Details

Ansys Maxwell: Coil Models pt 2 - Inductance and Magnetic Field

Hello everybody, Ian from Ozen Engineering here. Welcome back to the series on coil models in Maxwell. In the last video, we built the geometry that we'll be testing, and in this one, we'll be running the simulations and making comparisons.

Mesh Rule Creation

Jumping right in, I'll start by creating a mesh rule. I'll make it a length-based rule on the surface of the coil. I like to take the default value and divide it by a variable. This allows me to easily refine the mesh further if I want to.

Simulation Setup

Right-click on parameters and click on assign, then matrix.
Check the box for our winding and then click OK. This tells Maxwell to include the winding in the solution matrix, which is what we'll use to check the inductance.
Right-click on analysis and add a solution setup with the default settings.

Mesh Refinement

Before analyzing, I want to check the mesh. The simplified model has a mesh that looks way too coarse. So I'm going to set that mesh refinement variable that we made earlier to a higher value. Let's start with 2.5. Double-check the properties box of your mesh rule. You want to make sure that the units don't change. Sometimes that can happen when you use a variable in the definition. After regenerating the mesh, I can see that it looks much better. And I'm satisfied with the default settings of the other coil. Time to analyze.

Results and Comparisons

Taking a look at the results, I'm not too happy with how they line up. The simplified model is giving me an inductance of 2.89 microhenries, and the full model gave me an inductance of 3.05 microhenries, which is just over a 5% difference. My immediate thought is that the two leads that we added are adding to the inductance. So I'm going to try cutting those off and connecting the two ends.

Design Adjustments

I'm still worried that this connecting line will throw off the comparison. So I'll make just one more design. My plan is to use symmetry to avoid having to create any leads at all. I'll start by cutting the model in half along the YZ plane. Now I can add the ports. If you remember from the last video, coil terminals require a closed conduction path. Because our wires are separated, we'll need to add a coil terminal to each wire. I can add ports that face into the wires on this side and out of the wires on this side, creating a closed conduction path for each wire. But this would mean creating two lines, two coil terminals per wire. To reduce the number of terminals I need to make, I'm going to slice the coil along the XZ plane and add ports in the middle of the coil. Each of these needs to be assigned to the same winding. After finishing the setup, I'll simulate the new designs.

Simulation Time and Mesh Elements

Time to finally start making comparisons. Taking a look at the simulation time, the simplified geometry was the fastest by a wide margin, solving in 30 seconds. The design that we split in half solved in 4.5 minutes. The extended leads took just over 10 minutes, and the design where I connected the leads took 12.5 minutes. This order is to be expected when you consider the number of mesh elements for each design:

The simplified design had around 35,000 mesh elements.
The design with the connected leads had over 400,000 mesh elements.

It had just over 11 times the number of elements and took just over 11 times as long to solve. The final thing to note here is the simple coil solved in two passes, whereas the other coils solved in three.

Inductance and Loss

Moving on to the actual results, everything is in the same ballpark, which is good. Because we used the magnetostatic solver, I can't set a symmetry multiplier like we can with the eddy current solver. That just means that we'll have to double our values for inductance and loss. I'll double my matrix with some post-processing I set up earlier. To see if there's a difference, I made a copy of this design and solved it with the eddy current solver, which has an option to set a symmetry multiplier. And the inductance and the loss results agree. So I'm comfortable with this value of 2.91 microhenries.

Final Adjustments and Observations

Worth noting is the simplified coil has the smallest inductance out of all the coils. It's not surprising when you compare it to the coils that have the lead. But the coil without them should match up even closer than this. So as a final check, I took a closer look at the height of my approximator coil and tried to get it to match even closer. This put the inductance at 2.9081 microhenries, which I finally think is close enough to match. This last test shows you how careful you need to be when entering the height of your simplified geometry.

Conclusions

On to the solutions. We discussed the inductance and why the other coils have higher values. This is something that you'll want to consider. In many applications, you need to know about that extra inductance that the leads add. In many other applications, it doesn't make a big enough difference to be worth sacrificing simulation speed. Let's only take a quick look at the losses. The simplified geometry stands out as having a much lower value compared to the other coils. It's given a stranded loss, but the other three coils are given a solid loss. We're going to take a closer look at this in the eddy current video, so make sure that you're subscribed so you don't miss it.

Field Data

Now let's look at some field data. Plotting the field on a line through each coil shows that the maximum occurs at the same place. But the simplified geometry is still 30 to 40 microTesla lower than the other three. Looking at this in the vector plot, we see the same thing. Overall, my observations from this video are that the simplified geometry's matrix solution had high accuracy, but so far the field data has not been as accurate. I could have slightly adjusted the height of the simplified geometry to make the inductance match better, but we still would have had issues with the loss being off and the magnetic field not being accurate enough.

Next Steps

Still, I think we're on a path to really understanding how to use these things in the best way possible. Next up, we take a look at some transient simulations and see what else we can learn. I hope to see you there. Until then, this is Ian from Ozen Engineering. Take care.

Transcript

[This was auto-generated. There may be mispellings.]

Hello everybody, Ian from Ozen Engineering here. Welcome back to the series on coil models in Ansys Maxwell. Last video we built the geometry that we'll be testing and in this one we'll be running the simulations and making comparisons. Jumping right in, I'll start by creating a mesh rule.

I'll make it a length based rule on the surface of the coil. I like to take the default value and divide it by a variable. This allows me to easily refine the mesh further if I want to. Next up, we want to right click on parameters and click on assign and then matrix.

Check the box for our winding and then click ok. This tells Maxwell to include the winding in the solution matrix which is what we'll use to check the inductance. Right click on analysis and add a solution setup with the default settings. Before analyzing, I want to check the mesh.

Right click on your analysis and then click on the mesh. After checking the mesh, I will set the mesh refinement variable that we made earlier to a higher value if it looks too coarse. Let's start with 2. 5. Double check the properties box of your mesh rule.

You want to make sure that the units don't change. Sometimes that can happen when you use a variable in the definition. After regenerating the mesh, I can see that it looks much better. And I'm satisfied with the default settings of the other coil. Time to analyze.

Taking a look at the results, I'm not too happy with how they line up. The simplified model is giving me an inductance of 2.89 microhenries. And the full model gave me an inductance of 3.05 microhenries. Which is just over a 5% difference.

My immediate thought is that the two leads that we added are adding to the inductance. So I'm going to try cutting those off and connecting the two ends. I'm still worried that this connecting line will throw off the comparison. So I'll make just one more design.

My plan is to use symmetry to avoid having to create any leads at all. I'll start by cutting the model in half along the YZ plane. Now I can add the ports. If you remember from the last video, coil terminals require a closed conduction path.

Because our wires are separated, we'll need to add a coil terminal to each wire. I can add ports that face into the wires on this side and out of the wires on this side. Creating a closed conduction path for each wire. But this would mean creating two lines. Two coil terminals per wire.

To reduce the number of terminals I need to make, I'm going to slice the coil along the XZ plane. And add ports in the middle of the coil. Each of these need to be assigned to the same winding. After finishing the setup, I'll simulate the new designs. Time to finally start making comparisons.

Taking a look at the simulation time, the simplified geometry was the fastest by a wide margin. Solving in 30 seconds. The design that we split in half solved in 4.5 minutes. The extended leads took just over 10 minutes. And the design where I connected the leads took 12.5 minutes.

This order is to be expected when you consider the number of mesh elements for each design. The simplified design had around 35,000 mesh elements. And the design with the connected leads had over 400, 000. It had just over 11 times the number of elements and took just over 11 times as long to solve.

The final thing to note here is the simple coil solved in two passes. Whereas the other coils solved in three. Moving on to the actual results, everything is in the same ballpark, which is good.

Because we used the magnetostatic solver, I can't set a symmetry multiplier like we can with the eddycurrent solver. That just means that we'll have to double our values for inductance and loss. I'll double my matrix with some post-processing I set up earlier.

To see if there's a difference, I made a copy of this design and solved it with the eddycurrent solver. Which has an option to set a symmetry multiplier. And I'll show you how to do that in a second. And the inductance and the loss results agree.

So I'm comfortable with this value of 2.91 microhenries. Worth noting is the simplified coil has the smallest inductance out of all the coils. It's not surprising when you compare it to the coils that have the lead. But the coil without them should match up even closer than this.

So as a final check, I took a closer look at the height of my approximator coil and I tried to get it to match even closer. This put the inductance at 2.91 microhenries. And I'm happy with that. It's not surprising that the coil is 2.9081 microhenries. Which I finally think is close enough to match.

This last test shows you how careful you need to be when entering the height of your simplified geometry. On to the solutions. We discussed the inductance and why the other coils have higher values. This is something that you'll want to consider.

In many applications, you need to know about that extra inductance that the leads add. In many other applications, it doesn't make a big enough difference to be worth it. But I'm happy with that. I'm happy with the result. And I'm happy with sacrificing simulation speed.

Let's only take a quick look at the losses. The simplified geometry stands out as having a much lower value compared to the other coils. It's given a stranded loss, but the other three coils are given a solid loss. We're going to take a closer look at this in the eddy current video.

So make sure that you're subscribed so you don't miss it. Now let's look at some field data. Plotting the field on a line through each coil shows that the maximum occurs at the same place. But the simplified geometry is still 30 to 40 microTesla lower than the other three.

Looking at this in the vector plot, we see the same thing. Overall, my observations from this video are that the simplified geometry's matrix solution had high accuracy, but so far the field data has not been as accurate.

I could have slightly adjusted the height of the simplified geometry to make the inductance match better, but we still would have had issues with the loss being off and the magnetic field not being accurate enough. So I'm happy with that.

But still, I think we're on a path to really understanding how to use these things in the best way possible. Next up, we take a look at some transient simulations and see what else we can learn. I hope to see you there. Until then, this is Ian from Ozen Engineering. Take care. Bye. Bye. Bye. Bye.

Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye.

Bye. Bye. Bye. Bye. By

View on YouTube