Re: [Help-gnucap] newbie: edit and BJT's

[Al Davis]

On Friday 04 October 2002 03:37 am, Hugo van der Merwe wrote:
> > This is a known bug in 0.31, fixed in 0.32, due out soon.
> Cool, thanks! (May I ask what is "soon" in the gnucap project, a
> couple of weeks?)

There is a password problem that is holding things up.  For now, you 
can get 0.32 at ftp://progeny.isu.edu/pub/gnucap/gnucap-0.32.tar.gz  .

There are also some papers and presentations there.

> > The one that isn't fixed is what happens when reading an "NA"
> > in the input.
>
> Which probably doesn't matter, does anything generate netlists with
> NA, now that gnucap no longer does?

It really does matter, but is not urgent.  It is possible for user 
defined models to have the same problem.

The BJT, diode, and mosfets are all implemented through the model 
compiler, which is the way a user can add new models without a lot of 
hassle.  The model compiler bug^H^H^Hfeature that makes this possible 
is still there.

> > I can't answer this one without a netlist.  Do ng-spice and
> > gnucap agree?  That is, give the same "strange" result?  How
> > about older versions of spice?  If so, that leads me to believe
> > there is something wrong with the parameters.
>
> Gnucap and ngspice agrees.

That tells me that the Gnucap implementation of the model is 
"correct".

An implementation of a model is correct if it gives the same results 
as a reference implementation.

Whether the model itself is correct or not is another issue, and one 
I choose not to deal with.

>......... I could possibly try hspice later. I was
> more concerned about problems similar to what I had with ngspice:
> it being strictly spice3 based, Is parameters greater than 1 aren't
> adjusted... iirc. I'm happy to hear gnucap is aiming at full
> spice2/spice3 compatibility...

Actually ... there is a bit of conflict there.  The MOS 1,2,3 models 
in Gnucap were designed to be Spice-2 compatible.  The Spice-3 models 
are a little bit different.

> > > How far down the line might monte-carlo analysis be, years,
> > > or months?
> > Probably a long time, unless someone (you??) wants to help do
> > it.
>
> (What is a "long time" in the gnucap project, a couple
> of years, I'd think?)

That means I am hoping for someone else to do it.  There are other 
higher priorities.  A scripting language is one of them.