Re: [SI-LIST] : Differential pairs and place splits

Neven Pischl (Neven_Pischl@BayNetworks.COM)
Tue, 10 Feb 1998 10:43:47 -0800

At 09:50 AM 2/10/98 -0500, Young, Randy wrote:
>I understand that in general, routing high-speed digital traces across
>plane splits is a bad idea from both a signal quality perspective and an
>EMI perspective due to return current paths. My question is concerning
>differential pairs - specifically, USB and 1394. Do the same rules
>apply? Is routing a differential pair across a plane split bad from
>either signal quality or EMI perspective? My assumption is that the
>return current path for a differential pair is the other trace of the
>pair, and that this would greatly reduce any signal quality or EMI
>problems associated with crossing splits. Is this correct?
>Randy Young


Differential pairs take care of differential noise (or signal) return, if
they are really closely coupled and symmetrical, hence you are in principle
correct. However, they always carry some common-mode (CMM) noise, as any
other traces, level of which depends on many factors, but it is primarily
due to the CMM noise on the chip at which the pairs originate. You do not
want that noise to cross the moat, and that is more for EMI than for SI.
For instance, a CMM choke across the moat or a pair of capacitors from each
trace to ground (same one on which the chip sits), placed near the chip,
may do the job. The pairs can pick up significant amount of CMM noise if
they are routed next to power planes. Reference them to the ground plane,
especially if they leave the board.

Neven Pischl
Bay Networks
408 495 3261