Re: monotonicity

On Sep 5, 8:42am, edlund@hendrix.eng.pko.dec.com wrote:
> Subject: monotonicity
>
> > Why is their a restriction put on monotonicity? Is it because
> > some simulators cannot handle this case?
> >
> > In 2.1 it states that most simulators filter non-monotonic data.
> > Does this mean that it's OK to feed non-monotonic IBIS models to
> > most simulators i.e. it will generate results based on filtered
> > data?
>
> I believe non-monotonicity can be indicative of a feedback
> mechanism employed in the I/O buffer, for example a slew
> rate controlled driver or a driver that matches its output
> impedance to the transmission line it's driving. These
> kinds of buffers are inherently difficult to model using
> the behavioral approach. You wind up with things like
> models that are only accurate under certain loading
> conditions.
>
> I've heard people make this criticism of behavioral models
> for GTL drivers, but I've never actually compared a
> simulation using such a model to SPICE or lab data.
> Maybe someone can correct me if I'm off-base on this point.

Hi Greg,

I've got good agreement between HSPICE and behavioral GTL models from
a selection of silicon houses (+/- 5 mv and +/- 10 ps). Both compare
reasonably well to lab measurements. The catch is I design the layouts not
turn on protection diodes. Also I write my own SPICE to behavioral stuff.

... Rich Mellitz, NCR

>
> Greg Edlund
> Alpha Server Development
> Digital Equipment Corp.
> edlund@eng.pko.dec.com
>
>
> p.s. Come to think of it, you would need to get your
> I-V curve from a transient analysis to accurately capture
> the behavior of a slew rate controlled or impedance
> matching driver...
>-- End of excerpt from edlund@hendrix.eng.pko.dec.com
Received on Thu Sep 5 07:40:58 1996