Many thanks to Randy Wolff for his fantastic efforts in running the meeting and taking minutes!
 
======================================================================
IBIS INTERCONNECT TASK GROUP MEETING
 
http://www.eda.org/ibis/interconnect_wip/ 
 
Mailing list: ibis-interconn@freelists.org 
 
======================================================================
Attendees, Nov. 14 
 
Agilent Technologies                Radek Biernacki
ANSYS                               Luis Armenta
Cadence Design Systems              Brad Brim
Mentor Graphics                     Arpad Muranyi
Micron Technology                   Randy Wolff 
Signal Integrity Software           Walter Katz
Teraspeed Consulting Group          Bob Ross
 
Minutes

Randy Wolff ran the meeting in Michael Mirmak’s absence. No patents were declared.
 
Walter Katz presented his PowerPoint “Package EBD/EMD.”  He noted that IBIS-ISS is easier to implement in EMD than in native IBIS.  An MCM is defined as a package with multiple die.  Even though pins aren’t physical pins anymore, now more likely BGA balls, Pins in IBIS and the Pin List in EBD refer to similar things.  He showed an EBD modified by adding [IBIS-ISS Package] pointing to an IBIS-ISS file (.ipkg) and [Extended Nets] that show relationships of pins to individual component connections, essentially low impedance path connections.  The difference between EBD and EMD is calling a ‘board’ a ‘module’.  EMD could be written in a parameter tree format.
 
Walter continued describing how EMD could be used to describe connectors with minor changes to the syntax.  On slide 11, Radek Biernacki asked what the word ‘Full’ refers to.  Walter said this is just a name for the model, since you could have multiple models for connectors.  Walter described handling wire bonds.  They could be included directly in the package model, with wire bond lengths parameterized, or the wire bond assembly could be treated as an interposer.  
 
He ended by recommending defining a new EMD standard to model MCMs, connectors, cables and single die packages.  Then, independently we could enhance .ibs files to handle IBIS-ISS on-die interconnect and package interconnect.
 
Brad Brim commented that if you take the example of 3 dies stacked, Extended Nets usually helps when there are blocking caps and you need to define a connection on a net from one side of the blocking cap to the other.  For an instance of serial bondwire connections (cascade bonding), does Extended Nets still work?  Walter replied that yes, it is just providing info on low impedance connection points.  Brad also asked why is on-die and package modeling conceptually different?  Walter commented that package is modeling between pins of package and die pads.  On-die is modeling between die pads and buffers.  They all have ports.  
 
Bob Ross commented that Model_Ports is equivalent to a subckt call defining connections with a specific port order.  Walter noted that when getting to generic subckts, it is better to use parameter tree syntax than syntax for subckt connections as defined in IBIS-ISS.
 
Radek noted that by selecting a subset of terminals in a subckt, you should define more.  You would need a wrapper subckt to define the terminations of all the unused ports.  Walter said this is unnecessary as long as you carefully define how to pull out a .s4p from a .s72p.  
 
Walter then showed his presentation “Package Die Ports.”  He noted that a naming convention is needed for die buffer ports including pad and power/ground connections.  Diepad defines the bondwire pad or die bump pad.  PkgPad is the bondwire pad or die bump pad of the package substrate.  PkgPin is the bga ball or pin of the substrate.  Implicit port names come from [Pin], the on-die model and package model.  Legacy IBIS doesn’t allow distributed die PDN models, so we’ll need the ability to add die pads to .ibs files.
 
Brad asked if the second presentation essentially says that we need to define die pads. Walter responded that it describes the need to come up with naming conventions.
 
Time ran out, so Walter did not go through the presentation in full detail.  Walter proposed that we start creating a new EMD standard, and this could be the main discussion topic for next week.
 
The group decided that the next meeting should be Nov. 21.