IBIS Macromodel Task Group Meeting date: 18 Dec 2012 Members (asterisk for those attending): Agilent: * Fangyi Rao * Radek Biernacki Altera: * David Banas Julia Liu Hazlina Ramly Andrew Joy Consulting: Andy Joy ANSYS: Samuel Mertens * Dan Dvorscak * Curtis Clark Steve Pytel * Luis Armenta Arrow Electronics: Ian Dodd Cadence Design Systems: Terry Jernberg * Ambrish Varma Feras Al-Hawari Brad Brim Kumar Keshavan Ken Willis Cavium Networks: Johann Nittmann Celsionix: Kellee Crisafulli Cisco Systems: Ashwin Vasudevan Syed Huq Ericsson: Anders Ekholm IBM: Greg Edlund Intel: * Michael Mirmak Maxim Integrated Products: Mahbubul Bari Hassan Rafat Ron Olisar Mentor Graphics: * John Angulo Zhen Mu * Arpad Muranyi Vladimir Dmitriev-Zdorov Micron Technology: Randy Wolff Justin Butterfield NetLogic Microsystems: Ryan Couts Nokia-Siemens Networks: Eckhard Lenski QLogic Corp. * James Zhou SiSoft: * Walter Katz Todd Westerhoff Doug Burns * Mike LaBonte Snowbush IP: Marcus Van Ierssel ST Micro: Syed Sadeghi Teraspeed Consulting Group: Scott McMorrow * Bob Ross TI: Casey Morrison Alfred Chong Vitesse Semiconductor: Eric Sweetman Xilinx: Mustansir Fanaswalla Ray Anderson The meeting was led by Arpad Muranyi ------------------------------------------------------------------------ Opens: - Arpad: This is the last meeting this year - The next meeting will be 8 Jan 2013 - James: Is the recent email topic on the agenda? - Arpad: It is not planned - It might be brought up in interconnect tomorrow - Fangyi: The recent topic is more urgent -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Walter send BIRD 150.1 draft 7 to list - Done - Mike post BIRD 150.1 draft 7 to web - Done - Walter define Usage of dependency table parameters - Fangyi, Mahbubul, et. al., to discuss merging their "Redriver" proposal with BIRD 131. - No progress ------------- New Discussion: Interconnect task group report: - Michael M: We focused on SI2, details will have to wait until next year - We may need more people from this meeting to resolve some interconnect issues - We will be taking up both analog and digital interactions Walter showed "Backchannel with Repeaters": - slide 1: - Walter: A repeater can be a retimer or just a redriver - slide 3: - Walter: Redrivers generally are non-LTI - Retimers contain a CDR, maybe DFE, FFE, regenerate the stimulus - Fangyi: Redrivers can have filters and an FFE too - slide 4: - Walter: The industry uses the term "training" for backchannel - slide 5: - Walter: With a redriver the backchannel connect RX directly to TX - Michael M: It depends on what the redriver is capable of - They are controlled different ways, but they may be backchannel capable - slide 6: - Walter: Redrivers are simple, low power. - Retimers are more sophisticated - The RX trains the retimer, which trains the TX - slide 7: - Walter: Redrivers here are defined as not able to be trained - slide 8: - Walter: This is how people think it works - FAMOUS LAST WORDS - Walter: We need standardized terminology for the repeater pins - Fangyi: We call them input and output pins - With Walter's BIRD the output is associated with an input - My plan calls for the user to set it up - Ambrish: My BIRD has other useful features - Arpad: Should the BIRDs be combined? - Fangyi: The flows need to be different for redriver and retimer - Ambrish: Redrivers do not necessarily have to use AMI models - Fangyi: Then Walter's idea to have separate BIRDs is valid - Walter: We never defined the output of an RX model - On an [External Model] we do, but not legacy IBIS - The AMI RX is defined - Ambrish: We might work this into the BIRD - Michael M: One question is if we need probes for an analog-only repeater - Otherwise this can be a black box - Walter: What language would the black box be written in? - It is non-LTI so ISS can't be used, but BSS could - Arpad: Wouldn't it have to be LTI, as AMI requires? - Walter: Only the channel interaction is LTI - Only the algorithmic part is non-LTI - Michael M: AMS models could be used - Walter: Fangyi and I need to work together on this - We should have a report for the next meeting - Michael M: This is high priority - Fangyi: I don't agree that redrivers are not trainable - Walter: If the redriver is trained it has to retrain the TX - Arpad: Could the RX train both redriver and TX? - Walter: Not by 803.2bj - Michael M: The redriver is usually thought to be transparent - Ambrish: Should the protocol be handled in our spec? - Walter: The designer decides this in the BIOS setup - EDA tool might help automate it Analog models: - Walter: In IBIS 5.1 no analog model is defined - Michael M: You can have zeroed-out IV models - Walter: An IBIS Input requires no IV curves at all - In IBIS 5.1 the DLL can not affect impulse response creation - Ambrish: Correct - Walter: James proposed a way to have the DLL modify the channel response - James: Two issues: - What should models do in IBIS 5.1? - The emails have clarified that - The analog channel should have high Z input - What if models do not follow the convention? - Have seen models that do not, created by members of this group - From one vendor one TX assumes 0 ohm, other assume 50 ohm - Signals can only flow forward - We could let model makers use any declared impedance - If an s-param is used that impedance is always there - Walter: Agree, we need a specified on-die s-param analog model - It has to declare the termination parameters - Ambrish: Analog models have been put in the DLL? - Walter: No - James: My point is that we do not have to force people to use 0 ohms - Michael M: We are running into TX interoperability problems right now - In some cases ramp and IV have been embedded in the DLL - We must distinguish between analog info in DLL vs. AMI - I don't think anyone wants it in the DLL - Ambrish: Then the AMI model is not a legal IBIS model - Michael M: Some problems are cause by assumptions about where the VT data is - Embedding in the DLL may be bad for bandwidth reasons - Arpad: Ramp is required in IBIS - Fangyi: BIRD 116 might address the impedance issue - Arpad: It replaces C_comp and IV with ISS - Radek: We have had that discussion, we decided TX is a 0 ohm source - Defining source impedance should be doable right away - Walter: IBIS 5.1 says that - Fangyi: It just says "high impedance interface" - Arpad: It says "high impedance connection" - That is even more vague - Radek: There is no requirement that the input impedance is infinite - James: We agree on interpretation of IBIS 5.1 - Arpad: IBIS 5.1 does not allow s-parameters - James: If output is 0 ohms then the s-parameter is -1 - Michael M: There is no way to test this for current implementations Arpad: We might continue this tomorrow Arpad thanked all for their hard work during the year ------------- Next meeting: 08 Jan 2013 12:00pm PT Next agenda: 1) Task list item discussions ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives