Enhanced Package Models

Hello Fellow IBISans:

     I have been tossing around ideas for enhancing package model
descriptions by describing package pins as a series of transmission
lines. In some cases a package modeled as a series of uncoupled transmission
lines is adequate, in others a series of coupled transmission lines
is required. I was thinking about enhancing the existing '.pkg'
description to include both pin length info and the idea of each
'pin' (package stub) as having different sections. I've also spoken
with Kellee Crisafulli at Hyperlynx about this and he has offered the
idea of describing the package at a physical level using some sort of
a PCB like description. His points in brief:

 1. Simulator companies support coupled transmission line simulation
     by reading in PCB information. By treating the package as a
     'PCB' we have a 'natural' way of getting package info into a simulator.
 2. With this enhancement IBIS could then handle things like SIM modules,
     MCMs, etc.

      Some problems/questions:
  1. Do we have to create a standard 'PCB' like description? There is an EDIF PCB
     standard, but does anybody support it? What would it take
     to create our own?
  2. Getting CAD companies to output such a format (There is
     a company called Router Solutions whose business is to write reader/
     writers from any format to another).
  3. Presently, packages are usually described in a purely mechanical
     way without the required dielectric info, layer stack up, etc.
     How can the user get this mechanical info into our description.
  4. Are there possible proprietary issues involved with giving out
     physical package (construction) info?

     An alternative method is to supply the user with the electrical
information only. For simple packages that can be represented by
a series of uncoupled transmission lines it's a relatively
straight forward, simple enhancement to the '.pkg' description. For
coupled transmission lines it's a more complicated matrix style
description, but doable. Points in brief:

  1. It's a 'behavioral' description, which is consistent with what IBIS
      is all about.

  2. For the matrix description there are several ways of extracting
      and presenting the coefficients in the matrix. The forum will have
      to decide on a consistent methodology.

  3. For well behaved packages (i.e. ones that can be described with
      uncoupled transmission lines) a direct description for each
      pin would be simple and quick to implement, and for simulator
      companies to use.

     Of course, we can do both, or some combination of the two. I'd like
to discuss this more at the meeting tomorrow, with an eye towards getting
somthing concrete going.

                   Best Regards,
                   Stephen Peters
                   Intel Corp.
Received on Thu Apr 6 16:10:53 1995