RE: FW: [SI-LIST] : IBIS datasheets for PCI and DDR

About this list Date view Thread view Subject view Author view

From: Roy Leventhal (Roy_Leventhal@mw.3com.com)
Date: Thu Nov 18 1999 - 16:01:55 PST


My opinion:

IBIS and SPICE both have their place.

As I understand it, IBIS runs faster on signal integrity because it is
behavioral and is (will?) be supposedly easier to get because of the relative
ease of reverse engineering SPICE. Also, under limited conditions and direct
physical measurement it might turn out to be more accurate than SPICE because it
assumes nothing about the internal workings of a device and merely records its
behavior at the port of interest - input or output.

SPICE is more extensible over a range of bias, environment, etc., precisely
because of its interest in modeling the internal behavior of a part. And, as
regards input-to-output behavior, say as an amplifier, IBIS was never intended
to do that. Regarding really high frequency work - I would go to Scattering
Parameter tables - which when you think about it is behavioral.

It depends on what you want to do. I could go on. It has all been said before.

Regarding shoddy models. The problem has nothing to do with model choice and
everything to do with the character of the model makers and their company. When
I open up an IBIS model (simulated from SPICE) with a power clamp current of
1E+23, well, that's just a mistake and a propagation of a mistake and a
too-simplistic model of diode resistance. When I open up a several hundred pin
device with 2 to 5 pins in its pin list - sheer arrogance and laziness. No
excuse for it.

Whatever. I think it would be far more productive for IBIS and SPICE users to
start a discussion of who consistently produces trustworthy models and who
doesn't.

Lest we forget, SPICE is the name of a computer program. The program is based on
Gummel-Poon, Ebers-Moll, Schickman-Hodges, etc., and those models were improved,
simplified, expanded in different directions as devices changed, and were based
on earlier work that corrected previous mistakes over a period of many years.

The first BJT model was an input resistance and a dependent output generator.
Then along came narrow based devices, saturation characteristics, (Dr. Jim)
Early effect and 2-port models. These new models, empirically observed and
generated, prompted people to formulate more comprehensive theories and models
of what was going on inside a device. Beyond the first exploration of theory and
some serendipitous guesses measurement preceded theory and not the other way
around.

But, before we advance ideas about accuracy and inherent model goodness let's
understand:

The quality of the model source/modeling engineer.
The class of problems the model is targeted at.
And, that all models have their limitations - particularly when you try to push
them into new territory.

Regards,

Roy

**** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 11:39:00 PST