[SI-LIST] Re: Differential trace route question
- From: christopher.heard@xxxxxxxxxxxx
- To: levinpa@xxxxxxxxxxxxx
- Date: Sun, 20 Oct 2002 20:14:43 -0400
This is a classic signal integrity question, one that could extend to be a
very long thread.
Some will say this is exactly the reason why you should route broadside
coupled striplines. It keeps the lengths the same for traces that have a
lot of changes in direction. However broadside coupling drives thickness
of boards higher and aggravates problems with plated through holes at high
frequencies. There is always a debate as to whether broadside coupled
lines have more or less loss at high frequencies, there are equal amounts
of people who believe both sides of the argument.
Backplanes that I have designed have been "uncoupled" to solve this
problem. It allows me to add "trapezoidal jogs" to the trace without
changing differential impedance...I get perfect length matching and no
Zdiff change. The problem is the trace density is not especially high
which drives layer count and eventually thickness. Then there is the issue
of routing through connectors. If you start with an uncoupled transmission
line, you must deal with some coupling as you route through the pin field.
Not an open and closed case for sure,
Chris
**********************
Chris Heard
Teradyne
Work: 603-879-1031
Cell: 508-277-5780
Email: christopher.heard@xxxxxxxxxxxx
Paul Levin
<levinpa@earthlink To: James_R_Jones@xxxxxxxx
.net> cc: doug@xxxxxxxxxx,
si-list@xxxxxxxxxxxxx
Sent by: Subject: [SI-LIST] Re:
Differential trace route question
si-list-bounce@fre
elists.org
10/20/2002 03:34
PM
Please respond to
levinpa
Dear All,
In trying to decide how much skew you can tolerate, consider this:
To the extent that you have un-matched the pair, you have converted
part of your differential signal into common-mode, which your
differential receiver can probably tolerate. Conversely, you have
converted part of your common-mode signal, whether due to transmitter
mismatch *or to external pickup*, into differential signal which is
now indistinguishable from the signal you are trying to receive.
This is the crux of the matter.
I hope that this insight helps.
Regards,
Paul Levin
Senior Principal Engineer
Logic Innovations, a Xyratex company
_____________________________
James_R_Jones@xxxxxxxx wrote:
> Doug,
>
> I might think about using a loosely coupled differential pair. Say you
were
> trying to achieve 100 ohms differential, I might use two 50 ohm single
ended
> lines such that there was little coupling between the pair. This would
give
> you nearly 100 ohms differential. You can keep the traces close, but not
so
> close that there is coupling. This way, you still get some benefit of
> common mode rejection, but you can insert 'pimples' or 'snakes' without
> worrying about the trace spacing affecting differential impedance. Also,
> you may experience less loss with this type of setup due to increased
amount
> of 'skin'.
>
> Another thing that I might think about is the amount of skew that you are
> adding by turning your diff pair 180 degrees. If you calculate the
length
> difference, how much length mismatch is there really? What is the amount
of
> propagation latency that this introduces? Is this really a problem?
>
> If it is a problem, then I might consider making sure that the
complementary
> signals remain in phase along the length of the trace. This could be
> achieved by the dicretionary placement of 'pimples'.
>
> I have found that for the digital differential frequencies that I am
> concerned with that none of the above really matters all that much. But
I
> also realized that some designs are running much faster that what I am
> dealing with.
>
> James R. Jones
> Dell
>
>
>
> -----Original Message-----
> From: Doug Brooks [mailto:doug@xxxxxxxxxx]
> Sent: Friday, October 18, 2002 6:52 PM
> To: si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Differential trace route question
>
>
>
> Hi all,
>
> This question came up in a seminar and I'd be interested in your inputs.
>
> Assume I have a differential pair extending "North" from a driver for a
> ways. The pair turns "East" for a ways and then "South" for a ways before
> terminating at a receiver. The inner trace is shorter than the outer
trace.
>
> I am interested in controlling skew, so I want to equalize the trace
> length. I am also interested in controlling the differential impedance,
so
> I want to keep constant trace spacing. These requirements are mutually
> exclusive! I will compromise what I have to on impedance to obtain equal
> trace lengths.
>
> I visualize that I have three options for increasing the length of the
> inner trace.
>
> 1. I can "snake" the extra length of the inner trace with the constraint
> that the added length must be added as close as possible to the receiver.
>
> 2. I can "snake" the extra length of the inner trace without that
> constraint. That is, the snaking can take place anywhere that is
convenient
> along the trace.
>
> 3. I can add randomly placed little "pimples" along the inner trace (sort
> of like __/\__ ) (pointing away from the outer trace) each one of which
> would be small and insignificant but collectively they would add up to
the
> total length I need to add.
>
> Does anyone have any strong opinions why any one of these is better than
> any other, or why there is an even better alternative?
>
> Doug Brooks
>
>
>
____________________________________________________________________________
> ___
> UltraCAD Design announces availability of its new book "Signal Integrity
> Issues in PCB Design"
> Details at www.ultracad.com
>
>
> ------------------------------------------------------------------
> 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
>
>
>
------------------------------------------------------------------
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
Other related posts:
