I would like to know how the solver sub stepping differs from the dop sub stepping. It seems that the sub stepping on dop level is not a “scaling” of the various sub steppings in the different solvers.

What is the “secret” behind that topic?

The global substep setting drives all the solvers. If you set the substep to 2, then all the solvers will at a minimum will solve twice per frame. If you have two or more solvers in your DOP simulation, it's a good idea to increase the global substep so that all the solvers march forward together.

The substep settings on each solver relate to that solver explicitly. You have to read up on the help for each solver to see how the substep method is used. Many solvers support multiple sub steps per each global substep. Some support a CFL predictor to control min-max sub steps per solver such as the Smoke/Pyro Solvers and the RBD solver to remain stable and to prevent collision penetrations respectively.

thank you Jeff for your explanation.