## Parallelized Linear DC Analysis document.SUBSCRIPTION_OPTIONS = { "thing": "thread", "subscribed": false, "url": "subscribe", "icon": { "css": "fa fa-envelope-o" } };

srean
2012-04-25
2012-12-15
• srean - 2012-04-25

Hi,

I need to analyze a million node circuit of simple resistors and voltage sources. I am looking for something that will use multiple available cores. I need to solve for the currents several times (in a loop) where the value of the resistances change but not the topology. Should I be using qucs for this or are there other tools which are more appropriate.

Not educated as an electrical engineer, as you can see.

Thanks a lot for helping a newbie out

-srean

• magneto98 - 2012-04-25

Hi srean,

I guess, Qucs only uses one core. The problem would be putting all your data in a qucs schematic. Depends on the format of your data
Alternative to qucs:
If you only have resistors and voltage sources, you have to solve a simple system of linear equations. You could write down the equations (using Kirchhoff's circuit laws). You end up with
U = R*I (U, R, I being matrices)
To solve for the currents you have to invert R to get
R^-1*U = I

Then solve it with Matlab/Octave or something else.

Carsten

• srean - 2012-04-25

Hi Carsten thanks for the reply.

Yeah I understand that in principle. I was wondering if there was any tool that can form the linear equations to solve directly from some text or data structural representation of the circuit. I am interested in the effective resistance between all resistances too (this too is just a matrix pseudoinverse).

My use case does really need multicore support, but thanks for being helpful.