Discussion Forum

Hi

my approach (a generic one) would be to start with a 2D model, test it out carefully and then go to a 3D (if really needed) as your X-Y size wr.t. Z length is really quite excessive I believe

Good luck
Ivar

That the wire has a square cross-section is just a visualization error, I believe. If you mesh it with a fine mesh it will be circular.

If the wire is needed just to generate a magnetic field it is probably best to model the wire using an edge as a current source.

Thanks for the clarification.

I discovered something about modeling infinite elements. Basically, I drew a short cylinder and then changed the subdomain settings under the Physics menu to infinite element of type cylinder.

I am assuming that will help reduce the size of the domain and increase resolution of the mesh.

How do I use an edge as a sinusoidal current source ? I do just need to generate a magnetic field from a wire located at (0,0) and look at its effect on an induction coil with a complicated design (which I have drawn through extrusion) located some distance away laterally.

Hi

my approach (a generic one) would be to start with a 2D model, test it out carefully and then go to a 3D (if really needed) as your X-Y size wr.t. Z length is really quite excessive I believe

Good luck
Ivar

Thanks for your comment.

As I mentioned in my comment, I discovered something regarding infinite elements.

I do need 3D as the induction coil (please see other post) is oriented with variable direction cosines with respect to the wire. I am not sure that can be done using a 2D model.

Thanks for the clarification.

I discovered something about modeling infinite elements. Basically, I drew a short cylinder and then changed the subdomain settings under the Physics menu to infinite element of type cylinder.

I am assuming that will help reduce the size of the domain and increase resolution of the mesh.

How do I use an edge as a sinusoidal current source ? I do just need to generate a magnetic field from a wire located at (0,0) and look at its effect on an induction coil with a complicated design (which I have drawn through extrusion) located some distance away laterally.

I am not very familiar with the infinite elements, but I believe it is used for representing an infinite open domain. Basically, the boundary absorbs incoming waves instead of reflecting them, I think. I do not think you can use this on the wire.

Applying a current in an edge is easy: just replace the wire with a line (or several lines), then in Physics settings go to Edge Settings and select the chosen line(s). There you can specify the desired current. Just make sure it flows in the correct direction. This edge current then generates a magnetic field. It does not flow into or out of the line end points, but just vanishes when the line ends.

Helo

indeed I believe understand better your true final need for 3D,

Anyhow: starting in 2D with a simplified model to get the "physics" validated with a limited number of node is often the quickest approach, and then move to a full 3D with the additional meshing difficulties.

From the follow-up mail, I understand that you have passed this stage already, and that you have managed, with the infinite ements, to reduce your 3D meshing volume outside of your wire to an acceptable value with finer resolution.

And if you are only interested in what is happening outside of the wire you can drop the "inside" and use surface currents (hollow tube on 3D, the hole being the ouside wire surface)

Remains, that if you apply a sinus input current along your wire, you could work in the harmonic regime with the frequency as parameter and not the in the transient domain, no ?
I'm assuming your are in the AC mode with "low" frequencies, and an instantaneous constant current along the full wire length, or have I missed another point ?

Ivar

Helo

indeed I believe understand better your true final need for 3D,

Anyhow: starting in 2D with a simplified model to get the "physics" validated with a limited number of node is often the quickest approach, and then move to a full 3D with the additional meshing difficulties.

From the follow-up mail, I understand that you have passed this stage already, and that you have managed, with the infinite ements, to reduce your 3D meshing volume outside of your wire to an acceptable value with finer resolution.

And if you are only interested in what is happening outside of the wire you can drop the "inside" and use surface currents (hollow tube on 3D, the hole being the ouside wire surface)

Remains, that if you apply a sinus input current along your wire, you could work in the harmonic regime with the frequency as parameter and not the in the transient domain, no ?
I'm assuming your are in the AC mode with "low" frequencies, and an instantaneous constant current along the full wire length, or have I missed another point ?

Ivar

Thanks for your response.

I cannot say that I have even meshed the problem yet but I now have conflicting advice regarding the use of infinite elements.

The situation is as follows.

I have, to all intents and purposes, an infinite straight wire carrying a quasi-static current (frequency ~ 1 kHz) since the component of interest (the inductor) is only about 100 microns wide (at its widest extent). So, the electrical size of the component is about 1e-4/(c*1e-3) << 0.1.

All I am interested in are the induced currents and voltages in the inductor. I am not interested in what happens far away (or on the wire). Martin suggested that I use a line, and that seems simpler than a hollow cylinder with a hole "outside the wire surface" (?).

Martin's message earlier in the thread indicates that I cannot use the infinite elements with the wire (which was the point). You say that I can. Don't blame a newbie for getting confused. My reading from the manual was that it was precisely for situations like this that comsol had added the capacity for infinite elements. I could be mistaken, but I have a little more faith in you :)

As to the boundary conditions, how do I tell comsol to use periodic BCs outside of the domain of interest (about a 0.1 cm or so) ? Do I draw a hollow cylinder with that radius and apply the BCs there ? From my explorations so far, you have to have an object selected to apply any BCs to it. Can't select empty space :)

Working in the harmonic regime is a lot simpler provided I can get Irms and Vrms for the induced currents through a scale factor at the end of it.

What are my (comsol) options in case of a larger electrical size ? (I am not likely to encounter it, just curious.).

Hi Singh,
I think you may find the "Eddy currents in 3D" model interesting:
www.comsol.com/showroom/documentation/model/970/

It calculates the induced current in a plate from a nearby conductor. The wire is modelled as an edge current. Also, there is the use of impedance boundary conditions which perhaps is interesting too.

You mentioned the use of infinite elements; as Ivar said, if you use this it will be to reduce the size of the surrounding domain (air?), not to mesh the wire itself. I do not think it is necessary, though, since with a sufficiently large cylindrical volume I guess it is ok to assume magnetic insulation on the boundaries (like in the "Eddy currents in 3D" model).

Boundary conditions must be specified on the outside surfaces of your model, yes. If you have chosen a time-harmonic mode all BCs are automatically time-harmonic too. E.g. if you specify a magnetic field B, this magnetic field is time-harmonic B*exp(-jwt).

Note that comsol in the time-harmonic mode gives peak values. To obtain rms values simply divide by sqrt(2).
NB! When you specify the wire current, this should be given as peak value too, I=Irms*sqrt(2).

Hi

From your frequency domain I agree I would also start with the ACDC mode.

For me you can as weel trust Martin as myself (even we do not know each other, but from his name I suspect he's Norwegian, just as me, but we are, I believe working from two separate places/countries some 2000 km apart), I beleive the slight differences in the replies are from the different understandings of the problem, we do not have everything, and we fill in the missing with our own thoughts.

First he is proposing, as me, to use surface currents, he proposed as a simple line, I proposed a "hollow" tube, the line is obiously the simplest representation of a "tube". In this way you drop the volume inside the conductor, you gain in free memory for nodes for the doman of interest.

Second I do not read that he is stating that the infinite elements are not working, even for a line.
My understanding and use of infinite elements are, as Martin stated, to extend to infinite in a "short distance". The magnetic fields drop off "slowly", therefore to avoid to have to mesh a large volume of "empy-air" you add a thin infinite element that extends the space variable artificially, the result is that waves, or the field just "dissapears" in there, and any magnetic isolation of the outer edge is not influencing the shape of the field within the volume you are interested in.

Try it out, take i.e. the fist example chapter 2 in the Acdcmodlib document, they are proposing an external ring with an infinite element, run the example, save the image and then turn the infinite element "off" and re-run the simulation, compare the field lines contour plot, you clearly see the influence of the outer edge boundary condition.

For surface currents, you have an example in the following-on model (same book).

For the rms values of I, V, the harmonic development assumes I=I0*sin(2*pi*frequency*t) for the source and similarly the "receiver"(but with a different I0 often function of frequency), by default frequency is 50 Hz, but you can change it or better use it as a Parameter for a parametric sweep, so rms is just the integral of the sinus^2, you do not really need Comsol for that, but you can if you want to, however, Comsol is handy to get the I0 for complex cases and with unexpected transfer functions.

Forget for the time bein the DC motors examples, theay run in the time domin because the parts are moving in space, I understand yours are not, or wery slowly (?) compared to the 1 KHz, but jump to page 111 (doc V3.5a) and run trough the "induction current examples" or better the one at p130.

If I undestand you correctly, your inductor is on the "side" of the wire and not "around" as on the axisymmetric example of p130. This makes the conversion from the example to your case a bit trickier, and I agree, for a full representation requires 3D, you can only estimate 2 particular cases in 2D (coil axis perpendicular to the 2D plane, and in line with the 2D plane, am I correct ?.

Good luck
Ivar

Hello Martin,

Thanks for your response.

I am writing after a few days as I got very busy with something else.

Thanks for clarifying that about infinite elements. My understanding of the meshing process was that you have to have an object selected to actually mesh anything. However, to be able to plot the field in the environs of the inductor, I am guessing that comsol does some sort of internal meshing as well (at least that part is not user-driven as one cannot select empty space). Thus, infinite elements then permits one to truncate the region around an object of interest, if I am understanding you correctly, to save some memory for the solution.

I have looked through the eddy current link and studied the first document. My access to the remaining documents is complicated by the fact that I access comsol through a shared computer I do not have superuser rights to (and hence cannot upload any license file).

I am sure I will be able to figure something out from what I have read.

Hello Ivar,

Thanks for your response.

I just looked at the AC/DC modlib document. It is a treasure trove of information. Let me go through it and see if I still have any questions.

Applying a current in an edge is easy: just replace the wire with a line (or several lines), then in Physics settings go to Edge Settings and select the chosen line(s). There you can specify the desired current. Just make sure it flows in the correct direction. This edge current then generates a magnetic field. It does not flow into or out of the line end points, but just vanishes when the line ends.

I replaced the wire with a line. In Physics > Edge Settings, when I select the chosen line, the I0 specification is greyed out.

Maybe my mode selection is wrong ?

Hi

From your frequency domain I agree I would also start with the ACDC mode.

For me you can as weel trust Martin as myself (even we do not know each other, but from his name I suspect he's Norwegian, just as me, but we are, I believe working from two separate places/countries some 2000 km apart), I beleive the slight differences in the replies are from the different understandings of the problem, we do not have everything, and we fill in the missing with our own thoughts.

First he is proposing, as me, to use surface currents, he proposed as a simple line, I proposed a "hollow" tube, the line is obiously the simplest representation of a "tube". In this way you drop the volume inside the conductor, you gain in free memory for nodes for the doman of interest.

Second I do not read that he is stating that the infinite elements are not working, even for a line.
My understanding and use of infinite elements are, as Martin stated, to extend to infinite in a "short distance". The magnetic fields drop off "slowly", therefore to avoid to have to mesh a large volume of "empy-air" you add a thin infinite element that extends the space variable artificially, the result is that waves, or the field just "dissapears" in there, and any magnetic isolation of the outer edge is not influencing the shape of the field within the volume you are interested in.

Try it out, take i.e. the fist example chapter 2 in the Acdcmodlib document, they are proposing an external ring with an infinite element, run the example, save the image and then turn the infinite element "off" and re-run the simulation, compare the field lines contour plot, you clearly see the influence of the outer edge boundary condition.

For surface currents, you have an example in the following-on model (same book).

For the rms values of I, V, the harmonic development assumes I=I0*sin(2*pi*frequency*t) for the source and similarly the "receiver"(but with a different I0 often function of frequency), by default frequency is 50 Hz, but you can change it or better use it as a Parameter for a parametric sweep, so rms is just the integral of the sinus^2, you do not really need Comsol for that, but you can if you want to, however, Comsol is handy to get the I0 for complex cases and with unexpected transfer functions.

Forget for the time bein the DC motors examples, theay run in the time domin because the parts are moving in space, I understand yours are not, or wery slowly (?) compared to the 1 KHz, but jump to page 111 (doc V3.5a) and run trough the "induction current examples" or better the one at p130.

If I undestand you correctly, your inductor is on the "side" of the wire and not "around" as on the axisymmetric example of p130. This makes the conversion from the example to your case a bit trickier, and I agree, for a full representation requires 3D, you can only estimate 2 particular cases in 2D (coil axis perpendicular to the 2D plane, and in line with the 2D plane, am I correct ?.

Good luck
Ivar