Cross-talk between transmission lines and signal pads

varphi42varphi42 Member Posts: 1

I have a 7-layers PCB with the following layer stack-up: high voltage plates (1), segmented signal pads (2), ground shield traces (3), signal traces (4), ground shield traces (5), segmented signal pads (6), high voltage plates (7). The layer 4 (signal traces) is connected to layer 2 and 6 (segmented signal pads) with a via - see pictures attached - and is meant to extract a triangular current signal (1 ns rise time, ~300 ns decay time), induced on layers 2 or 6, outside of the PCB. There are several 'segmented' signal pads in layer 2 and 6, each connected to one single signal trace. I want to evaluate the cross talk due to the fact that the signal trace from a given segment run beneath other signal pad segments. The ground layers 3 and 5 are here to prevent this cross talk but they are traces, not plates and I want to see what width those shield traces should have to reach an acceptable level of cross-talk between a signal trace and the other signal pads.

I tried to use the cross-talk functionality within SIWave but it gives me 0 everywhere, I suspect it only considers cross-talk from trace to trace, not from pad to trace. I also tried to define ports and get the S-parameters but see an unexpected behavior: when I enlarge the width of the ground shields, the cross talk actually increases suggesting that the cross talk is actually occurring through the shields themselves (they all belong to the same net). Here are my questions:

1) Is SIWave the proper tool to do such a study or is there another one more suitable? Another possibility I see is to derive the equivalent circuit from Maxwell and then use Simplorer?

2) When I display the mesh in SIWave after solving the S-parameters, it seems to be defined only on a plane (and it is very coarse), even when checking the 'Q3D Solver' option. Can I force SIWave to solve things in three dimensions?

3) The ground shield traces are assigned to the 'power/GND' net list, is there something else to do in order to have SIWave considering these traces as perfectly grounded?

Thank you very much for any input!


Answers

  • pmunagapmunaga Posts: 140Forum Coordinator

    Hi @varphi42,

    Please use "upload image" option instead of attachments as Ansys employees cannot access these.

    Best regards,

  • varphi42varphi42 Posts: 5Member

    Hi @pmunaga ,

    I think this is what I did. I clicked on the "Upload image" button (below the "Leave a comment" window), maybe the fact that it is a pdf turns it into an attachment? On the other hand, I can access these images even when not being logged in, anyone should thus be able to see them. Can you confirm?

    Thanks

  • pmunagapmunaga Posts: 140Forum Coordinator

    Hi @varphi42,

    As I said earlier, we cannot access the attachments.

    You can use screenshots of your model to help us serve you better.

    Best regards,

  • varphi42varphi42 Posts: 5Member

    Then I guess I am confused with what you call the "Upload image" option, since the button I used is called "Upload image"... Let me then try to drag and drop the screenshots directly in the post:


    Thanks

  • mannymanny Posts: 23Forum Coordinator

    Hi Varphi,

    See my answers below.

    Q 1. No.

    Q 2. No. SIwave meshes the planes to see where voids and other things are. We can see the mesh only for DCIR analysis. For a rigorous 3D analysis, use HFSS 3D Layout.

    Q 3. No. Treating those traces as perfectly grounded wouldn't be physical, hence it is not normally desirable.

    Crosstalk scan in SIwave is designed specifically for signal lines. Not for power nets, planes, pads, polygons or arbitrarily-shaped geometries. Here’s an app note.  Try to convert the pads to planes and then the planes to traces. But this is laborious. Your best bet is to run SYZ solve in SIwave.  For a 3D analysis use HFSS 3D Layout.

    Best,

    Manny

  • varphi42varphi42 Posts: 5Member

    Hi Manny,

    Thank you very much for your answers and suggestions, I will thus try with HFSS!

Sign In or Register to comment.