Common Mode Choke Calculator

Common-mode inductance from turns and core, the leakage that is the only differential inductance the filter actually gets, the self-resonance above which the choke is a capacitor and attenuates nothing, and the flux the current imbalance puts in a core the load current was supposed to leave empty.

Last updated 24 Sep 2026

Inductance, leakage, self-resonance and what actually saturates it

Core + turns → CM inductance

How is the core given?A_L if you have the core data sheet, dimensions if you are working from a ring you measured. Either way the page needs the effective area as well, because that is what turns flux into flux density.

Core A_L (nH per turn squared)4,300 nH is the TDK/EPCOS R 25.3 × 14.8 × 10.0 ring in N30, a typical high-permeability MnZn ferrite for mains and DC-bus common-mode chokes. A_L carries a wide tolerance — ±25% on this one — and it is the number the whole page scales with. Use the figures from your part’s datasheet; typical values vary widely between manufacturers.

Effective core area A_e (mm²)51.26 mm² for the same R 25.3 × 14.8 × 10.0 core. It sets the flux density and therefore the saturation answer, and it is not derivable from A_L.

Ring outer diameter (mm)Only used when the core is entered by its dimensions.

Ring inner diameter (mm)

Ring height (mm)

Initial relative permeability of the material4,300 for N30, 10,000 for T38, 2,300 for N87. This is the LOW-frequency figure; above a few hundred kilohertz a MnZn ferrite’s permeability falls and becomes largely lossy, which is the reason a common-mode choke made of it still works up there — as a resistor, not as an inductor. Use the figures from your part’s datasheet; typical values vary widely between manufacturers.

Turns per windingBoth windings have the same turns count; that is what makes the differential flux cancel. The common-mode inductance goes as the square of this number, so twenty turns is four times ten.

How is the leakage inductance given?Leakage is the part of each winding’s flux that does not link the other, and it is the only differential-mode inductance the filter gets from this part. Measure it if you can: short one winding and measure the other.

Leakage as a percentage of the common-mode inductance (%)TDK quotes “approx. 1% stray inductance” for its B82721 current-compensated ring core chokes. Across Coilcraft’s 63-part CMT range of power-line chokes the published maximum leakage runs from about 0.7% to 2.0% of the minimum rated inductance. It depends on how the windings are laid out — sector-wound parts leak more than bifilar ones, deliberately — so it is a part property, not a constant. Use the figures from your part’s datasheet; typical values vary widely between manufacturers.

Measured leakage inductance (µH)Both windings in series opposing, which is what Coilcraft’s data sheets mean by “measured from pin 1 to pin 4 with pins 2 and 3 shorted”. That is the inductance the differential current sees.

How is the winding capacitance given?Most common-mode choke data sheets never print a capacitance, but they all plot impedance against frequency, and the frequency of the peak is the self-resonance — which gives you the capacitance for free.

Winding capacitance (pF)The distributed capacitance across each winding, a few picofarads to a few tens. It is set by how the turns sit next to each other; a single-layer winding with a gap between the start and finish is the usual way to keep it small. Use the figures from your part’s datasheet; typical values vary widely between manufacturers.

Frequency of the data sheet’s impedance peak (kHz)Read it off the impedance curve. For power-line common-mode chokes it is usually between 100 kHz and a few megahertz, and it falls as the inductance rises — which is exactly why the biggest choke is not always the best one.

Peak impedance at that resonance (Ω)The height of the impedance peak on the data sheet curve. It is finite because the core is lossy, and at and above resonance that loss is doing the attenuating — a high-permeability MnZn choke works as a resistor up there, not as an inductor. 7.76 kΩ at 0.48 MHz is Coilcraft’s figure for the CMT1-3.0-6L. Use the figures from your part’s datasheet; typical values vary widely between manufacturers.

Differential (load) current through the choke (A rms)The current the equipment draws. It flows out along one conductor and back along the other, so its flux cancels in the core and it produces almost no flux density at all — which is the whole reason a common-mode choke can carry it on a small core. It still produces copper loss.

Imbalance between the go and return currents (% of the load current)The part of the go current that does not come back on the return conductor. THIS is what magnetises the core. In compliant mains equipment the imbalance is the earth leakage current and is milliamps, so it is negligible; on an isolated DC platform where part of the return finds a path through structure it can be percent, and then it dominates everything.

Common-mode noise current (mA rms)Measured with a current probe, or estimated by a method you can state. It cannot be derived from the circuit diagram: common-mode current is set by parasitic capacitance from the switching nodes to chassis, which depends on layout, heatsink mounting and harness routing. This page will not invent it.

Frequency of interest (kHz)Where you want the impedance. 10,000 kHz is the top of the CE102 band and is where the difference between 2πfL and what the choke actually does is most brutal.

Saturation flux density at your operating temperature (mT)Use the HOT figure, not the 25 °C one. N30 is 380 mT at 25 °C and 240 mT at 100 °C; most high-permeability MnZn ferrites lose a third or more between room temperature and their operating temperature, and their Curie point can be as low as 130 °C. Use the figures from your part’s datasheet; typical values vary widely between manufacturers.

DC resistance per winding (mΩ)Both windings carry the full load current, so the copper loss is twice this times the current squared. On a mains or DC-bus choke this is the loss that matters; the core loss is negligible because there is hardly any flux in the core.

Thermal resistance of the choke to ambient (°C/W)20 to 30 °C/W for a 25 mm toroid in still air is a reasonable starting point; a potted or bracket-mounted part does better. The temperature rise matters twice over, because the saturation flux density falls as the core heats.

Two windings of the same turns count on one core. The load current goes out along the top conductor and back along the bottom one, so it passes through the two windings in opposite senses and its flux cancels — which is why a common-mode choke can carry a large load current on a small core. What does not cancel is the part of the go current that does not return on the return conductor: the imbalance, drawn here leaving through the chassis. That current sees the full common-mode inductance and sets the flux in the core. The core turns amber as the peak flux density approaches the saturation figure you entered and red past it.

1.72mHExample

20 turns per winding on a TDK R 25.3 × 14.8 × 10.0 ring in N30 (4,300 nH/N², A_e 51.26 mm²), with 18 pF of winding capacitance, 4 A of load current and 2% of it not coming back

One core, two windings, and only the current that fails to cancel

L_CM = N² · A_L L_leakage ≈ (0.5 to 2%) · L_CM, from the data sheet
B_peak = A_L · N · Î_imbalance ÷ A_e (the balanced load current contributes nothing)
f_SRF = 1 ÷ (2π√(L_CM · C_winding))
|Z| = 1 ÷ √( (1÷R_p)² + (2πfC − 1÷2πfL)² )
P_copper = 2 · I_load² · R_DC

N²: the common-mode inductance is the single-winding self inductance, because for common mode the two windings sit in parallel and each carries half. That is the number a data sheet quoting inductance per winding gives you
leakage: the flux that does not link the other winding. It is the only differential-mode inductance you get, it depends on the winding geometry, and it cannot be derived from the core — take it from the data sheet or measure it
Î imbalance: the PEAK of the current that does not cancel. The differential load current contributes nothing at all, however large it is
Rp: the loss resistance the data sheet’s impedance peak measures. Without it the model would predict infinite impedance at resonance; with it, the choke’s real behaviour up there — a lossy resistor — comes out right

Worked example

20 turns per winding on a TDK R 25.3 × 14.8 × 10.0 ring in N30 (4,300 nH/N², A_e 51.26 mm²), with 18 pF of winding capacitance, 4 A of load current and 2% of it not coming back
The common-mode inductance is N²A_L = 20² × 4,300 nH = 1.72 mH, and the leakage at the 1.0% TDK quotes for its ring-core chokes is 17.2 µH — which is all the differential-mode inductance this part gives the filter
With 18 pF across the winding it self-resonates at 904.5 kHz. At the 10 MHz top of the CE102 band an ideal 2πfL would promise 108.1 kΩ; the part delivers 886 Ω. That is 41.7 dB of attenuation the schematic says you have and the hardware does not
Now the flux. All 4 A of load current goes out through one winding and back through the other, so its flux cancels exactly and the core sees nothing — had it not cancelled it would have produced 9.491 T
What does not cancel is 2% of it: 80 mA rms, 113.1 mA peak, giving B = A_L·N·Î ÷ A_e = 189.8 mT. Add 11.86 mT from the 5 mA of common-mode noise current and the core is at 201.7 mT against the 240 mT entered — 1.19× margin, from a two per cent imbalance on a choke that carries 4 A without noticing
The copper is the only real heat: 2 × 4² × 40 mΩ = 1.28 W, and 22 °C/W of thermal resistance turns that into 28.2 °C of rise — which is why the saturation figure above should be the hot one

Published leakage inductance of real power-line common-mode chokes

Coilcraft’s inductance is a minimum and its leakage a maximum, so these fractions are pessimistic rather than typical; across the whole 63-part CMT range they run from about 0.7% to 2.0%. Coilcraft measures inductance per winding at 15.75 kHz and leakage “from pin 1 to pin 4 with pins 2 and 3 shorted” — the two windings in series opposing, which is the inductance the differential current sees. TDK states “approx. 1% stray inductance” for its B82721 ring-core chokes. None of this is a law: a sector-wound part leaks several times more, on purpose, so that the leakage can do duty as the differential-mode inductor.
Part	Rated L per winding	Leakage (max)	Leakage as a fraction
Coilcraft CMT1-3.0-6L	3.0 mH min	35 µH	1.17%
Coilcraft CMT1-15.0-1L	15.0 mH min	233 µH	1.55%
Coilcraft CMT3-8-4L	8.0 mH min	58 µH	0.73%
Coilcraft CMT1-.3-15L	0.3 mH min	6.0 µH	2.00%
Coilcraft CMT4-125-1L	125.0 mH min	1,400 µH	1.12%
TDK B82721 series	0.2 to 47 mH	2.5 to 500 µH	“approx. 1%”

What magnetises the core, and what does not

The middle two rows are the ones that catch people. A choke carrying 10 A of perfectly balanced load current sees no flux at all; the same choke with a couple of percent of that current finding its way back through the chassis instead of the return conductor can be close to saturation. In compliant mains equipment the imbalance IS the earth leakage current and is milliamps, so it never bites — but that is a property of the earth-leakage limit, not of the choke. On an isolated DC platform nothing enforces it.
Current	Direction in the two windings	Flux in the core
Differential load current	out on one conductor, back on the other	cancels exactly — zero, however large
Imbalance: go current that returns elsewhere	one winding only	full effect — this is what saturates the part
Common-mode noise current	the same way along both conductors	full effect, but at a frequency where the permeability is far lower
DC on one line only	one winding only	full effect, and it is a DC bias the AC flux sits on top of

Why the load current does not saturate it and two per cent of it might

A common-mode choke is two windings of the same turns count on one core, phased so that currents flowing the same way along both conductors produce aiding flux and currents flowing in opposite directions produce opposing flux. Everything follows from that one sentence.

Common mode. Both windings carry half of the total common-mode current and both are driven by the same voltage, so they sit in parallel and the inductance the total current sees is the single-winding self inductance, N²·A_L. That is the number a data sheet is giving you when it says “inductance per winding”, and it goes as the square of the turns — but so does the winding capacitance, roughly, which is why doubling the turns does not double the useful impedance.

Differential mode, and the leakage. The load current goes out through one winding and back through the other, so its ampere-turns cancel and the choke is almost invisible to it. Almost: the two windings do not couple perfectly, and the flux that fails to link the other winding is the leakage inductance. That leakage is the only differential-mode inductance the filter gets from this part, and for power-line chokes it runs at roughly 0.5 to 2% of the common-mode inductance — TDK states “approx. 1%” for its B82721 ring-core parts, and Coilcraft’s published maxima across the CMT range work out between about 0.7% and 2.0%. It is a property of the winding geometry, not of the core, so it cannot be calculated from anything on this page; take it from the data sheet or measure it by shorting one winding.

Self-resonance, and the blunt consequence. Every winding has capacitance across it, and L and C resonate. Below that frequency the choke is an inductor and 2πfL is roughly right. At it, the impedance peaks at whatever the core loss allows — a finite number that the data sheet plots and that this page takes as an input, because a lossless model would predict infinity. Above it the winding capacitance is in charge, the part is a capacitor, and it provides no common-mode attenuation at all; worse, it is a path across the choke for the noise the choke was fitted to block. Power-line common-mode chokes self-resonate somewhere between about 100 kHz and a few megahertz, which is inside the band everybody cares about. Where an ideal 2πfL predicts tens of kilohms at 10 MHz, a real part of the same inductance is often delivering a few hundred ohms. Above roughly 10 to 30 MHz the answer stops being component values and becomes layout: the loop the capacitor is mounted in, where the heatsink is bonded, how the harness is routed. A component-value prediction carried to 100 MHz without that caveat is wrong.

And saturation, which is the one that catches people. The core MMF is N times the sum of the two winding currents taken in the go direction, so a perfectly balanced load current contributes exactly nothing — a choke carrying 10 A sees no flux from it, which is the entire reason a 6 A common-mode choke can be the size of a thimble. What magnetises the core is the current that does not come back: B = A_L·N·Î ÷ A_e, using the imbalance and not the load current. In compliant mains equipment the imbalance is the earth leakage current and is milliamps, so it never matters. On an isolated DC bus, or anywhere part of the return finds a path through the chassis, a couple of per cent of the load current can put a high-permeability ferrite close to saturation — and a saturated common-mode choke has no inductance, so it has no impedance, so the filter is not there. That failure only appears under load, which is why it is so often missed on the bench and found in the chamber.

Use the hot saturation figure. N30 is 380 mT at 25 °C and 240 mT at 100 °C, and its Curie temperature is only just above 130 °C. The copper loss computed here — twice I²R, because both windings carry the load current — is what heats it, and the margin you think you have at room temperature is not the margin you have in the enclosure.

What this page leaves out. Core loss is not computed: with flux this small it is negligible beside the copper, and where the flux is large the right model is Steinmetz, on the core loss calculator. The winding’s own AC resistance is not computed either; see the litz wire calculator for skin and proximity effect. For the rest of the winding arithmetic see the toroid inductor turns calculator and the gapped core inductance calculator; for the capacitors either side of this choke, the Y capacitor limit calculator.

Frequently asked questions

Does the load current saturate a common mode choke?

Essentially not at all, as long as it is balanced. The load current flows out through one winding and back through the other, so its ampere-turns cancel and it produces no flux in the core. That is the whole point of the construction and it is why a small core can carry a large current. What does saturate it is the current that does not cancel — the imbalance.

How much leakage inductance does a common mode choke have?

For power-line parts, roughly 0.5 to 2% of the common-mode inductance. TDK quotes “approx. 1%” for its B82721 ring-core chokes; Coilcraft’s published maxima across the CMT range come out between about 0.7% and 2.0% of the minimum rated inductance. It depends on how the windings are laid out, so it is a part property — take it from the data sheet, or short one winding and measure the other.

What happens above a common mode choke’s self-resonant frequency?

The winding capacitance takes over and the part behaves as a capacitor. Its impedance falls with frequency instead of rising, so it provides no common-mode attenuation there at all, and it is in fact a path across itself for the noise. A smaller choke with a higher resonance, or two in series, is the usual fix; a bigger one lowers the resonance and makes it worse.

Why does my choke measure far less impedance than 2πfL says?

Because 2πfL ignores two things the part cannot: the winding capacitance, which drags the impedance down above resonance, and the fact that a high-permeability MnZn ferrite’s permeability falls and becomes largely lossy above a few hundred kilohertz. The page prints both numbers side by side and the difference in decibels, because that difference is a common reason a filter that predicted a large margin measures a small one.

Should I use the common mode choke’s leakage as my differential inductor?

It is a perfectly legitimate design and sector-wound chokes exist to make it work — they are wound to leak deliberately, several times more than a bifilar part. The risk is relying on it without checking: the leakage carries a wide tolerance, it is rarely specified as a minimum, and unlike the common-mode inductance it sees the full load current, so it can saturate on its own.

Related calculators

References

TDK/EPCOS, Ferrites and accessories — Toroids (ring cores) R 25.3 × 14.8 × 10.0, data sheet. The core used for the defaults: l_e 60.07 mm, A_e 51.26 mm², V_e 3,079 mm³, and A_L values of 2,200 nH in N87, 3,000 in N95, 4,300 in N30, 5,000 in T35 and T65, 6,500 in T37 and 10,000 in T38.
TDK/EPCOS, SIFERRIT material N30, data sheet. Initial permeability 4,300 ±25%, saturation flux density 380 mT at 25 °C and 240 mT at 100 °C (1,200 A/m, 10 kHz), Curie temperature above 130 °C, optimum frequency range 0.01 to 0.40 MHz. The derived relative permeability from the ring’s own A_L, l_e and A_e is asserted against this figure when the batch is built.
TDK, Current-compensated ring core double chokes B82721A/J/K, data sheet. Nominal inductance 0.2 to 47 mH measured at 10 kHz (100 kHz for parts at or below 1 mH) with an Agilent 4284A at 0.1 mA, stray inductance 2.5 to 500 µH measured at 5 mA, and the statement “approx. 1% stray inductance for symmetrical interference suppression”. Also the note that the inductance falls by less than 10% at DC magnetic bias with the rated current.
Coilcraft, CMT series power line common mode chokes, parametric data. Inductance per winding at 15.75 kHz and 0 A DC; leakage inductance “measured from pin 1 to pin 4 with pins 2 and 3 shorted”; peak common-mode impedance and the frequency at which it occurs. The CMT1-3.0-6L used in the table is 3.0 mH minimum, 35 µH maximum leakage, 27 mΩ DCR, 6.0 A, peak 7.76 kΩ at 0.48 MHz.
Würth Elektronik, application note ANP146a, WE-CMDC common mode chokes. The coupling coefficient definition k = M/√(L₁L₂) and its relation to the leakage, k = (L₀ − L_S)/L₀; the statement that differential currents generate fields that “ideally compensate each other completely” so that the differential inductance approaches zero; and that above self-resonance “the inductor’s impedance becomes capacitive, reducing its effectiveness”.
Coilcraft, Common Mode Filter Inductor Analysis, application note. The definition of leakage inductance as “the amount of inductance which is not coupled to any other windings through a shared core”, the self-resonant frequency 1/(2π√(LC)) and its origin in the distributed capacitance between turns, with measured self-resonances from 0.2 to 6.0 MHz across 23 parts.