Hi Per,
no, this conclusion is absolutely NOT correct!
The people who I have heard make these statements have not given the
subject enough thought. Typically an extremely simplistic example of a
single IC on a large board where the initial charge delivery is compared
between a low and high Dk material. Granted in this theoretical case, the
very initial charge delivery will be the same. However. after a given
time, the board with the smaller Dk will run out of board area to supply
charge (whereas the board with the large Dk will still have significant
board area remaining to supply charge). When the available charge between
the low Dk planes is exhausted, if the charge delivery is not sufficient
from another source, voltage droop will occur and noise will be generated.
This would not occur in a board with a larger Dk. Thus, a board with a
higher Dk provides longer charge delivery before charge must be delivered
from an alternative source. This means that other charge sources such as
decoupling caps can have a higher inductive route if desired (further away
for example).
On an actual typical board design, there are ICs switching at the same time
that are relatively close. These ICs will fight over the available charge
in the planes. Since the lower Dk materials have to go further to find the
available charge, there will be more fighting over the available charge
than a with a higher Dk material. For example, comparing a material with a
Dk of 4 and a Dk of 16, if there are two ICs a distance d apart, the ICs
will run out of full charge delivery as soon as the distance d/2 has
occurred. Let's say this starts in t nanoseconds. The material with the
Dk of 16 will still be able to supply the full charge delivery until a time
of 2t. Thus, being able to supply full charge delivery for twice as long
as a low Dk material.
Thus, in summary, on a typical board design, the very initial supply from a
low Dk material and a high Dk material will be the same. But considering
the speed in a low Dk material is going to be ~6 inches per nanosecond, it
will not be long until the full charge delivery from the low Dk material is
compromised. This could be from the lack of available charge from the ends
of sides of a board or from trying to get charge from an area of the board
that has already been exhausted from another IC. Thus, with the exception
of an extremely very short initial period, a high Dk material will supply
charge at a faster rate and longer than a low Dk material. This will
result in less noise being generated. High Dk materials will generated
less noise than low Dk materials with all other design parameters being
equivalent.
Above was a comparison of noise generation. The high Dk materials will
also do a better job of noise dampening since they will generate more loss
in the same distance (because they travel at a lower speed). The
combination of less noise being generated and better noise dampening will
mean less noise on a high Dk board than a low Dk material. For those who
have simulation and modeling software, this is easy to see when the data is
output from Mentor Hyperlynx PI or other design tools.
Istvan Novak at Sun did some work back about 10-12 years ago where part of
his study compared different dielectric thickness and Dk materials for
power-ground cores (in the same board design). In one example, a 25 um
FR-4 material (with a Dk of ~4) was compared to a 3M material (~24 um) with
a Dk of 16. The impedance of the 3M material was less at lower frequencies
(as expected due to higher capacitance) and the very significant improved
noise dampening ability was also easy to see as the impedance spikes from
the board resonances was very well damped in the high Dk board. ( I can
forward slides to anyone interested in seeing the data.)
High Dk materials will have lower noise levels compared to low Dk materials
on the same board design. Smaller boards or boards with smaller voltage
rails will tend to amplify the noise difference between low and high Dk
materials. Boards with multiple ICs in very close proximity will also tend
to amplify the difference in noise levels between low and high Dk
materials.
It is up to the board designer to determine whether the increased
performance level of a high Dk material is needed or not and if so, whether
the increased board cost (typically higher Dk materials cost more than
lower Dk materials) is justified.
Be extremely cautious of someone telling you that Dk does not make a
difference and that charge delivery will be the same regardless of Dk.
This is clearly not the case.
Joel Peiffer
Joel S. Peiffer
3M Electronic Solutions Division
3M Center, Building 201-1E-21
St. Paul, MN 55144
Tel: (651) 575-1464
Cell: (612) 327-1983
Fax: (651) 737-4601
[log in to unmask]
"Viklund, Per"
<Per_Viklund@MENT
OR.COM> To
Sent by: <[log in to unmask]>
EmbeddedNet cc
<[log in to unmask]
org> Subject
Re: [EM] Fw: MILAERO007: What
Became of the Embedded Passive
02/06/2012 05:32 Boom?
AM
Please respond to
D-50 Embedded
Devices Committee
Forum
<[log in to unmask]
ORG>; Please
respond to
"Viklund, Per"
<Per_Viklund@MENT
OR.COM>
Hi Rick,
Good to hear from you again -I bet you long to get back into our
industry ;-)
I seem to remember an article by you (in Circuitree?) where you elaborated
on the importance of high Er dielectrics for embedded capacitance planes
and the need for a circuit to obtain charge from the plane in a short time.
-You came to the conclusion that the higher the Er, the slower the charge
propagation and thus a smaller area of the plane around a circuit was
useful.
In fact, you found that the two canceled out and there was no benefit from
increasing Er.
http://tinyurl.com/7qqog63 (word doc)
...It sure fascinated me and I wonder if anyone over the years has found
reasons why this conclusion would not be correct and if not.... is there
really a need for extremely high k materials? (for decoupling)
/Per Viklund
-----Original Message-----
From: EmbeddedNet [mailto:[log in to unmask]] On Behalf Of Rick Ulrich
Sent: den 4 februari 2012 19:17
To: [log in to unmask]
Subject: Re: [EM] Fw: MILAERO007: What Became of the Embedded Passive Boom?
Hey Dennis -
As a guy who was at the forefront of embedded passives for some years I've
thought about this a lot and I think the answer is pretty simple. It was
the lack of embeddable capacitance that did it in. No one ever figured out
how to embed a significant amount of capacitance.
Unfilled polymers gave a maximum of about half a nF/cm2 and polymers filled
with ferro powders gave up to only 2 nF/cm2. There's just not a lot you
can do with that. Anodized thin films could go to 100 nF/cm2 but were
expensive, very fragile and could not be cheaply manufactured in a PCB. No
easy trimming technology has been demonstrated.
Many of us have spent a lot of time and money working on this with little
additional progress. A major company that many of us worked for at one
time put out an RFP for 10 micro-F/cm2 a couple of years ago. I told them
"no way" and I don't think they found anyone that could.
Embedded passives will always have a place in electronics, but there seems
no prospect of widespread replacement of surface mount. I think the issues
with capacitance is what prevents this and I frankly see no solution on the
horizon.
As for me, I'm now working in analyzing greenhouse gas emissions from
agricultural operations. I had to follow the money and, as long as we have
this crazy weather and a democrat in the White House, there will be funding
for environmental research.
Best to all of my old friends in the business. I try to keep up with
embedded passive development and I appreciate seeing Per's article.
- Rick
************************************************************************************
Dr. Richard Ulrich
Professor
Department of Chemical Engineering
BELL 3202
University of Arkansas
Fayetteville AR 72701-1201
[log in to unmask]
479-575-5645
************************************************************************************
-----Original Message-----
From: EmbeddedNet [mailto:[log in to unmask]] On Behalf Of Dennis Fritz
Sent: Friday, February 03, 2012 3:36 PM
To: [log in to unmask]
Subject: [EM] Fw: MILAERO007: What Became of the Embedded Passive Boom?
I know it's been a while since anyone posted to this forum. Here is an
article from Per Viklund that may stir discussion. Also, we are coming up
to IPC Expo the end of February, and there will be several committee
meetings on Embedded Components. Maybe the leaders of those committees
want to promote their sessions on this forum.
Denny Fritz
----- Forwarded by Dennis Fritz/MacDermid/MACDERMID/US on 02/03/2012 04:33
PM -----
Subject: MILAERO007: What Became of the Embedded Passive Boom?
What Became of the Embedded Passive Boom?
While the anticipated embedded passive boom never happened, several
industry segments have adopted the technology--even depending on it to be
successful. While new and exotic to many, EP technology is mainstream to
others, especially in the medical, automotive, and military/aerospace
industries.
To read this article, go to
http://www.milaero007.com/pages/zone.cgi?a=81481
______________________________________________________________________
This email has been scanned by the Symantec Email Security.cloud service.
For more information please contact helpdesk at x2960 or [log in to unmask]
______________________________________________________________________
______________________________________________________________________
This email has been scanned by the Symantec Email Security.cloud service.
For more information please contact helpdesk at x2960 or [log in to unmask]
______________________________________________________________________
______________________________________________________________________
This email has been scanned by the Symantec Email Security.cloud service.
For more information please contact helpdesk at x2960 or [log in to unmask]
______________________________________________________________________
______________________________________________________________________
This email has been scanned by the Symantec Email Security.cloud service.
For more information please contact helpdesk at x2960 or [log in to unmask]
______________________________________________________________________
|
