[SI-LIST] Re: general belief about Xtalk

  • From: "Howard Johnson" <howie03@xxxxxxxxxx>
  • To: <si-list@xxxxxxxxxxxxx>
  • Date: Wed, 4 Mar 2009 10:39:02 -0800

I will set down here in writing, for your benefit, some portions of the
little store of knowledge I possess about transmission-line crosstalk. 

Regarding the comment below by Mick zhou, he is correct that my first book,
High-Speed Digital Design, did include the statement on page 207, "The
forward crosstalk is never larger than the reverse crosstalk", *however* the
statement is (admittedly) misleading, due in part to my lack of
specification of what I meant by the term "larger".  A more precise
statement would be, "The (area under the) forward crosstalk (waveform) is
never larger than the (area under the) reverse crosstalk (waveform)."  That
statement about areas I believe to be true under all the conditions
(including connector and via crosstalk) that I can think of at the moment. 

The area relationship leaves open the possibility that the amplitude of a
long, low near-end crosstalk waveform (NEXT) could be exceeded by the
amplitude of a tall, narrow far-end crosstalk waveform (FEXT). As has been
pointed out by other respondents to this thread, that sometimes happens on
very long parallel microstrips (many, many signal risetimes long). At the
time my first book was published (1993) typical pcb traces were shorter than
half a signal risetime. That's not true any longer. My second book,
"High-Speed Signal Propagation", page 321, qualifies the discussion, saying
that for parallel microstrips having a length less than or equal to half the
signal risetime, the peak height of the FEXT never exceeds the peak height
of the NEXT. What I didn't say was that the length at which FEXT overtakes
NEXT depends on the trace geometry and dielectric constant.  

If I remember the math correctly, for parallel transmission lines at
extended lengths the FEXT amplitude eventually saturates, never growing
larger than 1/2 the signal amplitude. 

Typical microstrips have only a small FEXT coefficient. Striplines have an
even much smaller FEXT coefficient. As Yuriy correctly points out, the FEXT
coupling coefficient for a stripline may not be zero, but I have not in
practice observed significant amounts of stripline FEXT. Crosstalk from
packages, connectors, vias, and NEXT from other traces are more likely to
cause trouble in stripline connections than stripline FEXT.

It is my general experience that in well-terminated synchronous bus
architectures NEXT causes more trouble than FEXT for the simple reason that
the FEXT waveform arrives at the receiver along with the bus transitions--a
time when the synchronous bus is not, by definition, sampling its data
inputs. The NEXT waveform, on the other hand, can bounce off the drivers and
arrive at a much later time, interfering more heavily with data reception.  

Asynchonous serial-link designs are equally sensitive to NEXT and FEXT. In
an asynchonous serial-link design the FEXT from one channel can (and does)
interfere with adjacent channels and must be considered. 

One last point: in any system that produces most of its crosstalk within a
small space (small compared to the signal risetime), the NEXT and FEXT
waveforms have approximately equal durations. The NEXT waveform will always
be larger, but sometimes not by much. This point applies to many connector
and via situations. 

The NEXT and FEXT waveforms attain almost equal amplitudes when the
crosstalk is dominated by either a large lumped-element capacitive crosstalk
effect, or a large lumped-element inductive effect. The following article
shows a good example:  

  "Directionality of Crosstalk", (1997)
www.sigcon.com/Pubs/straight/crosstalk.htm 

I'll be teaching a public class in Orange County March 17-18 and would be
delighted to address this subject in person at that time. Even if you are
not registered, just show up at 5:00 pm after class the first day and we can
have a little after-hours discussion. 

Best regards,
Dr. Howard Johnson, Signal Consulting Inc.,
tel +1 509-997-0505,  howie03@xxxxxxxxxx
www.sigcon.com -- High-Speed Digital Design seminars, publications and films
 




-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On
Behalf Of Yuriy Shlepnev
Sent: Saturday, February 28, 2009 8:17 AM
To: 'Ihsan Erdin'; 'Mick zhou'
Cc: 'David Banas'; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: general belief about Xtalk

I agree with Ihsan in general, but would like to share some thoughts on the
TEM waves.
The waves in interconnects can be considered as TEM (no E and H along the
lines) only in cases of strip-line structures with homogeneous dielectric
and with lossless conductor. Even lossless micro-strip line has quasi-TEM
dominant mode with both E and H components along the propagation direction
due to the boundaries between dielectrics (the effect of these components
can be observed above about 5 GHz for typical PCB interconnects). In case of
lossy conductors, both strip and micro-strip lines have longitudinal
components of E and H that are growing with the frequency. Losses and
electromagnetic energy absorption by conductor generates electric field
along the propagation direction simply due to the Ohm's low E=r*j - this
violates the TEM nature of the fields in interconnects with lossy
conductors. Belief in TEM mode is probably based on the common use of static
field solvers for analysis of such multi-conductor interconnects. Though,
the static solution is typically just the starting point in such analysis.
Losses in conductors and dielectrics are added as post-processing of the
static solution (with perturbation analysis and analytical formulas for
conductor loss and with solution with complex dielectric constant at one
frequency for dielectric loss). It provides good approximation even for
non-TEM waves in cases if some additional conditions are satisfied - such as
the cross-section must be much smaller than wavelength at the highest
frequency of interest, smallness of conductor-related losses, identical loss
and dispersion in all dielectrics and so on - the detailed analytical
analysis is provided in F. Olyslager, Electromagnetic waveguides and
transmission lines, Oxford, 1999, p.29.

Note, that being non-TEM or quasi-TEM, the waves in multi-conductor lines
remain transverse even at very high frequencies (the phase of the fields
depends only on the position along the line) and Telegrapher's equations can
be still used with appropriate impedance and admittance per unit length
extracted with the full-wave (non-static) analysis. On the other hand, the
fields near the discontinuity are not transverse in general - that may be
distinguishing property of a discontinuity. Though, the fields can be still
expanded into a set of the transverse waves (both propagating and
evanescent) in any segment of a line close to the discontinuity - this is
just analogue of the Fourier expansion.

Best regards,
Yuriy Shlepnev
www.simberian.com 

-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On
Behalf Of Ihsan Erdin
Sent: Friday, February 27, 2009 7:39 PM
To: Mick zhou
Cc: David Banas; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: general belief about Xtalk

Before this issue is laid to rest I'd like to add a few comments to TEM mode
of propagation that was mentioned in this thread. The hallmark of TEM is
that the field component in the direction of propagation is zero. In a
linear and isotropic medium, the E and B fields will be perpendicular to
each other regardless of TEM or non-TEM mode of propagation. That condition
may be broken if the medium is nonlinear or anisotropic but this could be
hardly an issue for SI applications. For all practical purposes it is safe
to assume the electric and magnetic fields will remain perpendicular to each
other but there will be field component(s) in the direction of propagation
when TEM is violated at interconnect discontinuities, connectors and such.
Regards,

Ihsan


On Fri, Feb 27, 2009 at 6:36 PM, Mick zhou <mick.zhou@xxxxxxxxx> wrote:

> David,
> Thanks for your inputs and others. There is a similar description in
Howard
> Johnson's book, p.204-211. Actually I like Fig.5.16. On. p.207, he 
> even states " The forward crosstalk is never larger than the reverse
crosstalk".
> It might be the source of the belief. However it cannot be considered 
> as a proof at all, and is overtaken. His assumption is "Under normal
> conditions".   My point is we should use this statement to judge results
> carefully, especially when extend to other connectors, packages or 
> mixed modes.
>
> There are studies (peer-reviewed papers) showing the beliefs are
incorrect.
>
> Have a great weekend.
>
> Mick
>
>
> On Mon, Feb 23, 2009 at 5:22 PM, David Banas <david.banas@xxxxxxxxxx>
> wrote:
>
> > The beliefs you give, below, come from the practical observation 
> > that, for typical modern PCB designs, capacitive and inductive cross 
> > talk have similar magnitudes. Consider:
> >
> > For near end crosstalk:
> > - the capacitive and inductive components reinforce each other, but
> > - the received energy observed at the victim driver is spread out in 
> > time,  due to the fact that the crosstalk induced energy is 
> > traveling in a  direction opposite that of the advancing wave in the 
> > aggressor.
> >
> > For far end crosstalk:
> > - the capacitive and inductive components cancel each other, but
> > - the received energy observed at the victim receiver is localized 
> > in time,  because the crosstalk induced energy is traveling along 
> > with the advancing  aggressor wave.
> >
> > So, to know whether a larger magnitude of crosstalk will be observed 
> > at the near or far end requires knowing something about the relative 
> > significance of the inductive/capacitive cancelling effects vs. the 
> > spreading and localizing effects. In typical modern PCB designs, the 
> > inductive and capacitive components of crosstalk cancel rather well 
> > and, so, the cancelling ends up being more significant than the 
> > spreading, and we get larger magnitude at the near end.
> >
> > -db
> >
> >
> > > -----Original Message-----
> > > From: si-list-bounce@xxxxxxxxxxxxx
> > [mailto:si-list-bounce@xxxxxxxxxxxxx]
> > > On Behalf Of Mick zhou
> > > Sent: Monday, February 23, 2009 3:59 PM
> > > To: si-list@xxxxxxxxxxxxx
> > > Subject: [SI-LIST] general belief about Xtalk
> > >
> > > Hi,
> > > Recently, I run into some basic beliefs about Xtalk from my
> > colleagues.
> > > They
> > > seem match intuiations but may not be right.
> > >
> > > Statement 1: Single-ended near-end Xtalk is always greater than
> > far-end
> > > Xtalk, both in f and t domains.
> > > Statement 2: Differential/Common near-end Xtalk is always greater 
> > > than
> > the
> > > far-end, both in f and t domains.
> > >
> > > I do not remember any general proofs of the above statements. They 
> > > may
> > be
> > > true for many practical cases but may not be generally true. One
> > example
> > > in
> > > S. H. Hall's book, p.50. disproves the statement 1 in t-domain.
> > >
> > > Any more helps?
> > >
> > > Mick
> > >
> > >
> > > ------------------------------------------------------------------
> > > 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 technical documents are available at:
> > >                 http://www.si-list.net
> > >
> > > 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
> > >
> > >
> >
> >
> > This email and any attachments are intended for the sole use of the
named
> > recipient(s) and contain(s) confidential information that may be 
> > proprietary, privileged or copyrighted under applicable law. If you 
> > are
> not
> > the intended recipient, do not read, copy, or forward this email 
> > message
> or
> > any attachments. Delete this email message and any attachments
> immediately.
> >
> >
> >
>
>
> ------------------------------------------------------------------
> 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 technical documents are available at:
>                http://www.si-list.net
>
> 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 technical documents are available at:
                http://www.si-list.net

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 technical documents are available at:
                http://www.si-list.net

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 technical documents are available at:
                http://www.si-list.net

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: