Re: IBIS Package Model

gedlund@us.ibm.com wrote:
>
> Mike,
>
> Can you elaborate a little more? It sounds like you're trying to model the
> ground plane in a package as a non-ideal return path. Usually we make the
> assumption in transmission line analysis that the return terminals are
> connected to ideal ground, but this doesn't work when you're modeling power
> rail collapse on an IC due to package inductance. It sounds like you want
> a package model with two ports for each signal conductor and more than two
> ports for the ground conductor. Is this the case? If so, you're talking
> about a 3-D package model since a package model based on a 2-D cross
> section would only have two ports for a non-ideal ground.
>
> The quick answer is that IBIS can do coupled lumped package models or
> uncoupled distributed package models, but not coupled distributed package
> models. In my experience, I've used multiple lumped segments to describe a
> package; these were based on multiple unique 2-D cross sections with
> non-ideal grounds. However, I've never seen anyone derive a coupled
> distributed package model from a 3-D field analysis.
>
> Greg Edlund
> Advisory Engineer, Critical Net Analysis
> IBM
> 3650 Hwy. 52 N, Dept. HDC
> Rochester, MN 55901
> gedlund@us.ibm.com
>
> ---------------------- Forwarded by Gregory R Edlund/Rochester/IBM on
> 03/17/99 05:21 PM ---------------------------
>
> "Mike Lamson" <m-lamson@ti.com> on 03/17/99 07:46:27 AM
>
> Please respond to m-lamson@ti.com
>
> To: ibis-users@vhdl.org
> cc: (bcc: Gregory R Edlund/Rochester/IBM)
> Subject: IBIS Package Model
>
> Does anyone know how to represent RLC in IBIS for a ground plane with
> multiport nodes? The issues is the capacitance. The inductance and
> resistance can be described between any two ports of the plane but how to
> represent the capacitance in IBIS is a mystery. If the total plane
> capacitance is used, it gets repeated for each port which over estimates it
> in the model.
>
> Thanks,
>
> Mike Lamson
> m-lamson@ti.com
Full wave solvers will give you S parameters. These can be interpreted
into SPICE Laplace transfer functions that retain the frequency effects.
Therefore its possible to have the behavior of the system modeled in
a simulator accurately. Other approaches involve FDTD and similiar
methods. However Mike's question pertains to IBIS. In that regard
Greg's answer seems correct. Lumped RLC circuits from 3d analysis can
become large and hard to handle by time domain simulation.

Best Regards,

-- 
Fred Balistreri
fred@apsimtech.com
http://www.apsimtech.com