IBIS Macromodel Task Group

Meeting date: 17 June 2014

Members (asterisk for those attending):
Agilent:                      Fangyi Rao
                            * Radek Biernacki
Altera:                     * David Banas
ANSYS:                      * Dan Dvorscak
                            * Curtis Clark
Avago (LSI)                   Xingdong Dai
Cadence Design Systems:       Ambrish Varma
                              Brad Brim
                              Kumar Keshavan
                            * Ken Willis
Ericsson:                     Anders Ekholm
Intel:                      * Michael Mirmak
Maxim Integrated Products:    Hassan Rafat
Mentor Graphics:            * John Angulo
                            * Arpad Muranyi
Micron Technology:            Randy Wolff
                              Justin Butterfield
QLogic Corp.                  James Zhou
                              Andy Joy
SiSoft:                     * Walter Katz
                            * Todd Westerhoff
                            * Mike LaBonte
Teraspeed Consulting Group:   Scott McMorrow
                            * Bob Ross

The meeting was led by Arpad Muranyi.

------------------------------------------------------------------------
Opens:

- Bob: IBISCHK6 is now available.
  - Michael M will send an announcement.

- Arpad: Cadence requested that we postpone BIRD 147 discussion, they are not available
  - We can go through our list of outstanding items
  - Michael M might discuss BUG 150

- Walter: I have a short presentation to give

- Arpad: I have a BIRD draft for item #9

--------------------------
Call for patent disclosure:

- None

-------------
Review of ARs:

- None

-------------
New Discussion:

BUG 150:

- Michael M: Walter had suggestions in the past about this.
  - Both SE and DE signaling are permitted.
  - Some can be I/O, how do we identify that?
  - This might be a BIRD.
  - Flows strongly imply that models are either input or output.
- Radek: A receiver can be an output?
- Arpad: A [Model] can be I/O and have an algorithmic model.
  - The Executable subparam could have a selector.
  - Or a single executable could be passed flag to give the direction.
- Walter:
  - Three options:
    1 - No spec change, use a [Model Selector], DDR does this.
      - There would be a different AMI file, the DLL could be the same.
    2 - Single DLL and AMI file, BIRD for Reserved_Parameter to declare input or output.
- Michael M: Would the DLL be used for both?
- Walter: They are very different, but both models can be in one DLL.
  - Actually there only two options.
- Ken: It would be most convenient to have separate DLLs.
  - Do we need to model bus turn-around?
  - That might force a single DLL.
- Michael M: We are being pushed into modeling more architectural information like that.
- Arpad: The impedance may change when it turns around.
- Todd: This will not work using statically measured impulse responses.
  - There will be at least three modes.
- Michael M: Turn-around can be excluded.
- Walter: A new Model_type AMI parameter might be used.
  - A BIRD would be needed.
- Michael M: A single DLL approach seems less likely.
- Arpad: [Model Selector] doesn't .
- Mike L: There could be many Executable lines.
  - The identifier could have Platform_Compiler_Bits_InOut.
- Michael M: That would be a small change.
  - Could an executable and AMI file be in conflict?
- Arpad: A Reserved parameter could check against that.

AR: Michael M write BIRD for I/O AMI models.

Walter showed a Redriver Flow Problem presentation:
- Walter: Darshan Shah described this problem.
  - It is limited to redrivers where GetWave is missing.
  - There are real cases of this.
  - RX2 optimizes incorrectly because it does not have equalized RX1 output.
- Walter showed the redriver flow BIRD.
- Walter: RX2 Init receives output of TX2.
- Arpad: So RX2 has no chance to adjust taps.
- Walter: IBIS had a flaw 6 years, letting the TX optimize itself.
  - It was OK at 3GHz but not at 6GHz.
  - It becomes simpler if we assume the TX doesn't optimize itself.
- Arpad: DO you propose a flow change for all situations?
- Walter: Just repeaters for now.
  - Statistical analysis become easier.
  - If any TX lacks GetWave, all RXs should have it.
  - Today only 1 TX optimizes itself, we can deprecate that.
  - Fangyi has suggested combinations that are broken.
- Arpad: Should we require all Init or all GetWave?
- Walter: For TX models GetWave is academic.
  - It can be emulated by giving a unit impulse instead of the channel.
  - Once a DFE settles it is LTI.
- Arpad: How does crosstalk affect this?
- Walter: Time domain is unaffected, statistical is an open question.
  - It would be interesting to see if anyone is using crosstalk in Init.
  - Also if any TX out there optimizes.
- Ken: We have TX models that optimize.
  - It can be disabled.
- Walter: Preferably the default would be to disable.
  - One model out there can not disable optimization.

- Arpad: Kumar and Ambrish will be missing next week.
  - We will continue these topics.

-------------
Next meeting: 24 Jun 2014 12:00pm PT

-------------
IBIS Interconnect SPICE Wish List:

1) Simulator directives