[SI-LIST] Re: Stack up for EMI reduction, plane resonance and u-s trip radiation etc etc

  • From: "Michael E. Vrbanac" <vrbanacm@xxxxxxxxxx>
  • To: steve weir <weirsp@xxxxxxxxxx>, si-list@xxxxxxxxxxxxx
  • Date: Sat, 14 Feb 2004 15:54:35 -0600

I suppose this could go on forever.... I'll snip severely to shorten 
things... and then make them long again
... but then again isn't that what a "forum" is for... discussion?

>If your point is that we could arrange an inefficient coupling by 
>selectively choosing lambda, I agree.  But, I disagree that arbitrarily 
>putting a brick wall at the board edge does that, as lambda is then set by 
>the board geometry, and not by design against some excitation that we want 
>to present an inefficient coupling to.  And that's where I think the 
>position and reactance of that plate in your analogy is very important.

re: points of agreement
Yes, you can do it that way, too.  Generally, you don't get to choose 
lambda.  It gets chosen for
you based on the physical reality.  You choose the position based on that 
reality.  I think the value
of what we are talking about rests in the reasons why this analogy works.

re: brick wall at the board edge
Who's talking about that?  I thought we were talking about fences, etc. and 
the effects that they
had on things.  That is another matter.  Such things, like "brick walls" 
used judiciously are quite
effective and at least one very good paper was written about their use by 
some folks working for
a cell phone vendor.  I'll let you discuss it with them.

re: about arbitrariness.
For the record, I do not arbitrarily put things anywhere. So, if you 
please, let's not talk about arbitrariness
unless we are getting rid of it.

>I understand where you are trying to go with this.  I am sorry you felt 
>the need to go all the way back to Ampere's law for it.
>
>Let's see what we can agree upon, and what's left in our differences.
>
>We agree that Ampere and Biot-Savart still apply.
>We agree that mu is effectively 1.0
>So this means that we must agree that the distribution of B relative to 
>the trace position has not changed.
>I think we also agree that if we perform a 1.0000 meter measurement and a 
>1.0002 meter measurement the EMI numbers will be indistinguishable.
>
>Stopping right there, your posit is that B is the problem for radiation 
>and if we agreed on this, you would be absolutely correct.
>
>However, B is dominant to low impedance coupling, like nearby traces or if 
>there is some object that is going to reradiate.

re:  the easy stuff
Yes to the first three.  The fourth...yes, that's what I think will happen, 
but your measurements
will not be valid.  Let's not short-circuit a good answer by skimping on 
the procedure.  For the
measurement to be valid, the microstripline's distance to the plane must be 
that same in both cases.
Otherwise, that answer will be as bogus as any other myth I've seen.

re: wave impedance
I am sorry that this is not clear to you.  In many laboratory studies, I've 
proven this point ad nauseum
and will not argue it here with you.  I'll point you however to do some 
study about shielding, type of shielding,
why they work and when they won't and what are the operating 
characteristics of the materials, bonds, and
wave impedances involved.  We aren't going to go anywhere on this until 
this is understood.

re: object going to "re-radiate"
Sigh...  by definition (of "re-radiate") since there is no current loop, 
the re-radiation case is an E-field
induced structure.  It has to be. And its a "high-impedance structure", no 
circuit loop structure
except by parasitics, no/very low current flow, maximum voltage without 
load.  It will re-radiate in
proportion to its match to free-space following Gauss' Law for electric 
fields. Any reduction in charge
on the object will be due to small amounts of leakage current through the 
environmental parasitics.
It is not a B-field structure.

>>C = (area * permittivity) /  distance
>
>We disagree in that the above equation does not work for fringing.  It 
>works near the center of the "infinite plane of charge".

Hmph.  Are we changing physics now?  I am sorry that I provided the "at a 
point" version of the equation....
should I have expanded it to deal with vector quantities in three 
dimensions? I am stating the simplified version
just to point out the principle. But the equation still applies, its proven 
physics...  to make it work in the fringing
case we have to work on the spatial aspects of the problem.

We don't have to necessarily do that to get a general understanding of the 
problem.  Here's the facts:
1.  We assume that the microstripline of 4-5 mils width is closely 
associated with the reference plane
(4-5 mils) and we're using a dielectric constant of around 4.
2.  The order of greatest to least E-field flux density is as follows:
         a.  between the microstripline and the reference plane
         b.  off the middle to lower edge of the microstripline to the 
reference plane
         c.  off the middle to upper edge of the microstripline to the 
reference plane
           d.  off the backside of the microstripline
3.  (a) and (b) are already captured in the surface case and would be in 
the embedded case.
4.  (b) constitutes more than 50% available in all fringing fields due to 
dielectric presence and would
in the embedded case as well.
5.  (c) has some field lines captured but they are weakly coupled, the rest 
are not captured.
6.  (c) can be exploited by additional coverage by dielectric but is much 
less than 50% of total
fringing field lines.
7.  (d) cannot be exploited to any meaningful degree by additional coverage 
by dielectric
8.  if an interposed reference plane were added to make the structure a 
stripline then field capture
would be near 100% (assuming no apertures) depending on "top" dielectric 
characteristics and plane
distance and distance to the edge of the reference plane.  The flux density 
would more evenly spread
over the two surfaces of the "now stripline" center conductor.

Therefore, the few remaining flux lines in the fringing fields would have 
to constitute the entire change
of 15dB reduction or there was another mechanism, or there was a 
combination, heretofore unidentified.

OR

we really are not closely associated with the reference plane and loosely 
coupled.  Such a case might
deliver a larger than normal reduction due to the excessive fringing fields 
that would result.  In such a case,
the "rectangular" shape would begin to look more rounded from a greater 
distance and the fringing fields
would be must greater.  In such a case, it might be advisable to add the 
extra dielectric layer to contain the
fringing fields... downside... more cost.  This just might be the "trick" 
or the "other mechanism" I'm
talking about.  The problem is that even a 15dB reduction in that case may 
still result in very poor test
readings.  What is not said here is that why were they looking for a 15dB 
reduction in the first place?
Probably because something was really not working well for the loosely 
coupled case.

However, back to the tightly coupled model, is that is doesn't work that 
way.  Yes, I've tested this
before as a "cost reduction idea" for a company.  Zilch. Zip. Nada.

Also, as you probably know, its easy to see all sorts of 15dB or more 
reductions in a EMI debug lab
and yet see no change on the test range.

>>3. Permittivity only increased by a factor of four, a 12 dB change max. 
>>This sets the theoretical
>>top end of the performance range.  A dielectric constant greater than 9 
>>would likely be necessary
>>to bring us any hope of reaching 15dB.  FR-4 is only about 4.2 - 4.5 
>>(generally) depending on the
>>material makeup.
>
>Let's stick with Er = 4, for FR4 as close enough.  So the issue is the 12 
>dB value.
>So, the flux density of the electric field lines can only increase by a 
>factor of 4.  However as previously noted, we are already high influenced 
>by fringing.  The amount of total flux that is above the center of the 
>conductor is limited.

You're almost there.  The total flux above the center of the conductor is 
limited... and less than 50%
of the total value.  The majority of the rest are captured.

re: highly influenced by fringing
Only in the loosely coupled case is this possible... in the tightly coupled 
case (IMHO, proper design practice)... no.

>Agreed that this isn't a strip-line.  The key is to look at the 
>distribution of the lines, and the significant concentration near the 
>trace edges before and after the submersion.

Hey, gimme a break.  I'm not that stupid.  What have I been talking about 
all this time?  The
current density is greatest at the edges... hint, hint.  The flux lines off 
the signal conductor will
align themselves in proportion to the capacitance they see at their 
"attachment point".

The only way it will work is if the microstripline isn't hardly that at all 
and the signal conductor
is relatively distant to the reference plane.  Then perhaps the 15dB down 
might work... but then
again, I don't design like that.  I get rid of the problems and many others 
by tight coupling to
the reference plane in the first place.  I even do this with differential 
pairs at multi-gigabit but I
know how to tweak it to get precisely what I want.  It works, too. 
Wonderful BERs (years of
16 corner testing) and great signal shape, and performance.

>>5. If we were to assume half of the remaining field lines were totally 
>>removed from the problem by
>>completely immersion in the FR-4 dielectric, this would only amount to a 
>>6dB reduction over the
>>surface case... hence this was the number I quoted you earlier.
>
>I think the fallacy here is that if we removed the ground plane the power 
>would only go up by 6db.  I disagree that the converse is true that by 
>taking an arbitrary measure the power will only go down by 6db.

Its no fallacy.  Its just reasoning based on conservation of charge.  The 
"balance sheets of physics"
must always balance.  Half the field lines, half the E-field.  Pretty 
simple.  The only quibble is that
we usually calculate power transferred to the antenna so in that case we 
are dealing with E^2 so
a factor of 4 or 12dB just like I said, not 15.

>>6. One last thing we haven't thought about is, what happens now when a 
>>conductor is placed
>>above the embedded microstrip?  Say a component leadframe?  The added 
>>dielectric now works
>>in a different manner us when we make "conductive changes" in near 
>>proximity to the embedded
>>microstrip. Oops! We didn't account for that did we?  The field lines 
>>will change direction more
>>strongly than before!  Could it affect the measurements?  Of course.
>
>Absolutely, life does get more complicated.  But, still we are still far 
>better off than if the trace were on the surface.  Fringing, fringing, 
>fringing.

Oh, yes, it does... sometimes.  But I'd "choose wisely" (taking a line from 
Indiana Jones).  That chalice
you're holding might not be the "Holy Grail".  What seems like an advantage 
might just be a curse especially
in the loosely coupled case.  Fringing, fringing, fringing....

Regards,

Michael


------------------------------------------------------------------
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.org

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:

  1. Home
  2. si-list
  3. Archive
  4. 02-2004