Re: [ng-spice-devel] Patches


To ng-spice-devel@ieee.ing.uniroma1.it
From Alan Gillespie <alan.gillespie@analog.com>
Date Mon, 02 Oct 2000 14:09:06 -0400
Delivered-To mailing list ng-spice-devel@ieee.ing.uniroma1.it
Mailing-List contact ng-spice-devel-help@ieee.ing.uniroma1.it; run by ezmlm
References <Pine.LNX.3.96.1001002153608.24198B-100000@ieee.ing.uniroma1.it >
Reply-To ng-spice-devel@ieee.ing.uniroma1.it
Sender agillesp@epc.co.uk

Talking about giving you a lot of rubbish, I'd be interested in
feedback about whether anyone's noticed a significant
improvement or not. I think the mos3load.c glitch means
that the current CVS is probably a step back, but once the
latest fix is incorporated we should be up and running.

I notice that the make check thing still fails two out of three
tests (at least on my version) and I've a feeling I know
about one of the tests that fail. Has anyone looked at that ?
Is it fixed now ?

A number of my fixes will only make a difference if you're
simulating at a temperature other than nominal. Those
are fixes of junction capacitance models in a bunch of places.

The gmin fixes in MOS devices and maybe other places too,
improved convergence a lot for me, especially in circuits
which start off disabled.

The MOS3 gm/gds fixes mean that you can now trust the
results of an AC analysis of a MOS3 circuit. Previously
you'd get a lot more gain from the simulation than from
real life.

The "dynamic" gmin-stepping and source-stepping fixes
are "no-brainer" additions to what's already in Spice 3,
but after seeing PSPICE successfully converging after
stepping up the sources by tiny amounts at critical levels,
I decided that they can't do any harm, and indeed I think
they do catch all but the most temperamental circuits.

But if anyone notices something that doesn't work now
when it used to before, please let me know. I believe I
did find a circuit that previously converged when it
shouldn't have, but usually if a circuit stops converging
then it's the fixes that are at fault.

Cheers,

Alan

p.s. Paolo, any chance of getting my name on the web
page now ? I've told a couple of people that I've been
involved with ng-spice, but they'll be more convinced
when they see some evidence ;-)


Partial thread listing: