Why does a bigger choke often filter worse?

Because every winding has parasitic capacitance across it, and the pair resonates. Above that self-resonant frequency the choke is a capacitor, and its impedance falls with frequency instead of rising. Larger inductance usually means more turns and more parasitic capacitance, so the resonance moves down. A 1 mH choke with 5 pF of parasitic capacitance resonates near 2.25 MHz; a 10 mH choke with 20 pF resonates near 0.36 MHz. If the problem sits at 20 MHz, the 10 mH part contributes nothing useful and the 1 mH part contributes more. Choose by where the resonance is, not by the inductance printed on the part.

What limits the size of a Y-capacitor?

Safety, not EMC. A Y-capacitor bridges the mains to earth, so at mains frequency it passes a current straight to the protective conductor, and that current flows through anyone who touches the product if the earth connection is ever lost. It follows from I = 2 pi f C V: at 230 V and 50 Hz, 1 nF passes about 72 uA, 2.2 nF about 159, and 10 nF about 723. The permitted maximum comes from the safety standard that applies to the product, and it is a hard ceiling. EMC never gets to argue with it, which is why Y-capacitance is usually the first constraint a mains filter design meets.

Why does the same filter behave differently in two products?

Because a filter divides against the impedances around it, and the common-mode impedance of a real installation is neither known nor stable. Insertion loss follows 20 log10(1 + Z/(Zs + Zl)), so the same 1 kilohm choke gives about 21 dB in a 100 ohm environment and far less where the surrounding impedance is higher. Cable length, routing, chassis bonding and what the product is plugged into all move that impedance. This is why filters are selected by test rather than by catalogue, and why a filter that fixed the last product may do nothing for the next one.

Common-mode filtering: chokes, ferrites, Y-caps

Author Hugues Orgitello Last updated 30 August 2026 ~5 min read

Guide, EMC fundamentals

Common-mode current is the thing that radiates, so common-mode filtering is where most emission problems are finally solved. It is also where the most money is wasted, because the intuitive move, fitting a bigger part, frequently makes the result worse. This page explains what each component does, and derives the reason bigger is not better.

Why common mode is the target

From coupling mechanisms, common-mode current on a cable radiates roughly ten thousand times more efficiently than differential current in a loop of realistic area. A few uA of it produce as much field as tens of milliamps of the signal you meant to send.

So a filter that removes common-mode current while leaving the signal alone is worth a great deal, and that is exactly what the three components below try to do.

The common-mode choke

Both conductors are wound on one core, in the same sense.

Differential current flows out on one conductor and back on the other. The two fluxes oppose and cancel, so the choke presents almost no impedance to the wanted signal, and the core does not saturate on load current.

Common-mode current flows the same direction on both. The fluxes add, and the winding presents its full inductance.

One part, two behaviours, selected by which mode is passing. Ideally the impedance is:

Z = 2 pi f L

Inductance	150 kHz	1 MHz	10 MHz	30 MHz
0.5 mH	0.5 kilohm	3.1 kilohm	31 kilohm	94 kilohm
1 mH	0.9 kilohm	6.3 kilohm	63 kilohm	189 kilohm
10 mH	9.4 kilohm	63 kilohm	628 kilohm	1885 kilohm

Read that table and the answer looks obvious: fit the 10 mH part. It is the wrong answer.

Why bigger is worse: self-resonance

Every winding has parasitic capacitance across it, and inductance with capacitance across it is a parallel resonator. Above:

f0 = 1 / (2 pi sqrt(L Cp))

the choke stops being an inductor and becomes a capacitor. Its impedance falls with frequency instead of rising, and it filters progressively less.

Choke	Parasitic C	Self-resonance
1 mH	5 pF	2.25 MHz
1 mH	20 pF	1.13 MHz
10 mH	5 pF	0.71 MHz
10 mH	20 pF	0.36 MHz

More inductance normally means more turns, and more turns mean more parasitic capacitance, so the resonance moves down as the part gets bigger. A 10 mH choke with 20 pF is already past resonance at 360 kHz. If the failing harmonic sits at 20 MHz, that part contributes nothing and the 1 mH part contributes more.

Choose a choke by where its impedance peaks, not by its inductance. Manufacturers publish an impedance-versus-frequency curve for exactly this reason; the inductance on the label describes only the region below the peak.

Ferrites work by loss, not by reflection

A ferrite bead or sleeve is often described as an inductor, and below the material's loss band it behaves as one. That is not the useful part.

An inductive impedance reflects energy back towards the source, which relocates a problem rather than removing it, and can make things worse by feeding a resonance. Over the band where the ferrite material is lossy, its impedance becomes substantially resistive, and common-mode current is converted into a small amount of heat. The energy is gone.

Two consequences:

A bead is specified as an impedance at a stated frequency, typically 100 ohms at 100 MHz, not as an inductance. A part quoted only in nH is being sold on the wrong parameter.
Material grade decides the band. A ferrite optimised for 100 MHz does very little at 10 MHz. Fitting a larger bead of the wrong material buys nothing.

Turns matter more than mass: passing the cable through the core several times multiplies the impedance, roughly with the square of the number of turns, until parasitic capacitance between turns takes over.

Y-capacitors, and the ceiling that is not negotiable

A Y-capacitor connects a mains conductor to protective earth, giving common-mode current a low-impedance path home instead of out along the cable. It is the other half of a mains filter.

Its size is limited by safety, not by EMC. At mains frequency it passes a current to the earth conductor, and if the earth connection is ever lost, that current passes through whoever is touching the product. From I = 2 pi f C V, at 230 V and 50 Hz:

Y-capacitance	Current at 50 Hz
1 nF	72 uA
2.2 nF	159 uA
4.7 nF	340 uA
10 nF	723 uA

The permitted maximum is set by the safety standard applying to the product, and medical equipment is held far tighter than ordinary equipment. Whatever that figure is, EMC does not get to argue with it. Y-capacitance is usually the first constraint a mains filter design runs into, and the reason a filter that would solve the problem cannot be fitted.

Y-capacitors are also a safety-critical component class in their own right, built to fail open rather than short, which is what IEC 60384-14 governs. An ordinary capacitor of the same value is not a substitute.

Why the same filter behaves differently in two products

A filter is a divider, and it divides against the impedances on either side of it:

IL = 20 log10(1 + Z / (Zs + Zl))

Filter impedance	In a 100 ohm environment
100 ohm	6.0 dB
1 kilohm	20.8 dB
10 kilohm	40.1 dB

The difficulty is that the common-mode impedance of a real installation is neither known nor stable. It depends on cable length, routing, chassis bonding, and what the product is plugged into. The same choke that gave 21 dB on one product can give a fraction of that on the next.

This is why filters are selected by measurement rather than from a catalogue, and why CISPR 17 specifies filter characterisation in defined impedance systems: the published figure is a comparable number, not a promise about your product.

Key takeaways

Common mode is the target, because it radiates thousands of times more efficiently than differential mode.
A choke cancels differential flux and adds common-mode flux, so one part passes the signal and blocks the noise.
Bigger is often worse. More turns means more parasitic capacitance and a lower self-resonance: 10 mH with 20 pF is already a capacitor at 360 kHz.
Ferrites should be chosen for loss, specified as an impedance at a frequency, with material grade deciding the band.
Y-capacitance is capped by safety, not EMC, and that ceiling is not negotiable.
Insertion loss depends on the surrounding impedance, which is unknown and unstable, so filters are selected by test.

Frequently asked questions

How does a common-mode choke pass the signal but block the noise?: Both conductors are wound on the same core in the same sense. Differential current, which flows out on one and back on the other, produces two opposing fluxes that cancel, so the choke presents almost no impedance to the wanted signal and does not saturate on load current. Common-mode current flows the same way on both conductors, so the fluxes add and the winding presents its full inductance. One component therefore does two different things depending on which mode is passing through it, which is why it is the standard answer to a cable-radiation problem.
Why does a bigger choke often filter worse?: Because every winding has parasitic capacitance across it, and the pair resonates. Above that self-resonant frequency the choke is a capacitor, and its impedance falls with frequency instead of rising. Larger inductance usually means more turns and more parasitic capacitance, so the resonance moves down. A 1 mH choke with 5 pF of parasitic capacitance resonates near 2.25 MHz; a 10 mH choke with 20 pF resonates near 0.36 MHz. If the problem sits at 20 MHz, the 10 mH part contributes nothing useful and the 1 mH part contributes more. Choose by where the resonance is, not by the inductance printed on the part.
What is a ferrite bead actually doing?: Dissipating rather than reflecting. At low frequency a ferrite is mostly inductive and reflects energy back towards the source, which moves a problem rather than removing it. Over the frequency band where the material is lossy its impedance becomes substantially resistive, and common-mode current is converted to a small amount of heat. That is why a bead is specified by its impedance at a stated frequency rather than by an inductance, and why material grade matters more than size: a bead optimised for 100 MHz does very little at 10 MHz.
What limits the size of a Y-capacitor?: Safety, not EMC. A Y-capacitor bridges the mains to earth, so at mains frequency it passes a current straight to the protective conductor, and that current flows through anyone who touches the product if the earth connection is ever lost. It follows from I = 2 pi f C V: at 230 V and 50 Hz, 1 nF passes about 72 uA, 2.2 nF about 159, and 10 nF about 723. The permitted maximum comes from the safety standard that applies to the product, and it is a hard ceiling. EMC never gets to argue with it, which is why Y-capacitance is usually the first constraint a mains filter design meets.
Why does the same filter behave differently in two products?: Because a filter divides against the impedances around it, and the common-mode impedance of a real installation is neither known nor stable. Insertion loss follows 20 log10(1 + Z/(Zs + Zl)), so the same 1 kilohm choke gives about 21 dB in a 100 ohm environment and far less where the surrounding impedance is higher. Cable length, routing, chassis bonding and what the product is plugged into all move that impedance. This is why filters are selected by test rather than by catalogue, and why a filter that fixed the last product may do nothing for the next one.