Can I use common mode power line choke as 1:1 transformer in isolated buck-boost switched mode power supply at 20 - 100 kHz?

Question

I'm planning to build a multi-output power supply that allows output voltages between 0 - 30 V and currents below 5 A. The idea is that I will purchase a number of 30V 5.3A power bricks (unknown if they have floating output or negative ground) and create an adjustable isolated buck-boost converter (controlled by one 7805-supplied microcontroller per SMPS) and a linear high efficiency low dropout post-regulator to remove any remaining switching noise. The reason I'm choosing buck-boost topology is that buck converters have some voltage drop so 30V power brick could create only about 28V after the second self-built SMPS and 26V after the linear regulator, so I really need the ability to step voltage both up and down. Also buck-boost allows using 24V input power bricks as input in a pinch with full 0-30V output if they have enough current capacity. Also I'm choosing to build isolated SMPS stages because buck-boost converters convert negative ground to positive ground, which would be weird on the output, and I want the ability to use any 24-30V power brick, grounded (positive or negative) or not, and have the complete freedom to connect outputs in series in any manner, to create for example a dual-sided power supply.

I'm trying to find a suitable toroid core high frequency power transformer for my uses. The specs I need are:

• Operation way above audible frequencies
• About 7 amperes of average current handling
• About 1 mH inductance on both input and output, although this is not an exact requirement -- however, the input and output should have about the same number of turns

I think these specs are fairly odd. Most power transformers have 120-230V in input and a much lower voltage at output (but I want to use safe power bricks instead of building devices that directly work between 120-230V). Some are also intended for 50-60 Hz frequencies only. Also most inductors are just inductors, having only a single winding.

The best I could find is a toroidal common mode power line choke. Specifications here. It seems promising because the 1 mH inductance is measured at 10 kHz (although only at 0.1 mA), and because the DC resistance is very low, and because the rated current is 7 amperes. Furthermore, it has the same number of turns on both windings.

However, this doesn't seem to be an inductor specifically intended for my applications. I'm worried about the following properties:

• Is the high frequency AC resistance much higher than DC resistance due to skin effect? If the intended use case is for letting 50 - 60 Hz mains AC through and reject high-voltage common mode pulses, the skin effect probably doesn't matter at all. On the other hand, I would use PCB for the switched-mode power supply, and if skin effect is harmful on an inductor, it's probably harmful on the PCB too.
• Does the toroid core lose energy if used as a transformer between 20 - 100 kHz? (As the intended use case for this choke is in differential mode, and common mode only rejects momentary high-voltage pulses, and it probably doesn't matter if the pulses are fully rejected or only partially rejected and partially converted to heat)
• Does the inductance change much with frequency?
• Does the inductance change much with current?

Does it look like that this choke could be used as an isolated two-winding inductor in a buck-boost converter running between 20 - 100 kHz?

If not, how should I find or build a suitable two-winding inductor / 1:1 transformer?

• I could find a toroid core and wind both windings myself. If so, what are the specs I should look for in a toroid core given my application? What kind of wire should I use? Is skin effect harmful between 20 kHz - 100 kHz in wires capable of handing 7 amperes? If it is, how on Earth are all SMPSs built on printed circuit boards that should suffer from skin effect too?
• I could find an inductor intended for non-isolated switched mode power supplies and wind the second winding myself. If so, what kind of wire should I use for the second winding?

---

The circuit is essentially same as the circuit here: https://upload.wikimedia.org/wikipedia/commons/e/e6/Buckboost_conventions.svg

...but with the exception that the inductor is replaced with a 1:1 transformer, and the left side of the circuit is connected to the primary winding, and the right side of the circuit is connected to the secondary winding.

Answer 1

I can see that it is tempting to use an EMC common mode choke. I would not because the DC current rating of the choke is very low due to core saturation and often the ferrite is more lossey which is good for EMC but not as the main inductor of a SMPS. Also the common mode EMC chokes have significant leakage inductance which is generaly undesirable on hard switched topologies.

Answer 2

Hmm, it would probably work. The downsides to look out for on CMCs are:

• Core loss: they're intended for low CM voltages, just filtering.
• Leakage inductance: they're intended for some DM filtering as well, and the windings are placed far apart to maximize this (as well as making isolation easier to achieve).

The drawing is probably accurate as far as number of turns. So, five turns on whatever core that is. It's probably a VAC part, you can check their catalog for a core that would fit -- probably VITROPERM 400 to 800 -- unfortunately they don't give core dimensions, only outer dimensions after winding so it's a bit of a guess.

They do give CM and DM impedance (as a function of attenuation into 100 ohms), which suggests LL ~ 0.93µH. Note that these cores are ultimately laminated metal, so exhibit skin effect: the <30kHz region is inductance dominant; but 30k-10M, skin effect takes over, R ~= X, and the impedance is increasing more gradually (Z ~ sqrt(F)). So they tend to have high losses in this region -- still, the overall impedance is quite high, it might simply be low enough not to mind.

On the upside, there's very little hysteresis loss you need to worry about; the loss up here is essentially a linear process, eddy currents in the fine layers -- so you can calculate losses from the impedance plot. Somewhat roundabout, but representative.

As for the lower-frequency limit, that's determined by permitted maximum flux density, and acceptable core loss. Note that, since Z ~ sqrt(F) in the skin effect region, for a given applied voltage, it's better to run at higher frequency, to get lower core loss. Low frequencies may end up too lossy. But you also can't run too high, because LL will dominate (unless you move to a resonant topology to harness that inductance). Anyway, flux density for these materials is around 1.2T (nice and high: triple what you'd get in a ferrite the same size), but for core area, you need to find a datasheet and see.

As for the wire, sure, skin effect is a concern. It's solid wire, it's only intended for AC mains frequency, basically DC. It will get much hotter at 10s or 100s kHz, at similar currents. Note that you're asking for 30V 5A DC, in a half-wave topology, so half the time no current flows and thus 10A must flow during the on-pulse. There will be additional peak current due to the current ramp, due to series inductances.

As for topology -- as mentioned by others, "isolated buck boost" (aka flyback with a high-side primary switch) is just completely out. The discussion of loss should be enough hint already that this is impossible. The permeability of these cores is extremely high, giving saturation current of 10s of mA (probably 50-100 for this core) and storing essentially no energy.

You need a forward converter, preferably a full bridge topology. (Push-pull is out, without adding more windings; half-bridge would result in half output voltage. Likewise, you need a full-wave bridge rectifier on the output, costing additional voltage drop.) Such parts could still be used this way -- though again, LL is a problem, and likely limits maximum power output for square-wave converters like this. (1uH at 30kHz is 0.19Ω, or 1V drop at 5A, so it seems feasible still.)

If you're open to winding your own transformers, consider a twisted-pair construction (reduces leakage inductance; it can be reduced further by using better pairs like star quad, or multiple pairs in parallel). Ferrite toroids are readily available from the usual distributors, and parts with mu ~ 2000 are typical for power conversion.

Figure maximum flux density of, near saturation is acceptable at the lowest frequencies (low 10s kHz), say Bmax = 0.3T. Bmax drops with frequency, say 0.2T should still be okay at 100kHz, and maybe 0.1T at 300kHz, etc. (Above maybe 200kHz, you'd want to look for lower loss types; at lower frequencies, any generic power ferrite will do.) Flux density tells you how much, well, flux, you get per cross-section (Ae), per turn. Flux is simply voltage applied for some time. Use: $$ N = \frac{V}{4 B_{max} A_e F} $$ to solve for turns.

And then, obviously you can only get so many turns of so much wire onto a core; 16 or 18 AWG (or regional equivalents thereto) should do, preferably stranded (litz) to deal with the AC current; you need enough winding area (Aw) to fit all that wire. (Don't forget to count twice, because primary and secondary carry equal and opposite currents!)

If you're heart set on flyback, then you need a gapped core. A ferrite spool style inductor, or ferrite shapes (e.g. ETD) with generous air gap / shimming, are recommended. Powdered iron toroids are too lossy for flyback, for the most part. (You can run flyback with deep CCM (ripple fraction < 20% say), but control is more difficult as peak-current-mode controllers (primary side sensing) like UC3842 are unsuitable.)