Brad,

See below,

JaMi

----- Original Message -----
From: "Brad Velander" <[EMAIL PROTECTED]>
To: "'Protel EDA Forum'" <[EMAIL PROTECTED]>
Sent: Tuesday, September 17, 2002 7:35 AM
Subject: Re: [PEDA] Autorouter and Differential Pairs.


> Jami,
> you seem to be applying the term "differential" impedance as an
> equal to "controlled" impedance, it is not. At the end of paragraph 2 on
> page two Lee states. "...meaning that each is a well designed transmission
> line with a proper termination.  There is no need for differential
impedance
> to accomplish this." A transmission line is a "controlled" impedance line,
> but it is not a "differential" impedance line. So where in this article
does
> he state that controlled impedance is not required? All through the
article
> I can find nothing that even hints he is saying controlled impedance is
not
> required.
>

Let me try to break this down:

> you seem to be applying the term "differential" impedance as an
> equal to "controlled" impedance, it is not.

If required it can be.

>  . . .At the end of paragraph 2 on
> page two Lee states. "...meaning that each is a well designed transmission
> line with a proper termination.  There is no need for differential
impedance
> to accomplish this."

See discussion below.

>    . . . A transmission line is a "controlled" impedance line,

True

> but it is not a "differential" impedance line.

This statement is not necessarily applicable to the first part of the
sentence above. Here  you are beginning to mix "apples" and "oranges".

See discussion below.

>   . . .  So where in this article does
> he state that controlled impedance is not required? All through the
article
> I can find nothing that even hints he is saying controlled impedance is
not
> required.
>

The very premise of the title, the first paragraph, and also portions of the
article dealing with historical requirements of differential signaling, and
the stuff beginning about the middle of page two, all seem to be saying that
"differential signals" do not necessarily require "controlled impedance".

If you believe that he really IS saying "that controlled impedance IS
required" for all "differential signals", then I respectfully submit that
you really are misreading the article.

General Reply

I am not applying the term "differential impedance" as an "equal" to
"controlled impedance", but more properly, one form of, or one way to
achieve, "controlled impedance". I am also not saying that "differential
signals" require "controlled impedance".

Referring to your quote above, back up a sentence, and read what he is
saying . . . "This can be accomplished by routing the signals side by side
(differential impedance, one form of "controlled impedance"), (or) on
different layers or on the same layer spread apart, so long as the two
signals arrive in good condition at about the same time, meaning that each
is a well designed transmission line (another form of "controlled
impedance") with the proper termination."

Don't you see that everything he is talking about here is relating to how to
get 2 lines from "A" to "B" with "some" form of "controlled impedance", so
that either way, everything that he is talking about here is some form of
getting the two parts of a "differential signal" distributed or routed from
one point to the other as "controlled impedance transmission lines", either
individually, or side by side.

This is why I say he is talking about "apples" and "oranges", and this is
one of the parts of that conversation, which has absolutely nothing to do
with what I thought to be his original premise of whether simple
differential signals require a controlled impedance environment.

I mean I am having trouble even with his title. Take the first part,
"Differential signaling doesn't require Differential Impedance". That
statement is true enough, but based on that, and especially on the second
part of the title one would assume that he was going to discuss simple
differential circuits, and why they do not necessarily have to be impedance
controlled (this is what I would call "apples"). But in reality ,what he is
actually talking about and saying in certain parts of the article is that
"("controlled impedance ") Differential signaling doesn't require
Differential Impedance (to maintain "controlled impedance")" (and this is
what I would call the "oranges", which is not really necessarily related to
the "apples").

After about the middle of page two, he appears to get back on the subject of
"simple differential signaling, beginning with "How the current switch
works", but this is back to talking about "apples".

I hope this is making some sense and you can understand what I am trying to
say.

There are really two different things that he is talking about here.

The first thing he is trying to say is that differential signals do not in
fact have to be "controlled impedance" signals, this is "apples". I agree
with this. This is exemplified by the parts of the article which deal with
the aspect of differential signals being developed and used to overcome lack
of ground references, and the stuff beginning with "How the current switch
works".

The second thing he is trying to say is that a "pair of differential
signals" which have to maintain "controlled impedance" can do so either
differentially or as two individual signals, this is the "oranges". I agree
with this too, but they are not the same thing.

To me the term "differential impedance" speaks of the impedance that results
due to the "coupling" of the two lines in a close proximity structure, but I
don't think that I would actually want to call that a definition.

I actually reread the beginning of the article last night again, and was
going to write you and Ian this morning anyway, before I got this post.

This time I didn't get past the third paragraph on the first page before I
saw the basic problem with the article.

The very first sentence of the third paragraph states: "As it turns out,
differential impedance does not play a role in this form of signaling and is
not necessary." This follows a paragraph that discusses ECL, which is a
"controlled impedance" logic family, and LVDS, which in at least some
circumstances must maintain "controlled impedance" (as I stated in my
previous post, it may be wrong to say that LVDS always requires "controlled
impedance" in every case).

Under the normal rules of usage of the english language I would interpret
the "this" in the middle of the first sentence of the third paragraph, as
referring to that which was just discussed in the preceding paragraph, which
includes ECL (requiring "controlled impedance") and LVDS (at least sometimes
requiring "controlled impedance").

That being the case, the statement is wrong.

"Differential impedance" can be the method of maintaining the "controlled
impedance" of a differential pair when that differential pair also needs to
maintain "controlled impedance", although you do not have to use
"differential impedance" to maintain "controlled impedance", which can be
done individually.

Yeah, I know that is a tricky sentence, but read it carefully, it is a valid
statement.

I could accept his statement if you were to modify it so that it read: "As
it turns out, differential impedance does not ALWAYS HAVE TO play a role in
this form of signaling and is not necessary PROVIDING YOU USE OTHER MEANS TO
CONTROL THE IMPEDANCE." (please remember he is talking about "controlled
impedance" in the previous paragraph.

Beginning to see why I say he is talking about "apples" and "oranges".

The real issue that I believe is the problem, and I thought that he was
trying to address here, although he does not (in the two pages I read), is
that many people assume that just because signals are differential, that
does not in and of itself necessarily mean that the signals have to be
"controlled impedance".

A variant of this might be that while differential signals routed together
side by side, do exhibit some differential impedance, this is not
necessarily so (they may not need to be "controlled impedance").

We have two completely different things that are not necessarily related to
each other, and we have a third area where we use these two seperate things
come together in a certain application. First, we have "differential
signals". Second, we have "controlled impedance" transmission lines. These
two things are not necessarily related. In the Third area, we bring these
two things together, and we have "differential" and "controlled impedance"
transmission lines, and one of the ways of physically doing this results in
what I will for the moment call "differential controlled impedance".

One of the problems here is that with any two parallel wires or traces used
as "differential signals", and they are in close proximity to each other,
there is something that I have heard refered to as "self impedance"
(although I will not vouch for the legitamacy of this term), which is a
result of the "mutiual coupling" of the two conductors. This impedance is
also referred to as "differential impedance", and is usually there to some
extent with any "differential signals" when the conductors are in close
proximity. Unfortunately, this "differential impedance" often gets confused
with what I discribed above as "differential controlled impedance", but
which is actually correctly called simply "differential impedance" also. In
reality, both are really just "differential impedance", and the one happens
when you don't even care and your conductors are in close proximity, and the
other happens because you want it to and have to have it happen in a
controlled manner, but these are really two different ways of looking at the
same thing.

With that said (and with the realization that any "differential signals" in
close proximity may will exhibit some "differential impedance"), we need to
realize that thet that there are the two seperate and not necessarily
connected worlds of "differential signals" and "controled impedance", and
that there is the third world where these two worlds are combined together.

It is when we try to transport the term "differential impedance" around in
all areas, without proper understanding of the which of these three worlds
we we are discussing, that the term becomes unmanagable and confusing, and
while with certain limitations, it can be used in each of the three worlds,
it does not always carry the same meaning or sense.

I notice that this article states up at the top of the first page "Article
for Printed Circuit Design", and I am wondering if that is referring to
printed circuit design in general or to the magazine Printed Circuit Design.
If it is the magazine, it needs a lot of rework before it gets published,
and if it has been published already it needs a lot of "corrections", and I
would love to se the "letters" that it generated.

It is entirely possible that I missed something by not reading the article
completely thru, but with the problems I have already pointed out in the
third paragraph and on the second page, I think that the article is already
much too problematic to continue with, or rely on.

I know that by stating this in a post in this forum that I am going way out
on a limb, and almost inviting someone to saw off the branch, but this
really is an important topic to understand in printed circuit design in
general and here in this forum particular if we are going to get Altium to
include the capability to route this kind of stuff in DXP.

No, I don't think they would give it to me in SP7.

This whole topic began with Anand asking questions about routing a BGA back
on Friday evening, and I wonder if he ever got the help that he needed.

JaMi

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
*
* Forum Guidelines Rules:
* http://www.techservinc.com/protelusers/forumrules.html
*
* Browse or Search previous postings:
* http://www.mail-archive.com/proteledaforum@techservinc.com
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Reply via email to