Use constant background fields and classical particle trajectories as unit tests for characteristics
So far the integration test checks that the distribution function changes correctly according to the characteristics.
This test is a useful unit test for the characteristics. We could test the change of the distribution function against analytical results for constant background fields with a point wise comparison or L1,L2,LInf norm.
\lVert f_\text{analytical}(\vec x, \vec v, t) - f_\text{simulation}(\vec x, \vec v, t) \rVert \ll \epsilon
The following discussion from !125 (merged) should be addressed:
-
@mraeth started a discussion: (+1 comment) That's a very inacurate test, isnt it? Tolerance of 0.5?