rd1042

This appears to be related to how the centering is performed in the term (Z/T e^(-v^2) ) in the streaming scheme. The current scheme calculates this source term as: source_term...

Just an update that, with branch bugfix/Davies-linear-EM-terms-maxwellian-centering-streaming, numerical instability occurs if ky is reduced to e.g. 0.01. The numerical instability gets worse as the timestep increases, which is the opposite...

Not quite - running the same input files for both simulations, the (delta phi and h) approach is more instability-prone (i.e. for a given ky, instability occurs if the timestep...

 Attached are mode structures for master & em_stella (the latter with the bugfix) showing instabilities for different dt with ky=0.01. master is stable for dt=0.004 (and 0.04, but...

> Brief experimentation with the input file for the implicitly treated MTM above suggests that the apparent bug with drifts_implicit =.true. is in fact caused by a bug with the...