[ibis-macro] Re: Three new flow diagrams have been posted
- From: "Walter Katz" <wkatz@xxxxxxxxxx>
- To: <fangyi_rao@xxxxxxxxxxx>, <kumarchi@xxxxxxxxx>, <ibis-macro@xxxxxxxxxxxxx>, <Arpad_Muranyi@xxxxxxxxxx>
- Date: Mon, 12 Oct 2009 06:00:28 -0400
Fangyi,
It depends on how the Rx model is written. The Rx Init function can either
ignore the input impulse response, use it to initialize tap coefficients
that are refined in the Rx GetWave call, or determine the Rx tap
coefficients. If it ignores its impulse response input then the performance
of the Rx model is not affected. If it uses its input impulse response to
initialize the Rx GetWave processing then it may take additional hundreds of
thousands of bits for the Rx GetWave function to converge thus affecting
simulation time performance. If it uses its input impulse response to
determine the Rx tap coefficients, it will choose Rx tap coefficients that
ignore the Tx equalization, and therefore affect the electrical performance
of the Rx model.
Walter
Walter Katz
303.449-2308
Mobile 720.333-1107
wkatz@xxxxxxxxxx
www.sisoft.com
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx]On Behalf Of fangyi_rao@xxxxxxxxxxx
Sent: Sunday, October 11, 2009 7:23 PM
To: wkatz@xxxxxxxxxx; kumarchi@xxxxxxxxx; ibis-macro@xxxxxxxxxxxxx;
Arpad_Muranyi@xxxxxxxxxx
Subject: [ibis-macro] Re: Three new flow diagrams have been posted
Hi, Walter;
I don't understand why you said
Anyone who writes a Tx model that has a GetWave and does not return a
modified impulse response can significantly degrade the behavior of the Rx
model.
Tx GetWave + original channel impulse + Rx GetWave pretty much cover all
functionalities of the system. Where is the degradation from?
Thanks,
Fangyi
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Walter Katz
Sent: Saturday, October 10, 2009 4:22 PM
To: kumarchi@xxxxxxxxx; ibis-macro@xxxxxxxxxxxxx; Arpad_Muranyi@xxxxxxxxxx
Subject: [ibis-macro] Re: Three new flow diagrams have been posted
All,
Responding to Kumar's comments:
1. Agreed, as shown in the Flows_3W, everything works. It simply
requires doing a deconvolution for two of the flows. I suggested adding the
option for the AMI_Init to return the Buffer-Equalization so that
deconvolution would not be required.
2. Two different types of data in the same vector is not confusing
if you document it properly. We already have three different types of data
in the same array of vectors. Three comes from the Channel, the Modified
Channel, and Crosstalk.
3. There is no need for Init to return an impulse response when
GetWave exist, but is highly desirable. Consider the task of the person
writing the Rx model. Is the input to his AMI_Init call going to contain the
Tx equalization or not? Anyone who writes a Tx model that has a GetWave and
does not return a modified impulse response can significantly degrade the
behavior of the Rx model.
Walter
Walter Katz
303.449-2308
Mobile 720.333-1107
wkatz@xxxxxxxxxx
www.sisoft.com
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx]On Behalf Of C. Kumar
Sent: Saturday, October 10, 2009 6:43 PM
To: ibis-macro@xxxxxxxxxxxxx; Arpad_Muranyi@xxxxxxxxxx
Subject: [ibis-macro] Re: Three new flow diagrams have been posted
hello:
some flow diagram comments
1. I am not convinced about the need to return the buffer equalization
rather than the modified impulse response. At this point i do not see why we
have to break the established flows with modified impulse response
2.impulse response and buffer equalization is two different types of data
and using the same vector in ami init both of them can be confusing.
3. I also do not see the need for a model to return modified impulse
response when it has a getwave function. it should be free not to return
anything after init.
--- On Fri, 10/9/09, Muranyi, Arpad <Arpad_Muranyi@xxxxxxxxxx> wrote:
From: Muranyi, Arpad <Arpad_Muranyi@xxxxxxxxxx>
Subject: [ibis-macro] Re: Three new flow diagrams have been posted
To: ibis-macro@xxxxxxxxxxxxx
Date: Friday, October 9, 2009, 7:37 PM
Fangyi,
I think you are correct with your observations.
The main difference between the last two flows is that
mine will let the tool do the convolutions at all times,
even for the Init functions. This way there is no need
for the Boolean switches, and there is no need for that
deconvolution Walter has.
Other than that, I believe the two flows are pretty much
equivalent.
Arpad
=========================================================
-----Original Message-----
From: fangyi_rao@xxxxxxxxxxx [mailto:fangyi_rao@xxxxxxxxxxx]
Sent: Friday, October 09, 2009 6:14 PM
To: Muranyi, Arpad; ibis-macro@xxxxxxxxxxxxx
Subject: RE: [ibis-macro] Three new flow diagrams have been posted
Hi, Arpad;
Thanks for your efforts to clean up the flows. My understanding of them
is Flows_3A and Flows_3W are basically the same. In order to handle four
permutations of Tx/Rx GetWave_Exists and to support stateye, the
simulation must know different combinations of h_AC, h_TEI and h_REI
from Tx and Rx Init calls. In your flow (3A) Init simply returns h_TEI
or h_REI. In Walter's flow (3W) the simulator figures out h_REI by
deconvolution. This is the only difference between 3A and 3W. If that's
the case I'd favor 3A because it's cleaner.
Regards,
Fangyi
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Friday, October 09, 2009 2:46 PM
To: ibis-macro@xxxxxxxxxxxxx
Subject: [ibis-macro] Three new flow diagrams have been posted
Hello Everyone!
I would like to inform all of you that we just posted
three new flow diagram PDF files to the IBIS-ATM web
site.
http://www.vhdl.org/pub/ibis/macromodel_wip/archive/20091009/arpadmurany
i/AMI%20flows:%202009%20Sept%2029%20proposal%20-%20fixed/AMI_Flows_2fixe
d.pdf
AMI_Flows_2fixed.pdf is basically the same flow we discussed
in the October 6th, 2009 teleconference but I added the
correction that was requested based on Ambrish's question.
I also added some notes to the bottom of the last two
slides.
http://www.vhdl.org/pub/ibis/macromodel_wip/archive/20091009/arpadmurany
i/AMI%20flows:%20Walters%202009%20October%206%20proposal/AMI_Flows_3W.pd
f
http://www.vhdl.org/pub/ibis/macromodel_wip/archive/20091009/arpadmurany
i/AMI%20flows:%20Arpads%202009%20October%206%20proposal/AMI_Flows_3A.pdf
AMI_Flows_3W.pdf and AMI_Flows_3A.pdf are new flows
which came out of the discussions in the October 6th
ATM teleconference mostly based on Danil's comments to
try simplifying things. The file with 3W is Walter's
way of implementing the simplifications. The file with
3A is my way of simplifying things.
Please study them both and be ready for comments in our
next IBIS-ATM teleconference.
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:
