Discussion Forum

Hi

But are you sure what you demand is "physical" ? in MEF you solve for scalar V and vector A. Normally with magnetic fields you need to loop your current too, to get the physics correct. You cannot have spontaneous apparition and removal of voltage or current in the middle of your model, as for ES or EC where you ignore the vector A.

Or am I wrong ?

--
Good luck
Ivar

Hi again,

You might be right; but suppose I have a device providing my current, and it is sitting inside the system. It is fed with DC from outside, yes, but with it it provides me AC or anything else; I don't want to have to build the device itself in COMSOL, it's complicated and I don't know how it generates the current; but I do want the current it is generating, and I need it to physically be in a place where it is inside the cylinder of air in which I want my results..

Also, another idea perhaps, I don't know if that's easy to create but it's worth to offer - a 'test geometry' node, for which you feed the domains/objects you want to test from the geometry you have. it's like a shortcut for disabling alot of nodes, running and enabling; especially if the disabled ones are stuff like revolve or extrude, in which you need to re-create them after enabling.

Shoval

Hi

for the second point I agree, having means to group the sub-nodes with dummy title nodes could be handy, but as nodes position (top-down) is important you cannot do all sorts of grouping. I'm not COMSOL so I cannot do anything, but the developers are reading our exchanges so they might come with a hint.

I believe that if you want to solve for MEF (full vector A & scalar V) you need to be coherent with the physics, so there must be other ways to correctly get what you want

--
Good luck
Ivar

I have submitted requests to support before but I will state mine here:

1) Smith chart for the RF module - Such a tragedy it does not have one already.

2) Live cursors on 1D plots or anything with with xy axes of information. Why would any plot not have "live" data that you can read off directly.

3) Add a K coupling statement to the spice interface - do you realize how hard this makes it to build a transformer w/o it? It can be done but that requires knowing the ODEs/PDEs for the circuit. The s/w is supposed to make my life easier, not harder. If LT Switchercad can do it for free...

4) Geometry snapping in 3D. I don't know how many times I cannot get geometry objects to line up exactly and then only be off by fractional amounts (and then it won't mesh and simulate). I think having a way to snap faces together would be great.

I have to say that for point 4 of your post, I totally support and encourage that. more than once it has occured to me and not always I managed to work around it. For example, sweeping over pi/2 radians not always gets you the two end surfaces parallel to the axises.

I forgot to mention that I use 4.2, in case that matters.

Implementing GPU computation through CUDA or openGL.
GPU computation has way surpassed the CPU counterpart. Users like me without high end workstations or cluster access would save lot more time by implementing these feature.

a timer just to see how long the program has been computing could be good, or at least a timestamp on the log file.

Camera path for 3D animation; like click with the mouse and moving a little but; much more impressive!

Ralf