DP83867IR: Connection to MDI interface

Can you elaborate on this?

We have a combination of transformer'ed RJ45 links and a direct connect to a switch port that is transformer-less on a newly-arrived prototype board design. We erroneously omitted the AC coupling/DC block capacitors that we intended.

We believe that we saw that link come up and pass traffic at one point, but it quickly drops, presumably, due to drift. I searched the forum, and came across this thread, which makes me nervous about reworking in the capacitors.

Should we expect that transformer-less, capacitive coupling should not work? Under what circumstance could capacitive coupling be made to work?

We have a DP83867IR that runs across a high-speed connector system from a daughtercard down to the main board, where a switch exists. fixThe total length of the differential pairs is only on the order of a few inches and are 100Ω characteristic impedance throughout. We wish for this link to run at 1Gbps, either through auto-negotiation, or fixed, if that helps our cause.

Do you expect that our problem is the transformer-less direct connection, or that our "cable" is shorter than 1 meter? (Or a combination of each and/or something else entirely?) We do have a 12" unshielded twisted pair cable that succeeds from the transformer'ed RJ45 to another switch port (with its RJ45 and magnetics).

Any suggestions would be greatly appreciated. (...including a suitable transformer suggestion, should capacitive coupling be a complete dead-end, which we hope is not the case.) Thanks very much.

Well, for a small sample set, we have the Ethernets working (at room temperature, etc.) with a direct connect. (That is, no transformer, and no AC coupling.) So the heat is off in the short-term, but I'm not overly confident in our stability, reliability and longevity in the longer term.

(We had a mechanical issue with our mezzanine connectors not being fully seated into the PCB, press-fit on one side, and soldercharge on the other, BOTH being improperly mounted.)

Thank you, Geet.

I imagine the AC coupling might not be needed if both ends of the link are in the same system sharing the same ground potential, which explains why we're having success with it so far.  Interesting that TI chooses not to test this scenario.  But I understand.  Fair enough.  So we'll need to discuss our application internally, and see if this is a variable we can tolerate, versus adding a transformer for an application that doesn't really need isolation.

Many thanks,

Matt