[SI-LIST] Re: What frequency are IBIS models good to?
- From: "Muranyi, Arpad" <arpad.muranyi@xxxxxxxxx>
- To: "si-list" <si-list@xxxxxxxxxxxxx>
- Date: Wed, 15 Sep 2004 16:22:18 -0700
VS,
Excellent comments. Here are a few more...
1) Thanks.
2) Depending on who you ask, it can go well beyond 1 GHz.
I don't think there is a specific hard limit that
anyone can set. It is similar (but even less
definable in my opinion) to the question, how far=20
up can we use FR4?
3-4) Good point, and I have been talking about this for
years. The points can be fitted to an equation which
would probably be even faster than the PWL approach
plus it could be scaled instead of having three
distinct corners (typ., min., max).
The reason we started IBIS with the table format
is because we used exactly what you say, controlled=20
sources in SPICE. It was easier to use the PWL
sources with lab measured data, which is what we
did in many cases. We just got stuck with that
format in IBIS, but now you can do it through the
*-AMS extensions of the spec any whichever way you
please.
Part of the problem with using SPICE was (is) that
its equation capabilities was (is) quite limited
(depending on which flavor you have).
5-6) The original question was asking about IBIS being
nothing but DC IV curves, so I assumed that the
person asking this was talking about buffers only.
That's why I didn't bring in the package and EBD
aspects of IBIS, which I agree are quite limited.
If you need anything decent on that front fro high
frequency simulations, consider the (new) ICM
specification. It provides much more capabilities
and accuracy in many areas, including the ones you
mentioned (coupling, etc...).
It is a new spec, so there are not too many models
available yet, but just this week's HSPICE release does
have support for it, so we can now all start using it...
7) The reasons were IP, simulation speed, and the relative
ease of doing "what if" analysis by manipulating the IV
curves instead of the transistor level circuit design
and/or its process parameters. We could also add in tool
independence, and lack of SPICE models.
Sorry for my ignorance, but could you please clarify
what you meant by Linux-way, instead of Microsoft-way?
I don't see the connection there.
I am not sure if you mean the same, but if we had better
equation and behavioral modeling capability in most flavors
of SPICE (not just HSPICE), we would have probably never
started IBIS...
8) You are absolutely right. But IBIS became a standard for
different reasons. Probably mostly to have a common
format that everyone could use, independently from which
vendor's tool they use in their work.
9) That capability was there for a long time in the B-element
of HSPICE. They provide scaling coefficients for various
purposes. Unfortunately these are not features of legacy
IBIS, but now we could do it with the *-AMS extensions of
IBIS.
Thanks,
Arpad
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] =
On Behalf Of V S
Sent: Wednesday, September 15, 2004 1:24 PM
To: si-list
Subject: [SI-LIST] Re: What frequency are IBIS models good to?
Venu,
1. On a lighter tone - notice the difference between
Muranyi and Muryani.
2. The ibis is not bad even upto 1 GHz. There is good
correlation in most of the cases till that frequency.
3. I beleive that there could have been less verbose
and more accurate method than 200 lines of VI for each
of the pull up, pull down, power clamp, ground clamp
and 100 lines for each of dV/dt. I have found that
spice models created with voltage control sources,
resitors are equally good and more intuitive to
understand and work with - rather that the verbose
black box ibis model.
4. I will suggest to always try to create your own
simpler spice model after "understanding" ibis model.=20
5. The ebd model is not well supported by some vendors
including SigXplorer.
6. Crosstalk is difficult to simulate with ibis / ebd
for package crosstalk.
7. I beleive that the major reason that ibis was
developed was to protect the IP. A Linux way would
have been more desirable rather than Microsof's way.
Spice could have done better job to provide models. It
would also have provided more intuitive understanding
of the models. . =20
8. This is not to attack Arpad's and Intel's ibis. But
to provide an insight - All good SI engineer like to
play with the driver and receiver model by creating
their own intuitive model to understand drive strength
variation, edge rate variation, package variation,
package crosstalk, die and package capacitance, etc.
9. Sometimes an optimal drive strength setting is
required for a particular driver and receiver. IBIS
will provide only extreme settings of drive strength.
An intuitive model will provide better understanding
of "what if" scenario.
VS
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List FAQ wiki page is located at:
http://si-list.org/wiki/wiki.pl?Si-List_FAQ
List technical documents are available at:
http://www.si-list.org
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
