Re: gEDA-user: Question regarding 1st LTSpice simulation

[al davis <ad151@xxxxxxxxxxxxxxxx>]

On Monday 17 September 2007, Dan McMahill wrote:
> I'd comment, but I probably can't give a good measure of the
> learning curve.  From my perspective, if you've used any
> circuit simulators, ngspice and gnucap are both pretty easy.
>  But then again I first used spice nearly 2 decades ago and
> use circuit simulators daily so most of my learning curve
> memory is pretty distant.  gnucap has some neat capabilities
> like being able to directly get at some internal components
> of device models (junction current vs charging current in a
> diode for example).  gnucap also is quite a bit better than
> ng-spice for mixed mode sims since it was designed for that.
>  I can't recall though if gnucap has small signal noise
> analysis at the moment.

No.  no small signal noise analysis.  Use NG-spice for that.

There is also no small signal distortion analysis.  Again, use 
NG-spice for that.  On the other hand, I have never found the 
small signal distortion analysis to be very useful, because it 
doesn't show large signal distortion at all.  What I have found 
more useful is a real Fourier analysis.  For this, gnucap wins.  
The Fourier analysis and time stepping work together to 
significantly lower the noise, so it is actually useful for 
measuring distortion.

Gnucap is a lot faster for large circuits.  I have one that runs 
in 8 hours in NG-spice, 40 seconds in gnucap.  Quadratic time 
vs. linear time.

As to the learning curve, it depends on where you are coming 
from.  gnucap interactive commands are different from spice.  
As a teacher, I found that I could get students started faster 
on gnucap than any other, even the graphic ones.  If you are 
starting cold, the command line is really the easiest way!

The biggest trip point is the sequencing of attaching probes and 
doing an analysis.  Batch spice doesn't care about sequencing.  
As a result, it doesn't let you play.  Gnucap cares completely 
about sequencing.  You need to attach the probes before you 
turn the power on (run the analysis), like you would with a 
GUI, or a real circuit.  Gnucap is more like a breadboard 
metaphor.  Spice is more like a "declarative programming 
language" metaphor.


> I'm a fan of learning about netlists and doing the first
> couple of sims by typing in a netlist by hand.  Why?  Because
> even with expensive commercial CAD systems, problems come up
> where you have to dig into the netlist to debug.  Besides,
> it's one less thing to worry about when you're getting
> started.

That's another thing I ran into in teaching.  Other profs would 
teach only with a GUI, so they might not see a netlist ever, or 
until they hit a course I was teaching.  I start them with a 
netlist, then let them learn a GUI later if they want to.

With Spice, you need to make a file containing the netlist.  
Gnucap lets you type it in interactively, then make many cycles 
of change and simulate again, interactively.  With Spice, every 
change and simulate again is another edit of a file.

Too often, students are taught simulation as an afterthought.  
You do everything else, including actually build one, then 
simulate to appease the professor.  They don't learn that a big 
reason for simulation is that it is easier than the many 
rebuild cycles on a real circuit.  They also don't learn that 
simulation can give you data you can't measure, and can 
directly give you the numbers you want, as opposed to measuring 
what you can measure and calculating from there.  gnucap is 
better in this respect, because lots more "probes" are 
available.  You can directly probe things like the charge in a 
capacitor, the incremental capacitance of a junction, the 
transconductance of a transistor.

The biggest problem I ran into is that many students can only 
use a GUI.  They can't even type "ls" to get a file listing.  
For them any command line is incredibly confusing.  They need a 
few weeks of lessons in how to use a real computer first.

Many profs respond to this by only using a GUI, which at best 
only delays the awakening, at worst they never learn.



If you are just starting with gnucap, get the stable release 
(0.35).  When you have a need for something it doesn't do, the 
development version might do it.  There is a big difference 
between the latest stable version and the latest development 
version.

If you want to help us make a Free simulator that competes 
against the big bucks simulators, get the development version 
and dive in.





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