================================================================================ IBIS INTERCONNECT TASK GROUP http://www.ibis.org/interconnect_wip/ Mailing list: ibis-interconnect@freelists.org Archives at http://www.freelists.org/archive/ibis-interconn/ ================================================================================ Attendees from December 6 Meeting (* means attended at least using audio) ANSYS Curtis Clark Cadence Design Systems Bradley Brim Cisco David Siadat Intel Corp. Michael Mirmak* Keysight Technologies Radek Biernacki, Ming Yan Mentor, A Siemens Business Arpad Muranyi* Micron Technology Justin Butterfield*, Randy Wolff* SAE ITC Maureen Lemankiewicz, Logen Johnson Signal Integrity Software Walter Katz*, Mike LaBonte* Teraspeed Labs Bob Ross* University of Aveiro in Portugal Wael Dghais Michael Mirmak convened the meeting. No patents were declared. Justin Butterfield took minutes. Review of Minutes: - Michael called for review of the minutes from the November 29 meeting. Mike LaBonte moved to approve the minutes. Randy Wolff seconded. The minutes were approved without objection. Review of ARs: - Mike to post BIRD189.5_draft11 to the website. - Mike reported this is done. - Mike to have IBIS-ISS change procedure discussion in the Open Forum meeting. - Michael reported this is done. - Bob to add bus_label text and node 0 rule to the latest BIRD189 draft. - Bob reported that these are done. - Arpad stated that he started an initial draft of a BIRD on "explicit die pads" in IBIS Version 6.1 and their relation to BIRD189. Michael noted that this was an AR from the Open Forum meeting. Randy confirmed it will be in the minutes for the Open Forum meeting. Opens: - Michael will need to end the meeting a few minutes early. Summary of ATM: Bob had presented Groups and Sets in block diagram form. He noted that he does not advocate shorted paths, although this may not be clear in the diagrams. Arpad asked if Bob's stance is that there should not be automatic shorting. Bob said that this is correct. He also stated that he and Walter Katz have a syntactical difference. Another difference is his proposal does not assume that a missing Set means a short. Bob noted that if we allow a die only interface, then we have a way to merge pins for signal pins, but for rail terminals we would need to use bus_label. Bob commented that he and Walter still have some details to resolve between their proposals. Walter stated that there was not enough time in the ATM meeting to have discussion on his proposal. Interconnect Model Groups and Sets: Walter showed an email he sent out, which proposes to add a sub-parameter named "Group_type" with four possible selections: Pin_to_Pad, Pin_to_Buffer, Buffer_to_Pad, and Bare_die. He thought it would be useful to distinguish between Buffer_to_Pad and Bare_die, so it can be known whether to include a legacy package model or not. He also went over his Component and Bare_die proposal which he would prefer over the Group_type proposal. Walter commented that this makes clear what the EDA tool needs to do with the Interconnect Model and is a simple solution with the rules outlined in the email. Bob disagreed that it is a simple solution. Bob asked if we have a Group for pin to pad then we do not have buffer terminals. Walter thought the best approach is to have only two groups of Component and Bare_die. Bob commented that it is useful for the user to know what is in the Group. Bob noted that missing Sets open the door to bad models, as we don't know if a short should be applied or it is a mistake. Walter commented that all EDA tools today assume a short from pad to buffer. Bob stated he does cover this in his proposal through bus_label. Walter noted that power aware simulations today are done with pin to pad models for signal and pin to pad for power with a decoupling model at buffer pads. Bob stated that on die capacitance can be relabelled, and he agreed this is the approach most use. Arpad asked Bob, assuming we have the 4 Group_type subparameters, if a pad to pin model that does not provide a pad to buffer path would be illegal and would have a disconnect. Bob stated that the pad to pin replaces the existing IBIS package model, but we have Pin Mapping and bus_labels to make the connections. Arpad asked if that works for both supply rails and signals. Bob stated that for signals the pin_name is the same across pad and buffer. Arpad asked if we need to define the Pin Mapping for signals. Bob stated yes, but we require Pin Mapping and must use it. Michael asked if the Pin Mapping is required with the Interconnect BIRD. Bob thought that it is needed for the details of which buffer connects to which supply. Walter stated that you do not need Pin Mapping for his proposal if you tie the terminals directly to buffers. He noted that Pin Mapping is only need to tie the power buses together. Bob agreed. Arpad asked if we disallowed pin to pad only models, then we would require pad to buffer models. He has some doubt that Pin Mapping can be used to short the signals. Walter agreed with this. Bob disagreed with this statement. Bob noted that there is a difference in terminology. He agreed that you can connect with existing IBIS syntax to create the short for signals. But, for the PDN structures, you need Pin Mapping and bus_label. Arpad asked why we need to distinguish the connections at the Group level, and if we could do the same thing at the Set level. He wondered if there would be any advantages to defining the type of model at the Set level. He noted that the Group_type or similar subparameter would apply to Sets below the Group. Walter commented that he sees the Sets as generated from a specific tool. Bob noted that we should have flexibility at the Set level for the EDA tools creating the models. Arpad stated that we need to avoid the situation of including the wrong Sets in the wrong Groups. Bob noted we disagree on the way to define precise boundaries between the models. Walter thought we need to get closure on the Sets and Groups issue. He believes some of Bob's concerns are unfounded and would like to hold a vote. Arpad stated that he is not ready to vote. Michael cautioned against having a vote until we clearly define the two proposals. Michael asked Bob and Walter to each write a concise (one page) bullet list of their Sets and Groups proposals [AR]. Bob commented that there are other details to discuss such as the Aggressor rules being tied in to connection issues, which he disagrees with. Bob noted that he also does not like statements referring to legacy package models, as we are not clear on which format it applies to. The next meeting is scheduled for Friday December 8. Mike moved to adjourn. Randy seconded. The meeting adjourned without objection. Task List BIRD189.5 editorial additions/changes to be completed: 1. File_TS0 and File_TS referencing warning language. 2. Remove the word "reference" from the IBIS-ISS examples.