Then, you can always do it old school style, as it was done here https://www.researchgate.net/publication/228505595_Experiences_from_FTF_NanoScience_Lab._at_the_University_of_Lund_use_of_a_2-stage_isolation_system_with_very_low_natural_frequency
/Claes
Article Experiences from FTF NanoScience Lab. at the University of L...
I am probably not the best to answer your question as I have not used Abaqus since the mid 1990's, but I will give it a try in any case.
The Rho*C approach that you mention work only for compression waves in 1-D waveguides, e.g. pipes. If you apply Rho*C as impedance on your boundary it will reflect also compression waves that impinge at an angle. On top of this, for soil, there are more wave types than compression waves to consider.
Abaqus has Infinite elements.These are much better as they can absorb also waves that impinge at an angle. As far as I am able to gather, they support at least also shear waves.
As a rule of thumb, this type of elements work the better, the more closely their faces are normal to the impinging wave front. This observation has practical consequences to how you should mesh to get a quality interface boundary for such Infinite elements.
If you look at section 2.2.1 (p 996) here, you will find some example problems that you can try and repeat and find your way using such Infinite elements. http://www.maths.cam.ac.uk/computing/software/abaqus_docs/docs/v6.12/pdf_books/BENCHMARKS.pdf
Just as a word of warning. Acoustics involves only compression waves. Soils have more wavetypes. So make sure the elements you use are up to the job you intend to apply them for.
Then, you can always do it old school style, as it was done here https://www.researchgate.net/publication/228505595_Experiences_from_FTF_NanoScience_Lab._at_the_University_of_Lund_use_of_a_2-stage_isolation_system_with_very_low_natural_frequency
/Claes
Article Experiences from FTF NanoScience Lab. at the University of L...
you can use viscous dampers at the boundary nodes. The coefficients of the dampers should be density times shear wave or p wave velocity (according to the direction of motion) times tributary area over the node.
create set with those nodes. Then assign viscous damper from engineering property.
Just as information, adding viscous dampers adds damping for sure and it does absorb either P - or S- waves that impinge normal to the boundary.
Unfortunately waves that impinge at an angle or is of the 'other type', i.e. if dampers are set for, say, P-waves, the 'other type' would be S-waves and vice versa will be reflected to some degree.
In pure math language, adding dampers would imply adding a substructure matrix with terms on its diagonal, when you need a full matrix.
The above mentioned old school style works better but is computationally costly.
As food for thought - you cannot somehow use substructuring for your analysis?
I am agreeing with Claes. It is absolutely right that viscous dampers can work only for orthogonally propagating waves. To mitigate this problem lot of research has been done, such as Kelvin element. For different types of absorbing boundary any standard SSI book can be referred, for example,
2. Song, Ch., Wolf, J.P., (1994) Dynamic stiffness of unbounded medium based on damping solvent extraction, Earthquake Engineering and Structural Dynamics, vol-23, pp: 169-181
I have also tried to compare the performance of viscous type of boundary with fixed one in https://www.researchgate.net/profile/Arnab_Banerjee6/publications
Substructuring maybe a good idea to solve that type of problem.
FYI - this is exactly what we did in a project involving soil interaction for a large scale scientific structure.
As you are interested in this topic, a tweak we used was that Wolf's method specifies a matrix inversion of the interface coupling matrix, i.e. to dig a hole in the substructure (which usually is an analytical wave propagation model).
Now, our coupling matrix was so large that inversion of this matrix was a feat in itself. Not at all practical even on a supercomputer.
We circumvented this by the use of an old Nastran trick that is used to 'turn off' elements, e.g. in shape optimization. We used a homogeneous layer of elements with negative Young's modulus for the part of the analytical 'substructure' that should be 'dug out' from the analytical substructure and inserted in there a second substructure model, i.e. the large scale structure of interest.
There is a freeware code called Code Aster (CA). http://www.code-aster.org/V2/spip.php?rubrique2
CA supports anechoic boundary conditions by the use of a 'rubber band mesh'. I imagine that this mesh does provide some reflection. In any case, on top of this, there is an interface developed to an analytical code called MISS3D. I know little more than this, but if you are researching the topic, you might want to take a look at this interface. http://www.google.se/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CCcQFjAA&url=http%3A%2F%2Fwww.theses.fr%2F2013ECAP0006%2Fabes&ei=JxPMVLzCGsXZywPdg4CICQ&usg=AFQjCNH3T-V2n7PdKjWn7MzsAG8WudogeQ&bvm=bv.85076809,d.bGQ&cad=rja