Electromagnetic interference (EMI) is unwanted electrical noise radiated or conducted by one circuit or device that degrades the performance of another. In embedded systems, EMI can corrupt data buses, cause spurious resets, and introduce errors in analog measurements.
In practice
EMI appears in embedded work from two directions: your board as a victim (susceptibility) and your board as a source (emissions). High-frequency digital signals -- fast edges on SPI, SDIO, or PWM lines -- generate broadband noise that couples onto nearby traces, power rails, or cables. Switching power supplies and motor drivers are particularly aggressive emitters. On the victim side, ADC inputs and low-level analog front ends are especially sensitive to injected noise from nearby switching activity.
PCB layout is the first and most effective line of defense. Keeping return paths short and directly under their signal traces, using solid ground planes, and placing decoupling capacitors close to power pins all reduce both emissions and susceptibility. The blog post "Layout recommendations and tips for best performance against EMC" covers practical layout strategies in detail. Slowing edge rates to the minimum acceptable for reliable signaling reduces radiated energy significantly without changing logic function.
Regulatory compliance adds a formal dimension. Products sold in the EU must meet the EMC Directive, which mandates both emissions limits (EN 55032 for IT equipment, for example) and immunity requirements (EN 61000-4 series). Pre-compliance testing with a near-field probe and a spectrum analyzer is a low-cost way to identify problem areas before a formal accredited test. The post "Embedded Software Creation II - European Normative & Legislation" discusses the regulatory landscape for embedded products in Europe.
Software can also contribute to or mitigate EMI. Spreading the energy of a repetitive clock by applying spread-spectrum clocking (available on many modern SoCs) lowers peak emissions at a specific frequency. Conversely, firmware that rapidly toggles many GPIOs simultaneously, or that leaves unused peripherals clocked and switching, adds unnecessary noise. Disabling unused clocks and peripherals is a simple software measure that reduces both EMI and power consumption.
Frequently asked
What is the difference between radiated and conducted EMI?
Radiated EMI propagates through the air as electromagnetic waves and couples into nearby circuits or cables without a direct electrical connection. Conducted EMI travels along a physical conductor -- typically power supply lines or signal cables -- directly into connected equipment. Regulatory test standards address both separately; for example, CISPR 32 covers both radiated and conducted emissions limits for multimedia equipment.
How do I find the source of an EMI problem on my board?
A near-field probe (a small loop or tip probe connected to a spectrum analyzer or oscilloscope) lets you scan the board surface to localize hot spots. Fast-edge signals on switching regulators, clock lines, and
PWM outputs are common culprits. Correlate spectral peaks with known switching frequencies in your design -- a peak at exactly your
buck converter's switching frequency, or its harmonics, points directly at the likely source.
Does EMI filtering belong in hardware, software, or both?
Primarily hardware. Ferrite beads, LC filters, common-mode chokes, and shielding address EMI at the physical level where software cannot reach. Software can help at the margins: reducing unnecessary switching activity, enabling spread-spectrum clocking where available, and configuring
GPIO drive strength and
slew rate to the minimum needed. But software alone cannot fix a poorly laid-out board or an unfiltered power entry.
Why does a solid ground plane matter so much for EMI?
Return currents follow the path of least impedance, which at high frequencies means the path directly beneath the signal conductor. A solid, uninterrupted ground plane provides that low-impedance return path and minimizes the loop area formed by the signal and its return. Smaller loop area means less magnetic flux, which directly reduces both radiated emissions and susceptibility to external fields. Slots, cutouts, or split planes in the ground layer force return currents to detour, enlarging the loop and worsening EMI.
What is the relationship between EMI and EMC?
EMI is the interference phenomenon itself. EMC (electromagnetic compatibility) is the broader discipline and the regulatory goal: a device exhibits EMC when it neither emits EMI above permitted limits nor malfunctions due to EMI from its environment. Passing EMC certification means demonstrating acceptable behavior on both the emissions and immunity axes.
Differentiators vs similar concepts
EMI (electromagnetic interference) is the noise event itself -- the unwanted signal. EMC (electromagnetic compatibility) is the system-level property describing whether a device can operate correctly in its electromagnetic environment without causing unacceptable interference to others. RFI (radio-frequency interference) is sometimes used interchangeably with EMI but technically refers only to interference in the radio-frequency portion of the spectrum, while EMI covers the full range including lower-frequency conducted disturbances.
ESD (electrostatic discharge) is a separate but related phenomenon: a brief, high-energy pulse from static charge that can damage or latch up ICs; ESD and EMI share some mitigation hardware (TVS
diodes, filtering) but are tested and specified differently.
