EMI Filter Designer Calculator
EMI Filter Designer Calculator
Size the X capacitor, the Y capacitors and the common-mode choke of a single-stage input filter, working back from the attenuation you need in each mode. The two modes are designed separately because the same three components look completely different to them — and the differential inductance is not the choke’s inductance, it is its leakage. Every attenuation figure here is computed with the components’ parasitics in and in the impedances you state.
X cap, Y caps and the choke, by mode
a 250 kHz converter needing 40 dB of differential attenuation at 250 kHz and 60 dB of common-mode attenuation at 1 MHz, with a 1 mH choke whose leakage is 1% of that, 4.7 nF Y capacitors, and 100 ohm either side in differential mode
The same three parts, twice
fc = f / 10A/40 C = 1 / ((2πfc)²L) (second order, 40 dB per decade)
Lleak = k · LCM, k typically 0.005 to 0.02 — or measured
insertion loss = 20·log10 |(A·Zl + B + C·ZsZl + D·Zs) / (Zs + Zl)|
with the choke as (Rdc + jωL) ∥ 1/jωCw and the capacitor as Resr + j(ωLesl − 1/ωC)
- L leak
- the choke’s leakage inductance, and the only differential inductance most filters have. Common-mode flux from the two windings adds; differential flux cancels, and what survives the cancellation is the leakage
- 2 C Y
- the two Y capacitors in parallel, which is what common mode sees because both of them go from a line to the same chassis
- A
- the attenuation you need. The 40 in the exponent is the second-order roll-off, and it is an ASYMPTOTE: near the corner the real filter gives less, and the real figures below say how much
- Z s, Z l
- the impedances either side, per mode. They are not 50 ohm in your circuit and the answer moves with them, which is why both they and the 50 ohm / 50 ohm figure are printed
Worked example
a 250 kHz converter needing 40 dB of differential attenuation at 250 kHz and 60 dB of common-mode attenuation at 1 MHz, with a 1 mH choke whose leakage is 1% of that, 4.7 nF Y capacitors, and 100 ohm either side in differential mode
Differential mode sees the leakage, 1% of 1 mH = 10 µH. For 40 dB at 250 kHz a second-order filter needs a corner at 250 kHz / 10^(40/40) = 25 kHz, which with that inductance asks for 4.053 µF — 4.7 µF in E6
With 4.7 µF the differential corner is actually 23.22 kHz, comfortably 10.8 times below the switching frequency
What that filter really delivers at 250 kHz, solved as a network with the parasitics in and 100 ohm either side, is 51.60 dB — not the 40 the asymptote promised, because at 100 ohm the capacitor alone is already shunting hard. In 50 ohm / 50 ohm the same filter gives 45.88 dB, which is the figure a datasheet would quote
Common mode sees the full 1 mH against 2 x 4.7 nF = 9.4 nF, a corner at 51.91 kHz and 53.93 dB at 1 MHz — 6.07 dB short of the 60 asked for
The Y capacitor cannot grow, because its value is a leakage or chassis-current constraint rather than a filtering choice. The choke can: 2.695 mH with these Y capacitors would reach the requirement, against the 1 mH specified
Both capacitors have a ceiling of their own. The X capacitor self-resonates at 1.896 MHz on 1.5 nH of package and mounting inductance, and the choke at 1.592 MHz in common mode. Above those this is not an LC filter any more
What each mode sees, and what it does not
| Component | To differential mode | To common mode |
|---|---|---|
| Common-mode choke, both windings | Its leakage inductance only — flux from the two windings cancels | Its full rated inductance — flux from the two windings adds |
| X capacitor, line to return | Its full value | Invisible: both its ends move together, so no common-mode current flows through it |
| Two Y capacitors, each line to chassis | C_Y/2, the two in series across the line — usually thousands of times smaller than C_X and safely ignored | 2C_Y, the two in parallel |
| The chassis bond | Not in the differential loop at all | In series with both Y capacitors. A bond made with a wire puts tens of nanohenries there and undoes most of them |
| Capacitor ESL | Ends the X capacitor’s usefulness at its self-resonance | Ends the Y capacitors’ usefulness at theirs, which is usually higher because they are smaller |
| Choke winding capacitance | Self-resonates against the leakage, so quite high | Self-resonates against the full inductance, so much lower — often 1 to 10 MHz, inside the band you care about |
Measured parasitic inductance of multilayer ceramic capacitors
| Case size | Measured ESL | Note |
|---|---|---|
| 0603 | 0.87 nH | conventional termination |
| 0805 | 1.05 nH | conventional termination |
| 1206 | 1.20 nH | conventional termination |
| 1210 | 0.98 nH | wider than it is long, so the current loop is shorter |
| 0612 | 0.61 nH | reverse geometry: terminations on the long sides |
| 0508 | 0.60 nH | reverse geometry |
| Leaded film, 15 mm pitch | 15 to 20 nH | not measured here; the leads dominate and the self-resonance falls by more than an order of magnitude |
Two filters in one, and the inductance that is not on the datasheet
A single-stage input filter has three components and two jobs, and the two jobs see the three components completely differently. Get that separation right and the design is straightforward; miss it and you can build a filter that is excellent in one mode and does nothing in the other, which is the usual way a unit arrives at a test house having been carefully filtered against the wrong problem.
The choke. A common-mode choke is two windings on one core, wound so that common-mode current — the same direction in both — produces flux that adds. That is the inductance on the datasheet and it can be millihenries. Differential current goes up one winding and back the other, its flux cancels in the core, and what is left is the leakage: the flux that never linked both windings. That is typically half a per cent to a couple of per cent of the common-mode figure, so a 1 mH choke gives perhaps 10 microhenries differentially. In most real filters that leakage IS the differential inductor — there is no other one — and it is the number that decides the X capacitor. It is also the number least likely to be on the front page of the datasheet. The better way to get it is off the differential-mode impedance curve: read the impedance in its straight rising region and divide by 2 pi f.
The capacitors. The X capacitor goes across the line and is differential-mode only; in common mode both its ends move together and no current flows through it. The two Y capacitors go from each line to chassis, and in common mode they are in parallel, so common mode sees 2C_Y. Differentially they appear in series with each other across the line, giving C_Y/2 — which beside a microfarad-class X capacitor is four orders of magnitude down and is ignored here.
Why the Y capacitor is a constraint and not a choice. Every other component in this filter you can make bigger. The Y capacitor you cannot, because its value is set by something that has nothing to do with filtering: the current it is allowed to pass to chassis. On mains equipment that is a touch-current limit and it is a safety requirement. On a vehicle or a spacecraft it is a chassis-current and isolation question, and the single-point-ground architecture may decide it outright. Either way the number arrives from elsewhere, which is why this page asks for the Y capacitor you are allowed. When the common-mode requirement is not met with that value, the variable you have left is the choke, and the page prints the inductance that would do it.
Where the common-mode requirement comes from. Not from here, and not from any schematic. Common-mode noise is driven by dV/dt on switching nodes acting through parasitic capacitance to chassis — the switch tab against its insulator, the transformer’s interwinding capacitance, the harness passing the magnetics — and all of that is layout and mechanical design. It has to be measured with a current probe round the whole harness, or estimated with a stated method from a real layout. This page takes the required attenuation as your input and says so rather than inventing a model for it.
Once you have components, the insertion-loss page takes them and draws what they really do against frequency in both impedance environments — it is the same network solved by the same expressions, so the two pages agree by construction. The attenuation requirement itself comes from the margin page or, before hardware, from the converter estimator. This page says nothing about whether the differential stage will make the converter unstable, which is a separate and serious question: that is the EMI input filter and Middlebrook page, and an undamped LC in front of a regulating converter can oscillate however good its attenuation is.
Frequently asked questions
What inductance does differential mode actually see?
The common-mode choke’s LEAKAGE inductance, not its rated inductance. The two windings are phased so that common-mode flux adds and differential flux cancels, so differentially only the leakage survives — typically 0.5% to 2% of the common-mode figure, though it varies enormously with how the choke is wound. In most single-stage filters that leakage is the only differential inductor present, which makes it the number the X capacitor is sized against.
Why do the two Y capacitors add for common mode?
Because both of them run from a line to the same chassis, and in common mode both lines are at the same potential. Two capacitors from the same node to the same node are in parallel, so common mode sees 2C_Y. Differentially they are in series with each other across the line and give C_Y/2, which beside a microfarad X capacitor is negligible.
Why can I not just use a bigger Y capacitor?
Because its value is not a filtering decision. It is bounded by the current it is allowed to pass to chassis: a touch-current safety limit on mains equipment, a chassis-current and isolation requirement on a vehicle or spacecraft, and sometimes a single-point-ground architecture that forbids the connection altogether. When the common-mode requirement is not met with the value you are allowed, the choke is the variable you have, and this page prints the inductance that would reach it.
Why is the attenuation this page computes different from the one I asked for?
Because the sizing step and the answer are two different calculations. The sizing uses f_c = f / 10^(A/40), which is the second-order asymptote and assumes the filter is deep into its roll-off with impedances that do not matter. The figure reported is the real network solved with the parasitics in, in the impedances you entered. They agree only where the asymptote is valid, and the gap between them is the most useful number on the page.
Does it matter which end the capacitor is at?
A great deal. A shunt capacitor facing a low-impedance source barely does anything, because the source is already holding that node stiff; a series inductor facing the same source does a lot. The rule is to face the inductor toward the low impedance and the capacitor toward the high one. Turning a two-port round is exactly the same as exchanging its source and load impedances, which is what the selector on this page does.
Should I check the filter for stability as well?
Yes, and separately. A regulating converter draws constant power, so its incremental input impedance is negative, and an undamped LC in front of it can make a working converter oscillate at the filter’s resonance no matter how good its attenuation is. That is the Middlebrook criterion and the damping branch that satisfies it, and it has its own page on this site. Nothing here tests it.
Related calculators
References
- Würth Elektronik, 1-Phase Line Filter Design, application note ANP015, 2024-06-03. Gives the single-stage sizing step f_CO = f_sw / 10^(A/40 dB) and the fourth-order form with 80 in place of 40; states the common-mode equivalent capacitance as C_YG = 2 C_Y; and takes the differential inductance from the choke’s leakage, read off the datasheet’s differential-mode impedance curve in its inductive region as L_DM = X_L / 2 pi f rather than assumed as a percentage. Verified from the published application note.
- Ott HW. Electromagnetic Compatibility Engineering. Wiley, 2009. Chapter 13, 13.3 Power-Line Filters: 13.3.1 Common-Mode Filtering, 13.3.2 Differential-Mode Filtering, 13.3.3 Leakage Inductance, 13.3.4 Filter Mounting and 13.3.6 High-Frequency Noise — the separation of the two modes, the leakage inductance as the differential inductor, and why the mounting decides what the Y capacitors are worth. Section titles checked against the author’s own published contents listing.
- KYOCERA AVX, Parasitic Inductance of Multilayer Ceramic Capacitors, Cain J. Measured ESL by case size on 0.1 uF samples: 0603 at 870 pH, 0805 at 1050 pH, 1206 at 1200 pH, 1210 at 980 pH, and reverse-geometry 0612 and 0508 at 610 and 600 pH. The ESL defaults on this page come from these figures plus an allowance for the mounting loop.
- CISPR 17 Edition 2.0, 2011-06, Methods of measurement of the suppression characteristics of passive EMC filtering devices. The source of the 50 ohm / 50 ohm convention this page prints alongside the reader’s own impedances, and of the asymmetrical, symmetrical and unsymmetrical test circuits that correspond to common mode, differential mode and one line at a time.
- MIL-STD-461G, Requirements for the Control of Electromagnetic Interference Characteristics of Subsystems and Equipment, 11 December 2015, superseding MIL-STD-461F. A work of the US Government, distributed without charge. Paragraph 5.5.1 makes CE102 applicable from 10 kHz to 10 MHz on all power leads including returns that take power from a source outside the equipment, and Figure CE102-1 gives the limit: 94 dBµV at 10 kHz, falling at 20 dB per decade to 60 dBµV at 500 kHz and flat to 10 MHz. That band is what this page’s chart covers. Read from the standard itself; the curve is implemented on the margin page.
- Middlebrook RD. Input Filter Considerations in Design and Application of Switching Regulators. IEEE Industry Applications Society Annual Meeting, 1976, pp. 366-382. Not used for anything on this page, and cited because it is the question this page does not answer: a filter that meets its attenuation requirement can still destabilise the converter behind it.