I was reading the datasheet for an Ethernet interface layout from NXP and I found they didn't use a reference plane in the region where they routed the differential pairs. I couldn't understand why they do that and I didn't find an answer. The datasheet explains it that they wanted to reduce the capacitance but how does the signal couple? It couples only with each other?

enter image description here

Datasheet AN2759 page 16

  • \$\begingroup\$ The note says that it's done because the transformer requires it, and since this is only fast Ethernet maybe that is ok. In general that board layout, which routes a number of traces over a huge ground split makes me wary. \$\endgroup\$ – user1850479 May 15 at 14:47
  • 1
    \$\begingroup\$ It couples only with each other? - do you have a problem with that concept? \$\endgroup\$ – Andy aka May 15 at 14:55
  • \$\begingroup\$ @user1850479 So what about unshielded CAT cables, they carry Ethernet 100m while routing the signals over a huge ground split? \$\endgroup\$ – Justme May 15 at 16:17
  • \$\begingroup\$ @Andy aka no not realy… just asking if this is what happens \$\endgroup\$ – arisk4 May 16 at 15:32
  • \$\begingroup\$ @use1850479 yeah true gnd split is troublesome… but i think if the gnd planes are the same it is accepted. \$\endgroup\$ – arisk4 May 16 at 15:33

? It couples only with each other?

Capacitance is computed by a product of the inverse gap and trace length, so increasing the gnd plane gap reduces common mode C, while differential mode C exists with it’s gap and length to each other.

With a gnd pour, the common mode C coupling per cm = 1 to 2 pF/cm (ballpark) when copper poured with a gap same or smaller than Trace width.

  • consult Saturn PCB design.exe as this is a complex function.

By raising the CM impedance together with the PHY Balun transformer then there is less ingress from CM to DM from differences in the CM capacitance.

So by reducing Ccm the tolerances of mismatch have less impact on ingress by impedance ratios of error in CM impedance to trace DM impedance. However this also compromises EMI egress so the traces ought to be matched impedance over the spectrum of signal either 100MHz or 1GHz depending on speed.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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