Common mode noise mitigation in offline power supplies

[treez]

It won't cause one mode to transform into the other.

...Thanks, I agree with this and always have agreed with it, i am not sure why you assume i wouldnt agree with it.

I think this is the crux of the matter, from my above post

the thing is, noise that couples out of the product can couple back through a y capacitor and go to neutral and go through the second half of the common nmode choke.....and then it would go back via neutral, and it would be deemed diff mode noise.

..i think this is the key point, (from my post #18 above) unless this one is ironed out i think we are talking at crossed purposes. do you agree that this can occur?...or are you saying that this cannot occur?..i think this is the defining point.

 

[mtwieg]

...Thanks, I agree with this and always have agreed with it, i am not sure why you assume i wouldnt agree with it.

Because you said this:

Common mode noise is reduced by having a common mode choke…..the common mode choke “forces” noise which has coupled out of the product to come back into the live or neutral wires, via the common mode chokes other half……so it makes the noise into differential noise, which gets acted on by the differential filter.

I think this is the crux of the matter, from my above post


..i think this is the key point, (from my post #18 above) unless this one is ironed out i think we are talking at crossed purposes. do you agree that this can occur?...or are you saying that this cannot occur?..i think this is the defining point.

That post doesn't even make sense. Y capacitors force conducted EMI from the load side back into earth, away from the mains. If the filter is balanced, then common mode signals do not see hot or neutral as being different.

Try taking the figure posted earlier and drawing your separate common mode and differential mode signal sources on the load side, then go from there.

 

[treez]

Thanks, the whole process of common mode emissions is somewhat steeped in the “magic” of RF engineering…..recently, we had a problem where common mode noise was causing a big problem at around 150kHz in our offline 700mW buck converter.
The insulator pad between the noisy drain node and the earthed heatsink could only have been presenting about 100pF to the noise, but somehow we got a bad problem around 150kHz. As you know, 100pF has an impedance of 10600 Ohms at 150kHz.
What is the modus operandi of this?
Is it because the common mode noise was actually at a much higher frequency but this higher frequency was somehow modulated by a lower frequency, ie 150kHz?

 

[mtwieg]

Thanks, the whole process of common mode emissions is somewhat steeped in the “magic” of RF engineering

I disagree.

As you know, 100pF has an impedance of 10600 Ohms at 150kHz.

100pF is very substantial. 10K of impedance sounds high for an interference source, but it also has an amplitude in the hundreds of volts. You still need to attenuate it quite a bit to meed EMC specs.

Switchmode Power Supply Handbook by Billings has a very good description for how to analyze and design CM filters. I presume many other references use similar approaches as well.

 

[treez]

OK thanks,
Do you believe that page 17 of this….
**broken link removed**
..which states that..

Above
1 MHz, current emissions which exceed the desired specification
are usually common mode emissions caused by either ringing
waveforms identified earlier or resonances caused by parasitic
components themselves.

is at best “off the mark”?

Do you agree….it should say that “above 1MHz, conducted emissions are predominantly common mode, and below 1MHz, they can be either differential mode or common mode”

- - - Updated - - -

Also, isn’t it odd that when you have a common mode emissions problem, the live and Neutral EMC scans can be exactly the same as each other in amplitude, all along the frequency range?

I mean, due to the nature of common mode emissions, you would expect the live and neutral EMC scan plots of offline power supplies with common mode emissions problems to be different?

 

[mtwieg]

OK thanks,
Do you believe that page 17 of this….
**broken link removed**
..which states that..


is at best “off the mark”?

Do you agree….it should say that “above 1MHz, conducted emissions are predominantly common mode, and below 1MHz, they can be either differential mode or common mode”

I don't really like to generalize these things too much. The amount of emissions you see will depend on a lot of factors such as converter topology, operating frequency, transformer design, heatsink design, etc.

Also, isn’t it odd that when you have a common mode emissions problem, the live and Neutral EMC scans can be exactly the same as each other in amplitude, all along the frequency range?

I mean, due to the nature of common mode emissions, you would expect the live and neutral EMC scan plots of offline power supplies with common mode emissions problems to be different?

No, this is expected, since common mode is by definition equal on both lines.

 

[treez]

No, this is expected, since common mode is by definition equal on both lines.

Thanks, so in that case, if you have only a differential mode conducted EMC problem, then you expect your EMC scan plots of live and neutral to be different?

 

[mtwieg]

Thanks, so in that case, if you have only a differential mode conducted EMC problem, then you expect your EMC scan plots of live and neutral to be different?

That depends on how the data is acquired. If you've measured the current one each line individually, then converting those measurements to differential/common mode components requires that you add/subtract them, which means the phase of the measurements must be preserved. If you measure each line with a spectrum analyzer which only returns magnitude plots, then you can't really determine whether the interference is common mode or diff mode. Refer to your relevant regulations for details.

 

[treez]

Thanks, i presume the need to pass EMC in both live and neutral is the reason why we see inductors in AC filters in both the live and neutral lines?

 

[FvM]

if you have only a differential mode conducted EMC problem, then you expect your EMC scan plots of live and neutral to be different?

Basically not. Presuming a standard LISN setup with separate L and N decoupling and pure DM versus CM interferer, you get equal magnitudes in both lines. Superimposing CM and DM interferer or asymmetrical ground impedance results in different magnitudes.

I believe, the simple equivalent circuit illustrates why.

 

[treez]

Basically not. Presuming a standard LISN setup with separate L and N decoupling and pure DM versus CM interferer, you get equal magnitudes in both lines.

Thanks, so if its pure DM interference, and no CM interference, then Live and neutral scans would be the same?

- - - Updated - - -

Also, what is the ultimate reason for putting diff mode inductors in both live and neutral lines?, it surely cannot be about reducing diff mode emissions?
An inductor (well known inductor manufacturer) apps engineer said the reason for L1 and L5 is to help reduce common mode emissions

 

[FvM]

In your schematic, L1 and L5 are required to achieve any differential mode attenuation at all. The ideal common mode choke has zero differential mode inductance due to K=1.

In a real world, common mode chokes have at least some percent of common mode inductance as differential mode (leakage) inductance. It's usually increased by separating the windings. The practical reason for having L1 and L5 is
- better interference attenuation by multi stage filtering
- increasing attenuation at higher frequencies by supplementing inductors with higher SRF.

 

[treez]

The practical reason for having L1 and L5 is
- better interference attenuation by multi stage filtering
- increasing attenuation at higher frequencies by supplementing inductors with higher SRF.

Thanks, i am wondering why the top Europe apps engineer for xxxxx company (a very respected company) told us that the L1 and L5 inductors help reduce common mode noise?

- - - Updated - - -

The practical reason for having L1 and L5 is
- better interference attenuation by multi stage filtering.........

Thanks i agree, but for costs sake, i think you would agree that a single bigger inductor, in either live or neutral only , would be preferable?

 

[mtwieg]

Basically not. Presuming a standard LISN setup with separate L and N decoupling and pure DM versus CM interferer, you get equal magnitudes in both lines. Superimposing CM and DM interferer or asymmetrical ground impedance results in different magnitudes.

I believe, the simple equivalent circuit illustrates why.

View attachment 139163

Purely CM interference and purely DM interference would show up the exact same results in magnitude measurements on each line. When both CM and DM interference sources are present, then you will see different spectrums on L and N, however it's not possible to determine the relative contribution of CM and DM interference. In some tests, a measurement with a combiner can be used to extract just the common mode component, as discussed here.

- - - Updated - - -

Thanks, i am wondering why the top Europe apps engineer for xxxxx company (a very respected company) told us that the L1 and L5 inductors help reduce common mode noise?

They will have a common mode impedance (equivalent to L1||L5) which will act to filter CM signals, but usually their CM impedance will be much less than that of the CM choke, so their impact is much less.

Thanks i agree, but for costs sake, i think you would agree that a single bigger inductor, in either live or neutral only , would be preferable?

Having asymmetric filter design would bring up the possibility of differential interference being transformed into CM interference, possibly making some EMC regulations harder to pass.

 

[treez]

100pF is very substantial. 10K of impedance sounds high for an interference source, but it also has an amplitude in the hundreds of volts. You still need to attenuate it quite a bit to meed EMC specs.

Thanks, yes, the thing is, it still surprises me that we get such a high common mode peak at around 150kHz. This is because as you know, capacitive current flow (in eg the stray capacitance) goes as i = C.dv/dt
…for 150kHz, the dv/dt is not very quick……and in conjunction with such low stray capacitance, I am surprised we have this high common mode peak at 150kHz.
I would have thought that the switching edges, the nanosecond rise times of the LNK302 based buck converter would have been the main causators of our common mode noise problem, and would surely put the mains common mode emissions problem high in the MHz region?...not at 150kHz.

 

[treez]

I actually believe that the reason that our EMC scan has a peak due to common mode noise down at ~180kHz is because 180kHz is the “envelope” of the third harmonic of the switching frequency of the LNK302 Buck converter. Do you agree?
The common mode noise is caused ultimately by high dv/dt switching node transitions, and this represents a frequency of several MegaHertz…but the reason why we don’t get the peak in common mode noise at several MegaHertz is because it has this “envelope” of the switching frequency (60kHz) and its harmonics. Hence our Common mode noise problem appeared to manifest itself at 180kHz, when in fact, a 180KHz sine wave voltage would not have , in itself, caused anything like as bad a common mode noise problem as what we are getting.
Do you agree?
Henceforth, the way to tackle our common mode noise problem is to tackle the high MegaHertz frequencies, and not to put big filters in to tackle the 180kHz "problem".....the 150kHz part is just the "envelope" of the higher frequency problem?...do you agree?

 

[treez]

Having asymmetric filter design would bring up the possibility of differential interference being transformed into CM interference, possibly making some EMC regulations harder to pass.

..(from post #34 above.)
Thanks, so that is a declaration that DM noise can be transformed into CM noise....do you think that the reverse can happen?...ie CM noise getting transformed into DM noise?

No, this is expected, since common mode is by definition equal on both lines.

(from post #26 above)
Thanks, do you mean that CM noise is equal in phase and equal in magnitude in each line......?
Also, is DM noise opposite in phase and equal in magnitude on each line ?(L and N)

Y caps filter common mode noise because they are connected to earth ground.

(from post #14 above)
Thanks, but isn’t it strange that it is the stray capacitive coupling between the circuit and earth ground that causes common mode noise in the first place?
In other words, capacitance_to_earth (the stray capacitance between the circuit and the earthed heatsink) has caused common mode noise…..and is then being used (as in Y capacitors) to solve the common mode noise problem. Isn’t this a kind of slight contradiction?
:-|

Y caps filter common mode noise because they are connected to earth ground.

again from post #14 above.
Thanks, so you are saying that we can use Y capacitors from the circuit to earth ground to filter the common mode noise…..and presumably, the more capacitance we have from the circuit to earth ground, the more filtration we apply to the common mode noise….which is surely a good thing.
In order to really filter the common mode noise really well with capacitors to earth ground, shouldn’t we connect the Y capacitors from the switching node of the SMPS to earth ground?…after all, that way we will get much of this dreaded common mode noise to pass through the Y capacitance, and thence it will get well filtered by this extra Y capacitance that we have added in order to filter common mode noise.?
:-o
The thing about this, is that it is common in SMPS ( especially SMPS that sit on an earthed heatsink) , when common mode noise filtering is being done, -to actually connect Y capacitors from the earthed heatsink to a “quiet node” on the circuit (ie not the switching node).
How is connecting a Y capacitor from a “quiet node” on the circuit, to the earthed heatsink, going to filter common mode noise? I know this isn’t your statement, but as we know, it’s a common thing that gets done.
….Why does it get done?, I mean, the stray capacitance from the circuit to the heatsink is what caused the common mode noise problem in the first place, so why is more capacitance (ie Y capacitors) added from the circuit to the earthed heatsink in order to solve the common mode noise problem?

 

[treez]

What is clear is that there are 2 distinct uses of Y capacitors for mitigating common mode noise...

1...In "filtration" (LC filter attenuation of the common mode noise in conjunction with a common mode inductance)
2....In "diversion".....diverting emissions away from the LISN so that they can't be detected as common mode noise......ie emissions which would otherwise go back through the LISN are "Invited" to go back through a y capacitor , back on to the PCB from which they originated and run in a loop via that.....thus the emission is diverted back out of the earthed heatsink and does not go back through the LISN.....which mitigates common mode noise.

-This "diversion" is a purpose for which Y capacitors are often not given credit....do you agree?

 

[mtwieg]

..(from post #34 above.)
Thanks, so that is a declaration that DM noise can be transformed into CM noise....do you think that the reverse can happen?...ie CM noise getting transformed into DM noise?

(from post #26 above)
Thanks, do you mean that CM noise is equal in phase and equal in magnitude in each line......?
Also, is DM noise opposite in phase and equal in magnitude on each line ?(L and N)


(from post #14 above)

Review FvM's figures in post #30, they describe things as simply as possible.

Thanks, but isn’t it strange that it is the stray capacitive coupling between the circuit and earth ground that causes common mode noise in the first place?
In other words, capacitance_to_earth (the stray capacitance between the circuit and the earthed heatsink) has caused common mode noise…..and is then being used (as in Y capacitors) to solve the common mode noise problem. Isn’t this a kind of slight contradiction?
:-|

What is "the circuit?"

Y capacitors help because they allow the capacitively coupled signals to return to their source inside the PSU, rather than through the AC cable and source.

- - - Updated - - -

-This "diversion" is a purpose for which Y capacitors are often not given credit....do you agree?

This is a false dichotomy. Any filter has both shunt and series elements used in combination, and we just call that a "filter."