Module 4.2 - Conducted Emissions
Design review checkpoints for controlling noise conducted back to the AC mains or DC power bus through the power input
Design review checkpoints for controlling noise conducted back to the AC mains or DC power bus through the power input
Switch-mode power supplies generate conducted emissions at the switching frequency and its harmonics. The fundamental switching frequency typically falls in the 50 kHz to 2 MHz range, with harmonics extending to 30 MHz and beyond. These harmonics propagate back through the power input and must be attenuated below regulatory limits.
| Frequency Range | CISPR 32 Class B QP (dBuV) | CISPR 32 Class B Avg (dBuV) | FCC Class B (dBuV) |
|---|---|---|---|
| 150 kHz - 500 kHz | 66 to 56 (linear decrease) | 56 to 46 (linear decrease) | 66 to 56 |
| 500 kHz - 5 MHz | 56 | 46 | 56 |
| 5 MHz - 30 MHz | 60 | 50 | 60 |
Note: CISPR 32 limits are measured at LISN output (50 ohm/50 uH). FCC Part 15 uses similar LISN setup.
500 kHz buck converter with designed EMI filter: Two-stage LC filter: Stage 1 (closest to converter): L1=10uH, C1=1uF X7R ceramic. Stage 2 (closest to input): L2=47uH, C2=4.7uF ceramic + 100nF. Provides 80 dB attenuation at 500 kHz (switching fundamental). Common-mode choke (10mH) between stages handles CM noise. Result: All harmonics 10+ dB below Class B limits.
500 kHz buck converter with only bulk input capacitor: Single 47 uF electrolytic at input. ESR = 0.2 ohm, ESL = 5 nH. At 500 kHz, impedance is dominated by ESL: Z = 2*pi*500e3*5e-9 = 15.7 mohm (but ESR dominates = 0.2 ohm). Minimal filtering. First harmonic at 500 kHz: 148 - 20*log10(attenuation) = still 120+ dBuV at LISN. Fails by 64 dB!
Filter stability interaction: Adding an input LC filter creates a resonance that can destabilize the converter. The filter output impedance must be less than the converter input impedance at all frequencies. Use Middlebrook's criterion: Z_filter_out < Z_converter_in. Add damping resistor in series with a capacitor across the filter inductor.
Electrolytic capacitor limitations: Above 100 kHz, electrolytic capacitors are essentially resistive (ESR-dominated) and provide no filtering. Always parallel with low-ESR ceramics for high-frequency filtering.
The input EMI filter must be designed to work correctly with the Line Impedance Stabilization Network (LISN) used during conducted emission testing. The LISN presents a defined source impedance (50 ohm || 50 uH) to the equipment under test, and the filter must provide adequate attenuation as seen from this source.
Two-stage DM filter for 500 kHz SMPS, 12V/5A input:
Stage 1 (converter side):
L1 = 10 uH (Wurth 744770110, I_sat = 6A, DCR = 15 mohm)
C1 = 2.2 uF X7R ceramic (TDK C3225X7R1H225K, 50V, ESR = 5 mohm)
f_c1 = 1/(2*pi*sqrt(10e-6 * 2.2e-6)) = 33.9 kHz
Attenuation at 500 kHz: 40*log10(500/33.9) = 46.4 dB
Stage 2 (input side):
L2 = 47 uH (Wurth 744770147, I_sat = 5.5A, DCR = 35 mohm)
C2 = 4.7 uF X7R ceramic (TDK C3225X7R1E475K, 25V, ESR = 3 mohm)
f_c2 = 1/(2*pi*sqrt(47e-6 * 4.7e-6)) = 10.7 kHz
Attenuation at 500 kHz: 40*log10(500/10.7) = 66.8 dB
Total attenuation at 500 kHz: 46.4 + 66.8 = 113.2 dB
Noise at LISN: 148 - 113.2 = 34.8 dBuV (limit: 56 dBuV)
Margin: 21.2 dB -- PASSES with excellent margin
Damping: 1 ohm resistor in series with 10 uF across L2 to prevent resonance
Properly damped two-stage filter: Two LC stages with staggered corner frequencies (10 kHz and 30 kHz) to avoid resonance stacking. Each inductor has a parallel RC snubber (1 ohm + 10 uF) to damp the LC resonance Q. Input impedance remains low and flat from DC to 100 kHz, ensuring converter stability. Total insertion loss measured: 95 dB at 500 kHz, 110 dB at 1 MHz.
Undamped single-stage LC with high-Q resonance: Single 100 uH + 10 uF LC filter with no damping. Resonance at 5 kHz with Q = 50 creates 34 dB gain peak. Converter becomes unstable (oscillates at 5 kHz). Also, above resonance the inductor self-resonant frequency (SRF = 2 MHz) causes filter to become ineffective above 2 MHz, missing harmonics at 2.5 and 3 MHz.
LISN Measurement Setup:
Common-mode (CM) conducted emissions are typically the dominant emission mechanism above 1 MHz. CM currents flow equally on both power lines (L and N) and return through the ground/earth conductor or parasitic capacitance to ground. Identifying and controlling these paths is essential for conducted emission compliance.
Flyback converter with controlled CM path: MOSFET heatsink isolated with 1.5 mm thick alumina washer (C = 3 pF) instead of mica (C = 15 pF). Faraday shield between transformer primary and secondary windings reduces interwinding capacitance from 50 pF to 5 pF. Y-capacitors (2x 2.2 nF, Class Y2) placed symmetrically from L and N to chassis. CM choke (10 mH) at input. Result: CM emissions 12 dB below Class B limit.
Flyback with no CM attention: MOSFET mounted directly to metal chassis (C = 100 pF). No Faraday shield in transformer. Single 4.7 nF Y-cap from one line only (asymmetric). No CM choke. dV/dt at drain = 10 V/ns. CM current: 100pF * 10e9 = 1A peak! Even at 50% duty: CM harmonic at 100 kHz = 400 mA peak = 152 dBuV at LISN. Fails by 96 dB at fundamental!
Y-capacitor safety limits: Y-capacitors from line to earth must be limited for safety (touch current): max 4.7 nF for Class Y2 (medical: 470 pF for Class Y1). This limits CM filtering capability - cannot simply add more capacitance.
CM and DM interaction: Y-capacitor imbalance converts CM noise to DM noise and vice versa. Always use matched pairs of Y-caps (from same lot) placed symmetrically.
Differential-mode (DM) noise flows between the power lines (L to N) and is caused by the pulsating input current of switching converters. DM noise typically dominates below 500 kHz to 2 MHz, while CM noise dominates at higher frequencies. The DM filter uses X-class capacitors and series inductors.
X-capacitor classes (per IEC 60384-14):
Class X1: 4 kV peak pulse rating, for C_across-the-line after mains filter
Class X2: 2.5 kV peak pulse, for general across-the-line use (most common)
Class X3: 1.2 kV peak pulse, for general applications
Typical X2 capacitor values and SRF:
100 nF film: SRF ~ 15-20 MHz (useful to 10 MHz)
470 nF film: SRF ~ 5-8 MHz
1 uF film: SRF ~ 3-5 MHz
2.2 uF film: SRF ~ 2-3 MHz
For ceramic X-capacitors (higher SRF but lower capacitance):
100 nF C0G/NP0: SRF ~ 50-100 MHz
1 uF X7R: SRF ~ 10-20 MHz
Design rule: Place both film and ceramic X-caps in parallel:
Film (1 uF) for low-frequency DM filtering (150 kHz - 2 MHz)
Ceramic (100 nF) for high-frequency DM filtering (2 MHz - 30 MHz)
Boost PFC with proper DM filter: 65 kHz boost PFC converter, 400W, AC input. DM filter: Stage 1: 330 uH inductor (amorphous core, I_sat = 5A) + 2x 470 nF X2 film capacitors. Stage 2: leakage inductance of CM choke (50 uH) + 1 uF X2 film + 100 nF C0G ceramic. Corner frequency: 13 kHz. Provides 72 dB at 150 kHz (fundamental). Total system DM emissions: 42 dBuV at 150 kHz (limit: 66 dBuV QP). 24 dB margin.
Boost PFC with inadequate DM filter: Only the PFC inductor (150 uH) and a 100 nF X2 cap at input. Corner frequency: 41 kHz. Attenuation at 150 kHz: only 40*log10(150/41) = 22.6 dB. DM noise from discontinuous boost input current is 130 dBuV. Result: 130 - 22.6 = 107.4 dBuV at LISN. Limit: 66 dBuV. Fails by 41.4 dB!
The EMI filter corner frequency determines the frequency above which attenuation begins. Setting this correctly requires balancing between adequate attenuation at the lowest emission frequency (150 kHz for CISPR, or switching frequency) and practical component values/sizes. Too low a corner frequency requires impractically large inductors.
| Corner Freq | Inductor Size | Capacitor Size | Atten @ 150kHz | Atten @ 500kHz |
|---|---|---|---|---|
| 5 kHz | 1-10 mH (large) | 10-100 uF | 88 dB | 120 dB |
| 10 kHz | 0.5-2 mH | 2-10 uF | 70 dB | 102 dB |
| 20 kHz | 100-500 uH | 1-5 uF | 52 dB | 84 dB |
| 50 kHz | 50-200 uH | 0.1-1 uF | 28 dB | 60 dB |
Correctly sized Pi-filter for 200 kHz SMPS: Target: 80 dB at 200 kHz (switching fundamental). Pi-filter with f_c = 10 kHz: L = 1 mH (Wurth 744821001), C1 = C2 = 2.2 uF (X2 film). Attenuation at 200 kHz: 60*log10(200/10) = 78 dB. With 6 dB margin target met. Physical size: inductor 20x20x15mm, capacitors 18x7x12mm each. Total filter volume: 15 cm^3 -- acceptable for 200W supply.
Corner frequency too close to switching frequency: 200 kHz SMPS with f_c = 100 kHz. Attenuation at 200 kHz (first harmonic): only 60*log10(200/100) = 18 dB. Required: 80 dB. Insufficient by 62 dB. Designer would need to drop corner to 10 kHz, requiring 10x larger inductor -- should have been planned from the start to reserve board space.
Inductor saturation at rated current: EMI filter inductors must maintain inductance at the DC bias current. Iron powder cores lose 50-80% of inductance at rated current. Use inductance value at actual operating current, not the zero-bias value. Amorphous or nanocrystalline cores maintain inductance better but cost more.
Capacitor derating with DC bias (ceramics): X7R ceramic capacitors lose 50-80% of capacitance at rated voltage. A "4.7 uF" X7R at 80% of rated voltage may only provide 1.5 uF actual capacitance. Film capacitors do not have this issue.
A conducted emission pre-scan early in development verifies that the EMI filter design is effective before committing to final hardware. Pre-compliance measurement using a LISN and spectrum analyzer identifies problems while corrections are still inexpensive.
To separate CM and DM components, measure at both LISN ports:
V_L = voltage at Line LISN port
V_N = voltage at Neutral LISN port
V_CM = (V_L + V_N) / 2 (common-mode component)
V_DM = (V_L - V_N) / 2 (differential-mode component)
Practical method using two LISNs simultaneously:
1. Connect both LISNs to spectrum analyzer (use power splitter/combiner)
2. In-phase combination (0 degree): gives V_L + V_N = 2*V_CM
3. Anti-phase combination (180 degree via balun): gives V_L - V_N = 2*V_DM
Alternative: Commercial CM/DM separation networks (e.g., Schwarzbeck CMDM 8700)
If V_CM dominates: Improve CM choke, reduce parasitic capacitance, add Y-caps
If V_DM dominates: Improve DM filter (add inductance or X-capacitance)
Pre-scan at EVT stage identifies filter gap: Pre-compliance scan shows 3rd and 5th harmonics of 300 kHz switching (900 kHz and 1.5 MHz) are 4 dB and 2 dB above Class B QP limit respectively. CM/DM separation reveals DM dominance. Solution: Add 100 nF X2 capacitor at filter output (cost: $0.05, board space reserved). Re-scan shows 8 dB margin. Fix implemented before DVT build with zero schedule impact.
No pre-scan, first test at compliance lab: Product sent directly to accredited lab for conducted emission test. Fails at 5 frequencies, requiring board redesign to add filter components for which no space was allocated. Re-layout, re-fabrication, and re-test costs $35,000 and delays program by 6 weeks. A $2,000 pre-scan investment would have caught this at the prototype stage.
Spectrum Analyzer Settings for Conducted Pre-Scan: