=================================================================
IBIS INTERCONNECT TASK GROUP MEETING

http://www.eda.org/ibis/interconnect_wip/ 

Mailing list: ibis-interconn@freelists.org 

=================================================================
Attendees, Oct. 31 

Agilent Technologies                 Radek Biernacki
ANSYS                                Steve Pytel
Cadence Design Systems               Brad Brim
Intel Corp.                          Michael Mirmak
Mentor Graphics                      Arpad Muranyi
Micron Technology                    Justin Butterfield, Randy Wolff 
QLogic                               James Zhou
Signal Integrity Software            Walter Katz
Teraspeed Consulting Group           Bob Ross

Minutes

Michael Mirmak called the meeting to order.  No patents were declared.

Walter Katz made some comments per the discussion on multi-chip packages 
kicked off by Lynne Green on the reflectors.  He added a few details 
regarding his presentation from the previous meeting, where network 
connectivity was intermittent.  Bob Ross asked whether VDD is a 
reserved signal name in Walter's presentation.  Walter noted that 
it is not.  Walter suggested that additional subparameters within 
[Model] may be required to support his proposal.

Steve Pytel asked about the objective of the group.  If the purpose 
is to update IBIS for current package technologies, by the time we 
solve that problem, the industry will be doing interposer-to-die 
connections that will require additional updates. One of his 
customers is already facing multiple dies on a single interposer.  
Si2 organization already taken this problem on.  What is IBIS's 
position here?

Michael made some comments on Si2 3D modeling.  IBIS is simply 
trying to "catch up" with current technology, to enable better 
package definitions than the current limited IBIS package keywords 
permit.  Si2 is involved in 3D package modeling, from an extraction 
perspective, which is outside IBIS's scope at present.  
PCB-to-package-to-die connection mapping may also be covered by Si2, 
which is of greater interest to EDA vendors than other parts of 
IBIS today, but would be monitored for consistency with IBIS goals.  
This team is nearly finished with proposals to combine IBIS and 
improved packaging, including IBIS-ISS and Touchstone 2.0.

Walter added that the team is also trying to tie die pads to 
buffers, to include signal, power coupling; an EBD-like approach 
would work for this.  Arpad expressed partial agreement with 
Walter, noting that the team has had a long discussion about 
incremental change vs. fundamental changes.

Steve stated that 60-70% of packages he sees are S-parameter-based, 
with some W-elements seen.  Today's EBD not sufficient.  Bob asked 
how many ports are seen: 2, 4, other?  Steve replied that the models 
were n-port, with some models including thousands.   He added that 
the challenge is to perform circuit simulation with S-parameter data, 
where checking passivity, causality is critical. He asked whether 
this group will make recommendations or statements on best-known practices.

Michael answered that the policy of IBIS groups and IBIS itself 
was not to provide algorithmic solutions at a specification level 
that might compete with the solutions provided by its own members.  
Formats in common, not specific solution methods, were the main 
development objective of IBIS.  Arpad added that transportability, 
interoperability were the main points; Radek Biernacki agreed, 
adding that interconnect is only one usage method for S-parameters; 
active devices are another and passivity and causality are not 
concerns there.

Walter and Arpad agreed that an EBD-like process could be used for 
interconnect support.  Arpad suggested that EBD needs more updates 
to support IBIS's needs.

Michael asked if there is overlap between the current interconnect 
proposals now.  Arpad responded yes.  Walter suggested further that 
there's a clear direction to IBIS:
- enhancement to IBIS to support packages
- distributed on-die power delivery

EBD with interconnect improvements would be what we need, according 
to Walter.  EMD (Electrical Module Description) is preferred, with 
Touchstone 2.0 files and/or IBIS-ISS replacing interconnects.  
Syntax to connect package data to IBIS buffers is also needed.  Steve 
added that general improvements to packages are needed in baseline 
IBIS.  Michael suggested that port, node and pin mapping closure is 
needed, particularly for Touchtone 2.0.

Michael asked Arpad and Walter to identify the specific areas where 
their proposals required changes to the IBIS specification and where, 
if anywhere, the specific proposals differ.