• Date: Wed, 26 May 2010 16:36:52 -0600

```Walter,

BER can be found as a conditional integration of eye density. Assuming
that waveform pieces that correspond to sent '1' are red and others
('0') are blue, for any time/voltage sample point location, ber can be
estimated as a number of red waveforms below this point divided on the
total number of red waveforms plus the number of blue waveforms above
this point divided on the total number of blue waveforms. From here it
follows that BER is zero in the middle of an open eye but approaches to
0.5 if you go too far in any direction. We can imagine a small contour
at very large or very small voltage, such that nothing from the eye gets
into that contour. However, we cannot say that sampling at +1000V
threshold is BER-free.

-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Walter Katz
Sent: Wednesday, May 26, 2010 4:05 PM

First, you have to think of the eye as a series of eye contours of
probability that the eye will be inside that contour. Where the two
horizontal lines you suggest cross each eye probability contour is
essentially the bathtub curve.

Walter

Walter Katz

303.449-2308

Mobile 720.333-1107

wkatz@xxxxxxxxxx

www.sisoft.com

-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
Sent: Wednesday, May 26, 2010 5:22 PM
To: ibis-macro@xxxxxxxxxxxxx

Thanks Mike and Walter for the explanation.  This

reminds me to another question I asked a while ago

about eye templates.  It seems that this sensitivity

parameter could be interpreted as two horizontal

lines across the eye diagram serving as if they were

an eye template.  The only difference is that it

doesn't contain the timing component of the eye

template, only the voltage or amplitude portion.

Is my understanding correct?

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

________________________________

From: Mike Steinberger [mailto:msteinb@xxxxxxxxxx]
Sent: Wednesday, May 26, 2010 2:45 PM
Cc: ibis-macro@xxxxxxxxxxxxx

The Rx_Receiver_Sensitivity is needed for the EDA tool to do an accurate
job of estimating the bit error rate.

Let me describe one of my favorite experiments. It's one my former group
at Cray performed many times.
1. Take a signal directly out of a data generator and put it into a
- The BER should be indistinguishable from zero.

2. Using microwave attenuators, decrease the amplitude going into the
- Up to some attenatuation, the BER will remain zero. Then, with 1dB
increase in attenuation, the BER will go from zero to 10e-3 or so.

Anyone can perform this experiment with an eval board and a bag full of
attenuators. It doesn't take a lot of time, skill, or money, and the
same behavior will be observed for every SerDes on this planet, albeit

So, what happened in this experiment? The shape of the eye diagram
remained the same (because microwave attenuators are nearly constant
with frequency), but the amplitude decreased. At some minimum amplitude,
the receiver stopped working even though the eye diagram (for example,
as displayed on a sampling oscilloscope) is still absolutely beautiful.

An EDA tool must be able to reproduce this experiment, and in order to
do so accurately, it needs to know the Rx_Receiver_Sensitivity.

The definition we currently have for Rx_Receiver_Sensitivity is exactly
what we need and is successfully fulfilling the role it was originally
intended for.

Thanks.
Mike S.

On 05/26/2010 01:42 PM, Muranyi, Arpad wrote:

Hello AMI experts,

I know this was discussed before but it is still not
clear to me how the Rx_Receiver_Sensitivity parameter
is supposed to be used.

This parameter is Type Info, meaning that it is not
passed to the DLL, it is only information for the EDA
tool.  The question is, what is the tool supposed to
do with this?

I assume that if the DLL is written correctly, it would
account for this sensitivity parameter in its algorithms
(for example in CDR), so the EDA tool wouldn't need to
adjust the returned times for the threshold (in)sensitivity.
Is the EDA too expected to do anything with the returned
waveforms when displaying eye diagrams and/or BER plots,
etc...?  If yes, what should be done?  If no, what is
the purpose of this parameter?

Thanks,

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

---------------------------------------------------------------------
IBIS Macro website  :  http://www.eda.org/pub/ibis/macromodel_wip/
IBIS Macro reflector:  http://www.freelists.org/list/ibis-macro
To unsubscribe send an email:
To: ibis-macro-request@xxxxxxxxxxxxx
Subject: unsubscribe

```