Sometimes social networks (in vias) are not a good thing
1/3/2011 5:59 PM EST
To reach the inner layer of the signal lines on the striplines, vias are used from the top layer (pink) to the buried layer (teal). While the signal lines have vias providing the continuous path, the return vias are located some distance away, where the probe contacts for the return path are located. The distance between the signal and return vias is 75 mils in this launch.
Could the far end cross talk in the stripline structure be due to the cross talk between the two signal vias in the stripline structure? After all, the cross talk arises between the signal-return path loops, and the stripline launches have some pretty large loops.
If the launch region with the vias were the source of the cross talk, Jeff reasoned, there should be noise coupled into the victim line, only when the edge passes by the launches on both ends. Some of this noise would be coupled back to the source when the edge passes by. The near end cross talk spatially resolves in time, where the coupling happens.
On the other hand, if the far end noise in these stripline traces was really due to a non uniform dielectric material distribution, the noise would couple uniformly down the line, as is the case in microstrip.
The figure below shows the measured TDR and near end cross talk (NEXT) measurements for the stripline and microstrip traces.
Figure 4. NEXT showing the enhanced cross talk when the edge hits the beginning and ends of the lines.
Sure enough, he found the location of the coupling to be at the launches, due to coupling in the large loops between the signal and return vias.
"My vias are talking to each other too much," Jeff said. "I want anti-social vias." The solution, he found, was to be sure future designs used return vias in closer proximity to the signal vias. This is how to engineer "anti-social vias."