An industrial EMI filter should address an identified port and coupling path. A power inductor, common-mode choke, DC-link reactor and motor-output filter have different duties; one set of inductance and current ranges cannot select them all. This guide focuses on a DC supply input, using a hypothetical 24 V control-cabinet rail as context. It does not specify a mains or motor-output filter, an insulation system or an approved product.
Map the source, path and affected circuit
Record the converter or switching source, the affected power or signal port, cable and chassis arrangement, operating states and repeatable symptom. Separate differential voltage between supply and return from common-mode current returning through other paths. A large spectral peak at one location does not identify the mode by itself.
TI's AN-2162, SNVA489C, revised April 2013, distinguishes noise paths and describes conducted-EMI measurement and filtering. Its examples support a design method; their measured results and historical standards discussion are not results or current compliance requirements for your equipment.
Build a comparison that can be repeated
| Record | Why it matters | What to avoid concluding |
|---|---|---|
| Port, cables, return and chassis connections | Defines the coupling and measurement path | Every cable symptom is common mode |
| Load, input voltage, switching mode and frequency | Determines the operating condition being compared | One light-load test represents the full range |
| Probe or receiver, bandwidth, detector and fixture | Makes spectra and waveform comparisons meaningful | A spectrum-analyzer peak is a formal pass result |
| Exact part, effective L, DCR and temperature | Links noise changes to electrical and thermal behavior | Lower noise guarantees acceptable temperature or stability |
Keep the baseline and the changed setup otherwise comparable. Select a common-mode choke for the identified common-mode path and a differential filter for the corresponding supply-return path. Leakage and parasitic coupling can make the modes interact. Retest both modes and the affected circuit function after a change.
Derive insertion loss for a defined circuit
The following is an independent single-frequency calculation, not a measured filter. Let an ideal Thevenin voltage source V_s drive a real source resistance R_s and real load R_L. Insert one series impedance Z_f between them and change nothing else. H_0 and H_f are load-voltage/source-voltage ratios before and after insertion.
At 1 MHz, assume Z_f = j60 Ω. This is a hypothetical impedance, not a part specification. For the two different resistive terminations:
H_0 = R_L / (R_s + R_L)
H_f = R_L / (R_s + R_L + Z_f)
IL = 20 log10 |H_0 / H_f|
= 20 log10 |(R_s + R_L + Z_f) / (R_s + R_L)|
R_s = R_L = 50 Ω: IL ≈ 1.335 dB
R_s = R_L = 5 Ω: IL ≈ 15.682 dB
The same series impedance gives different voltage insertion losses because the terminations differ. The derivation uses the complex magnitude; omitting the load or treating the impedance as a real number changes the problem. It is not a common-mode four-port model and does not predict installed conducted or radiated emissions. For a practical filter with shunt capacitors, losses and parasitics, use the actual network and stated measurement impedances.
Check input-filter resonance and converter interaction
An added LC section can reduce high-frequency disturbance while creating an impedance peak. In TI's input-filter design note, SNVA538, the “Undamped L-C Filter” and following input-impedance discussion explain why an undamped filter may worsen behavior near resonance and interact with the converter control loop.
For a separate ideal example, assume effective values under the evaluated conditions:
L_eff = 10 µH; C_eff = 4.7 µF
f_0 = 1 / (2π√(L_eff × C_eff)) ≈ 23.2 kHz
This estimates a resonance frequency, not stability or compliance. Evaluate the filter output impedance and converter input impedance over the relevant frequencies and operating states. Model or measure damping, parasitics, startup and load transients. A cutoff number alone cannot select a damping resistor or capacitor, and a larger inductor is not automatically a better filter.
Use the actual bias-dependent L and capacitor effective value in that assessment. Coilcraft's rating note distinguishes saturation and heating criteria; check the exact manufacturer's Isat, Irms and temperature conditions. A common-mode part's heating rating should not be relabeled as a differential inductor's saturation limit. Verify loss and temperature on the intended board.
Define the standards scope before final acceptance
The official public summaries of IEC 61000-6-2:2016 and IEC 61000-6-4:2018 describe generic industrial immunity and emissions respectively, to be used when the relevant dedicated product or product-family standard is unavailable. An industrial installation alone is therefore insufficient reason to assign every listed standard to a product.
Only these public scope summaries were checked here; the paid full texts were not reviewed. A qualified EMC specialist must determine the applicable product rules, editions, ports, test levels, arrangements and functional acceptance criteria. This guide provides no universal frequency ranges, severity levels or conformity decision.
Before release, test the actual assembly with its intended cables, enclosure, grounding and relevant operating states. Check electrical function, thermal behavior and input stability alongside emissions and immunity. A filter substitution is not automatically evidence of motor-bearing protection, improved power factor or longer equipment life.
For mode identification see the common-mode choke guide; the inductor current check supports preliminary waveform calculations.
References and revision
AN-2162 sections 4, 4.4 and 6 support filter planning, damping and LISN context. SNVA538 PDF pages 2–4 explain the ideal LC response and converter interaction; its body carries a 2010 National Semiconductor copyright. The rating note and IEC public summaries have the narrower roles stated above. No laboratory test, certification or customer acceptance is claimed.
Revised 2026-10-28: removed unsupported material and current ranges, blanket test bands and compliance assurances; replaced the unqualified insertion-loss formula with a circuit-specific derivation and added a resonance/stability check. Original URL and publication date remain recorded. AI editorial self-review only; independent native-speaker engineering review and product-specific EMC approval remain outstanding.
