IBIS Macromodel Task Group Meeting date: 13 March 2018 Members (asterisk for those attending): ANSYS: Dan Dvorscak * Curtis Clark Broadcom (Avago): Xingdong Dai Bob Miller Cadence Design Systems: * Ambrish Varma Brad Brim Kumar Keshavan Ken Willis eASIC: David Banas Marc Kowalski Ericsson: Anders Ekholm GlobalFoundries: Steve Parker IBM Luis Armenta Trevor Timpane Intel: Michael Mirmak Keysight Technologies: Fangyi Rao * Radek Biernacki Ming Yan Maxim Integrated Products: Hassan Rafat Mentor, A Siemens Business: John Angulo * Arpad Muranyi Micron Technology: * Randy Wolff Justin Butterfield SiSoft: * Walter Katz Todd Westerhoff * Mike LaBonte SPISim: * Wei-hsing Huang Synopsys: Rita Horner Kevin Li Teraspeed Consulting Group: Scott McMorrow Teraspeed Labs: * Bob Ross TI: Alfred Chong The meeting was led by Arpad Muranyi. Curtis Clark took the minutes. -------------------------------------------------------------------------------- Opens: - None. ------------- Review of ARs: - Arpad to prepare a draft of a new BIRD to supersede BIRD158.7. - Done. -------------------------- Call for patent disclosure: - None. ------------------------- Review of Meeting Minutes: - Arpad: Does anyone have any comments or corrections? [none] - Bob: Motion to approve the minutes. - Walter: Second. - Arpad: Anyone opposed? [none] ------------- New Discussion: BIRD189 and BIRD158 related issues: - Arpad: [sharing his draft of a BIRD to supersede BIRD158.7] - Title - Simply added "Revised" in front of the original BIRD158 title. - Analysis Path and Background Information - Unchanged from BIRD158. - Open to suggestions on whether we should leave the full history in this new BIRD or start fresh. - At the start of the textual changes for the new BIRD, state that the text shall supersede the text in BIRD158.7. - Remove the following text from all three locations (Tx, Rx, channel figure): "Note: The triangle ground symbols in the Tx, Rx and channel circuits represent the same node. This would typically be the global ground, such as node 0 in IBIS-ISS." - In the Tx and Rx figures, replace it with: "The reference node, represented by the triangle ground symbol in the above figure is the local ground node, A_gnd of the IBIS [Component]." - In the full channel figure, replace it with: "Note: The triangle ground symbols in the above figure represent the local ground node, A_gnd of the IBIS [Component] for the buffer and package models. These local ground nodes would typically have to be connected to the channel model’s reference node in simulation, which may be the global ground, such as node 0 in IBIS-ISS. The channel’s Touchstone model is not limited to a single reference terminal (N+1) for all of its ports. If the channel’s model has independent reference terminals for its Tx and Rx sides, the local ground nodes of the Tx and Rx [Component]s may remain independent nodes." - Walter: I think it would be better to say: "The reference node represented by the triangle symbol in the above figure is the local reference node of the IBIS [Model]". - (Note: [Model] as opposed to [Component]) - The only local references we really know about are the [* Reference]s of the [Model]. - Bob: I agree it should be [Model]. - Arpad: Agreed. - Discussion: Arpad noted that the new text under the full channel figure was intended to call attention to the fact that the reference nodes need not all be the same. For example, if the channel model were an N+2 arrangement, then one could have independent reference nodes on the Tx and Rx sides of the channel. However, he then referred to an email he had subsequently sent to the ATM in which he had suggested it might be better to remove the entire section containing the full channel figure. Given that each BIRD189 model could itself be an N+2 or 2N arrangement, the independent reference node discussions could also apply to each BIRD 189 model. Text to explain all of the possibilities could get cumbersome, and it wasn't necessary since BIRD158 is really only concerned with buffer modeling. Walter agreed on removing the section. He noted that just stating that the triangle was the local reference node of the [Model] was enough. He noted that even if we left the figure in the BIRD, it was only necessary to say that the triangles represent the local reference. Arpad noted that he thought it was safer to remove the figure altogether and not get into channel discussions at all, since this BIRD is not about channel design. Radek was not in favor of removing the section entirely. He noted that the ports for each block are defined with the reference nodes for the block. Without proper connections for the references you wouldn't have proper port connections. He noted that what Arpad had described with his cascaded, N+2 channel, configuration was okay under certain circumstances, but that we gained nothing from having independent local references. He noted that he preferred what Arpad had written utilizing the A_gnd syntax. Then BIRD189 need only say to use A_gnd for referencing when used with AMI models. Radek noted that he would simplify the discussion of the two sides of the channel that accompanied the full channel figure, but he would not delete it altogether. Radek also noted that "The channel’s Touchstone model" should be replaced with different text because the channel model need not be a Touchstone file. Arpad took the AR to produce a new draft of this BIRD incorporating the feedback from the meeting. Ground Recommendations: [presentation reviewing recent email discussions] - Walter: [sharing his presentation] - [slide 3] IBIS ISS and "ground" - IBIS ISS does support Node 0. - Many uses of "reference" already. - Very few occurrences of "ground" - Slides 4-6 define changes to eradicate "ground" from IBIS-ISS. - [slide 4] IBIS-ISS page 17 - "ground" --> "simulator reference node" - [slide 5] IBIS-ISS page 36 - "Ground reference" --> "Reference node" - [slide 6] IBIS-ISS page 39 - "ground (reference line)" --> "reference line" - With these simple changes, we can rid IBIS-ISS of "ground". - [slide 7] Correct Voltage Measurements - Voltage measured between a point and a second reference point. - Reference point should be located as near as possible to the measured point. - Component manufacturer is required to specify those two points. - The data book should specify the signal names that are references for I/O voltage measurements. - Data books typically call these nodes "ground" or rail voltage = 0. - Bob: Data books are often for external, not internal, measurements. - Walter: Yes, they're typically for pins. - A proper voltage measurement at a pin is made relative to a nearby pin. - Arpad: Data books are usually written with pin thresholds, etc., assumed with respect to the ground pin you defined in the previous paragraph. - Walter: Yes. - [slide 8] IBIS is an Electronic Data Book - We need to accept that "ground" was the traditional way of saying that certain nodes were the reference nodes for measurements. - 20 years ago, "nearby" was on the chassis. - We know better now. - [slide 9] "Name" of Reference - Signal Ground - Local Ground - Local Reference - Reference Node - Reference - We want to eliminate use of "ground". I recommend "Local Reference" - [slide 10] - IBIS describes measurements. - It does state that when generating IV, VT, and threshold data the reference nodes are connected to a simulator reference node (e.g. Node 0). - IBIS does not describe how to measure voltages during simulation. - [slide 11] How to change "ground" to "reference" in IBIS 6.1 - Over 100 occurrences of "ground" in IBIS 6.1 - Rather than replace them all, I would prefer to add a statement at the beginning of IBIS regarding "ground". - Model Makers and EDA tools shall interpret "ground" in the IBIS 6.1 specification as a Reference Node for each voltage measurement. - [slide 12] A_gnd and Node 0 - A_gnd added as part of [External Model], where the analog signals for Stimulus and Enable terminals would typically have a different reference than the I/O and Rail terminals. - IBIS supports Berkley SPICE and IBIS-ISS, both of which support Node 0. - We can say A_gnd is an alias for Node 0 (???). - A_gnd (???) and Node 0 can be a reference for Interconnect Models, but this limits the way power aware simulations can be performed. - Arpad: The most recent changes to BIRD189 changed A_gnd from a global to a local reference node. - Therefore, I'm not sure A_gnd can be an alias for Node 0. - Walter: Then you have problems with [External Model], too. - [Walter added the question marks to the last two bullets of page 12]. - [slides 13, 14] "Mixed Reference" simulations. - BIRD189 elements can reference a Buffer, Pad, or Pin rail terminal, or A_gnd or Node 0. - Results may be incorrect if some models reference Node 0 and others do not. - Multiple vendors may generate models used in a given channel simulation. - What to do if one of these references Node 0? - [slide 15] What to do if one of these references Node 0? - Reference to Scott McMorrow quote regarding valid ground based power aware modeling. - [slide 16] Conclusions - Make Minor changes to IBIS-ISS to remove "ground." - Add section at the beginning of IBIS stating that voltage measurements are made with respect to a "Reference Node". The reference node is only Node 0 when describing methods to generate IBIS data. - BIRD189 should recommend that Interconnect Model elements reference local rail voltage terminals, and suggest that all reference terminals be connected to a common reference node if any reference (that current flows to) is A_gnd or Node 0. - Discussion: Bob noted this was a very useful summary document. Arpad noted that it contained good cleanup recommendations. Arpad noted that the IBIS-ISS portion was easy. With respect to the IBIS changes, Arpad noted that he preferred to undergo the editorial effort to remove the uses of "ground", rather than try to address it with an upfront statement that users might not see. He asked if Walter envisioned these changes in 7.0. He asked if it was even possible to push the changes off until after 7.0 when BIRD189 is intended for 7.0. Walter said he preferred to put the upfront statement into 7.0. Bob noted that we had previously decided to defer the referencing cleanup until 7.1, as the changes got too big and complicated. - Curtis: Motion to adjourn. - Walter: Second. - Arpad: Thank you all for joining. AR: Arpad to prepare a new draft of the BIRD to supersede BIRD158.7. ------------- Next meeting: 20 March 2018 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives