Hi,
 I'm an electronic engineer and student now student again for my
Master. I have a past experience in periodic Steady State analysis and
 I would like to hear from you about GSoC2008.
I was quite happy about the chance to work on Steady state analysis
again (I'm doing something completely different at the moment), and I
started to think about a would-be plan.
In the past I saw that Gnucap seems well suited for SS analysis in the
Time domain, mainly due the way the electric values at the model nodes
can be accessed.
When I previously studied the Transient Analysis Gnucap source code, I
wondered if  an S.S module could be developed as relatively simple
superset of the transient analysis module plus the Extrapolation
module.
>From there, the transient analysis module could be exploited to
calculate the little transient analysis time frame required to feed
the extrapolation and over, until convergence (hopefully).  Of course
some bits should have to be added to the net list interface too, but
not too many**.
Probably the worst point to overcome would be:
- to force the transient analysis to be computed at the specific times
required by the algorithm. Again, I expect this to be more easier in
Gnucap than how I found to be in Spice, both at command and code
level.
 - to develop a simple, good convergence stopping criterion.
The last one is scary; without one, even good code is probably going
to miss and hit the wall of nightmarish numerical issues.
Some strong points of the project over existing SS simulators could be:
- many of them are in the frequency domain, and are not very well
suited for some specific tasks where a TD simulator would work great.
-  author and source code are available and source code is documented,
so I should be easier to tweak the Transient analysis to assist the
Extrapolation and not the other way around.
- **net list should be backward compatible, and transient analysis
should be possible on it without/with little changes to permit user to
"Brute force" the circuit using  classical Transient Analysis, for
reference and comparison.
The whole project of course rely a lot on the Extrapolation Dr.Davis
would choose, and of course the mileage can vary a lot depending on
this.
My two cents are to restrict the project to the non-autonomous cases.
With a larger timespan, and again depending on the algorithm , the
autonomous case should be doable afterwards with moderate effort.
II'll like to hear the your ideas on this, and on time you think it
would require.
Of course feel free to contact me for any questions.
Regards,

Giuseppe Dia,
Milan


_______________________________________________
geda-dev mailing list
[email protected]
http://www.seul.org/cgi-bin/mailman/listinfo/geda-dev

Reply via email to