[ibis-macro] Re: Backchannel and the crosstalk simulation flow
- From: "Muranyi, Arpad" <Arpad_Muranyi@xxxxxxxxxx>
- To: "ibis-macro@xxxxxxxxxxxxx" <ibis-macro@xxxxxxxxxxxxx>
- Date: Tue, 23 Aug 2011 17:08:18 +0000
Walter,
I understand that backchannel is all time domain. The
point I am trying to make is that a model which supports
backchannel should probably not even modify the IR at all
in AMI_Init. The only reason this could be useful is for
FD simulations, but since the coefficients are not set up
yet at the time the AMI_Init function is executed (since
backchannel training comes later in GetWave), I don't see
how modifying the IR in AMI_Init could give correct results.
The other part of my question is whether we should say
something in the spec about the crosstalk and backchannel
flow in our time domain flow descriptions. Does everyone
know what to do the way this flow is described today in
BIRD 120.1?
Thanks,
Arpad
============================================================
-----Original Message-----
From: Walter Katz [mailto:wkatz@xxxxxxxxxx]
Sent: Tuesday, August 23, 2011 11:41 AM
To: Muranyi, Arpad; ibis-macro@xxxxxxxxxxxxx
Subject: RE: [ibis-macro] Backchannel and the crosstalk simulation flow
Arpad,
Backchannel is time domain, which does not use the modified impulse
response from the AMI_Init function. Since Tx_GetWave is responsible for
applying the Tx equalization, then backchannel with crosstalk will just
work. This would be very complicated if Use_Init_Output was not
deprecated.
Walter
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Tuesday, August 23, 2011 12:30 PM
To: ibis-macro@xxxxxxxxxxxxx
Subject: [ibis-macro] Backchannel and the crosstalk simulation flow
Hello everyone,
I took an AR last week to start this thread with a few questions on how
the backchannel communication proposal would affect the existing AMI
crosstalk flow.
Summary of the crosstalk flow:
===============================
1) The Tx tap settings of all aggressors and the victim channel are
determined from normal through channel simulations.
2) The AMI_Init functions of the aggressor Tx models will be used to
modify the crosstalk impulse response on the victim channel.
3) The AMI_Init function of the victim Tx model will be used to modify the
through impulse response on the victim channel.
4) The Rx model will further process the modified impulse responses in its
AMI_Init and AMI_GetWave functions.
Note that in case the Tx models do not modify the impulse response in
their AMI_Init function, or their AMI_GetWave function have a glorified
version of the algorithm in their AMI_Init function, the EDA tool will
have to use the GetWave functions of the Tx models on the time domain
waveform to get the correct crosstalk response.
Summary of the Backchannel flow:
================================
The communication between the Tx and Rx GetWave functions establish the
settings of the various coefficients for the Tx model. Note that this is
done in the GetWave functions, and the AMI_Init function will not be
executed after the backchannel communication finished.
Questions:
==========
1) Backchannel simulations render the AMI Init function unusable for
modifying the impulse response(s) of the channel.
Should we have a rule that in this case the model shall not return a
modified impulse response in the AMI_Init function?
2) The crosstalk flow relies on the AMI_Init function's modification of
the crosstalk (and through) impulse responses.
The outcome of the backchannel communication is not known at this time.
How are we going to incorporate the results of the training in the
crosstalk simulations? Again, should we have a rule that in this case the
model shall not return a modified impulse response in the AMI_Init
function?
3) In case the aggressor's AMI_Init function does not modify the crosstalk
impulse responses, the EDA tool needs to use the time domain waveforms and
use the time domain flow to figure out the crosstalk effects.
Our spec and BIRDs do not describe the time domain flow for crosstalk
simulations. BIRD 130 (Crosstalk BIRD) was written to clarify the rule
about the Tx model modifying the crosstalk impulse responses. We did not
write anything on how crosstalk is done in the time domain. Adding the
backchannel communication to the flow, I wonder if we should address these
in a more detailed time domain flow. Is our current time domain flow
description adequate for any EDA vendor and model maker to know how to
implement these features in the tools and models?
There may be more questions, but these might get some discussions going...
Thanks,
Arpad
=====================================================================
---------------------------------------------------------------------
IBIS Macro website : http://www.eda.org/pub/ibis/macromodel_wip/
IBIS Macro reflector: //www.freelists.org/list/ibis-macro
To unsubscribe send an email:
To: ibis-macro-request@xxxxxxxxxxxxx
Subject: unsubscribe
---------------------------------------------------------------------
IBIS Macro website : http://www.eda.org/pub/ibis/macromodel_wip/
IBIS Macro reflector: //www.freelists.org/list/ibis-macro
To unsubscribe send an email:
To: ibis-macro-request@xxxxxxxxxxxxx
Subject: unsubscribe
Other related posts:
