Is This Conducted or Radiated?#
Distinguishing how noise gets from source to victim. Conducted noise travels on wires β power leads, signal cables, ground connections. Radiated noise travels through the air as electromagnetic fields. The fix is completely different for each, so identifying the coupling path before reaching for a solution saves time.
The Four Coupling Mechanisms#
| Mechanism | Path | Frequency tendency | Fix approach |
|---|---|---|---|
| Conductive (galvanic) | Shared wire, trace, or ground | All frequencies, often low-frequency | Separate current paths, improve grounding |
| Capacitive (electric field) | Between traces/components with voltage difference | Increases with frequency and dV/dt | Shielding, distance, guard traces |
| Inductive (magnetic field) | Between current loops in proximity | Increases with frequency and dI/dt | Reduce loop area, distance, shielding |
| Radiated (far-field EM) | Through the air, antenna-to-antenna | Dominant above ~30 MHz | Shielding, filtering, layout |
At bench distances and PCB scales, most noise below ~30 MHz is near-field (capacitive or inductive coupling). Above ~30 MHz, radiated coupling becomes dominant.
Distinguishing Conducted vs. Radiated#
Distance test: Measure noise level, then physically move the noise source away from the victim. Noise that drops rapidly with distance is radiated. Noise that doesn’t change with distance is conducted.
Cable disconnect test: Measure noise with all cables connected, then disconnect cables one at a time. If noise drops when a specific cable is disconnected, that cable conducts the noise. If noise persists with all cables disconnected, it’s radiated.
Ferrite clamp test: Clip a ferrite around a cable suspected of conducting noise. If noise drops, common-mode conducted noise on that cable is the path. Ferrites attenuate common-mode current at high frequencies (typically > 1 MHz).
Shielding test: Place grounded conductive sheet between suspected source and victim. If noise drops, coupling is through electric or magnetic fields. For magnetic shielding at low frequencies (50/60 Hz), mu-metal or thick steel is needed β aluminum doesn’t help.
Common-Mode vs. Differential-Mode Noise#
| Type | Where noise appears | How to measure |
|---|---|---|
| Differential-mode (DM) | Between signal and return | Normal measurement β probe signal vs. ground |
| Common-mode (CM) | On both signal and return equally | Probe both wires vs. earth; or current clamp around entire cable |
Differential-mode noise is on the signal and must be filtered on the signal path. Common-mode noise is on both wires equally β a differential receiver rejects it if CMRR is high enough.
Measuring common-mode current: Clamp an AC current probe around the entire cable (all conductors together). If only differential-mode current flows, the fields cancel and probe reads zero. Any current the probe reads is common-mode.
Empirical Fix Verification#
- Before: Measure and record noise level (scope screenshot, RMS voltage, FFT)
- Apply one fix (ferrite, shield, filter cap, reroute, grounding change)
- After: Measure again under same conditions
- Compare: Improvement (dB) = 20 Γ log10(V_before / V_after)
| Improvement | Practical meaning |
|---|---|
| 6 dB (2Γ reduction) | Noticeable but modest |
| 20 dB (10Γ reduction) | Significant |
| 40 dB (100Γ reduction) | Excellent β essentially eliminated |
Tips#
- Change one thing at a time β if multiple fixes are applied simultaneously, which one worked is unknown
- Both coupling paths often exist simultaneously β fixing one may reveal the other
- Test under worst-case conditions, not best-case
Caveats#
- Moving cables changes both conducted and radiated coupling geometry β be systematic
- Ferrite clamps only help with common-mode noise β if noise is differential-mode, ferrite does nothing
- Common-mode noise converts to differential-mode at any asymmetry (different trace lengths, mismatched impedances)
- Some fixes interact β adding filter capacitor to switching regulator output can affect stability
In Practice#
- Noise that drops with distance is radiated β shielding and separation are effective fixes
- Noise unchanged by distance is conducted β filtering and ground improvement are needed
- Noise eliminated by ferrite clamp is common-mode conducted β ferrites at cable entry are the fix
- Noise eliminated by shielding is capacitive or magnetic field coupling β permanent shield needed
- Noise reduced by both ferrite and shield indicates both paths exist β address both
- A signal that becomes noisy only when a specific nearby IC is active, even though the two share no electrical connections, is frequently electromagnetic coupling through the PCB β the active IC’s bond wires or output traces are radiating, and the victim signal’s trace is acting as a receiving antenna.
- A design that works on revision A of a PCB but fails on revision B, where the only changes were “non-functional” layout improvements, is frequently revealing an unrecognized layout dependency β the “non-functional” changes modified a parasitic property that the circuit’s behavior depended on.