Common-Mode vs Differential-Mode Noise: Separate and Fix
Where common-mode and differential-mode noise come from, how to separate them with a LISN or current probe, and which chokes, X/Y capacitors and filters help.
Every noise current on a pair of conductors can be split into two parts: a differential-mode (DM) part that flows out on one conductor and back on the other, and a common-mode (CM) part that flows in the same direction on all conductors and returns through some other path, usually stray capacitance to ground. An X capacitor, a Y capacitor, a common-mode choke and a differential inductor each act mainly on one mode, so a filter chosen without knowing which mode dominates is a guess.
Where each mode comes from
Differential-mode noise is the switching current of the circuit itself leaking past the input capacitor. A switch-mode power supply draws its input current in pulses at the switching frequency. The bulk capacitor supplies most of each pulse, but its ESR and ESL let part of the ripple appear across the input terminals, so it flows out on the line and back on the neutral. The source is a current loop in the intended circuit, and DM noise is strongest at the switching frequency and its low harmonics. It typically dominates the conducted band up to a few MHz.
Common-mode noise is driven by voltage rather than current. Any node that swings rapidly, such as a MOSFET drain, couples through parasitic capacitance to the heatsink, the chassis or a secondary winding, and the displacement current I = C · dV/dt has to return to its source. It does so through the protective earth, the LISN and the reference ground plane, and then back up both supply wires together. For example, 50 pF between a drain tab and an earthed heatsink, with a 400 V swing in 50 ns, gives a peak of 0.4 A during each edge. CM noise typically dominates the upper conducted band and almost all cable radiation above 30 MHz.
Mode conversion
The two modes stay separate only while the two conductors are balanced, with equal impedance to ground. Any imbalance converts part of one mode into the other: a filter inductor in the line but not in the neutral, unequal Y capacitors, wires of different length, or a ground connection shared with one conductor. In practice a DM problem can reappear as CM on the far side of an unbalanced filter.
What the LISN actually measures
Each LISN measurement port terminates its line in 50 Ω to the reference ground. A total common-mode current I_CM splits equally between line and neutral and returns through both 50 Ω ports in parallel, while a differential current I_DM flows out through one port and back through the other. The port voltages are therefore V_L = 25 · I_CM + 50 · I_DM and V_N = 25 · I_CM − 50 · I_DM. The limit applies to each line separately, so depending on their relative phase the two modes add on one line and partly cancel on the other.
First, if line and neutral read almost the same at a failing frequency, one mode usually dominates there, but the readings alone do not tell you which. Second, if line and neutral differ by several dB, both modes contribute significantly, and suppressing only one of them can lower one line much more than the other, or even raise the other line. A spectrum analyzer shows only magnitudes, so CM and DM cannot be calculated from two separate L and N sweeps; the phase information is lost.
Four ways to separate the modes on the bench
| Method | How it is done | What it tells you | Limitation |
|---|---|---|---|
| CM/DM separator (mode rejection network) | Inserted between the two LISN ports and the receiver; it adds or subtracts the L and N signals with a 0° or 180° combiner | CM and DM spectra directly, comparable with the same limit line | Its rejection of the unwanted mode limits accuracy; check it across 150 kHz to 30 MHz |
| Current probe around both wires | Clamp the probe around line and neutral together, but not the earth wire | Total CM current; DM currents cancel inside the probe | The earth wire must stay outside the clamp |
| Current probe with one wire reversed | Pass the line through the probe in one direction and the neutral back through in the other | Twice the DM current; CM currents cancel | Keep the loop small and symmetrical |
| Substitution test | Temporarily add a large X capacitor, then separately a Y capacitor pair or a clamp ferrite over both wires | Which mode responds; a drop of several dB points to the dominant mode | Quick but indirect; a part that changes the circuit balance can mislead |
For both clamp methods, convert the receiver reading to current with the probe's transfer impedance: I (dBµA) = reading (dBµV) − Z_T (dBΩ). The CM contribution at each port is 25 Ω × I_CM, and the DM contribution is 50 Ω × I_DM, which equals 25 Ω × the reversed-wire reading. Adding 28 dB (20·log 25 = 27.96 dB) to either reading in dBµA therefore estimates that mode's share of the port voltage in dBµV. The estimate holds where the LISN impedance is close to 50 Ω, above roughly 1 MHz; at 150 kHz a 50 µH LISN presents only about 34 Ω.
Countermeasures mode by mode
| Measure | Mode | How it works | Practical limits |
|---|---|---|---|
| X capacitor, line to line | DM | Shunts DM current close to the source | Safety-rated X-class part; ESL limits it above a few MHz; large values usually need a discharge resistor |
| Y capacitors, line and neutral to earth | CM | Give CM current a short return path to the chassis before it reaches the LISN | Touch-current limits cap the total; 4.4 nF at 230 V, 50 Hz already passes about 0.32 mA; medical equipment allows far less |
| Common-mode choke | CM | Line and neutral fluxes cancel for DM, so load current does not saturate the core, while CM sees the full inductance | Winding capacitance bypasses it at high frequency; its leakage inductance adds a small DM inductance |
| Differential inductor, one per line | DM | Series impedance to DM current | Carries the full load current; inductance under DC bias or peak current can fall sharply |
| Clamp ferrite over the whole cable | CM | Adds lossy impedance to CM current; DM flux cancels | Low impedance below a few MHz; most effective from about 30 to 300 MHz |
| Lower dV/dt and parasitic capacitance | CM, at the source | Gate resistor or snubber, heatsink tied to the primary DC return instead of earth, shield winding in the transformer | Costs efficiency; recheck thermal design and safety insulation |
A typical single-stage mains filter combines both modes. With a 10 mH CM choke and two 2.2 nF Y capacitors (4.4 nF in parallel for CM), the CM section has a natural frequency of about 24 kHz. If the choke's leakage inductance is 100 µH and it works against a 470 nF X capacitor, the DM section sits at about 23 kHz. An ideal second-order section falls at 40 dB per decade above its natural frequency, which would give about 32 dB at 150 kHz for both. Real attenuation is usually lower and uneven, because source and load impedances are not 50 Ω, the choke's parasitic capacitance bypasses it at high frequency, and the filter can resonate.
Common mistakes
- Clamping a current probe around one wire and calling the reading common mode. One wire carries half the CM current plus all of the DM current.
- Adding a DM inductor and checking it unloaded. A powder or ferrite core that looks fine on an LCR meter can lose most of its inductance at full load current.
- Building an unbalanced filter, for example an inductor in the line only or Y capacitors of unequal value. The imbalance converts DM to CM, and the problem moves instead of disappearing.
- Putting a good filter in a bad position. If the unfiltered input wires run past the transformer or the switching node, noise couples onto the filtered side and bypasses the filter. Mount it at the entry point and keep input and output wiring apart.
- Fixing the conducted test and forgetting the radiated one. A mains CM choke that works at 1 MHz may offer little impedance at 100 MHz because of its winding capacitance, and the cable still radiates CM current above 30 MHz.
Calculators used in this post
More posts
Results and summaries are for reference. For certification and test reports use the latest official standard text and calibrated instrument data.