We introduced a multi-level conductor surface roughness model in ADS 2011. Multi-level means you model little hemispheres on bigger hemispheres like the scanning electron micrograph (SEM) shown here. You can use it for transmission lines such as our Multi-layer Models library or in our laminar 3DEM solver Momentum. There’s a joint Agilent-Cisco MTT MicroApps presentation about it titled A Multi-level Conductor Surface Roughness Model by Yunhui Chu, Amolak Badesha, Jing-Jiang Yu of Agilent Technologies, and Sammy Hindi of Cisco Systems.
The topic has an interesting history dating back to the era when computer meant a human computer, in this case Miss J. G. Asbury who solved Maxwell’s equations by hand to 1% precision for the case of a periodic roughness. She was working for Samuel P. Morgan Jr. in his work leading the publication of Effect of Surface Roughness on Eddy Current Losses at Microwave Frequencies in the Journal of Applied Physics, Vol. 20, p. 352-362, April, 1949.
Our model is an excellent compromise between accuracy and computation efficiency. Of course, to model the frequency response of transmission lines on FR4 PCB materials you need to model not just skin effect and conductor surface roughness but also dielectric effects. Our version of the Svensson/Djordjevic model is a good approximation for the last one but it doesn’t account for resonant peaks in the dielectric. (You can see two such peaks at ~18.2 and ~18.9 GHz on slide 10 of the MTT presentation cited above.) If you need a model that accounts for these, Prof. Paul Huray discusses an approach in his book The Foundations of Signal Integrity pp. 198-215. Maybe that’s a topic for another day and another posting.