• State TI Thinks Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 1 answer
  • Subscribers 46 subscribers
  • Views 235 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

AFE44I30: GND Plane under PD input pins & etch

Daniel Najemy
Prodigy 10 points
Part Number: AFE44I30


Tool/software:

What is the recommended approach for allowing the PD inputs (pin and etch) to have ground plan underneath?

In our application, the layout space is very tight (approx 4.4mm x 3 mm), slightly bigger than the part itself! With that said,
it's very difficult to keep the TX lines away from the PD input pins and etch. In fact, there are parallel runs where
the PD etch and TX etch are only a thin dielectric (in a flex circuit) away from each other.

One idea was to add another Cu layer in the flex assembly, and make that layer GND so that the TX and RX etches are
separated by that GND plane.

In discussions with partners of ours, advice came otherwise. There was apparently some history of noise coupling from
that GND plane into the PD inputs.

I do see GND under PD etches on your eval board, but that same run of etch drops down a layer onto the GND plane!

Could you please advise? Adding another layer to our stackup is not preferred, but if necessary to keep noise off the 
PD inputs we'll head in that direction.


  • 0
    TI__Guru 66706 points

    Hi Daniel,

    Thank you for your post.

    When the RX and TX paths need to run close to one another, or cross paths, it is best to include a layer or two of separation between them. You can sometimes get away with a 2-layer PCB if the form factor permits you to run the traces away from one another, but I understand your constraint. 

    In the EVM layout, you'll notice a few important techniques:

    1. The PD inputs are routed as differential pairs to ensure noise couples equally to P and N inputs.
    2. The TX signals are routed to the right of the PD traces. (Note: Other PD inputs from the second connector are routed into the AFE from the right-hand side, directly below the traces for TX1-4. Generally both connectors are not used simultaneously, but if they were, there is both a solid GND plane as well as a power plane between them to reduce coupling).

                 

    Regards,

    Ryan

  • 0 in reply to Ryan Andrews
    Prodigy 10 points

    Thanks for the quick response Ryan.

    1. Good point on the differential routing, we'll try our best to achieve that.

    2. Here's a pic of our latest routing:



    You can see the diff pair in blue coming up from the bottom (that's where the optics are). The brown-ish etch just to the 
    right of the right side via is one of the TX lines (again, all the optics are below!). Those are on adjacent layers in
    a small flex (3 layer stackup).

    3. Would there be a benefit to add a 4th layer to put a GND plane in between those layers? It's painful to the flex
    manufacturing process, but if necessary....or better to just try to route the diff pair away from the TX line?

    4. The feedback we were given was "don't put a ground plane under the PD lines, you'll couple noise onto them from
    the GND plane" That's mostly counter intuitive to what I know, since the GND plan usually provides a pretty good shield.

    5. One thing we have going is that we sample the PD results not on the edges of our pulses, rather when the signal is stable
    and not switching.


    Really interested in #3.

    ty

  • 0 in reply to Daniel Najemy
    Prodigy 10 points

    This might be helpful - 3 layer stackup, note: no GND plane anywhere.

          Top                             Middle                      Bottom

  • 0 in reply to Daniel Najemy
    Prodigy 10 points

    Hi Ryan,

    Any advice on the above ?'s and layout?

  • 0 in reply to Daniel Najemy
    TI__Guru 66706 points

    Hi Daniel,

    Thanks for sharing the three separate layer plots. The layout is more clear now. 

    Is it possible to shift the GND via to the right and move the two PD lines further from the TX line?

    Do you have access to our EVM layout files under your MySecure resources page? There you can view our complete EVM design file. We do use a ground pour to fill in the top, bottom, and third layers. The second layer is entirely dedicated to GND.

    If you need to request access, please use the following link: https://www.ti.com/drr/opn/AFE44I3X-DESIGN

    Regards,

    Ryan