ICM Issue Resolution Document  (IIRD)

IIRD ID#:        7.1
ISSUE TITLE:     Clarification of Off-Diagonal Matrix Terms
REQUESTOR:       Michael Mirmak, Intel Corporation
DATE SUBMITTED:  September 10, 2004
DATE REVISED:    October 5, 2004
DATE ACCEPTED BY IBIS OPEN FORUM: October 29, 2004 

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

STATEMENT OF THE ISSUE:

The language used in the ICM 1.0 specification to describe [Resistance Matrix] 
data does not clearly establish the meaning of off-diagonal terms, especially
for missing Sparse_matrix columns or where the matrix is not present.

Much of the specification language used to describe "Diagonal_matrix," "Banded_matrix"
formats, etc. involves the phrase "entries are guaranteed to be zero if they
are not on the main diagonal."  This implies that the resistance matrix
off-diagonal elements are to be interpreted as many leading tools actually do 
interpret them, namely as a relationship between the current on one conductor 
and the voltage on another.  A value of zero implies "no relationship."

However, language specific to the [Resistance Matrix] definition states that off-diagonals
are to be assumed "infinite."  This suggests that the off-diagonals describe
a real resistive connection between conductors.  Few, if any, industry
field solvers present off-diagonal resistance matrix data in this form.

Finally, the sign convention for mutual resistances is described as important
in one part of the specification, but the allowed number formats for
[Resistance Matrix] data later in the text is omitted, unlike the descriptions 
of the other matrix formats.

IIRD7 attempts to clarify the issue, by explicitly stating that off-diagonal
[Resistance Matrix] data refers explicitly to the relationship between
currents on one conductor and voltages on another, implying zero off-diagonal
values as defaults.  IIRD7 also adds language to explictly state that missing
matrix data for any matrix is assumed equal to zero.


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

STATEMENT OF THE RESOLVED SPECIFICATIONS:

Changes in IIRD7 are denoted by |*.
Changes in IIRD7.1 are denoted by |**.
 
The text of Section 8, under the definition of the [Resistance Matrix], etc.,
should be changed as follows:

Remove the following sentence:
|* Users should be aware that matrix data of value 0 represents an 
|* ideal short to reference ground and should not be associated 
|* with conductors which carry signal information (see below).

and replace it with:
|* For all matrix keyword data sets below, both diagonal and non-diagonal 
|* entries are assumed to contain actual values of resistance, conductance, 
|* etc.  For example, a [Resistance Matrix] is assumed to contain information 
|* in the form of ohms or ohms per unit length, depending on the use of the 
|* Mult or Len modifiers.  

|* Values for non-listed [Row] or column entries or for matrices not explicitly 
|** defined shall be assumed equal to zero.

----------------------------------------------------------------------

Under "Resistance Matrix Description," remove the following:
|* The resistance matrix is optional.  Simulation tools shall use
|* 0.0 ohms for all diagonal entries and "infinite" for all off-
|* diagonal entries if this matrix is not present.

and replace it with:
|* The resistance matrix is optional.  

|** Previous IIRD7 text
|* Users and model authors should be aware that [Resistance Matrix] data of 
|* value 0 in the diagonal represents an ideal short. 

|** IIRD7.1 text
|** [Resistance Matrix] data of value zero in the diagonal always represents
|** an ideal short. 

|** Note that a value of zero ohms in the off-diagonal of a [Resistance Matrix] 
|* does NOT imply a short circuit between conductors.  The off diagonal entries 
|* in resistance matrices relate the voltage generated in one conductor to a 
|* current in another. Using Ohm's law (V=I*R), an off-diagonal resistance 
|* value of zero indicates the absence of resistive coupling between the conductors.

----------------------------------------------------------------------

Also under "Resistance Matrix Description," remove the following:
|* Diagonal elements may be dominant.  Comparably, the off-diagonal elements
|* may be negligible thus leading to off-diagonal element values of infinite value.
|* Therefore, a "Diagonal_matrix" matrix type is the best choice for describing
|* a resistance matrix.

and replace it with:
|* In many cases, diagonal elements may be dominant and off-diagonal elements
|* (expressing the relationship between the current on one conductor and the voltage 
|* on the other) may be negligible.  Therefore, a "Diagonal_matrix" matrix type is 
|* usually the best choice for describing a resistance matrix.  Off-diagonal
|* terms, when other matrix types are used, may be positive real numbers, negative
|* real numbers or zero.


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

ANALYSIS PATH/DATA THAT LED TO SPECIFICATION:

The changes above were discovered as a result of an IBIS Futures Committee
discussion of IIRD 2.2.  The language above on off-diagonal [Resistance Matrix]
data is paraphrased from excellent SI-LIST discussions in February 2003 and from
comments by Bob Ross and Arpad Muranyi.

IIRD7.l is issued to make stylistic changes requested by Bob Ross, including
removal of "model author" language and use of "zero" as standard and 
consistent with the rest of the document.

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

ANY OTHER BACKGROUND INFORMATION:


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