EMI in Modern Power Electronics: Mechanisms, Modeling, Measurement, Mitigation, and EMC Compliance
Radiated Electromagnetic Interference in Modern Power Electronics: Sources, Measurement, Modeling, Limits, and Mitigation
The rapid transition from conventional silicon switching devices to silicon-carbide (SiC) MOSFETs and gallium-nitride (GaN) transistors has changed the electromagnetic behavior of power converters. Faster switching improves efficiency, reduces magnetic-component size, and enables higher power density, but it also increases the spectral content of switching waveforms and makes radiated electromagnetic interference (EMI) more difficult to control.
Radiated EMI is especially important in electric-vehicle power electronics, onboard chargers, traction inverters, industrial motor drives, renewable-energy converters, aircraft power systems, medical power supplies, and compact high-frequency adapters. In these products, the dominant radiator is often not the semiconductor package itself. High-dV/dt nodes and high-dI/dt loops excite heat sinks, transformer structures, PCB planes, enclosures, and attached cables, which then behave as unintended antennas.
This engineering guide focuses exclusively on radiated emissions in power-electronic systems. It explains near-field and far-field mechanisms, the influence of SiC and GaN switching behavior, standardized measurement conditions, practical emission limits, modeling methods, common failure mechanisms, and mitigation techniques. The technical foundation is the 2024 IEEE review paper A Review of Radiated EMI Research in Power Electronics Systems, reorganized here into a practical format for design and compliance engineers.
Core design principle: Radiated EMI is a system-level result. Switching waveforms create the excitation, parasitic impedances determine the propagation path, and cables, heat sinks, magnetic components, PCB structures, and enclosures determine how efficiently that noise is radiated.
1. Why Radiated EMI Has Become a Critical Power-Electronics Problem
Traditional power converters based on IGBTs and slower silicon MOSFETs often switched at frequencies from a few kilohertz to several tens of kilohertz. Their relatively slow voltage and current transitions limited the highest significant harmonics. Modern SiC and GaN converters can switch substantially faster, with transition times of only a few nanoseconds. The converter switching frequency may remain below 1 MHz, yet edge-related energy and parasitic ringing can extend beyond 100 MHz.
The radiated-emission spectrum is therefore governed by two different mechanisms. Low-frequency emissions are influenced mainly by switching frequency, waveform amplitude, and converter operating state. High-frequency emissions are influenced more strongly by rise time, fall time, package and PCB inductance, output capacitance, transformer leakage inductance, cable geometry, and ringing frequency.
| Technology | Typical converter use | Switching behavior | Radiated-EMI implication |
|---|---|---|---|
| Silicon IGBT | Medium- and high-power drives, traction, industrial conversion | Relatively slow transitions; switching frequency often in the low-kilohertz to tens-of-kilohertz range | Lower high-frequency spectral content, but large current loops and cable currents can still dominate radiation |
| Silicon MOSFET | Low- and medium-voltage converters and power supplies | Faster than IGBTs; moderate junction capacitance and switching speed | Radiated EMI is often driven by trapezoidal switching edges and parasitic ringing |
| SiC MOSFET | EV inverters, onboard chargers, industrial drives, high-voltage converters | High dV/dt and dI/dt; reduced reverse recovery; small capacitances | Stronger common-mode excitation and higher-frequency ringing; layout and heat-sink coupling become critical |
| GaN HEMT | High-frequency adapters, data-center power, aerospace, compact converters | Extremely fast edges and very small parasitics | Ringing and radiated spikes may move to hundreds of megahertz; gate-loop and PCB inductance become highly influential |
2. Near Field, Far Field, and the 30 MHz Transition
Radiated EMI measurements in power electronics cannot be interpreted correctly without distinguishing near-field and far-field behavior. A conventional transition boundary is approximately:
At 30 MHz, the free-space wavelength is approximately 10 m, and λ/2π is approximately 1.59 m. A monopole antenna placed 1 m from the equipment is therefore still in the near-field region at 30 MHz and below. At much higher frequencies, the same 1 m or 3 m distance increasingly approximates a far-field measurement.
Near-Field Radiation
In the near field, electric and magnetic fields are not necessarily linked by the free-space impedance of approximately 377 Ω. High-current, low-voltage structures behave primarily as magnetic-field sources, while high-voltage, low-current structures behave primarily as electric-field sources.
Magnetic-field sources
Typical examples include switching-current loops, transformer and inductor windings, magnetic-core air gaps, resonant inductors, and discontinuous-conduction current paths. Their radiation is strongly influenced by current amplitude, dI/dt, number of turns, loop area, and air-gap geometry.
Electric-field sources
Typical examples include switch nodes, heat sinks connected capacitively to switching devices, PCB traces carrying pulsating voltage, transformer structures, inductors with common-mode voltage, and input or output cables. Their field strength is influenced by voltage amplitude, dV/dt, parasitic capacitance, conductor height, and distance from the reference plane.
Far-Field Radiation
Above 30 MHz, attached power and signal cables frequently become the dominant radiating structure. Input and output cables can behave as an unintended dipole antenna; a single cable above a reference plane can behave as a monopole. The converter creates an excitation voltage, while the cable geometry and terminating impedances determine the resulting common-mode current and radiation efficiency.
Unless differential-mode current is extremely high and the loop area is unusually large, far-field radiated EMI above 30 MHz is often dominated by common-mode current flowing on external cables.
3. Physical Sources of Radiated EMI
3.1 High-dI/dt Current Loops
The magnetic dipole moment of a current loop increases with current, number of turns, and loop area:
This relationship explains why reducing loop area is one of the most effective radiated-EMI design actions. The relevant loop is not necessarily obvious from the schematic. It is the actual high-frequency current path formed by semiconductor packages, capacitors, vias, planes, busbars, and return conductors.
3.2 Magnetic-Core Air Gaps
An inductor or transformer air gap can behave as a localized magnetic radiator. The magnetic moment associated with the air gap is related to magnetomotive force and gap cross-sectional area. Fringing flux can couple into nearby PCB traces, enclosures, cables, and receiving antennas. Gap placement and orientation therefore matter, especially in compact high-current converters.
3.3 High-dV/dt Nodes and Electric Dipoles
A conductor with common-mode voltage above a reference ground carries time-varying charge. The electric dipole moment is:
The distance d between the energized conductor and its image in the reference plane increases the dipole moment. This is why a high-dV/dt heat sink or inductor located far above the chassis can produce strong electric-field emissions even when its physical area is modest.
3.4 Cables as Antennas
At frequencies below 30 MHz, a cable may be electrically short, yet its time-varying voltage can still produce a measurable near electric field through displacement current. Above 30 MHz, the same cable can operate as an efficient antenna if common-mode excitation is present.
A useful cable-antenna representation is:
Here, Rr represents radiated power, RL represents cable loss, and XA represents reactive near-field energy. The converter can be represented by a noise source and source impedance that drive this antenna impedance.
Important distinction: Below 30 MHz, time-varying cable voltage and displacement current can dominate a 1 m monopole measurement. Above 30 MHz, common-mode current on the same cable more often controls far-field radiation.
4. How Switching Waveforms Create the Radiated Spectrum
A practical switching waveform can be separated into three components: the intended trapezoidal waveform, the zero-voltage-switching transition component, and high-frequency ringing. Each component contributes differently to radiated EMI.
| Waveform component | Physical origin | Primary spectral effect | Engineering implication |
|---|---|---|---|
| Trapezoidal waveform | Normal converter switching between voltage states | Broad harmonic spectrum determined by amplitude, switching frequency, rise time, and fall time | Usually the major broadband radiated-EMI contributor |
| ZVS-related waveform | Voltage developed by reverse current during soft switching | Normally a smaller contribution for conventional MOSFETs | May become more relevant in GaN systems because reverse-voltage behavior differs |
| Ringing waveform | Resonance of parasitic inductance and capacitance | Narrow or clustered high-frequency peaks | Frequently produces isolated failures around a resonance, including frequencies above 100 MHz |
The roll-off corner frequency associated with the switching edge is approximated by:
| Shortest rise/fall time | Approximate corner frequency | Interpretation |
|---|---|---|
| 100 ns | 3.2 MHz | Edge-related spectral roll-off begins in the low-megahertz range |
| 20 ns | 15.9 MHz | Significant energy approaches the conventional 30 MHz radiated-emission boundary |
| 10 ns | 31.8 MHz | Broadband edge energy directly enters the standard radiated-emission range |
| 5 ns | 63.7 MHz | Strong potential for cable excitation and VHF radiation |
| 2 ns | 159 MHz | Typical of very fast WBG switching; layout and package parasitics become dominant |
| 1 ns | 318 MHz | High risk of radiated spikes in the VHF/UHF range |
Calculated from the corner-frequency relationship presented in the review paper. These values indicate spectral behavior; they are not compliance limits.
4.1 Ringing Frequency
The resonance created by parasitic inductance and device output capacitance can be approximated by:
PCB and package inductance can range from several nanohenries to tens of nanohenries. Because WBG devices normally have lower output capacitance than comparable silicon devices, their ringing frequency can move into the hundreds of megahertz.
| Parasitic inductance | Output/equivalent capacitance | Approximate resonance | Likely EMC effect |
|---|---|---|---|
| 20 nH | 500 pF | 50 MHz | Radiated peak in the lower VHF range |
| 10 nH | 200 pF | 113 MHz | Potential failure near FM/VHF bands |
| 5 nH | 100 pF | 225 MHz | Typical high-frequency ringing concern in compact WBG layouts |
| 2 nH | 50 pF | 503 MHz | UHF radiation sensitive to package, vias, and enclosure geometry |
5. Silicon, SiC, GaN, and Diode Reverse Recovery
Reverse recovery in a silicon p-n diode produces a large current transient before the diode regains reverse-blocking capability. The associated dI/dt and overshoot excite parasitic inductance and increase EMI. Silicon Schottky diodes reduce reverse recovery but are limited in high-voltage applications. SiC Schottky diodes provide high breakdown voltage with very low reverse-recovery charge and can reduce the diode-related portion of EMI.
However, replacing a silicon diode with a SiC Schottky diode does not guarantee low system-level radiation. The switching transistor, transformer, PCB layout, ground impedance, common-mode capacitance, and cable antenna may remain dominant.
| Diode type | Reverse-recovery behavior | Typical EMI implication | Limitation |
|---|---|---|---|
| Silicon p-n diode | Pronounced reverse-recovery current and longer recovery time | Higher current overshoot and stronger excitation of parasitic inductance | Higher switching loss and EMI at fast commutation |
| Silicon Schottky diode | Very low reverse recovery | Lower recovery-related EMI | Breakdown-voltage limitations restrict many high-power uses |
| SiC Schottky diode | Very low reverse recovery at high voltage | Lowest diode-related current spectrum in the reviewed comparison up to 100 MHz | Overall converter EMI may still be dominated by transistor switching and parasitic paths |
6. Standards, Frequency Ranges, Antennas, Distances, and Detector Settings
Radiated-emission requirements depend on product category and installation environment. The IEEE review emphasizes CISPR 25 for automotive electronics, CISPR 32/EN 55032 for multimedia equipment, RTCA DO-160 for aviation, and MIL-STD-461G for military equipment.
| Standard | Typical application | Relevant radiated-emission method | Important setup notes |
|---|---|---|---|
| CISPR 25 | Automotive components and modules; protection of onboard receivers | Low-frequency electric-field and higher-frequency radiated-emission measurements | The review describes 1 m measurement distance; monopole below 30 MHz and hybrid biconical/log-periodic antenna above 30 MHz |
| CISPR 32 / EN 55032 | Multimedia and digital equipment | Radiated emissions above 30 MHz, commonly evaluated at 3 m or 10 m depending on setup and limit set | The reviewed example uses a 3 m semi-anechoic chamber and hybrid biconical/log-periodic antenna |
| RTCA DO-160 | Airborne equipment | Radiated RF emissions over category-dependent frequency ranges | Installation category and aircraft environment determine the applicable limits and setup |
| MIL-STD-461G RE102 | Military subsystems and equipment | Electric-field radiated emissions from 10 kHz to 18 GHz, subject to platform applicability | Measurement distance and limit depend on the procedure and frequency range |
| FCC Part 15 / ANSI C63.4 | Unintentional radiators marketed in the United States | Radiated emissions typically above 30 MHz | Class A/B limits and measurement distance depend on equipment category and authorization procedure |
| ICES-003 / CISPR 32-based methods | Canadian digital apparatus and IT/multimedia equipment | Radiated emissions above 30 MHz | Canadian compliance remains separate from FCC and European market access |
| IEC 61800-3 | Adjustable-speed power-drive systems | Radiated emissions according to drive category and environment | Installation, cable length, grounding, and motor connection are integral to the result |
| Frequency range | Typical antenna in reviewed setups | Distance | Resolution bandwidth | Field interpretation |
|---|---|---|---|---|
| 150 kHz–30 MHz | Monopole antenna | 1 m | 9 kHz | Predominantly near electric field at the specified geometry |
| Above 30 MHz for CISPR 25-type setup | Hybrid biconical/log-periodic antenna | 1 m | 120 kHz | Transitions toward far-field behavior as frequency increases |
| Above 30 MHz for CISPR 32/EN 55032 example | Hybrid biconical/log-periodic antenna | 3 m | 120 kHz in the reviewed discussion | Approximate far-field measurement at sufficiently high frequency |
6.1 Selected Numerical Limits Discussed in the Review
The paper’s comparison plot provides representative CISPR 25 Class 5 and CISPR 32/EN 55032 Class B levels. The following values are a structured transcription of the plotted levels and should be checked against the exact edition and product configuration before use in a formal test plan.
| Frequency region | Peak level shown | Average level shown | Units |
|---|---|---|---|
| Approximately 0.15–0.3 MHz | 46 | 26 | dBµV/m |
| Approximately 0.53–1.8 MHz | 40 | 20 | dBµV/m |
| Selected narrow bands near 5.9–6.2 MHz and 26–28 MHz | 40 | 20 | dBµV/m |
| Standard / detector | Frequency region | Representative plotted level | Units |
|---|---|---|---|
| CISPR 32 / EN 55032 Class B quasi-peak | 30–230 MHz | 40 | dBµV/m at 3 m |
| CISPR 32 / EN 55032 Class B quasi-peak | 230–1000 MHz | 47 | dBµV/m at 3 m |
| CISPR 25 Class 5 peak | Selected bands from 30 MHz to 1 GHz | Approximately 28–46 depending on band | dBµV/m |
| CISPR 25 Class 5 average | Selected bands from 30 MHz to 1 GHz | Approximately 15–36 depending on band | dBµV/m |
Compliance caution: CISPR 25 limits are band-dependent and may depend on component class, antenna type, detector, and standard edition. Always use the purchased/current standard and customer test plan for certification decisions.
7. Why FFT Results Can Disagree With the EMI Receiver
A common modeling error is to compare an FFT directly with a receiver measurement without reproducing resolution bandwidth and detector behavior. The paper reports that the discrepancy can reach 35 dB under certain conditions.
A spectrum-analyzer or EMI-receiver model that includes RBW behavior becomes necessary when:
- The switching frequency is below twice the resolution bandwidth; or
- The modulation frequency applied to duty cycle, switching frequency, or phase is below twice the resolution bandwidth.
With a 120 kHz RBW, twice the RBW is 240 kHz. A converter switching or modulating below this value can therefore produce a receiver reading that differs substantially from the amplitude predicted by a simple FFT bin.
Modeling rule: Match the receiver—not just the circuit. A valid compliance prediction must reproduce RBW, detector, dwell time, frequency stepping, modulation, operating mode, and cable configuration.
8. Radiated-EMI Modeling and Prediction Methods
| Method | Best suited to | Main inputs | Strength | Important limitation |
|---|---|---|---|---|
| Near-field scanning with equivalent dipoles | Converters without long external cables | Measured E/H fields above the PCB | Identifies localized sources and predicts far field from equivalent source sets | Scan resolution, probe loading, and equivalent-source reconstruction affect accuracy |
| Method of Moments from near-field data | PCB-level radiation | Near-field scan and conductor geometry | Reconstructs equivalent sources and predicts distant fields | Requires a suitable electromagnetic representation of the structure |
| Radiation Transfer Function (RTF) | Converters with repeatable cable and chamber geometry | Measured transfer relationship between source/current and field | Efficient for design iteration once characterized | Valid only when geometry and boundary conditions remain representative |
| Hertzian dipole segmentation | Common-mode current distributed along cables | Measured or predicted current distribution | Flexible and physically intuitive for cable radiation | Can be inaccurate below 30 MHz when displacement current dominates |
| Finite-element or full-wave simulation | Complex converter, cable, enclosure, and ground structures | 3-D geometry and material properties | Can predict near and far fields directly | Computational cost and uncertain high-frequency parasitics |
| Equivalent radiation circuit model | Converter plus cable antenna | Noise sources, source impedance, component parasitics, cable impedance | Links circuit design variables directly to radiated EMI | Requires accurate parasitic extraction and validated antenna representation |
| Transmission-line and field theory | Motor cables and long power interconnects | Terminal voltages/currents and cable parameters | Predicts voltage/current distribution and field generation | Common-mode termination and installation geometry must be known |
| S-parameter extraction | Linearized transfer path from source to antenna/field | Vector-network-analyzer measurements | Captures complex high-frequency coupling paths | Operating-point dependence and nonlinear switching source must be treated separately |
| Deep learning | Repeated design families with large training datasets | Measured or simulated training data | Fast inference after training | Limited interpretability and uncertain extrapolation outside training space |
9. Isolated and Nonisolated Converters Radiate Differently
9.1 Nonisolated Converters
In buck, boost, and buck–boost converters, the voltage drop generated by discontinuous switching current across PCB ground impedance can become an excitation voltage between the input and output cable systems. This voltage drives common-mode cable current and radiation.
Capacitive coupling from the switching node to input or output cables can also bypass the intended EMI filter. A converter may therefore continue to radiate even after a larger common-mode choke is installed if the switch-node-to-cable capacitance remains uncontrolled.
9.2 Isolated Converters
Flyback, forward, LLC, dual-active-bridge, push–pull, and other isolated converters introduce transformer-mediated common-mode conversion. Unbalanced transformer parasitic impedances transfer primary differential excitation into a voltage difference between primary and secondary grounds.
The excitation voltage can be influenced by transformer capacitance, Y-capacitor impedance, primary and secondary ground impedance, switch-node coupling, and cable-antenna impedance. The EMI filter is not necessarily effective against a coupling path that bypasses it capacitively.
10. Low-Frequency Radiated EMI Below 30 MHz
Low-frequency radiated EMI deserves separate treatment because a 1 m monopole measurement is dominated by near electric field rather than conventional far-field radiation. The reviewed research identified several important effects:
- A power inductor can dominate the measured field below 30 MHz even when cables are present.
- Shielding the inductor produced a reported 13 dB reduction in one evaluated converter.
- CISPR 25 and DO-160-type setups can exhibit chamber resonance below 30 MHz, creating a measured spectral spike.
- In motor-drive systems, switching voltage on three-phase cables can drive displacement current directly toward the monopole antenna.
- At these low frequencies, conductive common-mode cable current may be less important than cable voltage and displacement current.
| Region | Typical dominant source | Primary coupling mechanism | Useful diagnostic |
|---|---|---|---|
| 150 kHz–30 MHz | Inductor common-mode voltage, cable voltage, magnetic air gap, switching voltage | Near electric or near magnetic field; displacement current | Monopole measurement, E-field probe, magnetic near-field probe, temporary inductor shielding |
| Above 30 MHz | Common-mode cable current, switch-node coupling, ground impedance, high-frequency ringing | Cable antenna and enclosure radiation | Current probe on cable, cable-position sensitivity, near-field scan, antenna maximization |
11. Radiated-EMI Mitigation Techniques
The reviewed techniques fall into three groups: reducing the noise source, attenuating the propagation path, and optimizing critical PCB traces and ground structures.
| Technique | Reported or highlighted result | Mechanism | Trade-off / caution |
|---|---|---|---|
| Increase or split ON/OFF gate resistance | Reduces switching speed and high-frequency spectral content | Lowers dV/dt, dI/dt, overshoot, and ringing | Increases switching loss and junction temperature |
| Active gate drive | Controls gate resistance by switching stage; reviewed example includes a 6.7 GHz GaN gate driver architecture | Maintains fast initial transition while increasing damping around overshoot/ringing | Added driver complexity, sensing, timing, and validation effort |
| Aperiodic pulse-position and pulse-width modulation | Up to 10 dB suppression reported | Spreads discrete harmonics over a wider band | May affect control, acoustics, losses, and detector behavior |
| Chaotic/random PWM | Approximately 10 dB reduction over a broad frequency range in a reviewed boost-converter study | Reduces concentrated spectral peaks | Average energy is redistributed rather than eliminated |
| Interleaved SiC inverter modulation correction | 9 dB EMI reduction reported | Eliminates pulse-width mismatch and improves cancellation | Requires accurate timing and current sharing |
| Spectrum steering | 44% EMI-filter volume reduction reported compared with conventional spreading | Moves spectral energy into regions where the filter provides greater attenuation | Requires coordinated modulation and filter design |
| Inductor shielding below 30 MHz | 13 dB reduction reported | Reduces near electric-field radiation from the inductor | Shield current, loss, thermal impact, and saturation must be assessed |
| Common-mode choke | Reduces cable common-mode current | Adds series common-mode impedance between source and cable antenna | Parasitic capacitance can limit high-frequency attenuation |
| Y capacitor | Reduces excitation voltage across cable antenna | Provides a controlled shunt path for common-mode current | Leakage-current and safety limits apply |
| PCB ground-impedance reduction | Can directly reduce cable excitation voltage | Lowers voltage developed by switching current across shared ground inductance | Requires layout redesign rather than component substitution |
11.1 Reduce the Source
Slowing the switching transition is effective but should be applied selectively. Separate turn-on and turn-off resistors allow independent optimization. Active gate drivers can use different resistance or current levels during the Miller plateau, overshoot interval, and ringing interval, maintaining efficiency while damping the most EMI-sensitive part of the transition.
Snubbers can reduce ringing by dissipating resonant energy, but they must be placed physically close to the resonant loop. A theoretically correct RC value may be ineffective when several nanohenries of connection inductance separate it from the switching device.
11.2 Control the Propagation Path
A common-mode choke increases the impedance between the converter and the cable antenna. A Y capacitor reduces the common-mode excitation voltage by creating a deliberate return path. Their effectiveness depends on parasitic capacitance, self-resonance, safety requirements, and where they are connected relative to the noise source.
Ferrites can be effective when placed at a cable exit or around the complete cable bundle carrying common-mode current. Placing a ferrite around only one conductor changes differential impedance and may not address the dominant radiation mechanism.
11.3 Optimize the PCB and Mechanical Structure
Critical PCB traces are those carrying discontinuous current or linking the switch-node electric field to the cable system. High-frequency ground impedance is particularly important. A trace or plane that appears equipotential at DC can develop several volts of RF noise when exposed to high dI/dt.
Mechanical design must be considered at the same time. Heat sinks, standoffs, shields, cable glands, painted seams, bonding straps, and enclosure apertures can all become part of the radiating structure.
12. Practical Troubleshooting During Chamber Testing
| Observed failure | Likely mechanism | Fast diagnostic action | Potential corrective direction |
|---|---|---|---|
| Broad increase across tens of megahertz | Switching edges too fast; common-mode excitation | Temporarily increase gate resistance | Gate optimization, common-mode path control, active drive |
| Narrow peak near a fixed frequency | Parasitic LC ringing | Probe switch node and current loop; add temporary damping | Reduce inductance, tune snubber, change device capacitance or layout |
| Emission changes strongly when a cable is moved | Cable-antenna radiation | Measure common-mode cable current | CM choke, 360° shield termination, lower excitation voltage, improved cable routing |
| Failure below 30 MHz that changes when an inductor is covered | Near electric field from inductor/common-mode voltage | Apply temporary grounded shield while monitoring thermal behavior | Inductor shielding, winding orientation, lower common-mode voltage, reduced height above plane |
| Filter change gives little improvement | Capacitive bypass around the filter | Identify switch-node-to-cable or switch-node-to-heat-sink capacitance | Electrostatic shield, geometry change, controlled return path |
| Different results from simulation and receiver | RBW/detector mismatch or inaccurate cable model | Reprocess waveform through receiver-equivalent model | Include RBW, detector, modulation, dwell, and full test geometry |
| Emission rises at heavy load | Higher switching current, altered transition speed, stronger ringing excitation | Compare waveforms and cable current at light and heavy load | Load-dependent gate optimization, lower loop inductance, damping |
13. Design Checklist for First-Pass Radiated-Emission Compliance
- Identify the applicable product standard, market, class, detector, frequency range, and cable configuration before schematic completion.
- Estimate edge corner frequency from the intended rise and fall times.
- Estimate resonant frequencies from likely PCB/package inductance and device capacitance.
- Minimize high-dI/dt loop area using tight placement and overlapping forward/return paths.
- Place DC-link and commutation capacitors directly at the switching stage.
- Use Kelvin-source or Kelvin-emitter connections where supported.
- Keep switch-node copper area only as large as needed for electrical and thermal performance.
- Control capacitance from switch nodes to heat sinks, shields, chassis, and cable conductors.
- Prevent switching current from flowing through shared PCB ground impedance.
- Design transformer interwinding capacitance and shielding intentionally in isolated converters.
- Locate common-mode filters at the cable interface and prevent parasitic bypass around them.
- Terminate cable shields with low-inductance 360° connections where the product architecture permits.
- Avoid long pigtail shield connections at VHF and UHF frequencies.
- Orient magnetic air gaps away from sensitive traces and cable exits.
- Evaluate low-frequency electric-field radiation from inductors and motor cables separately from high-frequency cable-current radiation.
- Perform near-field scanning with production-intent cables and operating modes.
- Measure common-mode current on complete cable bundles.
- Validate at minimum and maximum input voltage and representative load conditions.
- Use an EMI-receiver-equivalent post-processing model rather than a simple FFT when RBW conditions require it.
- Freeze cable length, routing, grounding, enclosure, hardware, firmware, and load configuration before formal testing.
14. Recommended Development Workflow
Phase 1 — Regulatory and Architecture Review
Define the applicable standard, test distance, frequency range, antenna, detector, limits, operating modes, and production-intent cable configuration. Identify all high-dV/dt nodes, high-dI/dt loops, galvanic isolation barriers, heat sinks, and external interfaces.
Phase 2 — Source and Path Modeling
Estimate switching-edge corner frequencies and LC resonance. Build a common-mode equivalent circuit that includes device capacitance, transformer capacitance, PCB ground impedance, Y capacitors, common-mode chokes, cable impedance, and chassis return paths.
Phase 3 — Prototype Diagnostics
Measure switch-node voltage, gate waveform, current commutation, and common-mode cable current. Use E-field and H-field probes to identify local hot spots. Compare multiple load and input-voltage conditions.
Phase 4 — Pre-Compliance Chamber Testing
Use the required antenna geometry, cable layout, grounding, and detector settings. Record worst-case orientation, antenna polarization, cable configuration, and operating mode. Apply controlled modifications one at a time.
Phase 5 — Accredited Compliance Testing
Test the finalized hardware and firmware configuration with documented production-intent accessories. Maintain configuration control so the tested sample remains representative of the product placed on the market.
Conclusion
Radiated EMI in modern power electronics is created by fast switching but shaped by the complete electrical and mechanical system. Below 30 MHz, near electric fields from inductors and time-varying cable voltages can dominate. Above 30 MHz, common-mode cable current, switch-node coupling, PCB ground impedance, and parasitic ringing frequently control the measured result.
Wide-bandgap devices intensify these mechanisms because their short transition times and small capacitances extend the noise spectrum into the VHF and UHF ranges. The solution is not simply to slow every switching edge or add a larger filter. Effective control requires coordinated optimization of the semiconductor transition, commutation loop, transformer, ground structure, cable interface, common-mode return path, magnetic components, and enclosure.
When source, path, and antenna are analyzed separately, radiated-emission failures become much easier to diagnose. This approach also allows engineers to preserve the efficiency and power-density benefits of SiC and GaN while achieving reliable, repeatable, and standards-compliant EMC performance.
Radiated-emission testing and troubleshooting support: Stancer Testing-Lab provides accredited radiated emissions testing, EMC testing, automotive EMC evaluation, pre-compliance measurements, near-field troubleshooting, and engineering support for SiC, GaN, motor-drive, renewable-energy, industrial, aerospace, medical, and multimedia products.
References
- Z. Ma, S. Wang, Q. Huang, and Y. Yang, “A Review of Radiated EMI Research in Power Electronics Systems,” IEEE Journal of Emerging and Selected Topics in Power Electronics, vol. 12, no. 1, pp. 675–694, Feb. 2024, doi: 10.1109/JESTPE.2023.3335972.
- CISPR 25, Vehicles, Boats and Internal Combustion Engines — Radio Disturbance Characteristics — Limits and Methods of Measurement for the Protection of On-Board Receivers.
- CISPR 32, Electromagnetic Compatibility of Multimedia Equipment — Emission Requirements.
- EN 55032, Electromagnetic Compatibility of Multimedia Equipment — Emission Requirements.
- RTCA DO-160, Environmental Conditions and Test Procedures for Airborne Equipment.
- MIL-STD-461G, Requirements for the Control of Electromagnetic Interference Characteristics of Subsystems and Equipment.
- IEC 61800-3, Adjustable Speed Electrical Power Drive Systems — EMC Requirements and Specific Test Methods.
- ANSI C63.4, Methods of Measurement of Radio-Noise Emissions from Low-Voltage Electrical and Electronic Equipment.
- H. W. Ott, Electromagnetic Compatibility Engineering, Wiley, 2009.
- C. R. Paul, Introduction to Electromagnetic Compatibility, 2nd ed., Wiley, 2006.
