EMC Is the Oldest Rule in the Book. It's Still Where Products Fail First.

Ask a hardware team which CE requirement keeps them up at night and you'll hear about the Cyber Resilience Act, the AI Act, maybe the Digital Product Passport. Nobody says EMC. It's the oldest, most boring pillar of CE marking. Which is exactly why it's still the one that fails.

The pattern is always the same. The product works beautifully on the bench. Firmware is stable, the app connects, everyone is proud. Then it goes to the EMC lab and the report comes back with a spur sailing over the emissions limit, or the touch interface ghost-activating the moment the immunity test injects a burst on the power line. Suddenly there's a four-week lab requeue and a board respin standing between you and launch.

EMC is a design property, not a test

The mistake underneath is treating EMC as an exam you cram for. It isn't. It's a property of decisions you made months earlier: where the switching regulator sits, how the ground plane is stitched, which cable leaves the enclosure where, whether the connector shell is actually bonded or just looks like it. By the time an accredited lab measures your product, the physics is already decided. The lab just tells you the score.

And half the requirement is the half nobody tests early. The EMC Directive demands two things with equal weight: don't disturb your neighbours, and survive your neighbours. Teams obsess over emissions because that's what labs flag first, then ship products that reset every time a contactor switches two metres away. Immunity failures don't show up in the lab report you skipped. They show up in warranty returns.

The radio trap

Here's the one that catches modern teams: the moment you add Bluetooth, Wi-Fi or any intentional radio, your product legally stops being covered by the EMC Directive. It becomes radio equipment, and the same EMC physics now runs through the Radio Equipment Directive instead, with a different standards set and a different Declaration of Conformity. I've seen products with a current test report declared against the wrong directive because a radio module was added in revision C and nobody told the compliance file.

The target moves even when your product doesn't

The other quiet failure mode: harmonised standards get new editions, and only the editions cited in the Official Journal give you presumption of conformity. Your product can be unchanged for three years and still lose its legal footing, because the paperwork under it expired. Checking the live OJ listing at every revision is unglamorous work. It's also the difference between a file that holds up and one that doesn't.

So: pre-compliance scans early, immunity testing before anyone asks for it, a compliance file that knows which directive it's actually declaring against, and a standing reminder to re-check standard editions. Boring, layered, unglamorous — like EMC itself. That's the job.