====================================================================== IBIS FUTURES/COOKBOOK TASK GROUP MEETING MINUTES Date: September 29, 2005 Attendees: ---------- Cadence Design Systems - *Lance Wang Cisco - *Syed Huq, *Zhiping Yang Intel - *Michael Mirmak Mentor Graphics - *Ian Dodd Micron - *Randy Wolff NCSU - *Ambrish Varma Teraspeed - *Bob Ross ====================================================================== Next Meeting: Thursday, October 6, 2005 9 AM - 11 AM US Pacific Time Telephone Bridge Passcode 916-356-2663 4 557-9368 Agenda: 9 - 9:10 AM Opens - Phase Delay Discussion in Open Forum BIRD95 Discussion 9:10 - 9:25 AM - Michael Mirmak: Case 1 & 2 Outcomes 9:25 - 10:00 AM - Lance Wang: Quiet Line Sim. Results; Summary of BIRD95 & Add'l Proposals 10 - 11 AM - Group Discussion, as needed ======================================================================== FUTURES --------- The team discussed the findings from Michael Mirmak comparing various cases with and without ground-connected parasitics. Ambrish Varma noted that the loop inductances do not match between the three cases (30 nH is used in case2, while 33 nH is used in case1). Michael stated that cases 1 and 3 were supposed to be equivalent, while case 2 was deliberately crafted to have a slightly different loop inductance, to gauge the impact. The mutuals were added by Zhiping. Zhiping stated that Case 1 and 3 are only identical in the self- inductance loops. Mutuals must also be counted. He added that the L33 mutual gets added into L11 and L22 in a matrix, and provided the matrix calculations. Zhiping noted that no mutual inductances were present in the BIRD95 Micron tests. An error is likely in his version of Michael Mirmak's test decks. He will send an updated version to Michael. Sam stated that a corrected version should be equivalent, even without the ground inductances. He added that the circuit can be modeled with S-parameters or time-domain RLCs. Arpad suggested that, for a good comparison, one has to measure at the local input and local ground, rather than versus "gnd" or "0." Lance Wang presented a number of materials related to C_comp splitting and also to power supply variation effects on output results. Arpad and Zhiping commented that the 0.9 V source used in some decks for output load biasing in Lance's examples should be referenced to die ground, not ideal ground; additionally, the same loading references should be used. Syed stated that Cisco's simulation deck and test cases are now publicly available. Lance has a different viewpoint on state-dependency of Cpu and Cpd C_comp; he suggests that BIRD95's approach may not be accurate if the state dependencies are not somehow included, if the C_comp splits don't properly account for them. Arpad noted that we cannot address results mismatches merely by making our C_comps state-dependent; phase delay is destroying the frequency response of our buffers. Arpad will clarify his point in a future presentation. To demonstrate this in Lance's decks, Arpad suggested Lance simplify the stimulus pattern he used and provide results for a series of sweeps. ARs ----- Michael Mirmak - update drawing with node names -- DONE Lance Wang - for phase delay issue, update sim buffer stimulus to transition only from 1.8 to 1.7 and "go horizontal"; sweep slope and watch output response; update decks to show sources and loads referenced to local ground John Angulo - analyze BIRD100 proposals, after Mentor discussion Sam Chitwood - analyze BIRD100 proposals, after Sigrity discussion Lance Wang - analyze BIRD100 proposals, after Cadence discussion