EMC/EMI Fundamentals

Intermediate11 min read

Assumes: 0.1 Anatomy of an SoC, 0.4 Power Architecture & PMIC

TL;DR

EMC (Electromagnetic Compatibility) means your electronics neither emit interference that disrupts other systems nor are susceptible to interference from them. In automotive, this isn’t optional — it’s a legal requirement (CISPR 25, ISO 11452). EMI (Electromagnetic Interference) is the noise itself; EMC is the discipline of controlling it. Every wire is an antenna. Every current loop radiates. OEMs routinely spend 30–40% of development schedules on EMC compliance. Understanding how noise couples, how to filter it, and how to design for quiet operation saves months of chamber rework.

If you’ve ever closed a software ticket as “cannot reproduce on bench” — then watched it reopen because QA drove past a cell tower with the rear defroster on — welcome to EMC. This is the page where “works on my desk” goes to die, and where a three-week CAN stack bug turns out to be a missing ferrite bead on the 12V feed.

You debug CAN frames and chase display glitches, but when a symptom appears only in the vehicle — never on the bench — EMC is often the culprit. This page gives you the vocabulary and mental models to recognize EMI problems, talk to hardware teams without guessing, and avoid the “works in lab, fails in chamber” trap.

EMI vs EMC

Term	What It Is	Automotive Context
EMI	The problem — unwanted electromagnetic energy (emission or susceptibility)	Switching regulator hash on 12V, CAN bit errors near a DC-DC, AM radio buzz from a display cable
EMC	The solution — design discipline to make systems coexist	Filtering, shielding, layout rules, harness routing, component selection

Mnemonic: “EMI is the disease, EMC is the vaccine.”

EMC splits into two directions you must pass in validation:

Category	Sub-types	What Gets Tested
Emissions	Radiated (antenna-like propagation through air)	Your module broadcasting noise that interferes with AM/FM, cellular, or other ECUs
	Conducted (noise on wires — power, ground, signal)	Switching hash riding the 12V harness into a sensitive receiver
Immunity	Radiated (RF fields injected at your module)	Display flicker, CAN errors, touch ghosting under RF stress
	Conducted (noise injected on power/signal lines)	Reset during BCI test, audio whine from load transients

Your PMIC buck converters (Page 0.4) are prime EMI sources — they switch at hundreds of kHz to MHz with sharp edges. Your CAN transceiver and Ethernet PHY are both sources and victims. EMC design is about keeping source energy contained and keeping victim circuits deaf to what’s left.

Mental Model

Think of your PCB as a festival main stage — clocks hammering like the kick drum, buck converters pumping like the subwoofer, CAN and Ethernet fighting for the mix. EMC is the FOH engineer’s nightmare: keep every instrument in its lane without the snare bleeding into the vocal mic. Shielding = iso booths. Filtering = EQ and compression. Grounding = the noise floor you can’t EQ out. The one rule that explains 80% of EMC: “Loops are antennas. Short loops = quiet boards.”

Think About It

It’s 2:47 AM. Pager: “CAN errors spiking on fleet vehicles — only when the 48V DC-DC is active.” You’re in the garage bay with a scope, half a cold coffee, and a CAN waveform that looks like someone added film grain to your eye diagram. Bench tests were clean all week. What’s happening?

Reveal Answer

Conducted coupling — switching noise from the 48V DC-DC injects common-mode current onto the shared ground/power return, raising the noise floor on the CAN transceiver differential input until bit sampling fails. On the bench, short cables and isolated supplies hide this; in the vehicle, a 2-meter harness running parallel to the converter’s return path completes the coupling loop. Scope CAN_H − CAN_L with a differential probe (not either line to ground), then fix harness separation and add a common-mode choke — not CAN stack code first.

Standards time — grab coffee. This is the part that shows up in tier-1 contracts and keeps compliance lawyers fully employed.

Key Standards

Automotive EMC isn’t one test — it’s a stack of international norms plus OEM-specific limits stricter than the baseline.

Standard	Scope	What It Covers
CISPR 25	Vehicle-level emissions limits	Radiated and conducted interference from components/modules — the benchmark for “does your module broadcast?”
ISO 11452	Component immunity testing	RF radiated/conducted immunity, BCI, ESD — “can your module survive the RF environment?”
ISO 7637	Electrical transients on 12V/24V	Load dump, cranking dips, reverse polarity — power-line abuse, not strictly RF but coupled to conducted EMC
ECE R10	European type approval	Legal gate for selling vehicles in EU markets — references CISPR 25 / ISO 11452
OEM specs	Stricter than baseline	Ford ES-XW7T-1A278-AC, GM GMW3097, VW TL 81000 — your tier-1 contract cites one of these

Rule of thumb: CISPR 25 tells you how quiet you must be. ISO 11452 tells you how tough you must be. The OEM spec tells you how much margin above both you actually need.

Now that you know what gets measured, here’s how the noise actually reaches your circuit — four paths, zero sympathy.

Noise Coupling Mechanisms

EMI reaches your circuit through four paths. Know which one you’re fighting before you add a ferrite bead.

Mechanism	How It Works	Typical Automotive Example
Conducted	Noise travels through shared power/ground — common impedance coupling	12V rail ripple from a nearby buck converter collapses CAN transceiver margins
Radiated (near-field)	Magnetic (current loops) or electric (voltage on high-Z nodes) coupling at short range	Unshielded display LVDS cable capacitively couples into an AM antenna trace
Radiated (far-field)	Full electromagnetic wave propagation — your module is an unintentional transmitter	Module fails CISPR 25 radiated emissions at FM broadcast band
Common mode vs differential	Differential = signal on pair (desired). Common mode = both lines move together (noise)	CAN works differentially; harness acts as common-mode antenna above ~30 MHz

Capacitive coupling = electric field, high-impedance victim (long parallel traces, unshielded cables). Inductive coupling = magnetic field, current-loop victim (power return sharing, un-twisted harness runs).

flowchart TB
    subgraph SOURCES["Noise Sources"]
        SW["Switching Regulator<br/>PMIC buck"]
        CLK["Clock / Data Bus<br/>harmonics"]
        MOT["Motor Driver<br/>PWM edges"]
    end

    subgraph PATHS["Coupling Paths"]
        COND["Conducted<br/>shared V/GND impedance"]
        CAP["Capacitive<br/>E-field, parallel runs"]
        IND["Inductive<br/>H-field, current loops"]
        RAD["Radiated<br/>far-field antenna"]
    end

    subgraph VICTIMS["Victims"]
        CAN["CAN Transceiver"]
        AUD["Audio Codec"]
        RF["AM/FM Receiver"]
        DISP["Display / Touch"]
    end

    SW --> COND
    CLK --> CAP
    MOT --> IND
    SW --> RAD
    COND --> CAN
    CAP --> RF
    IND --> AUD
    RAD --> DISP

Enough theory — here are the rules layout engineers tattoo on their forearms (metaphorically).

The Golden Rules of EMC Design

These rules apply whether you’re laying out a PMIC input filter or routing a CAN harness through a zonal gateway.

Rule	Why It Works	Practical Action
Minimize loop area	Loop area × current = magnetic dipole moment	Route return path adjacent to signal — never “ground later”
Filter at the source	Stopping noise at origin beats chasing it downstream	Ferrite beads + decoupling caps at switching nodes, not just at the connector
Shield when filtering isn’t enough	Enclosure blocks far-field radiation	Metal housing, EMI gaskets, grounded cable shields
Ground star topology	Single reference prevents ground bounce	One point where analog, digital, and power returns meet — avoid daisy-chained grounds
Separate domains	Mixed domains share noise	Split analog / digital / power ground planes with controlled bridge points

flowchart LR
    subgraph BAD["Bad: Large Loop"]
        S1["Signal"] --- L1["Long return path"]
    end

    subgraph GOOD["Good: Tight Loop"]
        S2["Signal"]
        GND["Return<br/>under/over signal"]
        S2 --- GND
    end

Common Gotcha

“My design works in the lab” means it’s EMC-clean. That’s like saying your mix sounds great with headphones in a silent room. Reality: Your lab bench has a solid ground plane, short cables, and no nearby high-power switching. The car has a harness running 2 meters alongside the DC-DC converter, shared ground return with the body ECU, and an AM radio antenna 30 cm from your module. Every lab pass needs chamber validation.

Common Gotcha

“Adding a filter fixes everything.” Slap a ferrite on it and call it a day — the EMC version of “did you try turning it off and on again?” Reality: Wrong filter placement or wrong cutoff frequency can make EMI worse. A ferrite bead that resonates at your switching frequency amplifies noise instead of suppressing it. Always check the impedance curve vs your noise spectrum — one size never fits all harmonics.

Rules in hand — time to find out if your module survives the chamber, or becomes a very expensive paperweight.

EMC Testing Basics

Validation happens in stages: pre-compliance (cheap, in-house, find problems early) → full compliance (certified chamber, legal/OEM sign-off).

Test	Setup	Pass/Fail Metric
Conducted emissions	LISN (Line Impedance Stabilization Network) + spectrum analyzer on power lines	Quasi-peak / average levels vs CISPR 25 limit lines
Radiated emissions	Anechoic chamber or OATS + calibrated antenna + spectrum analyzer	Field strength (dBµV/m) vs frequency
Immunity — BCI	Bulk Current Injection clamp on harness	Functional degradation thresholds (CAN errors, reset, display artifact)
Immunity — RI	RF field generator + antenna aimed at DUT	Operate-through criteria per ISO 11452
Immunity — ESD	ESD gun to connector shells and accessible surfaces	No latch-up, no permanent damage, recovery within spec

Pre-compliance uses the same topology with less expensive equipment — good enough to find 20 dB problems before you book the $5k/day chamber. Full compliance is the gate: fail here and you don’t ship.

flowchart LR
    DUT["DUT<br/>ECU under test"]
    LISN["LISN<br/>50Ω / 5µH"]
    SA["Spectrum<br/>Analyzer"]
    PS["12V Power<br/>Supply"]

    PS --> LISN
    LISN --> DUT
    LISN -->|"Conducted<br/>emissions tap"| SA

    subgraph RADIATED["Radiated Setup (separate run)"]
        ANT["Calibrated<br/>Antenna"]
        CH["Anechoic<br/>Chamber"]
        DUT2["DUT on<br/>turntable"]
        SA2["Spectrum<br/>Analyzer"]
        DUT2 --- CH
        ANT --> SA2
        CH --> ANT
    end

PCB Layout for EMC

Layout is where EMC is won or lost — software can’t filter a 500 MHz harmonic from a 3 cm loop.

Ground plane continuity — Never split a ground plane under high-speed traces. Return current follows the path of least impedance; a slot forces detours and creates slot antennas.

Decoupling cap placement — Place caps as close to power pins as physically possible. A 100 nF cap 2 cm away from the pin is decoration at switching frequencies.

Controlled impedance — High-speed signals (USB, Ethernet, MIPI) need defined trace width and reference plane. Mismatches cause reflections that radiate.

Via stitching — Ground plane islands connected by sparse vias act as independent resonators. Stitch vias every λ/20 around noisy regions and along board edges.

Clock routing — Shortest path, buried between ground planes when possible, guard traces tied to ground on outer layers. The system clock and its harmonics appear in almost every failed radiated emissions plot.

Where You'll See This

As a software engineer, you may not design the PCB — but you need EMC literacy when symptoms are environment-dependent:

Symptom	Likely EMC Mechanism
Display flickers at specific frequencies	Radiated immunity — RF coupling into display timing or touch controller
CAN errors only in vehicle	Conducted emissions from nearby switching converters on shared ground
Audio whine that tracks engine RPM	Ground loop / conducted noise — alternator ripple on analog ground

War story: A tier-1 passed all component-level EMC tests but failed vehicle-level radiated emissions at 88 MHz. The harness from cockpit module to trunk antenna acted as a tuned antenna at FM broadcast frequency. Fix: common-mode choke on the harness connector. Result: 6-week delay and $2M re-test cost. Component pass ≠ vehicle pass. Always validate at system level.

When bench tests pass and vehicle tests fail, ask “what’s the antenna?” before opening a software defect.

Connect the Dots

Page 0.7 (Physical Interfaces): CAN transceiver EMC performance (slew rate, common-mode range) and Ethernet PHY common-mode rejection determine how much harness noise becomes frame errors. Page 1.2 (ECU Networks): Twisted pair on CAN and Ethernet cancels differential-mode signal but rejects common-mode pickup — why OEMs mandate twisted harnesses, not zip-tied parallel runs. Page 9.7 (Validation): EMC is a DV/PV gate — fail radiated emissions or immunity and you don’t ship, regardless of feature completeness.

The short version

EMI is the noise; EMC is the engineering discipline to contain it — legally required, not optional.
CISPR 25 (emissions) and ISO 11452 (immunity) are the baseline; OEM specs add margin.
Loops are antennas — minimize current loop area; route returns with signals.
Filter at the source, verify the impedance curve, and never trust bench-only results.
Component pass ≠ vehicle pass — harnesses become antennas at the worst frequencies.

Check Your Understanding

1. What's the fundamental difference between EMI and EMC?

Recall

2. Your CAN bus shows errors only in the vehicle, never on the bench. Which coupling mechanism is most likely?

Applied

3. "Loops are antennas" — what's the practical design rule?

Recall

4. Which automotive standard primarily defines radiated and conducted emissions limits for in-vehicle receivers?

Recall

5. A tier-1 passes component-level EMC but fails vehicle-level radiated emissions at 88 MHz (FM band). The cockpit-to-trunk harness runs unfiltered for 2 meters. Most likely fix?

Applied

6. Radiated emissions fail CISPR 25 at 150 MHz. The PMIC buck switches at 2.2 MHz. A ferrite on the CAN harness reduces CAN errors but emissions still fail. What is the correct next step?

Synthesis

Connects to: Module 0.4 — Power Architecture · Module 9 — Safety, Security & Diagnostics