[SI-LIST] Re: Common-mode return path for differential signals.

  • From: Ihsan Erdin <erdinih@xxxxxxxxx>
  • To: michael.mirmak@xxxxxxxxx
  • Date: Fri, 28 Oct 2005 08:51:47 -0400

Although the term "differential impedance" is very commonly used in SI
circles, it may lead to some confusion. One can speak about "common mode" or
"differential mode" currents but technically even or odd mode impedance
related to them.
The common mode currents are simply the noise in the system and should be
eliminated not only for the noise immunity of the receiver but also to cut
down on radiated emissions from the system. The best way to accomplish this
goal is a symmetrical design, hence the main reason behind the differential
routing. If you target a "differential impedance" of -say 100 ohms, you may
route two 50 ohm single ended signals with the same bends and vias,
separated wide apart and that will achieve the goal as far as the impedance
requirements are concerned. But you have to provide a symmetric environment
to each element so that the noise not be unbalanced on them, which is very
difficult if not impossible in a complex PCB environment. By placing them
very close to each other, on the other hand, you'll approach the goal of
symmetrical routing with almost the same amount of common mode noise on each
element of the pair which will be eliminated by the CMRR of the receiver.
Otherwise, the noise immunity of the differential signaling may not be very
much different from that of a single-ended configuration.
From the routing perspective, the close proximity of the P and N elements of
a pair -when combined with the need to place them at a certain distance to
nearby signals- is always advantageous for real estate purposes. This is
more so for wide busses. Since the spacing has a secondary effect on the odd
mode impedance whence the "diff. impedance" is derived placing them closer
doesn't have much of an impact on the impedance. So, it's difficult to
justify not routing them as close as possible, except maybe the yield
concerns.
Regards.

Ihsan Erdin
Nortel

 On 10/26/05, Mirmak, Michael <michael.mirmak@xxxxxxxxx > wrote:
>
> Doug,
>
> I've seen a lot of debate on this lately. From what I can tell, there
> appears to be some misunderstanding about what the meaning of
> "differential" is in the context of system design.
>
> For most of the microstrip PC boards I've seen, the spacing between the
> nets of differential pairs never gets close enough in relation to the
> height above the dielectric to result in more than 5-10% EM coupling.
> The traces are effectively single-ended, as you noted for your
> situation. Routing these signals as a "differential pair" essentially
> means, in this context, routing the traces to (a) minimize the length
> mismatches and therefore timing skews, (b) equalize the effects of noise
> from surrounding signals and (c) maintaining the same reference plane
> for both. It does *not* appear to mean actually having strong EM fields
> between the traces. To gain the benefits of truly "differential" routes
> in an EM coupling sense, you would have to have something like a twisted
> pair, where one line really is the reference and return path for the
> other line and no other return path is nearby. Most high-volume PC
> board designs can't ever approach that kind of strong coupling.=20
>
> In general, the benefits for "differential" system design appear to be
> more oriented toward the receiver than the board itself (again, for the
> microstrip PC designs I commonly see). For a single-ended system, the
> receiver itself is likely a diff amp with its reference tied to a local
> voltage divider or other reference source connected to local
> power/ground supplies. Noise on the reference that doesn't occur on the
> incoming signal line will result in erroneous data going into the
> receiver logic core.
>
For a "differential" system, the receiver diff amp
> uses, as its reference, the signal on the inverted line of the diff
> pair. If this line was routed in nearly the same fashion as the line
> carrying the data signal, common-mode rejection will take care of the
> noise that affects both lines, resulting in cleaner data into the
> receiver.
>
Thus you can have a very good "differential" (noise immune) system
> operating using "differential" (highly parallel with clean referencing)
> routes on the board, while still having *effectively no coupling*
> between the trace pairs. The meaning of "the d word" changes depending
> on the context, making it very much misused in the industry today.
>
> For an example of such a system in action, see USB 1.1.
>
> My two cents; I hope it helps!
>
> - Michael Mirmak
>
> -----Original Message-----
> From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]
> On Behalf Of Doug Hopperstad
> Sent: Tuesday, October 25, 2005 8:23 PM
> To: si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Re: Common-mode return path for differential signals.
>
>
> When routing 100-ohm differential nets is there advantage to keeping the
> =3D
> two nets closely coupled other than for crosstalk issues? For current =
> =3D
> designs the space between the two nets is set for 1.5 times the width of
> =3D
> the net. For a 8mil trace, the gap is 12mils (20mil pitch). If the =3D
> differential pair needs to become uncoupled, the trace geometry is =3D
> adjusted to maintain the proper impedance. Length matching maintained =
> =3D
> for the route to reduce unwanted common-mode currents due to length =3D
> skew. All routing is done on balanced, striplines using ground reference
> =3D
> planes only.=3D20
>
> Some of the information I have read recently states to keep the =3D
> differential pair tightly coupled while others state that there is no =
> =3D
> major advantage other than crosstalk. With the 12mil gap design, the =3D
> nets are loosely coupled (8mil dielectric thickness) and has shown good
> =3D
> results.=3D20
>
> Any comments or feedback would be great,
> Thanks,
> Doug.
> ------------------------------------------------------------------
> 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 FAQ wiki page is located at:
> http://si-list.org/wiki/wiki.pl?Si-List_FAQ
>
> List technical documents are available at:
> http://www.si-list.org
>
> List archives are viewable at: =20
> //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
> =20
>
> ------------------------------------------------------------------
> 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 FAQ wiki page is located at:
> http://si-list.org/wiki/wiki.pl?Si-List_FAQ
>
> List technical documents are available at:
> http://www.si-list.org
>
> 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 FAQ wiki page is located at:
                http://si-list.org/wiki/wiki.pl?Si-List_FAQ

List technical documents are available at:
                http://www.si-list.org

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
  

Other related posts: