As shown in the following figure, the worst-case victim is in the middle column with four diagonal and two non-diagonal aggressors. The routing layers (via length) for the victim in the middle column and the two diagonal aggressors to the right of the victim are swept to visualize the effects of the via vertical parallelism to the crosstalk. The routing layer of the two diagonal aggressors to the left of the victim is kept to L24 (long vias) in the sweep analysis.
The following table lists the simulated power sum crosstalk for 56 Gbps PAM4 (at 14 GHz) for various routing cases. The data shows that the crosstalk amplitude correlates with the via vertical coupling length between victim and aggressors. A victim with four diagonal aggressors needs a via length < 45 mil to fulfill the –35 dB TX-to-TX power sum crosstalk requirement as stipulated in Table 1.
| Powersum TX Xtalk (dB) | Victim Via Length (mil) | Aggressor Left Layer | Victim Center Layer | Aggressor Right Layer | Comments |
|---|---|---|---|---|---|
| –28.7 1 | 102.3 | L24 | L22 | L22 | Routed two layers (L22 and L24) long vias |
| –28.9 1 | 102.3 | L24 | L22 | L20 | Routed three layers (L20, L22, and L24) long vias |
| –34.6 1 | 44.1 | L24 | L11 | L11 | Routed two layers (L11 and L24) short vias |
| –34.0 1 | 54.6 | L24 | L13 | L11 | Routed three layers (L11, L13, and L24) short vias |
| –34.5 1 | 44.1 | L24 | L11 | L22 | Routed three layers (L11, L22, and L24) short vias |
| –35.9 2 | 34.6 | L24 | L9 | L22 | Routed three layers (L9, L22, and L24) short vias |
| –37.8 2 | 26 | L24 | L7 | L22 | Routed three layers (L7, L22, and L24) short vias |
| –38.7 3 | 26 | L24 | L7 | L7 | Routed two layers (L7 and L24) short vias |
| –44.3 3 | 8.6 | L24 | L3 | L3 | Routed two layers (L3 and L24) short vias |
| –40.3 3 | 26 | L24 | L7 | L3 | Routed three layers (L3, L7, and L24) short vias |
|
|||||