Common mode choke sizing and saturation

[mrinalmani]

Hello everyone,
We're designing a 1KW AC-DC converter that would run at 70KHz to 100KHz range. I am completely unsure about sizing the common mode choke.
- Does it need to attenuate frequency in the 100KHz range or in the MHz range. I mean what is the primary frequency range of concern. Is it the switching frequency range or is it the MHz range?
- How much common mode current should it be able to handle without saturating? 10mA, 50mA or say 100mA? Is there some general rule for this?
- How much inductance do I need? Most application notes that I have encountered have CMC in the range of 3mH to 5mH.
- Last but important, if I wind a CM choke on a toroid, should I assume that the flux in the core under normal conditions is absolutely zero? Or some percentage will not cancel out? And what decides how much will not cancel out?
Thank you

 

[aliarifat794]

Select a core material with high permeability and high saturation flux density to ensure effective attenuation at both the switching frequency and higher harmonics.

 

[Magnethicc]

1. usually CM Chokes filter out noise in the MHz range.
2. usually CM Chokes cores are powdered alloys which does not saturate easily so a few mA of CM current will not saturate it not even close.
3. when in comes to the CM inductance then the higher the better just watch out of thermal behaviour.
4. since the same current goes in the first winding must return to its source then the same current goes in the second winding no net flux is *IN THE CORE*. there is a leakage inductance especially with toroid cores so there will be a net flux outside the core.

 

[mtwieg]

- Does it need to attenuate frequency in the 100KHz range or in the MHz range. I mean what is the primary frequency range of concern. Is it the switching frequency range or is it the MHz range?

This is something every engineer has to determine for themselves, usually after doing some precompliance conducted emissions testing.

- How much common mode current should it be able to handle without saturating? 10mA, 50mA or say 100mA? Is there some general rule for this?

Good question. The choke must be able to handle whatever earth leakage current you might see in the real world. Safety standards typically limit this to 10mA (refer to your relevant safety standard). Unfortunately, choke manufacturers typically never specify saturation behavior at all. I'd be surprised if 10mA were an issue, but have never verified this.

- How much inductance do I need? Most application notes that I have encountered have CMC in the range of 3mH to 5mH.

See first answer.

- Last but important, if I wind a CM choke on a toroid, should I assume that the flux in the core under normal conditions is absolutely zero? Or some percentage will not cancel out? And what decides how much will not cancel out?

This will depend mostly on core permeability. I would generally expect coupling to be 0.99 or higher.

Most engineers never even consider core saturation in common mode chokes. One tip I have is to never try and put multiple common mode chokes in parallel, either to increase the overall current capacity, or to handle additional power rails/phases. Doing so tends to result in unequal current distribution between windings, which will quickly saturate both cores. All windings must be around the same core to avoid this.

 

[KlausST]

Most engineers never even consider core saturation in common mode chokes.

True. for it to happen you need to staisfy both requirements
* the current in the two lines needs to differ. How?
* the frequency needs to be low enough to go beyond the I x t saturation limit.

If there is a current difference in low frequency ... it will trip the RCD.

Klaus

 

[mtwieg]

True. for it to happen you need to staisfy both requirements
* the current in the two lines needs to differ. How?

Earth leakage current is the most obvious source.
Beyond that it can increase further due to some novel aspect of the design (for example putting multiple chokes in parallel). I've seen some pretty wacky mistakes PDU/emi filter designs...

* the frequency needs to be low enough to go beyond the I x t saturation limit.

You mean v*t ?

If there is a current difference in low frequency ... it will trip the RCD.

Agreed, assuming there is an RCD, and the choke is implemented in a "normal" way.

 

[KlausST]

You mean v*t ?

Yes. Thanks for pointing it out.

Agreed, assuming there is an RCD, and the choke is implemented in a "normal" way.

Where I live RCDs are mandatory.

What is a "not normal way"?

Klaus

 

[cupoftea]

With common mode choke, you must have the coils separated to either half of the torroid, otherwise the emissions
will just couple from one coil to the other and you will have a worse common mode choke especially
at high frequency...and then when you separate them you will get high leakage L.

You should "wind and measure" for the inductance values.

But you sound very conscientious...most just shovel something in. And, there is no real way of calc'ing
exactly what combo of choke and y caps you will need. Common mode noise filtering is a little bit of a "try it and see" thing.

I once worked with an EE that made SMPS for the UK military...and they said that adding CMC's , y caps, ferrite beads etc, for common mode
noise fighting, is "Like chucking currants into a cake mix"

But with 1kW I reckon you will need two common mode chokes of some 10mH+ in cascade. Specially if is is hard switched SMPS....and you will
need even more common mode filtering if you have fets screwed to earthed metal heatsinks.

If you don't like common mode choking...then remember that adding shield windings in the transformer also reduces common mode noise.
Also, nice tight layout. Also, slowing up the fet switching transitions is good at reducing comm mode noise.

Some people have one big dirty great common mode choke..and then they cascade this with a smaller one which has its windings (within
each coil) well separated so as to reduce inter-winding capacitance...this gets rid of the high frequency common mode emissions
quite well.
..Especially if on a NiZn core...because Nizn handles higher frequency then Mnzn.....typical MnZn core for torroid comm mode choke isn't too good....
and most offtheshelf CMCs are made of MnZn. And in fact...it is a fact that above a certain (quite low) frequency, if wound on Mnzn, then
it will just be like an air core inductor!...and you wont get any of the nice common mode effect at all.

 

[mrinalmani]

Thank you everyone for the reply.
So the first question is clear that there is no simple theoretical method to approximate the common mode choke sizing. It is also apparent that hard fired topologies need more aggressive CMS and more so if the FETs are coupled to the heatsink.
As for the saturation, I understand that most engineers may not worry about saturation current because they are using off the shelf design. We have to wind a CMC on a toroid. We already have purchased toroid winding machine. One of the options available is T2615 core (26mm OD and 15mm thickness) with 10K relative permeability. This gives 15uH per turn. With 18 turns this gives around 5mH, no turns overlapping. Calculations show (B = LI/NA) that the core would saturate at 100mA CM current. Now if saturation was not an issue then I would use an thinner core, say 8mm instead of 15mm. Or should I consider that the currents in the two windings will not generate the exact same magnetic flux, perhaps due to winding geometry mismatch, and some residual flux will circulate in the core. This will limit the use of thinner cores. Question is what could lead to imbalance in flux apart from geometry mismatch. Of course I am talking about design parameters and not actual CM noise. Do you think leakage inductance, say 2% will lead to 2% residual flux?

 

[cupoftea]

The flux of the leakage inductance will not saturate the core...it is not coupled into the core as such.
The leakage inductor is like an air cored inductor.
Basically if your coils wire is ok to handle the rms current that it will see, then you shouldnt need to worry about saturation.

You tend to need a high value common mode choke inductance, because the C that you have is the poxy little y cap, which are usually only 1nF or so....so to get good attenuation...you need the L to be big.
You can get common mode noise right down at 150kHz.......it is often common mode noise that is of several MHz in frequency but it bursts on and off in packets at 150khz....so the low value frequency of common mode emissions is thus explained.

And tiny 1W SMPS can cause masses and masses of common mode noise....specially if eg a 1W high voltage (mains) switcher with a switching node right near an earthed heatsink...and a very fast fet turn on, with super high dv/dt switching node.

 

[mtwieg]

Where I live RCDs are mandatory.

But CM chokes are used for more than just mains emi filters.

What is a "not normal way"?

Basically gross design errors. True story: somebody thought it would be a good idea to use a three phase choke for a single phase power entry. They also had the idea to connect two of the windings in parallel to decrease resistance, but did so with opposite phase. Result: core always saturates.

With common mode choke, you must have the coils separated to either half of the torroid, otherwise the emissions
will just couple from one coil to the other and you will have a worse common mode choke especially
at high frequency...and then when you separate them you will get high leakage L.

This is not why coils are separated. It's to help meet creepage/clearance requirements.

 

[KlausST]

Hi,

But CM chokes are used for more than just mains emi filters.

OP:

Hello everyone,
We're designing a 1KW AC-DC converter

I hope you don´t mind me keeping on the OP´s topic.

Basically gross design errors. True story: somebody thought it would be a good idea to use a three phase choke for a single phase power entry. They also had the idea to connect two of the windings in parallel to decrease resistance, but did so with opposite phase. Result: core always saturates.

no comment..

Klaus

 

[mrinalmani]

Thank you everyone for the feedback.
After some searching and also taking help from Chat-GPT, it appears that CM chokes are designed to saturate between 2% to 4% of rated current. For for a 1KW converter running from 230V, the current is around 6A RMS and 8A peak. Accordingly the saturation current should be between 160mA to 350mA. The reason I found for leaving margin for saturation current is because of geometry mismatch and of course actual CM noise. I also found a datasheet of a CM choke that mentioned maximum mismatch between winding = 2%.
Although N is same in both windings, the spatial distribution of each turn may not be exactly the same in both the winding and therefore the flux linked by each winding may differ slightly.
For now I am winding T2615 toroid, ui = 10K
N = 18, wire thickness = 1mm diameter. This should give approximately 5mH with saturation current of 100mA

 

[mtwieg]

After some searching and also taking help from Chat-GPT, it appears that CM chokes are designed to saturate between 2% to 4% of rated current.

Absolute nonsense.

If I take a CM choke and replace the core with a different material, current rating will not change, but saturation current may change drastically.

If cut the wire diameter in half, the current rating will drop but saturation current will not.

There is no direct relationship between rated current and saturation current.

chatgpt is not designed to give correct answers. It's designed to provide answers that convince a majority of its audience that it's correct. On special topics like power electronics, it should never be trusted.

 

[danadakk]

chatgpt is not designed to give correct answers. It's designed to provide answers that convince a majority of its audience that it's correct. On special topics like power electronics, it should never be trusted.

Its a learning system, so users, such as yourself who have accurate and broad knowledge, must be and should be part of the training process.
Big data is the underlying principle, and provides ultimately (when trained/corrected) outstanding and fascinating results. I started using
simple big data techniques in mid 70's, and it found relationships/clues/answers I did not know existed. Semiconductor processing/manufacturing.

AI is a modern day encyclopedia.. We are the trainers. The medical field exploding in solutions now for doctors, and so many other
fields. My primary concern is it will figure out who is doing all the destruction on Earth and conclude humans the primary danger
to the machine, the planet, and then conclusion is......bye bye humans,

I do sims a lot, and imagine if we release it on model creation, feed it design and process info, ultimate result I think is much better
more complete models and jillions of them in short order.

Ref material CMC designs for OP :

https://abracon.com/uploads/resources/Common-Mode-Chokes-Basics-and-Applications.pdf

https://www.mouser.com/pdfDocs/doc191_CMFiltDesign.pdf

http://inductor.thayerschool.org/papers/ias2007a.pdf

Regards, Dana.