RE: ibischk3 V3.2.8 [Warning]


Subject: RE: ibischk3 V3.2.8 [Warning]
From: Muranyi, Arpad (arpad.muranyi@intel.com)
Date: Mon Feb 04 2002 - 13:11:40 PST


Andy,

The IBIS rules for the clamp curves are not saying that you
cannot provide more data for your clamps (than what the rules
say). This has been discussed a long time ago when I first
saw the need for the POWER clamp curve needing to cover a
wider range to include the currents of an on die pullup
device that is used as an always on terminator.

At that time the consensus in the Open Forum Committee was
that we can make models covering -Vcc to 2Vcc in the clamp
curves also, the spec doesn't prohibit it, it only says
that the required range is what it is. So you should be
able to make models for what you describe in your posting
without any difficulty.

Arpad Muranyi
Intel Corporation
============================================================

-----Original Message-----
From: Ingraham, Andrew [mailto:Andrew.Ingraham@compaq.com]
Sent: Monday, February 04, 2002 12:55 PM
To: ibis-users@server.eda.org; ibischk-bug@server.eda.org
Subject: Re: ibischk3 V3.2.8 [Warning]

> The I-V tables are supposed to include data for all of the
> "interesting"
> voltages of the buffer.
 
I wish this were so.

I really do wish that IBIS had required ALL I-V tables to completely
cover all normal voltages of the buffer.

What is the current from the [POWER Clamp] table when the buffer voltage
is 0V? Or +1.0V, or -1.0V? It doesn't say. The [POWER Clamp] table
doesn't cover these voltages. It apparently assumes that the current
due to the POWER Clamp should be negligible there and doesn't need to be
tabulated.

But what if it isn't?

Or what if ... for whatever reason ... the values in the [POWER Clamp]
table do not asymptotically approach zero as the buffer voltage
approaches POWER?

There are no IBIS rules for extrapolation. Therefore, the results are
entirely unpredictable. Ugh!

As noted, one simulator might continue the last current value
indefinitely, another might extrapolate with a constant slope. Yet
another might assume ZERO current beyond the defined I-V tables, causing
a rather large discontinuity.

Sure, the IBIS spec claims that "this data must not be left for the
simulator to provide," but that is, unfortunately, a false statement.



This archive was generated by hypermail 2b28 : Mon Feb 04 2002 - 13:39:44 PST