[SI-LIST] Re: Dk/Df value is increased very much after fabric progress
- From: "Yuriy Shlepnev" <shlepnev@xxxxxxxxxxxxx>
- To: <Cristian_Filip@xxxxxxxxxx>, "'Bert Simonovich'" <bertsimonovich@xxxxxxxxxx>, <dmarc-noreply@xxxxxxxxxxxxx>, <Gert.Havermann@xxxxxxxxxxx>, <si-list@xxxxxxxxxxxxx>
- Date: Fri, 18 May 2018 09:34:24 -0700
This is all about the predictability of the interconnect behavior. The key
is the process to achieve it systematically and the most important step is
to benchmark the model with the measured data.
Here is a simple test structure with all details - slide #9 at
http://www.simberian.com/AppNotes/Material_Identification_StepByStep_Diff_20
17_03.pdf
The cross-section geometry is well defined and both S-parameters and TDR are
available for that structure - no room for adjustments.
Transmission line model should predict insertion loss and phase delay of the
differential/odd and common/even modes. This is necessary condition to be
able to predict FEXT and NEXT.
Another condition is the correlation of the differential and common mode
impedance observed on TDR - that includes the lowest impedance and the
impedance trend. That should provide correlation in the reflection
parameters. More cross-sections with all geometry details and measured data
are available at
http://www.simberian.com/Presentations/SLIDES_14_40GHzPCBInterconnectValidat
ion_Marin.pdf - this is complete report from our DesignCon2018 paper. In all
cases, the Dk/Df reported by the laminate manufacturer were very close to
the "identified" values (taking into account resin content and fabric
style). That should be the common sense test.
Note that 2D solvers may give very different answers for a simple problem,
as was shown in the thread on the accuracy of the solvers. A field solver
error is usually translated into wrong Dk/Df that are far off the specs.
There is no problem with that if such model provides consistent correlation
in the magnitude and phase of the S-parameters - transmission, reflection
and cross-talk for any length of the transmission line. This is usually not
the case. Moreover, such models may work only for a particular cross-section
and it will be off if geometry changed.
Best regards,
Yuriy
Yuriy Shlepnev, Ph.D.
President, Simberian Inc.
2629 Townsgate Rd., Suite #235, Westlake Village, CA 91361, USA
Office +1-702-876-2882; Fax +1-702-482-7903
Cell +1-206-409-2368; Virtual +1-408-627-7706
Skype: shlepnev
www.simberian.com
Simbeor Accurate, Productive and Cost-Effective Electromagnetic Signal
Integrity Software
2010 and 2011 DesignVision Award Winner, 2015 Best In Design&Test Finalist
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On
Behalf Of Filip, Cristian
Sent: Thursday, May 17, 2018 10:16 AM
To: Bert Simonovich; dmarc-noreply@xxxxxxxxxxxxx;
Gert.Havermann@xxxxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Dk/Df value is increased very much after fabric
progress
Hi Bert,
If you can clarify which release of Hyperlynx includes a causal model,
then I would endorse that to.
Starting with HyperLynx VX2.3, that was released in February 2018, the
default model type for Surface Roughness is "Cannonball" as described in the
joint DesignCon 2018 paper. In addition to it we support several other
models such as Hammerstad and Modified Hammerstad.
My appologies if I offended anyone. It was not intentional.
I personally did not felt offended, but I think that it is a matter of
fairness here. If you want to learn more about HyperLynx please feel free to
contact us any time and we will always be happy to discuss with you...
Moreover, you can check for updated information on our website:
https://www.mentor.com/pcb/hyperlynx/
Thank you,
Cristian
-----Original Message-----
From: Bert Simonovich [
mailto:bertsimonovich@xxxxxxxxxx] ;
Sent: Thursday, May 17, 2018 10:55 AM
To: Filip, Cristian; dmarc-noreply@xxxxxxxxxxxxx;
Gert.Havermann@xxxxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: RE: [SI-LIST] Re: Dk/Df value is increased very much after fabric
progress
Hi Cristian,
At the time of our paper, Hyperlynx did not have a causal metal model for
released version of software, and as of my posting, I did not know if
current release of Hyperlynx included a causal metal model, so that's why I
did not include it. If you can clarify which release of Hyperlynx includes a
causal model, then I would endorse that to.
My appologies if I offended anyone. It was not intentional.
Thank you,
-Bert
Bert Simonovich
Signal/Power Integrity Practitioner | Backplane Specialist | Founder
Email:Lsimonovich @lamsimenterprises.com
LAMSIM Enterprises Inc.
Web Site:
http://lamsimenterprises.com
Blog:
http://blog.lamsimenterprises.com/
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [
mailto:si-list-bounce@xxxxxxxxxxxxx] On
Behalf Of Filip, Cristian
Sent: 17-May-18 11:49 AM
To: dmarc-noreply@xxxxxxxxxxxxx; Gert.Havermann@xxxxxxxxxxx;
si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Dk/Df value is increased very much after fabric
progress
Hi Bert,
Maybe it would be fair to include HyperLynx in the list of tools that are
using causal metal models. After all, both of the coauthors of the DesignCon
2018 paper that you mention are HyperLynx developers...
Thank you,
Cristian
(720) 494 -1120
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [
mailto:si-list-bounce@xxxxxxxxxxxxx] On
Behalf Of Bert Simonovich
Sent: Thursday, May 17, 2018 6:30 AM
To: Gert.Havermann@xxxxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Dk/Df value is increased very much after fabric
progress
Even when you do everything mentioned by seveal people here and use the
phase as a basis to measure phase delay, and ultimately effective Dk, you
will find that the effective Dk will be higher in proportion to the
roughness of the foil.
You must apply the correction factor described in the paper for core and
prepeg separately using 10-point mean roughness (Rz) of the Matte and Drum
side of the foil, then use these numbers instead of data sheet numbers in
your field solver simulator.
But this is only one part. The simulator must also use a causal metal model,
like that found in Simbeor or HFSS to get best correlation.
In our subsequent DesignCon 2018 paper, I coauthored with Vladimir
Dmitriev-Zdorov and Igor Kochikov, found on my web site
(
http://lamsimenterprises.com/), describes a causal metal model and showed
that when we applied effective Dk due to roughness and use a causal metal
model to measured data we got excellent correlation with insertion loss,
phase delay and impedance. This continues to hold true with other stackups
and measurements I have done since then.
Bert Simonovich
Signal/Power Integrity Practitioner | Backplane Specialist | Founder
Email:Lsimonovich @lamsimenterprises.com LAMSIM Enterprises Inc.
Web Site:
http://lamsimenterprises.com
Blog:
http://blog.lamsimenterprises.com/
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [
mailto:si-list-bounce@xxxxxxxxxxxxx] On
Behalf Of Havermann, Gert
Sent: 17-May-18 3:00 AM
To: si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Dk/Df value is increased very much after fabric
progress
In addition to the other comments you have to check if your process of
measuring dk is correct. Do you measure it based on impedance or phase? The
impedance based measurement isn't precise because surface roughness and
geometrical shape of the impedance cell take a big part of it. Phase
measurement (two different length traces and measure the phase difference)
will give more accurate results.
Alexanders point is very important too because some prepregs are available
with different Resin content but same glass style, so if you only specify
the glass style, you might be surprised in the end.
And maybe the supplier simply used a different material (happened to me some
times).
BR
Gert
----------------------------------------
HARTING Stiftung & Co. KG | Postfach 11 33, 32325 Espelkamp |
www.HARTING.com Persönlich haftende Gesellschafterin:
HARTING Führungsstiftung | Amtsgericht München | HRA 108479 | München
Vorstand: Dipl.-Kfm. Philip F. W. Harting (Vorsitzender), Dipl.-Kffr. Maresa
W. M. Harting-Hertz, Dipl.-Kfm. Dr.-Ing. E. h. Dietmar Harting, Dipl.-Hdl.
Margrit Harting, Dr. rer. nat. Frank Brode, Dipl.-Ing. (FH)
Dipl.-Wirtsch.-Ing. (FH) Andreas Conrad, Dr. iur. Michael Pütz Sitz der
Gesellschaft: Espelkamp | Amtsgericht Bad Oeynhausen | HRA 9021 | UST-ld Nr.
DE812136745
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx <si-list-bounce@xxxxxxxxxxxxx> On Behalf
Of Hung Dang
Sent: Thursday, May 17, 2018 8:17 AM
To: Alexander Ippich <Alexander.Ippich@xxxxxxxxxxxxxxx>
Cc: istvan.novak@xxxxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Dk/Df value is increased very much after fabric
progress
Thank you so much, Experts.
As Mr Alexander said that, I often get stack-up from Fab house. And I see
that Dk/Df value is often higher than Dk/Df in datasheet.
But in this time, Dk/Df from Fab house is very smaller than datasheet and
actual measurement value.[Dk = 3.5 (Fab house) --> Dk = 4.2(Datasheet) -->
Dk = 4.7 (measurement) at 4Ghz].
The big different Dk will make the gap between design and actual board such
as trace match length in DDR, estimated SerDes loss channel...
By the way, Thanks much Bert, I have read your paper and understand more
Roughness, Fabric process...It helps me configure more accurate stack-up in
3D simulation.
Thanks.
2018-05-17 12:11 GMT+07:00 Alexander Ippich <
Alexander.Ippich@xxxxxxxxxxxxxxx>:
Folks,
I would guess, that the reason is much simpler:
Hung mentioned, that he has picked the Dk/Df values from the data sheet.
Typical datasheets list values that are applicable to a certain
laminate thickness / construction. Very often, this is for a 0.060"
laminate (as lot of IPC-TM650 tests are based on that thickness) and
also a heavy glass weave (like 7628).
If he is now measuring on an actual board that is using a different
construction, Dk/Df values can differ significantly. This is why a lot
of laminate suppliers (including the company that I work for) offer
tables that show Dk/Df values for different constructions.
Let me just pick values from our 370HR material: we show resin
contents as low as 41% (Dk 4.44, Df 0.0134) and as high as 71% (Dk
3.89,
Df 0.0181).
Bottom line, talk to your PCB supplier or a technical contact from the
laminate supplier that provided the material. Lot of things can be
explained, when understanding which materials / construction /
processes were used.
Best regards,
alex
------------------------------------------------------------
------------------------------------------------------------
----------------
Alexander Ippich
Technical Director, Signal Integrity & Advanced Technology Product
Manager RF/Microwave OEM Marketing Europe
e-mail: alexander.ippich@xxxxxxxxxxxxxxx
web: www.isola-group.com
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx]
On Behalf Of Istvan Novak
Sent: Thursday, May 17, 2018 4:19 AM
To: hungdn.hcmut@xxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Dk/Df value is increased very much after fabric
progress
If you search the SI-list archives you will find several threads on
related topics.
Because of unavoidable tolerances and second-order effects, you can
never expect very good agreement unless you extract the Dk/Df values
very carefully yourself. However, a 28.7% difference seems to be too
big, which begs the question what is the basis of your measured Dk and
Df values?
Regards,
Istvan Novak
On 5/16/2018 9:21 PM, Hung Dang wrote:
Dear Expert,
When I run SI simulation, I apply Dk/Df of PCB material from datasheet.
After fabric progress, I measured on the actual board, Dk is
increased 28.7% (3.65 --> 4.7) and 41.6%(0.012 --> 0.017) for Df
when comparing
with
value in datasheet.
So I have concerns below:
1/. What is the Dk/Df worst value which I should apply to simulation
phase?
2/. Could you help to list some application note or paper which talk
about
this issue?
Thank you so much.
------------------------------------------------------------------
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 forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Disclaimer:
Any quotation or order confirmation provided with this communication
is subject to Isola's standard terms and conditions of sale which can
be found here http://www.isola-group.com/about-us/terms-conditions/
--
----------------------------------------------------------------------------
---
MS Hung-Dang Ngoc
Email: hungdn.hcmut@xxxxxxxxx <phandainghiamnc@xxxxxxxxx>
Phone: +84 1682 405 564
Skype/Facebook: hungdn.hcmut
------------------------------------------------------------------
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 forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
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 forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
---
This email has been checked for viruses by AVG.
http://www.avg.com
------------------------------------------------------------------
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 forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
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 forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
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 forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
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 forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
