EMC

Why products fail EMC testing: the 7 most common causes (and how to fix them)

Most products fail their first formal EMC test, and the reasons are remarkably repetitive. After enough chamber sessions the same seven causes explain nearly every red plot we see. This guide covers the honest failure statistics, how to classify your failure, the seven causes with their symptoms and fixes, and the debug workflow that turns a failed plot into a passing product.

· 11 min read

An EMC failure rarely announces itself during development. The product works, the firmware is stable, everyone is proud, and then the chamber plot shows a forest of peaks above the limit line. None of this means the team did bad work. Electromagnetic behaviour emerges from the interaction of layout, cables, grounding and software, and that interaction only reveals itself under measurement. The encouraging part: the causes are not mysterious, and neither are the fixes.

Quick answers

  • How many products fail their first EMC test? Published industry figures range from roughly half to the large majority; high-voltage and power-electronics products fail most often
  • Most common failure? Radiated emissions, usually driven by cable routing, with switching power stages as the underlying source
  • Fastest fix path? Classify the failure (emissions or immunity), find the frequency, find the structure that radiates it, break the path
  • Best prevention? A design review plus pre-compliance measurements during development, with at least 6 dB margin before the formal test

How often do products really fail?

Numbers circulating in the industry range from around fifty percent to as high as ninety percent first-attempt failure, depending on the product category and who is counting. Exact percentages matter less than the pattern behind them: failure rates climb with switching power, cable count and system complexity. A battery-powered sensor usually passes; a high-voltage drivetrain in a machine with twenty metres of harness usually does not, at least not the first time. Treat a first-attempt failure as the expected outcome to plan around, not as a surprise to be embarrassed by. The teams that pass at the first attempt are almost always the teams that measured before the formal session.

First, classify the failure

Every EMC failure belongs to one of two families, and the fix strategy differs fundamentally between them:

Family

What the lab observed

Where the fix usually lives

Emissions failure

Your product exceeds the noise limits, radiated (through the air) or conducted (through cables)

Filtering, cable routing, shielding, layout: stop the noise at its source or its exit

Immunity failure

Your product misbehaves when noise is applied: resets, glitches, sensor errors, lost communication

Grounding, protection components, robust firmware: keep the noise out or survive it gracefully

Emissions failures are about three times more common in our experience, and radiated emissions lead the list. That is where the seven causes below concentrate.

The 7 most common causes, with symptoms and fixes

1. Cable routing and harness coupling

What happens: cables are excellent antennas. A noisy signal routed alongside a long harness couples its noise onto every neighbouring wire, and the harness radiates it into the chamber. How it shows: broad emission humps between roughly 30 and 300 MHz that shift when the test engineer repositions a cable. That cable sensitivity is the giveaway. The fix: routing discipline. Separate noisy and clean runs, keep high-current loops short, twist supply pairs, and shield only where separation is impossible. On the drawing board this costs nothing; after the harness is tooled it costs a redesign.

2. Grounding and bonding mistakes

What happens: ground is not a magic drain but a return path with impedance. When return currents cannot flow back under their signal, they find creative routes through the chassis and turn the whole structure into a radiator. How it shows: emissions that change when a mounting bolt is touched, results that differ between test setups, immunity failures that depend on where the machine stands. The fix: a deliberate grounding concept: defined bonding points, short low-inductance straps, and one documented strategy instead of historical accident.

3. Missing or mis-specified input filters

What happens: conducted emission failures almost always trace back to the power entry: a filter omitted to save cost, chosen from a datasheet without measurement, or bypassed by layout. How it shows: sharp peaks at the switching frequency and its harmonics on the conducted plot, marching up the spectrum in regular steps. The fix: a filter designed against measured noise, mounted where the cable enters the enclosure, with a clean, low-impedance ground reference. Ferrites damp; a properly designed and placed filter decides.

4. Enclosure openings and seams

What happens: a metal housing only shields while currents can flow across it. Every slot, display cutout, plastic panel and badly gasketed seam becomes an antenna at the frequency whose wavelength fits it. How it shows: high-frequency failures, typically above 200 MHz, that survive every cable fix you throw at them. The fix: conductive gaskets, shorter slots, connectors moved to one wall, and honest acceptance that a plastic housing means the shielding must happen at board and cable level instead.

5. PCB layout: loop areas and return paths

What happens: on the board, emissions scale with loop area, the space enclosed between a signal trace and its return current. Split ground planes, traces crossing plane gaps and connectors far from their reference all enlarge those loops. How it shows: stubborn broadband noise that survives filtering and shielding, because the source itself is the board. The fix: continuous ground planes, returns routed under their signals, and decoupling close to the switching pins. The cheapest fix of all before layout freeze and the most expensive after tooling, which is why an EMC design review belongs in every development plan.

6. Switching power stages and HV drivetrains

What happens: inverters and DC-DC converters switch hundreds of volts in nanoseconds, and every edge is a broadband transmitter. In electric vehicles and machines this is the number one failure source by energy: high-voltage cables carry the noise through the entire platform. How it shows: strong, broad emission signatures tied to drive modes; the plot changes character the moment the drivetrain enables. The fix: a combination of slower switching edges where efficiency allows, shielded HV cabling with proper 360-degree terminations, filtering at the source and attention to the coupling between HV and LV harnesses. This is specialist work, and exactly where a lab with high-voltage experience earns its keep.

7. Software and operating-mode surprises

What happens: software decides when hardware makes noise and whether a disturbance becomes a harmless blip or a reset. The product that passed in idle fails in drive mode; a display PWM lands exactly on a protected band; error handling turns a microsecond glitch into a lockup. How it shows: failures that appear only in specific operating modes, or immunity results that differ between firmware versions. The fix: test in realistic modes, spread or shift clocks where possible, and treat recovery behaviour as part of the EMC design, not an afterthought.

The debug workflow that actually works

01

Read the plot before touching the product

Note every failing frequency and its character: narrowband spikes point to clocks and switchers, broadband humps point to cables and layout. The plot narrows the suspect list before any probing starts.

02

Correlate frequencies with the design

Match failing frequencies against known sources: switching frequencies and their harmonics, clock rates, communication bit rates. Most peaks identify their own culprit this way.

03

Localise with near-field probes

A near-field probe over boards, cables and seams shows where the energy actually leaves the product. Minutes of probing replace hours of guessing.

04

Change one thing, measure again

Ferrite here, cable moved there, gasket added: one change per measurement, or you will never know what worked. In a fix-and-measure session this loop takes minutes per iteration.

05

Verify margin, not just compliance

A fix that scrapes half a dB under the limit is a coin flip at the formal session. Aim for at least 6 dB of margin so setup differences and production spread cannot flip the result.

The pattern behind all seven

None of these causes is exotic, and almost none is visible in a schematic review alone. They emerge under measurement. That is the entire argument for EMC pre-compliance testing: the same chamber, months earlier, while every fix is still a layout change instead of a recall.

Prevention: the checklist that avoids the retest

Before you book the formal session

  • EMC design review held before layout freeze (grounding concept, filter positions, cable plan)
  • Pre-compliance emissions scan done on the first working prototype
  • All operating modes defined and tested, not just idle
  • At least 6 dB margin below the limits at the dress rehearsal
  • Fixes verified in the chamber, not assumed from the datasheet

What a failure costs in money and calendar time, and what catching it early costs instead, is quantified in our guide on EMC testing costs. The short version: one failed formal session usually costs more than the entire original test budget. Our EMC testing services are built around that reality: engineering first, chamber second, certificate once.

Every EMC failure is cheap exactly once: the first time you measure it. After that, the price only goes up.

ABOUT THE AUTHOR

IntegratR Engineering Team

EMC & e-mobility engineering, Innovatiecampus Nijmegen

IntegratR runs a heavy-duty EMC test facility and engineering team at the Innovation Campus in Nijmegen, the Netherlands. The team takes zero-emission vehicles and machines from prototype to certified, working with OEMs, Tier-1 suppliers and scale-ups across Europe. Articles are written from the test floor, not from theory.

1. What percentage of products fail EMC testing the first time?

Published figures range from roughly fifty percent to as high as ninety percent depending on product category and source. The pattern is more useful than the number: failure rates rise with switching power, cable count and complexity, so high-voltage and power-electronics products fail most often.

2. What is the most common EMC failure?

Radiated emissions above the limit, most often driven by cable routing and harness coupling, with switching power stages as the underlying source in electric vehicles and machines.

3. How do I find the cause of an EMC failure?

Read the plot first: narrowband peaks point to clocks and switching frequencies, broadband humps to cables and layout. Correlate frequencies with known sources in the design, then localise with near-field probes and change one thing per measurement.

4. Does adding a ferrite or filter always solve it?

No. Ferrites damp a resonance; they do not repair a broken return path or an oversized loop. Filters only work when placed and grounded correctly. Fixes that address the cause survive production tolerances; band-aids often fail the retest.

5. Can software cause an EMC failure?

Yes, in two ways: it decides when noisy hardware runs, and it determines whether a disturbance becomes a harmless blip or a system reset. Immunity results are as much firmware as physics, which is why testing in realistic operating modes matters.

6. Do electric vehicles fail EMC testing more often?

They fail harder, statistically: high-voltage inverters and cabling add powerful broadband sources that petrol-era designs never had. That is why high-voltage EMC experience matters when choosing a lab.

7. How much margin should I aim for below the limits?

Common engineering practice is at least 6 dB below the limit at pre-compliance, to absorb setup differences and production spread. Passing by half a dB is a coin flip at the formal session.

8. What happens after a failed EMC test?

The lab documents the failures, engineering analyses root causes, the design is modified and the affected tests are repeated. That loop costs weeks to months, which is why moving it forward into development via pre-compliance is the highest-return decision in the EMC budget.

Ready to accelerate your zero-emission innovation?

IntegratR offers everything on one campus: EMC test facility, certification support and production site in Nijmegen.

Get in touch