Re: IBIS modeling questions

Hi Greg,

You wrote:

>To: ibis@vhdl.org
>From: grege@chensys.com (Greg Edlund)
>Subject: IBIS modeling questions
>
>Anybody care to take a stab at these?
>

I'll give it a try.

>1. I was discussing IBIS model support with one of
> our IC vendors, and they raised a concern about
> ESD diode SPICE models not being accurate at
> -5 V and + 10 V. Any clues about how to get
> around this without setting up a test lab? Why
> do you need such a wide range in the first place,
> since nobody operates out there?

I think that some simulators use the last couple of points to extrapolate out
to infinity or some asymptotic value, so if your data is not available over the
full range, you can probably do the same. Maybe it doesn't matter whether the
full range is provided. It might depend on a simulator's interpretation of the
IBIS specification. If I remember correctly, using the full range is optional
in the latest IBIS revision. (Someone help me here.)

>2. When will s2ibis 2.0 be available on Windows?
> Actually, I'd be happy with DOS, but that's
> probably not in the plan...
>

I believe someone would be glad to make the source code available to you. Since
you seem to have some experience with C compiling, you might be the right
person to create the DOS/Windows executable and upload it to ibis@vhdl.org.

>3. Has anybody written a "theory of operations"
> paper on behavioral simulators? For instance, if
> I have a simple behavioral driver model (no ESD
> diodes, no package) driving a 50 Ohm transmission
> line terminated at the far end into 50 Ohms, how
> do I calculate the current and voltage vs. time at
> the output of the driver? At the far end? This
> must be an easy problem to solve. I'd like to
> write a little C program to do this so I can better
> understand IBIS and critically compare it to SPICE.
> (For the sake of accuracy in system design issues,
> not so I can write another simulator!)
>

Maybe, I can't really say.

I believe the way to approach this is to use piece-wise linear dependent
functions to control a conductance to Vdd, simulating a pull-up switch and one
to Gnd simulating a pull-down switch. The independent variable controlling the
dependent conductance functions would be the voltage at the driver's output
pad. Both conductance switches also need to be dependent on the pin's logic
state and need to make a transition between states at a controlled rate using
the dV/dt information in the IBIS. The state transition would have to occur
based on some threshold value and the buffer's input level.

Sounds simple, only the details get in the way.

>4. Has anybody thought about organizing an IBIS
> users group and mailing list?
>

This is it, I think. I would be happy to discuss the "theory of operations" of
behavioral simulators in more detail.

I agree that for the "sake of accuracy in system design issues" it is good to
know issues and limitations involved with behavioral model simulations. It may
be difficult to discuss IBIS models independent of a specific simulator, but
might be interesting to try.

Regards,

Fred Vance
FirePower Systems, Inc.
(415) 462-3055 Phone
(415) 462-3051 Fax
Received on Thu Nov 9 12:52:13 1995