[SI-LIST] Re: Spectre S-parameter based modeling- Stability issues
- From: "Taha Amiralli" <thamiral@xxxxxxxxx>
- To: si-list <si-list@xxxxxxxxxxxxx>
- Date: Mon, 4 Dec 2006 22:54:43 -0500
Bratfest,
The explanation you have was pretty much the direction I was thinking in. I
assumed that since a processor on an integrated circuit die can switch in a
very large number of permutations, the common practise was to either take a
worst case estimate, or some kind of approximation. But I guess it makes
sense to conduct a transient analysis to get more accurate values. Thanks
for the info :)
Ihsan,
Thanks for clearing out what Spectre meant by using convolution with regards
to their S-parameter data. I was under the impression they were doing an
IFFT, which didn't make sense. I kept thinking that they were somehow
including the frequency response/IFFT by adding extra equations into the
newton-rhapson/integration formula of the transient solver.
Sincerely,
-------------------
Taha Amiralli
thamiral [A] uwo [D] ca
thamiral [A] gmail [D] com
MESc Candidate 2007, Computer Engineering
The University Of Western Ontario
BESc, BSc. 2005,
Computer Engineering & Computer Science
The University Of Western Ontario
---------------------------------------------------------------
On 12/4/06, Ihsan Erdin <erdinih@xxxxxxxxx> wrote:
>
> Taha,
>
> Spectre (or any circuit simulator that uses a time evolution
> technique) fails because its iterative algorithm builds the error in
> the S-parameter data at every time step. If the data is unstable the
> result shoots to the sky, if non-passive, it oscillates. The
> convolution they mention is an intermediate step to get the
> s-parameter data into the admittance matrix. Keep in mind that the
> original frequency domain data can't be directly included in the
> simulator's time domain algorithm. The data is first approximated in a
> rational form and then expanded in decaying exponentials, which saves
> storage and computational time by recursive convolution. But the end
> result is still obtained by marching in the time, hence the passivity
> requirements.
> For the analysis of power delivery systems, I don't really see a big
> incentive to use a transient simulator as long as the linear load
> assumption holds. This applies to most practical cases. Usually the
> frequency domain solution will give enough information for the
> decoupling efficiency etc. A time domain response may be desired just
> to see the peak-to-peak noise level but then one can easily get it
> through IFFT as you put it.
>
> Regards
>
> Ihsan
>
>
> On 12/4/06, Taha Amiralli <thamiral@xxxxxxxxx> wrote:
> > Ihsan,
> > Your comments are definitely correct. No matter how bad the S-parameter
> > matrices are, with respect to passivity/causality, you will still get a
> > frequency response and thus, an IFFT should do the trick.
> >
> > The thing is that I have read that spectre does use convolution when
> > computing the time domain response from S-parameter matrices. So I guess
> the
> > question then becomes, why would spectre fail to compute the transient
> > response of a non passive S-parameter dataset, as is the case with
> Madhu?
> >
> > Bratfest,
> >
> > I definitely did not consider all those options... but I am still not
> sure
> > why a transient response would be required... If the design is lets say
> an
> > Integrated circuit, with lots of different loads, wouldn't it suffice to
> > just find out what the max power levels required are and provide a power
> > supply that is suited to the task? I guess I am still unsure of how a
> > transient analysis would apply since the loading conditions are
> different
> > depending on what the integrated circuit is doing at a certain moment in
> > time.
> >
> > Thanks,
> >
> > Sincerely,
> >
> > Taha
> >
> >
> >
> > On 12/3/06, Dmitriev-Zdorov, Vladimir <
> vladimir_dmitriev-zdorov@xxxxxxxxxx>
> > wrote:
> > >
> > >
> > > Hi Madhu,
> > >
> > > We do have a tool that can solve the problem you describe, unless your
> > > original touchstone files are of an extremely bad quality (in terms of
> > > resolution, causality, passivity etc.) For example, if instead of
> having
> > > L2 norm of the matrix below one, it reaches 1.1...1.5, it is possible
> to
> > > make the model passive without changing the dependences too much.
> > > However, if the norm goes up to hundreds, this is less likely.
> > >
> > > Anyway, we may produce either passive/causal/smooth data in a form of
> > > the touchstone file, re-sampled with sufficient resolution, or, in a
> > > form of an equivalent SPICE-compatible circuit, whichever you prefer.
> > >
> > > Please contact off-line for more technical details.
> > >
> > > Vladimir
> > >
> > >
> > >
> > > ------------------------------
> > >
> > > Msg: #5 in digest
> > > Date: Sat, 02 Dec 2006 00:16:01 -0500
> > > From: Madhusudanan K Sampath <mkeezhve@xxxxxx>
> > > Subject: [SI-LIST] Spectre S-parameter based modeling- Stability
> issues
> > >
> > > Hi All,=20
> > > I am using n-port component in spectre to perform time domain
> simulation
> > > using touchstone s-parameter files for power planes which are obtained
> > > from
> > > measurement/full wave tools. In many cases, the simulation explodes
> due
> > > to
> > > stability reasons. What are the possible causes of this and how can
> one
> > > solve this issue. I think Spectre n-port component uses convolution
> > > methods
> > > to perform time domain simulation from frequency domain s-parameter
> data
> > > and
> > > hence are prone to convergence issues. If so how does it enforce
> > > passivity
> > > in the model. I verified the s-parameter matrix for eigenvalues and it
> > > proves the data is not passive. I would like to compensate for
> passivity
> > > and
> > > then re-try the time domain simulation. There are quite a few
> > > interesting
> > > passivity enforcements techniques in literature by Dr. B. Gustavsen
> and
> > > Dr
> > > S.G. Talocia . However they are applied on the rational functions
> which
> > > are
> > > obtained by vector fitting the data. My question is if I vector fit
> the
> > > data, enforce the passivity using these techniques and then calculate
> > > the
> > > S-parameter matrix for each frequency point from the passive rational
> > > function model. Will spectre give stable results with this new data
> ?=20
> > >
> > > Please share your ideas and views on this. Thanks in advance.=20
> > >
> > > Madhu=20
> > >
> > > =20
> > >
> > >
> > >
> > >
> > > ------------------------------
> > >
> > > End of si-list Digest V6 #422
> > > *****************************
> > >
> > > ------------------------------------------------------------------
> > > 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 FAQ wiki page is located at:
> > > http://si-list.org/wiki/wiki.pl?Si-List_FAQ
> > >
> > > List technical documents are available at:
> > > http://www.si-list.org
> > >
> > > 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
> > >
> > >
> > >
> >
> >
> > --
> > -------------------
> > Taha Amiralli
> > thamiral [A] uwo [D] ca
> > thamiral [A] gmail [D] com
> >
> > MESc Candidate 2007, Computer Engineering
> > The University Of Western Ontario
> >
> > BESc, BSc. 2005,
> > Computer Engineering & Computer Science
> > The University Of Western Ontario
> > ---------------------------------------------------------------
> >
> >
> > ------------------------------------------------------------------
> > 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 FAQ wiki page is located at:
> > http://si-list.org/wiki/wiki.pl?Si-List_FAQ
> >
> > List technical documents are available at:
> > http://www.si-list.org
> >
> > 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 FAQ wiki page is located at:
http://si-list.org/wiki/wiki.pl?Si-List_FAQ
List technical documents are available at:
http://www.si-list.org
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:
