[ibis-macro] Re: What part of Legacy IBIS is used by IBIS-AMI

  • From: "Walter Katz" <wkatz@xxxxxxxxxx>
  • To: <dan.dvorscak@xxxxxxxxx>, "'IBIS-ATM'" <ibis-macro@xxxxxxxxxxxxx>
  • Date: Tue, 17 Jul 2012 14:44:54 -0400 (EDT)

Dan,

 

The Tx model writer assumes that the impulse response input is the impulse
response of the channel including the Tx and Rx analog model. As long as
the model does not change its equalization based on the impulse response
of the channel then the EDA tool can use the AMI_Init function to
determine the Ideal impulse response of the model. Currently, one would
need to read the documentation that comes with the model to determine if
the model AMI_Init will change its equalization based on the impulse
response input to the model.

 

One could propose a new Reserved Parameter (e.g.
Disable_AMI_Init_Optimization) that when True, would force the AMI_Init
function to determine the equalization of the models without regard to the
impulse response inout to the AMI_Init function. In this case, the output
impulse response of AMI_Init will be an impulse response strictly based on
the configuration (e.g. peaking filter selection, tap coefficients) that
are input to the AMI_Init function.

 

Walter

 

From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Dan Dvorscak
Sent: Tuesday, July 17, 2012 2:24 PM
To: IBIS-ATM
Subject: [ibis-macro] Re: What part of Legacy IBIS is used by IBIS-AMI

 

Hi Arpad,

 

OK, I may have been asking something in a roundabout manner.  I think what
I'm really trying to determine if there is ever a reason for the EDA tool
to ignore the IBIS analog model in order to create an "idea" impulse
response.  Yes, the EDA tools are free to do so if needed, but when do we
need to do so? 

 

Or if the vendor would ever want the EDA tool to ignore parts of the
analog model, should there be some kind of flag for that? 

 

Thanks,

 

-Dan

 

On Tue, Jul 17, 2012 at 1:39 PM, Muranyi, Arpad <Arpad_Muranyi@xxxxxxxxxx>
wrote:

Dan,

 

There are several issues that need to be considered in this topic.

 

The biggest issue in my mind is the question of time step.  An

ideal step or impulse function has zero width.  In time step based

simulators the best we can do for a step function is a slope that

is one time step wide, and for an impulse function two slopes

(one up and one down) that are a total of two steps wide.  No

matter how small you make your time step, these functions are

never going to be what they should be.  This introduces errors

and the amount of error will depend on the time step, which may

depend on the EDA tool's internal algorithms.

 

Regarding:  "Another way to look at it would be to say that if a model
writer intended for a model to work with an ideal impulse response, they
could essentially ship and IBIS file that pretty much only contains a
c-comp, "zeroed out" V-I tables, and sharp edge rate in the ramp data to
make it a fairly ideal driver."

 

This is basically a suggestion to implement an ideal step or impulse

function generator in the IBIS [Model].  But where is the buffer's

model then?  You would either have to put the buffer's impedance

model in series or in parallel with this function generator.  And the

C_comp would belong to the buffer's model, not the ideal function

generator.  But why would you go this route, when the EDA tool can

do the function generator in the stimulus, and the model maker can

do the model of the device in the IBIS [Model]?

 

Thanks,

 

Arpad

=======================================================================

 

 

 

 

From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Dan Dvorscak
Sent: Tuesday, July 17, 2012 11:49 AM


To: IBIS-ATM
Subject: [ibis-macro] Re: What part of Legacy IBIS is used by IBIS-AMI

 

Hi all,

 

Sorry to reply late to this thread.  I had been keeping an eye out for it,
but seem to have missed it. 

 

Per what Greg and Arpad were discussing, should it not be up to the model
writer to decide whether or not an ideal or an IBIS-generated impulse
response is to be used?  My expectation would be that a given IBIS buffer
that points to an AMI model will contain all of the IBIS analog model
features that the model writer intends to be included in the channel
characterization.  In that sense if a user or EDA tool chooses to ignore
any of those features they are using the model in a way that it was not
characterized for.  

 

Another way to look at it would be to say that if a model writer intended
for a model to work with an ideal impulse response, they could essentially
ship and IBIS file that pretty much only contains a c-comp, "zeroed out"
V-I tables, and sharp edge rate in the ramp data to make it a fairly ideal
driver. 

 

Does that seem like a fair assessment?

 

Thanks,

 

-Dan

 

 

On Wed, Jul 11, 2012 at 10:37 AM, Muranyi, Arpad
<Arpad_Muranyi@xxxxxxxxxx> wrote:

Greg,

 

The IBIS specification is intentionally silent about

how to generate the impulse response for the channel,

because there are different ways to do it, and EDA

vendors might want to use their own "smarts" to do it

the "best" way.

 

You are correct, if someone wants to extract such an

"ideal" IR, you would only make use of the C_comp and

the impedance provided by the I-V curve from the analog

model.

 

BIRD 120 doesn't give any more detail (or restrictions)

on how to generate the impulse response.  So the EDA tool

is free to do it the way they think it is the best to do

it.

 

I hope this helps,

 

Arpad

===========================================================

 

From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Gregory R Edlund
Sent: Wednesday, July 11, 2012 9:15 AM
To: 'IBIS-ATM'; ibis-macro-bounce@xxxxxxxxxxxxx


Subject: [ibis-macro] Re: What part of Legacy IBIS is used by IBIS-AMI

 

What if an EDA programmer wanted to discard the VT and ramp data and
compute an "ideal mathematical" impulse response assuming a delta
function?  Is that allowed by BIRD 120?  I took a quick look and didn't
see anything specific about how to calculate the impulse response.  In the
"ideal mathematical" case, I would think all you need is C_Comp and some
resistance extracted from the IV tables.

Greg Edlund
Senior Engineer
Signal Integrity and System Timing
IBM Systems & Technology Group
3605 Hwy. 52 N  Bldg 050-3
Rochester, MN 55901



Inactive hide details for "Walter Katz" ---07/11/2012 03:53:30 AM---I
agree."Walter Katz" ---07/11/2012 03:53:30 AM---I agree.

From: "Walter Katz" <wkatz@xxxxxxxxxx>
To: <Arpad_Muranyi@xxxxxxxxxx>, "'IBIS-ATM'" <ibis-macro@xxxxxxxxxxxxx>
Date: 07/11/2012 03:53 AM
Subject: [ibis-macro] Re: What part of Legacy IBIS is used by IBIS-AMI
Sent by: ibis-macro-bounce@xxxxxxxxxxxxx

  _____  




I agree.
 
Walter
 
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Tuesday, July 10, 2012 11:44 PM
To: IBIS-ATM
Subject: [ibis-macro] Re: What part of Legacy IBIS is used by IBIS-AMI
 
That is correct.
 
Arpad
================
 
From:  <mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
ibis-macro-bounce@xxxxxxxxxxxxx
<mailto:%5Bmailto:ibis-macro-bounce@xxxxxxxxxxxxx%5D>
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Mirmak, Michael
Sent: Tuesday, July 10, 2012 6:07 PM
To: Walter Katz
Cc: IBIS-ATM
Subject: [ibis-macro] Re: What part of Legacy IBIS is used by IBIS-AMI
 
Walter,
 
Thank you!  So, someone using IBIS-AMI models with valid traditional data
in a workflow should expect changes in the IBIS TX C_comp or V-t (or
[Ramp]) data to have an impact on the results provided by the receiver in
an IBIS-AMI-based flow, assuming nothing else changes?
  

-          MM

 
From: Walter Katz  <mailto:%5Bmailto:wkatz@xxxxxxxxxx%5D>
[mailto:wkatz@xxxxxxxxxx] 
Sent: Tuesday, July 10, 2012 12:20 PM
To: Mirmak, Michael
Cc: IBIS-ATM
Subject: What part of Legacy IBIS is used by IBIS-AMI
 
MM,
 
The Legacy part of an IBIS model is used to generate the Impulse Response
of the channel. 
 
One method of generating the impulse response is to do a step response,
and take the derivative of the waveform at the Rx input.
 
A simulator uses a combination of the Tx [Voltage Range], C_Comp, IV, VT
and Ramp data, and the Rx C_Comp and IV data. Of course, [External Model]
can also be used by a simulator.
 
The following required Tx IBIS fields are not used in the AMI flow: 

Vmeas

The following required Rx IBIS fields are not used in the AMI flow: 

Vinl
Vinh

 
Walter
 
 
Walter Katz
wkatz@xxxxxxxxxx
Phone 303.449-2308
Mobile 303.335-6156
 





 

-- 
Daniel Dvorscak
ANSYS, Inc.
67 South Bedford St.
Suite 202E
Burlington, MA 01803
Tel:  781.229.8900 x352 <tel:781.229.8900%20x352> 
Fax: 781.229.8624





 

-- 
Daniel Dvorscak
ANSYS, Inc.
67 South Bedford St.
Suite 202E
Burlington, MA 01803
Tel:  781.229.8900 x352
Fax: 781.229.8624

GIF image

Other related posts:

  1. Home
  2. ibis-macro
  3. Archive
  4. 07-2012