[SI-LIST] Re: Remove Ground underneath Differential signal is deserved or not?
- From: jeff_latourrette@xxxxxxxxxxx
- To: xiaoning.ye@xxxxxxxxx
- Date: Thu, 1 Aug 2002 11:27:50 -0600
Xiaoning Ye:
My analysis covered B vs C, and for microstrip lines close together I found a 6
dB improvement using guard trace, while for lines farther apart, the
improvement with ground stitched guard traces dropped to 3 dB. I haven't any
experimental data to back this up or a compete intuitive understanding of why
it would drop to 3 dB. I did also do some A vs C, no guard traces, where I
varied the spacing in order to quantify the coupling vs spacing.
I definitely agree that an unstitched guard trace would aid instead of
defeating cross-talk and some simple coupled line analysis on ADS or other
simulator tool (don't even need 3D or 2.5D) would show that, I'm sure. I
didn't simulate it, but I think that merely stitching guard traces on the ends
isn't good enough, unless line is very short (electrical length, compared to a
wavelength).
Details:
As far as details, I did a three-line, six-port simulation, with and without
the guard trace, maintaining the spacing of signal lines. I put two traces,
tightly coupled as a (100-ohm) differential pair and then a third 50-ohm trace
was the target trace, spaced a little farther away. I looked at the amount of
signal on the 3rd trace ports vs amount of signal launched at either of the
differential lines. Then I repeated it with a grounded guard trace in between.
Width of guard trace was between that of a single differential trace and the
50-ohm trace. I spaced 6 vias evenly, but since they were spaced at lambda/20,
any resonances were well beyond my band of interest and beyond where I
simulated. As stated before, my line was 1/4-wave at 6 GHz and I simulated
0-10 GHz, using ADS (3-line only) and Momentum (3-line and 3-line with guard).
Remaining Questions:
I haven't yet been able to figure out how to simulate a true differential,
although Momentum does allow you to define differential pairs. I would be
interested if people feel there is a significant difference between a
single-ended and differential simulation for this problem. In other words, if
I launch a differential signal and my lines are tightly coupled so that I'm
getting some differential behavior, does this significantly reduce the amount
of cross-talk coupled over to my target trace ? I understand that if you had
true differential (which you can never get with microstrip), that far away from
the pair, radiation/coupling should cancel. I think this effect would have to
be reduced as your target trace gets very close to the pair.
In any case, I'd like to re-run this problem using a differential drive and
compare results.
Thanks for sharing your experience-I'd be interested to hear any other results
or any suggestions.
-----Original Message-----
From: Ye, Xiaoning [mailto:xiaoning.ye@xxxxxxxxx]
Sent: Wednesday, July 31, 2002 5:43 PM
To: 'jeff_latourrette@xxxxxxxxxxx'
Cc: si-list@xxxxxxxxxxxxx
Subject: RE: [SI-LIST] Re: Remove Ground underneath Differential signal
is deserved or not?
Jeff and all,
I did some simulation (3-D fullwave) quite a while ago and my results show
that guard trace helps little for microstrip line. It seems there are
distinct oppinions on the conclusion, but first I would like to make sure
that all of us are making apple-to-apple comparisons. Suppose we have the
following three configurations (each line represents a trace):
Configuration A: _ _ (no guard trace, agressor and victum closely
spaced)
Configuration B: _ ___ _ (guard trace in between, and stitched to
ground)
Configuration C: _ _ (no guard trace, but the aggressor and the
victum keep the same spacing as Config B)
My results show that:
* there is significant reduction of xtalk from A to B
* there is only minor (~2dB) improvement from C to B (even when the guard
trace is completely shorted to the ground by a vertical plane). On the
other hand, if the guard trace is not stitched to ground by vias by close
spacing, Config B can be worse than Config C. For example, if the stitch
spacing is 1", then we will see a resonance at around 3 GHz (which
correspond to half wavelength of 1").
In summary, the guard trace appears to reduce xtalk because by "squeezing"
guard trace between the aggressor and the victum, the spacing between
aggressor and victum becomes larger. It is the "larger spacing", not the
"extra shielding" that does the work.
If any of you have the results showing that Config B is much better than
Config C (for microstrip line), can you share with me a little bit more
detail of your configuration?
Thanks,
Xiaoning Ye
Intel Corp.
-----Original Message-----
From: jeff_latourrette@xxxxxxxxxxx [mailto:jeff_latourrette@xxxxxxxxxxx]
Sent: Wednesday, July 31, 2002 11:20 AM
To: inovak@xxxxxxxxxxxxxxxxxxx; leeritchey@xxxxxxxxxxxxx
Cc: scott@xxxxxxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Remove Ground underneath Differential signal is
deserved or not?
All:
I have to agree with Istvan's numbers-I've run analysis of cross-talk
(coupling) between microstrip lines (1/4 wavelength at 6 GHz) with and
without a via grounded guard trace using Agilent Momentum (2.5D solver) out
to 10 GHz. My finding was around 6 dB improvement for tightly coupled lines
(line-to-line spacing=around a board thickness).
I also saw the improvement drop to just 3 dB for loosely coupled lines
(line-to-line spacing>2 board thickness), however this behavior should be
much more dependent on the relative spacing of boundaries (walls) set-up for
the problem. If in fact a correct result, this seems to support the finding
that if you have decent spacing to begin with, guard traces usually don't
add much benefit. I expect that the accuracy of measuring very loose
coupling decreases substantially as spacing is increased, so that
experimentally, this could be hard to verify conclusively.
Anyone else have similar or different findings ??
Regards,
Jeff LaT.
-----Original Message-----
From: Istvan Novak - Board Design Technology
[mailto:inovak@xxxxxxxxxxxxxxxxxxx]
Sent: Wednesday, July 31, 2002 8:54 AM
To: leeritchey@xxxxxxxxxxxxx
Cc: scott@xxxxxxxxxxxxx; si-list@xxxxxxxxxxxxx;
inovak@xxxxxxxxxxxxxxxxxxx
Subject: [SI-LIST] Re: Remove Ground underneath Differential signal is
deserved or not?
Lee,
I think the source of confusion is that if a field solver is used to extract
the parameters of
the coupled traces, those extracted parameters will not directly show what
happens when you use
the shield traces properly connected. You are correct in saying that the
coupling capacitance
and inductance is primarily dictated by the relative separation, and the
presence of trace(s)
in between has minimal (but not zero) affect on the coupling parameters.
Having said this, it is also true, as pointed out by others, that shield
traces CAN reduce the
crosstalk beyond what we would get just from spreading the traces to make
room for the shield
traces. Below the half-wave resonance of shorted shield-trace segments,
there is a minimum of
about 6dB extra reduction of crosstalk. You get this approximately 6dB
improvement on
microstrip, with the shield traces having the same geometry as the signal
traces. If you have
a wider shield trace, the EXTRA crosstalk reduction is more. In stripline,
the numbers are
different, but show the same trends.
This extra crosstalk reduction of shield traces can be simulated and
measured, but in terms of
simulations, we have to go beyond the field-solver excercise and have to
simulate the real
scenario: hook up a source and make sure that the shield trace is connected
to ground.
It is obvious that on a very wide bus, and using fast edges, the many
stitching vias create a
real routing problem, so spreading the traces a little bit further is a
better choice. For a
few sensitive signals, though, shield traces can offer real benefits.
Regards
Istvan
Delivered-To: si-list@xxxxxxxxxxxxx
Date: Wed, 31 Jul 2002 07:56:50 -0700
From: Ritchey Lee <leeritchey@xxxxxxxxxxxxx>
MIME-Version: 1.0
To: scott@xxxxxxxxxxxxx
Cc: silist <si-list@xxxxxxxxxxxxx>
Subject: [SI-LIST] Re: Remove Ground underneath Differential signal
is deserved or not?
Content-Transfer-Encoding: 8bit
X-archive-position: 3574
X-ecartis-version: Ecartis v1.0.0
X-original-sender: leeritchey@xxxxxxxxxxxxx
X-list: si-list
Scott,
If you just space the two traces in question the same as would be
required to
make room for the guard traces, the cross talk will be the same as
with the
guard traces and you don't need to add those extra structures. This
is
relatively easy to show with a field solver.
In the bargain, there is no risk of creating unwanted resonant
strucutres.
Lee
Scott McMorrow wrote:
> Lee,
>
> > What I show in my class is that guard traces are always LC
networks that
> > resonate at some frequency and can and do turn into bandpass
filters
> > at some
> > frequencies. I demonstrate this will actual failed circuits.
> >
> >
> This is exactly why Mike Conn and myself recommend stitching guard
> traces to ground at random intervals across the length. The
average
> spacing of the via sites should be 1/10 of the wavelength of the
fast
> signal frequency or equivalent risetime that will be present on
this
> trace. These stitch vias to ground will eliminate the problems
which
> you have alluded to. They are backed by extremely good analytical
and
> experimental science.
>
> Best regards,
>
> Scott
>
> --
> Scott McMorrow
> Teraspeed Consulting Group LLC
> 2926 SE Yamhill St.
> Portland, OR 97214
> (503) 239-5536
> http://www.teraspeed.com
>
> --
> Scott McMorrow
> Teraspeed Consulting Group LLC
> 2926 SE Yamhill St.
> Portland, OR 97214
> (503) 239-5536
> http://www.teraspeed.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
-- Binary/unsupported file stripped by Ecartis --
-- Type: text/x-vcard
-- File: vcard.vcf
-- Desc: Card for leeritchey@xxxxxxxxxxxxx
------------------------------------------------------------------
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
Istvan Novak Sun Microsystems, Inc.
Istvan.Novak@xxxxxxx Workgroup Servers, BDT Group,
One Network Drive, Burlington, MA 01803
Phone: (781) 442 0340
------------------------------------------------------------------
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:
