[SI-LIST] Re: Guard traces for differential pairs

  • From: Duane Takahashi <duanet@xxxxxxxxxxxxxxxxxxxxxx>
  • To: si-list@xxxxxxxxxxxxx, chris.cheng@xxxxxxxxxxxx
  • Date: Thu, 02 Oct 2003 15:56:25 -0700

Hi Chris:

Nice to hear from you again.

I wasn't talking about a non homogeneous tline model or SSO noise.  I 
was trying to explain the conditions under which I've seen inductive 
xtalk appear in the lab.

Duane

> Can you explain what does the presence of guard trace for differential 
> pairs
> has any thing to do with driver impedance and load impedance ? You are
> mixing up a non-homogeneous transmission line model with SSO noise and
> Lij/Lii ne Cij/Cii has nothing to do with inductive or capacitive xtalk but
> due to the non-homogeneous dielectric boundary. If the driver is
> differential, the differential switching noise cancels each other out. 
> Don't
> keep throwing terms out just to confuse people.
> 
> -----Original Message-----
> From: Duane Takahashi [mailto:duanet@xxxx]
> Sent: Tuesday, September 30, 2003 9:25 AM
> To: damonjbowser@xxxx; si-list@xxxx
> Subject: [SI-LIST] Re: Guard traces for differential pairs
> 
> 
> Hi Damon:
> 
> I don't think you can look at the L and C matrices, and determine if the
> xtalk will be capacitive or inductive. As Scott points out, they move
> together.
> 
> IMHO, you need to take into account the driver's impedance and the
> load's impedance. To me, packages look inductive: bond wires with
> traces over perforated reference planes. Flip chip helps a lot, but you
> still have the perforated planes.
> 
> When the driver fires, the inductance forces the voltage wave to lead
> the current wave. This phase relationship travels down the line. To
> the load, the line's impedance is no longer purely real, and has an
> imaginary component to it.
> 
> In this case, the capacitive and inductive forward xtalk components do
> not cancel, and the inductance xtalk dominates.
> 
> Duane
> 
>  > For TEM and quasi-TEM transmission line systems, the Inductance and the
>  > Capacitance matrices are related. For lossless and low loss systems,
>  > the inductance matrix is:
>  > L = mu * epsilon * (C inverse)
>  >
>  > They are linked as long as the magnetic and electric fields remain
>  > orthogonal. Thus, whenever you change the capacitance of a system of
>  > traces, you also change their inductance.
>  >
>  > A good quasi-static field solver will compute the correct L and C
>  > matrices for any given trace configuration. In fact, many of the
>  > currently available field solvers assume that the conductors are perfect
>  > and therefore only solve for the capacitance matrix. The inductance
>  > matrix is computed using the expression above. Clayton Paul has a good
>  > discussion of this in his book, "Analysis of Multiconductor Transmission
>  > Lines."
>  >
>  > As you change capacitive coupling, you also change inductinve coupling.
>  >
>  > regards,
>  >
>  > scott
>  >
>  > Hi Duane,
>  >
>  > So how does one determine if it is magnetic or
>  > capacitive crosstalk that is dominating in a
>  > particular configuration?
>  >
>  > Let's say I get two matrices from a field solver that
>  > look like this:
>  >
>  > C11 C12
>  > C21 C22
>  >
>  > and
>  >
>  > L11 L12
>  > L21 L22
>  >
>  > Do I just look at how C12/C11 compares to L12/L11 and
>  > how C21/C22 compares to L21/L22 ?
>  >
>  > Thanks
>  > Damon
>  >
>  >
>  >>X-Original-To: si-list@xxxx
>  >>Delivered-To: si-list@xxxx
>  >>From: Duane Takahashi
>  >><duanet@xxxx>
>  >>User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0;
>  >>en-US; rv:1.4)
>  >>Gecko/20030624 Netscape/7.1 (ax)
>  >>X-Accept-Language: en-us, en
>  >>MIME-Version: 1.0
>  >>To: si-list@xxxx
>  >>Subject: [SI-LIST] Re: Guard traces for differential
>  >>pairs
>  >>Content-Transfer-Encoding: 8bit
>  >>X-archive-position: 8392
>  >>X-ecartis-version: Ecartis v1.0.0
>  >>X-original-sender: duanet@xxxx
>  >>X-list: si-list
>  >>
>  >>It also depends on the type of crosstalk: electric
>  >>or magnetic (
>  >>capacitive or inductive ). If it's magnetic, guard
>  >>traces won't help.
>  >>If it's capacitive, guards will help but will leave
>  >>you with the
>  >>'magnetic remainder' of the xtalk.
>  >>
>  >>I've seen high impedance lines, more than 60 ohms,
>  >>exhibit inductive
>  >>xtalk. My guess is guard traces would not have
>  >>helped in this case.
>  >>
>  >>Regards,
>  >>Duane
>  >>
>  >>
>  >>>Sogo,
>  >>>
>  >>>It is true that guard traces can INCREASE
>  >>
>  >>crosstalk, but
>  >>
>  >>>this can be avoided (and crosstalk in fact can be
>  >>
>  >>REDUCED)
>  >>
>  >>>if the via stitching is frequent enough. Between
>  >>
>  >>each
>  >>
>  >>>stitching, there is a half-wavelength resonator,
>  >>
>  >>with a
>  >>
>  >>>lowest resonance frequency of fres=1/(2*tpd),
>  >>
>  >>where
>  >>
>  >>>tpd is the propagation delay between two stitching
>  >>
>  >>vias.
>  >>
>  >>>As long as the bandwidth of the signal on the
>  >>
>  >>structure
>  >>
>  >>>is safely below this resonance frequency, there
>  >>
>  >>should be no
>  >>
>  >>>problem with signal integrity.
>  >>>
>  >>>Regards,
>  >>>
>  >>>Istvan Novak
>  >>>SUN Microsystems

-- 
Duane Takahashi              phone: 408-720-4200
Greenfield Networks            fax: 408-720-4210
255 Santa Ana Court          email: duanet@xxxxxxxxxxxxxxxxxxxxxx
Sunnyvale, CA 94085

* MOVING!  Please note new numbers and address *

------------------------------------------------------------------
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: