Re: New package model proposal by Steven Peters

Arpad,

Yes you understand me correctly. How else am I going to model a DRAM SIMM connected
to and loading my main logic board? Without having the vendor's PCB design
database, I have to guess how long each control, address, or data line is and what
the impedance and delay characteristics are, and I have to assume that there is no
coupling between traces.

If the SIMM were treated as a package it would be possible to define the minimum
and maximum delay and loading for each signal. This would allow the vendor to
provide information about his SIMM without revealing his PCB design. From my point
of view, a 72 pin SIMM is no different than a 100 pin PGA. Unless I want to dissect
the package, I only know what the vendor tells me.

If I were a SIMM vendor, I would use IBIS to describe my SIMMs if possible. If IBIS
allows bond wires and lead frames to be described and if Steven Peters' Egg will
enable IBIS to add coupled or uncoupled transmission lines to the package
description or some equivalent matrix, it seems to me that it would be possible for
a SIMM vendor to provide the information that I need without revealing his PCB
design.

Where do we draw the line between components and circuit boards? My car has an air
flow sensor circuit board with two die on it. I have read in news reports, that the
P6 has two die in one package. I believe the difference lies only in the need for a
person to successfully design the part into an application. If the success of the
design depends on knowning the I/O buffer and package characteristics, to me at
least, it is a component.

Regards,
Fred Vance

Begin forwarded message:

Date: Wed, 10 May 95 13:09:00 PDT
From: Arpad Muranyi <Arpad_Muranyi@ccm.fm.intel.com>
To: ibis@vhdl.org, fvance@FirePower.COM
Subject: Re: New package model proposal by Steven Peters

Fred,

Do I understand you correctly? Are you suggesting that the PCB portion of SIMMs
should be described as if they were packages?

Even though I see the similarities between PCBs and MCMs or other types ofpackages,
I strongly disagree with your wish to be able to handle SIMMs as ifthey were
packages. SIMMs are clearly printed circuit boards with multiplecomponents on them
and they should be handled as such. The best way of doing so
is to make full-board simulation tools understand multiple board situations, sothat
a motherboard with a SIMM (or more) could be simulated as two or more PCBsconnected
together.

As Stephen said, IBIS is a component modeling format and not a PCB
descriptionlanguage.

The question in my mind is whether MCMs and the like should be handled as
PCBs,because from a user's point of view, an MCM package is one piece of
somethingand from the outside it looks like a single component.

It is true, that vendors have a hard time to release models of their SIMMs. Ifyou
were a SIMM vendor, which format would use to create a model? Quad, xSPICE,
Interconnectix, EDIF, etc...?

As I see it, a generic interconnect description language standard is missinghere,
which could be used by any simulation tool. This standard should have the
capabilities to describe PCBs as well as packages, and then it could be used for
SIMMs, motherboards, MCMs, complex packages, etc...

Arpad Muranyi
Intel Corporation,
Folsom, CA
Received on Wed May 10 14:54:27 1995