[SI-LIST] Re: DesignCon presentations posted

Hi Bill,

Thank you for referencing our paper - there were more experimental
identification data in our presentation available at
http://www.simberian.com/Presentations/MaterialParameterExtractionWithSG_Des
ignCon2010_Final.pdf 
(short reference is http://kb.simberian.com/Presentations.php - see #4).
As you noticed we have investigated possibility to use wider lines to make
sure that the conductor and roughness models are correct. In theory, the
proposed extraction method works with different types of lines with
impedances that may be substantially different from 50 Ohm. In theory it
should allow identification of both conductor resistivity and roughness
model in addition to the dispersive dielectric model. In practice, the
result depends on the quality of measured data and the physical identity of
the connectors, launches and cross-sections for two lines used for the
extraction. Well optimized launches for 50-Ohm single-ended strips and
micro-strips and 100-Ohm differential strips provided more reliable data.
Non-identity of launches or cross-sections for 25-Ohm lines caused some
problems in the identification and the outcome on the roughness
identification was not clear (there are multiple ways to achieve good
correspondence of measurements with simulation). Together with Teraspeed we
are planning to spend more time on the refinement of the dielectric and
conductor roughness identification methodology. Though the existing
procedure based on generalized modal S-parameters is probably the simplest
possible and can produce models for practical applications up to 50 GHz or
100 Gb/s.

The subject of the dielectric and conductor characterization for
multi-gigabit applications may be really confusing at this point. Dozens of
papers and dozens of IPC standards is clear sign of a problem that has to be
solved. For practical applications we need not just measurements at one or
two points, but causal frequency-continuous models like the wideband Debye
(WD) model for instance. The WD model is simple and available in practically
all evolving SI tools, but unfortunately it does not work in many cases. In
particular, it does not look right for low-loss dielectrics for instance.
Multi-pole Debye models on the other hand can be used to fit Dk/Df
dependency for any laminate. Such models is more difficult to construct and
in addition they may partially include loss due to the conductor roughness
(it brings us back to the roughness identification). More sophisticated
models based on the mixtures of the dispersive epoxy and glass dielectrics
identified separately and on separation of layers with different content of
epoxy and glass may be needed. In all cases such models can be considered as
the macro-models of complex materials and the bottom line is how well they
correlate with the measured results for transmission lines over a given
frequency range and what is the impact on the data in time domain. Note that
some existing static t-line models start failing from 1-2 GHz even with the
causal WD models that is visible both in frequency and time-domains. The
reasons are roughness modeling and high-frequency dispersion.

Best regards,
Yuriy Shlepnev
www.simberian.com 


-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On
Behalf Of Bill Dempsey
Sent: Tuesday, February 16, 2010 9:15 PM
To: si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: DesignCon presentations posted

Istvan,
Ok, still absorbing a lot of great papers here... and
while doing so I seemed to have missed this one:
http://www.simberian.com/AppNotes/DesignCon2010_Paper2807.pdf  
which elaborates a method "similar" to what I proposed.б= I use the word
"similar" very loosely here. :)

In
that paper, the authors state "Conductor resistivity and RMS
measurements of roughness and roughness factor make it possible to
separate all metal losses with high confidence Б-? it is impossible to
identify the dielectric properties without such separation in the
model"б= You seem to indicate that this might be an issue (re:
cross-correlation).

So what do you (read:Sun) do with regards to
Dk/Df frequency dependence?б=б=б= What about everyone else out there?б=
What's next?б= Seems like Simbeor is addressing this in their tools --
anyone else?

...thinking...

Bill


--- On Tue, 2/16/10, Istvan Novak <istvan.novak@xxxxxxx> wrote:

From: Istvan Novak <istvan.novak@xxxxxxx>
Subject: Re: [SI-LIST] Re: DesignCon presentations posted
To: "Bill Dempsey" <cadcocat@xxxxxxxxx>
Cc: si-list@xxxxxxxxxxxxx, Istvan.Novak@xxxxxxx
Date: Tuesday, February 16, 2010, 10:02 PM

Hi Bill,

Yes, all those things you mention are doable, but the challenge is that
eventually we need a trusted model to fit to, because beyond a few hundred
MHz there is no direct way to measure the contributing factors separately.
AND the more cross-correlation we do (for instance different line trace
length, different line widths, etc) the more we rely on the elusive
assumption that everything else stays the same except the single
 parameter we intend to change.

Regards,

Istvan Novak
Oracle-SUN


Bill Dempsey wrote:
> <second posting...apologies>
> Hi Istvan,
> I just read the DesignCon paper on the loss tangent measurement and had a
question with regards to other possible techniques to measure Df (and
similarly Dk).
> Previously you built a stripline test board for measuring glass weave
effect.ц-б= The cool thing about this board was that you launched from
"inside" the board and minimized the effects of the launch structure.ц- ц-б=
Couldn't you build a similar test board with various copper line widths and
measure insertion loss on a VNA?ц-б= Once you have the insertion loss,
couldn't you then subtract out the Rdc+Rac component from the overall
insertion loss to give you a loss curve from which you could derive Df?ц-б=
Use the various line widths to minimize any weave effect.
> 
> Using this same
 board, can you not look at group delay to also measure the Dk over
frequency?
> 
> I know your paper talks about cured and uncured resin variations but in
the end, isn't it just the cured value we need?ц-б= Most tools don't address
a frequency dependent Df/Dk at this time so what do we do once we get a
frequency dependent Dk/Df anyway? I'm curious to see how others are
addressing this issue and how much their lab measurements have been off from
simulations.
> 
> And I'd like to see some posts from tool vendors who are addressing the
frequency dependence of these values.
> 
> Regards,
> Bill
> From: istvan Novak <Istvan.Novak@xxxxxxx>
> To: SI-List
>б= <si-list@xxxxxxxxxxxxx>
> Sent: Mon,
 February 15, 2010 10:52:42 PM
> Subject: [SI-LIST] DesignCon presentations posted
> 
> 
> FYI:
> 
> For those of you who could not make it to DesignCon, the Oracle-SUN
presentations are posted at http://www.electrical-integrity.com/
> 
> For the three Best Paper Award winner presentations, both the written
materials and the slides are posted:
> ц-Б--е?Introduction and Comparison of an Alternate Methodology for
Measuring Loss Tangent of PCB Laminatesц-Б--О©+
> ц-Б--е?Additional Trace Losses due to Glass-Weave Periodic Loadingц-Б--О©+
> ц-Б--е?Accuracy Improvements of PDN Impedance Measurements in the Low to
Middle Frequency Rangeц-Б--О©+
> 
> Also posted are the slides for the panel discussion "Making Sense out of
Dielectric Loss Numbers, Specifications and Test Methods"
> 
> Regards,
>
 
> Istvan Novak
> Oracle-SUN
> 
> 
>б=б=б=




      


      
------------------------------------------------------------------
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:
http://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.net

List archives are viewable at:     
                http://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:
http://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.net

List archives are viewable at:     
                http://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: