EMC Troubleshooting Part 3 — Broadband Footprints: 2‑Layer Boards & Plane Slots
- Francesco Poderico
- Sep 16
- 4 min read
Scope: Why some products show broadband radiated emissions from 30 MHz–1 GHz, why issues often persist up to 400 MHz–1 GHz even on multilayer boards, and how to prove and fix the root cause fast.
1) What “broadband” really means in RE
On the plot, broadband looks like a wide hump or plateau with many modest peaks rather than a few tall, razor‑thin lines. It often indicates large current loops or apertures that couple energy efficiently over a wide span of frequencies. Cables then act as efficient antennas that export this energy.
2) Why 2‑layer PCBs are frequent offenders (30 MHz–1 GHz)
On many 2‑layer designs, “ground” is not a solid plane but a collection of traces around the board. At low frequency that works; at high frequency it does not.
The return‑path story
High‑frequency return current wants to flow directly under the signal trace (lowest loop inductance). Without a plane, it must detour through whichever ground traces or capacitive paths exist.
The result is multiple, competing return loops of different lengths and geometries → each loop has its own resonance. The sum of many loops produces a broad continuum rather than a single sharp peak.
Those loops couple easily to attached cables (I/O, power), which radiate efficiently in the 30–300 MHz band and beyond.
Above you can see the typical EMC emission of a 2 layer board, as you can see is very brodband .
Why these designs also fail RI/CE/CI
Radiated immunity (RI): big loops = big induced voltages V=−dΦ/dtV = -\mathrm{d}\Phi/\mathrm{d}t. The same geometry that radiates well also receives RF well.
Conducted emissions (CE) / Conducted immunity (CI): poor return control promotes common‑mode conversion, pushing noise onto external leads and making the system sensitive to RF on those leads.
Tell‑tales on the plot: a wide “hill” spanning tens to hundreds of MHz; improving one cable or loop reduces the whole hill a few dB.
3) Broadband above 400 MHz on multilayer boards
Even with a solid ground plane, you can still get a broadband plateau in 400 MHz–1 GHz.
The slot/aperture mechanism
Unintended slots (long cuts, via farms, cutouts, plane splits under signals) break the local return path and create apertures.
A slot behaves like a slot antenna; its fundamental resonance is roughly:
fslot ≈ c2 L εefff_{\text{slot}} \;\approx\; \frac{c}{2\,L\,\sqrt{\varepsilon_\text{eff}}}
where L is the effective slot length. Real boards exhibit multiple modes and fringing fields, so instead of one razor‑thin line you often see a cluster or broad region of elevated energy.
If the slot sits near a fast edge (clock, switch node), the aperture gets driven over a wide spectrum, lifting the whole 400 MHz–1 GHz region.
Common culprits:
Long plane cuts for isolation that accidentally sit under high‑speed traces.
Stitched copper fills with too‑wide via pitch (insufficient stitching → effective apertures).
Connector keep‑outs that carve the plane near I/O where cables exit.
4) Fast proofs you can run at the lab
For 2‑layer boards:
Copper tape jumpers: bridge suspected gaps in “ground” to create a short return; the broadband hill should drop several dB if you hit the right spot.
Short the loop: add a temporary ground strap (wide braid or foil) from the noisy sub‑circuit to the nearest chassis/plane reference. (see figure below)
Cable length swap (e.g., 10 m → 1.5 m): broadband dominance shifting with length confirms cables are the radiator.
Above you can see some details of me, trying to "shorter" GND loops
For multilayer slot/aperture cases:
Copper tape over the slot (or add a via “staple” if you can): even a temporary bridge often knocks down the 400 MHz–1 GHz plateau.
Route‑detour test: if possible, re‑run with the sensitive/high‑edge net temporarily rerouted or slowed (slew control) to see if the plateau collapses.
Chassis‑return capacitors near the I/O: add small safety‑rated caps from shield/0 V to chassis; if the plateau falls, you’re exporting common‑mode via the aperture to the cable.
Document before/after traces and the exact physical change—this is your causality proof.
5) Design fixes that work (quick to deep)
For 2‑layer boards:
Prefer 4‑layer with solid GND plane on L2 whenever possible.
Pair forward and return (twist where practical in wiring; keep PCB traces closely coupled) to shrink loop area.
Add common‑mode chokes on external leads and provide a clean HF return to chassis at the connector (short, low‑inductance).
For multilayer (400 MHz–1 GHz slot cases):
Keep planes continuous under fast signals and clock trees; no splits. If a split is mandatory, do not cross it with high‑edge nets.
Use via fences/stitching around perimeters and along split boundaries (pitch ≲ λ/20 at the highest relevant frequency; e.g., ≤ 7 mm at 1 GHz in FR‑4).
Maintain short, low‑inductance chassis bonds at I/O (Do not use long pigtails).
6) Lab checklist (broadband edition)
Mark suspect return gaps/slots on the layout; correlate with hotspots. (a near field probe may help here)
Run copper‑tape bridges and ground straps; log dB changes. (see the picture above)
Try to change the cable lengths and log the results.
Recap
2‑layer “ground by traces” → many return loops → broadband radiation across 30 MHz–1 GHz + poorer RI/CE/CI.
Multilayer with plane slots/apertures can still show a 400 MHz–1 GHz plateau.
Prove with copper‑tape bridges, ground straps, and cable length swaps; fix with solid planes, stitched grounds, controlled returns, and low‑inductance chassis bonds.
Coming next — Part 4
Semi‑narrowband combs: when clocks, SMPS ripple, and cable resonances mix—and how to identify the spacing (Δf) to find the engine.