IBIS Macromodel Task Group Meeting date: 04 Sep 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 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 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 Sigrity: * Brad Brim Kumar Keshavan Ken Willis 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: - None -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Greg provide examples to show what IC vendors need - Done - Michael M provide examples to show what IC vendors need - In progress - Arpad prepare example BIRD 125 IBIS file set - Done ------------- New Discussion: Arpad showed an example IBIS file: - Arpad: The [Package Model] keyword looks the same - The [Define Package Model] keyword also looks the same - The [Pin] keyword calls MODELCALL in the model column, a reserved word - Comments in [Pin] show {} substitution - This examples makes no use of [Diff Pin] - [Node Declarations] give diff pair nodes, both signal and pad - Port_map parameters within [Model Call] make the node connections - There are multiple Diff_Signal_pins, but IBIS doesn't allow that yet - [Circuit Call] makes the package connections - [External Circuit] is used to connect the on-die model - Subckts are parameterized - In [Define Package Model] the [Pin Numbers] keyword is retained - In [Package Circuit] the connections are made to declared nodes - The pins and pads are matched positionally even if list lengths do not match - A parameterized variable is used for RLGCMODEL to make it more flexible Arpad showed a presentation: - slide 28: - Arpad: This shows ports connected to pins - Also buffer nodes are connected to pad nodes Back to the example IBIS file: - Arpad: This makes it possible to have "sliding pins" - Michael M: This is assigned explicitly to a pair of pins - This is a simplified swath scheme, but it does not cover cases in between - Arpad: This is only for a diff pair model - If there were more it could show coupling - Michael M: Somehow a subset must be intelligently selected, the user or the tool - Arpad: This is more a sliding mechanism than swathing - No edge effects are involved here - Michael M: We have to look at where the burden of computation falls - Does the EDA tool do it or the model maker? - Arpad: The tool has to do a reduction before we get to this point - James: That is too restrictive - It would be better to start from a syntax with no assumptions - How are model names resolved from the [Pin] table? - In the RX model there are no node names - Arpad: The model has implicit legacy analog and digital IO ports - Using BIRD 145 we can use their predefined names to make the connections in [Model Call] - Walter: This shows 3 TX and 3 RX with different lengths - [Model Call] would need 128 lines to cover the full chip, twice - [Port Map] would need 256 entries, twice - The lengths are not swathed, unless a parameter handles that - It would take quite a bit to represent a fairly standard package - Arpad: Parameters can handle the length variations - James: The connection mechanisms are complicated, still not clear to me - Mike L: A diagram might help - Arpad: The slide is for that, but it does not have the needed labels Walter showed "Morphing IBS to EMD and IBISx": - slide 3: - Walter: We will show legacy.ibs, incremental.ibs, package.emd and bare_die.ibsx for a real part - slide 5: - Walter: Everything in red is new - [Tstonefile] is added, same as BIRD 122 - slide 6: - Walter: This has keywords to call an IBIS-ISS package - The indentation is for readability, although IBIS might not like it - slide 7: - Walter: [Circuit Call] is better than [External Model] for calling Touchstone - Touchstone is LTI - For [Model] we require IBIS-BSS - Touchstone is all that is needed today for on-die models - slide 8: - Walter: We only need one pin list - Parameters allow us to add other information - EMD files do not contain silicon, they call silicon models - slides 9 & 10: - Walter: In the future EMD could call IBISx - Extended nets make it easy to connect pins in one to another - slide 11 & 12: - Walter: IBISx is completely in parameter tree format - It could have equation based models too - slide 13: - Walter: We should table [External Model] BIRDs - They might be needed when on-die models beyond Touchstone are needed - James: Package pad nodes are connected to silicon pads? - Walter: For silicon the pads appear to be pins - James: [Extended Nets] is a required keyword? - Walter: Yes - The concept is important for power modeling where power pins don't map 1-1 to pads - Michael M: The number of things modeled is set by the model maker - There is no swathing - If the model has an s50p the user has ti figure out how to reduce it - Walter showed a 12 port example - The EDA tool should do it transparently - At least one EDA tool does this - James: How to designate when crosstalk is desired? - Walter: Aggressors are marked with A, others are victims - James: This predetermines the aggressors - Walter: That is the way vendors are delivering models - Some vendors may come up different requirements that will need other keywords - These have been discussed before and they will come up again - We really need IC vendors to say what their needs are Arpad encouraged all to provide feedback and more information. ------------- Next meeting: 11 Sep 2012 12:00pm PT Next agenda: 1) Task list item discussions ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives