RE: IBIS at >400MHz

Hi folks,

Can IBIS help at > 400 MHz. Maybe.

Here's my opinion (mostly babbling):

At < 100 MHz we can talk about register to register
transactions. We can talk about a simple algebraic
equation. Clock to out, flight time, setup etc.
Simple, eh? Maybe we even through in fudge
factors for clock uncertainty and SSO.

At > 100 MHz there are more players; statistical
com theory, feedback, and crazy types of dispersion
to name few. You need to understand these to
develop design guidelines. For these you need just
about every tool and trick you can muscle up. No tool
alone is sufficient. You need to do some "real hard
to make measurements" too. I guess this validation
stuff is your basic a multi-dimensional task ;-)
Who knows? You may get lucky and find a case
where simple IBIS alone will work for you.

After all that, the IBIS tools can be relegated the
mammoth work horse task at the system level to
being the "bell ringer" for problems you may want
to look at closer. The real trick is to figure out
what bells to ring and how loud to ring them.
Anyone got any bells?

Richard Mellitz, NCR
 

>----------
>From: bob@icx.com[SMTP:bob@icx.com]
>Sent: Thursday, October 09, 1997 3:14 PM
>To: ibis-users@eda.org; pgregory@hpbs2933.boi.hp.com
>Cc: ericb@cadence.com
>Subject: Re: IBIS at >400MHz
>
>Paul:
>
>You raise some very good points and questions. I don't
>have definitive answers. I support using Spice simulation
>and measurement as part of the design methodology along
>with IBIS based behavioral simulation. I would expect
>that as specific IBIS format limitations are identified
>with respect to higher frequency simulation accuracy,
>proposals to ammend the IBIS format will be made.
>
>Eric Blomberg in "IBIS "models" come of age" in the October 1997
>issue of Printed Circuit Design. p. 20 provides a very concise
>paragraph that addresses your points and provides some thoughts.
>Eric's paragraph follows:
>
> "Can I use IBIS models for my 600 MHz bus?" The simplest
> answer is yes - if the IBIS provider has validated the
> model at 600 MHz. It is true that the device effects which
> are simplified out of the IBIS model format start to become
> visible about 300 MHz. It is also true that other effects,
> such as dielectric loss, and the modeling engine's way of
> handling this become significant as well. Device package
> models play a larger role. The expectation answer is that
> above 300 MHz, validation becomes a part-by-part issue for
> the entire modeling system.
>
>Bob Ross
>Interconnectix
>
>
>> From: Paul Gregory <pgregory@hpbs2933.boi.hp.com>
>> Message-Id: <9710031834.AA20168@hpbs2933.boi.hp.com>
>> Subject: IBIS at >400MHz
>> To: ibis-users@eda.org
>> Date: Fri, 3 Oct 97 12:34:01 MDT
>
>> In future designs, we plan to use interfaces that
>> reach beyond 400 MHz. As I think about the tools
>> that we need to help us successfully design at that
>> speed, simulation is going to be a big part of it.
>> But I am concerned that IBIS descriptions and
>> simulation techniques may not be adequate to accurately
>> simulate at that speed. Perhaps SPICE simulation
>> will be required.
>
>> Should IBIS based simulators be capable of accurate
>> simulation at speed of 400MHz to 1000MHz?
>
>> When device parasitics play a larger part of the
>> equation, is IBIS sufficiently robust to accurately
>> describe these parasitics?
>
>> What are the key factors to consider when simulating
>> at these higher frequencies?
>
>> What are your thoughts?
>
>> -- Paul Gregory
>
>> phone: (208) 396-5086 USmail: Hewlett-Packard
>> fax: (208) 396-4122 M/S 143
>> email: paul_gregory@hp.com 11311 Chinden Blvd.
>> Boise, ID 83714
>
>
>
 
Received on Mon Oct 13 06:32:24 1997