IBIS Open Forum Minutes Meeting Date: May 16, 2012 Meeting Location: SPI-E IBIS Summit, Sorrento, Italy VOTING MEMBERS AND 2012 PARTICIPANTS Agilent Radek Biernacki, Yoji Sekine, Fangyi Rao Altera David Banas, Hsinho Wu, Masashi Shimanouchi AMD (Nam Nguyen) ANSYS (Ansoft) Flavio Calvano* Apple Computer (Matt Herndon) Applied Simulation Technology Norio Matsui Cadence Design Systems Terry Jernberg, Jilin Tan, Dennis Nagle Ambrish Varma Cisco Systems David Siadat, Mike Sapozhnikev Ericsson Anders Ekholm, Zilwan Mahmod, Mattias Lundquist Foxconn Technology Group (Sogo Hsu) Freescale (Jon Burnett) Green Streak Programs (Lynne Green) Huawei Technologies Xiaoqing Dong IBM Adge Hawes, Greg Edlund Infineon Technologies AG (Christian Sporrer) Intel Corporation Michael Mirmak, Udy Shrivastava, Stewart Gilbert Eddie Frie IO Methodology Lance Wang* LSI Brian Burdick Maxim Integrated Products Hassan Rafat, Mahbubul Bari Mentor Graphics Arpad Muranyi, Vladimir Dmitriev-Zdorov Micron Technology Randy Wolff*, Aniello Viscardi*, Francesco Madonna* Antonio Prisco* Nokia Siemens Networks GmbH Eckhard Lenski*, Evis Minga* QLogic (James Zhou) Signal Integrity Software Walter Katz, Mike LaBonte, Todd Westerhoff Akalu Lentiro Sigrity Raymond Chen, Yingxin Sun, Sam Chitwood Ben Franklin*, Srdjen Djordjevic* Synopsys Andy Tai, Scott Wedge, Hany Elhak, Ted Mido Teraspeed Consulting Group Bob Ross, Tom Dagostino Texas Instruments Casey Morrison, Pegah Alavi, Valerie Chang Modesto Garcia, Karl Muth Toshiba (Yasumasa Kondo) Xilinx Harry Fu ZTE (Huang Min) Zuken Michael Schaeder OTHER PARTICIPANTS IN 2012 Aruba Networks Amir Motamedi Avago Technologies Sanjeev Gupta, Amolak Badesha Bayside Design Elliot Nahas Cavium Johann Nittmann Enterasys Robert Haller Exar Corporation Helen Nguyen Granite River Labs Jiang Xu, Johnson Tan Hewlett-Packard Ting Zhu High Speed Design Center Ben Chia Hitachi Yutaka Uematsu* Instituto de Telecomunicações Wael Dghais*, Hugo Teixeira* KEI Systems Shinichi Maeda Luxshare-ICT Alan Kinningham, Steven Wong MK Resolutions Olga Chervyakova Nvidia Eric Hsu Politecnico di Torino Stefano Grivet-Talocia* Research In Motion Yi Cao Sony Kenji Yasoda ST Microelectronics Davide Pandini* TechAmerica (Chris Denham) Université de Brest Mihai Telescu*, Cherif Diouf* University of Illinois Thomas Comberiate, José Schutt-Ainé* Vitesse Semiconductor Sirius Tsang In the list above, attendees at the meeting are indicated by *. Principal members or other active members who have not attended are in parentheses. Participants who no longer are in the organization are in square brackets. UPCOMING MEETINGS The bridge numbers for future IBIS teleconferences are as follows: Date Meeting Number Meeting Password June 1, 2012 205 475 958 IBIS For teleconference dial-in information, use the password at the following website: https://ciscosales.webex.com/ciscosales/j.php?J=205475958 All teleconference meetings are 8:00 AM to 9:55 AM US Pacific Time. Meeting agendas are typically distributed seven days before each Open Forum. Minutes are typically distributed within seven days of the corresponding meeting. When calling into the meeting, follow the prompts to enter the meeting ID. For new, local international dial-in numbers, please reference the bridge numbers provided by Cisco Systems at the following link: http://www.cisco.com/web/about/doing_business/conferencing/index.html NOTE: "AR" = Action Required. ---------------------------------------------------------------------------- --------------------------------------------------- OFFICIAL OPENING The IBIS Open Forum Summit was held in Sorrento, Italy at the Parco dei Principi – Sorrento Hotel following the 2012 SPI conference. About 18 people representing 11 organizations attended. The notes below capture some of the content and discussions. The meeting presentations and other documents are available at: http://www.eda.org/ibis/summits/may12/ Lance Wang welcomed all the participants and thanked the co-sponsors Micron Technology, Nokia Siemens Networks and Zuken. He asked all the participants to introduce themselves. There were a wide variety of people from many countries and organizations including academia and industry. IBIS MODELING USING LATENCY INSERTION METHOD (LIM) José Schutt-Ainé*, Jilin Tan**, Ping Liu**, Feras Al-Hawari**, Ambrish Varma**, *University of Illinois at Urbana-Champaign, **Cadence Design Systems, USA José Schutt-Ainé presented. He noted the use of a different approach to simulation. The usual Newton-Raphson method makes a guess for a solution and iterates until the solution converges. Sometimes this method locks into oscillations and the solution will not converge. The Latency Insertion Method (LIM) has convergence advantages. It does not use matrix formulations. Each branch must have an inductor, and each node must have a shunt capacitor. LIM is easy to code and is fast. LIM has no convergence issues, but it suffers from stability issues. LIM can be applied to IBIS. First, Ku and Kd coefficients must be solved from the V-t waveforms. Then, a LIM formulation can be used for the circuit simulation. José also showed an example of modeling gate modulation effects (BIRD98.3). Eckhard Lenski asked where the problems with non-convergence of Newton- Raphson on slide 28 came from. José responded that he did not discover specific issues with the IBIS models, just that the solution would not converge. A question was asked if the package parasitics might cause problems with Ku/Kd solutions, leading to failures of the Newton-Raphson method. José noted that Newton-Raphson non-convergence is very common. SSO NOISE AND CONDUCTED EMI: MODELING, ANALYSIS, AND DESIGN SOLUTIONS Patrice Joubert Doriol#, Aurora Sanna#, Akhilesh Chandra##, Cristiano Forzan#, and Davide Pandini#, STMicroelectronics, #Italy, ##India Davide Pandini summarized that he would discuss the use of IBIS models to solve signal integrity and power integrity challenges. Conducted EMI coming from signals is a problem at ST. EMC measurements for compliance were traditionally done at the end of the design flow, and if the design had issues, new mask sets would be very expensive. EMC-aware design is the goal, creating EMC models and doing EMC simulation before fabrication. Some noise can be injected on a stable victim from a switching aggressor. This causes conducted emissions on the victim I/O signal. Davide summarized the full SSO analysis flow. He showed a micro-controller design for an automotive application. Skewing of the I/O timing achieved a significant reduction in the amplitude of several harmonics. Skews of 10ns were possible due to the slow working frequencies of the technology. Reducing the I/O drive strength also further reduced the SSO noise. The gate-level SSO flow compared well to the Spice-level simulation. IBIS 4.0 models did not work well for the SSO simulations, but IBIS 5.0 models compared well to Spice up to 2GHz in frequency domain analysis. He concluded that IBIS 5.0 can be used for reliable SSO and EMI analysis. Eckhard Lenski asked what kind of package models were used. Davide replied that the models were RLC. Lance Wang asked how the frequency domain analysis waveforms were generated. Davide responded that an FFT of the time domain waveforms was used. Stefano Grivet-Talocia asked why the frequency matched only to 2GHz. Lance commented that it might be from poor modeling of C_comp frequency dependence. Davide added that the V-t waveforms look like they match well, but zooming in, there are some differences. VERSATILE SURROGATE MODELS FOR IC BUFFERS Cherif Diouf*, Mihai Telescu*, Igor Stievano**, Flavio Canavero**, P. Clostre*, N. Tanguy*, *Université Européenne de Bretagne, Université de Brest, France; **Politecnico di Torino, Italy Mihai Telescu presented modeling work of the authors. IC buffers are highly non-linear systems. Behavioral modeling techniques include IBIS and Mπlog. Previous approaches to modeling include Mπlog plus the use of Volterra series and constraints to improve static behavior. He introduced the theoretical framework for a new modeling approach. The model for a driver was implemented in Spice and validated against two different system loads. A model of cascaded drivers was then used to have a more non-linear system and was compared against the previous loads. Correlation was acceptable with room for improvement. There is ongoing research on more complex and recent buffers. They are working to tune the method to the needs of the industry. They would like feedback on the complexity versus the versatility desired. Wael Dghais asked if the weighting factors on slide 12 were related to the K coefficients in IBIS algorithms. Mihai responded that they were not related. Mahai clarified for Lance Wang that a least-squares method was used to determine the coefficients to fit the equations to the Spice results of the training simulations. Mihai clarified for Randy Wolff that the training simulations on slide 11 must be given ranges that will incorporate desired system loads, but more feedback is desired in this area. TOWARDS REAL TIME S-PARAMETER QUALIFICATION AND MACROMODELING Stefano Grivet-Talocia, Politecnico di Torino, Italy Stefano Grivet-Talocia introduced S-parameters and their use in SI/PI simulations. S-parameters must be qualified by looking for issues with passivity and causality. Circuit simulators assume that models are causal, so when non-causal S-parameters are used, they lead to unpredictable results. A generalized Hilbert transform can be used for generating a causality check error function. Macromodeling can be used to very accurately model S-parameter responses with rational function fitting of poles and zeros. He described a method developed to take advantage of multi- core processors to create the vector-fit macromodels quickly and efficiently. Passivity of the macromodel can be checked by looking at the eigenvalues. Computation of the eigenvalues works well with parallel processing. Stefano described the HPM service developed by Politecnico di Torino that is an online resource for creating S-parameter macromodels. Randy Wolff asked about support of IBIS-ISS as an output. Stefano said that support of this would be easy to add if there is a need. Lance Wang asked about support of Touchstone 2.0. Stefano responded that mixed mode support is not available yet, but it is being worked on. Mihai Telescu asked about use of the modeling to fix passivity and causality issues. Stefano responded that the macromodeling can fix issues, but it doesn't guarantee that the model is a better model of the original system than the original Touchstone file with its issues. Also, if the input data is highly non-causal, the macromodel will not be a good fit to the original S- parameters. IMPACT OF ACCURATE PDN MODEL ON IBIS SI AND PI SIMULATIONS Antonio Girardi and Aniello Viscardi, Micron Semiconductor Italia S.r.l., Italy Randy Wolff presented for the authors. An investigation of several power supply decoupling models was carried out on a LPDDR2 memory device. 20 I/Os were switched in an SSO simulation that included a coupled package model and a system load model. IBIS 5.0 driver models included BIRD95 and BIRD98 data tables. Four cases of PDN models were simulated including a case with no power supply decoupling, a capacitive-only model, an RC model, and a vector-fit decoupling model. The C-only model created a resonance with the package inductance that led to unrealistic ringing in the voltage waveform at the power pad. The RC model was created manually, and the addition of the ESR to the C-only model led to good agreement between behavioral simulation and Spice simulation. The vector-fit decoupling model was created automatically and led to a closer match than the RC model. The vector-fit model was a Spice subcircuit created from the S11 PDN response. Randy noted that there was not a large difference between the RC and the vector-fit models in the case shown, but there could be cases where the vector-fit model provides much better results than the RC, manually-fitted model. Randy clarified that the vector-fit subcircuit model could be easily made IBIS-ISS compliant and currently would need to included externally to an IBIS 5.0 model. He noted in response to a question from Eckhard Lenski that Micron is currently looking into providing IBIS 5.0 models for certain memory devices. DDR4 IBIS POWER INTEGRITY SIMULATION Randy Wolff*, Lance Wang**, *Micron Technology, **IO Methodology, USA Randy Wolff began with an overview of the new DDR4 memory technology. He noted that he wanted to look at potential issues with Power Integrity modeling of DDR4 with IBIS 5.0. Overclocking was a concern with the faster bit rates compared to DDR3. An automated IBIS 5.0 buffer extraction tool was used to create the I-V, V-t and I-t data tables. Randy noted that the tool used a best-points-fit algorithm, although this was not a major benefit with 1000 points available for the V-t and I-t tables. However, it is inevitable that customers will ask for an IBIS 3.2 compliant model, and in that case it will be necessary to remove the I-t data and limit the V-t data to 100 points. A best-points-fit algorithm will be needed for this. Overclocking was a concern, because the bit length at DDR4-1600 is 625ps and the [Composite Current] data was 750ps long in the typical corner and 850ps long in the slow corner. These lengths also required windowing of the data from tables that were 1.2ns in length. Data windowing may be automatic in some simulators, but may require manual adjustment in others. Simulations did reveal overclocking problems at DDR4-1600. The model was then simulated at DDR4-1866, and the problems got much worse. Manual windowing of the data tables was necessary, but a bug was found where the voltage waveforms looked ok, but the current waveforms were shifted in time. Other simulations revealed reasonable correlation between Spice and IBIS results. Randy also noted that a vector-fit decoupling model might improve correlation. Eckhard Lenski asked if it was now necessary to warn customers about overclocking issues. Randy responded that one problem is that he can't test the models in every simulator, so the vendors need to step forward and provide more testing and validation. Known overclocking issues should be mentioned in model quality reports. Randy added that even in overclocking situations, the pre-driver current will always be there, so it must not be ignored. MORE EXPERIENCES WITH IBIS-AMI MODELS Eckhard Lenski, Nokia Siemens Networks GmbH, Germany Eckhard Lenski gave an overview of IBIS-AMI model usage. He noted that a good documentation file must accompany the model to help the user. He described the boundaries between the Tx, channel and Rx models. He also showed the IBIS-AMI parameters and Tx and Rx elements. His experiences using the models have varied, including issues such as models not working in both Windows XP and 7, models with the package model inside the .dll file, incomplete .ami files, mis-alignment of values between the .ibs file and the .ami file, and documentation that does not match the model. Randy Wolff added that his first and recent experience using an IBIS-AMI model revealed very poor documentation as well. Stefano Grivet-Talocia asked about LTI assumptions with AMI modeling. Discussion confirmed that the convolution used in AMI model processing requires LTI assumptions. Modeling of non-LTI effects in the AMI models is not perfect and the methods are hidden within the .dll file. EFFICIENT TABLE-BASED I-Q BEHAVIORAL MODEL FOR HIGH-SPEED DIGITAL BUFFERS/DRIVERS Wael Dghais*, Hugo Teixeira*, T.R. Cunha, J.C. Pedro, Universidade de Aveiro, *Instituto de Telecomunicações, Portugal Wael Dghais presented. He showed the structure of a digital I/O buffer. A new model creates an output impedance I-Q model. Q-V curves are extracted using AC simulation. I-V/C-V and I-V/Q-V models showed similar accuracy, but the I-V/Q-V model was faster to simulate. The new models showed better accuracy and matching to the transistor level model than the IBIS model. The models contain data tables very similar to IBIS models. Lance Wang asked about implementing the model in simulation. Wael confirmed that it requires changing the algorithms in the simulator as well as extracting a new data table. The approach does complement existing IBIS, using similar I-V data extracted from simulation. Mihai Telescu made some comments about this approach versus the Mπlog modeling. Mπlog modeling works with continuous time, whereas IBIS and this approach work in discrete time. AD HOC PRESENTATION Stefano Grivet-Talocia on request from Lance Wang presented on details of rational curve fitting algorithms. The slides used were a subset of his presentation “Passive Macromodeling” given at EMC Europe on 8/9/2008 in Hamburg, Germany. CLOSING REMARKS Lance Wang closed the meeting by thanking the co-sponsors and the presenters. He also thanked all the attendees for making the meeting a success. The meeting concluded at approximately 6:00 PM. NEXT MEETING The next IBIS Open Forum teleconference will be held June 1, 2012 from 8:00 to 10:00 AM US Pacific Standard Time. ======================================================================== NOTES IBIS CHAIR: Michael Mirmak (916) 356-4261, Fax (916) 377-3788 michael.mirmak@intel.com Data Center Platform Applications Engineering Intel Corporation FM5-239 1900 Prairie City Rd., Folsom, CA 95630 VICE CHAIR: Lance Wang (978) 633-3388 lwang@iometh.com President/CEO, IO Methodology, Inc. PO Box 2099 Acton, MA 01720 SECRETARY: Randy Wolff (208) 363-1764, Fax: (208) 368-3475 rrwolff@micron.com SI Modeling Manager, Micron Technology, Inc. 8000 S. Federal Way Mail Stop: 01-711 Boise, ID 83707-0006 LIBRARIAN: Anders Ekholm (46) 10 714 27 58, Fax: (46) 8 757 23 40 ibis-librarian@eda.org Digital Modules Design, PDU Base Stations, Ericsson AB BU Network Färögatan 6 164 80 Stockholm, Sweden WEBMASTER: Mike LaBonte mikelabonte@eda.org IBIS-AMI Modeling Specialist, Signal Integrity Software 6 Clock Tower Place Maynard, MA 01754 POSTMASTER: Mike LaBonte mikelabonte@eda.org IBIS-AMI Modeling Specialist, Signal Integrity Software 6 Clock Tower Place Maynard, MA 01754 This meeting was conducted in accordance with the TechAmerica Legal Guides and TechAmerica Manual of Organization and Procedure. The following e-mail addresses are used: majordomo@eda.org In the body, for the IBIS Open Forum Reflector: subscribe ibis In the body, for the IBIS Users' Group Reflector: subscribe ibis-users Help and other commands: help ibis-request@eda.org To join, change, or drop from either or both: IBIS Open Forum Reflector (ibis@eda.org) IBIS Users' Group Reflector (ibis-users@eda.org) State your request. ibis-info@eda.org To obtain general information about IBIS, to ask specific questions for individual response, and to inquire about joining the IBIS Open Forum as a full Member. ibis@eda.org To send a message to the general IBIS Open Forum Reflector. This is used mostly for IBIS Standardization business and future IBIS technical enhancements. Job posting information is not permitted. ibis-users@eda.org To send a message to the IBIS Users' Group Reflector. This is used mostly for IBIS clarification, current modeling issues, and general user concerns. Job posting information is not permitted. ibis-bug@eda.org To report ibischk parser BUGs as well as tschk2 parser BUGs. The BUG Report Form for ibischk resides along with reported BUGs at: http://www.eda.org/ibis/bugs/ibischk/ http://www.eda.org/ibis/bugs/ibischk/bugform.txt The BUG Report Form for tschk2 resides along with reported BUGs at: http://www.eda.org/ibis/tschk_bugs/ http://www.eda.org/ibis/tschk_bugs/bugform.txt icm-bug@eda.org To report icmchk1 parser BUGs. The BUG Report Form resides along with reported BUGs at: http://www.eda.org/ibis/icm_bugs/ http://www.eda.org/ibis/icm_bugs/icm_bugform.txt To report s2ibis, s2ibis2 and s2iplt bugs, use the Bug Report Forms which reside at: http://www.eda.org/ibis/bugs/s2ibis/bugs2i.txt http://www.eda.org/ibis/bugs/s2ibis2/bugs2i2.txt http://www.eda.org/ibis/bugs/s2iplt/bugsplt.txt Information on IBIS technical contents, IBIS participants and actual IBIS models are available on the IBIS Home page: http://www.eda.org/ibis Check the IBIS file directory on eda.org for more information on previous discussions and results: http://www.eda.org/ibis/directory.html To create an account on the TechAmerica KAVI workspace, check out: http://workspace.techamerica.org/kwspub/join/ Other trademarks, brands and names are the property of their respective owners. IBIS CURRENT MEMBER VOTING STATUS I/O Buffer Information Specification Committee (IBIS) |Organization |Interest |Standar|March |April |May 11,|May 16,| | |Category |ds |30, |20, |2012 |2012 | | | |Ballot |2012 |2012 | | | | | |Voting | | | | | | | |Status | | | | | |Advanced Micro |Producer |Inactiv|- |- |- |- | |Devices | |e | | | | | |Agilent |User |Active |X |X |X |- | |Technologies | | | | | | | |Altera |Producer |Active |X |- |X |- | |ANSYS |User |Inactiv|- |- |- |X | | | |e | | | | | |Apple Computer |User |Inactiv|- |- |- |- | | | |e | | | | | |Applied Simulation|User |Inactiv|- |- |- |- | |Technology | |e | | | | | |Cadence Design |User |Inactiv|X |- |- |- | |Systems | |e | | | | | |Cisco Systems |User |Inactiv|- |- |- |- | | | |e | | | | | |Ericsson |Producer |Active |X |X |X |- | |Foxconn Technology|Producer |Inactiv|- |- |- |- | |Group | |e | | | | | |Freescale |Producer |Inactiv|- |- |- |- | | | |e | | | | | |Green Streak |General |Inactiv|- |- |- |- | |Programs |Interest |e | | | | | |Huawei |Producer |Inactiv|- |- |- |- | |Technologies | |e | | | | | |IBM |Producer |Active |X |X |X |- | |Infineon |Producer |Inactiv|- |- |- |- | |Technologies AG | |e | | | | | |Intel Corp. |Producer |Active |X |X |X |- | |IO Methodology |User |Active |X |- |X |X | |LSI |Producer |Inactiv|- |- |- |- | | | |e | | | | | |Maxim Integrated |Producer |Inactiv|- |- |- |- | |Products | |e | | | | | |Mentor Graphics |User |Active |X |X |X |- | |Micron Technology |Producer |Active |- |- |X |X | |Nokia Siemens |Producer |Active |X |X |- |X | |Networks | | | | | | | |QLogic |Producer |Inactiv|- |- |- |- | | | |e | | | | | |Signal Integrity |User |Active |X |X |X |- | |Software | | | | | | | |Sigrity |User |Active |X |X |- |X | |Synopsys |User |Inactiv|- |- |- |- | | | |e | | | | | |Teraspeed |General |Active |X |X |X |- | |Consulting |Interest | | | | | | |Texas Instruments |Producer |Inactiv|- |- |- |- | | | |e | | | | | |Toshiba |Producer |Inactiv|- |- |- |- | | | |e | | | | | |Xilinx |Producer |Inactiv|- |- |- |- | | | |e | | | | | |ZTE |User |Inactiv|- |- |- |- | | | |e | | | | | |Zuken |User |Inactiv|- |- |- |- | | | |e | | | | | Criteria for Member in good standing: • Must attend two consecutive meetings to establish voting membership • Membership dues current • Must not miss two consecutive Meetings Interest categories associated with TechAmerica ballot voting are: • Users - Members that utilize electronic equipment to provide services to an end user. • Producers - Members that supply electronic equipment. • General Interest - Members are neither producers nor users. This category includes, but is not limited to, Government, regulatory agencies (state and federal), researchers, other organizations and associations, and/or consumers.