Re: Spice Model of IBIS

From: Fred Balistreri <fred@apsimtech.com>
Date: Mon Mar 30 1998 - 16:08:49 PST

Kellee Crisafulli wrote:
>
> Hi John,
>
> You asked basically why IBIS was created when SPICE already existed and why
> can't it be used in SPICE.
>
> Really you should think of IBIS as somewhere between a true model and
> specification.
> It is really more a specification for what the parts do then a model.
> Think of it
> like a data book describing the characteristics of the part. For example
> there are
> V/I curves just like in a data book.
>
> IBIS can be used with SPICE. Many of the SPICE tool vendors have a
> SPICE sub circuit for running the IBIS specification.
> Just ask your SPICE vendor for a sub-circuit model. If your vendor
> doesn't have one, find a better simulator package.
>
> To answer your questions about IBIS it was specifically developed to
> fix the major problems not addressed by spice for doing IC signal integrity
> simualtion. A large number of SPICE users were involved in the development
> of IBIS to resolve problems that SPICE alone could not address i.e.
>
> 1) SPICE models were not portable even between spice tools since IC designers
> are very often using special transistor models not available in most spice
> packages used by systems engineers.
> 2) SPICE models gave away all the design tricks the IC companies wanted
> protected.
> This meant they would not distribute models without non-disclosure
> agreements
> which must be signed by a board member or V.P. or President level to be
> valid.
> 3) SPICE has no rules so there is no minimum requirements to be a spice model.
> IBIS models provides sufficient detail about the component
> characteristics to
> allow signal integrity, EMC analysis and high speed design work. Since
> SPICE
> models don't have any agreed upon standards there is no way to tell if the
> information needed will be present and where to find it.
> 4) There is no way to embed testing specifications in SPICE. IBIS contains
> testing
> constraints as well as device characteristics.
> 5) SPICE has no way to indicate best/worst case limits the manufacturer
> wants you
> to design to, IBIS does.
> 6) IBIS has been extended to include resistor and capacitor packages as
> well as
> diode packages. This can be done in SPICE, but there are again no rules
> about
> what information is present and how it is organized. This works OK
> (though a
> bit of extra work in a schematic level simulator) however it does not work
> very well if the model must be used at the board level and automatically
> loaded.
> 7) IBIS is also being extended to do connector models. This is an area
> than can
> be done in SPICE, but lack of standards in SPICE has prevented any standard
> method for modeling connectors from being developed. IBIS will provide a
> set
> of rules that everyone can follow to insure portable models that will
> automatically
> load without user intervention.
> 8) There isn't any standards organization supporting SPICE format changes.
> IBIS
> is an EIA specification and is a standard that everyone can use and
> design to.
>
> At 03:50 PM 3/28/98 -0600, John Synesiou wrote:
> >I would like to know if there is any reason why an IBIS model of a gate
> >cannot be modeled using a SPICE simulator, specifically a SPICE 3 compatible
> >simulator. Since IBIS, refers to the I/O characteristics of the gate, why
> >not use SPICE to model these characteristics? The package parasitics are
> >easy to model in SPICE, and the V/I tables can be modeled using if-then-else
> >statements.
>
> >I wonder why a new syntax was developed for IBIS, rather than using SPICE
> >syntax.
> >
> >I would like to think that there was some very compelling reason why a new
> >syntax was necessary and why it is necessary for me to purchase yet another
> >simulator. I see SPICE to IBIS converters are available, what about an IBIS
> >to SPICE converter?
> >
> >Regards
> >
> >John Synesiou jsynesio@us-power.com
> >U.S. Power, Inc Phone (612)826-1111
> >6497 City West Parkway Fax (612)826-1003
> >Eden Prairie Date: 03/28/98
> >MN, 55344 Time: 2:24 PM
> >
> >
> >
> >
> >
My company supports both IBIS and Spice models. There are a variety of
mistakes and misconceptions in what Kellee says about spice models.

Statements 1 and 2 are true. In essence this is the whole reason for
IBIS.

Statement 3 is misleading. Spice is a simulator not a syntax or
standard.
The rules need to be developed by the people making the model not the
simulator. Spice models levels have been around for years and the
various
levels are clear about what is modeled and what is not. Berekey
documents
the models very well and they are public. IBIS as it stands has not been
approved for EMC and it's untested. Since there is no information about
the
power supply and gnd currents and their waveforms its doubtful that IBIS
could be used in EMC at all. Repeated attempts to include this in the
spec
by Cadence and other IBIS knowledgeble people have been down played by
other
vendors who would be at a disadvantage technically by including the
current
specifications. Needless to say that Spice models do not suffer this
fate.

4. I don't have a clue what Kellee is talking about. Is the embeded test
the
   load resistance to test the model under? If so this has been a sore
spot
   in IBIS for a long time. John Powel can tell you why Quad requires 2
sets
   of TR/TF data in their simulator.

5. This is bogus since IC desigers have been doing this for at least 30
years
   using Spice. Most if not all currenty avaliable IBIS models are made
from
   Spice in the first place. Now how is that possbible if Spice doesn't
or
   can't indicate best/worst case limits?

6. Not only can Spice to this but it can do this elequently, both at the
   schematic level and board level. Kellee you should learn more about
using
   Spice subckts. It works wonderfully well at the board level thankyou,
we've
   been doing this since 1992.

7. Spice connector models can be more robust than IBIS models. IBIS
cannot
   handle a matrix type solution since the committee was
unwilling/unable to
   come up with a coupling standard everybody would agree to. The result
is
   pin based parasitics which are inaccurate and practically useless in
some
   applications. It's true though that there is not set standard in
Spice.
   Using lumped LCR networks will work in any Spice package though.

8. True. There is an upcomming EIA proposal from Japan that would
formally
   standardize Spice models without giving out the propietary
information.
   If it doesn't get killed by IBIS bigets it would be a good standard.

Now there is good reasons for having IBIS models. As mentioned
statements 1
and 2 are true. And there is some truth in most of the statements Kellee
made.
However Spice and Spice models should not be compared to IBIS data
sheets.
They are NOT the same thing. There are applications where IBIS has
advantages.
Those advantages are in simulation speed and simplicity. There are areas
where
IBIS falls short. Those areas are in flexibility, modeling, and
accuracy.
Syntax is the strength and weakness of IBIS. Since it takes more than
one
year to change the syntax/parser and everything that means it becomes a
weakness. IBIS is an emerging standard but it has not withstood the test
of
time. On the other hand I don't know of a main stream EDA or IC company
that
does not use SPICE. That's something to think about. How about it Bob
Ross
and company when are we going to get a SPICE standard?

I'm not the most popular guy around and have no plans to be in the
future.
But somebody should stand up and tell the truth in this industry once
and
awhile.

The opinions stated above are my personnal ones and do not reflect the
views
of Applied Simulation Technology.
 
   
Best regards to all!
Received on Mon Mar 30 17:05:55 1998

This archive was generated by hypermail 2.1.8 : Fri Jun 03 2011 - 09:53:46 PDT