====================================================================== IBIS FUTURES/COOKBOOK SUBCOMMITTEE MEETING MINUTES Date: May 19, 2005 Previous Meeting Attendees: ---------- Cadence - Lance Wang Cisco Systems - Syed Huq, Zhiping Yang Intel - Arpad Muranyi Mentor - Ian Dodd Siemens (Munich) - Katja Koller Teraspeed - Bob Ross Green Streak Programs - Lynne Green- ====================================================================== 1 PM - 3 PM US Pacific Time Telephone Bridge Passcode 916-356-2663 4 747-9612 Agenda: 1 - 1:05 PM Opens - Bob/Syed 1:10 - 1:30 BIRD97.1 - Arpad 1:30 - 2:00 BIRD95.5 - Syed 2:00 - 2:20 BIRD94.2 - Arpad 2:20 - 2:40 Cookbook - Arpad 2:40 - 2:50 Other items 2:50 - 3:00 ?? ------------------------------------------------------------------------- Upcoming Futures Meeting: Date: Thursday, June 2, 2005 Bridge: 1 Passcode: 495-7208 ====================================================================== Futures Minutes (abbreviated discussion) Syed Huq ran the meeting since Michael Mirmak was out. PREVIOUS MINUTES CORRECTIONS Bob Ross noted these corrections: Under BIRD97.1, the statement: k = [1 + Pc*(Vinst-Vnom)/Vnom]**a Katja stated that the power a is between 1 (long channel) and 2 (short channel). ... should be corrected to: Katja stated that the power a is between 1 (short channel) and 2 (long channel). ... Under FSA UPDATE (new item), Lynne Green added the following correction/clarification: The group performing the IC (transistor & interconnect) modeling is the Device Technology Council, a subgroup of Si2, http://www.si2.org/ The slides were presented at a meeting of the FSA modeling group. The FSA is working on the SPICE Quality document. OPENS No new items were proposed. BIRD97.2 Arpad Muranyi stated that at the last IBIS teleconference meeting, Donald Telian suggested adopting a table format rather than an equation format. Arpad asked whether he and Katja Koller should proceed in this direction. A few advantages of each format were noted. Ian Dodd mentioned that fixed equation physics are less reliable for emerging smaller geometry technologies since fringe effects are becoming more dominant. Possible table processing slowdown is not a big factor for EDA tools. Arpad and Katja mentioned that the tables did not have to be long. A table format avoids the problem of fitting equations. The equation format in BIRD97.1 is more concise. The group suggested holding BIRD97.1, but introducing BIRD98 with a slightly modified title. Otherwise the changes would be hard to follow. Bob Ross stated that typ-min-max table entries would still be supported. The I-V limit of 100 rows would still exist, although fewer rows (10 to 15) might be suggested as adequate. Optionally, separate tables would be provided for each I-V table. The extraction method needs to be documented. It would document output current shorted to Vcc and in low-state for the pulldown table gate width modulation data. The Vcc voltage is swept, but the temperature and process are constant. The output is connected to ground and in the high-state for the pullup table modulation data. Bob indicated that the BIRD should suggest how to use the table data (convert the differential currents versus voltage in to scaling factors for the corresponding pulldown/pullup tables) to convey the intended usage of the table data. Lance Wang emphasized that this BIRD is still separate from BIRD95.5. In fact, his experience is that gate modulation is a more significant factor in SSN. AR - Arpad will issue a new BIRD. BIRD95.5 Lance Wang questioned whether the [Composite Current] includes the [Power Clamp Reference] current when the [Power Clamp Reference] is not connected to the [Pullup Reference]. Bob Ross stated that the [Composite Current] captures only the current through the [Pullup Reference] (power reference terminal). Only when the [Power Clamp Reference] is connected is it included. The EDA tool might have to do some bookkeeping to deal with whether the clamp current is included based on connectivity. The tool might have to process the information differently based on the [Power Clamp Reference] connected to or different from the [Pullup Reference]. Arpad Muranyi questioned the [External Reference] entry in the context diagram. Bob mentioned that it is included in the reference diagram completeness since it could be a separate power supply entry point. However, it could also be an internally generated reference voltage or an a reference point set up by an external resistor divider to other supplies. In any case, it is used for non-driving modes (Input) and sould not be a factor in SSN analysis. Lance commented on using Arpads tutorial class SPICE models. Lance noted that the current shape remains roughly constant down to 2.5 V. At 2 V, the shape changes. Arpad mentioned that the SPICE model was based on 5 V technology, and the threshold voltages might be above 1 V. Lance is continuing the investigation of using [Composite Current]. BIRD94.2 Discussion was deferred since no one has done any further work. COOKBOOK Arpad Muranyi sent out a revision -AM3a of the draft Cookbook with highlights in red. His main addition was to show how to do a V-T table extraction with circuitry added to cancel out the differential C_comp (C_diff). It is canceled in the same manner as the differential R component (R_diff). Arpad amended the existing Section 3.2.5.10 and added four lines of SPICE code to cancel the differential impedance effects. So the actual differential impedance is factored out, but is added in later to the model with [R Series] and [C Series]. The V-T tables based on the canceled differential impedance should produce simulations that match the actual device characteristics when the assumed impedances are added to the differential model. Arpad asked people to review these pages around 26-27. We look forward to Michael Mirmak issuing an cleaned up version 8 draft with the noted corrections and deletions applied.