====================================================================== IBIS INTERCONNECT TASK GROUP MEETING MINUTES http://www.eda.org/ibis/interconnect_wip/ Mailing list: ibis-interconnect@freelists.org Archives at http://www.freelists.org/archive/ibis-interconnect/ ====================================================================== Wednesday, September 4, 2013 8 AM US Pacific Time Agenda: 1) Opens 2) Call for any related patent disclosures 3) Roll call 4) Review of ARs 4) Look at Walter's "Case 5" example using BIRD 125 concepts 5) Try to decide between EMD vs. BIRD 125 style 6) BIRD 145 7) Next Meeting Dates/Agenda ====================================================================== Attendees: ---------- (* denotes present) Agilent - Radek Biernacki Ansoft - Denis Soldo Cadence Design Systems - Terry Jernberg, Brad Brim, Dennis Nagle I/O Methodology - Lance Wang Intel - Michael Mirmak Mentor Graphics Corp. - John Angulo, Arpad Muranyi* Micron Technology - Justin Butterfield*, Randy Wolff* SiSoft - Walter Katz*, Mike LaBonte Teraspeed Consulting Group - Bob Ross* ======================================================================== Patents: - No patents were declared. ======================================================================== Opens: - None ======================================================================== ARs: Walter: Send Arpad a "Case 5" example - done Arpad: Implement Walter's example with BIRD 125 syntax - done (see attached PDF file) ======================================================================== Walter's "Case 5" example using BIRD 125 concepts: Walter showed the example sdram.ibs file with the additional sdram.pkg file. The sdram.ibs file has a package with 4 DQs and power and ground (6 terminals). Each DQ victim has two agressors, but the agressors are unknown. The sdram.pkg file is IBIS EMD syntax showing connections to an s6p file. Two ports are the DQ victim and the other 4 ports are the DQ agressors. Another style could have a different s6p for each DQ, where the victim trace length is uniquely modeled, but the agressors are still generic. Bob asked if it is the intent that the EDA tool configures the model to determine which signal becomes victim, agressor 1 and agressor 2. Walter responded that for pre-layout, yes. For post layout, the tool can look at layout on the PCB and figure out the worst case agressors, then assign the same worst case agressors to the package model. Bob asked what would happen if the victim post layout isn't in between the agressors. Walter responded that a much larger S-parameter for the full package could be used later on to get the most accurate coupling. IC vendors don't often support the full package models. Arpad asked Walter to clarify that the first column numbers are the same terminal numbers as in the Touchstone file. Walter responded yes, they match but there is no requirement to use all the terminals in the Touchstone file. Radek has brought up the issue of still needing to determine how to terminate unused terminals. Arpad showed his example using BIRD125 syntax. The legacy [Package Model] keyword points to the package file. The PackageModel.cir file has a .subckt wrapping a Touchstone file using the S element. He did not change any terminal order. BIRD125 makes a change in the .pkg file where the [Pin Numbers] section matches names from the [Pin] section to names in the [Pin Numbers] section. Then implicit die pads are defined. Arpad added extra text that isn't yet defined in BIRD125. The Language parameter defines IBIS-ISS as the language and the Corner subparameter defines a case pointing to the PackageModel.cir file and the S_pkg subcircuit within the file. The Touchstone file name is passed as a parameter to the subcircuit. The Ports subparameters define the same 6 port names corresponding to Walter's example. Assigning the subcircuit in arbitrary ways required the additional syntax. Arpad defined 6 different port maps (Port_map subparameter). Each Port_map has a position number. If the Position number is 1, the terminal name connects to the first position in the pins list. The position parameter allows the user to shift the mapping around. He developed the syntax from the Spice .connect syntax. Walter asked Arpad how he shows the DQ with specific victim and aggressor assignments. Arpad responded that he did not list all the permutations. Walter stated that for the controller, Arpad has too many permutations for this syntax to be useful. Arpad agreed that the syntax gets too cumbersome, but also the syntax he showed is not in the BIRD yet, and it could be changed to make it more automated. Walter stated that he's claiming that what he is proposing solves the problem of "Case 5" with an efficient syntax. The only justification for the BIRD125 method is to re-use some existing IBIS keywords. We must make a decision on a path forward. Arpad responded that he is not debating yet which way is the smarter way to do it. We must solve the user's needs. The decision point is how much to leave behind existing syntax and move to new syntax or try to re-use exisitng syntax as much as possible and add more baggage to it. He agreed that Walter's syntax is more elegant. One could rewrite the IBIS syntax in a more elegant way too. Walter added that he thinks that going with the simple, elegant method and a methodology compatible with EMD is the right approach. Randy added that he favors EMD style syntax due to its reduced complexity that makes it easier for model creators. It also nicely builds upon the EMD syntax. Arpad asked how we decide how to move forward. We don't have enough people in attendance in the Wednesday meetings to take a vote. Bob responded that we need to take some recommendations to ATM or the Open Forum. Arpad stated that Walter's syntax also requires users to somehow assign the proper connections in the GUI. Walter responded that actually, the EDA tool can do this automatically by determining the worst case coupling and assigning the model to those signals. Arpad asked if this is done automatically, why would BIRD125 syntax need to define all possible permutations if only 3 pins would need to be simulated. Walter responded that the BIRD125 syntax must describe at least the nearest neighbor of every signal. Arpad added to remember that the proposed syntax can be changed. He would still like to have the data presented in front of more people. Walter stated that the ATM task group is probably the best place for this, with lots of technical people present. Arpad took the AR to schedule this topic for Tuesday's ATM meeting. ======================================================================== Next Meeting: Wednesday, September 11, 2013