[SI-LIST] Re: Students - matching 1 mil IEEE1394/ethernet guidelines and DM to CM conversion
- From: steve weir <weirsi@xxxxxxxxxx>
- To: Bill Owsley <wdowsley@xxxxxxxxx>
- Date: Sat, 02 Jun 2007 19:48:13 -0700
Bill, if stipulating such tight matching works for you, keeps your boss
happy and you happy, I am happy. I see no harm provided the cost is
free. Maybe it provides the competitive advantage that you express it
does, but absent either a theoretical basis, or verifiable experience I
just don't see it. Until I see one or the other I'll pass on accepting
this practice as anymore of a competitive advantage than using 0.01%
tolerance termination resistors.
When we connect up to magnetics, even good magnetics, the isolation
transformer is going to give us 1% mode conversion on a good day, more
on a bad one. This pretty much forces us into CM suppression, most
commonly implemented with iron. I don't care how closely you match
those trace lengths, they won't make up for what the transformer is up
to. Drop in the intersting physical paths inside the RJ, especially
RJ's with IM, and the trace match just isn't on my RADAR.
What you may be experiencing is that the ultratight match requirement is
causing your designers to be more careful with the routes themselves,
and avoiding doing things that would create unwanted coupling and mode
conversion along the path between the PHY's and the RJ's. If that
constraint is an effective and economical way to get there, great. But
given my understanding of physics and experience I just don't see any
reason to believe it has anything to do with the route length matching
to such a fine precision in itself.
Best Regards,
Steve.
Bill Owsley wrote:
> Steve,
> I would if I could, but alas I have been mandated (pay check
> dependent) to produce product that goes out the door with the minimum
> fuss possible and with no returns for redesigns. There has been no
> interest (funding) to pay for demonstrating in a public forum what is
> considered by some employers as a competitive advantage. Let rational
> engineering rule.
> ps. And you mentioned in the note below the very area I was thinking
> of for my assertion, and the students apparent area of work. And
> smaller route, the more the effect, such as within the integrated
> magnetic RJ's and the wiring to them, but I found the effect first in
> the discrete magnetics, some good and some not so good. Remove the
> package, or x-ray the internals, to see the routing.
> Curious how the original problem limitations dropped away during the
> discussions that followed. I found it to be one of the more
> stimulating discussions on the list, and as one contributer suggested,
> one ripe for research in the many areas of diff-pair use.
> As for those atrocious integrated magnetic RJ's, I'm afraid I may have
> provided the reason for developing them when the choice of discrete
> magnetics manifested a remarkable ability to convert DM to CM. A
> different vendor had a much better product, but a different pin out
> and so could not be substituted. We shipped a most embarrassing
> product wrt EMC, but not for long for as soon as the integrated RJ
> ramped up volumns, in it went.
>
>
> */steve weir <weirsi@xxxxxxxxxx>/* wrote:
>
> Bill, Scott delivered a very good paper on this subject at DesignCon
> 2005. While there are methods to reduce the effects, what was found
> from a good test platform is that the random behavior only serves to
> determine the distribution of the skew. It does not normalize it
> towards zero. Unless specific measures are taken: choice of weave,
> routing pattern wrt the weave etc ( some of which are or were
> recently
> under patent dispute ), the 3-10ps / inch skew that Scott
> mentioned is
> very real. When best methods are employed, that can be reduced to
> perhaps 1.5-2ps/inch. On top of this there is the bandwidth
> limitation
> of any particular transmission line design with whatever materials we
> choose. What this says to me is that only very short lines can
> benefit
> from a pretty tight match. But most cases I am familiar with, those
> would be lines going to magnetics. When dealing with manufacturing
> variation atrocities like many of the RJ's out there with integrated
> magnetics, 1 mil, 10 mils, even 100 mils mismatch is pretty hard
> to for
> a guy like me to see. If it's really free I see no harm. I am really
> pressed, and would be delighted to see a real statistically valid
> study
> that showed that a particular physical path match length provides a
> measureable and meaningful benefit.
>
> Best Regards,
>
>
> Steve.
> Bill Owsley wrote:
> > Most systems are quite functional (SI) when presented to EMC for
> testing, and quite often fail. And as so many have very eloquently
> (where's my spell check) explained, there is not likely any 1 mil
> mis-match in a matched pair that caused it. So given my realm of
> influence, (which certainly is not the weave of FR4. Is there a
> felt or random pattern FR4?) I ask for continuous incremental
> improvements (that don't get me necktie party). I suspect that the
> weave variation of FR4 over any run of interest would have a plus
> and minus shift that on the average would come out near the
> nominal - remember odd/even number of twists for a pair, that odd
> number twist would unlock pandoras box. And certainly the other
> sources mentioned that are not the diff-pair trace length can be
> the dominate source of problem, but they were not part of the
> trace length constraint.
> >
> > Thanks to those that brought up the BER, RJ, any other jitter,
> eye diagrams and that stuff of SI world. All those numbers scare
> me in some fashion, they are so big.
> > And it seems that a better match in trace lengths given the
> phase percentage mentioned in another note and other descriptions
> of the effects, or lack of, for excessive constraints for
> differential signalling, that the little X mark in the middle of
> eye diagram seems to shrink, the supposed flat segments of the eye
> diagram are a little bit flatter. In a tightly couple pair, the
> forward crosstalk to the other signal of the pair is a little bit
> closer in phase with better matching and so does not cause as much
> of a slight shift in the crossover or switching point, leading to
> less jitter. Some of the multilevel signalling has such small
> differences in the discrete levels that any small improvement in
> the little effects that degrade these levels would seem to be
> better. Does a 1 mil request/constraint do that and at what cost?
> Well now, that all depends on where you would like to be in the
> market.
> >
> >
> >
> > Chris Cheng wrote:
> > Show me a case where 1 mil difference will break SI.
> > Then.
> > Show me a case where 1 mil difference will break EMI but not SI.
> >
> >
> > -----Original Message-----
> > From: si-list-bounce@xxxxxxxxxxxxx
> > [mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Bill Owsley
> > Sent: Friday, June 01, 2007 2:29 PM
> > To: si-list@xxxxxxxxxxxxx
> > Subject: [SI-LIST] Students - matching 1 mil IEEE1394/ethernet
> > guidelines and DM to CM conversion
> >
> >
> > If any students are still with us, the ongoing interchange might
> indicate that these subjects are indeed interesting and can be
> somewhat complex in that there are a number of variable to keep in
> mind - all at once.
> > And maybe enough information to get your project done well.
> > You're welcome... < > really stupid grin within the brackets
> >
> > And this all started with a simple help me with my project question.
> >
> > Since this is an SI list, the EMC aspects seem a little less
> important. I'm reminded of a class on how to use one of those CAD
> tools for schematic capture, layout, SI, EMC. The SI guys got to
> go home a day early since their concern in class was millivolts.
> The EMC guys had to stay over a day to work on the microvolts part.
> >
> > And there is at least one layout group that has for me, a short
> rope and tall tree. But I love them anyway.
> >
> > ---------------------------------
> > Fussy? Opinionated? Impossible to please? Perfect. Join Yahoo!'s
> user panel and lay it on us.
> >
> > ------------------------------------------------------------------
> > 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
> >
> >
> >
> >
> >
> > ---------------------------------
> > It's here! Your new message!
> > Get new email alerts with the free Yahoo! Toolbar.
> >
> > ------------------------------------------------------------------
> > 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
> >
> >
> >
> >
> >
>
>
>
> ------------------------------------------------------------------------
> Get the Yahoo! toolbar and be alerted to new email
> <http://us.rd.yahoo.com/evt=48225/*http://new.toolbar.yahoo.com/toolbar/features/mail/index.php>wherever
>
> you're surfing.
------------------------------------------------------------------
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:
