THE MODEL
Source, path, victim.
A useful electromagnetic-compatibility investigation starts with three questions:
What is generating energy? A switch-mode converter, clock, motor commutator, PWM edge, relay, radio, cable current, ESD event, lightning transient, processor I/O bank, or something outside the product.
How is that energy getting across? Through a shared conductor, electric field, magnetic field, common-mode cable current, parasitic capacitance, mutual inductance, or radiated field.
What circuit is sensitive? An ADC input, reset line, oscillator, communication pair, sensor reference, amplifier, radio receiver, comparator, memory bus, or external interface.
Without those three pieces, “EMI problem” is a category, not a diagnosis.
CONDUCTED COUPLING
Sometimes the antenna is just a wire you already drew.
Noise can travel through power rails, grounds, signal conductors, shields, cable returns, and chassis bonds. A switching converter can inject ripple or high-frequency current into the supply network. A motor can dump current into a shared return. A digital interface can move common-mode current through cable capacitance.
Conducted problems are often easier to reproduce because there is a direct electrical path. They are also easy to mislabel as “radiated EMI” when the actual path is a shared rail or reference.
Filters, decoupling, impedance control, local energy storage, ferrites, common-mode chokes, topology changes, and better return routing can all help, but only when matched to the dominant current and frequency range.
CAPACITIVE COUPLING
Fast voltage change pushes current through parasitic capacitance.
Any two conductors separated by insulation form some capacitance. If the voltage on one conductor changes rapidly, displacement current can flow through that capacitance into the victim.
A simple relationship is:
i = C × dv/dt
That means even a tiny parasitic capacitance can carry meaningful transient current if the voltage edge is fast enough.
Long parallel traces, switch nodes, high-voltage transitions, relay contacts, cable conductors, heatsinks, and nearby planes can all create electric-field coupling. Increasing spacing, reducing parallel overlap, slowing unnecessary edge rate, shielding with a low-impedance reference, and controlling switch-node geometry can reduce the path.
INDUCTIVE COUPLING
Changing current creates magnetic fields, and loop area decides how much you care.
A current loop produces magnetic flux. A nearby victim loop can intercept some of that flux and develop a voltage. Mutual inductance is the formal description; practical layout language is simpler: high di/dt current loops and large victim loops are a bad combination.
The induced voltage scales with changing current:
V ≈ M × di/dt
Switching-regulator hot loops, motor phases, relays, transformer windings, high-current PWM paths, and fast GPIO return loops can all be magnetic-field sources.
Reducing loop area on both source and victim sides is one of the most powerful fixes. Keep high-current outgoing and return paths close together. Keep sensitive loops small. Rotate or separate magnetically coupled structures where useful. Do not route sensitive traces around a board perimeter and then wonder why every current transient gets a vote.
RADIATION
A board becomes a better antenna when geometry and wavelength start cooperating.
At sufficiently high frequency, traces, cables, seams, heatsinks, and enclosures can radiate and receive electromagnetic energy efficiently enough that a near-field lumped approximation stops telling the whole story.
Cables are especially important because a modest common-mode voltage can drive current along a long conductor that radiates far more effectively than the compact PCB that created the disturbance.
This is one reason passing emissions tests is often less about the differential signal itself and more about controlling common-mode current on external conductors.
A product can therefore have pristine logic waveforms and still fail radiated emissions because the cable shield, reference, connector transition, or chassis bond lets common-mode current escape onto a good antenna.
EDGE RATE
Frequency content follows transition time, not the marketing clock.
A 10 MHz clock with sub-nanosecond edges contains significant harmonic energy far above 10 MHz. The same is true for PWM, serializers, memory buses, and fast logic outputs.
If the receiving device does not need the fastest available slew rate, slowing the edge can reduce high-frequency energy, ringing, crosstalk, and radiation while preserving the required timing margin.
Series resistors near a driver can sometimes damp transmission-line ringing and reduce edge energy. Programmable slew-rate controls can help. The correct value is a signal-integrity design choice, not a universal “add 33 ohms” charm.
RETURN PATH
Every fast signal carries its electromagnetic field with its return path.
A continuous reference plane provides a low-impedance nearby return for high-frequency current. When a signal crosses a split, void, connector discontinuity, or layer change without an appropriate return path, the current must detour.
That increases loop area and therefore both susceptibility and emissions. Layer changes can be especially troublesome if the reference changes from one plane to another and there is no nearby stitching path for return current.
Signal integrity and EMI are not separate planets. Poor return geometry that creates ringing and crosstalk often creates radiation too.
COMMON MODE
Differential systems still fail through common-mode current.
Differential signaling is powerful because equal and opposite currents can cancel external fields and reject shared noise. Real implementations are never perfectly balanced.
Skew, asymmetrical routing, connector geometry, unequal parasitics, reference discontinuities, transformer imbalance, and driver mismatch convert some differential energy into common mode. That common-mode energy can then excite cables, shields, and chassis structures.
This is why Ethernet, USB, LVDS, CAN, and other differential interfaces still require careful layout and connector treatment. “Differential” reduces a coupling mechanism. It does not repeal electromagnetism.
SHIELDING
A shield only works if current has somewhere intentional to go.
A conductive enclosure can attenuate electric fields and, depending on material and frequency, magnetic fields. But apertures, seams, cable penetrations, poor bonds, and long pigtail shield connections can undermine the result.
At high frequency, a shield connection needs low inductance. A wide 360-degree connector bond can behave very differently from a long thin wire attached to the shield at one point.
Shielding also interacts with safety earth, chassis, signal reference, and cable return strategy. The design goal is to route unwanted common-mode current through a controlled low-impedance path rather than through the sensitive electronics.
FILTERING
A filter is an impedance network, not a decorative capacitor.
Filters work by presenting different impedances to wanted and unwanted spectral components. Their behavior depends on source impedance, load impedance, component parasitics, placement, and layout.
A capacitor intended to shunt high-frequency noise is useful only if its path to the reference is low impedance at the relevant frequency. A ferrite bead can attenuate some high-frequency content but can also interact with capacitance to create resonances. A common-mode choke addresses a different current mode than a differential LC filter.
The correct filter begins with identifying whether the noise current is differential or common mode and where the energy should be diverted.
SUSCEPTIBILITY
Passing emissions is only half the job.
Electromagnetic compatibility includes both what a device emits and what it can tolerate.
External fields, ESD, EFT/burst, surge, nearby radios, contactors, automotive transients, and utility disturbances can inject energy into inputs and cables. Protection components need a current path to work. A TVS diode placed far from the connector with a long trace to ground may allow the transient to travel through the board before the protection finally acts.
Robust interface design therefore considers connector entry, current diversion, clamping, filtering, reference strategy, isolation, and the failure energy the protection parts themselves must survive.
NEAR FIELD DEBUGGING
Sometimes the fastest way to find EMI is to deliberately eavesdrop on it.
Near-field electric and magnetic probes can help localize high-frequency sources on a PCB. A current probe can reveal common-mode cable current. A spectrum analyzer can show frequency structure that correlates with clocks, converters, radios, or PWM.
These measurements are relative diagnostic tools unless calibrated for a specific purpose. A strong near-field reading does not by itself establish regulatory failure. It identifies where energy exists and how it changes when the circuit changes.
Likewise, moving a cable, touching a shield, adding a temporary ferrite, slowing a clock edge, or changing converter load can be useful controlled perturbations. The goal is to learn which path matters.
FIELD METHOD
How to turn “EMI” into a mechanism.
1. Correlate in time and frequency. Does the failure align with a PWM edge, radio burst, converter frequency, clock harmonic, motor event, or external transmitter?
2. Separate conducted from radiated paths. Change power source, disconnect cables, use temporary shielding, and alter physical spacing.
3. Reduce edge rate or switching activity. If the problem changes, high-frequency content is implicated.
4. Inspect return paths. Look for plane splits, connector transitions, large loops, and shared impedance.
5. Measure common-mode cable current. External conductors frequently turn board noise into radiated noise.
6. Probe locally. Use near-field tools to identify hot loops, switch nodes, oscillators, and interfaces.
7. Change one coupling parameter at a time. Distance, orientation, capacitance, loop area, bond inductance, and source impedance are more informative than adding random ferrites until the symptom retreats.
BOTTOM LINE
Interference follows physics, even when the symptom looks supernatural.
An EMI failure becomes tractable once you identify source, path, and victim.
Reduce the source, break the path, harden the victim, or do all three.
The worst debugging strategy is treating every electromagnetic problem as a request for more shielding. The best is discovering exactly where the unwanted current or field is going and giving it a better place to go.
SOURCE TRAIL
Technical starting points.
Texas Instruments — Sources of EMI in Ethernet Applications
Texas Instruments — High-Speed Layout Guidelines for Reducing EMI for LVDS SerDes Designs
Analog Devices — Basic Switching-Regulator Layout Techniques