What does electromagnetic compatibility actually mean?

EMC is two promises made about the same product. The first is that it will not put more electromagnetic disturbance into its surroundings than the rules allow. That half is called emissions. The second is that it will keep doing its job when the disturbances that normally exist around it arrive at its own terminals and enclosure. That half is called immunity. A product is electromagnetically compatible when both hold at once, in the environment it is sold into. Neither half implies the other: a very quiet product can be fragile, and a very robust product can be noisy.

What is the source, coupling path and victim model?

Every EMC problem has three parts. Something generates energy (the source), that energy travels by some route (the coupling path), and something is disturbed by it (the victim). The model is useful because it tells you there are always three places to intervene, and you only need one of them to work. You can quiet the source with a slower edge rate or a spread-spectrum clock, break the path with a filter, a shield or better board layout, or harden the victim so the energy no longer matters. In a real fix the path is usually the cheapest of the three to attack, and the victim the most expensive, because hardening often means someone else's silicon.

What is the difference between conducted and radiated?

It is a question of how the energy travels. Conducted disturbance rides along a wire: a power cord, a signal cable, an interface lead. Radiated disturbance crosses open space as a field, with the cables and the enclosure seams acting as unintended antennas. The split matters because it decides the instrument. Conducted emissions are measured through a LISN clamped on the mains lead, typically over 150 kHz to 30 MHz. Radiated emissions are measured with antennas in a chamber or on an open site, typically over 30 MHz to 1 GHz and higher for fast digital products. The same underlying noise often shows up in both, which is why fixing one sometimes fixes the other.

Why do products pass pre-compliance and then fail the formal test?

Most often because something changed that nobody considered part of the test. Cable routing is the classic one. A harness dressed slightly differently between the two sessions changes the length and geometry of the structure that is doing the radiating, and several dB can move with it. The other frequent causes are a different operating mode, a firmware build with different clock behaviour, a sample with a pre-production shield or gasket, and pre-compliance equipment whose own uncertainty is wider than the margin being claimed. Pre-compliance answers "roughly where am I", not "will I pass".

What is EMC? Emissions, immunity and coupling

Author Hugues Orgitello Last updated 29 August 2026 ~6 min read

Guide, understanding EMC

Electromagnetic compatibility is two promises about one product: that it will not disturb its neighbours, and that its neighbours will not disturb it. Every EMC standard, every test chamber and every filter component exists to hold one or other of those promises. This page explains the idea from the ground up, without assuming you have met a spectrum analyser, and then points at the guides that take each part further. If you have arrived here because a test house has just sent you a failure report, the section on coupling paths is the one that tells you what to change.

EMC is two promises

Split the subject in half and it stops being intimidating.

Emissions is the outgoing half. Any circuit that switches current generates electromagnetic energy, and some of that energy escapes, along the cables or straight out through the case. Emissions requirements cap how much may escape, so that the product does not degrade radio reception, corrupt a neighbouring instrument, or raise the noise floor of the building it sits in.

Immunity is the incoming half, sometimes called susceptibility when spoken of as a weakness. The product is going to be handled by someone carrying a static charge, plugged into a mains supply carrying switching transients, and stood next to a phone transmitting at full power. Immunity requirements set how much of that it must absorb while still working.

The two are independent. A product that emits almost nothing may still reboot when someone touches it, and a product that shrugs off every disturbance thrown at it may still be radiating far too much of its own. Both must be shown before the product is compliant.

	Conducted (travels on a wire)	Radiated (travels through space)
Emissions (going out)	Noise leaving on the mains or signal leads, measured through a LISN	Fields leaving via cables and case seams, measured with antennas
Immunity (coming in)	Disturbance injected onto cables: surge, fast transients, conducted RF	Fields applied to the product: radiated RF, ESD, magnetic fields

Those four boxes are the whole of routine EMC testing. Every test on a typical report belongs in one of them.

Source, coupling path, victim

Every EMC problem, without exception, has three parts:

a source that generates the energy,
a coupling path that carries it,
a victim that is disturbed by it.

This is the single most useful idea in the subject, because it tells you that there are always three places to intervene and you only need one of them to work.

Quieting the source means reducing what is generated in the first place: a slower switching edge where timing allows it, a spread-spectrum clock to smear a sharp harmonic into a broader and lower one, a soft-start on a converter.

Breaking the coupling path means stopping the energy getting from one to the other: a filter on the cable, a shield around the noisy block, a ground plane that gives return current a tight path home instead of a wide loop. This is usually the cheapest intervention and the one most fixes actually use.

Hardening the victim means making the disturbance stop mattering: more decoupling, a watchdog, input protection, firmware that tolerates a corrupted sample. It is often the most expensive route, because the victim is frequently someone else's silicon that you cannot change.

The coupling path itself comes in four flavours worth knowing by name. Conductive coupling shares a physical connection, typically a common power or ground path. Capacitive coupling passes energy between two conductors at different voltages that are close together. Inductive coupling passes it between two current loops sharing magnetic flux. Radiative coupling is a true electromagnetic wave crossing open space, and dominates once the aggressor structure becomes an appreciable fraction of a wavelength.

Why the law gets involved

EMC is not left to good intentions, because the cost of getting it wrong lands on someone other than the manufacturer. A noisy power supply degrades reception for a whole building, and the building has no way to identify or refuse it.

In the European Union the instrument is 2014/30/EU, the EMC Directive. It applies to most electrical and electronic equipment, and satisfying it is a precondition of CE marking. See the EMC Directive guide for scope, exclusions and the conformity route.

In the United States the equivalent obligation for unintentional radiators sits in FCC Part 15. The philosophy differs in places and the limits are not identical, which is why a product sold on both sides is designed to the tighter envelope band by band rather than to one regime. CE versus FCC on EMC sets the two side by side.

Radio products do not escape any of this. Their EMC requirements are absorbed into the radio regime, the RED in Europe and the relevant FCC part in the US, but the emissions and immunity questions are the same questions.

What a beginner usually gets wrong

Treating EMC as a test rather than a design property. EMC is decided at layout, in the stackup and the return paths, long before anything reaches a chamber. A product that fails by 20 dB rarely has a filter-shaped problem; it has a geometry-shaped problem. PCB design for EMC covers the decisions that matter.

Assuming a passing pre-compliance scan means a passing formal test. Pre-compliance tells you roughly where you stand. It does not reproduce the chamber, the antenna calibration, or the measurement uncertainty that the formal result must absorb.

Adding filters at the end. A filter fitted to a cable that is radiating because of a bad return path treats the symptom at the point where it is most expensive to treat.

Confusing EMC with RF exposure. Human exposure limits are a different subject with different instruments, different bodies and different numbers. ICNIRP publishes the exposure guidelines most jurisdictions build on; EMC standards say nothing about human safety.

Assuming CE covers it. CE marking is a declaration against every applicable directive. EMC is one of them, and passing EMC does not address electrical safety, radio, or chemical restrictions.

Key takeaways

EMC is emissions plus immunity, and neither half implies the other. Both must be demonstrated.
Every problem is a source, a path and a victim, which means three places to intervene and only one of them needs to work.
Conducted versus radiated is a question of how the energy travels, and it decides which instrument measures it: LISN over 150 kHz to 30 MHz, antennas from 30 MHz upward.
It is a legal requirement, via 2014/30/EU in the EU and 47 CFR Part 15 in the US, not a quality target.
The fix is usually geometry, not components. Return paths and loop areas decide more than filters do.

Frequently asked questions

What does electromagnetic compatibility actually mean?: EMC is two promises made about the same product. The first is that it will not put more electromagnetic disturbance into its surroundings than the rules allow. That half is called emissions. The second is that it will keep doing its job when the disturbances that normally exist around it arrive at its own terminals and enclosure. That half is called immunity. A product is electromagnetically compatible when both hold at once, in the environment it is sold into. Neither half implies the other: a very quiet product can be fragile, and a very robust product can be noisy.
What is the source, coupling path and victim model?: Every EMC problem has three parts. Something generates energy (the source), that energy travels by some route (the coupling path), and something is disturbed by it (the victim). The model is useful because it tells you there are always three places to intervene, and you only need one of them to work. You can quiet the source with a slower edge rate or a spread-spectrum clock, break the path with a filter, a shield or better board layout, or harden the victim so the energy no longer matters. In a real fix the path is usually the cheapest of the three to attack, and the victim the most expensive, because hardening often means someone else's silicon.
What is the difference between conducted and radiated?: It is a question of how the energy travels. Conducted disturbance rides along a wire: a power cord, a signal cable, an interface lead. Radiated disturbance crosses open space as a field, with the cables and the enclosure seams acting as unintended antennas. The split matters because it decides the instrument. Conducted emissions are measured through a LISN clamped on the mains lead, typically over 150 kHz to 30 MHz. Radiated emissions are measured with antennas in a chamber or on an open site, typically over 30 MHz to 1 GHz and higher for fast digital products. The same underlying noise often shows up in both, which is why fixing one sometimes fixes the other.
Is EMC a legal requirement or just good engineering?: Both, and the legal half is not optional. In the European Union the EMC Directive 2014/30/EU applies to most electrical and electronic equipment, and a product that does not satisfy it cannot lawfully carry the CE marking or be placed on the market. In the United States the equivalent obligation for unintentional radiators sits in 47 CFR Part 15. Radio products are handled by a separate instrument on each side, the RED in the EU and Part 15 subpart C or the licensed parts in the US, but the EMC obligation does not disappear, it is absorbed into that regime.
Why do products pass pre-compliance and then fail the formal test?: Most often because something changed that nobody considered part of the test. Cable routing is the classic one. A harness dressed slightly differently between the two sessions changes the length and geometry of the structure that is doing the radiating, and several dB can move with it. The other frequent causes are a different operating mode, a firmware build with different clock behaviour, a sample with a pre-production shield or gasket, and pre-compliance equipment whose own uncertainty is wider than the margin being claimed. Pre-compliance answers "roughly where am I", not "will I pass".