Re: [Gnucap-devel] SPICE to gnucap transition

[Dan McMahill]

al davis wrote:
> On Sunday 18 March 2007 18:54, Dan McMahill wrote:
>
> > I wonder what it would take to catch gwave up a little or to
> > teach gtkwave to read the ascii data that gnucap is currently
> > able to produce.
> >
> > A better waveform viewer (including calculator) is something
> > I think is really needed.  Too many projects, not enought
> > time....
> >
> > gwave is really lacking in any analysis features like octave
> > but the octave plotting tools really aren't good for
> > interactive waveform viewing (lack of scrolling, cursors,
> > etc).
> >
> > This is an area where I think it wouldn't even be that hard
> > to be way better than some of the high end commercial tools.
>
>
> I think a complete replacement for gwave, with the existing 
> functionality, could be done in a summer, so it might be a good 
> project for the "Summer of Code".

Whats somewhat amazing to me is I think in a summer one might even be 
able to approach or exceed some commercial tools.  It has amazed me how 
truely bad some of the commercial offerings are.  Some like cadence's 
analog environment are so close but yet so far from being really good. 
They have the extensibility (very useful) but you can't define your own 
gridlines or your own cursor readout function (think smith chart -- 
which their tool does -- or a nichols chart which their tool does not 
do).  Also it is a major pain to change the plot styles (thin line, 
thick line, type of marker, etc).


> That and a replacement for "gnetlist" are the biggest needs.
>
> The viewer part of qucs is much better, but qucs is too 
> integrated to be really useful without huge investment.
>
> At Bell Labs, we did a lot of that stuff in "S".  For those 
> unfamiliar with "S" .. it is a statistical language with 
> graphics.  It spawned a commercial product "S-plus", and a Free 
> project "R".  The original developer of "S" is now affiliated 
> with "R".  "S" and "R" have the ability to do math on tables 
> and equations easily.
>
> What I would like to see is a new waveform analysis tool, with 
> analog emphasis, entirely written in an interpreted language 
> like python or ruby.  It is important that it is "entirely" 
> written in the interpreted language, not a hybrid, because of 
> installtion and extendability issues.  gwave is a hybrid.
>
> Then ... someone could write a C++ wrapper that would interface 
> between the new analysis tool and gnucap, which would make it 
> fully integrated.


And I'll add that it is very important I think for it to be easily 
extensible.  I needs to have a waveform calculator available that can be 
extended.  I want to be able to plot v(1) + v(2).  Then I'll want to 
plot db20(v(x)/(1 + v(x)).  Then I'll want smith chart gridlines and 
nichols chart gridlines, and some other less common ones.  Then I'll 
want fft's, and more complex calculations...  You get the idea.  If it 
is design with extensibility in mind, you can grow the built in 
functionality simply.


-Dan