Re: Connector spec swathing

I would like to add an illustration of Chris Reid's point, as I see it.
Here is an ASCII picture of a connector. The pins marked as 'X'
are those carrying the signals that the simulation product has
determined MUST be characterized and simulated. The '0' pins are
the other pins of the connector:

0 0 0 0 0 X O 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
 0 0 0 0 X 0 0 0 0 0 0 0 X 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X 0 0 0 0 0 0

Assume that the connector model uses a 3x3 swath with 1 left edge
and 1 right edge (5x3 total matrix size). Here is one possible
pattern for using the swath, where 'M' represents matrix middle pins,
and 'E' represents pins in an edge column, which must be terminated
by Cn_Z. The '?' pins are those that could be either an edge or a
middle, depending on which nearby swath it is considered to be "in":

0 0 E M M X E 0 0 0 E M M ? ? M M E 0 0 0 0
 0 0 E M X M E 0 0 0 E M X ? ? M M E 0 0 0
0 0 E M M M E 0 0 0 E M M ? ? X M E 0 0 0 0

There is no single swath that can contain the 2 pins on the right.
But 2 swaths, each centered around one pin, will overlap. Most likely
the simulator will have to make some choice that does not fulfill the
intent of the model developer. Even the 2 pins on the left have some
ambiguity; I chose the pin in the middle row as my center point, but
the choice is not always obvious.

On the subject of centering a swath around a pin, a swath with an even
number of columns has four possible locations, giving four possible
simulation results, depending on which is chosen. This is because the
swath has no pin at the center, but has 4 pins equally close to the center.

Since simulators have to handle a full matrix anyway, I would have to
join Chris Rokusek in hoping for a solution that turns a connector model
into a full matrix.

Mike LaBonte
Cadence

Chris Reid wrote:
>
> Hello,
>
> Chris is absolutely correct. We need an unamiguous method of
> building an arbitrary submatrix from the connector swath. At
> ICX we wrote a program that translates SPICE connector models
> into full matrix models which our tool can use. This is a
> difficult process and needs significant input from the user.
> There are many cases with ambiguities, sometimes without even
> one clearly right answer let alone the case where there might
> be more than one right answer.
>
> The problem is best illustrated by considering what the SI
> tool must accoomplish. Given a particular situation to simulate
> (which may include coupling to other nets) certain pins on each
> component must be included in the simulation. If any two (or
> more) of these pins are situated such that using the swath for
> both of them results in an overlapping set of pins (but not
> coincident) then what is the coupling matrix that should be used?
>
> Chris Reid
>
> Chris Rokusek wrote:
> >
> > Kellee & IBIS connector committee,
> >
> > The DAC IBIS meeting was one of the best I've attended...well done!
> >
> > I would like to request that the description of the "swath" matrix also
> > describe how to properly construct a fully-coupled matrix from the swath
> > matrix since there appears to be more than one "right" way to do it.
> >
> > It would also be nice if the parser itself could perform this algorithm so
> > that a simulator may (if it so desires) just query the full matrix
> > regardless of how it was specified (swath, full, sparse, ...).
> >
> > Best Regards,
> >
> > Chris Rokusek
> > Innoveda
Received on Wed Jun 14 07:40:13 2000