EMC issues: a quick overview

Every time you face a problem, you should find the root cause rather than looking for a patch. In this article I'll try to give you an idea about most common electromagnetic noise sources and how related problems can be mitigated.

EMC problems often arise from a poor layout, superficial or wrong hardware design, low-quality components choice or insufficient countermeasures adopted during the design phase. However, a lot of compliance engineers still point the finger against electronics when a problem is met, expecting that all EMC problems are solved at PCB or electrical design level. Well, this approach is correct whenever the issue is caused by a very poor electronic design but in many other cases, EMC failures are due to mechanical problems, bad shielding, high-impedance paths between metallic cases and PCBs and so on.

In this article the most common causes of EMC issues, that could lead to a failure during a compliance test, will be analysed, followed by some possible solution or mitigation.

Electrical, RF and EMC engineers identify two types of noise generated by electrical equipments: common mode and differential mode noise. They are generated in different ways and their severity depends on many aspects of both mechanical and electrical design.

Electromagnetic noise is generated by switching circuits and periodic signals whose energy is high; on the contrary, static circuits usually are electromagnetically compatible by design and, even if linear, static, time-invarying circuits generate some noise too, it is often negligible for many applications.

Electromagnetic noise can be caused by two types of modes that are identified by the sign of the current flowing between the various conductors in an electrical system; take a look to the following image.

This kind of noise is generated by circuits that sink currents from one or more poles in a time-variant way and source the same amount of current from another one or more pole. This is true for circuits where it is possible to identify inputs (both signals and/or supply lines) and they are made of just two conductors, a forward and a return path. In real-world applications it is not so easy to understand how and where these currents flow, making the path to solution quite impervious.

Common mode noise takes origin from parasitic paths between electronics and mechanics: ideally, there is no common mode noise if there is no coupling between two conductors.

Refer to the following image to better understand what capacitive coupling is.

Two conductors separated by a non-conductive material always have some sort of interaction: in many cases, one of the two capacitively injects noise on the other one, disturbing a signal, a supply or generating an antenna exploiting mechanical parts and stray elements. Capacitive coupling is something to pay attention to every time you have a circuit (typically on a PCB) face to face with a mechanical support (a metal plate, a face of the case, a flat cable...): if the circuit is switching, you can see that the stray capacitor can conduct some current, creating an unwanted parasitic path between electronics and mechanics. This is one of the most common sources of common mode noise: electronics to mechanics interfaces should be deeply analysed since the earliest stages of the design!

On the left you can see a real application circuit where a power MOSFET is used in a switching circuit. Right under the SMD component there is a metal shape used to dissipate it on a metallic plate, but this leads to the generation of a stray capacitor whose dielectric is made of the thermal pad under the board, and the metal shape on the bottom layer (connected to the component tab) with the metallic plate work as the two armatures of a planar capacitor.

When the DCDC is switched on, you'll see a lot of noise on power lines: if you perform a conducted emission scan, you'll see a lot of harmonics of the converter switching frequency on all the wires! The main cause is exactly this mechanism.

Inductive coupling is caused by stray inductances that share some coupling path for magnetic field. This means that some conducting elements are not electrically coupled like in the previous mechanism but they are so close that one of the two (the aggressor) induces some unwanted noise in the other one (the victim).

From this kind of coupling, many problems could show up: if two or more wires are hit by the electromagnetic wave, the magnetic field could pass through a turn, inducing a current in the wires; if they go out the system, they will behave like a loop antenna and the wire will suffer from differential mode noise. If, instead, the magnetic field affects all the wires, the system will suffer from common mode noise.

Noise can easily pass through low impedance paths, so, for this reason, you should always force the current to flow in the best way possible, to avoid unwanted loops or undesired coupling between conductors.

Finding a valid solution to an EMC fail can be challenging exhausting. However, there are some attempts you can do to quickly fix the most common issues; these solutions works well if the problem is exactly the one reported here and, if it is not giving good results, probably your problem is not exactly that one but maybe a combination of more phenomena. However, the path toward the solution is always the same, and can be summarized with the following procedure:

When you enter the anechoic chamber, you must already know everything about all the switching circuits in your design, their frequency, the routing of input and output traces on the PCB, filters position, topology and cutoff frequency. Only then you can perform the first scans: if you have a lot of switching circuits, you should see many superimposed harmonics of all of them, exhalted in some frequency ranges and attenuated in other ones. If you can't properly see harmonics at high frequencies, reduce the resolution bandwidth (RBW).

Once you see the harmonic content in all the sub-ranges, you can proceed to the next step.

Let's separate broadband from narrwband noise.