[Ktechlab-devel] Grounded ground.

[Alan Grimes]

One of the deeper issues in Ktechlab is grounding.

The issue about the glitiching out to giga-volts, is comparatively
trivial, requiring a better understanding of linear algebra, in that
trying to patch a singular matrix with a tiny value instead of 1 is
bound to make things a lot worse. (the problem being the singular matrix
in the first place...)

Go research an app-note on power-factor correction circuits and try to
implement the basic parts of the topolgy in ktechlab. Most of the
schematics assume that the ground-symbol means "COMOMON" which is
distinct from "Earth", (which is a different symbol).

Worse, Ktechlab inserts hidden earth-grounds where ever it feels like
it. In fact, every line of code in the simulation classes assumes that
these hidden grounds exist. I think if ktechlab were to adopt an "All
voltages add to 0" approach, the simulation matrices it generates will
be much easier to solve.

(the battery circuit in the attached would have a +2.5 volt side and a
-2.5 volt side, equalling 0)

Another major change to the simulation engine I've been thinking about
is the model for reactive components. Right now, ktechlab solves
reactive components seperately from regular components because of the
way the linear algebra system is coded. I think it would work much
better if reactive components such as capacitors and inductors
(including transformers) could request additional state variables in the
 main simulation matrix for charge and magnetic flux, respectively.

An issue arises when you have iron core inductors, in that they saturate
at a certain current... -- similar to how capacitors always have a
finite breakdown voltage. But neither of these issues is yet addressed
anyway.

/me just tried ktechlab 0.3.6. damn that thing gets slow after only a
few minutes... =\

-- 
Opera: Sing it loud! :o(  )>-<
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/ktechlab-devel/attachments/20070531/ab5a3968/attachment.html>