Electrical – Why are there different ways of winding a common mode choke

Even though this questions looks like very specific, it can be treated indeed as a much more general case filtering question: "How can one filter out electrical noise coming from power electric motors?".

The first information data we need to gather in advance is the type of noise our circuit is exposed to. Sometimes it is really difficult to get this data in advance, sometimes it is even harder to measure the noise without prior experience and high-end laboratory equipment.

In general, we can assess our noise sources in terms of:

• Intrinsic or extrinsic. I.e.: does the noise comes/is generated inside our own system? Or does it comes outside of our system?
• Coupling mechanism: capacitive coupling, inductive coupling, ground loops, EM radiation...
• Characteristics of the noise: switched, thermal (gaussian), shot, flicker...
• Frequency band and Q. How narrow or wide band is our noise? Does it fall/dissapears abruptly outside that band (quality factor)?

The above is a partial list, incomplete, which may serve as a starting point only.

Then, there are a lot of techniques, I mean literally hundreds of tricks and broader approaches depending on the case.

Delving into the specifics of the original question, this is my best guess on the sort of noise that may be originated by the system,

1. The noise is coming mainly from system itself, power motors and driver circuits. 30A of peak switching current is high enouch to generate pulses which can easily couple to the rest of the circuit.
2. Capacitive coupling, inductive coupling and ground loops can be all source of trouble here, due to the high current pulses of the drivers.
3. The noise is switched, I guess in the sub 1MHz region, however, armonics in the 1-10MHz range could be easily generated/radiated.

Some practical hints and techniques for dealing with the noise in the system above:

• If possible, separate physically the motors and drivers from the rest of the circuits. This is obviously not possible in all cases, for instance, if you have a single board for all the electronics. However, if you can afford having two separate boards, one for driving the motors, another one for the rest of the system, it is helpful doing so.
• Avoid ground problems and loop coupling of the noise by using a carefully thought star ground connection for all your circuits, including power drivers, batteries and chassis.
• Do not let any chassis or big metallic part floating, as this will interact with the EM fields generated by the motors and power drivers, reflecting, propagating and/or re-emitting the EM fields as additional noise.
• Regarding the motors themselves, and depending on the type of motor, you can certainly apply noise filters near/attached to your motors. For DC motors, which may not be your case, it is wise to solder small ceramic capacitors across each phase, as near as possible to the motor. Rugged (high voltage) 0.1uF capacitors are a good rule of thumb to start with. Depending on the application, you could also add another pair of ceramic capacitors from each of the phase leads to the chassis. Beware of checking the exact motor type and driver before going this route.
• The cabling connecting the drivers and motors should be as close as possible and be twisted.
• Decoupling/bypass capacitors should be generously added to your driver power lines, in two flavours: bulk capacitors (maybe in the hundreds of uF, for low frequency filtering) and high frequency capacitors (typically 0.1uF).

Returning to the circuit you posted, my initial approach would be:

• Not using a common-mode choke, as it is more indicated for capacitive coupling noises generated from outside of your system.
• Applying dual LC filtering for both lines (power and GND return) or even better, a dual L pi filter. This is the most effective filter for KHz to low MHz noise. A big inductor (in the mH range) in series with each of the battery terminals will improve dramatically the noise entering the digital part of your circuit. Ferrite beads, on the contrary, are dissipative by its own nature and best suited for higher (dozens of MHz frequencies).
• Substituting the standard zener & unidirectional TVS for a bidirectional rugged (high energy) TVS. The zener in your circuit can be kept, however, if your input regulator cannot withstand small peaks of overvoltage.
• Adding a pair of small ceramic capacitors in parallel with the bulk capacitor: for instance 1uF and 0.1uF MLCCs, rated conservatively (>100V). This will increase your filter effectiveness for higher frequencies (>1MHz).

Last but not least, devise a simple way to measure your circuit at critical points, in order to verify the effectiveness of the different approaches. Do, please, try to test under similar circumstances as the real device will operate under.

If neeeded, I can provide more references (books, articles) to the approaches aboves. If you can specify in greater detail some parts of your system, additional filtering techniques will surely apply.

Maybe. But the probable answer is 'not for your application', unfortunately. And furthermore, a better question might be, "Should I use a common mode choke instead of a coupled inductor?" And the answer to that question is always no.

Common mode chokes typically have two ratings, the differential mode current, and the common mode current. If you have found some massive choke the size of an obese house cat that 'has the right specs' (it has the high common mode current that you need), then sure, it will 'work'. It will certainly not be the best choice, but it will work.

If you're looking at the current ratings of common mode chokes and describing them as anything like 'high' or the unit is whole amperes, then that is the differential mode current rating. This rating is meaningless for any applications where one would use a coupled inductor. That current rating is a rating how many amperes of differential mode current can be handled. There is very little (in other words, there is current that is perfectly balanced, equal, in phase between the windings, but in opposite directions, like the power and ground return for it for example).

This current cancels its own magnetic flux, so it will only see the leakage inductance worth of inductance. In other words, the differential mode current is the maximum current only if you're not actually 'using' the inductance. It's essentially resistance limited. Because its the rating for currents that don't store energy magnetically, and keep in mind, inductance is a measure of energy stored in a magnetic field.

What is important for SEPIC, or really any circuit that uses an inductor for, well, its inductance, is that inductor's saturation current. This is the maximum current that the magnetic core can withstand before some chosen drop of inductance occurs (20-30% is frequently used as this drop). Or put a different way, saturation current is how much energy the magnetic core can store in a magnetic field before it's 'full'. When it's 'full', then the magnetic core isn't unable to store more energy, so increasing current beyond this point will quickly only store as much additional energy as an air-core, which presents as a rapid loss of inductance.

How this actually plays out is highly dependent on the core material. Ferrite of all kinds saturate like hot jello hitting a silicon carbide wall at 100mph. Going too close to the saturation point is simply not done, too risky, and the drop-off is too sudden.

Iron powder cores, or my favorite, carbonyl iron, saturate linearly, so you'll still have 40% of the inductance left even after you've gone to twice the saturation current. You will also have ridiculously high core losses using a powder core like that at any useful frequency, but it can be useful for peak currents in certain situations.

Chokes could be either core, both types get used for common mode chokes frequently. But it doesn't really matter, because there aren't common mode chokes rated for high current at 1mh. There are no chokes that meet your specifications - because you're using the differential mode current rating as if its the saturation current, and it isn't. A large 1mh choke rated for anything more than hundreds of milliamps saturation/common mode current would be extremely fat house cat sized (as mentioned earlier). 1 millihenry is a ton. You need sheer milliliters/inches^3 of magnetic core material if you want to store that much energy. No way around it.

Take this guy for example. It's already a behemoth at least for PCB-level stuff, and it has your inductance rating and no, it will not handle 16A of current before saturating. It will handle 240mA. For use as a coupled inductor, it's peak current 240mA. I wouldn't call that 'high current', but you didn't really mention what sort of currents you needed, so maybe that is enough. Probably not though.

This brings me to what is not going to be the answer to the question you asked, but the answer you need. I highly doubt you will find a cheap, mass-produced coupled inductor (or a choke that can be mis-used as one) that meets your specs. If you actually need 1mH at 10A or whatever you have in mind, then expect to have such a thing custom made, and expect it to be very expensive.

The reason there aren't any is because there is no need for coupled inductors that large and no reaon to mass produce them and get the cost down like chokes and reasonable coupled inductors. What I am trying to say is, if you think you need a 1mH high current coupled inductor, then your design is inherently flawed. The only reason I think think of that would require so much inductance is you want to convert currents that are far too high for a switching frequency far too low.

That design is wrong. There is no reason to do that. I suspect you have chosen some specific controller or driver that has a relatively low switching frequency, and want to build a inferior and impractical DC/DC conveter at extreme cost and no advantage, save for you don't have to learn how to use a chip that is actually appropriate for your end-goal. I suspect this because I've been there, we probably all have at some point. I make no judgement here, and freely admit to being guilty of it in my past. And what I know now is that if you think you need such a large inductor and at high power, then you don't know enough about switching converters to build one that high power.

Don't give up that goal, but work towards it by taking some intermediate steps and make smaller stuff. Learn your way around a bunch of topologies and controllers. Figure out how to select your own mosfets. Learn why electrolytics are just glorified resistors above 100KHz, or what happens to class II ceramic capacitors under DC bias (hint: they lose capacitance. Sometimes most of it. Fun! =P). Learn why you'll be optimizing a layout for every millimeter and how much a couple of nanohenries of parasitic inductance can cost you. Learn how to snub switch nodes' ringing voltages. Most of all, learn why a SEPIC converter is neither appropriate or needed for something high power that can have an input above or below the output. You would do much better with a 4-switch true buck-boost.

Ignoring all that, you do not even need a coupled inductor at that - you can just use two inductors. They need not be on the same core. The only thing sharing a core gets you is reduced ripple current. Or you could do the same by doubling the switching frequency, or using two phases at the current frequency. Either of these would be much easier, cheaper, effective, and doable. In fact, doubling the frequency gets you all sorts of other good stuff too, like reduced input ripple, reduced inductance needed, lower size, lower cost.

It's not 1990, we have switching elements that can have such low losses as to have the resistive losses from higher-inductance inductors, as well as core/hysteresis losses, outweigh switching losses until the hundreds of KHz. And even then, going faster can cost you a watt or two if you're doing it right. Take a look at the LT8705 or dozens of other 4 switch buck boost converters. They'll let you do everything a sepic could do but with 10µH of inductance, less EMI, more efficiency, be smaller than a deck of cards, and can be built using components that actually exist. If you're using something that switches at 52KHz or 70Khz or 100KHz, then you're about 26-27 years out of date. If you want to convert high power levels, well, we couldn't back then easily, not without it being more expensive than simply using big iron and big copper linear transformers/boat anchors. There is a reason switch mode power supplies began appearin when the did. High power switch mode power supplies were larger than linear supplies (but possibly a bit lighter) up until relatively recently. The power density you seem to be imagining was not possible with whatever chip it is you've selected. But that's ok, there are much better alternatives now.

So, I know you never asked about that.

But if I am to give you the most useful answer that goes beyond what you actually asked it is simply that when you say you need a 1mh high current coupled inductor....no. You don't.