RE: IBIS comments to the EDIF committee

Stephen,

You're correct about RLCs for discretes...I had active parts on the
brain for some
reason. I will propose a location for storage of these values, since
there isn't one
now. Is there anything else other than the following that would possibly
need to be
described for a 'physical' netlist :
- active model (IBIS link)
- passive model (RLC values in EDIF)
- physical topology (layers, traces/vias, part geometries, etc. in EDIF)

Note that we don't have to be all-inclusive here...but if we don't get
something into
the spec now we will end up adding generic name/value properties that we
can't
require anyone to recognize.

David Wiens

>----------
>From: Stephen Peters[SMTP:speters@ichips.intel.com]
>Sent: Thursday, May 23, 1996 10:33 AM
>To: Wiens, David
>Cc: ibis@vhdl.org
>Subject: RE: IBIS comments to the EDIF committee
>
>Hello David:
>
> Regarding your questions below I'm going to give you
>my interpretation, but I think the simulator vendor folks
>need to give you the final word.
>
> 1. Yes, an IBIS file should be associated with a 'part' rather
>than each component instance.
>
> 2. The RLC values we are talking about refer to descrete components.
>A board my have descrete resistor, capacitor and inductor components
>connected as part of a net. The simulator vendors must have a way to
>clearly and uniquely extract from the EDIF description the values of
>these components (i.e. 50 ohms, 20pF).
>
> Stephen
>
>
>
>> On Thu, 23 May 1996 11:06:10 David Wiens wrote:
>
>Stephen,
>
>In working up the EDIF proposal, I came up with a few things I want run
>past you...
>
>1. In EDIF, a 'component' is an instance of a 'part'. Given the choice,
>I'm assuming we want to attach IBIS models to parts, not components.
>
>2. Regarding the issue of one location for RLC values...
> - There aren't ANY places I can find within EDIF that define RLC,
>so
> the only way to add them would be using the 'property' name/value
>type.
> Unfortunately, this can
> be done under about any element type.
> - The real point I'm missing is why we care about RLCs within EDIF
> if we can properly point at IBIS and get them there for package
>and die
> pins.
>
>David Wiens
>
Received on Thu May 23 10:09:20 1996