Conductance (was: BIRD 42.3)

Bob Ross wrote:

> (2) It is also interesting that you pose the problem
> in terms of conductance variations (which could
> be implied by the IBIS architecture.) It could
> have been presented in terms of resistance
> variations. However, there would be an implied
> difference. Pulldown one, for example goes
> linearly from 0 to 20 mS conductance. If it
> were expressed as resistance, it would go
> non-linearly from infinite to 100 ohms to 50 ohms
> resistance.

In general I dislike dividing by zero; I'd much rather
deal with tens of siemens at one end of the scale than
(fictitious) millions of ohms at the other. Also, it
turns out to be a more-nearly-realistic time-domain
process with the output device ramping up its conductivity
as its controls change state. ("Hello, students, can
you say 'transconductance'?" Sorry -- Friday.)

> (3) I understand that there are other real world
> effects that can enter into the bounce analysis
> that may not show up in the IBIS model processing.
> I would still think that a delta current could
> still be used for some of these effects that
> are caused just by the device switching itself.

I've been dealing with conductivity modulation in output
devices for more than a year and don't see a clean solution
short of full SPICE simulation. Life just isn't fair
sometimes. Until someone smarter than me comes up with a
solution, we'll have to deal with the sort-of-linear
model that's implied by IBIS. It's really not all that bad
in principle but there are all sorts of uglies hiding in the
details. If we're going to get serious about doing SSO
analysis in an IBIS context the biggest obstacle isn't
crowbar current (sorry; cheap shot) it's the lack of useful
power-distribution information. Supply connections are
extremely local, so it's not helpful to know the total
number of power and ground contacts on a part -- you have to
know which ones feed which pins and their relationship to
other supplies less near. Once we have *that* we can make
a pretty good guess about the rail potential changes and IBIS
v/i curves do a fair job from there.

-- 
D. C. Sessions
dc.sessions@tempe.vlsi.com