Re: [ng-spice] Predictor code


To ng-spice@ieee.ing.uniroma1.it
From Michele Quarantelli <quarante@legs.ing.unibs.it>
Date Mon, 30 Aug 1999 17:25:58 +0200
Delivered-To mailing list ng-spice@ieee.ing.uniroma1.it
In-Reply-To Your message of "Mon, 30 Aug 1999 14:53:08 +0200." <Pine.LNX.3.96.990830145100.8066A-100000@ieee.ing.uniroma1.it >
Mailing-List contact ng-spice-help@ieee.ing.uniroma1.it; run by ezmlm
Reply-To ng-spice@ieee.ing.uniroma1.it

> Michele,
> 
> Can you tell me why CAD Group decided to take out from the compilation so
> many options or enhancements: bypess, predictor, sense2, etc. ?
> 
> Paolo
To be honest I have no idea, but most probably the code was not reliable. I'm
going to test the different options and I will report results.

I can only tell about sense2 (at least a good guess): the transient 
sensitivity algorithm was developed as part of a student project and it 
didn't 
find a sequel since it was not very efficient in the multi-parameter case 
(which today
is "the" case). May the other cases suffered the same story...

BTW:We are currently developing a new efficient algorithm for multi-parameter 
time domain sensistivity and we it have partially implemented in spice3f4. As 
soon as final tests will make good sense we will provided you the code.

Michele




Partial thread listing: