IBIS Macromodel Task Group

Meeting date: 13 May 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:

- None

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

- None

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

- Ambrish to send presentation to Mike to be posted.
  - Done

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

Back-channel:
- Walter showed Backchannel, Training and Co-optimization BIRD Introduction and Flows.
- slide 3:
  - Walter: 
    - Need to know:
      - Is there a solution?
      - What is the solution?
    - DFE here applies to all EQ, not necessarily a DFE.
- slide 3:
  - Walter: the 802.3 specifications will make your head spin.
- slide 5:
  - Walter: Training makes assumptions about the TX.
    - Must be able to convert coefficients to tap indexes, among other things.
- slide 6:
  - Walter: Taps are represented three ways:
    - hardware registers.
    - indexes.
    - coefficients.
  - Training/Co-optimization deals with coefficients and indexes.
- slide 7:
  - Walter: The TX can't optimize itself.
- slide 8:
  - Walter described a 6 step training procedure.
  - Walter: Interval between commands to TX may be thousands of UI.
- slide 9:
  - Walter: Can we use existing TX models?
    - It would be helpful.
    - Reserved_Parameters can do this.
    - Some TX DLL changes may be needed,maybe not.
    - Can use existing tap names.
    - TX may not be able to make tap changes during TD simulation, but Init can.
- slide 10:
  - Walter: Peak-peak voltage should be controllable,if it exists.
    - Tap_Conversion would convert from index to coefficient, etc.
- slide 11:
  - Walter: Different TX optimization modes are possible.
- slide 13:
  - Walter: Boolean parameters to indicate RX optimization capabilities.
    - Vendors use PRBS training patterns, only need register length.
    - Tap_Coefficient would communicate actual tap.
    - Need a Boolean to indicate training status.
- slide 14:
  - Walter: Tap Coefficient Ranges are passed from TX to RX.
- slide 16:
  - Walter: Multiple protocols can be handled by branch labels.
- slide 17:
  - Walter: Presets are needed, but not sure where they should be.
    - Expressed in dB.
- slide 18:
  - Walter: The EDA tool picks presets, or tries them all.
    - 802.3bj tries all coefficients.
    - Doesn't take long because the operation is quick.
- slide 19:
  - Walter: It would be good to have TX models that have both of these.
- slide 20:
  - Walter described RX Init training sequence.
  - Ambrish: Who provides the input information here?
  - Walter: The EDA tool gives TX tap coefficients and ranges to RX Init.
- slide 21:
  - Walter: TX Init can be called again,to:
    - Verify tap coefficients.
    - Convert coefficients to indexes.
    - Create a new impulse response.
- slide 22:
  - Walter described TX GetWave sequence.
- slide 23:
    - Walter: Not all modes like coefficients may be supported.
- slide 24:
  - Walter: Walter described RX GetWave sequence.
- slide 25:
  - Walter: Tap coefficients from TD could be used for statistical processing.
- slide 28:
  - Walter described the procedure after training is done.

- Ambrish: Designers say they do a lot of changes.
  - It is not true that a generate TX can be used.
  - We do pattern dependent coefficients.
- Kumar: Standards only set minimum requirements.
  - The TX can be more complicated.
  - It can depend on power for example.
  - We can't legislate what the TX architecture should be.
- Walter: I'm saying some TX architectures can be used.
  - My approach can do everything BIRD 147 does.
  - A TX could have only Tap_Increment and do the same thing.
- Ambrish: On slide 12 what is the Model_Specific parameter?
- Walter: The Reserved Parameter points to a Model_Specific parameter that implements it.
- Ambrish: How does the RX know the ranges?
- Walter: The TX has Reserved_Parameters for that.
- Ambrish: The TX may reduced its range during training.
- Walter: The silicon has a range that can't change.
- Ambrish: Tap resolution won't change?
- Walter: Only if you change amplitude.
  - Some protocols don't allow that.
- Ambrish: How is private training done.
- Walter: A text field could be passed for that.
- Ambrish: BIRD 147 can handle things this can't.
- Walter: Private_Training could be a Reserved_Parameter.

- Ambrish: We have enough information now, we need to move forward.
- Ambrish motioned to vote on this next week.
- Walter seconded.
- There were no objections.
- A vote to either promote BIRD 147 or Walter's proposal will be held in the next meeting.

-------------
Next meeting: 13 May 2014 12:00pm PT

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

1) Simulator directives