[SI-LIST] Re: Fw: Re: S-parameter simulation in Time Domain
- From: Steve Corey <steve@xxxxxxxxxxxxxx>
- To: si-list@xxxxxxxxxxxxx
- Date: Mon, 19 Nov 2001 13:07:55 -0800
Hi Larry -- In my experience, the pivotal issues in good eye simulation
are accurate driver models and accurate interconnect models. It sounds
like you have the former covered in SPICE. Our IConnect software helps
you extract accurate SPICE models for any interconnect elements for
which you have measured data, including lossy transmission lines.
As you have stated, using SPICE clearly allows you to rigorously
incorporate the nonlinearity inherent in your drivers into your
simulation, as long as you have accurate models of each. An earlier
statement that it is necessary to go to a linear simulation in the
presence of loss and dispersion is not accurate -- nonlinear drivers and
lossy, dispersive interconnects can definitely be simulated together in
SPICE, resulting in accurate eye prediction.
Unfortunately, I'm not sure where my reply fits on your requested A-F
grading scale...
-- Steve
-------------------------------------------
Steven D. Corey, Ph.D.
Time Domain Analysis Systems, Inc.
"The Interconnect Modeling Company."
http://www.tdasystems.com
email: steve@xxxxxxxxxxxxxx
phone: (503) 246-2272
fax: (503) 246-2282
-------------------------------------------
atspeed wrote:
> Hello Mr. Smith.
>
> Of all the many ways that I have seen people ask this question, you've asked
> it the most clearly. Let me try to give you an equally clear answer, while
> trying to be brief.
>
> First of all and fundamentally, S-parameters and nonlinear models DO NOT
> mix. The reason for that is that NONLINEAR means nonlinearly dependent on
> magnitude. When a signal is decomposed into its Fourier frequency
> components, which is what you do to invoke S-parameter modeling and
> analysis, it loses its time domain MAGNITUDE until the series is regrouped
> back into time domain. Solid state devices are nonlinear devices and, for
> that reason, do not lend themselves to S-parameter representation
> accurately.
>
> Having said that, you know very well that there is an approximation to
> everything. There are black boxes and gray boxes, there are iterative
> procedures, there are so-called small signal models and a myriad of
> techniques that mathematics has explored to give a good approximation to
> this fundamental incompatibility of frequency dependence of characteristics
> and nonlinearity of devices: active devices (transistors, diodes) in their
> large signal mode (digital as opposed to microwave) do not have an accurate
> S-parameter representation; passive components (which are linear) have
> characteristics (skin effect, dielectric loss, dispersion, etc) that do not
> lend themselves to time domain representation but are very accurately
> represented and characterized in the frequency domain (S-parameter).
>
> Because it is fundamental, the dilemma will always exist to a degree but
> therein also lies the answer. The answer is in making a choice:
>
> 1) When a net is largely dominated by reflections (even frequency dependent
> reflections such as short transmission line discontinuities, capacitors,
> inductors, etc), you can remain in the time domain. Your
> transmitter/receiver models have been made for this. Even mild skin effect
> is approximated by small resistors. S-parameters for these components have
> RLC and transmission line equivalents and everything is hunky dory in the
> time domain. That's where SPICE works best. You can ask for SPICE models for
> everything: connectors, packages, vias, bond wires, discretes (of course),
> etc. The accuracy of your sim is substantially determined by the
> approximations of the frequency dependent RESISTANCES by simple resistances
> and, of course, by the accuracy of the models themselves.
>
> 2) When a high speed net is largely dominated by loss and dispersion (skin
> effect, fr4 loss, cable loss {combination of dielectric and skin effect
> losses}), you need to migrate to an S-parameter analysis. Incidentally, all
> discrete components (RLC) have exact S-parameter equivalents. Thus, whereas
> the accuracy of the simulation in 1) is determined by the validity of the
> lossy and dispersive models, the accuracy within a full S-parameter sim is
> determined how well is the source isolated from the receiver with respect to
> high frequency components. Large backplanes (larger than a foot) and cables
> operating at high speed fit this class. Depending on the propagation
> lengths, these effects start dominating at a few hundred Mb/s. As an
> example, at a Gb/s, 10 to 12" of a 7 mil stripline trace starts being in
> this regime. With cables, the effects are prominent even below 100 Mb/s.
>
> To make this story of reasonable length, let me steer you first to
> www.atSpeed.net (and the literature therein) and to the internet broadcast
> that should still be posted at the ENEN studio: www.
> netseminar.com/index.cgi?sem_num= 51 4&source=AGILENT. This will introduce
> you to the Oculus software which has been specifically designed for
> accurately simulating loss and dispersion with means for various forms of
> signal conditioning, exactly the class of problems described in 2). What's
> more, it's the most inexpensive solution to a difficult class to simulate
> accurately. Please write to info@xxxxxxxxxxx (or to me, directly) for any
> clarification and on how to use Oculus with the nonlinear models of devices
> that you have. If you can give additional (non confidential details) of the
> cases that you are trying to sim, that will help. There are many validations
> of designs that exist that are not posted on the site.
>
> I hope this helps and I apologize for being so professorial and tutorial on
> items that you may already know from way back when.
>
> Henri Merkelo, Director
> Ultrahigh Speed Digital Electronics
> hmi@xxxxxxxx
> atSpeed@xxxxxxxxxxx
>
>
> ----- Original Message -----
> From: "mmunroe" <To: "Bill MacKillop" <>; "Jason Roe"
> <jroe@xxxxxxxxxxxxxxxxxx>; "Thomas Gneiting" <Thomas.Gneiting@xxxxxxxx>;
> "Roland Mödinger" <RMoedinger@xxxxxxx>
> Cc: "Henri Merkelo" <atSpeed@xxxxxxxxxxx>
> Sent: Sunday, November 18, 2001 11:35 PM
> Subject: [SI-LIST] Re: S-parameter simulation in Time Domain
>
>
> Bill and Jason,
>
> There have been several recent questions and responses relating to mixed
> s-parameter and SPICE modeling.
>
> -------------Forwarded Message-----------------
> From: INTERNET:samir_aboulhouda@xxxxxxxxxxx,
> INTERNET:samir_aboulhouda@xxxxxxxxxxx
> To: [unknown], INTERNET:Date: 11/16/101 12:32 PM
> RE: [SI-LIST] Re: S-parameter simulation in Time Domain
>
> Bonjour,
>
> With ADS, it is possible to use a black-box-S-parameter-model and to
> perform
> a time domain analysis (or vice-versa).
>
> The black-box-S-parameter-model could be anything: from a highly non-linear
> device such as: a Laser diver model or Lasers models (something I do very
> often), or linear devices (parasitic, Transmission lines, cap,
> inductance...).
>
> I have to say that sometimes (euphemism) the convergence is not easy to
> obtain without playing with the convolution parameters (ADS2001 version
> looks like more stable than the 1.5). It is "necessary" to set the correct
> set of parameters in order to obtain (physically correct results I meam).
> But over all, the method gives a good sims results. Actually, without the
> inclusion of the black-box-S-parameters-model, I found it very difficult
> (may be "impossible" is the correct word) to fit the Sims to the
> measurements, and indeed the eye openness (or closure) is affected by the
> black-box-S-parameters model. In my case, the data rate is much higher that
> 1G.
>
> I'll be glad to give more details if necessary..
>
> Rgds
>
> Samir
>
> -----Original Message-----
> From: Larry Smith [mailto:]
> Sent: Friday, November 16, 2001 4:56 PM
> To: si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] S-parameter simulation in Time Domain
>
> A similar question was asked on SI-list several weeks ago but did not
> get a clear answer. Does anybody have a good way to simulate
> S-parameters in the time domain?
>
> We often have silicon driver and receiver models that are composed of
> proprietary (BSIM or otherwise) transistor by transistor nonlinear
> elements best handled by HSpice, Smart Spice or similar software. We
> want to know the openness of an eye diagram at the far end of a
> transmission path. The transmission path involves two electronic
> packages, printed circuit boards with vias and a connector between
> boards. We may have S-parameter models for some of the components and
> lumped element models for others. The characteristics of nearly all of
> the components are a function of frequency.
>
> The response to the question several weeks ago involved methods for
> obtaining S-parameter models. We can get S-parameter models form
> measurement, extraction from materials and geometries, the guy down the
> hall or the company across the street. The question is, "what do we do
> with the models after we get them?" We want to know the openness of
> the eye diagram at 1 GB/sec or more. We want to know the response of a
> receiver built from nonlinear circuit elements to that eye diagram.
> Does anybody have a good success story in this area? Does anybody have
> software that claims to do this?
>
> Grading scale:
>
> A - Has model to hardware correlation for eye diagram.
> B - Can simulate but does not exactly match lab waveforms.
> C - Above, but not with proprietary spice models.
> D - Simulator has capability but nobody has ever tried it.
> F - Software available in next release.
>
> regards,
> Larry Smith
> Sun Microsytems
>
> ------------------------------------------------------------------
> 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 archives are viewable at:
> //www.freelists.org/archives/si-list
> or at our remote archives:
> http://groups.yahoo.com/group/si-list/messages
> 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 archives are viewable at:
> //www.freelists.org/archives/si-list
> or at our remote archives:
> http://groups.yahoo.com/group/si-list/messages
> Old (prior to June 6, 2001) list archives are viewable at:
> http://www.qsl.net/wb6tpu
>
>
>
>
>
>
> ----------------------- Internet Header --------------------------------
> Sender: si-list-bounce@xxxxxxxxxxxxx
> Received: from avenir.dhs.org (168-215-193-75.dslindiana.com
> [168.215.193.75] (may be forged))
> by siaag1ab.compuserve.com (8.9.3/8.9.3/SUN-1.12) with ESMTP id
> MAA08860
> for <mmunroe@xxxxxxxxxxxxxx>; Fri, 16 Nov 2001 12:32:52 -0500 (EST)
> Received: from avenir.(none) (localhost [127.0.0.1])
> by avenir.dhs.org (FreeLists Mail Multiplex) with ESMTP
> id 5D205108BC9; Fri, 16 Nov 2001 12:26:04 -0500 (EST)
> Received: with LISTAR (v1.0.0; list si-list); Fri, 16 Nov 2001 12:22:50
> -0500 (EST)
> Delivered-To: si-list@xxxxxxxxxxxxxx
> Received: from msgbas1.net.europe.agilent.com
> (msgbas1x.net.europe.agilent.com [192.25.19.109])
> by avenir.dhs.org (FreeLists Mail Multiplex) with ESMTP id
> D4D7F108A7E
> for <si-list@xxxxxxxxxxxxx>; Fri, 16 Nov 2001 12:22:49 -0500 (EST)
> Received: from msgrel1.germany.agilent.com (msgrel1.germany.agilent.com
> [134.40.2.57])
> by msgbas1.net.europe.agilent.com (Postfix) with ESMTP id B406B81C
> for <si-list@xxxxxxxxxxxxx>; Fri, 16 Nov 2001 18:28:38 +0100 (MET)
> Received: from pollux.germany.agilent.com (pollux.germany.agilent.com
> [134.40.35.20])
> by msgrel1.germany.agilent.com (Postfix) with SMTP id 66A3755E
> for <si-list@xxxxxxxxxxxxx>; Fri, 16 Nov 2001 18:28:40 +0100 (MET)
> Received: from 134.40.35.20 by pollux.germany.agilent.com (InterScan E-Mail
> VirusWall NT); Fri, 16 Nov 2001 18:28:40 +0100
> Received: by pollux with Internet Mail Service (5.5.2653.19)
> id <W8FWS3C7>; Fri, 16 Nov 2001 18:28:40 +0100
> Message-ID:
> <A7C6F8C32798D311A4E20090276508B302ACA7CB@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
> From: samir_aboulhouda@xxxxxxxxxxx
> To: si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Re: S-parameter simulation in Time Domain
> Date: Fri, 16 Nov 2001 18:28:38 +0100
> MIME-Version: 1.0
> X-Mailer: Internet Mail Service (5.5.2653.19)
> Content-type: text/plain; charset=iso-8859-1
> Content-Transfer-Encoding: 8bit
> X-archive-position: 1325
> X-listar-version: Listar v1.0.0
> Sender: si-list-bounce@xxxxxxxxxxxxx
> Errors-To: si-list-bounce@xxxxxxxxxxxxx
> X-original-sender: samir_aboulhouda@xxxxxxxxxxx
> Precedence: normal
> Reply-To: samir_aboulhouda@xxxxxxxxxxx
> X-list: si-list
>
>
>
> ------------------------------------------------------------------
> 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 archives are viewable at:
> //www.freelists.org/archives/si-list
> or at our remote archives:
> http://groups.yahoo.com/group/si-list/messages
> Old (prior to June 6, 2001) list archives are viewable at:
> http://www.qsl.net/wb6tpu
>
>
>
>
--
-------------------------------------------
Steven D. Corey, Ph.D.
Time Domain Analysis Systems, Inc.
"The Interconnect Modeling Company."
http://www.tdasystems.com
email: steve@xxxxxxxxxxxxxx
phone: (503) 246-2272
fax: (503) 246-2282
-------------------------------------------
------------------------------------------------------------------
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 archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
