RE: help with ibischk3 warnings and min data

Another point worth noting ....

If you're using a recent copy of IBIS2SIGNOISE (with SPECCTRAQuest),
the -curvedir option will cause the translator to create "before and after"
views of all the curves in a format you can look at with SigWave. That
makes it easy to determine when the translator removes or adds a point
whether the change was a significant one or not.

The -d option will keep the translator from changing the curves at all.

Todd.

-----Original Message-----
From: Kim Helliwell [mailto:khelliwe@acuson.com]
Sent: Tuesday, November 21, 2000 12:38 PM
To: Betty Luk
Cc: 'ibis-users@eda.org'
Subject: Re: help with ibischk3 warnings and min data

> Betty Luk wrote:
>
> Hi,
>
> I appreciate all the help I have been getting so far.. thanks!
>
> I have built my IBIS model using s2ibis2, typ, min and max data. The data
for my min curve seem to be a bit strange
> (when I looked at the curves using s2iplt), and the IBIS golden parser is
complaining about non-monotonic data. I am
> not sure whether I should try to fix these warnings or not, because they
are not just isolated occurences of
> non-monotonic data (ie. I cannot just discard the problematic point). I
have included my IBIS model for
> convenience. These are the warnings that I get:
>
> WARNING (line 52) - Pulldown Minimum data is non-monotonic
> WARNING (line 134) - Pullup Minimum data is non-monotonic
> WARNING (line 307) - POWER Clamp Typical data is non-monotonic
> WARNING (line 308) - POWER Clamp Minimum data is non-monotonic
> I also get another warning. What does this warning mean?
>
> WARNING - Model 'tristate_driver': MIN AC Rising Endpoints ( 0.00V,
0.89V) not within
> 0.018V (2%) of (-0.02V, 0.84V) on VI curves for 50 Ohms to 0V
>
> I would really appreciate any feedback!
>
> Thanks,
> Betty Luk
>
>
>
>
>
>
> Name: test.ibs
> test.ibs Type: unspecified type (application/octet-stream)
> Encoding: quoted-printable

1. I believe that most simulators currently available filter the IBIS table
data
   so as to remove nonmonotonicity from the data. That's not to say it's
not
   better to remove it yourself first. It might be a good idea to check the
   actual curves from the simulator you're using with s2ibis2 and see
whether
   the nonmon behavior is present in the data, or whether it's introduced by
   the interpolation routines in s2ibis2. If it is present in the actual
   simulation output, does the nonmonotoncity appear to be severe? If so,
then
   perhaps there are problems with the models you are using. If it's minor,
   and if it's outside the normal supply range, it probably won't hurt
things
   too much.

   I would say: do your best to eliminate this problem, but if it's not
possible,
   you may have to rely on the filtering in your customer's simulators.

2. The problem with your AC warning is that the start and end values of the
   V-T curves do not match the start and end values of the VI curves. You
need
   to understand why this is happening, which will involve looking at the
SPICE
   netlists and output for the VI and VT curves. The lower endpoint is
probably
   close enough to ignore, but the upper one is a bit further off. There is
a
   risk that some simulators your customers use will not tolerate this too
well
   and simply fail or (worse) quietly give the wrong answer.

In any case, any warnings you don't fix should be mentioned in comments in
the
resulting IBIS file, so your users don't panic when they see the warnings.

--
Kim Helliwell
Senior CAE Engineer
Acuson Corporation
Phone: 650 694 5030  FAX: 650 943 7260