[ibis-macro] Response to AMI BIRD comments

From: "Walter Katz" <wkatz@xxxxxxxxxx>
To: "IBIS-ATM" <ibis-macro@xxxxxxxxxxxxx>
Date: Tue, 12 Jan 2010 09:52:38 -0500
All,


Thanks for everyones response.

I am combining all questions and responses chronologically in this E-Mail,
and inserting the comments and responses in the document that will be
reviewed later today.

Walter


Fangyi, 1/11


  1.       On page 11, statement ?All reserved parameters must contain
sub-parameters Usage, Type, and one of the Allowed-Values sub-parameters.?
is contradictory to statement ?If a Reserved Parameter must have one and
only one Usage, Usage is optional. If a Reserved Parameter must have one and
only one Type, Type is optional.? on page 7 and some examples.


  2.       On page 18, should the order of array values in the parameter
string follows the order of the leaves appear in the branch? If yes, how
does the ibis parser make sure such order information is passed to EDA
tools?


  3.       On page 24, the document states that ?The impulse values are in
volts?. I think the unit of impulse should be volt/sec because by definition
impulse*dt is in volt.


  4.       On page 25, should ?element[n]? be ?element[n,0]??

  Response,

  I added the following. If we agree to this change, then contradictory
statements do need to be corrected.
  ?If a Reserved Parameter must have one and only one Usage, Usage is
optional. If a Reserved Parameter must have one and only one Type, Type is
optional.?

  The order of the vlues of the leaves shall be in the order of increasing
tap number.

  I believe that the impulse values are in Volts.

  ?element[n,0]? is correct.


Fangyi,. 1/11

  Can you clarify the functionality of the new parameter
Rx_Clock_Recovery_Mean? My understanding is that


  If its Usage is In, signal is sampled at 0.5*bit_time after a clock time
as stated on page 29 in the BIRD.

  If its Usage is Info or Out, signal is sampled at (0.5*bit_time +
Rx_Clock_Recovery_Mean) after a clock time.


  Response,
  If Rx_Clock_Recovery_Mean is Info or Out, it is used by the EDA tool to
predict BIT Error rate in statistical analysis, or in time domain analysis
if Rx GetWave does not return clock ticks.
  If Rx_Clock_Recovery_Mean is In, then is it is used by Rx GetWave to
generate clock ticks.

Kumar, 1/11

  page 4:

  environment variables : Does the environment variable names need to be
standard?
                          only one standard example name is provided:
"AMISearchPath"


  page 11:

  Init_Returns_Filter
  in WMK comment:
  (WMK
    what is the combination of impulse response and filter; is it same as
modified impulse response?
    How does this combination prevents deconvolution/other techniques to
derive the filter response?)

  page 14

  Rx_Receiver_Sensitivity:
    How the EDA/model is expected to use this information in a standard way
during simulation/post processing?

  page 18

  tap input examples
      (mySampleAMI (txtaps .05 -.1 .7 .1 -.05) (strength 6))


  tap output through parameter_out
      (mySampleAMI (txtaps .06 -.05 .8 .05 -.04) (framis "Overloaded"))


   there are two additional parameters here, 'strength' and 'framis'

    does that mean parameter_in and parameter_out need not be consistent?

  page 22:

   null **AMI_memory_handle

    it is proposed that with null memory handle, the model still can return
**msg and **AMI_parameters_out.
    The present standard says this memory is owned by the model. Is that
still correct?

  Response,

  Environment variable do not need to be standard, only AMISearchPath will
be standard.

  Init_Returns_Filter/Convolution (This was discussed in the new flow
presentations)

  Rx_Receiver_Sensitiviy (as an Info, In, or Out):  The EDA tool will
declare the value of a bit as 1 if the waveform value is >=
Rx_Receiver_Sensitivity or 0 if the waveform value is
<= -Rx_Receiver_Sensitivity. The EDA tool will declare the value unknown
otherwise. The DLL can use the value of Rx_Receiver_Sensitiviy in any way it
chooses to.

  parameter_in and parameter_out need not be consistent!

    it is proposed that with null memory handle, the model still can return
**msg and **AMI_parameters_out.
    The present standard says this memory is owned by the model. Is that
still correct? Yes.

Ambrish, 1/11

  With regards to the parameter string going to the model, the EDA tool will
also have to strip the reserved parameters from the string as well. We would
need to put in language that signifies/clarifies this as well, else model
makers will start accounting for all (reserved and model specific) the
parameters in the ami files in their models.

  Response, I added the following



  | The string passed into the model only contains Usage In, and InOut
parameters. The string passed

  | back from the model only contains Usage InOut and Out parameters.
Bob, 1/11
  "sub-parameter" has different meaning in AMI then in IBIS, would like to
use a different term (WMK how about qualifying-parameter?)

  String always defined with quotes (This change already aggreed to)

  Does not want to introduce NA for undefined value.

  Would like to require Usage and Type for all parameters.

  AMI Version, (required?)
[ibis-macro] Response to AMI BIRD comments

Other related posts:
