What are the tricks for winding a tight air-core inductor?

[niconiconi]

For a project, I need some inductors around 10 uH to 33 uH. Because they're used in a pulse-forming network, the peak current is extremely high, up to 500 A but just one microsecond. Due to core saturation, I think that an air-core inductor is the only acceptable choice, so I decided to pick up this lost art and wind my own. Now the prototype has successfully generated the waveform I need.

But I found it's difficult to wind a professional-looking, tight inductor. If you look at an inductor from classic shortwave radio (see attachments), you can see the winding is tight and evenly spaced, and they're often mechanically stable without additional support. What are the tricks for winding a good inductor coil?

In my attempt, I tried winding my roll of enameled copper wire on the body of a RF torque wrench or a whiteboard marker, 10 uH was around 40 turns. If I don't push the wire, the spacing would be too large. But when I tried to push the coil, it creates a strong mechanical stress. When the coil is eventually released, the windings suddenly spread out and jump like a spring (the mechanical version of inductive kickback?  ).  The lengthening significantly reduces the inductance value. After it's soldered onto the board, I have to squeeze it significantly while watching a meter to tune its value, the final result is a completely deformed coil. The shape and size of the final inductor is unpredictable.

[Benta]

How about using a bobbin and isolated copper wire? That'll also prevent microphonics and other mechanical influences (vibration, shocks).

[mag_therm]

+1 to Benta.
Hold the bobbin rigidly on a mandrel held in a vice or similar.
Tape or solder the start leadout so the turns are wound with tension, by hand.
A better way is on a lathe, but by hand is ok too.

[radiolistener]

1. to get the best Q: use the same length and diameter for the coil

2. to avoid parasitic resonances: keep fixed step distance between neighbor windings of the coil

3. to avoid deformation: use coil frame from good material. Good material should have minimum losses at working frequency.

You can use microwave oven for quick and dirty test of material for coil frame. Just put piece of that material into microwave oven and turn on at full power for a minute. Good material should stay cold. More heat means more worse material.

[niconiconi]

I just learned that there are Material #0 toroid cores for sale, basically just a polymer core. I think I'll try this solution instead. It needs more turns, but it comes as a standardized part, good for PCB use. Parasitic capacitance and Q are not critical for this circuit.

[RoGeorge]

Nice radio! 

The coil in the second pic has more personality, I like that one more.  The ones in the first pic are too perfect, have no soul.  If it has to be uniform, maybe use a piece of plastic pipe as mechanical support, a drill to shape the coil, or cut a piece of PET bottle and roll the plastic sheet into a support cylinder.

For the high current coils, a donuts magnetic will keep the field in the core, while an air coil will radiate a lot of RF parasites, especially with high current and sharp pulses. 

[niconiconi]

Yes, a closed magnetic path in a toroid core is definitely a plus for EMI/EMC... Though I don't mind going back to a solenoid if I can find a more suitable option. The purpose of this entire device is to generate a pulse for immunity testing, so the DUT would also radiate even if the generator does not, so I'm not particularly concerned about inductor radiation (though one can argue a large solenoid inductor is a more effective radiator).

Also an observation during my search for parts - off-the-shelf (not custom) air-core power inductors are surprisingly still being manufactured, by some "boutique" Hi-Fi audio vendors for loudspeaker crossover networks. Due to well-known reasons, Hi-Fi parts are often overpriced. But in this case, $6 for a high-quality power air-core inductor is pretty reasonable to me. Unfortunately, though they have what I need, my local distributor doesn't sell the uncommon low-inductance coils, otherwise I'll just get a premade one.

[RoGeorge]

Another thing for microseconds pulses, that would mean the frequencies in the coil are in the MHz range and higher, where the skin effect might become important.  If high Q factor is needed, a coil made with isolated strands of wires (instead of single wire solid copper) would be better.

[jonpaul]

Bonjour à tous

1. high current PFN pulse forming network are very well studied since 1940s for wwii radar.

Yes aircore is the gold standard.  At 500A most magnetic  matériels will saturate

2. Use a fiberglass or other tube, wind single layer, hold temporary with tape, apply varnish, let dry or bake out, remove tape.

3. Audio Speakers voice coils use a Bondease insulated magnet wire.
With curing  heat the adjacent turns of Bondease wires will bond to each other.

4. Depending upon the ambient temperature, airflow, duty cycle, ohmic and skin effet losses, minimum wire guage is determined.

5. in 1980s  our 4 kV 500..1000 A aircore MRI test coils were ~ 1m dia, 1m l, wire was #8 rect, HPTZ on a custom fiberglass cage  frame.

6. The pulse causes mechanical forces that tend to disassemble the coil. They need to be well fixed by varnish or encapsulation.

Just the ramblings of an old retired EE

Jon

[T3sl4co1l]

Mind that Q factor is worse for tight spacing, than for pitch ~ 2 * wire dia.  Worse still for multilayer, where the current crowding occurs between layers (particularly the inner ones), and as mentioned, worse at high frequencies compared to using litz cable -- but Q also generally goes as sqrt(F), so you might not care if the frequency content is high enough (solid wire is fine for >= 10MHz).

Aspect ratio (coil length / diameter) also matters, best Q being close to 1.

For point of reference, with litz cable, it's possible to make coils with Q factor of 2000 or more, at medium frequencies (100s kHz to few MHz).  Single layer, even spaced, solid wire coils will have Q in the 100 ballpark (give or take say 3x), in the same frequency range.  Tight, multilayer coils can be down in the 10-30 range.  Probably, one of these ranges will be adequate for your application, thus determining the geometry.


I guess magnetic materials could still be used, 500A isn't completely insane; but it will require quite a wide air gap, even for single-turn inductors.  So the result will look more like a string of beads, than a winding as such.  And rather than solid (ring) beads, they would be split for air gap, or even just flat plates arranged either side of the conductor.  Preferably, using powdered iron (for the higher Bsat).

Alternately, a shielding box could be arranged around a solenoid coil, which roughly doubles its inductance while shielding the surroundings -- maybe helpful if you need to pack things together tightly.  The flux density outside of a coil is lower than in the center (where a core would otherwise go), so this can be reasonable even with ferrite; though it may have to be fairly thick still.

There's also machinable powdered iron / composite materials, such as Fluxtrol.

The main application for core materials here, would be if your pulse rep rate / duty cycle is high enough that the air-core inductor is getting too hot.  You can either increase its size (wider and longer, thicker wire, fewer turns), use Litz, or add core pieces.  Can also use tubing and circulate cooling water -- this is the only reasonable solution for 500A continuous, as for induction heating coils.

Tim

[jonpaul]

Rebonjour

This  giant  1980s  is huge but  I think similar specs to the OP requirement.

1.2 mHy 500A 5 kV MRI pulse inductor.

Dimensions ~ .85 m dia 1 m l

Weight !  75 kG

Fabricated many 100s over some years.

HUGE winding machine, as big as a car...

Your feedback?

Bon Soiree!

Jon

[mag_therm]

Hi Tim and Jonpaul,  comparing Induction heating with the present case, coils need low copper R series for best efficiency.
Freq is low and dimensions are large, however mostly, the dimensional scaling is linear considering an empty coil with Ur=1

So:
Efficiency rolls down when length < 3* diameter.

Any non mag shielding closer than about 1.5 dia increases Rseries and reduces jwL hence reducing Q_empty.

Multilayer coils are rarely used  due to the inner layers being in the longitudinal H component.

The current density distribution in turn cross section shows highest density at the inner radius with a fall- off based on reference depth in copper.

Turn space factor is kept high and ranges up to about 0.8 to approach current sheet if voltage and refractory resistivity permit.

If designing Nico's coil at 1 Mhz, I would just follow the above , with a mini version using enameled wire. However the pulse duty is needed to get the size.
I might do a 2D when I get home from Ozzie and check the effect of L/D ratio etc.

[niconiconi]

Rebonjour

This  giant  1980s  is huge but  I think similar specs to the OP requirement.

1.2 mHy 500A 5 kV MRI pulse inductor.

Dimensions ~ .85 m dia 1 m l

Weight !  75 kG

This is impressive but overkill for my purpose. My peak current looks high, but the total energy involved is extremely low: 3 joules. Repetition rate is equally low, once per minute. Similarly, efficiency is something of no concern for this simple experiment device. A "shortwave radio" style air-core coil is sufficient. (An experienced forum member in the field of EMI/EMC can even guess exactly the generator type that I'm building here, given all my previous hints, but I'll save the details for a full article later).

[coppercone2]

Well not every inductor has a bobbin, if you take apart a TV tuner or something you will see alot of free standing inductors.

Not sure if its done for a cost or performance reason.

[niconiconi]

I'm glad to report that I've successfully made the needed air-core inductors for 500-amp, microsecond pulse generation. I finally decided to use Material #0 (= plastic) toroid cores. The part number is T130-0 by Amidon Inc. Its advantage over a homebrew solenoid includes mechanical stability, standardized part for reproducibility, and a closed path to reduce EMI. Only the inductor at the top is the pulse-shaping inductor, the two additional ones at the bottom are di/dt limiting inductors to protect the thyristor (an air core is unnecessary since they don't handle the actual high current pulse, they'll be replaced with a T130-1 core in the next revision).

The actual circuit I'm building is a 1.2/50-8/20 μs impulse generator for testing surge protection circuitry. My breadboard prototype has successfully generated the required 1000 V / 500 A surges, and its waveforms conform to IEC 61000-4-5 Combination Wave Generator's specification.

Open-Circuit Voltage:
    - 1.02 kV, Front time: 1.42 μs, Duration: 58.90 μs
Short-Circuit Current:
    - 496 A, Front time: 7.93 μs, Duration: 22.27 μs

The next step is transferring the design to a real circuit board. When the project is completed, I'll publish all the design files along with a complete article to describe its operation and uses. For now, you can just read Wikipedia, I'm the author of the "IEC 61000-4-5" article. The old version didn't have any details, now it has been updated with all the information I've learned during the construction of this prototype.

[T3sl4co1l]

Aha, nice.

Anyone else looking to make a circuit like that, might also find these of interest...
https://www.allelectronics.com/item/mkp-204/20.0-uf-400-vdc-polypropylene-capacitor/1.html
(At least in the US; YMMV.)

Tim

[jonpaul]

Rebonjour, bravo pour l'effort. But your breadboard has neither the voltage nor current capabilityes required.

  We were consulting for decades in power electronics, HV and compliance/EMI/transient protection. We built similar test fixt in 1980s..1990s for testing of transient protection eg from Furman Sound, Monster, Tripp-Lite.

1/ The toroid's have very close spacing of each ends of wdg,

2/ solenoids are much better topology than toroid's for HV pulse inductors.

3/ Fine wire has high resistance, skin effect, use much thicker wire eg #18..16 AWG

4/ The large film caps have way too much series L and too low a pulse i rating.

5/ Tiny film resistors will be subject to change in R or blowouts.

6/ PCB may not be a preferred way to build. We used FR4/G10 0.125 thk fiberglass perfboard ("Vectorboard/VeroBoard) and Teflon standoffs.

Bon chance!

Jon

[niconiconi]

Rebonjour, bravo pour l'effort. But your breadboard has neither the voltage nor current capabilityes required.
[...]
Jon

The repetition rate of this generator is once per minute, discharging just a 6 uF capacitor, the overall energy involved is very low. I don't expect the current handling capability to be the problem for the board, inductors, wiring or the thyristor.

My most concerning problem is the pulse capacitor. The latest version, IEC 61000-4-5 Edition 3 requires a mandatory 18 uF AC coupling capacitor. It means every time the generator is fired, the AC coupling caps will experience enormous stresses, the stress on the AC-coupling caps is even higher than the storage caps. Properly rated pulse capacitors will be extremely expensive. Thus, I've omitted it for this simplified build. For testing unpowered protection devices, AC-coupling or not makes little difference. Another idea to reduce costs is to simply use a few normal capacitors in parallel - this circuit is built just for some personal experiments (or a quick pre-compliance test, perhaps), it's not meant for production in an EMC test lab working from 9-to-5. So even a lifetime of a few thousands pulses is perhaps acceptable.

3/ Fine wire has high resistance, skin effect, use much thicker wire eg #18..16 AWG

This impulse generator, by design, requires a low-Q inductor. It uses a 0.7 Ω resistor in series with the inductor to kill its Q. High wiring resistance is not a problem. But it's true that dissipating the heat into the resistor is always better than dissipating it into the inductor's ESR. Though I did a quick calculation, at once per minute, the average power is pretty low and I think the wires and cores can definitely handle that.

[jonpaul]

We used a PRF 1/sec for easier manipulation of the DUT.

The surge V and I are affected by coil design.

Photo is typical HV solenoid we manufactured and used in such transient pulse gen.

Note perfect layer, Wind  square wire, fiberglass  tube

Jon

[RoGeorge]

Why the square wire?

[jonpaul]

RoGeorge, this arc lamp series injection transformer generated 35 kV to ignite a 1 kW Xénon lamp

After ignition the current can be 10s..100 A.

A square wire has more Cu cross section and holds the shape better than round.

The 1 M coil previous photo used #4 recto  HPTZ, Belden, either square or rect is special order.

Jon

[coppercone2]

why is the rod inductor better then a torroid for pulse generation?

I am wondering just because I conceived that it could be possible to wind a air coil helical torroid inductor a while back but no one here seemed to know why it would be made. Also wondering about inductors wound on aluminum or copper (less then 1 relative permeability). I never saw one made. I thought it might be good for a LISN if you could make a big enough one for the 250uH but its some what absurd.

Square wire might get more popular now I think because EVs.

[mag_therm]

I did a few core-less toroids , big water cooled ones.
The OD, ID and turn diameter have ratios. I can't find the recipe in my old book now.
They greatly reduce stray flux problems compared to air cored solenoids.
One way to do it, a solenoid is wound on a round or square mandrel, then the turns are curled around another mandrel.
And if the inductance calc is wrong, a complete rebuild usually, so you get dis-honorable mention by the winding department in the morning meeting.

[coppercone2]

Well if the stray flux is better (the only benefit I know), then why is it worse for this pulse gen?

I always figured it would be more stable if the field does not happen to go through the chassis, stuff on the chassis, etc. And you could revise the chassis/system without worrying about the magnetic field. For instance when I built my 50uH LISN prototype with 12 AWG wire air cores, the impedance dropped by 10uH when I put it in the steel box (1 foot long seperated segment coil inside of a ~2x1 foot steel box out of 10AWG steel (made from old storm cover for a basement, it warped to shit from welding and the project is stuck unless I want to do alot of flame straitening).

 And if the inrush current is bigger for a torroid, is it not better for discharging for a pulse also? Could the problem be that the dIdT is too big for the control circuit to handle, where the rod provides some current limiting?

[niconiconi]

4/ The large film caps have way too much series L and too low a pulse i rating.

This prototype used Metallized Polypropylene (MKP) capacitors, the capacitor of choice for pulsed power applications. The pulsed current ratings of these capacitors range between a few hundreds to a few thousands amps. The ESL is not really a big problem for this particular impulse generator, since it's a pretty slow impulse, the frequency content is negligible above 2 MHz. For IEC 61000-4-4 Electrical Fast Transient tests (which I also plan to build in the future), ESL will be a big problem.

Anyone else looking to make a circuit like that, might also find these of interest...
https://www.allelectronics.com/item/mkp-204/20.0-uf-400-vdc-polypropylene-capacitor/1.html
(At least in the US; YMMV.)

I already investigated the available pulse capacitors on the market. The original goal was a 1500 V pulse (now already reduced to 1 kV), I looked at 1600 and 1700 V capacitors. There are three major series: Kemet R75 & R76 series, Vishay MKP385 series, and Cornell-Dubilier 940C & 941C series (the last one also has a cult following by Tesla coil builders), 941C is currently the cheapest, 1600 V, 1.5 µF for just $11. There are also cheaper alternatives, usually sold as "snubber capacitors" for suppressing transient voltages for IGBTs, and thus also has a high surge rating and use the same MKP technology, a few models can be found at LCSC at $5 or so.

The downsides of using proper pulse capacitors are enormous size and high price. I ordered some 941C and I found their size is unreasonably huge for a perfboard, so I used some cheaper 1.2 kV MKP capacitors from China. Without AC coupling, the whole circuit only requires 4 capacitors, so even huge and expensive capacitors are still acceptable. The 4 huge CDE ones cost $44, with a large but still manageable board space. But if 18 uF AC coupling is used, the circuit will be really enormous with 10 huge MKP capacitors just for output coupling, and costs over $100, and be the single most expensive item in this circuit.

Since the goal of this circuit is to make this test easy and accessible for other experimenters, long-term reliability or AC coupling can be sacrificed, as I mentioned previously. I plan to do some lifetime testing of "improper" capacitors and see long they can survive.