I'm curious as to why you think tubing would have a higher Q than edge wound. Seems to me the current would be concentrated on the inner facing edges of the tubing due to proximity effect, and also concentrated on the inner edge as it would be with edge wound anyways. so if you compare an edge wound squished tube of the same dimensions and inductance as a helix of tubing.. seems to me the edge wound should have lower proximity current effect producing higher resistance. It should also have slightly lower turn to turn capacitance due to the increased space between turns.
Yes, exactly: curvature matters. Magnetic fields more or less "want" to be round, and a square boundary condition makes friction (so to speak, but in this case kind of literally as well).
Somewhat obscure but visceral illustration: if you put a steel cylinder inside a moderately-close-fitting induction coil, made with square tubing, and hit it with a good couple 10kW, you know, like you do... you see the image currents from the corners of the tubing lighting up much faster. Before the whole cylinder melts away like 20s later or whatever, but in those first moments when it starts glowing, yeah.
Whereas with a round tubing coil, you see the cylinder light up under the tubing, in a fairly rounded manner.
The real zinger is, the image current from each corner is visible. You get two peaks from both edges of each turn, rather than just one peak right underneath the round.
So, you get more current density in the corners, than you do the inside-ish edge-ish of the tubing, and for conductor of the same cross-sectional area (based on the outer perimeter, i.e. ignoring if it's hollow tubing or solid) you get more loss with the square corners than round.
This doesn't change when cores are used; the popularity of edge-wound inductors is a matter of optimizing DCR, and while they can offer pretty nice Q at AC, it's still something you want to use at a lower ripple fraction to keep efficiency up. The main limitation is the air gap: the fringing field intersects the windings, and the windings being solid, there's nowhere for the field to go, it crashes right into the solid metal surface, inducing eddy currents. I did an LLC converter with planar transformer recently, where the Lm-mode Q factor went from ~150 with a single gap, to ~400 with a proprietary solution.
This website suggests he's measured Q as high as 1100 for tubing and 900 for edge wound, but this may have to do with just what he's been able to get hands on and measure, and I presume it is very difficult to measure the difference between an 1100 Q and a 900. https://www.w8ji.com/loading_inductors.htm
I wonder if one issue might be edge winding work hardens the copper and increases its resistance.
That sounds about right. The other thing is, you need the winding pitch about twice the diameter (or for strip, effective diameter: imagine something like, a bit smaller than the bounding circle) to optimize Q: permit enough space for field to flow around wire, between turns, and yes this necessarily increases leakage (more leakage means less mutual inductance, total L drops), but it reduces eddy currents faster at first, so Q is improved.
We don't do this [modest pitch winding] for standard PCB and SMPS components, because size and cost dominate over Q, and there's no space to spare inside a ferrite core anyway. Again, DCR generally gets priority.
Work hardening may be relevant; it's a smaller fraction though, I think -- for copper, more like 5%, so a contributor to the total ~20% difference, but the geometry is still dominant.
Or if that was comparing to one with finer pitch, or smaller diameter wire, it could go either way. A 20% difference feels very neither-here-nor-there; both inductors are quite good in general (~1k is bragging-rights territory), and making a subtle change like pitch, conductor diameter (or strip aspect ratio for that matter), or even just a modest change to diameter or length, has a comparable impact on Q.
I wonder what a practical Q be if you were to take a bundle of say,24 gauge wires, twist them together, and fit them into a coil form to make an edge wound, continuously interposed inductor.
24AWG isn't very much -- crowding dominates over interleave pretty quickly, but that would be okay for a few kHz. Count matters: 10 or 20 strands may well be worse than 1-3 strands of equal cross section; once you get into the hundreds of strands count, litz gets properly effective. (Or for something you might otherwise choose tubing, tens of thousands, easily! We used broom-handle-sized litz cables to wire up induction power supplies, back at PPoE. That was the floor-standing power supplies, 600A IGBT modules, which ran up to 50kHz or so, and I think used 36AWG strands.)
Also hard to twist wires into a rectangular form -- especially as stiff as 24AWG. Practical problem, perhaps,
![Smiley :)](https://www.eevblog.com/forum/Smileys/default/smiley.gif)
but there are actually rolled, crushed or otherwise formed litz cables available. Fairly specialized I think, not something you'll just run across, but I have a random sample in my collection that came from NEWT; it is what it says it is.
Litz probably isn't relevant here [in this thread], as it becomes ineffective above some MHz: the necessary stranding becomes too fine (it's harder to get e.g. 46AWG, or more, not to mention handling it), and the voltage between strands increases to such a point that the skin effect around the cable itself -- now due to dielectric as well as resistive losses -- starts to support eddy currents, and the cable looks like a round wire again, no not wholly so, not suddenly, but that there is some shielding effect again, where magnetic field becomes excluded from the middle, and you get skin effect and (small loop) eddy currents.
To be more clear about that: what litz is, is, we want a conductor transparent to magnetic fields, yet which supports longitudinal current. It's an anisotropic conductor. At lower frequencies, we get a large anisotropy ratio, and performance is good: we can pack a lot of metal into a small space and not worry about eddy currents. As frequency rises, the ratio shrinks, and we can only try so hard before wires get too fine, or insulation too thin, or packing density drops anyway.
At this point (somewhere in the mid 1s to low 10s MHz), a solid-wire inductor with judicious pitch, and maybe a powder core loading it, does very well already, and it isn't a big deal to use the solid wire, versus putting in heroic effort to use litz.
Tim