[SI-LIST] Re: package SSN model accuracy requirements, now B ehavioral Modeling
- From: Chris Cheng <Chris.Cheng@xxxxxxxxxxxx>
- To: "'lgreen'" <lgreen22@xxxxxxxxxxxxxx>, Chris Cheng <Chris.Cheng@xxxxxxxxxxxx>
- Date: Mon, 21 Mar 2005 13:41:02 -0800
Uhh,
That's a private response to your private mail.
Since when did I ask you to forward it to the Si-list ?
-----Original Message-----
From: lgreen [mailto:lgreen22@xxxxxxxxxxxxxx]
Sent: Monday, March 21, 2005 8:55 AM
To: 'Chris Cheng'
Cc: si-list@xxxxxxxxxxxxx
Subject: RE: [SI-LIST] Re: package SSN model accuracy requirements, now
Behavioral Modeling
Hi, Chris,
With IBIS 4.1, "spice" is one of the allowed [Model] formats. Some of the
EDA tools already support HSPICE/SPICE 3f5 models through netlist export.
Please keep on kicking the vendors to provide what you (and all of us) need.
They might eventually catch on...
- Lynne
-----Original Message-----
From: Chris Cheng [mailto:Chris.Cheng@xxxxxxxxxxxx]
Sent: Sunday, March 20, 2005 6:49 PM
To: 'lgreen '
Subject: RE: [SI-LIST] Re: package SSN model accuracy requirements, now
Behavioral Modeling
Lynne,
Point taken, what I have always been advocating is start from the top.
If you hand out SPICE model, which is what I/O designers really use, you can
always abstract it down to whatever you need as an end user.
But if you start with the lowest denominator like IBIS, you can't go back up
if things go wrong (and they do, like SSO problems).
I have given up my position on "just hand out spice" already because I know
there are careers built purely based on IBIS as the greatest tool for SI as
a premise. And it is unlikely these people will yield to the fact that IBIS
is inadequate for high performance analysis. But why not do a parallel track
? Hand out encrypted spice and whatever the current favor of IBIS. That's
all I am asking.
-----Original Message-----
From: lgreen
To: Chris.Cheng@xxxxxxxxxxxx
Sent: 3/20/2005 5:40 PM
Subject: RE: [SI-LIST] Re: package SSN model accuracy requirements, now
Behavioral Modeling
Hi, Chris,
I agree with you that simulators suffer from convergence problems under
some
conditions.
What I was trying to say is that engineers should verify models against
hardware, whether they are "spice" or "ibis" or "*-ams", whenever
possible,
since simulation results are never any better than the models used.
My example was not that the simulator failed to converge, but that plots
of
Id vs Vds had discontinuities (between 0 and Vcc, no less). Likewise, I
have seen discontinuities in Id vs width. Neither of these is
physically
reasonable for a CMOS device, so these were fundamental modeling
problems.
- Lynne
PS: Admittedly many companies have AEs who can generate IBIS modes, but
some
companies push this back on the I/O buffer designer. Been there, done
that.
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]
On
Behalf Of Chris Cheng
Sent: Sunday, March 20, 2005 2:05 PM
To: 'si-list@xxxxxxxxxxxxx '
Subject: [SI-LIST] Re: package SSN model accuracy requirements, now
Behavioral Modeling
Lynne,
I don't think I/O designers provide IBIS models. I believe I/O designers
provide SPICE deck that can be abstracted into IBIS by the app engineer.
We have to be fair here, while I keep hearing this "I remember this one
case
the SPICE circuit doesn't converge, so it must be bad". The IBIS
standard
community failed to provide a correct solution for SSO for the pass few
years. I am aware of the late Larry Rubin's proposal for handling SSO
noise
in TLC (not even XTK) since the early 90's. What can be done and how
long
does it take to become a standard are two different things. Let's not
get
into pure academic discussion here by keep say "I think behavioral model
can
do this and that". Of course give it enough time and resources it can,
but
are you expecting the end user to ignore the problem until the standard
committee settle on the solution ? Worst, to even recognize there is a
problem ?
There is a de facto standard to describe circuits other than i-v curves,
encrypted HSPICE. It may be slow but it gives the end users a chance to
discover potential problems that IBIS can't. That's the bottom line.
-----Original Message-----
From: lgreen
To: weirsi@xxxxxxxxxx; kumarchi@xxxxxxxxx; si-list@xxxxxxxxxxxxx
Sent: 3/19/2005 7:58 PM
Subject: [SI-LIST] Re: package SSN model accuracy requirements, now
Behavioral Modeling
Hi, Steve,
The real sticking point is that IBIS models is what I/O designers
provide,
and IBIS does not model tightly coupled diff pairs. If you get a SPICE
netlist, you are better off, of course. But even SPICE models are not
foolproof - back when I used to design I/O buffers, I remember seeing
some
flaky BSIM models from some of the foundries we dealt with. ("Little"
things like discontinuities in I vs Vds.)
The links under Peter Lauritzen's page show a number of devices and
circuits
where no macromodels could match the transient circuit characteristics,
while a well-built behavioral model provided a better match to data
(i.e.
accuracy) and ran faster.
I guess the bottom line is to do whatever modeling it takes to get the
design out the door, while also verifying models whenever possible.
- Lynne
------------------------------------------------------------------
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:
