Re: new ibischk3 V3.2.6

Kim,

The problem may be due to ideal diodes. Or it may be due
to a voltage sweep which is beyond the operational limits of
the device, where the device and the spice model breakdown.
In essence, a spice to Ibis extraction can easily drive the
transistor models of the device beyond their assumptions that
the models are based upon.

At SiQual, when we create models from Spice, we make sure
to limit these overcurrents in the final model by truncating the
curves at a reasonable point, well beyond the normal operating
range of the device.

Then, to be sure, we correlate the ibis simulation results obtained
with the model to spice. (In a "real" reflected wave transmission line
circuit.)

In the future, we will begin "closing the loop" by performing
correlations of real devices in test boards.

regards,

scott

--
Scott McMorrow
Principal Engineer
SiQual, Signal Quality Engineering
18735 SW Boones Ferry Road
Tualatin, OR  97062-3090
(503) 885-1231
http://www.siqual.com
Kim Helliwell wrote:
> I have another perspective. This problem is almost certainly caused
> by diode models in the SPICE netlist that are "ideal" models; that
> is, there is no series parasitic resistance (RS) specified.
>
> And I think it's more than reasonable for the IBIS community to demand
> that suppliers of SPICE or IBIS models do not use such ideal components
> in the model.  This is a well-known SPICE trouble spot, and anyone who
> still perpetrates that sin appears to me to be a rank amateur. It causes
> me to wonder what other inaccuracies exist in the models that exhibit this
> problem.
>
> Kim
>
> "Dunbar, Tony" wrote:
> >
> > Aubrey,
> >
> > First of all, let me just set the scene. In this e-mail, I am ONLY referring
> > to the situation of gross currents in the V-I tables.
> >
> > >From a purist stand-point, I agree with you. Unfortunately, in my experience
> > the reality is that many, many IBIS models derived from SPICE exhibit these
> > giga-amp characteristics. I think what Bob means is that the model is
> > correct in that it reflects what the SPICE model has. A further reality is
> > that the IBIS forum is not going to change the world; these decks and models
> > are not going to change to satisfy this anomoly. Fortunately, they usually
> > occur well away from the normal operating region and normal clamping region
> > so, in actual operation, they don't give us a problem.
> >
> > Given that this is reality and it's not going to change, I think we (the
> > IBIS forum) need to look at what, if anything, we are going to change to
> > deal with it? Maybe we need to change things a little to be closer to this
> > normal operation. One question is, 'what is the reasoning behind the
> > (somewhat large) range of (2xVCC to -1xVCC) for the V-I tables?'; can this
> > be truncated? Or, better (IMHO), check that the clamp currents are
> > reasonable(?) within a tighter range; i.e. closer to the normal operating
> > region and to a limit more aligned with an expected clamping event; e.g.
> > VCC+1.0V and GND-1.0V.
> >
> > Yes, it sounds like capitulation, but I think it's the only practical
> > course.
> >
> > -----Original Message-----
> > From: Aubrey_Sparkman@Dell.com [mailto:Aubrey_Sparkman@Dell.com]
> > Sent: Wednesday, January 10, 2001 11:01 AM
> > To: bob_ross@mentorg.com; shuq@cisco.com
> > Cc: ibis-users@eda.org
> > Subject: RE: new ibischk3 V3.2.6
> >
> > Bob,
> > I'm not sure I agree with your statement that a model with end point I-V
> > currents that are "extremely large (such as 1e20)" "might actually be
> > correct" even if those data points are produced from a valid spice deck.
> > The purpose of a model is to reflect reality where possible and 1e20
> > amps????
> >
> > Aubrey Sparkman
> > Signal Integrity
> > Aubrey_Sparkman@Dell.com
> > (512) 723-3592
> >
> > > -----Original Message-----
> > > From: Bob Ross [mailto:bob_ross@mentorg.com]
> > > Sent: Tuesday, January 09, 2001 7:26 PM
> > > To: Syed Huq
> > > Cc: ibis-users@eda.org
> > > Subject: Re: new ibischk3 V3.2.6
> > >
> > >
> > > Syed:
> > >
> > > The new ibischk3 changed the Warning message to an Error
> > > message when the mismatch exceeded 10%.  In your example,
> > > the mismatch between 0.41 and -0.71 exceeds the 10% value
> > > of the range (.21v).  There may exist a real problem that
> > > needs to be examined.  This change is documented as BUG47:
> > >
> > >   http://www.eda.org/pub/ibis/bugs/ibischk/bug47
> > >
> > > However, the -0.71 value is suspicious.  I have seen a
> > > similar problem when some of the end point I-V currents are
> > > extremely large (such as 1e20) and cause ibischk3 to
> > > not properly converge to the correct DC endpoints.  You
> > > might check this and try smaller values if such large
> > > values exist.  Your model might actually be correct.
> > >
> > > Bob Ross
> > > Mentor Graphics
> > >
> > >
> > > Syed Huq wrote:
> > > >
> > > > I ran a model with the NEW ibischk3 ver3.2.6 and get this:
> > > >
> > > > new version:
> > > > ERROR - Model XYZ_IO: The [Rising Waveform]
> > > >       with [R_fixture]=50 Ohms and [V_fixture]=2.5V
> > > >       has TYP column DC endpoints of  0.41V and  2.50v, but
> > > >       an equivalent load applied to the model's I-V tables yields
> > > >       different voltages (-0.70V and  2.50V),
> > > >
> > > > In the earlier version(V3.2.5),this would show up as a
> > > WARNING. Since now
> > > > it shows up as ERROR, the file fails.
> > > >
> > > > old version:
> > > > WARNING - Model 'XYZ_IO': TYP AC Rising Endpoints ( 0.41V,
> > > 2.50V) not within
> > > >           0.042V (2%) of (-0.70V,  2.50V) on VI curves for
> > > 50 Ohms to 2.5V
> > > >
> > > > Why was this changed to ERROR ?
> > > >
> > > > Syed
> > >
>
> --
> Kim Helliwell
> Senior CAE Engineer
> Acuson Corporation
> Phone: 650 694 5030  FAX: 650 943 7260