Discussion Forum

Hi

do you mind push over the (or better a simplified) model just so we can speak about the same issue.
V4.1 OK which patch 4.1.0.154 ? which physics MEF, ES, EC ... OK solver is transient ...
Not obvious for us out here to understand ;)
--
Ivar

Sorry for the missing Information.

My Version is 4.1.0.112
I attached a small Example Model. I hope you will understand what I tried to do there:

Pass a Voltage between Border 1 and 2, 2 and 3 and 3 and 1 while the conductancy is changing over time.
In a later step I want to plot the conductancy as a lumped parameter over time. (not in the Examplemodel yet)

I thing I got something wrong about the identity pairs. Probably just some general missunderstanding how to use these.

Hi

first of all you are in 2D-axi do you have a block or a torus ? (because you are simulating a torus)

Indeed I do not believe the identity pairs was foreseen to be used in this way ? Does not seem natural for me, anyhow to start with put already an epsilon_r to 1 or whatever is correct, I doubt that "0" is what you want, anyhow you will have many persons out here interesting in a material with epsilon_r really = 0.

Then you need to define by hand the three cases (in a future release I have heard you will be able to link different solvers to different BC enable/disable but not yet). By the way, I can only advise you to update to latest patch for v4.1, see COMSOL's main site

Set a port/terminal with a voltage on one boundary and add a "gnd" on the other + solve it works quite OK

Then finally if you have a cube/square and apply a voltage to one side and ground the lateral side, then the "point"/edge at the frontier is rather ill defined, floating somewhere between V and GND. This means that your solution will have a singularity in this corner and that the impedance will be mainly driven by this singularity. Easily checked if you go from a mesh of "normal" to "extremely fine" => an impedance (ec.V0_1/ec.I0_1) change of about 25-30% , from some 35.8 Ohm to 25 Ohms!
For an V electrode and GND opposite to each other you have R=100 Ohms, as expected, all for t=0 and 50% thereof at t=10sec, also as expected no ?

My model is a v4.1.0.154
--
Good luck
Ivar

Yes, I was aware that its a torus. The numeric problem in the one corner doesnt apply for my real model anyway. I just wantet a very simple model to show you my problem.

My Problem is, that in my real Model, I ve got 9 cases, and I would like to be able to make adjustments to different parameters later on without having to change the whole Model nine times to get the solving. Is there any other way to do this, even if portsweeping is not the right way to do this?

Hi

OK my model was 2D, but the comments remain.

Portsweep might be possible, it's just that I havent used it so I do not know exactly how it works ;)
BUT I do not belive the identity pairs should be used how you tried them out

Hopefully there are others (or support) that can give more advice

--
Good luck
Ivar

But thanks anyway. I will try a different approach.

If anyone else has got an idea about how to use port sweeping in my example, please let me know.