does CFDEM handle particle collisions analytically or by timestep?

I see two ways:

1) timestep integrating all the contact forces (normal and tangential - according to e.g spring dashpot model described in Kloss, Goniva, Hager et al papers)
which gives the velocity changes, new normal and tangential displacements and hence new values of forces

2) assuming the selected model and knowing the impact velocities, the loss of energy during the impact, collision duration and other parameters are calculated analytically which gives the after impact values of linear and angular velocities.
(looks like much faster way)

It looks like 1 is implemented however I am struggling to understand what should be the right timestep for dealing with collisions.

My apologies if my question is obvious, I am a complete beginner in CFDEM.