Re: Question about new Warning from ibischk2+

Kai,

Here are answers to your questions...

> I understand the concept of the warning from ibischk2.
> But now I have some questions.
>
> 1) the value in the expression
>
> > abs(.95 - .74) > 2% * (3 - .74)
>
> I think .95 means an ideal voltage of endpoint of the waveform.
> If it is true, is the right expression 2%*(3-.95) ?
>

>> If you were to plot the intersection of a 50 Ohm to 3V load line (slope
>> = 50 V/I) and the given DC VI curve you would see that they interesect
>> at .95V (this is fist value in [3] below).
 
The .95 in the example is mathematically determined from the DC VI
curves in the model and the given load. When this value differs from
the waveforms endpoints there is a discrepency in the data.

Currently the checker takes the 2% times the Vac_high - Vac_low. I
agree it might be better to trust the DC vi curves to get the range as
your suggested formula says and will suggest this change for next
release. However, practically speaking, just thinking about the values
it really doesn't appear to matter which range (AC vs. DC) is used.

> 2) Where is "2 percent" from ?

The IBIS committee agreed that 2% of the range is a reasonable tolerance
for IBIS model builders to comply with.

The spec says something to the effect "The first and last AC waveform
point should be equal to the stable DC values"

2% just allows some tolerance to "equal"

 
> 3) Is the value of "2" rightful ?

I ran tests on all .ibs models and found that about 98% of model's
passed the test within 1% and they other 2% were off by more than 5%.

You may need to lengthen the run time of the AC analysis so that the
output becomes stable. If you have too many points, you may need to
carefully (in an interlaced fashion) remove some of the points toward
the end of simulation (not the last one).

It would be a nice feature of the new visual ibis building tools to
allow visual editing of the waveforms as well as an automated
"redundant" point removal based on some linearization tolerance.

See http://www.eia.org/EIG/IBIS/ibis.htm and hit tools for more info.

Chris Rokusek
Viewlogic Systems
Received on Fri Apr 3 08:55:23 1998