3
\$\begingroup\$

More specifically, does it matter on a PCB if routing differential pairs if they cross reference planes?
Do they need a reference plane?
Is there even a reference plane with a differential pair?

I don't do many high speed designs, when I have in the past, I make the routing as short as possible between high speed connectors and PHY's so I don't have to worry as much about matching, transmission line effects, ect. I have also used a continuous ground plane in all cases. But now I'm wondering...

My intuition tells me that a continuous ground plane in the case of a differential pair is not a big deal because there is little or no return current, so the only thing that matters is ensuring that the lines are matched.

A reason not to would be cross coupling of noise through capacitance into the differential pair, but noise will appear equally on both pairs (if the capacitance is the same on both pairs) and subtracted out.

\$\endgroup\$
  • \$\begingroup\$ Can you back up an answer with some examples? \$\endgroup\$ – Voltage Spike Jun 4 '19 at 19:46
  • \$\begingroup\$ lack of a continuous reference-plane will inject common-mode trash into the differential signals; this will, due to limited high-frequency CMRR in the differential receiver and the +/- package (consider BGA with mismatched routing), degrade the data-eye; for certain jitter and noise components, the extra trash may be significant. Would 20dB SNR (maybe BER of 1e-20) degrading to 15dB SNR (maybe BER of 1e-6) be a big deal ? \$\endgroup\$ – analogsystemsrf Jun 5 '19 at 3:59
3
\$\begingroup\$

A reference plane is not a requirement but most designs use one. Normally some of the return current will flow on the reference plane depending on how closely the differential pair is spaced. The reference plane also plays a role in the impedance of the differential pair. Achieving 100 ohms differential impedance on a PCB might be difficult without one. Reference planes provide other benefits like EMI reduction and isolation from other signals.

\$\endgroup\$
2
\$\begingroup\$

Sounds right to me for the most part. If there is a very strong aggressor causing voltage fluctuation between the two reference planes, it could be an issue depending on the frequency relation to your signal. It depends on what level of spurious performance you need. I would honestly not worry about it much myself.

The impedance calculation of a differential pair does depend on the distance to the reference plane (only weakly if the distance is much greater than the distance between the pair), so you probably wouldn't want to transition halfway through the route between having a close reference plane and not having one.

\$\endgroup\$

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.