Improving residual error, magnetostatic

       Hi  everyone,

I come back to the magnetostatic module (of FreeFem) and i discovered a residual error too much high for an another problem that i’m working on.

An analytical formula give the field at the center of the coil:
Bcenter = Jfil* (asinh((2* Rext)/length)/2-asinh((2* Rint)/length)/2)* length* u0
Jfil=5E6 A/m^2 , length=1 ; R1=0,1 m R2=0,2 m , Bcenter = 0.601128704 T

But with my files, i have still an error of 1-4% except if i increase drastically the mesh with more than 700 000 tetrahedrons and it take time to solve.

Could you give me any advice to decrease the error ? I would need something inferior to 1E-5 in relative.


First a error of 1E-5 is very small, this imply use order 2 finite element at least.

Yes, it’s very small, but I want to find the result from two opposite vectors. The result is between 1E-1 - 1E-2 ( in relative error ), so i need a sufficient precision to examine.

The module’s example use the P2 element ?