Tool/software:
Below is the SGMII signal at the input to the DP83867IS.
At first glance, this looks like a reflection. However, our trace lengths have a transit time of approximately 240ps, so a reflection would take 480ps to make a round trip. The "reflection" seen here is too fast for that. A reasonable argument is that there could be some other feature in the PCB traces that would correspond to the timing seen here. However, on the same circuit board, we have two of these PHYs with differing traces and placement of vias, yet when I overlay the signal to one of the PHYs with the signal to the other PHY, I see this:
The identical timing of this extra wiggle implies that a common feature in the circuit for each PHY is the cause.
This is a representation of the traces, with dimensions and calculated impedances of traces and vias. Each color represents a different layer of the board. The ethernet switch is represented by the purple rectangle at the bottom of the image:
The only common feature between the two signal paths is the vias next to the decoupling capacitors. However, a reflection at that location would occur much faster than what we observe.
Since our line impedances and the terminating resistors of the DP83867 match well, this would also tend to argue against reflections. Could it potentially be a ringing? If so, is there some impedance at the pins of the PHY that we could account for?
This signal is strikingly similar to signals seen in these prior two forum issues, neither of which identified a final solution before being closed:
https://e2e.ti.com/support/interface-group/interface/f/interface-forum/1030982/dp83867e-sgmii-reflections
https://e2e.ti.com/support/interface-group/interface/f/interface-forum/1009804/dp83867is-sgmii-eye-diagram-issue/3755302?tisearch=e2e-sitesearch&keymatch=dp83867%252525252520%252525252526%252525252526%252525252520eye#3755302
The signals from the PHYs back to the switch do not exhibit this ringing/reflection
