Conducted versus radiated emissions sounds like two different faults with two different cures. In most products it is one noise current viewed through two instruments. The measurement changes at the frequency where the attached cable stops behaving like a wire and starts behaving like an antenna, and the underlying source is usually identical on both sides of that line.
Two measurements, one current
Conducted emissions are measured by inserting a line impedance stabilization network between the product and the supply, presenting a defined impedance and tapping off the radio-frequency voltage that the product pushes onto the mains. Radiated emissions are measured with an antenna at a set distance while the sample rotates. Below roughly thirty megahertz, ordinary mains cords are short compared with a wavelength and radiate inefficiently, so the conducted measurement captures the problem. Above that, the same cord approaches a useful fraction of a wavelength and becomes a genuine antenna driven by the common-mode current on it. The regulatory frequency split exists because of that transition, not because two independent phenomena suddenly swap places at a round number.
Why the cable is nearly always the radiator
Circuit boards are small. A trace a few centimeters long is a hopeless antenna at a hundred megahertz, radiating a tiny fraction of what a meter of cable does at the same current. What boards are very good at is generating common-mode voltage between their ground system and the outside world, and that voltage drives current onto every attached cable. The cable then does the radiating. This is why an enclosure that passes with all cables removed can fail badly with a printer lead attached, and why swapping one cable for a shielded type sometimes moves a failing peak into compliance. It also explains why the most productive fixes are filters and shield terminations at the connector, right where cable current can still be controlled.
Harmonic combs point back to their source
Switching converters and clocked digital circuits produce energy at a fundamental and at every harmonic of it, spaced evenly across the spectrum. Reading the spacing of the peaks identifies the source without any probing: peaks every hundred kilohertz suggest a converter running at that rate, peaks every twenty-five megahertz suggest an oscillator. The envelope of those harmonics carries more information. Faster switching edges push energy higher in frequency, which is why an improvement in efficiency from a faster switch can be paid for in emissions. Deliberate edge-rate control, a small series resistor on a clock line, or a snubber on a switching node all trade a little loss for a lot of high-frequency content, and they attack the problem at the source rather than at the cable.
Fixing one measurement often fixes the other
Because the same common-mode current feeds both, a change that reduces it helps in both bands. A common-mode choke at the mains inlet lowers the conducted voltage and also starves the cord of the current it needs to radiate. Better still are changes further upstream: reducing the capacitance from a switching node to the chassis, giving return current a tight low-area path, or bonding a shield properly at the point of entry. Beware of fixes that only move energy around. Adding line-to-chassis capacitance can pass a conducted test while increasing chassis current, and lengthening a ground strap can shift a resonance so a peak moves out of one measured band and into another. Verify both measurements after every change.
Detectors and bandwidths change the number you read
Emission limits are specified with particular detectors and resolution bandwidths, and those choices are not cosmetic. A peak detector records the highest instantaneous value, a quasi-peak detector weights by repetition rate to approximate how annoying a disturbance sounds on a radio, and an average detector smooths it. An impulsive noise that repeats rarely can sit well above the limit on peak and comfortably below on quasi-peak. Resolution bandwidth matters too, because broadband noise scales with it while a narrow carrier does not. This is why a bench scan with the wrong settings can produce panic or false confidence, and why comparing your own measurement to a published limit requires matching the detector and bandwidth the standard specifies.
What to check before you act
- Measure with and without each cable to find the real radiator.
- Read harmonic spacing to identify the source frequency.
- Re-test both conducted and radiated bands after any fix.
- Match detector and bandwidth settings to the standard you are citing.
Common questions
Why did my product pass conducted and fail radiated?
Usually because a cable resonates in the upper band. The noise source may be unchanged; only the efficiency of the attached antenna has increased.
Does a metal enclosure fix radiated emissions?
Only if every cable is filtered or shield-bonded at the wall. Otherwise the cables carry the current straight past the enclosure.
The practical takeaway
Stop treating the two scans as separate projects. Find the common-mode current that feeds both, reduce it at the source or at the connector, and confirm the change in both bands with the correct detector settings. That approach fixes causes instead of chasing individual peaks.
References and further reading
- FCC: electromagnetic compatibility division reference material
- FCC: radio frequency energy, emissions and equipment authorization
- NFPA: electrical safety guidance for equipment and wiring
Numerical scenarios are illustrative unless identified otherwise. Follow the exact product instructions; component ratings and local installation requirements can differ.
