From: bkemp@lexmark.com
Date: Tue Jun 20 2000 - 11:36:55 PDT
You may have missed my comments earlier because I left out the "Guard Traces" in
the subject line. Here it is again. I have done the simulations with the guard
traces as well as no guard traces with separations. I simply like to call them
ground or signal return traces, not guard traces.
I have gotten numbers using xfx that indicate guard traces do work. More
importantly, I have seen positive effects on real boards. It is not important
to via to ground every "X" intervals of distance. It is critical that the
"guard traces" are grounded at the source and receiver. It is also necessary
for the guard traces to via whenever the signal does. For instance, if you have
a clock trace that vias from layer 1 to layer 4 and is routed as
ground-signal-ground (GSG), the ground (guard) trace should via from layer 1 to
layer 4 very closely.
See below for some quick results. Compare numbers in bold for effects on
crosstalk.
Brandon Kemp
EMC/SI Engineer
Lexmark International, Inc.
Trace width = 6 mils
Edge to edge separation = 6 mils
Height above ground plane = 7 mils
Er = 4.5
1.
S S
___ ___
___________________________GP
Signals separated by 6 mil
Configuration Name: SS_GP_7mil Conductors: 2
#INTEGRAL solution#
Conductor index: 0 name: $$GND$$
Conductor index: 1 name: 1
Conductor index: 2 name: 2
i j Lij Cij Ze Zo Se So Fwdx Rvsx
from to (nh/in) (pf/in) (ohms) (ohms) (ns/ft)(ns/ft) (s/s) (v/v)
--------------------------------------------------------------------------
1 1 9.791 2.176 68.57 - 1.71 - - -
1 2 2.735 0.334 82.47 53.02 1.82 1.60 0.132 0.217
2 2 9.791 2.176 68.56 - 1.71 - - -
;
2.
S S
___ ___
_____________________________GP
Signals separated by 18 mil
Configuration Name: S_S_GP_7mil Conductors: 2
#INTEGRAL solution#
Conductor index: 0 name: $$GND$$
Conductor index: 1 name: 1
Conductor index: 2 name: 2
i j Lij Cij Ze Zo Se So Fwdx Rvsx
from to (nh/in) (pf/in) (ohms) (ohms) (ns/ft)(ns/ft) (s/s) (v/v)
--------------------------------------------------------------------------
1 1 9.842 2.117 68.27 - 1.73 - - -
1 2 0.928 0.061 72.37 63.97 1.79 1.67 0.066 0.062
2 2 9.842 2.117 68.28 - 1.73 - - -
;
3.
G S G S
___ ___ ___ ___
_____________________________GP
Signals separated by 18 mils, one trace routed GSG
Configuration Name: GSGS_GP_7mil Conductors: 2
#INTEGRAL solution#
Conductor index: 0 name: $$GND$$
Conductor index: 1 name: 1
Conductor index: 2 name: 2
XFX V6.5.0 Report 16 Jun 13:55 2000 Setup File =
F:\bkemp\xfx\Scarecrow\Clocks\clks.xfx
i j Lij Cij Ze Zo Se So Fwdx Rvsx
from to (nh/in) (pf/in) (ohms) (ohms) (ns/ft)(ns/ft) (s/s) (v/v)
--------------------------------------------------------------------------
1 1 8.206 2.253 60.35 - 1.63 - - -
1 2 0.237 0.022 61.51 59.18 1.65 1.62 0.019 0.019
2 2 8.928 2.184 63.94 - 1.68 - - -
;
.xfx
CONFIG SS_GP_7mil e: 1.00
INTEGRAL
DIELECTRIC e: 4.50 HORIZ_PLANE y: 0.0 7.0
CONDUCTOR 1 RECTANGLE x: 0.0 6.0 y: 7.0 9.0
CONDUCTOR 2 RECTANGLE x: 12.0 18.0 y: 7.0 9.0
GROUND HORIZ_PLANE y: -1.5 0.0;
CONFIG S_S_GP_7mil e: 1.00
INTEGRAL
DIELECTRIC e: 4.50 HORIZ_PLANE y: 0.0 7.0
CONDUCTOR 1 RECTANGLE x: 0.0 6.0 y: 7.0 9.0
CONDUCTOR 2 RECTANGLE x: 24.0 30.0 y: 7.0 9.0
GROUND HORIZ_PLANE y: -1.5 0.0
;
CONFIG GSGS_GP_7mil e: 1.00
INTEGRAL
DIELECTRIC e: 4.50 HORIZ_PLANE y: 0.0 7.0
GROUND RECTANGLE x: -12.0 -6.0 y: 7.0 9.0
CONDUCTOR 1 RECTANGLE x: 0.0 6.0 y: 7.0 9.0
GROUND RECTANGLE x: 12.0 18.0 y: 7.0 9.0
CONDUCTOR 2 RECTANGLE x: 24.0 30.0 y: 7.0 9.0
GROUND HORIZ_PLANE y: -1.5 0.0
;
**** To unsubscribe from si-list or si-list-digest: send e-mail to
majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****
cadpro2k%dacafe.com@interlock.lexmark.com on 06/20/2000 12:15:59 PM
Please respond to cadpro2k%dacafe.com@interlock.lexmark.com
To: arap%jdc.paso.fujitsu.co.jp@interlock.lexmark.com,
si-list%silab.eng.sun.com@interlock.lexmark.com
cc: (bcc: Brandon Kemp/Lex/Lexmark)
Subject: Re: [SI-LIST] : Guard traces
Hi Tadashi,
Did you run a simulation with only an airgap between the traces (same
distance you have with the guard traces on the board) to show you what
your crosstalk would be with only the added separations? Try that, and
report back what you find. (Pssst, guard traces aren't necessary) :)
-M-
---------Included Message----------
> From: "Tadashi ARAI" <arap@jdc.paso.fujitsu.co.jp>
> Subject: Re: [SI-LIST] : Guard traces
>
> I'm standing in opposite.
>
> My calculation using XTK showed that there are benefit to reduce
> crosstalk with additional guard trace. Wider the guard trace,
smaller
> noise we can get. Also vias intervals are very effective.
>
> Radiation from clock traces are reduced by guard traces according
to
> MOM simulations (but not by me), too.
>
========================================================
> Mail from dmckean@gte.net described as below:
>
> > Personally?
> > They're unnecessary.
> > If done incorrectly, can lead to all sorts of problems.
> > If done correctly, have to be stitched every inch or so
> > with vias driving up the cost. Adding guard traces leads
> > to separating the traces you wish to separate anyway.
> > Distance is the best cure for crosstalk. Therefore, IMO,
> > it's sort of redundant to use them.
> >
_____________________________________________________________
Tired of limited space on Yahoo and Hotmail?
Free 100 Meg email account available at http://www.dacafe.com
**** To unsubscribe from si-list or si-list-digest: send e-mail to
majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****
**** To unsubscribe from si-list or si-list-digest: send e-mail to
majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****
This archive was generated by hypermail 2b29 : Wed Nov 22 2000 - 10:50:41 PST