Re: gEDA-user: A little puzzled about the purpose of gschem

[al davis <ad252@xxxxxxxxxxxxxxxx>]

On Saturday 24 April 2010, Link wrote:
> I hadn't intended for anyone to interpret it that way, and
>  I'm sorry if you interpreted that as bashing gEDA. Perhaps
>  my choice of words was rather unfortunate.

Apology accepted.  

> What I intended to is that one component (the simulator) of
>  LTSpice tends to perform better _for me_ than gEDA's
>  equivalents, in sheer terms of the simulation results being
>  what I expect. I'm not sure if that is even a problem with
>  gnucap/ngspice or if I'm simply doing something wrong
>  myself, but I do know that I find LTSpice easier to use for
>  simulation. That does not mean I dislike gEDA, or think it's
>  bad, or anything of the sort - in fact, I find it to be
>  absolutely brilliant for schematic capture and PCB design.

There is a real problem with the way "gnetlist" works.  
(contrary to what some others here say).  I have known this for 
years.  Often people have trouble with it.  I don't know what to 
say.  It seems I always need to hack the netlist.  I have tried 
to recruit help on this, specifically making gschem and pcb 
plugins for the gnucap translator system.  No takers.  Maybe 
when I get my other pile of work out of the way I will do it 
myself, for both geda and kicad.

Then "gspiceui" doesn't help.  I don't like that style anyway. 
For a beginner, combined with how gnetlist works, I don't know 
what to do.

In this environment, it's not gnucap or ngspice that is the 
trouble spot.  It's gnetlist, documentation, and communication 
between tools.  So the situation here is that people don't see 
the capability that ngspice and gnucap have because of problems 
elsewhere.


The other problem people run into is that nobody here has 
collected the hundreds (thousands?) of models (of all kinds) 
that come from all over.  A commercial organization can pay a 
junior person full time to maintain the collection.  That's what 
it takes.

Not in a commercial environment, this kind of thing has to be 
done by the community, a shared effort. 

But really "google for it", and "check to see if it makes sense"  
is the correct answer.  A lot of those models don't make sense 
for the particular application people are asking about.  The big 
collection that LT and P spice have, that is still incomplete 
because it is impossible to be truly complete, lulls users into 
a false confidence.


As far as simulation results go ..   If it gives you trouble, 
please report it.  I'm well aware of the Spice3f5 false 
convergence issue, and that it still exists in NGspice.  I'm 
well aware of the NGspice step control bugs, where an attempt to 
fix 3f5 problems really made it worse.  I am not convinced that 
LTspice is any better in these issues than NGspice.  As far as I 
know, Gnucap doesn't have these problems. 

Convergence and step control are never perfect.  It seems to 
work for me, but if it doesn't for you, the only way it can get 
better is if you let me know about the problem, and send a test 
case that demonstrates it.

The benchmarks I have tend to show that Gnucap usually 
outperforms NGspice on difficult circuits, and that LTspice 
usually performed about the same as NGspice.  NGspice is often 
faster than Gnucap on small circuits, where both are fast enough 
that it doesn't really matter.  The last time I made any of 
these comparisons was a few years ago, so some things might have 
changed.  Gnucap has improved, perhaps LTspice has improved too.  
NGspice hasn't really changed in this regard.




_______________________________________________
geda-user mailing list
geda-user@xxxxxxxxxxxxxx
http://www.seul.org/cgi-bin/mailman/listinfo/geda-user