Minutes, IBIS Quality Committee 15 May 2007 11-12 AM EST (8-9 AM PST) ROLL CALL Adam Tambone Barry Katz Benny Lazer Benjamin P Silva * Bob Ross, Teraspeed Consulting Group Brian Arsenault * David Banas, Xilinx * Eckhard Lenski Eric Brock Gregory R Edlund Hazem Hegazy John Figueroa John Angulo Katja Koller Kevin Fisher * Kim Helliwell, LSI Logic Lance Wang, IOMethodology Lynne Green * Mike LaBonte, Cisco * Moshiul Haque, Micron Technology Peter LaFlamme Radovan Vuletic, Qimonda Robert Haller * Roy Leventhal, Leventhal Design & Communications Sherif Hammad Todd Westerhoff Tom Dagostino Kazuyoshi Shoji Sadahiro Nonoyama Everyone in attendance marked by * NOTE: "AR" = Action Required. -----------------------MINUTES --------------------------- Mike LaBonte conducted the meeting. New items: - Barry Katz has offered to join the meeting next week to help clarify issues regarding [Receiver Thresholds]. - We agreed to discuss [receiver Thresholds] next week. - We should also find a spec to show example timing requirements. AR: Moshiul send a datasheet for SSTL2 AR: Mike invite Barry Katz to next meeting for [Receiver Thresholds] discussion AR: All review IBIS [Receiver Thresholds] and SSTL2 for next call AR Review: - Mike prepare IQ status presentation for IBIS summit - TBD - Will not be reviewed by the group ahead of the summit. - David research [Receiver Thresholds] issues - Will discuss next week Continued review of IQ specification 1.1p: 3.2.5. {LEVEL 3} [Pin] RLC parasitics are present and reasonable - Mike inserted a new clause with his last edit: "To pass this check the RLC values must be present for all signal pins in the [Pin] section, even if [Package Model] is present." - Roy pointed that requiring lumped RLC when a better model is present is "going backward". - In theory it is possible to extract an RLC model from any [Package Model]. - Kim is writing a "lumpifier" to do that. - We voted to reverse the sense of the requirement so that RLC is not required when [Package Model] is present. AR: Mike change 3.2.5 to exempt RLC requirement when [Package Model] present 3.3.3. {LEVEL 3} [Diff Pin] Vdiff and Tskew correct - "Tskew" is actually "Tdelay, Tdelay_min, Tdelay_max" in IBIS - We decided to eliminate the "corresponds to datasheet" language. - Discussion of how the Tdelay is measured: - David suggested that it should be the difference in Vmeas crossing - Question: is it measured at the pads or at the pins? - Mike said that EDA tools simply change the launch times of + and - signals. - Therefore it has to be measured at the pads. - Discussion of how Vdiff is measured: - Moshiul pointed out that [Receiver Thresholds] has Vdiff_ac, similar - Bob said that this overrides [Diff Pin] Vdiff - ibischk may not check Vdiff_ac conflicts thoroughly - Roy pointed out that vendors make Vdiff datasheet limits wider than necessary. - IBIS Vdiff should be the maximum value, which leads to pessimistic timing - 3.3.2 is the basic check, 3.3.3 adds only a datasheet check - We voted to eliminate 3.3.3 AR: Mike delete 3.3.3 from IQ spec 3.3.2. {LEVEL 3} [Diff Pin] Vdiff and Tdelay_* complete and reasonable - This check is not clearly written, in light of the previous discussion. AR: Roy clarify 3.3.2 3.3.4. {LEVEL 3} [Diff Pin] referenced pin models matched - Kim said the "check to make sure you intended this" is not good language for a requirement. - It was felt that using the same buffer for both sides of a diff pair is normal. - Mike said that bench measured IBIS files tend to have separate models for each side, even if the 2 have nearly identical data. - If buffer models referenced by the two physical pins of a [Diff Pin] are different, comments should be required to explain why. AR: Mike change 3.3.4 to require a comment for unique +/- diff pair models Next meeting: 22 May 2007 11-12 AM EST (8-9 AM PST) Phone: 1.877.384.0543 or 1.800.743.7560 Passcode: 90437837 Meeting ended at 12:16 PM Eastern Time.