Author Topic: Production and design of high Q inductors with 3D printed hardware  (Read 1832 times)

0 Members and 1 Guest are viewing this topic.

Offline sourcechargeTopic starter

  • Regular Contributor
  • *
  • Posts: 221
  • Country: us
Hi there,

I've been working on design and production of inductors of Qs between 200 and 300, and I have been working on trying to squeeze every last bit of efficiency into the design of these things.

So far, I have used a 2inch MPP toroid core( 60u 128al), wound in 24 sections of about 960 turns of 28 AWG polyimide heavy mag wire, using a 1/16 thick electrical tape strip "fence" to hold the beginning and end apart.  Using my 3D printing, I make molds of sections that are screwed together to make these sectional wound toriods as well as a winding bobbin that resembles an "H" to increase the rigidity, reduce the wall thickness, and provide protection of the wire while hand winding. This inductor is measured to have a Q of about 250 to 300 using my mastech LCR meter at 10k hz.  Under resonant load, it drops to about 190 Q (why, is another discussion). 

I am currently designing another inductor to increase the Q by increasing the number of sections to 28, with two 12 degree seperations that are 180 degrees from each other giving the gaps and splitting the inductor in half for a total of 30 sections.  Using my 3D printing, I am going to make the hardware to hold on to the toroid core and maintain the gap between the halfved inductor.  I know that I can increase the Q of the inductor by only using 1 section as a gap, but spliting the inductor in half will allow aditional measurements that I wish to do.

I am wondering about the coating of the toroids though, as the MPP cores from arnolds are diped already.  My question is if there is an extra value in taping the cores with a layer of polyimide tape?

Also, does anyone have any opinion on the inductor core hardware that uses PLA as seperation and covering material as I have not been able to find any datasheets on the electrical properites of PLA for it's dissipation or dielectric breakdown limit.  I could use ABS but I'm not sure if ABS is bettter than PLA.

Thanks.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #1 on: August 25, 2019, 10:13:00 am »
To what end?  What are you resonating with/for?

960 turns is an insane amount of wire for most purposes; that's a sizable fraction of a henry, total! :)

If this is for low frequencies, you might try to avoid using inductors, for power transfer, if possible; either shift the frequency up, or use alternate means (capacitors? piezos??).  Something of an X-Y problem.

Why use bobbin sections?  Why not wind freehand, back and forth, covering the toroid evenly (adding tape layers for spacing if needed)?  (The gaps will tend to reduce how much wire you can use, and increase stray fields, probably reducing Q overall.  Maybe it's not much effect?  I'm not sure!)

Why not use a larger core, or a different material?

This may be of interest:



These Q values are derived from the core loss curves; real Q as-wound may vary.  Usually downwards due to copper resistance, but maybe upwards also, depending on relative permeability and winding distribution (i.e., particularly for the #2 material, which has much higher inductivity when the turns are grouped together, instead of spread out evenly).

The Sendust core seems poorer than your MPP core, I'll have to add some MPPs to the plot some time.  (The core you have may well be one of the best picks already.)

Ferrite may be worth consideration too, but it's harder to plot in this way, because airgap is key.


More insulation on the base layer, probably doesn't matter: presumably, you're already operating at such a high frequency that the core looks like a[n electric] short circuit, its resistivity doesn't much matter.

You may also consider using litz wire.  Probably the 28AWG wire is subject to proximity effect, deep within those windings, which I'm guessing go many layers deep.  Even at fairly low frequencies.  Easy enough to try with some, say, 7 to 50 strand wire, of the same total cross section.  (For example, 16 strands x 40AWG would be equivalent I think.)

Regarding large-signal losses, it probably has something to do with initial versus average permeability, the losses that accompany it, and also the change in reactance.  It's pretty common for ferromagnetic materials to have a lower mu_r for small signals, say for fractional A.t/m.  It may be that the lower mu acts like more airgap, and therefore gives lower losses; alternately, it may be that the smaller hysteresis loop thus traveled, happens to enclose a relatively smaller area and therefore there's less loss due to that property.

Conversely, for large signals (which for this many turns, may require hundreds of volts!), mu_r is larger, and while this also gives higher reactance, it comes with a disproportionately larger hysteresis loop, and so the Q is overall lower.

It could well end up the other way around, that despite the core having higher losses, the copper loss was instead dominant, and the higher reactance improves things overall.  That might be the more likely outcome for lower-mu cores, where copper resistance is more significant.  But such cores have less mu_i/mu_avg change, so it would be harder to see as well.

In any case, not much you can do about it, just mentally adjust downward the measurement seen on the meter.  By how much, depends on signal level, number of turns, and material.  So... :-//

Tim
« Last Edit: August 25, 2019, 10:18:47 am by T3sl4co1l »
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline sourcechargeTopic starter

  • Regular Contributor
  • *
  • Posts: 221
  • Country: us
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #2 on: August 25, 2019, 12:44:20 pm »
1st let me thank you for responding to me in such a very professional and concise questioning.

To what end?  What are you resonating with/for?

Research. Multiphase series resonant converter.

960 turns is an insane amount of wire for most purposes; that's a sizable fraction of a henry, total! :)

It's about 130mH, but compared to my 2000H and 80H toroid inductors, not so much...

If this is for low frequencies, you might try to avoid using inductors, for power transfer, if possible; either shift the frequency up, or use alternate means (capacitors? piezos??).  Something of an X-Y problem.

Higher frequencies mean better switching requirements and lower frequencies have cheaper components with less resistance.

Why use bobbin sections?  Why not wind freehand, back and forth, covering the toroid evenly (adding tape layers for spacing if needed)?  (The gaps will tend to reduce how much wire you can use, and increase stray fields, probably reducing Q overall.  Maybe it's not much effect?  I'm not sure!)

I have tried freehand winding, and making mulitple inductors with tolerances between them to be somewhat resembeling each other requires a technique that dictates a more repetitive productive procedure.   It's true that I "could" do this free hand, but doing it over and over would be very, well, hard....The stray parasitic capictance is what I was asking about, but I'm thinking that the 12 degrees seperations for the gaps would be large enought to decrease this capacitance.  My concern is about the dissipation, dielectric constant, and the dielectric breakdown of PLA vs ABS.  One of the problems with progressive winding and using tape to cover the previous layer would be in production, as this would be extremely difficult to do, and you would know this if you have ever wound a toroid by hand, let alone, multiple toroids.

Why not use a larger core, or a different material?
MPP cores are the most efficient cores on the market, aside from no core and using LN2, but this is not an opition for my "lab."  MPP cores are extremely expensive and the last time I checked on the price of a 4 inch MPP core, which BTW would fit in my toroid winder, was about 50 bucks a core.


This may be of interest:



These Q values are derived from the core loss curves; real Q as-wound may vary.  Usually downwards due to copper resistance, but maybe upwards also, depending on relative permeability and winding distribution (i.e., particularly for the #2 material, which has much higher inductivity when the turns are grouped together, instead of spread out evenly).

The Sendust core seems poorer than your MPP core, I'll have to add some MPPs to the plot some time.  (The core you have may well be one of the best picks already.)

Ferrite may be worth consideration too, but it's harder to plot in this way, because airgap is key.


More insulation on the base layer, probably doesn't matter: presumably, you're already operating at such a high frequency that the core looks like a[n electric] short circuit, its resistivity doesn't much matter.

You may also consider using litz wire.  Probably the 28AWG wire is subject to proximity effect, deep within those windings, which I'm guessing go many layers deep.  Even at fairly low frequencies.  Easy enough to try with some, say, 7 to 50 strand wire, of the same total cross section.  (For example, 16 strands x 40AWG would be equivalent I think.)

Regarding large-signal losses, it probably has something to do with initial versus average permeability, the losses that accompany it, and also the change in reactance.  It's pretty common for ferromagnetic materials to have a lower mu_r for small signals, say for fractional A.t/m.  It may be that the lower mu acts like more airgap, and therefore gives lower losses; alternately, it may be that the smaller hysteresis loop thus traveled, happens to enclose a relatively smaller area and therefore there's less loss due to that property.

Conversely, for large signals (which for this many turns, may require hundreds of volts!), mu_r is larger, and while this also gives higher reactance, it comes with a disproportionately larger hysteresis loop, and so the Q is overall lower.

It could well end up the other way around, that despite the core having higher losses, the copper loss was instead dominant, and the higher reactance improves things overall.  That might be the more likely outcome for lower-mu cores, where copper resistance is more significant.  But such cores have less mu_i/mu_avg change, so it would be harder to see as well.

In any case, not much you can do about it, just mentally adjust downward the measurement seen on the meter.  By how much, depends on signal level, number of turns, and material.  So... :-//

Tim

Q, depends on resistance, resistance of the wire in both AC and DC, resistance of the core, and resistance of the parasitic capacitance of the wire.

The AC resistance of the 28 AWG wire at 10khz is nil.  Litz wire is expensive and would not contrubute to increasing the Q at such low frequencies.  The MPP cores from arnolds magnetics are "the most efficient cores on the market." 

Parasitic capacitance of the wire can be limited by the winding technique and can include progressive winding and sectional winding techiques.  Progressive winding that is the same from one inductor to another, can be done using toroid winders.  My DIY toroid winder is able to use about 3 inch cores and higher.  Secontional winding allows the production of multiple inductors with relativly the same winding technique characteristics.   I already own 24, 2inch MPP cores that I bought at a electronics depot for a reduced price.

Core resistance can be calculated by "Legg's equation"

R(core) = u(r) x L x F (a x B(max) x f + c x f x e x f^2)   (edited due to mistake, see later post)
a, c, and e are listed within the arnold's datasheets and therefore the core resistance and be approximated for testing purposes
Notice as the permablity increases, the core resistance increases..That is the main drawback for using solid iron cores.

BTW, do you have much experience with PLA or ABS 3d printed hardware for inductor cores?
« Last Edit: August 26, 2019, 01:18:30 pm by sourcecharge »
 

Online ch_scr

  • Frequent Contributor
  • **
  • Posts: 885
  • Country: de
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #3 on: August 25, 2019, 02:00:33 pm »
Kerry Wong has tested the dielectric strength of PLA, also linked in some papers http://www.kerrywong.com/2019/01/06/pla-dielectric-strength-measurement/
Any chance we get to see some pictures / renders of your coil winding aids?
 

Offline sourcechargeTopic starter

  • Regular Contributor
  • *
  • Posts: 221
  • Country: us
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #4 on: August 25, 2019, 03:02:14 pm »
Kerry Wong has tested the dielectric strength of PLA, also linked in some papers http://www.kerrywong.com/2019/01/06/pla-dielectric-strength-measurement/
Any chance we get to see some pictures / renders of your coil winding aids?


45kV/mm, nice, now all I need is the dielectric constant and the dissipation factor.

Here are screen captures of one of the sectional mold wedges, the core hardware and the H winding bobbin.

I can't upload the freecad files, but anyone really can make these with the right dimensions for individual toroid cores using freecad.

Did you want some boring pics of the wound core?
 

Online ch_scr

  • Frequent Contributor
  • **
  • Posts: 885
  • Country: de
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #5 on: August 25, 2019, 04:10:00 pm »
These guys http://poseidon2.feld.cvut.cz/conf/poster/proceedings/Poster_2018/Section_PE/PE_043_Vesely.pdf seem to have measured the loss tangent and dielectric constant of FDM PLA and ABS samples. You can upload anything as a zip file btw, but a picture of how those parts go together while winding a core would be more interesting I suppose.
 

Offline sourcechargeTopic starter

  • Regular Contributor
  • *
  • Posts: 221
  • Country: us
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #6 on: August 25, 2019, 06:21:28 pm »
These guys http://poseidon2.feld.cvut.cz/conf/poster/proceedings/Poster_2018/Section_PE/PE_043_Vesely.pdf seem to have measured the loss tangent and dielectric constant of FDM PLA and ABS samples. You can upload anything as a zip file btw, but a picture of how those parts go together while winding a core would be more interesting I suppose.

Thanks for the link...I will review and get back..

I am currently cleaning up a batch of 96 sections of toroid molds.  I only need 60 for a 30 sectional wound toroid, but it was more efficient to print this many just in case some break.  Cleaning of the molds involve smoothing the wedge straight edges and removing the flashing.  It takes some time, but I will post a pic of the completed mold composite with the core visable, along with the H bobbin shown being able to fit within the open window area for winding.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #7 on: August 25, 2019, 10:24:53 pm »
The AC resistance of the 28 AWG wire at 10khz is nil.  Litz wire is expensive and would not contrubute to increasing the Q at such low frequencies.  The MPP cores from arnolds magnetics are "the most efficient cores on the market."

Okay but those are just quoted things, what argument or data support them? :D

Looking at the curves, I see higher frequency for the same loss, comparing the Sendust 60u of my plot to MPP 60u, say 140mT vs. 250mT at 100 mW/cm^3 and 10kHz.  So MPP should be up there with the #2 core on the plot, but with far higher mu, which is quite nice indeed.  But also comparable to what you're seeing -- so we can conclude that is currently the limiting factor.

For sine waves, the formula is:
\[ Q = \frac{ \pi F {B_{pk}}^2 }{ \mu_r \mu_0 p_c } \]
Where p_c is the core loss (remember to use consistent units!).

Most of these materials have a core loss best-fit formula with a small exponent of Bpk (near 1), which implies they're better at high frequencies.

A metal has a bulk time constant: \$\mu_0 / \rho\$.  This has units of ohms/m^2, so there's still a geometry factor in there.  It involves two lengths: the magnetic loop length, and the length equivalent volume of the wire (cross section divided by wire length, as you would use to calculate conductance).  So the full time constant is:
\[ \tau = \frac{ \mu_0 l_e A_\textrm{Cu} }{ \rho l_\textrm{Cu} } \]
where the area is the wire cross-section.

The core further improves this, but too much core (high mu) and you can't store much energy before it saturates, and the core properties themselves probably aren't real great (too low R_core).

What you aren't getting, with more turns or higher inductances, is a longer time constant in the copper.  As wire gauge gets finer, cross section goes down while length (number of turns) goes up; resistance goes up as the square.  Inductance goes as the square too, so they go at the same rate.

Note this means a [physically] larger inductor has a longer time constant.  So you can always get better performance with a bigger and bigger core and winding (until transmission line effects stop you)... but, well, that's the rub, isn't it?

Of course you'll need large inductances for very low impedances or very high voltages, and I don't know which way you're going with this.  But it's quite possible you're below the peak Q frequency, down in this kHz range.  If you can measure at higher frequencies (this will probably require windings with fewer turns, of larger wire; adjust for proximity effect to get equivalent Q) that should be illuminating.

https://en.wikipedia.org/wiki/Proximity_effect_(electromagnetism)
This shows how to estimate proximity effect.


Quote
Parasitic capacitance of the wire can be limited by the winding technique and can include progressive winding and sectional winding techiques.  Progressive winding that is the same from one inductor to another, can be done using toroid winders.  My DIY toroid winder is able to use about 3 inch cores and higher.  Secontional winding allows the production of multiple inductors with relativly the same winding technique characteristics.   I already own 24, 2inch MPP cores that I bought at a electronics depot for a reduced price.

Nice score! :)

Hmm, 980 turns of 28 AWG is around a 15% winding factor on a, say, C055716A2 (which would only be 70mH); with sectional dividers, that might look more like, 25, 30% per section?  So, around 61mm/turn average, which should be around 12.7 ohms DCR?  And that's going to be about, what, ehhh, maybe 8-10 layers deep?

Then, proximity effect.  h/δ is around 0.4 here (assuming 10kHz), and the plot gives for m=10, Rac/Rdc around... 1.2ish.  So if these assumptions are right, it's just getting started, but it's not significant enough to be a problem.

If you're running at higher frequencies after all, or the layers are much deeper (Rac/Rdc ~ 2 at 20 layers), it will get more significant, but shouldn't be a big deal otherwise. :-+

Interestingly, 70mH and 12.7 * 1.2 ohms at 10kHz is a Q of 289, comparable to what the core Q should be.  But both can't be this high; the parallel combination must be -- well, if they're comparable, then about half, or 140.

I wonder if I dropped a constant in my core Q formula?  I derived it years ago, and no one's told me, but no one ever reads or checks these kinds of things...


Quote
Core resistance can be calculated by "Legg's equation"

R(core) = u(r) x L x F (a x B(max) x f + c x f^2 x e x f)
a, c, and e are listed within the arnold's datasheets and therefore the core resistance and be approximated for testing purposes
Notice as the permablity increases, the core resistance increases..That is the main drawback for using solid iron cores.

I'm not sure what all is going on here (two F's? what is the function u?) but that will be the parallel equivalent resistance, no?  So a higher resistance is better?


Quote
BTW, do you have much experience with PLA or ABS 3d printed hardware for inductor cores?

Yes; it's small either way at low frequencies, but PLA is preferred.  ch_scr's link looks useful.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline sourcechargeTopic starter

  • Regular Contributor
  • *
  • Posts: 221
  • Country: us
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #8 on: August 26, 2019, 01:17:28 pm »
Ya, it looks like PLA beats ABS hands down.  It's dielectric constant is about the same as ABS but the dissipation fact is much lower in PLA.

Legg's equation is great.  You should try it..

https://en.wikipedia.org/wiki/Magnetic_core

Looks like I typed it in wrong above, and I will edit it so it's not used wrong, but the Legg's equation depends on knowing the A C and E coefficents, and arnold's MPP cores have datasheets that give these values at low flux densities.

the calc on the DC resistance is almost exact, it's about 13 ohms.  If this was the only resistance, that would be great. The inductor is about 130mH, but the core resistance and parasitic capacitance adds just enought to drop the Q at 10khz to about 298, but for some reason, it's not always 298.  I've measured the SRF of the coil to be about 70Khz, so the R(di) is only about 1 ohm or so...

Still working on cleaning up the molds, and I will post some pics later.

 

Offline sourcechargeTopic starter

  • Regular Contributor
  • *
  • Posts: 221
  • Country: us
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #9 on: August 27, 2019, 06:53:51 pm »
Well, After filing down the straight surfaces of the toroid molds, I wound up with about a 1/4'' space that could not be filled by an extra mold section.  I think this might have happened before with the 24 sections, and that is why I used the 1/16 tape fence to begin with instead of using a section.  Before I filed down the straight surfaces, the mold fit pretty good, but I think I might have redesgin them to include some extra width for post processing of the molds' straight surfaces to account for removing burs and making the straight surfaces smooth.  This is going to take some time for design, printing, and post processing.  I'll get those pics out when I can though...
 

Offline sourcechargeTopic starter

  • Regular Contributor
  • *
  • Posts: 221
  • Country: us
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #10 on: September 01, 2019, 11:01:43 pm »
yada yada yada, winding toroids sucks...
 

Online ch_scr

  • Frequent Contributor
  • **
  • Posts: 885
  • Country: de
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #11 on: September 02, 2019, 07:51:54 pm »
Looks nice! I might be a bit dense, but how does a "sectional winding" go on the core?
I suppose you do all the windings in this sector, then somehow (how?) fix them to the core, remove the next mold sector rinse & repeat?
If so, are  all the sectors then connected in series or individually brought out? And how does the "core hardware holder" fit into the equation?
 
The following users thanked this post: sourcecharge

Offline sourcechargeTopic starter

  • Regular Contributor
  • *
  • Posts: 221
  • Country: us
Re: Production and design of high Q inductors with 3D printed hardware
« Reply #12 on: September 02, 2019, 08:59:31 pm »
Looks nice! I might be a bit dense, but how does a "sectional winding" go on the core?
I suppose you do all the windings in this sector, then somehow (how?) fix them to the core, remove the next mold sector rinse & repeat?
If so, are  all the sectors then connected in series or individually brought out? And how does the "core hardware holder" fit into the equation?

Well, each section is independent of each other.  All of the sections are first aligned correctly to the core, and then each section is tightened down with the 4-40 screw and nut.  This toroid of the 22 toroids was the shortest height of all of them and if and when I get around to another toroid, I'm thinking of using washers between the top and bottom of the molds to fill the gap. 

Regarding the change from one section to the next, on my first sectional mold of 24 sections that I did, I thought this might be a problem.  When winding, I would keep constant tension on the wire, and when the next section was due, I kept a small amount of tension on the wire as I removed the next section. After the first turn in the next section, the previous section as essentially locked down.  The movement of the wire in the previous section when starting the next section was little, as I remember there may have been movement of only the last turn that may have slipped, but it was easily put back.  The key is to always have tension on the wire.

This may not be the case with this new one though because of the smaller width of the sectional winding due to using more sections.  I haven't actually started yet, but I'm hoping that the 30 sectional mold will have the same winding characteristics as the 24 sectional mold.  I'm also hoping that the added sections will increase the Q.

On a separate note, the molds don't really have a great alignment unless some care is used to make sure the inner smaller top wedge part aligns with the bottom, and that I think if I were going to make more molds in the future, I would include some type of keying of the top and bottom to more easily keep the alignment.  Also, the 1/4'' gap I thought was a problem was actually due to how I put the molds on.  Basically when the sections were loosely on, they didn't fit perfect but when they were tightened down in the correct position, everything fit together nicely.

The core hardware holder is basically what is going to hold the toroid inductor away from damage, and create connection points for the inductor wire acting as terminals.  I've actually thought about changing it a bit by making the center strip a circular piece that has a hole so that later, a cover for the inductor can be bolted on, completely shielding it from damage.  These core hardware holders have not been printed yet, as I was going to check the window area dimensions after the core was wound, and adjust this circular piece to fit this dimension.

The hardware holder is really only a prototype, as I'm thinking that it's better to hold onto the core, than the windings.  Just typing this actually has my mind already coming up with a better design. (Thanks)  Maybe I might include a extra piece that spans across the holder on it's top to reinforce the holder.  Anyways, if you have any more question or maybe suggestions I'm up for that.

Lastly, I still haven't come up with a definitive answer if taping the core with a thin layer polyimide tape would help increase the Q of the inductor.

 
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf