100nF

EMI input filter calculator: buck LC filter and damping

For 45 dB at 420 kHz with a 12 µH inductor, the filter capacitor must be at least 2.13 µF, damped by a capacitor of at least 38.5 µF with about 1.12 Ωof ESR. This calculator sizes a buck converter's differential-mode input filter by TI's AN-2162: the attenuation needed at the switching frequency, entered or estimated from the input current or a scope reading, the filter capacitor from Equations 4 and 5, and the damping branch. It then checks the capacitor you fit, and the filter's output impedance against the converter's input resistance.

0 dB−45.0 dBf_s420 Hz4.20 MHz20 dB−100 dB
Fig 1 — the fitted stage gives 52.3 dB at 420 kHz against 45.0 dB required; its corner is at 25.5 kHz.
|R_in| = V_in²/P_in, 38.4 Ωdampedundamped 6.66 Ω494 Hz444 kHz100 Ω10 mΩ
Fig 2 — the output impedance peaks at 6.66 Ω undamped and 987 mΩ damped, under the converter's 38.4 Ω input resistance: 15.2 dB of margin undamped.
Duty cycle D = V_OUT/V_IN
0.208
Required attenuation at f_s
45.0 dB
C_fa, Equation 4 · C_fb, Equation 5
1.37 µF · 2.13 µF
Filter capacitor C_f, at least
2.13 µF (C_fb: the attenuation sets it)
With 4.92 µF fitted: corner
25.5 kHz (f_s/10 is 42.0 kHz)
Attenuation at f_s, fitted stage
52.3 dB (+7.3 dB against the requirement)
Damping capacitor C_d, Equation 6
≥ 38.5 µF, electrolytic, rated 25 % above the highest source voltage
Damping resistance ESR_d, Equation 7
≈ 1.12 Ω
Output impedance peak at V_B
6.66 Ω undamped at 14.8 kHz; 987 mΩ damped
Converter input resistance |R_in| = V_in²/P_in
38.4 Ω (margin 15.2 dB undamped, 31.8 dB damped)

How this is calculated

Standard: TI SNVA489C, AN-2162 Simple Success With Conducted EMI From DC-DC Converters (rev. April 2013): §4, Equations 2–7, §5; TI SNVA755, Simple Success with Conducted EMI and Radiated EMI for LMR160X0 (June 2016): §2

∣Att∣dB=20log⁡(Vin-ripple-p2p1 μV)−Vmax|Att|_{dB} = 20 \log\left(\frac{V_{in\text{-}ripple\text{-}p2p}}{1\ \mu\mathrm{V}}\right) - V_{max}
Equation 2 (p. 5), Method 1: a peak-to-peak ripple measured at the switching frequency with a wide-bandwidth scope, against the limit V_max in dBµV.
∣Att∣dB=20log⁡(Iπ2fsCINsin⁡(πD)1 μV)−Vmax,D=VOUTVIN|Att|_{dB} = 20 \log\left(\frac{\dfrac{I}{\pi^2 f_s C_{IN}} \sin(\pi D)}{1\ \mu\mathrm{V}}\right) - V_{max}, \qquad D = \frac{V_{OUT}}{V_{IN}}
Equation 3 (p. 5), Method 2: the first harmonic of a square-wave input current, (2I/π)·sin(πD), across the impedance 1/(2π·f_s·C_IN). I is the DC inductor current, the output current for a buck. D as V_OUT/V_IN is the first-order buck duty cycle; with V_IN(min) it reproduces the C_fb values AN-2162 §5 prints.
Cfa=CINCINLf(2πfs/10)2−1C_{fa} = \frac{C_{IN}}{C_{IN} L_f \left(2\pi f_s / 10\right)^2 - 1}
Equation 4 (p. 6): "ensures that the resonance frequency of the EMI input filter is at least one decade below the switching frequency" — the resonance of L_f with C_f in series with C_IN. It has no positive solution when C_IN·L_f·(2πf_s/10)² ≤ 1.
Cfb=1Lf(10∣Att∣dB/402πfs)2,Cf≥max⁡(Cfa,Cfb)C_{fb} = \frac{1}{L_f}\left(\frac{10^{|Att|_{dB}/40}}{2\pi f_s}\right)^2, \qquad C_f \ge \max\left(C_{fa}, C_{fb}\right)
Equation 5 (p. 6), "derived from an approximation that ensures proper attenuation of the EMI filter". "Select the higher value of Cfa and Cfb because both conditions must be met."
Cd≥4 CIN,ESRd≈Lf/CINC_d \ge 4\, C_{IN}, \qquad ESR_d \approx \sqrt{L_f / C_{IN}}
Equations 6 and 7 (p. 6). "The dc voltage rating of the electrolytic should be at least 25% larger than the worst case maximum source voltage."
∣VAVB∣=∣11+(RDCR+jωLf)jωCf∣,Zout=(RDCR+jωLf)∥1jωCIN∥(ESRd+1jωCd)\left|\frac{V_A}{V_B}\right| = \left|\frac{1}{1 + \left(R_{DCR} + j\omega L_f\right) j\omega C_f}\right|, \qquad Z_{out} = \left(R_{DCR} + j\omega L_f\right) \parallel \frac{1}{j\omega C_{IN}} \parallel \left(ESR_d + \frac{1}{j\omega C_d}\right)
Derived from the circuit of AN-2162 Figure 4: the attenuation of the fitted stage, and the output impedance at V_B looking back toward a stiff source (C_f is then across the source and drops out). The damping term is present only for the damped curve. The peak is found numerically.
∣Rin∣=VIN2PIN,PIN=VOUT Iη|R_{in}| = \frac{V_{IN}^2}{P_{IN}}, \qquad P_{IN} = \frac{V_{OUT}\, I}{\eta}
Derived: a regulated converter draws constant power, so dI/dV = −P/V² at its input. The margin is 20·log(|R_in|/Z_out,peak). AN-2162 §7 lists Middlebrook and Erickson for the full stability treatment; no threshold is drawn here.

Assumptions

What it computes

A buck converter draws its input current in pulses, and the ripple those pulses make across the input capacitor is the conducted emission a compliance test measures. TI's AN-2162 (SNVA489) designs the differential-mode filter that stops it in four steps — "Calculate the required attenuation", "Select filter inductance Lf", "Calculate filter capacitance Cf", "Calculate damping capacitance Cd" — and SNVA755 runs the same procedure on a real board. The calculator does all four and then checks what AN-2162 only warns about: whether the filter's output impedance comes near the converter's input resistance.

The required attenuation is the height of the unfiltered fundamental above the limit at the switching frequency. Measured with a LISN, it is a number to enter. Before that, AN-2162 gives two estimates: Equation 2 converts a scope's peak-to-peak ripple into dBµV, and Equation 3 derives the fundamental from the input current itself, "extracting the first harmonic of current from the Fourier series of the input current waveform and multiplying that value by the impedance at the input port". The harmonics fall faster than the fundamental, so the fundamental is the design point.

The capacitor comes from two conditions, and AN-2162 is explicit: "Select the higher value of Cfa and Cfb because both conditions must be met." Equation 4 places the resonance of Lf with Cf and CIN "at least one decade below the switching frequency"; Equation 5 is "an approximation that ensures proper attenuation". Which one wins depends on the design, and the three published examples split: Equation 4 sets AN-2162's two, Equation 5 sets SNVA755's.

Worked example: SNVA755's LMR160X0 filter

The calculator's defaults, from SNVA755 §2: 24 V to 5 V at 3 A, switching at 420 kHz with 9.62 µF of input capacitance, and 45 dB required "in order to pass CISPR 22 Class B". The inductor is the 1234AS-H-120M, 12 µH and 0.19 Ω.

Required attenuation    SNVA755, CISPR 22 Class B             = 45 dB
C_fa  (4)   9.62 µF / (9.62 µF·12 µH·(2π·42 kHz)² − 1)       = 1.367 µF
C_fb  (5)   (1/12 µH)·(10^(45/40) / (2π·420 kHz))²           = 2.128 µF
C_f         the higher                                        = 2.128 µF
Fitted      4.7 µF ∥ 0.22 µF                                  = 4.92 µF
Corner      12 µH with 4.92 µF in series with 9.62 µF         = 25.5 kHz
At 420 kHz  |V_A/V_B| with 0.19 Ω DCR                         = 52.3 dB
C_d   (6)   4 × 9.62 µF                                       ≥ 38.5 µF
ESR_d (7)   √(12 µH / 9.62 µF)                                ≈ 1.12 Ω
Z_out peak  undamped · damped                                 = 6.66 Ω · 987 mΩ
|R_in|      24² / (5 V × 3 A)                                 = 38.4 Ω

SNVA755 prints Cfa = 1.37 µF and Cfb = 2.1 µF, which the equations reproduce, and fits 4.92 µF — more than twice the requirement, and 7.3 dB more attenuation than asked for. It also leaves the damping capacitor out: "in order to reduce the cost, no electrolytic is used". The impedance lines show what that costs: an undamped peak of 6.66 Ω set only by the inductor's DCR, 15 dB below the converter's 38.4 Ω input resistance at full load. With the branch AN-2162 prescribes it would be 987 mΩ.

The three published designs, reproduced

DesignAttenuationC_faC_fbRequiredFitted
AN-2162 Example 1, LMZ2360536.2 dB
printed 40 dB
4.93 µF
printed 4.9 µF
2.55 µF
printed 2.5 µF
4.93 µF4.7 µF
AN-2162 Example 2, LMZ1420242.6 dB
printed 40 dB
8.34 µF
printed 8.3 µF
6.40 µF
printed 6.4 µF
8.34 µF10 µF
SNVA755, LMR160X045.0 dB
printed 45 dB
1.37 µF
printed 1.37 µF
2.13 µF
printed 2.1 µF
2.13 µF4.7 ∥ 0.22 µF

Every printed capacitor value is reproduced, but AN-2162's two tables hide one step. They give the estimated noise as "≈ 80 dBµV" and the required attenuation as 40 dB, and Cfb at a flat 40 dB would be 3.96 µF and 4.74 µF — neither printed value. The printed Cfb of 2.5 µF and 6.4 µF come from Equation 3's unrounded attenuation, 36.2 dB and 42.6 dB, with the duty cycle taken as 3.3 V / 30 V. The table's 40 dB is the rounded target; the column beside it was computed from the equation. The calculator, in the input-current mode, follows the column.

The fitted values disagree with the rule in both directions. Example 1 fits 4.7 µF against a calculated 4.9 µF, a little under Cfa; Example 2 fits 10 µF against 8.3 µF. A standard value one step below the requirement moves the corner slightly above fs/10, which the attenuation line shows the cost of: 0.43 dB at 800 kHz for Example 1, against 5.2 dB of margin over Equation 3's requirement.

Filter capacitor by inductor, SNVA755's operating point

45 dB at 420 kHz with 9.62 µF of CIN. Below 1.49 µH no capacitor can satisfy Equation 4, because Lf with CIN alone already resonates above fs/10; the calculator says so rather than return a negative capacitance. The 12 µH row is the worked example.

L_fC_fa (4)C_fb (5)C_f, at leastESR_d (7)
2.2 µH20.30 µF11.61 µF20.30 µF478 mΩ
4.7 µH4.48 µF5.43 µF5.43 µF699 mΩ
6.8 µH2.71 µF3.76 µF3.76 µF841 mΩ
10 µH1.69 µF2.55 µF2.55 µF1.02 Ω
12 µH1.37 µF2.13 µF2.13 µF1.12 Ω
15 µH1.06 µF1.70 µF1.70 µF1.25 Ω
22 µH0.70 µF1.16 µF1.16 µF1.51 Ω

Read down the table and the trade AN-2162 describes is visible: a bigger inductor needs less capacitance on both counts, which is why its advice is to "Choose the highest value in compliance with amperage and physical size requirements". The damping resistance rises with it, as √(Lf/CIN).

Damping and the converter's input resistance

The resonance the filter is built around does not disappear because it sits a decade below fs. AN-2162 §4.4 is about its other effect: "The output impedance of the filter must be sufficiently small at point VB so that input filter does not significantly affect the loop gain of the SMPS", and damping "is needed when the output impedance is very high at the resonant frequency (that is, Q of filter formed by CIN and Lf is too high.)" The calculator computes that impedance — (DCR + jωLf) ∥ CIN, looking back from VBtoward a stiff source, so Cf drops out — with and without the branch of Equations 6 and 7 across it.

What it is compared with is derived rather than quoted. A regulated converter holds its output power constant, so its input current falls as its input voltage rises, and incrementally its input looks like a negative resistance of magnitude Vin²/Pin. AN-2162's references list Middlebrook's "Design Techniques for preventing Input Filter Oscillations in Switched-mode regulators" and Erickson's work on input-filter damping for the full treatment; the calculator reports the margin between the peak and that resistance in dB and does not draw a pass line of its own. Check it at the lowest input voltage and highest load, where Vin²/Pin is smallest.

Where the model stops

Further reading