1
\$\begingroup\$

This question already has an answer here:

Problem:

We are designing a small payload which will contain several ethernet capable devices connected via a backplane we must design.

All guidelines concern the scenario where the PHY is connected to an RJ45 connector, but in my case I am connecting ethernet accross a backplane.

The Major Components in payload enclosure:

  • Embedded Gigabit Switch (Example: Embedded Switch) (includes Ethernet Magnetics)
  • 2x Embedded Computer & Expansion module (Example: COMe-mAL10 (includes Ethernet Magnetics)
  • Interface Board to connect to enclosure connectors. Industrial connectors with/ twisted cables)
  • Backplane connecting all the above boards

Note: Both the computers and the switch have integrated magnetic and are designed for 1000BASE-T so implementing 1000BASE-KX is not really an option. I realize that 1000BASE-KX is made for back-plane connection but the PHY transceiver needs to be modified to support that (and magnetic removed)

Usual PCB PHY to RJ45 Guidelines:

The usual guidlines are for PCB routing between the transceiver (PHY) and the magnetic, some common ones listed below:

  1. Signal after magnetics should be above chassis ground plane, no overlap with board ground plane.
  2. Signal should be routed as differential pairs (100Ohms controlled impedance)
  3. The use of vias is to be minimized. If vias are used, keep them to a minimum and always match vias so the differential pairs are balanced.
  4. ideally the magnetics are included in RJ45 connectors or that the RJ45 connector is very close to the magnetics.

Example Guideline Documents

LAN7500/LAN7500i Layout Guidelines (Section 5)

Gig PHYTER V 10/100/1000 Ethernet Physical Layer Design Guide

Similar Questions:

  1. Ethernet Pass Through over PCB

Answer: Small PCB section, wont have much effect, follow general layout guidelines.

  1. Are there any special internal Ethernet connectors other than RJ45?

Answer: Implement backplane ethernet (1000BASE-KX)

  1. How to route Gigabit Ethernet, remote PHY or remote RJ45 connector?

Answer: If the correct reference plane is under the signals the PHY<->Magnetics or Magnetics<->RJ45 then longer length can be run, else keep as short as possible.

  1. Ethernet Without Magnetics?

Answer: Use capacitors and implement 1000BASE-KX.

Questions:

  1. From all of the above my take home message is that I can successfully connect my different devices directly via back-plane using 1000BASE-T as long as I have the separate ground below each signal and the traces are of controlled impedance, is this correct?

    1. If the above is correct, The impedance controlled trace on the backplane will be run above a ground plane which is only capacitively coupled to the center tap of the magnetic at each end? I guess this is the equivalent of the shielding in shielded cables?

    2. Does anyone have a link to a document/guideline describing how 1000BASE-T can be implemented on backplanes?

\$\endgroup\$

marked as duplicate by Marcus Müller, winny, DoxyLover, Sparky256, Andy aka Jul 9 '18 at 13:12

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • \$\begingroup\$ Marcus Muller: Ethernet without magnetic will require me to remove the magnetic circuit and modify the PHY on my embedded computer and the switch. The answer says it is essentially implementing 1000BASE-CX which requires different transceiver or transceiver config. \$\endgroup\$ – Alexis Jun 27 '18 at 9:41
1
\$\begingroup\$

1) yes, you can. 1000BASE-T (MDI) is suitable for doing this. but keep in mind that:

  • you does not need a separate ground "under" each 1000BASE-T MDI (= a couple of four pairs matched), one plane "under" all of them may be enough. the main tip here is that plane must be tied to chassis and topologically separated from any local ground despite of such a ground is tied to chassis too.
  • the switch you referred to is a board level switch, "fanning" its 1000BASE-T MDIs out of the case directly (via passive MDI to "RJ" connectors) could be erroneous because of a possible lack of necessary isolation level (port-to-port, port-to-common_ground, and/or port-to-chassis) provided by the switch board itself.

2) UTP gives about 5-7 pF/m, STP gives about 50-70 pF/m. how much the board gives... i don't know and think you can calculate it yourself based certainly on your design. that will give you an answer about equivalence.

3) it is nearly the same as for 100BASE-TX MDI (the key word there is MDI) because both 100BASE-TX and 100BASE-T PHYs work over the same cabling - Cat.5 compatible mediums. yes, 1000BASE-T MDI has a stricter requirements - but slightly not dramatically.

\$\endgroup\$

Not the answer you're looking for? Browse other questions tagged or ask your own question.