As I understand differential signaling, return current goes via the "-" (minus) line, so why we do we need to provide a reference ground plane for differential signals?

  • \$\begingroup\$ because voltage is a potential. You never have "minus" without a point to refer where it is minus to \$\endgroup\$
    – PlasmaHH
    May 11, 2016 at 9:27
  • \$\begingroup\$ What PlasmaHH says and imagine this: I provide you two wires containing a differential voltage of 1 Vpeak which is what you need. But I do not tell you that these signals are lifted +1000 V DC from your groundlevel. Would you still be happy ? No you would not as it will probably fry whatever you connect to my signals. You want that reference level to be around 0 V or maybe a couple of volts. Now you know why the reference level is important. \$\endgroup\$ May 11, 2016 at 9:35
  • \$\begingroup\$ Many differential signalling systems are not current loops. RS485 is an example. \$\endgroup\$
    – brhans
    May 11, 2016 at 12:49
  • \$\begingroup\$ @FakeMoustache Just want to understand: if the signals from the source are lifted +1000V from the GND level at the other end, doesnt this imply that there's somehow already a common reference between the 2 ends? After all, 1000V on one end is only 1000V on the other end if the GNDs are connected? So there should be no reason why the receiving end should get fried, since it has no way of "appreciating" the 1000V+ that its receiving? \$\endgroup\$ May 12, 2016 at 4:18
  • \$\begingroup\$ I guess because differential impedance is a function of characteristic impedance and for maintaining a fixed characteristic impedance you need a ground plane. \$\endgroup\$ May 17, 2017 at 9:16

2 Answers 2


More to the point the common edge coupled pairs that everyone uses on PCB layouts are not usually particularly tightly coupled (Think about the geometry, the reference plane is the same order of spacing as the two lines, and the plane has a dielectric in the gap!

They are a lot better then nothing, but the impedances are mainly defined by the coupling to and from the reference plane rather then the coupling between the lines, and line balance is mainly a common mode impedance thing, the reference plane helps here.

73 Dan.

  • \$\begingroup\$ I think this is the key point, personally. \$\endgroup\$
    – user57037
    May 12, 2016 at 3:27
  • \$\begingroup\$ Thanks Dan, So if I simplify what you say the reason for having reference plane is to reach specified impedance (100Ohm perhaps) easier. Now it make sense \$\endgroup\$
    – pazel1374
    Sep 21, 2018 at 8:18

Old fashioned telephone wires were just that - two wires and the circuitry of the phone was isolated from ground and also the user in the house. Never really a problem and, this is an example of a two wire differential system not having a "reference plane". I say it didn't need one but cross talk was a common problem and the good old post-office used to randomly swap the wires between consecutive telephone poles. A bit of a nightmare if you have something like this: -

enter image description here

This is basically the same in CAT 7 wiring - four pairs with different twist lengths and hence it lowers the possibility of cross-talk.

But, adding a screen (or reference ground) is also a good idea because it avoids E-field contamination and also "protects" the consistency of the cable's characteristic impedance from the influence of external objects. The screen maintains a good electric shield (Faraday cage) and keeps the "pair's" characteristic impedance fairly constant in widely differing environments.

So, for high speed data links using twisted pair, you need to terminate the wire at one end of the cable (or both) and if your characteristic impedance was NOT reasonably constant along the length you would get high speed reflections and possible data corruptions.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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