Hi Hung,
If the extraction leads to non-causal and non-passive S-parameters, there is
something wrong with either process or tools. You have to inspect both -
even small violation should not be allowed.
If you get a connector or cable model that have substantial violation of
passivity and reciprocity, request another model. Validate it before use, it
ma be fixed by brute force approach with substantial degradation of
accuracy.
Considering the quality metrics, they are specified by IEEE P370 standard
committee https://standards.ieee.org/project/370.html
There are metrics both in frequency and time domains.
Metrics for the passivity and reciprocity in frequency domain has the
following ranges:
Good (green) - [100, 99.9);
Acceptable (blue) - [99.9, 99);
Inconclusive (yellow) - [99, 80);
Poor (red) - [80, 0];
Models with the Inconclusive and Poor metrics have to be further inspected -
at which frequency bandwidth the violations take place and whether the
bandwidth can be reduced, to restore the model quality. See examples at
http://www.simberian.com/Presentations/Shlepnev_S_ParameterQualityMetrics_Ju
ly2014_final.pdf
Similar metrics in time domain are introduced in the standard.
The causality quality of the models can be evaluated with the preliminary
rotation metric, metric based on the rational approximation as well as
metric based on single bit response. See more in the standard description -
the P370 committee also provides matlab scripts, to compute some metrics.
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 to Design Predictable Interconnects!
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 Hung Dang
Sent: Monday, July 22, 2019 4:35 PM
To: shlepnev@xxxxxxxxxxxxx
Cc: farhilltramper@xxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Bandwidth of channel sparameter vs EH/EW
Dear Yuriy,
Sometimes, when I extract S-parameter of multi-layer PCB(DDR4 or PCIE4),
the S-para can be non-causality or non-passive.
or I also meet a similar issue when I get connector or cable models from
the vendor.
What do I need to be noted so that I can get the good S-parameter? and how
to verify the quality of S-para in both frequency and time domain?
Thank you so much.
Vц=o Th 3, 23 thg 7, 2019 vaл-o luл│c 05:33 Yuriy Shlepnev <
shlepnev@xxxxxxxxxxxxx> д-цё viА│©t:
Hi Tramp,on
There are three contributing factors that may explain the outcome.
First, the actual power spectral density of the signal you used in the
experiment. The spectrum will depend not only on the data rate, but also
the rise time. If you set the rise time unrealistically small, it willused
extend the signal spectrum and will create dependency on the bandwidth
in the simulation.affect
Second, the channel losses. Shorter links with lower losses have more
dependency on the simulation bandwidth. If you are experimenting on a very
short of idealized link, the results will be very sensitive to the
bandwidth.
Third, algorithm used to compute the eye diagram. If it is based on DFFT,
the discretization as well as the extrapolation of S-parameters will
the result. You will never know what it is. If it is based on the rationalhttp://www.simberian.com/Presentations/SLIDES_112G_Panel_EMAnalysis_rev3.pdf
compact modeling, check the model quality and investigate how the rational
model behaves outside of the original bandwidth - resample it to higher
frequencies and see if it extracts the delay and attenuates the harmonics
(good) or not (bad).
Another important factor is the bandwidth of every component model that
goes
into the channel model. Extending the simulation bandwidth beyond the
bandwidth of the component models does not make much sense.
Actually, if everything done properly, the investigation of the eye (or
single bit response) should be done to identify the simulation bandwidth.
The outcome may be not so obvious - see example at
snp
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 to Design Predictable Interconnects!
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 Tramp
Sent: Sunday, July 21, 2019 12:29 PM
To: si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Bandwidth of channel sparameter vs EH/EW
Dear SI experts,
It seems that the bandwidth of channel sparameter has influence on the eye
result.Taking 10GKR for example, when comparing the eye obtained w/ the
of the BW of 50GHz, 60GHz, and 100GHz, the trend I observed so far is as
follows:
The 50GHz snp gets the smallest EH + largest EW, and vice versa for the
100GHz snp. The eye obtained w/ 60GHz has the mediate EH/EW in between,
Does that make sense? If, anyone can explain why?
--
Best Regards,
Tramp
------------------------------------------------------------------
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
