Bode Plot of Pi-Filter with CMC? Does this make sense to you...

[Control:Eng]

Hello,

I face some serious EMC problems in one of my applications and may need your help.

In the Recom Application Notes:
http://www.recom-international.com/fileadmin/Media/Folder-Flyer/App_Notes_16062014.pdf
on page A31 they suggest a  second order Pi-filter that uses Common Mode Chokes to eliminate Common Mode Noise...


Normally, if I think about filters I'm thinking in the frequency domain since most time you want to attenuate high frequency Noise.

So I did a simulation of the filter and LTSpice gave me the Bode Plot that you find as an attachment.

Does the Bode Plot make sense to you? Could you maybe explain me why the filter shows this response?

It seems that it introduces a pole and a zero since the phase converges to 0° if f->oo.
So at least maybe we can say that there are as much poles as zeros....

How could a transfer function of a CMC look like? I really have no clue and don't find many informations in the net.


I'm sure that I messed up some EMC things in my brain...

Maybe you guys could help me to tidy up

Thanks in advance!

[moffy]

This article: www.pulseelectronics.com/download/3100/g019/pdf
Describes common mode noise and filtering. It might help.

[T3sl4co1l]

How exactly did you model that?

A CMC filter is tested like any other: with controlled impedances and real components.  I'm guessing your circuit contains very little resistance at all, hence the high Q and low attenuation.  A typical CMC can be modeled as a nonideal transformer with a few uH equivalent leakage inductance (or k ~ 0.998, give or take), winding inductance of 100uH-10mH+ (as rated), parallel capacitance (maybe 20pF, depending), series resistance (DCR, plus AC skin effect if necessary) and parallel resistance (eddy currents and core loss, usually in the kohms equivalent).

The source is generally a capacitively coupled voltage source: internal switching nodes (via heatsinks, transformers, etc.) are the noisiest points in a switching supply, and couple across the isolation barrier (common mode).  Hardly a 50 ohm source, which is important to keep in mind as well.

Tim

[Control:Eng]

Hey,

sorry for the late response

Well I didn't model the CMC on my own.

It's the model that Würth Elektronik provided me. I'm quite sure, that the model is very accurate.

My problem is the following:

- I know that my filter is able to filter Common Mode Noise
- But is it also able to filter Differential Noise? I doubt that if I look at the Bode Diagramm?

Any thougths?

Thanks in advance!

[T3sl4co1l]

You didn't mention what part, so I'll provide a random example:
http://www.murata-ps.com/data/magnetics/kmp_5100.pdf
The 51505C shows common mode starting to drop at 1kHz, peaking to -50dB just below 400kHz, then rising more or less from there (possibly with an antiresonance at 20-30MHz... which is pretty crappy of them, since most customers will be interested in the 0.15-30MHz range -- marketing shame, anyone?).  Common mode is tested with both windings in parallel, connecting a 50 ohm signal generator in series with the choke, to a 50 ohm terminated load, which senses the attenuation.

Differential is tested with inverse signals applied to each winding, 50 ohm terminated.  The total impedance as seen by the choke is therefore 100 ohms.  The input and output voltages are sensed differentially.  The 51505C starts dropping off around 1MHz, hits a peak at 8MHz and rises from there (probably hitting numerous resonances and anti-resonances in the 20-100+ MHz range).

Note that the resonances aren't even in the same place, so you can't guess, based on common mode response, what the differential response may be.

You may also see "normal mode", which is either synonymous with differential, or measured with one winding shorted (roughly equivalent, but won't be exactly the same around the higher resonances), or one winding alone (the other(s) open circuit).  Hopefully, they provide a definition.

If you see a choke listed by impedance rather than attenuation, the difference is simply the attenuation as an impedance divider in the system impedance (50 ohms or so).

Measurements are similar for ready-made filter modules, except that it's always by attenuation, and the measurements are always relative to chassis ground (leaving three modes: common, differential, and single line).

Tim