Re: Rising Waveform Loading Effects

I think I get the point Dileep is trying to make. It is true that with the V-t data there is no UNIQUE solution possible. This problem will baloon when we add multi stage models.

The real issue is IBIS is ONLY data. And the data is subject to different interpolation and extrapolation schemes. In other words this data by in itself cannot guarantee that different simulators will produce the same result. Neither can it say which one is correct. The only locus is the golden waveform. In this respect IBIS data is fundamentally different from a macro model , where you should expect the same results from different simulators. IBIS data is rather a precursor to a macro model, which we do not have any standard for.

- kumar

> From owner-ibis@vhdl.vhdl.org Tue Oct 29 17:31 EST 1996
> Received-Date: Tue, 29 Oct 1996 17:31:23 -0500
> From: dileep@contec.Apsimtech.COM (Dileep Divekar)
> To: ibis@vhdl.org
> Subject: Re: Rising Waveform Loading Effects
> Cc: dileep@contec.Apsimtech.COM
> X-Sun-Charset: US-ASCII
> Content-Type: text
> Content-Length: 2417
> X-Lines: 46
>
>
> Arpad Muranyi, Intel Corporation wrote:
>
> > What I mean by this is, that if the simulator is smart enough, it can take the
> > data and modify it, derive new curves, etc. So, if you have an I-V curve and a
> > V-t curve which was generated with 50 Ohm resistive load in the model, a good
> > simulator should derive a new V-t curve that matches the different load
> > resistance in that particular simulation. Taking this further, the simulator
> > could make V-t curves with reactive loads also. In general, then, the simulator
> > can simulate at each iteration with a different set of curves that matches any
> > loading condition, even if the "load" is another non-linear I-V curve. The
> > point here is that the data in the IBIS model is not supposed to be a limiting
> > factor to what the simulator does with it.
>
> This is not quite true. Simulator cannot CREATE additional information.
> If V-t data is given only for one resistive load, how can the simulator
> create V-t data for arbitrary loading conditions? To illustrate the point,
> let us assume that the so called "load" can be described by only one
> independent variable. (In real life, it is an n-dimensional problem.)
> Let us assume that the dependent entity is the V-t curve. Thus, in this
> hypothetical two dimension space, the ibis specification gives one data
> point. From this single data point, there is no way to produce a straight
> line, or at least there is no UNIQUE way. If the ibis specification were
> to give two data points, then one can say that the simulator should be
> smart enough to fit a straight line thru these two points and then
> interpolate and extrapolate(although extrapolation is dangerous in many
> situations).
> [In practice, ibis specification has room to provide V-t curves for
> multiple loads, but the simulator is still limited by the information
> provided in the ibis model. It cannot CREATE V-t curves for the
> general n-dimensional case. The simulator may be able to do this if it
> makes certain assumptions, but since these assumptions are not specified
> in the ibis model, the simulation results will be unpredictable (and even
> unreliable).]
>
> ------------------------------------------------------------
> Dileep Divekar
> Applied Simulation Technology, Inc.
> 2188 Bering Drive
> San Jose, CA 95131
>
> Phone - (408)-434-0967 x 100
> Fax - (408)-434-1003
> Email - dileep@apsimtech.com
> ------------------------------------------------------------
>
>
Received on Thu Oct 31 11:43:50 1996