The viscosity solver implemented is from this paper: Accurate Viscous Free Surfaces for
Buckling, Coiling, and Rotating Liquids
. One of the authors of the paper has a GitHub repository showing an implementation of the method: https://github.com/christopherbatty/VariationalViscosity3D
This method is quite accurate for high viscosity liquids, but there are still some limitations and it can’t simulate some more complex viscous behaviour such as honey coiling. This is an interesting video that shows a more accurate viscosity method with some comparisons to the method that we implement: Simulating Honey Coiling | Two Minute Papers #158.
Some viscosity solvers require many substeps and this depends on the viscosity simulation method. Sometimes a simulation method needs to take substeps to prevent the simulation from ‘exploding/blow-up’ due to loss of accuracy when the time between frames is long.
The viscosity method that we use is described as ‘unconditionally stable’. This means that it is stable and will not explode even if the time between frames is long. But you still might want to add substeps in our simulator if the viscosity is very high. More substeps puts less stress on the viscosity solver and a small timestep can help improve accuracy and can help the viscosity solver find a better solution.
The surface tension method that we implement is not ‘unconditionally stable’, meaning that it can explode if the timestep between frames is too large. That is why high surface tension simulations require more substeps.
There is a surface tension method that I would like to implement that requires much fewer timesteps: A Multiscale Approach to Mesh-based Surface Tension Flows.
The Mantaflow project is planned to replace the current default fluid simulator in Blender 2.81 and will be a huge upgrade from the default fluid. You can try it out now in the ‘Mantaflow Blender branch’. I think there are also builds for Blender 2.79. More info can be found in the Blender Artists thread here: Lets talk about Mantaflow
The progress of the Mantaflow code review can be found here: https://developer.blender.org/T59995
Recently I have seen some discussion of improving simulations within Blender here and here. I think that Blender still has a long way to go in terms of development to create a simulation environment that can match what Houdini is capable of.