[ibis-interconn] Re: Sparse matrix/label treatment, draft 3...
- From: "Walter Katz" <wkatz@xxxxxxxxxx>
- To: "IBIS-Interconnect" <ibis-interconn@xxxxxxxxxxxxx>
- Date: Wed, 14 Oct 2009 05:18:17 -0400
John,
I would like to clarify one point. The EDA tool only needs to simulate the
S-Parameter data in the full reconstructed matrix. The Integer_Labels
(Framis) in the [Network] data, along with the [Sparse Matrix Mapping]
Integer_Label/Index_Pair data tell the Touchstone Parser (or EDA platform)
how to populate the full reconstructed matrix. So the critical thing for an
EDA tool is not that there are any Integer_Labels that are not used in the
full reconstructed matrix, but that the full reconstructed matrix is
complete. By complete I mean that any Index_Pair that is not in any
Integer_Label is supposed to be zero (i.e. no coupling, or no reflection, or
no transmission between that Index_Pair).
You suggest that we might allow a Touchstone 2.x file might have multiple
[Sparse Matrix Mapping] sections (like an IBIS model has multiple
[Components]. One option is to create a ?Virtual Touchstone 2.x File?
dynamically for the pins in a connector or package that are being used in a
simulation. This ?Virtual Touchstone 2.x File? simply needs to reference the
real ?Touchstone 2.x File? that contains the [Network Data]. We do this
today and it is very affective. We do not need to write a spec for a
?Virtual Touchstone 2.x File?, each EDA tool can do it as it sees fit, but
allowing a real ?Touchstone 2.x File? that is simply a container of
Integer_Labels allows this to work.
I certainly would not object if the Touchstone 2.x parser would generate a
warning, listing all of the Index_Labels that do not have any Index_Pair,
but it should not be an error in the parser.
Walter
Walter Katz
303.449-2308
Mobile 720.333-1107
wkatz@xxxxxxxxxx
www.sisoft.com
-----Original Message-----
From: ibis-interconn-bounce@xxxxxxxxxxxxx
[mailto:ibis-interconn-bounce@xxxxxxxxxxxxx]On Behalf Of Angulo, John
Sent: Wednesday, October 14, 2009 3:25 AM
To: IBIS-Interconnect
Subject: [ibis-interconn] Re: Sparse matrix/label treatment, draft 3...
To reiterate my comment in last week's meeting on the use of [Sparse Matrix
Mapping] labels without "index pairs": If the [Sparse Matrix Mapping]
keyword does not map a column of [Network Data] into the full matrix, then
the EDA tool does not have a way to simulate the S-parameter data in that
column without further information.
If the model maker provides suitable comment text in the Touchstone file, or
if a third party consultant knows how the [Network Data] columns
corresponding to the empty labels in a particular file are to be treated,
then the user of an EDA tool can modify the [Sparse Matrix Mapping] keyword
text by hand to select the active subset of [Network Data] columns.
More automated ways to give further information might include having
multiple [Sparse Matrix Mapping] keywords in the Touchstone file, each with
a subparameter identifying the alternative mapping by some name. Another
way might be to somehow override or supplement [Sparse Matrix Mapping]
through a new keyword in the [Begin Information]/[End Information] section.
However it is done, it seems certain that some additional information is
necessary to make use of [Sparse Matrix Mapping] labels without index pairs.
When should this committee consider whether and how the Touchstone format
should convey this information? Should the proposed language allowing empty
[Sparse Matrix Mapping] labels, and hence unused columns of [Network Data],
go into a Touchstone 2.x specification without such consideration?
Thanks,
John
_____
From: ibis-interconn-bounce@xxxxxxxxxxxxx on behalf of Mirmak, Michael
Sent: Wed 10/7/2009 8:56 AM
To: IBIS-Interconnect
Subject: [ibis-interconn] Sparse matrix/label treatment, draft 3...
For review and discussion...
- MM
Other related posts:
