====================================================================== IBIS INTERCONNECT TASK GROUP MEETING http://www.eda.org/ibis/interconnect_wip/ Mailing list: ibis-interconn@freelists.org ====================================================================== Next meeting: Wednesday, Oct. 3, 8 AM US Pacific Daylight Time Agenda: - Call for Patents - Opens - Topic: modify existing IBIS structures or start with/fork to new, similar specification? - Progress to Schedule - Next Meeting Logistics and Agenda Teleconference Logistics For international numbers, please contact Michael Mirmak. 916-356-2663, Bridge: 2, Passcode: 1450744 Join online meeting (https://meet.intel.com/michael.mirmak/S1580TK9) ====================================================================== Attendees, Sept. 19 Agilent Technologies Radek Biernacki Cadence Design Systems Brad Brim Intel Corp. Michael Mirmak Mentor Graphics Arpad Muranyi Micron Technology Justin Butterfield, Randy Wolff Signal Integrity Software Walter Katz Teraspeed Consulting Group Bob Ross Minutes Michael Mirmak called the meeting to order. No patents were declared. During Opens, Walter Katz noted that no responses had been received to his recent e-mailed questions to IC vendors. He concluded that enough information was available, however, to draw conclusions about how to structure IBIS package and buffer relationships. Michael summarized the rationale for the Task Group and the policies it will follow. The Task Group acts as a relief valve for the IBIS-ATM task group, which has numerous BIRDs to consider. As some critical BIRDs related to IBIS-AMI were not being discussed while packaging direction was being decided, the Interconnect Task Group was reactivated to handle packaging discussions and provide direction. Arpad Muranyi noted that multiple meetings per week might represent too great a workload for some participants. Michael stated that the IBIS Open Forum charter rules would apply to this meeting: organizational membership is used for voting eligibility and a quorum of five paid members in good standing is required to conduct business. Minutes will be more rigorously published and distributed. Walter Katz suggested that straw votes be used to make decisions. Bob Ross moved to conduct business using formal votes; the motion was seconded and approved without objection. Michael then asked the participants to identify the key issues requiring disposition by the Interconnect Task Group. The following summarizes these points: * Arpad: Do we decide specific proposals for packaging/on-die interconnect, or decide the general direction of IBIS first? * Bob: What is the status of EMD? A unified solution is needed that incorporates Touchstone 2.0, IBIS-ISS and IBIS models. * Walter: Several items need support o Die pads and package pins, both single-ended and differential, with coupling; power effects are secondary o Multi-chip modules with single-ended and differential support and coupling; power effects are secondary o Connector and cable models o Optical interconnects (cables) o On-die interconnect, including on-die pads (bumps) and buffer pads; this is the only non-traditional interconnect in the list o Not supported: actual buffer model data/curves * Michael: Do we maintain a structure and keyword hierarchy that is component oriented? * Michael: Do we maintain a single-file structure (all called through .ibs) or use multiple files that are hierarchically equivalent? * Bob: Do we maintain IBIS as a baseline, or branch off to form a new but related specification? A common way of checking models vs. the specification would be needed. Post-layout simulation support should also be discussed. * Arpad: Definitions are needed. Is a component a bare die? Is it a bare die plus on-die interconnect plus other aspects? During IBIS development, this wasn’t needed as everyone has fairly common understandings of these terms. Walter noted that the simplest and quickest way to solve the IBIS packaging problem is to keep the existing structure: component/model within a given file (.ibs or other). For example, with IBIS-ISS, a component may reference separate files. Walter added that he thought of components as EBDs, but deferred on how to approach multi-chip modules. Randy Wolff suggested that a multi-chip module be defined as two pieces of silicon but with one-to-one connections to pins. Walter added that this was an exception, not a rule; an enable signal might connect two pieces of silicon inside a package that are otherwise “hidden”. He suggested that IBIS be modified to support: - Single package, single chip designs - EBD enhanced with IBIS-ISS - Package modeling structure and content could be the focus of the remaining discussions Bob asked whether this means providing a separate file for the package model. Walter responded affirmatively. Bob noted that IBIS packages are overloaded today, with pin-centric lumped models, path descriptions without coupling and matrix descriptions with coupling. Walter suggested that the critical issue is whether to assume multiple supply nodes per port, or one supply node per port (meaning, power rail or buffer power connection). Brad Brim asked whether, in a 16 buffer structure, all buffers would be connected to a single power node. Walter confirmed this. Brad suggested that the resulting impedance variation could be as much as 3 times the value of one path, depending on the stimulus and interactions. Walter noted that on-die interconnect could be handled with Touchstone within IBS-AMI files. Arpad suggested that the fundamental question to decide first would be whether to proceed with incremental changes to IBIS or a more radical overhaul. Radek Biernacki asked whether this would be limited to packages or would “hit everything”. Michael stated that all decisions regarding packaging be concluded by the end of the year. Walter suggested that six months would be possible, adding that minimizing EDA vendor and IC vendor impacts are goals. Michael asked for preferences on meeting times and schedule and suggested several options. Walter noted his unavailability for the next meeting. Brad stated that no proposed time was optimum for him, while Bob suggested Fridays from 8-9 AM US Pacific time when the IBIS Open Forum was not holding a teleconference. Michael took the AR to determine via e-mail the best regular time for the meeting. The meeting adjourned.