IBIS Macromodel Task Group Meeting date: 21 April 2015 Members (asterisk for those attending): ANSYS: * Dan Dvorscak * Curtis Clark Avago (LSI) Xingdong Dai * Bob Miller Cadence Design Systems: * Ambrish Varma Brad Brim Kumar Keshavan Ken Willis eASIC * David Banas Ericsson: Anders Ekholm IBM Steve Parker Intel: Michael Mirmak Keysight Technologies: * Fangyi Rao * Radek Biernacki * Nicholas Tzou Maxim Integrated Products: Hassan Rafat Mentor Graphics: * John Angulo * Arpad Muranyi Micron Technology: * Randy Wolff Justin Butterfield QLogic Corp. James Zhou Andy Joy eASIC Marc Kowalski SiSoft: * Walter Katz Todd Westerhoff * Mike LaBonte Synopsys Rita Horner Teraspeed Consulting Group: Scott McMorrow Teraspeed Labs: * Bob Ross (Note: Agilent has changed to Keysight) The meeting was led by Arpad Muranyi. ------------------------------------------------------------------------ Opens: - None -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Arpad, Randy and Radek prepare BIRD 175 draft ready for Open Forum. - Still discussing details. - Walter find all places in IBIS specification affected by PAM4 BIRD. - Done and distributed. - Michael M update AMI Directionality BIRD - No update - Arpad to review IBIS specification for min max issues. - In progress. ------------- New Discussion: Bob Miller introduction: - Design engineer and IBIS-AMI modeler at Avago. BIRD 175: - Arpad: An update was supposed to go out Thursday but we are still working on it. - The deadline has been missed for a May 1 Open Forum vote. - It will be discussed next Tuesday. PAM4 BIRD: - Walter showed the BIRD draft. - Walter: All known updates have been made. - This is waiting for review comments. - The next IBIS will have to support PAM4 as well as NRZ. - Many places in IBIS talk about zero crossings. - This is really about transitions. - Bit time is really symbol time. - It would help for others to check this work. - Bob R: "Rest of IBIS" means "rest of IBIS-AMI"? - Walter: The key changes are in AMI. - Arpad: Legacy models will not support PAM4. - Walter: PAM4 still assumes an analog model from legacy curves. - It can only go from symbol 0 to symbol 3, without EQ. - Arpad: Then no changes are needed outside of IBIS-AMI. - Walter: Agree. - An introductory paragraph should appear somewhere. Back-channel BIRD: - Walter showed two draft BIRDs. - Arpad: Was this supposed to be a modification of BIRD 147? - Walter: That was not my intention. - Ambrish: At DesignCon we agreed to look at BIRD 147 again. - Walter: The change to BIRD 147 as structured would be massive. - The BIRD on the left defines AMI parameters for back-channel. - The BIRD on the right describes a BCI protocol for TX-RX communication. - Ambrish: Why are they not one BIRD? - Walter: The management should be agnostic to the protocol. - I'm defining a "basic" protocol. - The BIRD on the left must be approved first. - The basic protocol is a superset of BIRD 147. - It does not prohibit other protocols. - The EDA tool does not need to understand the messages. - Radek: The word "not" is missing. - Walter corrected a sentence. - Walter: Models can use a private protocol. - Public protocols can use the BIRD process. - Nicholas: This will support models from different vendors? - Walter: Yes, they just need to support the same protocol name. - Public protocols should be well supported. - Two vendors could support a private protocol. - The EDA tools would handle it because they don't care what is being passed. - Backchannel_Protocol gives the protocol name. - BCI_State has fixed allowable values. - BCI_GetWave_Block_Size lets the RX set a block size requirement. - Mike: Isn't block size in samples? - Walter: Yes but it should be given in UI quantities. - BCI_GetWave_Training says if time domain training is supported. - BCI_Init_After_GetWave says if Init should be trained after GetWave is done. - An informative table is included to describe allowed combinations. - Training flows are shown in diagrams. - Radek had suggested we do need a new function for a second Init call. - Radek: I did suggest a separate function. - Walter: We can go either way on that. - Arpad: The SiSoft logo should not go in the IBIS specification. - Walter: We need to decide if we want the graphics at all. - The EDA tool call sequence is listed in text too. - This is informative, other variations are possible. - There is a note about number of bits which need not be known in advance. - The RX will terminate when done. - Ignore_Bits should be replaced with waiting until training is done. - Only the PRBS7 pattern is called for by every known standard. - If a protocol name ends in .bci that means there is a .bci file. - There is no need for the EDA tool to read that. - Bob R: If that is standardized does that mean the .bci file is standardized? - The parser would have to check it for compliance. - Walter: That would be up to the protocol being incorporated into IBIS. - RX eye quality metrics should be output. - Silicon programming tools should be provided, as a best practice. - An 802.3KR example AMI file is provided. - Ambrish: What is a .ibc file? - Walter: That should be .bci. - Ambrish: How does a model prepare for what it will receive from other models? - Walter: A standard protocol must be written. - My "basic" protocol would be one. - Bob M: The extra AMI_Impulse function will be needed. - Otherwise neither TX nor RX has saved any history. - Ambrish: Models w=know when they are in BC mode. - Bob M: AMI_Init will start from scratch each time. - Walter: I had proposed that when the allocated memory handle is passed back in it does not start fresh. - It doesn't matter much which way we go. - Walter switched to the basic protocol document. - Walter: This describes a Back-Channel Protocol called "basic". - BCP here could stand for Black Capped Pigeon, a "messenger" BIRD. - The TX would output a description of itself. - Bob R: So BCI is a reserved branch name. It can be repeated several times? - Walter: No there can be only one. - Bob R: It is in the AMI file? - Walter: No it is only in the DLL parameter string in our BNF format. - Ambrish: The TX would be restricted to just tap parameters? - Walter: The TX describes whatever parameters it will allow changes to. - The example that looks like AMI file format is just for documentation. - Neither the tool nor model reads that. - Bob R: It sends a sequence of values? - Walter: Each bottom level parameter has only one value here. - The RX will create a string with an instruction in it. - The instructions use one of several methods such as "increment" and "decrement". - The TX needs to support each method. - Nicholas: How would a third party optimizer work? - Walter: That is not in this document. - Nothing prevents the EDA tool from updating tap coefficients. - The command does not have to come from the RX. - Only published protocols could be supported by tools. - Bob M: If the EDA tool is agnostic do the TX and RX communicate by file I/O? - Walter: The EDA tool finds the BCI branch and copies it. - Radek: This is a good argument for the special API function. - That is a good way to do it. - Walter: I can change the BIRD if we decide that. - Ambrish: How does one create a private protocol? - Walter: The contents of the BCI string can be anything. - Bob R: Does this depend on BIRD 128? - Walter: Yes, if the existing AMI_GetWave function is used. - Bob R: There is no Protocol_Specific branch? - Walter: There is not. - This should be easy for AMI model developers. AR: Walter send back-channel BIRD drafts to Mike for posting. ------------- Next meeting: 28 Apr 2015 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives