Doug, I'm not an SI expert on the level of you and others, but I think you're right. I think much of the differential-mode current return during transitions will be from one trace, across the ground plane, and back the other (anitphase) trace to the differential driver. Since there is generally much more coupling trace-to-ground than there is trace-to-trace in a typical "edge-coupled" differential pair, I would think that this cross-plane coupling would account for most of the differential coupling within the pair. And as you mention, there would be no differential-mode return current through the ground plane back to the driver chip, because to the extent that the complimentary signals are perfectly balanced (the definition of differential mode), there will be no return current to any other pin on the driver chip. If the above is all true, then a differential pair crossing a plane split, even without benefit of any vias to tie the plane edges back to a common (unsplit) plane, would not suffer any significant degradation: The plane current, being nearly orthogonal to the traces, would not even "see" the split. Also, if the above is true, the total coupling between the traces in an edge-coupled differential pair would not be strongly affected by trace spacing, provided the trace widths were much greater than the trace thickness, and the trace spacing was not small compared with the dielectric thickness from trace to plane. This is because most of the overall coupling would not really be "edge-coupled" at all, but rather coupled broadside from trace, to plane, to trace. Common mode current is of course quite another matter. Whatever unbalance current there is at the differential driver would have to return longitudinally back to the driver, and crossing a loosely-coupled split with that return current would have to cause EMI and ground bounce. Ken Hayden Doug Brooks wrote: > I would like the group's opinion on something. > > One view is that the currents on the plane are "return" currents, as they > would be if we treated the traces simply as a pair of microstrip traces. > But what if we don't treat the differential pair as simply two microstrip > traces? After all, microstrip traces have their return currents leaving > (for example) the ground pin of the device and then traveling on the plane. > With a differential pair, there is no current that flows through the ground > pin. > > What if we took the view that the currents on the plane are NOT return > currents. What if we believed that the *return* current is on the other (-) > trace and the currents on the plane are circulating, coupled, eddy currents > (formed by something analogous to transformer action). The further apart > the traces are, the further apart the current loops are. And if we move the > plane very far away, the main signal current loop doesn't change, but the > coupled signal on the plane gets smaller (because the coupling gets weaker > with distance.) This view also explains why differential circuits can > perform without planes and with planar discontinuities. > > The distinction between these two views is *very* significant, because the > *current loop* definition (relevant for EMI) turns out to be very different > depending on which view you consider to be correct. > > The situation is actually even more complicated than this, because there is > a combination of effects at work here. If we had a *perfect,* very high > frequency square wave, the signal during the transitions would couple to > the plane through the "distributed caps" along the transmission line. But > during the "flat" parts of the waveform, the return signal would not couple > to the plane (the "distributed caps" would be fully charged) but would be > on the other trace. > > Doug Brooks > > At 10:55 PM 10/14/03 -0700, Charles Grasso wrote: > >Referring to Eric Bogatins book (SI simplified) > >there are nice pictures that show the current distribution > >in a pair of coupled transmission lines over a ground plane. > >Fundemetally, when the coupling between the signal line > >and the return plane is much larger than the the coupling > >to the adjacent signal, there are seperate and distinct > >return currents with very little overlap in the plane. > > > >(to paraphrase from the book) > > ------------------------------------------------------------------ > To unsubscribe from si-list: > si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field > > or to administer your membership from a web page, go to: > //www.freelists.org/webpage/si-list > > For help: > si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field > > List archives are viewable at: > //www.freelists.org/archives/si-list > or at our remote archives: > http://groups.yahoo.com/group/si-list/messages > Old (prior to June 6, 2001) list archives are viewable at: > http://www.qsl.net/wb6tpu > ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu