A product can work perfectly every time you use it and still fail certification the first time it goes into the test lab.
That’s because the lab checks things your own testing never does, like how much electrical noise your product gives off and whether it survives a zap of static electricity.
When a product fails, you’re usually looking at another board spin, another round of lab fees, and weeks of delay before you can legally sell a single unit.
And most of those failures come from design decisions that look completely harmless, like which edges of the board your connectors are on, or trusting that a pre-certified part has you covered.
These are the 11 PCB design rules behind the most common certification failures, and for each one I’ll tell you exactly how to fix it before your product ever gets to the lab.
Table of Contents
Rule #11 – NEVER Ignore Creepage and Clearance on Mains-Powered Boards
Any product that connects directly to AC wall power has to meet safety requirements, and the spacing on your board is a big part of that.
Clearance is the gap through the air, and it’s set by the peak voltage, since it has to stop a sudden voltage spike from arcing across.
Creepage is the gap along the surface of the board, and it’s set by the RMS voltage, which is the steady everyday voltage, since it has to stop current from slowly creeping across as dust and moisture build up.
For two traces side by side, those distances are the same, but a slot cut in the board between them adds creepage without changing the clearance, as long as the slot is wide enough to count.
The safety standard for your type of product, like IEC 62368 for most consumer electronics, sets both distances between the high-voltage side of your board and anything a person can touch, and they can be 5 mm or more.
So look up the required distances before you place a single part, and keep the high-voltage section in its own zone.
Even better, use a pre-certified external power adapter, so the AC side never touches your board at all.
Rule #10 – NEVER Route Traces Under or Beside the Crystal
The crystal sets the clock for your chip, and the area around it is both a source of noise and one of the most sensitive spots on your board.
Fast digital clocks create harmonics, which are copies of the clock frequency at every multiple, and traces running right next to the crystal circuit can pick up that noise and carry it out to a cable.
Noise from those nearby traces can also couple into the crystal and push it slightly off frequency, which is a big deal if it’s clocking a radio.
So place the crystal as close to your chip’s clock pins as you can, and keep its two traces short.
Keep every other trace away from the crystal circuit, and follow your chip maker’s layout guide for the copper underneath it, since some call for solid ground there and others want the copper cleared.
And keep the crystal away from the board edge and your connectors, since those are the easiest ways for its noise to escape.
Rule #9 – NEVER Skip the Filter on the Power Input
Conducted emissions testing measures the electrical noise your product pushes back out through its power cord and into the wall outlet.
If your product runs from the wall, even through a separate power adapter or while it’s charging, it gets this test, and only products with no connection to the wall skip it.
The main source of that noise is your switching regulators, which chop current on and off hundreds of thousands of times per second or more.
A pre-certified adapter doesn’t protect you here, because it was tested on its own, not with your noisy board plugged into it.
On a low-voltage DC input, the usual fix is a filter right at the power connector, made from a ferrite bead or a small inductor with a capacitor on each side of it.
Then check that the filter doesn’t make your regulator unstable, since some combinations of parts can resonate.
Keep the input and output sides of that filter physically apart, because if the noisy side runs right next to the clean side, the noise just jumps across and skips the filter.
Rule #8 – NEVER Let the Switching Regulator’s Hot Loop Grow
Every switching regulator has a hot loop, which is the path where current switches on and off in just a few nanoseconds.
In a buck converter, which is a regulator that steps voltage down, that loop runs from the input capacitor, through the switching transistors, and back to the capacitor’s ground.
A loop of fast-changing current acts like a small antenna, and the bigger the area it encloses, the more noise it radiates.
The sharp switching edges also ring at much higher frequencies, right in the range the lab scans.
The switch node is a second antenna, since it’s the copper connecting the switching transistors to the inductor, and its voltage jumps up and down with every cycle.
Place a small ceramic input capacitor right at the regulator’s input and ground pins with no vias in between, and make the switch node copper only as big as it needs to be.
Most regulator datasheets include a recommended layout, and copying it exactly is the easiest way to get this right.
Rule #7 – NEVER Drive an External Line With a Full-Speed Edge
Most people assume the noise from a digital signal depends on how fast it’s clocked, but the bigger factor is usually how fast each edge rises and falls.
A slow signal with sharp edges still puts out noise at frequencies hundreds of times higher than the signal itself.
Microcontroller pins can switch in just a few nanoseconds, which is far faster than an ordinary control signal running to a connector or another board needs.
The first fix is a small series resistor right at the pin that drives the line, usually somewhere between 22Ω and 100Ω, which cuts ringing and slows the edges down.
The second fix is your microcontroller’s GPIO speed setting, since many chips let you choose how fast each pin switches, and picking the slowest setting that still works costs nothing.
Fast interfaces like USB are the exception, so for those, follow the chip maker’s requirements instead of adding resistors.
Rule #6 – NEVER Let a Signal Leave the Board Without ESD Protection
ESD immunity testing is required for many products sold in Europe, and the lab will zap your product with thousands of volts of static electricity on purpose.
The lab hits the spots a person can touch, like connectors, buttons, and exposed metal, and if your product gets damaged or locks up and doesn’t recover, it fails.
Adding an ESD protection diode to your schematic is only half the job, because where you place it on the board matters just as much.
A protection diode only works if the zap reaches it before anything else, so it needs to sit right at the connector.
Route the signal straight through the diode’s pad on its way in, instead of branching off with a separate trace, so the zap can’t take a shortcut to your chip.
Then give the diode’s ground pin the shortest possible path to the ground plane, ideally with its own via right at the pad.
A long ground path adds inductance, and that inductance lets a big voltage spike get past the diode and into your circuit.
Rule #5 – NEVER Put Connectors on Opposite Edges of the Board
Connector placement can decide whether your product passes radiated emissions testing, even though it looks like a purely mechanical choice.
Even a solid ground plane has a little resistance and inductance, so noise currents flowing across it create tiny voltage differences from one side of the board to the other.
With one cable on the left edge and another on the right, that voltage difference drives the two cables like the arms of an old rabbit-ear TV antenna.
A difference of just a few millivolts across the board can be enough to push a product over the limit.
Group your cable connectors close together on one edge, so all the cables sit at nearly the same ground voltage, and keep noisy circuits like regulators and clocks from sitting between them.
If your connectors truly have to go on opposite edges, pay extra attention to the ground and shield connections near each one, and add filtering where those lines need it.
Connector locations usually get locked in during enclosure design, so make this decision then, not during layout.
Rule #4 – NEVER Let Return Current Change Layers Without a Nearby Path
Every signal on your board travels in a loop, out along the trace and back to where it started, and that current flowing back is called the return current.
At high frequencies, the return current flows on the plane right next to the trace, which is called its reference plane.
When a fast signal drops through a via to another layer, its reference plane often changes too, and the return current has to jump over to the new plane.
If there’s no ground via nearby, that current spreads out looking for one, and the bigger loop it creates radiates noise.
On fast lines like clocks, put a ground via right next to the signal via whenever the signal’s reference plane changes from one ground plane to another.
If its reference plane changes from a ground plane to a power plane, a via can’t connect those two, so place a small stitching capacitor between the planes right next to the signal via.
And add a row of ground stitching vias along the board edges, which ties your ground layers together and keeps noise from leaking out between them.
Rule #3 – NEVER Break the Conditions of Your Pre-Certified Module
A pre-certified wireless module can save you most of the expensive testing on the radio itself.
In the US, you can usually reuse the module’s FCC approval, and in Europe, you can include the module maker’s radio test reports in your own CE paperwork, but only if you use the module exactly the way its documents say.
A module with a built-in antenna has a keep-out area where nothing else is allowed, because copper, traces, parts, or even a metal screw inside it change how the antenna performs.
Swapping antennas is another trap, since modules with an antenna connector are only certified with specific antenna types, and only up to a certain gain.
Some approvals also limit how close the antenna can be to a person’s body, which matters a lot for wearables and handheld products.
Place the module where its datasheet shows, usually with the antenna right at the board edge, and copy the keep-out exactly on every layer.
Break any of those conditions, and you can end up paying for the full radio testing the module was supposed to save you.
And even when you follow every rule, your complete product still needs its own emissions testing.
Rule #2 – NEVER Route a Fast Signal Over a Gap in the Plane Below It
A fast signal’s return current flows on the ground plane right underneath its trace.
Any gap in that plane forces the return current to detour around it, and that detour forms a loop that radiates noise.
Gaps show up in a lot of ways, like a split between ground sections, a slot cut for a trace, or a row of vias whose clearance holes merge into one long opening.
Normal clearance checks in your layout software won’t catch any of these, but a slot under just one clock line can be enough to fail the radiated emissions scan.
So keep one solid ground plane under your fast signals, and separate noisy and sensitive circuits by placement instead of cutting the plane apart.
If a signal has to cross a gap, first try rerouting it or closing the gap, and if neither works, put a stitching capacitor across the gap right where the trace crosses.
Just never do that across a safety isolation gap on a mains-powered board, since that gap is there to protect people and needs its own rated parts.
Then before you order boards, view your ground layer by itself and look for any gap sitting under a fast trace.
Rule #1 – NEVER Let a Cable Leave the Board Without a Filter
Cables cause more radiated emissions failures than anything else on this list, since a cable around 1 m long is a nearly perfect antenna for the frequencies the lab cares about most.
Noise from your board leaks onto every wire in the cable as common-mode current, which flows the same direction on all the wires at once, so it doesn’t cancel out.
Just a few µA of common-mode current on a cable can be enough to fail FCC testing.
So treat every cable as a noise path, and put any filtering right at the connector.
On fast differential pairs like USB, a common-mode choke can help when you need it, but use one made for that interface, since the wrong one can damage the signal.
On slower signals and power lines, a ferrite bead with a small capacitor to ground works well, as long as the values still let your signal through cleanly.
If the cable is shielded and your enclosure is metal, bond the shield to the enclosure right where the cable enters, and in a plastic enclosure, follow the shield connection the interface’s datasheet recommends, keeping it short.
And if you’re not sure a line needs filtering, leave a spot for a ferrite bead with a 0Ω resistor in it, or a spot for a choke on USB, so you can add a filter at the lab instead of spinning a new board.
If you’d like experienced engineers to check your layout for these exact certification problems before you pay for a lab test, you can get that help from me and other engineers inside my Hardware Academy.