I tought it may be dtmax, and I can get the Julia solver to give me a result that doesn’t have oscillations by reducing dtmax. But then Julia takes longer than Matlab (3x as long), while it normally would be 10x faster.
I was wondering if the speed-up should be that dramatic, since I assume ode45 is implemented in c, so should be fast as well.
Could it be that ode45 is quicker in lowering the time step to avoid oscillations?