in order to retain all three components of the vector potential ([Ax Ay Az]) and make the toroidal inductor & current ring appear continuous? I need the vector potential at the 5 degree boundary to be equal to the 0 degree boundary vector potential rotated 5 degrees about the y axis:

[Ax Ay Az]_5 = [Ax*cos(rev)-Az*sin(rev) Ay Ax*sin(rev)+Az*cos(rev)]

where rev is the 5 degree angle of revolution.


My options, seem to be:

(1) Magnetic or Electric insulation (nxH=0 or nxA = 0):
Problem: These conditions kill the contribution of the current ring or toriodal inductor, respectively.

(2) Periodic Condition (Adst=Asrc): AC/DC module Boundary conditions drop-down option
Problem: This works for the Ay component (axis of symmetry), but the Ax & Az components must be rotated 5 degrees.

(3) Arbitrary periodic condition with rotation: menu Physics->Periodic Condition->Periodic Boundary Conditions
This is the option that I think should work. I set the source variables on the 0 degree plane to be the vector potential times the rotation matrix
[Ax*cos(rev)-Az*sin(rev) Ay Ax*sin(rev)+Az*cos(rev)]
I set the destination variables on the 5 degree face to the vector potential
[Ax Ay Az]

Unfortunately, my setup for (3) (in the attached model) gives complete garbage for the magnetic fields on the boundaries, and I do not understand why. Is my implementation wrong, or am I simply completely wrong in assuming that there is a simple way to do a rotated continuity condition on the boundaries?

Thanks,
Nick