RE: [ng-spice-devel] convergence
> Al> 3. That's all!!
>
> Yes, that's how it appears to me also.
There's also that thing Jon mentioned about switches
changing state. That CKTnoncon flag could be incremented
anywhere, so unless you know all the code, it's difficult
to tell what's checked. (and what was intended to be checked).
>
> Sounds like ACS doesn't directly check KCL either. True?
>
What exactly do you mean by a KCL check ? I can only think
of two possibilities :-
A) Check that all the currents in each devices add up to zero
B) Check that all the currents into each node add up to zero.
I'm no expert, but from what I saw of HSPICE, I got the
impression that the KCLcheck option did A, but I can't be
sure. I regularly saw "convergances" in Hspice, where it
seemed that currents were not appearing out of the terminals
of some devices. Whether that was because the code for the
device just wasn't coming up with the right currents, or
whether the matrix wasn't adding it in, I couldn't tell.
The KCL check stopped that, but it also stopped it converging
a lot of the time. (But I could even make HSPICE stop converging
by moving a capacitor up one line in the "deck").
But surely B must be satisfied if the nodes haven't changed
since the last iteration. If the matrix stuff is correct, surely
an imbalance of current would move the node voltage in the
solution. Or are we worrying about currents going into very low
impedance nodes.
That's a point. Are we saying that device currents aren't
checked, only inductor and voltage source currents kept in
the matrix ?
Alan
Partial thread listing: