Issue link: https://iconnect007.uberflip.com/i/1071356
82 DESIGN007 MAGAZINE I JANUARY 2019 models used in the analysis. To have accurate interconnect models, the material models must be either confirmed or identified [1] . Thus, the things observed on the graphs here cannot be used to draw any design guidelines for any other cases. In this particular case, one can observe approximately 0.01-dB/mm differences in attenuation and a0.4 ps/mm in phase delay for the investigated structures. This is the case of a high-loss dielectric. The conductor roughness was neglected too, and this assumption alone can ruin your design [1] . The conductor rough- ness can be easily accounted for in the sim- ulations, but there were no numbers to define the model for this case. The outcome of such investigation may be substantially different if the dielectric is very low loss and conductor roughness is taken into account. The point is that all of that should be simulated and taken into account. S-parameters Scattering or S-parameters and compliance metrics derived from them are becoming popular in designing interconnects for digital systems. S-parameters of a trace segment with the meshed reference plane can be easily computed as soon as the modal and per- unit-length parameters of the corresponding periodic structure are extracted. Note that this type of analysis is not an approximation; it follows from the physics of the periodic structures [2 & 3] . An alternative to this analysis is a simulation of a complete link in a 3D EM solver, which could be a very time- and resource-consuming process. Note that most of the flexible interconnects have micron- sized conductors that bring the onset of skin effect into the multi-GHz bandwidth. Thus, simulation of such links requires meshing of the conductor interior, as it is done with Trefftz finite elements in all examples in this article (Simbeor 3DTF solver). As an example, insertion and reflection losses of a 10-cm segment of the traces over a meshed plane were computed and compared with the trace over the solid reference plane (Figure 7). The insertion loss differences are within 1 dB at 20 GHz for the investigated configurations. However, reflection losses vary dramatically. This is the result of differences in the characteristic impedances. Coupling and Mode Transformations Finally, what if the meshed plane is used to shield traces on the opposite sides? You can expect almost ideal isolation in case of the solid plane; in other words, the traces can be considered as not coupled with the solid plane. The fields do not penetrate through the solid planes at the frequencies with well-developed skin effect. Coupling at lower frequencies below the skin effect does not matter for the signal integrity analysis (interconnects become electrically short). When you cut the holes, the Figure 6: Attenuation (left) and phase delay (right) for traces over a solid and meshed plane.