ICM Issue Resolution Document  (IIRD)

IIRD ID#:        1.1
ISSUE TITLE:     Clarification of Mult for S-parameters
REQUESTOR:       Michael Mirmak, Intel Corporation
DATE SUBMITTED:  April 16, 2004
DATE REVISED:    May 11, 2004
DATE ACCEPTED BY IBIS OPEN FORUM: June 4, 2004

****************************************************************************
****************************************************************************

STATEMENT OF THE ISSUE:

S-parameters, in Touchstone(R) files, may be used to describe the electrical
behavior of sections within an ICM model.  S-parameters may only be used
if the following conditions are also met:

- a [Nodal Path Description] is used
- ICM_model_type is "S-parameter"
- [Derivation Method] is "Lumped"
- LEN is not used with any N_section

The language of the specification implies that the MULT token must be used 
with S-parameter section descriptions, and states that any integer value of 
MULT is legitimate.  This is misleading, especially for cases
where S-parameter data sets feature odd numbers of ports

The specification text is changed below to limit Mult to "1" when 
S-parameter data is used under ICM 1.0.


****************************************************************************

STATEMENT OF THE RESOLVED SPECIFICATIONS:

Changes are denoted by *.
Changes in IIRD1.1 are denoted by **.
 
The text of [Nodal Path Description] under Section 7 should be changed
as follows:

The scale factor "Len" is applied to each parameter of all matrices of a
section resulting in scaled RLGC matrices.  When S-parameters are used
to describe a section, the usage of scale factor "Len" is not allowed.

** The "Mult" factor must be set to 1 for any N_sections which are 
** described using S-parameters.


****************************************************************************

ANALYSIS PATH/DATA THAT LED TO SPECIFICATION:

A value of MULT other than 1 is meaningless for S-parameter data sets
with odd numbers of ports.  

For example, imagine the case of a "Y" splitter, with two input ports and
one input port.  In this case, [ICM S-parameters] will properly map the
ports for each section to the nodes of the N_section of interest.  
However, if Mult=3, for example, how can a tool properly copy the 
S-parameter data and attach the duplicated section models together?
ICM includes no explicit mapping for a section multiplied in this way;
such explicit mapping within a multiplied section would be required to
connect the output stage of the "Y" with one of the input stages of
the next duplicated section.

As odd numbers of ports may be useful and the parsing limitations involved 
would be prohibitive, limiting Touchstone(R) S-parameter data files to 
even numbers of ports is not be a preferred solution.

RLGC matrices do not have this problem, as the number of input and output
ports is assumed to be identical within each section.

IIRD1.1

Small modification made to language, to remove misleading justification.

Original change text read:
* As S-parameter data may contain odd numbers of ports, the "Mult"
* factor must be set to 1 for any N_sections which are described using
* S-parameters.



****************************************************************************

ANY OTHER BACKGROUND INFORMATION:

A separate question raised by this BIRD is whether Mult=0 is valid.  The
specification permits integers, but without limit, to be used with Mult.

Thanks to Shulong Wu of Synopsys for highlighting this issue in April
messages on the SI-LIST.

****************************************************************************