PFC with non linear inductor

[opampsmoker]

Hi,
Non-linear inductors have higher peak B capability, and higher delta B capability than ferrite inductors.
What is the PFC power level (240VAC input) for a Boost PFC, where you would stop using a ferrite inductor and use a non linear inductor instead?

Talking of one PFC...ie not several PFCs in parallel.

[Yansi]

Every  cored inductor is pretty darn non-linear if driven hard enough, so what kind of question is this?

If you drive the material so hard into saturation, core losses become significant, if not extreme.

[T3sl4co1l]

The classical application is a "swinging choke".  Start with a choke of adequate size for the supply voltage and current.  Reduce the core gap (usually with an odd shape, so that saturation occurs somewhat stepwise).  The higher initial inductance gives lower critical load current, for a given size of inductor.

This is only relevant to choke-input rectifiers, typically used with mains supplies (and at that, typically only in vintage (tube) equipment).

The biggest problem elsewhere is dynamics.

Indeed, the dynamic problem still exists even in the filter -- namely that your LC filter network exhibits a variable cutoff frequency, and you want to avoid resonating at mains frequency for example.  This limits the minimum (saturated) inductance you can choose.

For switching circuits, the inductance is a term in the control compensation.  The loop must be compensated for the worst of either condition.  Overall bandwidth suffers.  A more stable inductance is preferable, allowing the loop to be tuned tighter.

For this reason, it's common to design a system using a powder inductor up to maybe 20 or 30% saturation, rather than say 50% or more.  Designing with powder cores, this tends to prefer low-mu types (20-60 mu).  Designing with off-the-shelf inductors, just follow the given saturation curve (and keep shopping if one is not provided).

Note that the exact same limitation applies to variable (type 2 dielectric) capacitors, in regards to the voltage loop, rather than the inductor current loop.

Tim

[TimNJ]

My two cents: Do you need to meet current harmonics requirements over the entire operating range of the converter? If not, and if you have the space for a gapped ferrite inductor that won't saturate at full power, use it. Ferrite will yield lower losses and lower temperatures, at the maximum load condition. As power level increases, PFC inductance requirement decreases, generally. Suppose you have a 500W design requirement. At 500W, maybe you need 300uH to meet some current harmonics requirements. But, at 100W, you might need 1000uH. So, 1000uH is the minimum required value, which forces an overall larger inductor design if you don't want saturation at 500W. You can use a sloped-gap or two discrete gap lengths. I think two discrete gaps is preferred for manufacturing, but someone can correct me if I'm wrong.

I've seen one or two manufacturers do 500W+ on powder alloy cores (two of them stacked), but I don't think it's the optimal solution for efficiency in most cases. Ferrite still beats even molypermalloy (in terms of core loss) by quite a bit. From my understanding, I'd only use a powder core if you have a specific line harmonics (or perhaps conducted EMI) requirement to meet over a wide range, and can simultaneoulsy accept higher inductor losses.

[TimNJ]

The added distributed air gaps give the soft saturation behavior and in some cases, lower loss behavior, at the cost of reducing permeability.

Can you describe the conditions under which a powder core will yield lower losses? I've seen people suggest this, but any testing I've done has shown worse efficiency compared to ferrite.

[Wolfram]

I've seen one or two manufacturers do 500W+ on powder alloy cores (two of them stacked), but I don't think it's the optimal solution for efficiency in most cases. Ferrite still beats even molypermalloy (in terms of core loss) by quite a bit. From my understanding, I'd only use a powder core if you have a specific line harmonics (or perhaps conducted EMI) requirement to meet over a wide range, and can simultaneoulsy accept higher inductor losses.

I've used Sendust alloy powder cores in a 30 kW SiC boost converter (380 V in, 750 V out), and reached a measured 99.17 % efficiency, using a regenerative test bench. This was using two cores weighing around a pound each, with a cost of around 4 euros per core in moderate quantities. Whether ferrite or alloy powder cores are the best option in a given situation is a pretty tricky question to answer, with many tradeoffs. Alloy powder cores have higher losses for a given B swing, but they also handle DC bias better than ferrite, and the distributed gap reduces eddy current losses in the windings.

[TimNJ]

Can you describe the conditions under which a powder core will yield lower losses? I've seen people suggest this, but any testing I've done has shown worse efficiency compared to ferrite.

I almost exclusively use toroidal cores, so there's not a good way of introducing air gaps to ferrite cores, thus the lower power density, or higher loss if you drive B higher.

With powder cores, you can introduce however much distributed air gaps as you wish.

This seems like a case-to-case thing (as always). Obviously, the vast majority of commercial power supplies use air-gapped ferrite cores..so there clearly is a way to reduce the permeability this way. Is your point that there's no good way to gap ferrite toroids? If so, then yes, gapping toroids kind of sucks. But then my question is: Why are you using almost exclusively toroids when you can use RM, PQ, POT, EE, etc.? I could see that small toroids may have better space usage than core+bobbin.

I suppose fringing field related losses may be relevant at some point, but I do not have a good handle on that.

Anyway, not trying to derail the OP's thread, but I feel this is still on topic. Thanks.

[T3sl4co1l]

Just to clarify: ferrite doesn't much care how it's gapped, just that flux density amplitude is reasonable for whatever power dissipation is allowed.

Where it does matter, is the winding and what wire is used.  The fringing field around the gap is extremely aggressive, inducing eddy currents even in fairly fine wire.

Typically, the center limb of a shape core is cut short, leaving an internal air gap.  The width of this gap is the total air gap in the core, all other core faces are flush.  This is typically a sizable amount, 1mm say, so the fringing field extends quite far, enough that a simple straight bobbin won't save you.

Bobbins can be specially constructed to maximize distance between inhomogeneous fields and the windings, and the windings themselves (proximity effect), to minimize copper losses.  These are easily made in these days of 3D printing, but I haven't seen any used in production at least yet (but I also don't take apart a lot of stuff anymore these days, so that's not a great reference!).

The alternative is to use extremely fine wire, which is cost-prohibitive, but hey, if you have some on hand, you absolutely can do it, you'll also get reduced ACR in general for even higher efficiency overall.

There is no reasonable way to gap a toroid; ferrite toroids are, to my knowledge, never gapped in any production-adjacent context, for purposes of distributing air gaps (exceptions for very specific circumstances, like Hall effect sensors -- but that's not for distributing air gaps or reducing losses).  You could fracture or cut a toroid into a bunch of arc segments and gap them individually, and the fringing field around each gap can be smaller than the wire diameter wrapped over it (bonus points for adding a layer of spacer tape over the gaps, to increase that distance).

Heh, I've even had a toroid do this to me, automatically.  When pulse testing one, which might've been NiZn material for what it's worth, as I swept frequency, at one range it seemed to do something a bit different, and going back over it again more slowly, tick, and suddenly the inductance was much lower.  I poked at it, and the core fell apart in the winding: ten or so fragments of very regular size.  Apparently I had hit an acoustic resonance, excited by magnetostriction, driven by the unidirectional pulses I was using.  Not that I would recommend cutting ferrites in this way.

So yeah, powder cores are desirable where the size and shape of a toroid is, at least not detrimental, but perhaps even preferential; and where copper losses are best avoided, while remaining economical.  Gapped ferrite shapes tend to give more compactness at higher frequencies, but tend to be more expensive.

A common strategy is to avoid the issue entirely: use the best of both worlds.  A typical DC-DC module might be an active clamped forward converter, using a planar ferrite transformer (a transformer has maximized inductance, zero air gap) and either a planar gapped ferrite inductor (in a planar winding, the wiring can be relatively far from the gap, and can also be customized on every layer), or a composite molded inductor (like toroids, these use powdered metals, but achieve higher density and lower cost while maintaining high Q).

Tim

[mag_therm]

But then my question is: Why are you using almost exclusively toroids when you can use RM, PQ, POT, EE, etc.?

Mainly for power density reasons. As you've said, the bobbins can be quite large (talking about 3 digit W/in3 all inclusive, every bit of space matters).

Also, there are no sharp corners, meaning flux distribution is more even, and that means it takes quite higher average flux density to get one point to saturate (since less concentration), that allows me to drive the cores harder.

Another geometry benefit is that a toroidal core is extremely easy to model in cylindrical coordinate system, which is the de facto natural habitat of integral Maxwell's equations.

It makes finding analytical solutions easier for math noobs like me, and makes computing numerical solutions easier even without unaffordable tools like Ansys Maxwell or Q3D.

I do all with SciPy (homebrew differential solver with core material support, for cored transformers and inductors) and FastHenry (multipole solver without core material support, for interconnections).

Further, using laser direct patterning (using a laser to burn epoxy mixed with CuO powder to selectively deposit copper) and electroplating, I can make thin and wide windings to further reduce volume as well as to reduce leakage inductance (since there is no cramming at ID and spacing at OD, I can make perfect sector shaped conductors), and I can make easy shielding layers to eliminate the needs for Y caps.

When money is not the No. 1 priority and all you care is how much can you lead ahead of Vicor, a lot of interest design decisions can be made.

In QuickField, the toroid can not be modelled in axi-symmetrical plane.
Not sure but I think the other general purpose  2D sims are similar.

The current has to go at 90 degrees to the plane so toroid is set up in X-Y plane,
looking at  annulus.

Earlier this year I did some weeks of modelling a large (2000kVA) toroidal AC reactor
which identified its failure mode and led to modelling a replacement with a double E
configuration with distributed gaps.

AC reactors with magnetic cores are difficult and troublesome,
compared to 2 winding transformers I have learned to avoid them where feasible.

[mag_therm]

Hi Blueskull,
There is a way to use a 2D solver to solve in the plane orthogonal to its intended plane.
That is to exploit the duality of the orthogonal solutions and transpose B [T] and J [A/m^2] .

That can give the field distributions visually, but in the packages, the post processing does not work, and the solution can not be coupled to the thermal PDE.

I have done it to obtain visually, distributions in complex shapes, for example the effect of rounding sharp corners.

Your method for skin effect is OK for conductors without ferromagnetic permeability.

[Wolfram]

This seems like a case-to-case thing (as always). Obviously, the vast majority of commercial power supplies use air-gapped ferrite cores.

This is true for lower power applications, but toroidal core inductors are more common in high power applications. To list a few examples off the top of my head, the Tesla Model S and Model 3 11 kW onboard chargers use toroidal PFC chokes, so does the onboard charger for the prototype BMW electric Smart. Also the bigger Elektro-Automatik lab power supplies I've taken apart use toroidal PFC chokes, while the TDK Lambda lab power supplies and Chevrolet Volt charger use gapped E cores.

There is no reasonable way to gap a toroid; ferrite toroids are, to my knowledge, never gapped in any production-adjacent context, for purposes of distributing air gaps (exceptions for very specific circumstances, like Hall effect sensors -- but that's not for distributing air gaps or reducing losses).

There is one industrial product on the market that uses multigap ferrite toroid inductors as part of an interleaved multilevel buck converter. Five or six gaps made by sawing, if I recall correctly. I'm not going to argue whether or not this is a good idea, but it does exist.

[mag_therm]

There is one industrial product on the market that uses multigap ferrite toroid inductors as part of an interleaved multilevel buck converter. Five or six gaps made by sawing, if I recall correctly. I'm not going to argue whether or not this is a good idea, but it does exist.

I have seen that made by a specialized manufacturer in USA on wound steel toroids.
A slight misalignment with the gap smaller at the outer might be beneficial.
But smaller at the inner adds to inherent toroidal non-uniformity of flux density.

[opampsmoker]

Thanks,
From  some of the kind answers, it appears that ferrite cores are just not available in large sizes. And when a large core is needed, its best to use a torroid....and a ferrite torroid cant easily be gapped, so using the "integrated gap" method of the powder core is needed.
As such, it appears that, would you agree, above approx 600W for a Boost PFC, and you are going to be almost forced to use a powder core torroid?...and not a gapped ferrite core, for the boost inductor?

[Wolfram]

I have seen that made by a specialized manufacturer in USA on wound steel toroids.
A slight misalignment with the gap smaller at the outer might be beneficial.
But smaller at the inner adds to inherent toroidal non-uniformity of flux density.

Now that I think of it, the inductor I was thinking of might have used a segmented tape-wound nanocrystalline core. It was a pretty long time ago and there was alcohol involved, so I'm not entirely sure. This would make sense, as toroids are one of the few geometries in which you can get nanocrystalline cores.

Thanks,
From  some of the kind answers, it appears that ferrite cores are just not available in large sizes. And when a large core is needed, its best to use a torroid....and a ferrite torroid cant easily be gapped, so using the "integrated gap" method of the powder core is needed.
As such, it appears that, would you agree, above approx 600W for a Boost PFC, and you are going to be almost forced to use a powder core torroid?...and not a gapped ferrite core, for the boost inductor?

No, I don't agree with any of that. Don't try to generalize, because answers are rarely black and white, engineering is all about tradeoffs. Ferrite cores are available in sizes for up to tens of kilowatts, and ferrite blocks are available that allow you to assemble cores for hundreds of kilowatts and up. Note that for larger size ferrite cores, dimensional resonances can be an issue. As mentioned in my previous post, some big PFCs use ferrite E cores, some use toroidal alloy powder cores, and some use neither. The last power converter I took apart used alloy powder E cores with edge-wound copper strip, as part of a very compact air cooled 40 kW Vienna rectifier PFC weighing about 8 kg including the casing.

[T3sl4co1l]

Interesting about strip cores, is they can be gapped nearly seamlessly.  Or more on the nose: the seam is the gap.  Consider a cylindrical cross-section: every turn of core-strip material makes a spiral outward; there's a fine gap from one to the next, through which flux must flow in order to stay radially symmetric.  (Which means there's a radial component to the flux, through the plane of the strip, which should limit the ultimate losses possible for a stripwound core, regardless of its sheet thickness?  Hmm, hadn't thought of that before, neat.  The radial component goes as the sine of the pitch angle (which is very shallow indeed), so losses should indeed go ~proportional to material thickness.  This effect is separate from skin effect acting in the sheet cross-section.)  Well, simply add a plastic film backing to the strip, and there's your air gap -- perfectly distributed, or about as perfect as you can get in such a geometry!

Might not be so reasonable in a cut core, where the faces won't line up perfectly from strip to strip; but maybe that doesn't matter so much (I rather doubt it matters much at all, but I don't have a justification for that offhand, hmm).

But not that this matters much in the present thread -- nanocrystalline materials would be reasonable I suppose, but steel is basically only relevant at line frequency (50-400Hz say).  I have certainly seen toroidal reactors, but I don't know their construction; it could well be they are cut and gapped after all, but they could be using this method instead, and should have lower losses this way.

Regarding ferrites -- you can get inches-scale bricks with ground faces, for stacking into whatever kind of build you like.  Induction heating power supplies regularly use transformers of this sort, having capacities into the MVA.  A 1MVA 20kHz transformer is only the size of a cube fridge or so (though with the mass of a compact car ).

These guys make the windings out of some combination of rectangular copper tube and plate, brazed together:
http://www.jacksontransformer.com/products-page/transformer/khz-high-frequency-transformer/
The facing plug-board is quite handy as you can select the tap (turns ratio) with just a few bolts.  It does add a few uH of stray inductance.  But so do your dick-sized litz cables, so it's not a big deal.

I'm not kidding, it's not a bad good anatomical reference, especially when the litz is the nice flexible rubber-jacketed type.    #000 AWG (metric shitton x 36AWG or finer stranding) cable is a typical choice to connect from a 600A IGBT inverter module.  Inverter modules can be wired in parallel using 0° power combiners (which are woven inline with loose toroids).

Tim

[TimNJ]

It does add a few uH of stray inductance.  But so do your dick-sized litz cables, so it's not a big deal.

I'm printing this out and framing it.

[mag_therm]

Hi Tim, I read your spiral comments with interest.

The first image below is of a double wound toroid loaded on the secondary.
A we expect, most of the action is outside the core.
Here I did not put the color graph to amplify the changes inside the core.
And it just shows H [A/m] rms value, not instantaneous.
The post processor could be re-run to show the  spatial vectors at instantaneous electrical angles.

But it does show the ripple extending all the way into the core, due to the discrete turns.

The next two images are of the Lorentz forces on the outside lamination.
In this case it is a single winding AC reactor with distributed gaps.
In this model the outer laminations and insulation gap were dimensioned according to a manufacturer's typical data.
(I recall the lams were about 0.1 mm ( 0.004 inch)
The two images are of instantaneous forces on the outer lamination at 45 and 135 degrees electrical  respectively.

The final image is of the Joule heat in an outer lamination at the gap, in the single winding AC reactor.

These models are of cores in high power electronics and I might scale them down for ferrites at small power high frequency when I get time.