Power Electronics

Conducted EMI (Power)

Q: My converter passes below 2 MHz but fails from 10 to 20 MHz, which noise mode should I chase?

That spectral signature points to common-mode (CM) noise. Differential-mode (DM) noise flows in opposite directions on the two power conductors, in the same loop as the load current, is driven mainly by the converter ripple current, and dominates the lower part of the conducted band, roughly 150 kHz to 2 MHz. Common-mode noise flows in the same direction on both conductors and returns through the safety ground or parasitic capacitance to chassis. It is driven by fast dv/dt nodes such as the switch drain coupling through heatsink and transformer interwinding capacitance, and dominates the upper band, roughly 2 MHz to 30 MHz. A failure concentrated from 10 to 20 MHz is therefore almost always CM, so attack it with a larger common-mode choke and Y-capacitors, a transformer Faraday shield, and slower gate slew rather than touching the X-capacitor and series differential inductance that target DM.

Q: Why do CISPR and FCC stop the conducted measurement at 30 MHz and start it at 150 kHz?

CISPR and FCC define the conducted emission measurement range as 150 kHz to 30 MHz because below 150 kHz noise is dominated by mains harmonics and is addressed by separate harmonic-current limits (IEC 61000-3-2), while above 30 MHz the dominant coupling mechanism shifts from conduction along wires to radiation, which is covered by radiated-emission limits. Within this band the quasi-peak and average detectors of a CISPR 16 receiver scan the LISN output. Switching converters with fundamental switching frequencies of 50 kHz to a few hundred kHz produce harmonics that fall squarely in this range, and the steep edges of the switch node generate spectral energy extending to tens of megahertz, which is why this band is the critical compliance region for power electronics.

Q: If I move to a GaN design and raise the switching frequency, will my conducted EMI improve or get worse?

Raising the switching frequency shifts the fundamental and its harmonics higher in the spectrum. If the fundamental moves above 150 kHz, the first several harmonics that previously appeared inside the conducted band may move out of or to the high end of it, sometimes easing low-frequency differential-mode limits and allowing smaller filter components. However, faster switching, especially with wide-bandgap GaN and SiC devices that have very fast dv/dt and di/dt, increases the high-frequency spectral content and excites parasitic resonances, worsening common-mode emissions in the 5 MHz to 30 MHz region. The net effect depends on the slew rates, layout parasitics, and whether spread-spectrum or soft-switching techniques are used. In practice GaN converters often trade a smaller magnetic filter for a more demanding common-mode and radiated EMI design above 10 MHz.

/kuhn-DUK-tid E-M-I/

Inside any switch-mode power converter, the rapid on-off transitions of the power devices generate broadband noise current that travels back out along the AC mains or DC input cabling rather than radiating into free space. This conducted noise splits into common-mode and differential-mode components and is the primary reason a power supply must pass the 150 kHz to 30 MHz emission limits of CISPR 32 and FCC Part 15. Engineers quantify it with a line impedance stabilization network and suppress it using a multistage EMI filter built around a common-mode choke, X-capacitors, and Y-capacitors. Limit levels run 56 to 66 dBµV (quasi-peak) for Class B equipment, and filter attenuation of 40 to 60 dB is typical for an off-line converter.

Category: Power Electronics

Measurement Band: 150 kHz to 30 MHz

Class B QP Limit: 56 to 66 dBµV

How Switching Converters Generate Line Noise

A switch-mode converter is an efficient power processor precisely because it chops the input with hard, fast transitions instead of dissipating energy linearly. Each transition is a small step of voltage and current with rise and fall times in the tens of nanoseconds, which by Fourier decomposition contains harmonic energy reaching well past 30 MHz. Two distinct paths carry this energy back onto the supply lines. The differential-mode path follows the same loop as the load current and is driven by the trapezoidal ripple drawn through the input capacitor and its equivalent series resistance and inductance. The common-mode path is more subtle: the high dv/dt at the switch node couples through parasitic capacitance, the heatsink, and transformer interwinding capacitance into the safety ground, returning to the source through the green-wire and chassis.

Because the two mechanisms have different source impedances and different frequency signatures, they must be measured and filtered independently. Differential-mode noise tends to dominate from 150 kHz up to a couple of megahertz, where it is set by ripple amplitude and capacitor parasitics. Common-mode noise tends to dominate from a few megahertz to 30 MHz, where it is set by node slew rate and stray capacitance. A LISN presents a defined 50 ohm impedance to the equipment under test at RF while still passing mains power, so that the noise voltage developed across that impedance is repeatable from lab to lab. The receiver then applies CISPR 16 quasi-peak and average detectors, which weight the measurement by pulse repetition rate so that infrequent spikes are penalized less than continuous tones.

Suppression is a filtering problem. The classic off-line input filter places a common-mode choke in series with the lines, a pair of Y-capacitors from each line to ground to shunt common-mode current, and X-capacitors across the lines plus the choke leakage inductance to attenuate differential-mode current. Good layout, gate-drive slew control, shielding the switch node, and adding a Faraday shield in the transformer all reduce the noise at the source so the filter has less to do.

Common-Mode and Differential-Mode Governing Relations

LISN line-to-ground decomposition:
V_CM = (V_L1 + V_L2) / 2 V_DM = (V_L1 − V_L2) / 2

dBµV conversion (across 50 Ω LISN):
Level (dBµV) = 20 × log₁₀( V_noise / 1 µV )

Common-mode noise current from switch-node dv/dt:
i_CM ≈ C_par × (dv/dt)

CM choke corner frequency:
f_c = 1 / (2π × √(2 × L_CM × C_Y))

Where V_L1, V_L2 = the two LISN line voltages, C_par = parasitic switch-to-ground capacitance, L_CM = choke common-mode inductance, C_Y = line-to-ground capacitance. Example: C_par = 50 pF with dv/dt = 20 V/ns gives i_CM ≈ 1 A peak of common-mode drive.

Conducted Emission Limits and Filter Elements

Standard / Element	Frequency Range	Quasi-Peak Limit	Average Limit	Targets	Notes
CISPR 32 / FCC Class B	0.15 to 0.5 MHz	66 to 56 dBµV	56 to 46 dBµV	CM + DM	Residential, sliding limit
CISPR 32 / FCC Class B	0.5 to 5 MHz	56 dBµV	46 dBµV	CM + DM	Flat region
CISPR 32 / FCC Class B	5 to 30 MHz	60 dBµV	50 dBµV	Mostly CM	Slew-rate driven
Class A (industrial)	0.15 to 30 MHz	79 to 73 dBµV	66 to 60 dBµV	CM + DM	~13 dB looser than B
Common-mode choke	0.5 to 30 MHz	n/a	n/a	CM current	1 to 30 mH, high μ ferrite
Y-capacitor	0.15 to 30 MHz	n/a	n/a	CM shunt	1 to 4.7 nF, leakage limited
X-capacitor + leakage L	0.15 to 2 MHz	n/a	n/a	DM current	0.1 to 2.2 µF across line

Common Questions

Frequently Asked Questions

My converter passes below 2 MHz but fails from 10 to 20 MHz, which noise mode should I chase?

That spectral signature points to common-mode noise. Differential-mode dominates the low end, roughly 150 kHz to 2 MHz, where it is set by input ripple and capacitor ESR/ESL. Common-mode dominates the upper band, roughly 2 to 30 MHz, because it is driven by switch-node dv/dt coupling through parasitic and transformer interwinding capacitance to safety ground. A failure concentrated from 10 to 20 MHz almost always means CM, so attack it with a larger common-mode choke, better Y-capacitor placement, a transformer Faraday shield, and slower gate slew rather than touching the X-capacitor and differential elements.

Why do CISPR and FCC stop the conducted measurement at 30 MHz and start it at 150 kHz?

Below 150 kHz the spectrum is governed by mains harmonics, handled by separate IEC 61000-3-2 current limits. Above 30 MHz the dominant coupling shifts from conduction along cabling to radiation, covered by radiated-emission limits. Switching fundamentals of 50 kHz to a few hundred kHz drop their harmonics directly into this band, and fast edges spread energy to tens of megahertz, so 150 kHz to 30 MHz is the critical compliance window for power converters.

If I move to a GaN design and raise the switching frequency, will my conducted EMI improve or get worse?

Pushing the fundamental higher shifts the harmonic comb upward; if it moves above 150 kHz the first harmonics can leave the low end of the band, easing differential-mode limits and shrinking the magnetics. But faster edges, especially with GaN and SiC devices having very high dv/dt, raise high-frequency content and excite parasitic resonances, worsening common-mode emissions in the 5 to 30 MHz region. The net result depends on slew rates, layout parasitics, and whether spread-spectrum or soft-switching is used, and GaN often trades smaller magnetics for a harder CM and radiated fight above 10 MHz.

What does the LISN actually present to my power supply during a conducted scan?

The equipment under test is powered through a 50 ohm/50 µH line impedance stabilization network that presents a defined RF impedance to each line while still passing mains power. Each line-to-ground port feeds a CISPR 16 receiver applying quasi-peak and average detectors across 150 kHz to 30 MHz. The quasi-peak detector weights the reading by pulse repetition rate, and both the QP and average results must sit under the applicable Class A or Class B limit lines with adequate margin.

Related Terms

← Conditional Stability Conducted Emissions →