[ibis-macro] Re: Question about Rx_Receiver_Sensitivity

  • From: Mike Steinberger <msteinb@xxxxxxxxxx>
  • To: ibis-macro@xxxxxxxxxxxxx
  • Date: Thu, 14 Feb 2013 10:34:24 -0600


You pose an interesting question that we hadn't considered before.

First of all, I suggest that we do _not_ assume that the DC value of the model output is the threshold value. There are a whole lot of different ways that that assumption could get violated. For example, a PRBS7 data pattern will produce a DC offset that is not the decision threshold.

To date, we've always assumed that the received signal is differential, and so the threshold _should_ be zero. You are quite correct, however, that we never stated this assumption, and there is no absolute reason why this assumption has to be valid. This assumption does happen to be valid for most receiver designs, and any differential receiver with a nonzero threshold will be sub-optimal. However, that doesn't mean it can't happen.

It is true that there will be an on-chip variation (OCV) of the actual decision threshold, and we've generally lumped that in with the receiver sensitivity.

I suggest that we define a new Reserved parameter Rx_Receiver_Threshold, and make its default value zero. That would mean that all models published to date remain compatible and use the unstated assumption we've made to date, and new models can be more accurate when they have an offset threshold.

Thanks for the question.
Mike Steinberger

On 02/14/2013 10:04 AM, Muranyi, Arpad wrote:
Hello Everyone,

We ran into an Rx AMI model which produces an output
from its GetWave function that is not centered around
the horizontal axis.  This raises a question about
Rx_Receiver_Sensitivity which defines an offset around
a threshold line.

Is the threshold assumed to be 0 V, i.e. the horizontal
axis, or can it be anything?  If it is anything, how do
we know what it is?  I didn't find a definition for this
in the specification.  Am I missing something?

This is problematic, because if the model doesn't return
clock ticks, the EDA vendor would not know where the
threshold crossings really are.

It seems that we need to write a BIRD to define this
threshold level, or state that it is assumed to be zero.

Any comments, suggestions?


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: