Discussion Forum

I found an online inductance calculator and plugged in what I thought were reasonable values for your parameters, and it estimated L to be less than 1.5 mH. So, based on that quick check, I'm not convinced that the Comsol model is wrong, if working with your inputs to it. Is there any chance your coil actually has a larger number of turns? That's just a guess.

Hi Robert,

This is definitly a good thought! That said, I'm fairly confident that the experimental inductance meausrment is (more or less) correct. That's becuase (A) I have a manufacturer specification, and (B) I have two independent measurements that agree really closely with the specification.

I definitely could be a little off with counting the turns, but I'm pretty confident within 20%. In order to force console to agree with the manufacturer specification, I would need to more than double the number of turns.

Any chance you could send me the online calculator you used? I'd love to check it out.

See https://electronbunker.ca/eb/InductanceCalcML.html, among others.

Ok, this is helpful. As you can see from the attached screenshot, I also find the theoretical coil inductance to be quite close to what COMSOL predicts.

Now I have matching theory and simulation, and I separatly have three matching experimental measurments. But the experimental measurments are all much higher than theory/simulation. I can think of several reasons (e.g., self capacitance) that would cause the experimental measurments to be too LOW, but too HIGH is weirder. I recongize that this may be more of a physics question than a COMSOL question, but if anyone has other ideas to think about, I'm all ears!

Care to show us a photo of your real-world coil? How do you know the number of turns? Did the manufacturer tell you how many, did they put that in their datasheet, or did you unwind the coil and count them yourself? Also, might there be any non-air but somewhat magnetic material in the core? Finally, are the turns wound around the symmetry axis (like a cylindrical spool of thread, as the model assumes, and which makes this a "short length solenoid" coil) or are they actually wound perpendicular to that, i.e., radially looping in/and out of the hole in the ring, and making this a toroidal coil?

I can offer one more option. Use a circular coil instead of numeric and select inner and outer circular edges to define the wire path and you get 5.2 mH.

Good call. Here are some photos with a cm/mm scale.

Ok, I'll admit that I have NOT unwound the coil in order to count turns. I am just assuming it's uniformly packed throughout. There is definetly nothing in the core, but could there be some sort of anular magnetic material hiding within the coils? Yes, I suppose so... Since the calculator you linked very helpfully gives the total wire length, I'll check that against the resistance of my coil to see if things check out and report back.

Are the photos helpful, or should I take other angles?

No, the resistance per unit length more or less checks out for a 30 m-long piece of 20 AWG wire. Perhaps the photos above will give some other clue. Thank you for your time!

Edgar--

Wow! Indeed you do. Would you be able to elaborate on the differences in what COMSOL assumes for the numeric and circular coil types? In general I am quite confused on the proper use of the latter. Are you supposed to always select both the outside and inside edge on the top and the bottom?

Ben,

I couldn't tell more than the documentation does. As far as I understand things, the circular coil geometry analysis needs a path that corresponds to the average path length in the coil, and selecting inner AND outer edges accounts for that. If you only select an outer edge, the inductance gets too high. It is all in all pretty confusing and what I learn from your case is that coils need to be treated with great attention. And as always: measurement is gold, and I don't see any flaws in your measurements, using an LCR meter and determining resonance. I presume you weren't using huge loops of lead wires to connect things in the lab.

Cheers Edgar

I played a little more with your model and I honestly do not understand the discrepancy between the circular and the numeric coil. I encourage you to submit the case to the COMSOL support. It would be nice to let the community know about any outcome in this case.

Cheers Edgar

Yes... and it would also be nice to know if the on-line impedance calculators (well, at least one) for short solenoids are not to be trusted.

I am still curious about the difference between circular and numeric coils in this case, and so I have reached out to COMSOL support for clarification. If I learn anything helpful, I will relay the message here.

Putting that asside for now though and setting the coil back to the numeric type, I beleive I may have found the solution... and I must admit that this whole problem appears to be due to a mistake on my part rather than any technical problems.

It is very easy to count the number of turns in the axial direction for my coil. It also APPEARED to be easy to count the number of turns horizontally as well--I could only see 9 turns on the top layer and 8 on the bottom layer. However, digging into the coil a little bit, I realized that the wires are actually better approximated by "hexagonally packed" circles. Per my calculation (see attached image), each column of conductors should only take up about 83% of the circle's diameter. Thus, I now realize that the coil has about 16 layers! Feel free to make fun of me--I had really convinced myself I was right, but I was missing about half of the turns. If I plug 16x30 = 480 turns into COMSOL, it gives me an inductance value of 4.8 mH, which is very much in line with what I would expect. (5.0 mH). Updated simulation is attached.

With my updated geometric picture, I decided to calculate the length of copper wire by hand. I calculated the radius for each layer, found the circumfrance of a loop in that layer, and then multiplied by 30 loops axially. I added up all 16 results and determined the wire length to be about 54.7 m. Assuming 20 AWG wire has 10.15 Ohms/1000ft, I get a resistance of 1.8 Ohms. The actual resistance is 1.95 Ohms--not too bad.

I will say that I can't really get the electronbunker.ca simulation to match as well. I beleive this is because it does not appear to allow for hexagonal packing. It does not permit the radial pitch to be anything less than the total wire diameter.

Thank you both for your help!

No worries, stuff like that happens. I am still curious about the numeric-circular discrepancy.

Edgar and anyone who's interersted--

Here's what COMSOL technical support had to say about the numeric vs circular coil type. I have tried out what they said, and indeed I now get results in perfect agreement. That said, it looks like "Numeric" is still a slightly better choice, so I'm unclear on why "Circular" is ever advantageous.

When using the "Coil" feature of type "Circular" select a single edge (not the inner and outer radius) and that single edge should have the average radius of the coil. In this situation, what you can do is add a workplane (x-y plane) and draw a circle of radius: (48.03[mm]+24.6[mm])/4 and then use those four edges in the Coil Geometry feature. With that change, you'll get about the same L=4.8mH.

In general, the numeric option is actually a bit better, since you there are going to be using a current distribution that is computed on the actual mesh. The circular option assumes that the mesh resolves reasonably well the circular shape.

Then in a follow-up message:

Actually, one other thing: You can (in your original model, with the Circular type) alternatively uncheck the option "Use robust geometry analysis method (requires a complete circle)" and that will also give the expected inductance. (This is actually the default setting, at least in 6.2.)

In any case, yes, using the Numeric coil is (at least in my opinion) the "safest" since it will still be valid if you make the coil, say, slightly elliptical or something like that.

Please feel free to pass this along in whatever way you see fit.

Hi Ben,

thanks for the relay.

Cheers Edgar