How to simulate a coil in ansoft maxwell 3d?

  1. Hello,

    I want to simulate a coil of a transformer in Maxwell (3D), but my problem is: the coil design is not a closed path. Imagine that you draw a ring and then you cut part of it. You get something like a "letter C". In these two terminals I`m gonna put an excitation (100A, for instance), but it looks like maxwell do not accept a open path, just a closed one. Can anyone help me with this problem?

  2. jcsd
  3. gerbi

    gerbi 163
    Gold Member

    Seems like the general concept is fine (winding as C-shaped element). Please double chceck your setup:
    1) materials (conductive for winding, non conductive for environment),
    2) terminals (input=-output, check value and vector direction),
    3) region (I mean box containing experiment. All terminals must be on outer surface of region box).

    If this won't help, post maxwell error message regarding this problem or a screenshot maybe.
    1 person likes this.
  4. Hi, thanks for your help.

    I tried to import an external circuit and worked fine, but only the current density(J) is weird. I'm gonna put a screenshot here as soon as possible, but first: you said "2) terminals (input=-output, check value and vector direction)". what do you mean?

    3D in Maxwell is something very new to me, sorry about the silly questions,

  5. gerbi

    gerbi 163
    Gold Member

    Current density in not uniform ? Have extreme values on bottom and top ? This is due to fact, that you selected "solid" as conductor type (meaning winding is a solid piece of copper). Set this to stranded (a lot of thin wires) and it will be fine.

    What I mean is Kirchoffs Current Law - what goes in must go out. You set a current by three parameters : value of current, it's sign (+ or -) and vector orientation of terminal. Maxwell won't work if KCL requirements are not met (but this is not your problem here, I think).
  6. If I am interested in the Magnetostatic Solution? Can I simute this coil of the picture?

    This coil is not a closed-path. It's an open one. How do I attach the excitations here?


    Attached Files:

  7. gerbi

    gerbi 163
    Gold Member

    Like I said: create a region. It must border two faces of your coil (where it is cut in half). Do this in region creation menu. Select: Z+/-200%; Y+/-50%; X+0%,-100%.
    Next step: assign excitation to two faces of coil (in Y-Z plane).
    1 person likes this.
  8. Why it must border the two faces? If I don't put my region box "bordering" these two faces I can not have a correct simulation? I'd put the excitation in the middle of the path of this coil, is it wrong? In the middle I say something like that: I assigned the excitation in a section of the coil, without any kind of terminal... I think Maxwell help says it isn't necessary to create any kind of it...

    What do yout think about my solution? I couldn't do this way with these terminals(look at the picture to see the problems), can you do that in Maxwell? If so, can you just create a coil like that, assign the excitations and send me, please?

    Thanks, You are helping a lot!

    Best Regards.

    Attached Files:

    Last edited: Apr 8, 2014
  9. gerbi

    gerbi 163
    Gold Member

    Why terminals must border "region" object: because place where current leaves your coil is some kind of boundary condition, don't you think ? You assume symmetry of your model there (no matter if you assign terminal there or not).

    Your approach to assign excitation to section of coil is correct (quite problematic in some cases but ok). But still, "region" should border your coil ends. Mazwell automatically assigns boundary conditions (magnetic field tangent), so please rethink using zero tangential field boundary. Please search for "Frequency Domain (Eddy Current) Solver" in Help.

    One more thing, is your coil a conductor ? I mean, did you assigned a conducting material to it ? This error (".. is not on any conduction path") is one of typical errors when you forget about this.
  10. In this case the problem is that Maxwell says there is a "overlap" when I assign excitation in the same faces that are bordering the region...Look, my really challenge is to simulate a Double Pancake coil. I've drawn this part of toroid just to find a method to simulate a coil without closed path. As you may know, the design of a double pancake coil doesn't have a closed path...That's why I'm here, trying to find this solution. Do you have any suggestion for this task? In Maxwell 2D is okay, but in 3D is really difficult for me. Thanks!
  11. gerbi

    gerbi 163
    Gold Member

    I've said already, rethink using boundary conditions. Do not assign them to faces where your excitations are. Excitations overlap ? Then you are doing something wrong (coils should not touch each other).
    It can be done for coils without closed path. Draw coils (check if they do not ovelap). Coils must border "region" box (in places where they are cut in half). Assign proper materials. Assign excitation (no matter, one per coil in it's middle section or two per coil at their ends). Do not use boundary conditions. Simulate.
  12. Accuracy in maxwell!!!

    Thanks again for all your help. Do you have any information about the accuracy in Maxwell? For example, I have a problem to solve, and I attach some mesh operations, number of steps, % error..etc...How do I know if this configuration is good enough to be accurate?? FOR EXAMPLE, I think the VERY IMPORTANT things are the *variance* and the *order of magnitude* if I simulate a coil with 1 turn, 2 turns, 3...Am I right? It doesn't matter so much if I find 8.2E-2 Testa, or 8.7E-2 or 9E-2...Right?

    Do you have a pdf to send me, or any tips about it?

    (if so,

  13. gerbi

    gerbi 163
    Gold Member

    FEM will be always inaccurate to some degree. For static solvers you can open Maxwell/Results/SolutionData. Here you can observe accuracy of your simulation in terms of energy error (Convergence tab). In most cases my target to achieve is to get below 0,1 % of energy error - I assume it means simulation is accurate. In some coarse simulations I accept energy error between 0,5-1%. Naturally, this is my view resulting from experience (it's hard to obtain scientific explanations on how to determine accuracy). In addition, always watch your mesh. Look for ill shaped elements and increase meshing there.

    Regarding your example: like you said, it is important to watch parameters e.g. magnitude and variance. On the other hand, it's hard to say if 8.2E-2 Tesla is big or small variance. It all depends on design. Only way to judge if design is fine or not is by your expierence.
  14. What do you mean with ill shaped elements? I just try to find very small elements in the mesh...Is that?

    Man, thanks again. You always solve my doubts about it.

    Best Regards
  15. gerbi

    gerbi 163
    Gold Member

    By ill shaped I mean irregular ones (e.g. when lenghts of triangle arms differ more than 2:1 you may consider re-meshing. But again, judgment is based on users experience).

    Typical element is close to a equilateral triangle in terms of shape (for 2D mesh). Presence of highly irregular elements will cause results to be.. not good :)
  16. What do you think about this mesh? It's an HTS tape section. 0.4mm x 0.1mm. In the top, they are smaller...

    Attached Files:

  17. gerbi

    gerbi 163
    Gold Member

    HTS ? Whats that ?
    What kind of phenomena occurs in this element ? (current conduction, eddy currents etc.).

    This mesh ? Generally its ok. You won't achieve perfectly distributet triangles, this is not needed in FEM.

    What is the problem here ? results are not fine ?
  18. High temperatute superconductor. That's what I'm trying to simulate so far...

    I'm still simulating in Magnetostatic Solver. But I intend to do the same simulation in Transient too.

    My question about the mesh was made because sometimes I wanna check if my results are good enough...Sometimes I think "This simulation may be not good and I can be wrong with my conclusions, right?". The results are ok then... This mesh has triangles very weird in the boundary, that's what I doubt... Thanks for your help. If you need anything, tell me. Do you like to trade some Maxwell Simulations any time?
  19. gerbi

    gerbi 163
    Gold Member

    Superconducting material.. I guess that for static simulation this mesh is ok (it is easier to judge from results: if current or losses distribution looks fine then mesh is fine).

    It's smart to check mesh when looking at results. That's good approach.
    Thanks for the offer but I am not allowed to do that.
  20. Not allowed? Do you work for CIA? kidding

    What do you think about this mesh? I think shape of the elements is pretty good, right?


    Attached Files:

Know someone interested in this topic? Share this thead via email, Google+, Twitter, or Facebook

Have something to add?

Draft saved Draft deleted