Arpad, Ambrish,
Do you really think that a PAM4 model written by someone who cannot determine
what the thresholds are is capable of writing a PAM4 model that can be trusted?
How did this model maker optimize the equalization without being able to
determine the thresholds? How does this model maker handle non-linearities
which are critical to PAM4 operation.
IEEE has documented a method of determining PAM4 eye metrics for compliance
tests based on a simple analysis of a PAM4 eye histogram generated from either
an impulse response at the latch in statistical mode or a waveform at the latch
in time domain. (See the CEI_56G_PAM4 standard).
I implemented this function in MATLAB for SerDes Toolbox users. Every AMI PAM4
model development tool should offer that function to their customers if their
customers cannot do it for themselves.
I think any model that cannot determine the PAM4 thresholds in either
statistical or time domain should be rejected for the reasons above. This
simple test is part of our model certification suite.
Walter
Walter Katz
Work 508.647-7633
Cell 720.417-3762
[Description: Description: Visit MathWorks.com]
From: ibis-macro-bounce@xxxxxxxxxxxxx <ibis-macro-bounce@xxxxxxxxxxxxx> On
Behalf Of Muranyi, Arpad
Sent: Monday, October 26, 2020 4:28 PM
To: 'IBIS-ATM' <ibis-macro@xxxxxxxxxxxxx>
Subject: [ibis-macro] Re: PAM4 reserved parameter issue/question
Hello,
I tend to agree with Ambrish. We had cases when the model returned the *wrong*
values, maybe
because they are required and the model maker didn’t know what to return. The
model should only
return a value if it is correct, and shouldn’t be required to return something
even if it has nothing to
return or doesn’t know what to return. Would returning “NA” be a solution for
that situation?
Thanks,
Arpad
==============================================================================
From: ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Ambrish Varma
Sent: Sunday, October 25, 2020 9:38 PM
To: 'IBIS-ATM' <ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx>>
Subject: [ibis-macro] PAM4 reserved parameter issue/question
Hi All,
We have an issue in the spec about the PAM4 reserved parameters where it
mentions that
“If the AMI Reserved Parameter Modulation lists “PAM4” (either as a Value or as
a List selection),
PAM4_UpperThreshold and PAM4_LowerThreshold are required for Rx AMI parameter
definition files.”
If the AMI model want to return those values – it should be allowed to – but I
don’t understand why it should be required to.
The EDA tool has the final analyzed waveform and can know the
PAM4_UpperThreshold and PAM4_LowerThreshold values.
This requirement is an unnecessary burden on the model maker. I am wondering if
that was our intention..?
Thanks,
Ambrish.