[SI-LIST] : differential pairs

bknott@wlgore.com
Wed, 17 Sep 1997 17:12:30 -0600

The broad side coupling is how I've always done differential pairs. The
conductor losses are lower because the whole line width (.008) is coupling
energy not just the trace thickness ( .0005"). More copper equals less
loss. The impedance variation due to the core thickness changes is a much
smaller problem than the attenuation encountered with the edge coupling.
Also make sure you account for the impedance due to the coupling to the
shield layers as shown on the sketch. The shield coupling could dominate
the differential impedance - two time the differential thickness is safe.
---------------------- Forwarded by Brad D Knott/WLGORE on 09/17/97 04:42
PM ---------------------------

jmueller@ca.newbridge.com on 09/17/97 12:43:43 PM

To: si-list@silab.Eng.Sun.COM
cc: jmueller@thor.ca.newbridge.com, andrew_roberts@qmail.ca.newbridge.com
(bcc: Brad D Knott/WLGORE)
Subject: [SI-LIST] : differential pairs

Hi,
I would like to get some opinions on how to layout differential pairs for
high speed (1 Gbit/s) data signals. I am not quite sure if it is preferable
to use broad side or edge side coupled differential pairs?
For data signals at that speed loss effects needs to be considered too.
What
is a minimum trace width to avoid significant skin effect losses?
edge side coupled: broad side coupled:
----------------------- -------------------------
--- --- ---
---------------------- ---
-------------------------

Advantages for the broad side coupled pairs:
============================================
Transmission line delay deviations between both lines can be controlled
better
with a broad side coupled pair.
I assume that the broad side coupled pair is also a better field guide
because
the field lines are more straight and better contained between the
conductive
bodies as it is the case for the edge side coupled pair.

Advantages for the edge side coupled pairs:
===========================================
The differential impedance of the edge side coupled pair can be designed
independend from the characteristic impedance of the cross section.
The distance between two traces might can be controlled more accurate than
the thickness of the core material?

All comments are appreciated
Joachim

--
+---------------------------------------------------------------------+
|  voice:  +1 613 591 3600       |  Joachim Mueller, EMC-Approvals    |
|  fax  :  +1 613 599 3654       |   600 March Rd. P.O. Box 13600     |
| e-mail: jmueller@newbridge.com |  Kanata, Ontario, (CA)  K2K 2E6    |
+---------------------------------------------------------------------+
|                  Newbridge Networks Corporation                     |
|                                                                     |
|                                __o                                  |
|                              _`\<,_                                 |
|                             (_)/ (_)                                |
|                            ~~~~~~~~~~                               |
+---------------------------------------------------------------------+