RE: confusion about c_comp

Hello,

To sum this problem up, I usually say that C_comp needs to
be visible from the outside looking in, but invisible from
the inside looking out.

Here is how I did it in my HSPICE implementation (before the
B-element came out). I used PWL sources to contain the I-V
and V-t tables and some additional controlled sources to do
the math between these. C_comp was a normal capacitance
connected to the end of all this. If I would have left it
this way, I would have indeed double counted for C_comp,
because the shape of the V-t curve in the PWL source already
includes its effect. Yet I knew I had to have that capacitor
there, because a reflection that hits this (driving) buffer
will bounce back from it differently if it was not there.

This is what I did to remove the double counting. I connected a
duplicate C_comp on the output of the PWL source(s) which had the
V-t curve(s). Then I just simply included an additional term
in my math which combined the I-V and V-t curves so that the
current of this duplicate capacitance was subtracted from the
total result. I know this will have some error as the actual
waveform deviates from the V-t curve, but this error is smaller
than double counting C_comp (zeroing it out).

Of course, when you write the code of your simulator you have
much more control over writing an algorithm that does a better
job than this.

Arpad Muranyi
Intel Corporation
================================================================

-----Original Message-----
From: Christopher Reid [mailto:chris_reid@mentorg.com]
Sent: Thursday, December 21, 2000 9:02 AM
Cc: ibis-users@eda.org
Subject: Re: confusion about c_comp

Hello IBIS land,

I don't see this as confusing at all. The VT curves are what the
simulator should produce under the load conditions specified in
the IBIS model for each curve. The simulation vendor must devise
an algorithm to "make it so."

The C_comp must be in place during the actual simulation because
it will affect the results in different ways for different loading
conditions.

Chris Reid
Mentor Graphics

"Lorang, David D" wrote:
>
> This is a great thread! I think many of us, both SI tool users and
> simulator developers, have struggled with this Ccomp problem.
>
> It is possible to match up the V-T curves using a piecewise linear ramp
> waveforms, various controlled voltage or current sources, and perhaps some
> resistors, but no storage elements. If the simulator produces its
internal
> model to match the waveform in this manner, then afterwards adds Ccomp to
> the model, then the Ccomp has been double counted and the final model
would
> be in error.
>
> What should happen is that the simulator should match the IBIS waveform
> using all of the ramps, sources, resistors, but with Ccomp in place also.
> Then, because Ccomp is already in place in the simulator's internal model,
> it must not be added in again.
>
> I have seen this problem, when trying to correlate IBIS based models with
> the transistor level simulations they were developed from. You see the
> problem when waveforms do not correlate unless the Ccomp is manually
"zeroed
> out" in the model.
>
> I'm not sure we can fault the vendors too much because it is easy to get
> wrong and the IBIS Specification standard does not spend a lot of words
> explaining all of this. Perhaps some more clarification in the IBIS
> Specification standard is in order. Further thoughts?
>
> Best regards,
> Dave Lorang
>
> -----Original Message-----
> From: Ingraham, Andrew [mailto:Andrew.Ingraham@compaq.com]
> Sent: Thursday, December 21, 2000 7:14 AM
> To: ibis-users@eda.org
> Subject: RE: confusion about c_comp
>
> > > Any simulator that uses the V/T data, must be careful enough to use
> > > it correctly!
> > > ...
> >
> > It is actually very difficult to implement it correctly.
>
> Yes, I'm sure it is!
>
> > C_comp is included in the V/T data, but it is not included in the
> > pullup/pulldown data. This means that to be truly correct, some kind
> > of add and remove algorithm is needed,
>
> ? The pullup/pulldown (I/V) data is DC data. C_comp is irrelevant, it
has
> no effect on the pullup/pulldown data. Maybe I misunderstood you.
>
> Regards,
> Andy

 
Received on Thu Dec 21 12:51:01 2000