==============================================================================

IBIS INTERCONNECT TASK GROUP
Mailing list: ibis-interconnect@freelists.org 

================================================================================

Attendees from June 11, 2025 Meeting (* means attended at least using audio)

ANSYS                                Curtis Clark, Wei-hsing Huang, Juliano Mologni
Arista Networks			     Jim Antonellis
Broadcom                             James Church
Chipletz			     Stephen Newberry*
Intel Corp.                          Michael Mirmak*, Xiaoning Ye
Keysight Technologies                Ming Yan
Marvell                              Steve Parker
MathWorks                            Walter Katz*
Micron Technology                    Justin Butterfield
Siemens EDA                          Weston Beal, Arpad Muranyi*, Randy Wolff*
Simberian 			     Yuriy Shlepnev
ST Microelectronics                  Aurora Sanna
Synopsys                             Ted Mido, Edna Moreno
University of Illinois               Jose Schutt-Aine

Michael Mirmak convened the meeting.  No patents were declared.

Stephen Newberry introduced himself.  He is located in New Jersey, 
working for Chipletz, an advanced packaging company located in
Austin, Texas.  A coworker recommended he join the Interconnect Task Group.
Michael welcomed Stephen to the group.

Michael reviewed the minutes of the June 4 meeting.  Randy Wolff moved to
approve the minutes; Arpad Muranyi seconded.  The minutes were approved
without objection.

The team reviewed the latest draft of the port mapping proposal, with a focus on 
the examples.

Randy noted that Example 9 still lists buffer_io among the terminals for C_comp Model; 
this appears to be an error.  The buffer_I node is the point for making measurements vs. 
buffer_io, which is the external connection point

Randy suggested that buffer_io should be required.  Arpad asked whether buffer_i 
is a measurement point, an unconnected node, or something else. Randy noted that the 
structure assumes a differential input buffer would be connected at this point.  
Arpad asked whether I-V tables, such as for a clamp, etc. could be connected there.  
Is the buffer_io node in the IBIS C_comp Model diagram only an output?

Randy replied that it is a high-impedance probe node. In industry, it is common to 
have a series resistor before the measurement location.  Arpad noted that it is similar
to an RC structure, which is otherwise impossible to represent in IBIS.  Randy suggested
it was like any other Touchstone data set.  Walter Katz suggested that it also is assumed
not to draw much current.  Can the gnd_clamp_ref be a port for the Touchstone data?  
He suggested it was likely not possible, due to the need for a return current.

A new rule was proposed: buffer_I/O should always be connected when Touchstone files 
are used.  Walter suggested that the assumed circuit structure needs a series component 
on the output, due to customer request.  This would be buffer_o (and would split up the 
buffer_io path to the pad).  This would be a series element.

Randy noted that this situation had been considered during C_comp Model development, but
was complicated by C_comp compensation assumptions.  This is a classic case where SPICE 
is being replaced by Touchstone.  What does the specification and/or a model maker do with 
A_gnd, etc.?

Randy suggested the specification maybe only allow A_gnd.  The new Example 9 being added
to the drat is a 2-port S-parameter data block, really.

Michael suggested that using A_gnd and a Physical declaration seems to imply allowing the 
Port line to use A_gnd.  Should this be allowed?  

Walter replied that this is a 2-port network, but a three-terminal "device", similar to 
an S-element.

The nodes are 1 0 2 0, with A_gnd used as 0.

Arpad asked about connecting a 0 V source.  Randy suggested that would be handled at 
instantiation, not definition.  Arpad replied that this information covered how the data
would be used.  Randy cited the example of a 20-port Touchstone S-parameter data set where
only four ports are used.

A_gnd was removed from list of nodes supported in this case.

Arpad cited the previous examples of one versus two RLC paths (VCC vs. VCC and VSS) on a circuit. 
Walter replied that, due to return path needs for AC analysis, it's a different situation.

Randy suggested that, for Figure 9 in IBIS, A_gnd can be applied to any (SPICE circuit) terminal.  
No change is needed.

Arpad moved to adjourn the meeting; Randy seconded.  The meeting adjourned without objection.