Re: [ng-spice-devel] convergence


To ng-spice-devel@ieee.ing.uniroma1.it
From Al Davis <aldavis@ieee.org>
Date Fri, 27 Apr 2001 14:56:48 -0700
Delivered-To mailing list ng-spice-devel@ieee.ing.uniroma1.it
In-Reply-To <15081.54410.217878.772132@pearls >
Mailing-List contact ng-spice-devel-help@ieee.ing.uniroma1.it; run by ezmlm
References <CF654D964573D311A1CA0090278A36FF4CBE4C@EDIN_EXM1 > <01042712572007.10628@hobbes > <15081.54410.217878.772132@pearls >
Reply-To ng-spice-devel@ieee.ing.uniroma1.it

On Fri, 27 Apr 2001, Steve Hamm wrote:
> The MNA equations are I(Y1) + I(V) = 0 and V(1) - V = 0.

To do this, you need to solve F(x) again.  This is an expensive 
operation.  That was my point.  How do you get it to be cheap with 
the proper formulation?  CPU time is usually dominated by model 
evaluation.

If the inputs to the model have not changed significantly (verified) 
then even after trying it again, the outputs and derivatives have not 
changed significantly, wouldn't you expect to get the same result in 
a KCL check as it got when it solved the original system?

I suppose it checks that you have set reltol and abstol reasonably, 
but maybe not.  Due to roundoff, it won't match exactly.  With 
iteration, you don't get an exact solution.  You get it to some 
tolerance.  So, one would expect that KCL would give you within some 
tolerance of zero, as opposed to zero itself.  What tolerance?  
reltol and abstol????  How is this different from comparing 
successive iterations?

Partial thread listing: