[ibis-macro] Minutes from the 22 and 29 September 2009 ibis-atm meetings
- From: "Mike LaBonte (milabont)" <milabont@xxxxxxxxx>
- To: "IBIS-ATM" <ibis-macro@xxxxxxxxxxxxx>
- Date: Tue, 6 Oct 2009 14:14:48 -0400
Minutes from the 22 and 29 September 2009 ibis-atm meetings are
attached. Also the Minutes page at
http://www.eda.org/ibis/macromodel_wip/ is now fixed and up to date,
having for the past several months listed minutes only through July as
the result of a code bug.
Mike
IBIS Macromodel Task Group
Meeting date: 22 September 2009
Members (asterisk for those attending):
Adge Hawes, IBM
Ambrish Varma, Cadence Design Systems
Anders Ekholm, Ericsson
* Arpad Muranyi, Mentor Graphics Corp.
Barry Katz, SiSoft
* Bob Ross, Teraspeed Consulting Group
Brad Brim, Sigrity
Brad Griffin, Cadence Design Systems
Chris McGrath, Synopsys
David Banas, Xilinx
Deepak Ramaswany, Ansoft
Donald Telian, consultant
Doug White, Cisco Systems
* Eckhard Lenski, Nokia-Siemens Networks
Essaid Bensoudane, ST Microelectronics
* Fangyi Rao, Agilent
Ganesh Narayanaswamy, ST Micro
Gang Kang, Sigrity
Hemant Shah, Cadence Design Systems
Ian Dodd, consultant
Jerry Chuang, Xilinx
Joe Abler, IBM
* John Angulo, Mentor Graphics
John Shields, Mentor Graphics
Ken Willis, Cadence Design Systems
Kumar Keshavan, Sigrity
Lance Wang, Cadence Design Systems
Luis Boluna, Cisco Systems
* Michael Mirmak, Intel Corp.
* Mike LaBonte, Cisco Systems
Mike Steinberger, SiSoft
Mustansir Fanaswalla, Xilinx
Patrick O'Halloran, Tiburon Design Automation
Paul Fernando, NCSU
Pavani Jella, TI
Radek Biernacki, Agilent (EESof)
* Randy Wolff, Micron Technology
Ray Komow, Cadence Design Systems
Richard Mellitz, Intel
Richard Ward, Texas Instruments
Samuel Mertens, Ansoft
Sam Chitwood, Sigrity
Sanjeev Gupta, Agilent
Shangli Wu, Cadence Design Systems
Sid Singh, Extreme Networks
Stephen Scearce, Cisco Systems
* Steve Pytel, Ansoft
Syed Huq, Cisco Systems
Syed Sadeghi, ST Micro
Ted Mido, Synopsys
Terry Jernberg, Cadence Design Systems
* Todd Westerhoff, SiSoft
Vladimir Dmitriev-Zdorov
Vikas Gupta, Xilinx
Vuk Borich, Agilent
Walter Katz, SiSoft
Zhen Mu, Mentor Graphics
------------------------------------------------------------------------
Opens:
- Fangyi: Any update on the parser?
- Michael M: We received a new draft code
- It is behaving as expected
- Minor typos in error strings
- Need volunteers to compile for platforms other than DOS and Linux
- Bob: Would like to have Reserved_parameters relaxed
- That will take a BIRD
- Michael M: We need to know what platforms people want soon
- Arpad: Email from Eric Monteiro
--------------------------
Call for patent disclosure:
- No one declared a patent.
-------------
Review of ARs:
- Walter to send his presentation to Mike LaBonte.
- Done
- Arpad to add flow changes to his block diagram presentation and send to
Mike LaBonte.
- Done
- Mike change web page to make items easily linkable
- in progress
- Arpad Write a clarification BIRD to discuss accuracy issues related to the
various AMI clock_tick algorithms in an IBIS-AMI DLL
- TBD
- Todd: Update the BIRD for IBIS S-parameter box based on feedback from
discussion
- No update
- Arpad: Write parameter passing syntax proposal (BIRD draft)
for *-AMS models in IBIS that is consistent with the
parameter passing syntax of the AMI models
- TBD
- TBD: Propose a parameter passing syntax for the SPICE
- [External ...] also?
- TBD
- Arpad: Review the documentation (annotation) in the macro libraries.
- Deferred until a demand arises or we have nothing else to do
-------------
New Discussion:
Arpad showed an email from Eric Monteiro:
- It contains 2 links referencing SiSoft presentations
- Eric sees conflicting statements:
- One page says to measure Rx input
- Other pages say to measure Rx output
- Arpad showed the presentation to verify the pages
- Todd: The first slide does not represent what AMI does
- It says "Traditional SerDes Simulator Flow"
- Arpad: It should be measured at the Rx pad for channel characterization
- Todd: Agree, the receiver's network should load the channel
- Fangyi: There is no other choice
- Arpad showed email from Eric asking if measured at point D
- Slide 3 IBIS-AMI Serial Channel has arrows that do not denote measure points
- Fangyi: So we are arguing the definition of a channel?
- Arpad: There are separate pictures for channel characterization and AMI flow
- Fangyi: How does this reflect the asymmetric relationship between rise and
fall?
- Arpad: That would be a long discussion
- Todd: We have to fix Tx_GetWave
- It is left to model vendors
Arpad showed Walter's document "New Reference Flow":
- page 12:
- One bullet has (if both Tx and Rx Use_Init_Output are true)
- Arpad: What about the other cases?
Arpad showed his document AMI Flows:
- This shows the flows graphically
- Page 1 is Walter's proposal, page 2 is current
- Only difference between pages 1 & 2 are order between Tx_GetWave and
convolution
- Fangyi: Tx_GetWave is inside Tx, so why is channel response included?
- Todd: The original idea was to let filters accommodate the channel
- But currently that makes no sense
- Page 3:
- Walter's proposal with true/false branching shown
- Fangyi: The first implementation does not have Tx_Init Return Filter
- Those booleans are not needed
- Arpad: Walter's presentation explains the need
- For Rx the convolution output is the same regardless of the boolean
- The truth table for Tx_GetWave operation has 4 outcomes
- John: Is Walter depending on the exact times of the bit stream?
- Arpad: The wavy equal sign indicates there may be non-LTI effects
- Fangyi: Why can't we put filter into channel response?
- Todd: There is no good reason to do it
- The model should just return the impulse response
- Fangyi: It would eliminate one flag
- Arpad: The flag is there for backward compatibility
- We could say if Use_Tx_Init_Output is true then Use_Init_Returns_Filter
is true
- Fangyi: There should then be a note about that
- The "or" in the notes would have to go away
- Page 4:
- Tx_GetWave convolution truth table has only 2 entries
- Page 5:
- Same as page 4 but with Rx_GetWave, no Tx_GetWave
- Page 6:
- Neither Tx_GetWave nor Rx_GetWave
- There is only 1 truth table entry
- John: So given LTI the wavy equal is now an equal
- Page 7:
- Waveform comes from some other tool such as SPICE
- Use_Tx_Init_Output must be false
- Fangyi: So we do not need Init_Returns_Impulse
- Arpad: This is for backward compatibility
- Fangyi: Old models will not work with the new flow
- Todd: I wouldn't want to write that parser
- Now that we see the flow we can see things we might get rid of
- Todd: I need to be able to say that Init does nothing
Next meeting: 22 Sep 2009 12:00pm PT
--------
IBIS Interconnect SPICE Wish List:
1) Simulator directives
IBIS Macromodel Task Group
Meeting date: 29 September 2009
Members (asterisk for those attending):
Adge Hawes, IBM
Ambrish Varma, Cadence Design Systems
* Anders Ekholm, Ericsson
* Arpad Muranyi, Mentor Graphics Corp.
Barry Katz, SiSoft
* Bob Ross, Teraspeed Consulting Group
Brad Brim, Sigrity
Brad Griffin, Cadence Design Systems
Chris McGrath, Synopsys
* Daniel Kirsanov, Ansoft
David Banas, Xilinx
Deepak Ramaswany, Ansoft
Donald Telian, consultant
Doug White, Cisco Systems
Eckhard Lenski, Nokia-Siemens Networks
Essaid Bensoudane, ST Microelectronics
* Fangyi Rao, Agilent
Ganesh Narayanaswamy, ST Micro
Gang Kang, Sigrity
Hemant Shah, Cadence Design Systems
Ian Dodd, consultant
Jerry Chuang, Xilinx
Joe Abler, IBM
John Angulo, Mentor Graphics
John Shields, Mentor Graphics
Ken Willis, Cadence Design Systems
Kumar Keshavan, Sigrity
Lance Wang, Cadence Design Systems
Luis Boluna, Cisco Systems
* Michael Mirmak, Intel Corp.
* Mike LaBonte, Cisco Systems
Mike Steinberger, SiSoft
Mustansir Fanaswalla, Xilinx
Patrick O'Halloran, Tiburon Design Automation
Paul Fernando, NCSU
Pavani Jella, TI
Radek Biernacki, Agilent (EESof)
* Randy Wolff, Micron Technology
Ray Komow, Cadence Design Systems
Richard Mellitz, Intel
Richard Ward, Texas Instruments
Samuel Mertens, Ansoft
Sam Chitwood, Sigrity
Sanjeev Gupta, Agilent
Shangli Wu, Cadence Design Systems
Sid Singh, Extreme Networks
Stephen Scearce, Cisco Systems
* Steve Pytel, Ansoft
Syed Huq, Cisco Systems
Syed Sadeghi, ST Micro
Ted Mido, Synopsys
* Terry Jernberg, Cadence Design Systems
* Todd Westerhoff, SiSoft
* Vladimir Dmitriev-Zdorov, Mentor Graphics
Vikas Gupta, Xilinx
Vuk Borich, Agilent
* Walter Katz, SiSoft
Zhen Mu, Mentor Graphics
------------------------------------------------------------------------
Opens:
- Mike L said he needs to get caught up on posting meeting minutes
- Arpad: What should EDA tools do with parameter_string_out?
- Walter: OUT or INOUT can be on any model-specific parameter
- Reserved parameters can be INFO
- Fangyi: How does the tool use model-specific parameters?
- Walter: For example, if tap_coeff is OUT, the model will return the value
- Fangyi: But the name may not be the same in all tools
- Walter: We discussed this 2 years ago
- New parameters are bound to appear
- If they are published, other vendors could adopt the same names
- Then they could become reserved parameters
- Fangyi: The EDA tool should only observe reserved parameters
- What does the specification imply the tools should do?
- Todd: Nothing
- Arpad: Would like to discuss:
- List of new BIRDs
- What is needed to release IBIS-ISS?:
--------------------------
Call for patent disclosure:
- No one declared a patent.
-------------
Review of ARs:
- Mike change web page to make items easily linkable
- Arpad dropped this from the AR list
- Arpad Write a clarification BIRD to discuss accuracy issues related to the
various AMI clock_tick algorithms in an IBIS-AMI DLL
- TBD
- Todd: Update the BIRD for IBIS S-parameter box based on feedback from
discussion
- No update
- Arpad: Write parameter passing syntax proposal (BIRD draft)
for *-AMS models in IBIS that is consistent with the
parameter passing syntax of the AMI models
- TBD
- TBD: Propose a parameter passing syntax for the SPICE
- [External ...] also?
- TBD
- Arpad: Review the documentation (annotation) in the macro libraries.
- Deferred until a demand arises or we have nothing else to do
-------------
New Discussion:
Walter's AMI proposal:
- Fangyi: PMC has reservations about the new flow
- They are trying to do statistical analysis
- They add DFE to the impulse by simple addition
- Assuming Init includes DFE breaks their flow
- The have to convolve, not add
- Walter: Tx_Init_Returns_Filter option can be set to do it the old way
- Vladimir: This flow is good because it corresponds to a physical process
- It has the flexibility to control if convolution is used
- There should be a different flow if there is crosstalk noise
- Walter: So sometimes it would be better to convolve inside Init so EDA tool
doesn't have to
- But that makes it difficult to process for noise
- For this the filter should be an input to Tx_Init
- Vladimir: The question is about GetWave
- We combine 3 responses into one for performance
- But for crosstalk we can't combine them
- Walter: Tx_Use_Init_Output is flawed, and should be false
- We should let Tx_GetWave do the convolution, instead of EDA tool
- Arpad: Is that a new flow?
- Walter: No, the option parameters allow it now
- Tx_Use_Init_Output = false
- When you call Tx_GetWave it has the filter impulse response
- This is what the dotted line represents
- Arpad: I did not expect the dotted line to mean that
- Walter: The DLL saves it in memory
- Cadence did not want to convolve inside GetWave
- They wanted an overlap and save method
- Fangyi: You suggest Init should never modify the impulse response
- Walter: Init returns the modified Response
- But it also has it in memory
- Fangyi: The specification says RX_Init can modify the impulse response
- Walter: That was a fundamental error
- We said 2 years ago that would break the crosstalk flow
- The new proposal fixes that
- Fangyi: The digital input can be recovered?
- Walter: Yes
- Fangyi: The Rx has to process only the main channel, not crosstalk
- That breaks the flow
- The proposed new flow is cleaner
- Arpad: I just modified the presentation to reflect what we have discussed
- Fangyi: We need to settle the -0.5 to +0.5 or 0 to 1 question
- Walter: It is -0.5 to +0.5
- Transitions occur nominally at 0
- Putting the old flow through a lossless channel results in -0.5 to +0.5
output
- Vladimir: The point is that the magnitude has to be 1
- Fangyi: The DC bias is important
- Vladimir: Can the specification be simplified?
- Did IC vendors require all these options?
- Walter: We could deprecate the old modes in 5.1
- Arpad show IBIS 5 page 188 that says samples are "centered around zero"
- Arpad: If we eliminate the convolution options can the arrows go away?
- Walter: They can go away
- Arpad corrected math expressions for the Tx_GetWave and Rx_GetWave outputs
- Walter: If Tx_GetWave is missing it must be replaced with a convolution
- Arpad: I will clean this up and Mike L will post it
Walter showed an outline of AMI documentation
- The string format was an error and should be deprecated
- Table has been a problem
- We will want to make some model_specific params reserved
- They should be deprecated
- We should support env vars
- We should support vectors, labels, scale, and limit
- There should be naming rules for parameter leafs
- Propose new reserved params:
- published
- tstonefile
- true_diff_analog_model
- Computed parameters:
- Some tools support more than 3 corners
- A language like Perl with if/else might do it
- It will take a compromise of EDA vendors
- We should make many modifications in one block
- Walter could modify the IBIS specification with these changes
- Arpad: We should discuss it before it goes to the Open Forum
- Walter: We should have a small EDA vendor subcommittee for this
- Walter, Arpad, Fangyi, ...
- Bob: It has to be open to anyone
- Want to see 6 vendors participate
- Bob: There should be no need to deprecate
- Walter: The EDA vendors should decide that
AR: Walter email Arpad on status of proposed AMI specification rework
Next meeting: 06 Oct 2009 12:00pm PT
--------
IBIS Interconnect SPICE Wish List:
1) Simulator directives
Other related posts:
- » [ibis-macro] Minutes from the 22 and 29 September 2009 ibis-atm meetings - Mike LaBonte (milabont)
