I have seen that some switch mode power supplies are injecting noise to the electrical power network. I guess this noise is common mode noise. That switching noise then propagates through and enters to another power supply and even can appear at the DC output.

1) Regarding the switching noise originates from an SMPS, does it propagate on neutral, line or earth wire?

2) And in the affected SMPS, how does it pass this switching noise all the way to its DC outputs? From its line, neutral or earth terminal?

3) Why is this type of noise is called common mode noise?

I would like to see the actual trace/path the noise travels from the injector to the receiver SMPS.

  • \$\begingroup\$ A lot of the noise is differential; the switching mechanism in an SMPS creates a diff signal that gets conducted to the L and N wires at the top and bottom of the bridge rectifiers conduction points. Plenty of SMPS have diff filters. \$\endgroup\$ – Andy aka Dec 11 '18 at 9:03
  • \$\begingroup\$ @Andyaka Thanks for the comment but I dont understand in real. Why do you call it differential if it is appearing both at neutral and line? And where does the noise current flow through all the way from the smps to the other smps. A illustrative answer would help a lot. \$\endgroup\$ – pnatk Dec 11 '18 at 10:04
  • 1
    \$\begingroup\$ Appearing on both lines simultaneously and having the same amplitude and phase is called common-mode. If either the amplitude or the phase are unequal then a combined differential and common signal is produced. The main switching element produces a signal that is largely differential. If you don't understand this then you should do some research and then adjust your question above. \$\endgroup\$ – Andy aka Dec 11 '18 at 10:09
  • \$\begingroup\$ What I understand from your words the noise initially common at both lines but because of in balance it will become as differential. If the noise at line is 10 and the noise at neutral is 8 the difference is 2. Th CM filter still helps because if it has attenuation factor of 10 the line noise will be 1 and the neutral noise will be 0.8 resulting s differential noise of 0.2. Thats what I have in mind. Is that totally wrong. If it is I will go and do more research as you recommend. \$\endgroup\$ – pnatk Dec 11 '18 at 10:16

The simple drawing below illustrates the difference between the differential- (DM) and common-mode (CM) conducted perturbation:

enter image description here

Usually, but it is not always the case, a DM perturbation appears in the low-frequency portion in an EMI plot while CM noise is located in higher frequencies. If you go deeper in the analysis of a switching converter such as a flyback converter, you can imagine the below mechanisms pertaining to conducted perturbations:

enter image description here

In DM, it is the circulation of the high-frequency primary-side current signature in the bulk capacitor which generates the perturbation. The harmonic content depends obviously on the operating conditions and, in particular, in the operating mode, discontinuous (DCM) or continuous (CCM). You fight these perturbations with a differential-mode filter, usually built around the leakage inductance of the common-mode component.

In CM, the coupling mechanism is mainly linked to high dV/dt switching nodes and fast switching components: the stray capacitance between the MOSFET tab and the heatsink, the inter-winding capacitance of the transformer, the rectifying diode and so on. All these parasitics go across isolation barriers via stray capacitances that must be minimized during PCB layout and component construction (transformer). Switching noise on the output is inherently due to the switching nature of the rectified waveform. On top of that, you add the recovery effect of the output diode in CCM which pollutes conducted but also radiated signatures. Minimizing loops area in which high currents circulate and damping all these oscillations are key to solving EMI issues.

Finally, you measure all these conducted perturbation with a line impedance stabilization network or LISN configured in such a way that you can isolate DM and CM noises. You have to separate CM and DM as each requires a different cure. The LISN offers a constant 50-\$\Omega\$ output resistance along the analyzed frequency band. The final EMI plot to check PASS or FAIL mixes them of course.

enter image description here

  • \$\begingroup\$ @ Verbal Finally, a good explanation. With abstracted and detailed diagrams. And consistent with my own "The electric fields are the big problem." conclusion. \$\endgroup\$ – analogsystemsrf Dec 11 '18 at 15:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.