Good point. With chip scale package, top two layers of the PCB in those
pictures are acting like the core layer in multi-layer packages.
As such it will be the same discussion as replacing IDC at the top of the
package (aka the others).
Chris Cheng
Distinguished Technologist , Electrical
Hewlett-Packard Enterprise Company
+1 510 344 4439/ Tel
chris.cheng@xxxxxxx / Email
4209 Technology Dr
Fremont, CA 94538
USA
-----Original Message-----
From: Smith, Larry [mailto:larrys@xxxxxxxxxxxxxxxx] ;
Sent: Tuesday, March 22, 2016 7:43 PM
To: Cheng, Chris <chris.cheng@xxxxxxx>; alfred1520list@xxxxxxxxx; SI-LIST
<si-list@xxxxxxxxxxxxx>
Subject: RE: [SI-LIST] Re: X2Y equivalent circuit model
Chris - Your comments make a lot of sense for servers with thick boards and
through-hole vias. But there is a big difference between the server-space and
the mobile-space. On mobile platforms, power and ground planes might be on the
1st two layers of the board. Dielectric thickness is often 2 mils or less.
Microvias are only needed on one terminal of the cap to establish mounting
inductance. X2Y and 3-terminal caps certainly have a place on the mobile
platform boards.
Please do a google search on "iphone s6 teardown" or go to link
https://www.ifixit.com/Teardown/iPhone+6s+Teardown/48170 . Scroll down to ;
step 19 and 20 and zoom in. A picture is worth 10,000 words. You can see how
precious board area is. You can see the 3-terminal caps all over the place.
Or, do a google search on "Galaxy S7 teardown" or go to link
https://www.ifixit.com/Teardown/Samsung+Galaxy+S7+Edge+Teardown/56845 . Scroll ;
down to step 8 and zoom in to see the 3 terminal caps on the backside of the
Qualcomm Snapdragon 820.
Board area is incredibly valuable real estate. We need to get the very lowest
inductance path from die circuits to decoupling caps in order to make the most
efficient use of board area. This is why X2Y caps belong on a board.
Best regards,
Larry Smith
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On ;
Behalf Of Cheng, Chris
Sent: Tuesday, March 22, 2016 5:28 PM
To: alfred1520list@xxxxxxxxx; SI-LIST <si-list@xxxxxxxxxxxxx>
Subject: [SI-LIST] Re: X2Y equivalent circuit model
Thank you, can we move on to discuss the merits of using X2Y vs. IDC on
packages then ?
How does it compare with IDC at the bottom of packages (aka Intel style) ?
How does it compare with IDC at the top of packages (aka the others) ?
Chris Cheng
Distinguished Technologist , Electrical
Hewlett-Packard Enterprise Company
+1 510 344 4439/ Tel
chris.cheng@xxxxxxx / Email
4209 Technology Dr
Fremont, CA 94538
USA
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On ;
Behalf Of Alfred Alfred
Sent: Tuesday, March 22, 2016 5:07 PM
To: SI-LIST <si-list@xxxxxxxxxxxxx>
Subject: [SI-LIST] Re: X2Y equivalent circuit model
At the risk of taking words out of context, can we draw a conclusion to the
thread by these:
Chris Cheng> if you properly reference your high speed signals together
with return vias and you do an ok job on package decoupling, you only need
to worry about decoupling below 1-200MHz for caps in PCB. Let the package
decoupling and on-die decoupling handle frequencies above that. No fancy
embedded planes nor exotic capacitors need on PCB.
Scott McMorrow> Chris. Yep, I agree with everything that you have said.
Lee Ritchey (wrote 8 days earlier)> From a PDS point of view, they don't
offer anything more than an 0402 as you point out.
So while there were debates on how the conclusion was reached, there is no
debate on the conclusion?
Best Regard,
Alfred
On Tue, Mar 22, 2016 at 4:42 PM, Scott McMorrow <scott@xxxxxxxxxxxxx> wrote:
Chris
Yep, I agree with everything that you have said.
Scott
Scott McMorrow
R&D Consultant
Teraspeed Consulting - A Division of Samtec
16 Stormy Brook Rd
Falmouth, ME 04105
(401) 284-1827 Business
http://www.teraspeed.com
On Tue, Mar 22, 2016 at 7:21 PM, Cheng, Chris <chris.cheng@xxxxxxx> wrote:
Larry and Scott,target
Somehow I never quite got my answer. I always thought X2Y should be
to substitute IDC on packages where the built up layers has micro-viasand
it is on the right side of the package resonance to even necessitate a >1-200MHz
1-200MHz target decoupling frequency.
I may sound like a broken record, but I believe if you properly reference
your high speed signals together with return vias and you do an ok job on
package decoupling, you only need to worry about decoupling below
for caps in PCB. Let the package decoupling and on-die decoupling handleneed
frequencies above that. No fancy embedded planes nor exotic capacitors
on PCB. So why are we spending a 100 threads to discuss X2Y applicationsin
PCB rather than packages where it is supposed to be used ?capacitors.
And before you say some thousand plus pins processor packages with 0.8mm
pitch BGA that has Swiss cheese planes need it, I can't fix that if you
shot yourself in the foot.....
Chris Cheng
Distinguished Technologist , Electrical
Hewlett-Packard Enterprise Company
+1 510 344 4439/ Tel
chris.cheng@xxxxxxx / Email
4209 Technology Dr
Fremont, CA 94538
USA
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]
On Behalf Of Smith, Larry
Sent: Tuesday, March 22, 2016 12:59 PM
To: scott@xxxxxxxxxxxxx; Sebastian Garcia <sg-listas@xxxxxxxxxxxx>
Cc: SI-LIST <si-list@xxxxxxxxxxxxx>
Subject: [SI-LIST] Re: X2Y equivalent circuit model
Hi Scott - I need to comment on this one. There are some subtleties that
will help us envision the inductance mechanisms for decoupling
For horizontal structures (i.e. power planes and MLCC capacitors) weas
should think of inductance in terms of the "number of squares" for the
material. The standard geometry capacitors have about 2 squares. Common
sizes are 0201, 0402, 0603, 0805, 1206, etc. They are all twice as long
they are wide and contribute about 2 squares of inductance. The currentto
path is through the capacitor and the return current path is somewhere on
the board. BTW, this works for ESR also. The horizontal contribution
both inductance and resistance goes as the number of squares.the
But horizontal is not the whole story. Current must go vertical in the
capacitor as well as horizontal. The larger capacitors tend to be taller
and some of the current has to go farther away from the return path on
board, therefor larger size capacitors have somewhat larger inductance.might
The increase in capacitor inductance with size is not as much as you
think because of the "number of squares" argument.terminal
The X2Y capacitors really do have fewer squares. They have a Vss
in the middle of the cap structure and Vdd terminals on the two ends(these
can be reversed if you want). This separates the cap into two portionsterms
(two squares) which are in parallel, effectively making 1/2 square. This
should be compared to the 2 squares of a standard geometry cap. BTW,
reverse geometry caps ( 0204, 0306, etc) are 1 square long and 2 squares
wide, also resulting in 1/2 square. They are on par with X2Y caps in
of the intrinsic inductance associated with the cap body.so
As you and other SI-list posters have pointed out, it is really all about
the vias. In many circumstances, the via loop inductance dominates and
the intrinsic inductance associated with the cap body becomeshave
insignificant. This is often the case in server boards that tend to
through-hole vias. But in the mobile space where microvias are the rule0402
rather than the exception, the board mounting inductance becomes small
making the intrinsic capacitor inductance a relatively bigger part of the
problem. This is where alternative cap geometries (X2Y, 3 terminal,
reverse geometry) can really pay off. You have to earn the right to use
low inductance capacitors by having low inductance mounting structures.
Regards,
Larry Smith
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]
On Behalf Of Scott McMorrow
Sent: Monday, March 21, 2016 8:56 PM
To: Sebastian Garcia <sg-listas@xxxxxxxxxxxx>
Cc: SI-LIST <si-list@xxxxxxxxxxxxx>
Subject: [SI-LIST] Re: X2Y equivalent circuit model
sebastian,
This is just physics. An 0201 capacitor is half the physical size of an
0402. The loop area through the capacitor is half the loop area of an
capacitor. Therefore, the capacitor contribution to bypass inductancefor
an 0201 capacitor is one half of an 0402.of
Loop inductance below the capacitor will depend on the via configuration
used to connect to the power and ground planes, along with the placement
the planes on the z-axis. I can engineer pretty much any inductance youabsolutely
want with enough vias and optimized spacing. The capacitor has
nothing to do with what happens inside the board.sg-listas@xxxxxxxxxxxx>
When the vias dominate inductance ... you need more vias, or better
placement of the planes and capacitors with respect to the load. No
capacitor fixes that problem.
What an X2Y capacitor does do is to have some amount of mutual inductance
cancellation near the capacitor terminals, which does lower inductance
above the plane, but, the biggest benefit of the X2Y capacitor is the
terminal configuration, which facilitates the use of 6 vias in a nearly
minimum energy configuration that is low inductance.
Scott McMorrow
R&D Consultant
Teraspeed Consulting - A Division of Samtec
16 Stormy Brook Rd
Falmouth, ME 04105
(401) 284-1827 Business
http://www.teraspeed.com
On Mon, Mar 21, 2016 at 4:01 PM, Sebastian Garcia <
wrote:tests.
Dave, the post you linked is technically enlightening.
Scott: Could you recommend a doc reference about more recent tests
with
0201 caps?
Best regards,
Sebas.
David Anthony wrote:
No disparaging comments here, just a clean Engineering discussionwith
about inductance and John's results:
https://www.freelists.org/post/si-list/6-layers-stackup,31
-Dave
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx]
On Behalf Of Lee Ritchey
Sent: Monday, March 14, 2016 1:00 PM
To: sg-listas@xxxxxxxxxxxx; 'SI-LIST'
Cc: John Zasio
Subject: [SI-LIST] Re: X2Y equivalent circuit model
Todd,
Well said.
As to the tests that John Zasio did, they were done on a 16 layer
PCB
plane capacitance designed into it as one would do for a high
performance design.
The testing was done as one would do to see how the PDS impedance
looks with either of the two capacitors connected. The results show
that, in this configuration, there is little difference with respect
to PDS performance.
From a PDS point of view, they don't offer anything more than an
0402 as you point out.
It has been a long time since I looked at the way Steve did his
Engineering.I'll have another look and see if I can find how he gets different
answers.
Lee
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx]
On Behalf Of Sebastian Garcia
Sent: Monday, March 14, 2016 9:23 AM
To: SI-LIST <si-list@xxxxxxxxxxxxx>
Subject: [SI-LIST] Re: X2Y equivalent circuit model
Well said, Todd. No more black magic, please, just well done
interests.
Without solid engineering documents describing repeatable tests, it
turns to be a battle of professional egos spiced with commercial
bus decoupling.
Sebas.
Todd Hubing wrote:
I think what Sebastian and others would like to see is a paperconfiguration.
describing the board and the measurements that were made. The slide
presentations on the website are interesting, but they do not
describe the board stackup or the measurements that were made to
justify the conclusions. In fact, in the 10 years since this work
was done, it has been effectively demonstrated that the conclusions
related to capacitor ratio and via ratio are not justified. The
impedance of the board in this study was not significantly affected
by the connection inductance of the decoupling capacitors above
about 200 MHz. Also, the extra vias on the 0402 capacitors have
relatively little effect on the mounting inductance and do not
represent a typical or recommended mounting
I don't mean to disparage Steve's work or X2Y capacitors. I talked
to Steve when he exhibited this at the IEEE EMC Symposium. His
board does a great job of presenting X2Y capacitors in a favorable
light, which is exactly what it was designed to do. It's actually a
very clever design.
Also, the X2Y capacitors themselves represent a well thought-out
design with many useful applications.
My point is that an optimum decoupling strategy depends on many
factors including the board stackup, active device requirements and
the intended application. 0402 MLCC capacitors do an excellent job
of high-frequency decoupling in most situations if they are
properly configured for the given application. If lower inductance
is called for, there are other options including embedded
capacitance, IDC capacitors, and 0201 capacitors. I have not seen
an application where X2Y capacitors, are the best option for power
caps used.Nevertheless, I've been in this business long enough that I knowthan to say "always" or "never"
better
and I would be very interested to review any published paper that
describes measurements and/or simulations on this topic.
Todd
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx]
On Behalf Of David Anthony
Sent: Sunday, March 13, 2016 11:24 PM
To: tom@xxxxxxxxxxxxxxxxx; sg-listas@xxxxxxxxxxxx
Cc: 'SI-LIST' <si-list@xxxxxxxxxxxxx>
Subject: [SI-LIST] Re: X2Y equivalent circuit model
Sebastian, the design papers Tom referenced showing system
measurements by Steve can be found here:
http://www.x2y.com/bypass.htm#examples
Dave Anthony
X2Y
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx]
On Behalf Of Tom Dagostino
Sent: Sunday, March 13, 2016 10:52 PM
To: sg-listas@xxxxxxxxxxxx
Cc: SI-LIST
Subject: [SI-LIST] Re: X2Y equivalent circuit model
Sebastian. I'm traveling right now and don't have access to those
archives. But look at the work by Steve on the X2Y web page. I
think you will find what you are looking for there.
Tom Dagostino
Teraspeed LabsOn Mar 13, 2016 8:13 PM, Sebastian Garcia
<sg-listas@xxxxxxxxxxxx> wrote:
Tom,
Could you shed some light on this, providing the specific Steve
Weir's documents you mention, please?
Best regards,
Sebas.
Sebastian Garcia
Slabs Electronics
Buenos Aires - Argentina
+54 911 3865 1770
Tom Dagostino wrote:
Joel
The layout I saw and Steve Weir commented on had the vias so
far away from the capacitor that the measured inductance was
dominated by circuit board traces. Steve fitted about 8 proper
layouts within the confines of the layout that Lee mentioned.
I asked Lee about this and he said the layout was something he
found on the X2Y web site I think it was in about 2007.
Since then X2Y has updated the layout and the Work Steve Weir
did has shown that with fewer X2Y caps the power supply noise
was lower compared to 0402 caps. This was shown with two
layouts of FPGAs running the same code and differing only in the
My issue here is the dragging out of data that does not
represent current recommended applications as proof anything.
I think I still have one of Steve's boards. His method was well
documented with measurements. He set up unused I/O cells high
and low and attached SMAs to them for ease of measurement. So
he showed the supply as the chip saw it.
Regards
Tom Dagostino
Tom Dagostino
Teraspeed LabsOn Mar 13, 2016 11:42 AM, Todd Hubing
<hubing@xxxxxxxxxxx>
wrote:
I wouldn't characterize the Zasio paper as bad information. It
is a rare (on this mail list) example of someone supporting an
argument with actual data from a test described well enough
that anyone else could repeat it. Yuriy Shlepnev's app
note is another example of this. Both of these studies
demonstrate that, for the configurations evaluated, two
0402 capacitors are probably a better choice than an X2Y cap.
Of course, it is possible to come up with configurations or
test criteria that favor the X2Y also.
For printed circuit board decoupling, the X2Y via configuration
is not optimum. The two capacitors share a pair of ground vias
that must be relatively close to each other. If the power
planes are more than about
0.5 mm from the board surface, an IDC capacitor or a pair of
two-terminal capacitors can achieve better flux cancelation by
alternating closely spaced power and ground vias. That said, if
you're REALLY concerned about high-frequency decoupling, take
Scott McMorrow's advice and use 0201 caps.
Todd
---------------------------------------------------------------
--
Todd H. Hubing
LearnEMC.com
---------------------------------------------------------------
--