1. Common Mode Currents
Definition
Common mode (CM) currents are equal-amplitude, equal-phase currents flowing in the same direction on both conductors of a pair (or on a cable and its reference ground). In a two-wire system, common mode current is defined as the average of the two conductor currents:
Where I1 and I2 are the currents on each conductor,
measured in the same direction.
Common mode currents are the dominant source of radiated EMI in most electronic systems. Even though CM currents are typically much smaller than differential mode (DM) currents (often 1000x smaller), they radiate far more efficiently because:
- CM currents flow on both conductors in the same direction, creating a large effective antenna (the entire cable length)
- DM currents flow in opposite directions and their fields tend to cancel
- Cable lengths are often comparable to a wavelength at troublesome frequencies, making them efficient radiators
At 200 MHz with a 1-meter cable, only 4.8 microamps of common mode current is needed to exceed FCC Class B radiated emission limits (at 3 meters). This is an astonishingly small current that is virtually impossible to measure directly on a PCB.
Why CM Currents Dominate EMI
Consider a 1-meter cable carrying both differential and common mode currents. The cable acts as a monopole antenna for CM currents. The radiation from a CM current on a cable of length L is:
Where: f = frequency (Hz), I_cm = CM current (A), L = cable length (m), r = distance (m)
For f=200 MHz, I_cm = 5 uA, L = 1 m, r = 3 m:
E_cm = (1.257e-6 * 200e6 * 5e-6 * 1) / 3 = 0.42 mV/m = -7.5 dBuV/m
Animated SVG: CM Currents on a Cable
Common Mode Current Sources on PCBs
There are several mechanisms by which CM currents are generated on a PCB and couple to attached cables:
| CM Source | Mechanism | Typical Amplitude | Frequency Range |
|---|---|---|---|
| Ground bounce | Shared inductance in ground path creates voltage between local and chassis ground | 10-500 mV | 10 MHz - 1 GHz |
| Power supply ripple | Switching regulator noise couples equally to both signal conductors | 1-50 mV | Switching freq + harmonics |
| Differential pair skew | Length/impedance mismatch converts DM to CM | 1-100 mV | Signal bandwidth |
| Reference plane discontinuity | Return current detour creates voltage drop that drives CM onto cable | 10-200 mV | Broadband |
| IC package coupling | Internal IC noise couples through package parasitic capacitance | 0.1-10 mV | Clock freq + harmonics |
| Board edge radiation | Fringing fields at PCB edge couple to cable shield | Variable | 100 MHz+ |
Cable Length and CM Radiation Efficiency
A cable's radiation efficiency as a CM antenna depends on its electrical length relative to the wavelength. The efficiency peaks at quarter-wave resonance (lambda/4) and has additional peaks at odd multiples (3*lambda/4, 5*lambda/4, etc.).
| Cable Length | Lambda/4 Resonance | Lambda/2 Resonance | Peak Radiation Band |
|---|---|---|---|
| 0.3 m (1 ft) | 250 MHz | 500 MHz | 200-600 MHz |
| 1.0 m (3.3 ft) | 75 MHz | 150 MHz | 60-200 MHz |
| 1.8 m (6 ft) | 42 MHz | 83 MHz | 30-100 MHz |
| 3.0 m (10 ft) | 25 MHz | 50 MHz | 20-60 MHz |
USB cables (typically 1-2 meters) are efficient CM radiators in the 50-200 MHz range. HDMI cables (1-3 meters) cover 30-200 MHz. Ethernet cables (up to 100 meters) can radiate efficiently across a very wide range. This is why connector-level CM filtering is essential for every I/O interface.
2. Differential Mode Currents
Definition
Differential mode (DM) currents are equal-amplitude, opposite-phase currents flowing in opposite directions on the two conductors of a pair. This is the intended signal current in most digital systems. The differential mode current is:
Where I1 flows forward on conductor 1,
and I2 flows in the return direction on conductor 2.
DM currents create a small loop antenna. The radiation from a DM current loop is proportional to the loop area and the square of frequency:
Where: f = frequency (Hz), I_dm = DM current (A), A = loop area (m^2), r = distance (m)
Note the f^2 dependency makes DM radiation increase faster with frequency,
but the small loop area (conductor spacing * length) keeps it manageable.
Why DM Radiation is Usually Smaller
For a typical cable with 2mm conductor spacing (separation s) and 1 meter length:
- The DM loop area is only A = s * L = 0.002 m^2
- At 200 MHz, a 50 mA DM current radiates: E_dm = 1.316e-14 * (200e6)^2 * 0.05 * 0.002 / 3 = 0.018 mV/m
- Compare this to CM: even 5 uA of CM current radiates 0.42 mV/m (23x more!)
DM currents are 10,000x larger than CM currents, but CM radiation can still be 20-40 dB higher because the CM antenna (full cable length) is vastly more efficient than the DM antenna (tiny loop). This is why EMI engineers focus primarily on controlling CM currents.
Animated SVG: DM Currents on a Cable
Quantitative Comparison: CM vs DM Radiation
The following table provides a direct numerical comparison of CM and DM radiation for a practical example: a 1-meter cable with 2mm conductor spacing, at 200 MHz.
| Parameter | Common Mode | Differential Mode |
|---|---|---|
| Current amplitude | 5 uA (0.005 mA) | 50 mA |
| Current ratio | 1x (reference) | 10,000x larger |
| Antenna type | Monopole (full cable length) | Loop (spacing x length) |
| Effective antenna dimension | 1 meter | 0.002 m^2 loop area |
| E-field at 3m (200 MHz) | 0.42 mV/m | 0.018 mV/m |
| E-field ratio | 23x larger | 1x (reference) |
| FCC Class B limit (3m) | ~0.3 mV/m | ~0.3 mV/m |
| Compliance status | FAIL by 3 dB | PASS by 24 dB |
The numbers above reveal the fundamental paradox of EMI: currents too small to measure on the PCB (microamps) dominate the radiated emissions. You cannot measure 5 uA on a PCB trace with a current probe. This is why near-field probing and current injection techniques are essential EMI debugging tools. The CM current is invisible until you measure the far-field radiation with an antenna.
Sources of DM Current
Unlike CM currents which are parasitic and unwanted, DM currents are the intentional signal currents in the system. Understanding the DM current path is important because DM-to-CM conversion creates the problematic CM currents:
- Signal current: The desired data signal flowing from driver to receiver through the signal conductor, returning via the ground/return conductor
- Clock current: Clock distribution signals have high spectral purity and predictable harmonics
- Power current: DC and AC power delivery currents (these are DM when flowing in the intended supply-return loop)
DM Radiation Frequency Dependence
DM radiation has a critical difference from CM radiation in its frequency dependence. DM radiation from a small current loop increases as f^2 (6 dB per octave), while CM radiation from a monopole antenna increases as f (3 dB per octave) up to the quarter-wave resonance. This means DM radiation becomes relatively more significant at very high frequencies, but the absolute levels remain lower than CM for typical cable lengths.
E_dm ~ f^2 * I_dm * A (quadratic with frequency)
Crossover frequency (where DM overtakes CM):
f_cross = (I_cm * L) / (I_dm * A * 1.05e-8)
For typical values: f_cross > 10 GHz
(CM dominates for virtually all practical cases below 10 GHz)
3. CM vs DM Comparison
Side-by-Side Visualization
Antenna: full cable length. Fields add constructively. Dominant EMI source even at microamp levels.
Antenna: small loop (conductor spacing). Fields cancel. Radiation proportional to loop area x f^2.
Key Formulas
V_dm = V1 - V2 (Differential Mode Voltage)
Equivalently:
V1 = V_cm + V_dm/2
V2 = V_cm - V_dm/2
I_cm = (I1 + I2) / 2
I_dm = (I1 - I2) / 2
Interactive Signal Decomposition Tool
Enter two signal voltages and see their decomposition into common mode and differential mode components.
CM / DM Decomposition Calculator
Interactive Waveform Decomposition
Practical Measurement: Separating CM and DM
In the lab, separating CM and DM components requires specialized measurement techniques:
| Measurement Method | What It Measures | Equipment | Frequency Range |
|---|---|---|---|
| CM Current Probe | Net CM current on cable (I1+I2) | Fischer F-33 or similar clamp-on probe | 100 kHz - 1 GHz |
| DM Current Probe | DM current (difference I1-I2) | Differential current probe | DC - 500 MHz |
| LISN (Line Impedance Stabilization Network) | CM and DM conducted emissions | LISN per CISPR 16 | 9 kHz - 30 MHz |
| Balun/Hybrid | Separate CM/DM on a differential pair | 180-degree hybrid coupler | 100 MHz - 10 GHz |
| Mixed-mode VNA | Full Sdd, Scc, Sdc, Scd matrix | 4-port VNA | 10 MHz - 40+ GHz |
CM Current Probe Technique
The most common EMI debugging tool for CM analysis is a clamp-on current probe placed around a cable bundle. The probe senses the net current through all conductors simultaneously. If the currents are perfectly balanced (equal and opposite for DM), the net current is zero. Any non-zero reading is CM current.
Probe transfer impedance (Z_T): typically 1-5 ohm from 1-500 MHz
For a probe reading of 50 dBuV into 50 ohm with Z_T = 5 ohm:
I_cm = 10^(50/20) * 1e-6 / 5 = 316 uV / 5 = 63 uA
This 63 uA of CM current is easily enough to cause EMI failures above 100 MHz.
Understanding Mode Imbalance in Power Distribution
Power distribution systems also have CM and DM components. The DM component delivers useful power (VDD-GND potential difference), while the CM component represents common noise on both rails relative to chassis/earth ground. In a well-designed system:
- Power supply output: High DM (desired VDD), very low CM (< 1 mV relative to chassis)
- After DC-DC converter: DM is the regulated output, CM can be significant (switching noise couples to both rails equally through parasitic capacitance to chassis)
- At IC: DM is the clean supply, CM noise from switching couples through package parasitics
Differential receivers inherently reject common mode signals. The Common Mode Rejection Ratio (CMRR) expresses how well the receiver ignores CM while responding to DM. A typical LVDS receiver has CMRR > 60 dB at low frequencies, meaning CM signals are attenuated by 1000x. However, CMRR degrades with frequency (often dropping to 20-30 dB above 1 GHz), which is why high-frequency CM noise can still cause problems even on differential links.
4. Mode Conversion Mechanisms
What Converts DM to CM?
In an ideal differential system, there would be zero common mode content. In practice, multiple mechanisms convert differential mode energy into common mode, creating the CM currents that cause EMI problems:
| Mechanism | Cause | CM Generation Level | Frequency Sensitivity |
|---|---|---|---|
| Trace length mismatch (skew) | Unequal routing length in diff pair | High | Increases with frequency |
| Impedance imbalance | Different Z0 on each trace | Medium-High | Broadband |
| Asymmetric connector pinout | Different path lengths through connector | Medium | High frequency |
| Via stub length mismatch | Different stub lengths on each via | Medium | At stub resonance |
| Ground plane gaps | Split or slotted reference plane | Very High | Broadband |
| Asymmetric loading | Different parasitic capacitance per trace | Low-Medium | High frequency |
Mode conversion is quantified by the mixed-mode S-parameters Scd21 (DM-to-CM conversion in forward direction). A value of Scd21 = -20 dB means 10% of the differential signal energy converts to common mode. For good EMI performance, target Scd21 < -25 dB across the frequency range of interest.
Skew-Induced Mode Conversion
Trace length skew is the most common source of DM-to-CM conversion. The CM voltage generated by skew is:
Where delta_t = skew (time difference between traces)
For small skew: V_cm approximately = V_dm * pi * f * delta_t
Example: V_dm = 800 mV, f = 5 GHz, skew = 10 ps:
V_cm = 0.8 * sin(pi * 5e9 * 10e-12) = 0.8 * sin(0.157) = 0.125 V = 125 mV
Animated SVG: Mode Conversion at an Asymmetry Point
Quantifying Mode Conversion: Scd21
Mixed-mode S-parameters provide the formal framework for measuring mode conversion in differential systems. The 4x4 mixed-mode S-matrix decomposes the standard S-parameters into DD (differential-to-differential), CC (common-to-common), DC (differential-to-common), and CD (common-to-differential) sub-matrices:
| Sdd21 Sdd22 Sdc21 Sdc22 |
| Scd11 Scd12 Scc11 Scc12 |
| Scd21 Scd22 Scc21 Scc22 |
Key parameters for mode conversion:
Scd21 = Differential-to-Common mode conversion (forward)
Sdc21 = Common-to-Differential mode conversion (forward)
Target: |Scd21| < -25 dB across frequency range
Measurement Techniques for Mode Conversion
Measuring mode conversion requires a 4-port VNA or a 2-port VNA with a balun. The measurement procedure involves:
- Calibrate the VNA at all four ports using proper calibration standards
- Measure the full 4-port single-ended S-parameter matrix [S_se]
- Apply the mathematical transformation to convert to mixed-mode S-parameters using the mode conversion matrix M
- Extract Scd21 (DM-to-CM) and verify it meets the specification limit
Practical Guidelines for Minimizing Mode Conversion
- Match trace lengths within 2 mil (5 ps) for USB 3.0+
- Match trace lengths within 0.5 mil (1 ps) for PCIe Gen5+
- Use symmetric connector pin assignments (P on one side, N on adjacent)
- Match via stub lengths within 2 mil
- Maintain symmetric impedance (equal Z0 on both traces)
- Keep differential pair spacing constant (no spreading at vias)
- Continuous reference plane under both traces
- Large length mismatch (> 20 mil for GHz signals)
- Asymmetric connector pinout (P and N on opposite rows)
- Different via stub lengths on P and N
- One trace routed over a plane gap, other trace not
- Different trace widths on P vs N (impedance imbalance)
- Unequal parasitic capacitance (one trace near a via pad)
- Breakout routing that separates the pair by different amounts
Mode Conversion Budget Example
For a USB 3.0 link at 2.5 GHz with a target Scd21 of -25 dB, the total mode conversion budget might be allocated as:
| Element | Scd21 Allocation | Design Rule |
|---|---|---|
| PCB routing skew | -35 dB | Length match within 5 mil |
| TX package asymmetry | -35 dB | IC vendor controlled |
| Connector | -30 dB | Symmetric pinout selection |
| Via transitions | -35 dB | Matched stubs, back-drilling |
| RX package asymmetry | -35 dB | IC vendor controlled |
| Total (RSS) | -25 dB | Meets specification |
5. Common Mode Chokes
How They Work
A common mode choke (CMC) is a transformer-like device with two windings on a single magnetic core. Both conductors of a pair pass through the same core. The operating principle is elegantly simple:
- Differential mode (wanted signal): Currents flow in opposite directions through the two windings, creating opposing magnetic fluxes that cancel in the core. The net impedance to DM signals is very low (ideally zero). The wanted signal passes through unimpeded.
- Common mode (unwanted noise): Currents flow in the same direction through both windings, creating additive magnetic fluxes. The core presents high impedance to CM currents, blocking them. This is the filtering action.
Z_dm = j * 2 * pi * f * L_leakage (Differential Mode: low impedance)
Typical: L_cm = 100 uH, L_leakage = 100 nH
--> Z_cm / Z_dm = 1000:1 rejection ratio
Interactive CM Choke Impedance Plot
CM Choke Selection Guide
| Application | Data Rate | Recommended L_cm | SRF Target | Key Spec |
|---|---|---|---|---|
| USB 2.0 | 480 Mbps | 90-100 uH | 100-200 MHz | Low DM insertion loss < 1 dB at 240 MHz |
| USB 3.0 | 5 Gbps | 10-30 uH | 300-500 MHz | DM IL < 0.5 dB to 2.5 GHz |
| HDMI 2.0 | 6 Gbps | 10-20 uH | 400-600 MHz | Return loss > 15 dB |
| Ethernet 1G | 1.25 Gbps | 50-100 uH | 100-300 MHz | DM IL < 1 dB at 625 MHz |
| Ethernet 10G | 10.3 Gbps | 5-10 uH | 500 MHz+ | DM IL < 0.5 dB to 5.15 GHz |
| General EMI filter | N/A | 100-1000 uH | 10-50 MHz | High Z at 30-200 MHz |
CMC Design Considerations
Designing with common mode chokes requires careful attention to several parameters beyond just impedance:
| Parameter | Importance | Design Impact |
|---|---|---|
| CM Impedance |Z_cm| | Primary spec | Higher Z = more CM attenuation. Target > 100 ohm at problem frequency. |
| DM Insertion Loss | Critical for data integrity | Must be < 0.5-1.0 dB across signal bandwidth to avoid degrading the wanted signal. |
| Self-Resonant Frequency (SRF) | Determines useful range | Peak impedance at SRF. Above SRF, impedance drops. Choose SRF near EMI problem frequency. |
| DC Resistance (DCR) | Power loss, voltage drop | Keep DCR < 0.5 ohm for power lines, < 1 ohm for signal lines. Affects PoE and USB power delivery. |
| Rated Current | Reliability | DC bias current must not saturate the core. Saturation reduces L_cm dramatically. |
| DM Return Loss | Signal integrity | Poor return loss (< 15 dB) causes reflections that degrade the data eye. |
| Leakage Inductance | DM impedance | Too much leakage adds DM impedance, affecting signal quality. |
Common Mode Choke Materials
The core material determines the frequency response of the CMC. Different materials are optimal for different frequency ranges:
- Manganese-Zinc (MnZn) ferrite: Optimal for 0.1-30 MHz. High permeability (1000-10000). Used for AC power line filtering, Ethernet magnetics.
- Nickel-Zinc (NiZn) ferrite: Optimal for 10 MHz - 1 GHz. Lower permeability (50-1000) but maintains impedance at higher frequencies. Used for USB, HDMI, high-speed data.
- Nanocrystalline: Broadband response from 0.1-100 MHz. Very high permeability (20,000-100,000). Excellent for EMI filters on power cables.
Ferrite permeability is temperature-dependent. MnZn ferrites lose permeability above their Curie temperature (~200C). At elevated operating temperatures (85-125C automotive), the CM impedance may drop by 10-30%. Always verify CMC performance at the maximum operating temperature of your application.
CMC Placement: Before or After the Connector?
The CM choke must be placed between the noise source (PCB) and the antenna (cable). Ideally, it should be as close to the connector as possible:
- Best: CMC pads directly adjacent to connector pins, within 3mm. Minimal unfiltered trace between connector and choke.
- Acceptable: CMC within 10mm of connector, with matched-impedance routing between choke and connector.
- Poor: CMC more than 20mm from connector. The unfiltered trace between choke and connector can radiate and couple to other signals, reducing the choke's effectiveness by 5-10 dB.
6. Filters and Mitigation Strategies
CM Filter Topologies
Beyond CM chokes, several filter topologies are used to suppress common mode noise at different points in the system:
| Filter Type | Mechanism | Frequency Range | Insertion Loss | Placement |
|---|---|---|---|---|
| CM Choke | Inductive blocking of CM | 1 MHz - 1 GHz | 20-40 dB CM | Near connector |
| Y-Capacitors | Shunt CM to chassis GND | 10 MHz - 1 GHz | 10-30 dB CM | At I/O boundary |
| Pi Filter (CLC) | Combined shunt + series | 1 MHz - 500 MHz | 30-60 dB | Power lines, I/O |
| Ferrite Beads | Resistive loss at high freq | 10 MHz - 2 GHz | 5-20 dB | Signal/power lines |
| Feed-through Caps | Shunt at panel boundary | 100 MHz - 10 GHz | 20-50 dB | Enclosure penetration |
Interactive Filter Response Simulation
Placement Strategies
- CM choke placed close to the connector (I/O boundary)
- Y-caps connected directly to chassis ground with short trace
- Filter components on the same layer, minimizing via transitions
- Ground plane continuous under filter section
- CM choke placed far from connector (long unfiltered cable antenna)
- Y-caps connected through long traces to ground (adds inductance)
- Filter components spread across layers with vias
- Ground plane split under filter area (defeats purpose)
Ferrite Bead Selection for CM Filtering
Ferrite beads provide resistive loss at high frequencies, converting noise energy to heat. Unlike inductors which store and release energy, ferrite beads dissipate it. This makes them effective for broadband noise suppression when the noise frequency is known.
| Ferrite Bead Type | Impedance @ 100 MHz | DC Rating | Best Application |
|---|---|---|---|
| Low-Z (0402) | 30-60 ohm | 100-300 mA | Signal line filtering (SPI, I2C) |
| Medium-Z (0603) | 100-300 ohm | 200-500 mA | Power line filtering (sensor ICs) |
| High-Z (0805) | 300-1000 ohm | 500 mA - 3A | Power supply decoupling, high-current |
| Array (4-element) | 100-600 ohm | 200 mA per line | Multi-line bus filtering (USB, SPI) |
Ferrite beads lose their impedance when the DC current approaches or exceeds their rated current. At saturation, a 600 ohm ferrite bead can drop to less than 10 ohm, providing essentially no filtering. Always check the impedance vs. DC bias current curve in the datasheet and derate by 30% for reliable operation.
Y-Capacitor Safety Considerations
Y-capacitors connect signal or power lines to chassis ground (earth). They are critical for CM filtering but must comply with safety regulations because they provide a path for AC mains leakage current. Key safety requirements:
- Y1 rated capacitors: Maximum 4.7 nF for medical equipment (IEC 60601 leakage < 0.1 mA)
- Y2 rated capacitors: Maximum 4.7 nF for ITE equipment (IEC 62368 leakage < 0.25 mA)
- Working voltage: Must exceed the peak AC mains voltage plus any transients
- Failure mode: Y-capacitors must fail open-circuit (not short) to prevent electric shock hazard
PCB Layout for CM Filter Effectiveness
The physical PCB layout of CM filter components is as critical as the component selection. Poor layout can reduce filter effectiveness by 10-20 dB:
- CM choke placed within 3mm of connector pins
- Y-caps grounded through short, wide traces to chassis ground pad
- Filter components on the same PCB layer as the connector
- No signal traces routed between filter input and output (prevents coupling around the filter)
- Separate ground regions: "dirty" (before filter) and "clean" (after filter) connected only at the filter point
- CM choke 20mm+ from connector (long unfiltered antenna)
- Y-caps grounded through 10mm traces with via transitions (adds 2-5 nH inductance)
- Filter components on different layers requiring via transitions
- Unfiltered traces running parallel to filtered traces (EMI couples back in)
- Single ground pour with no isolation between dirty and clean sides
Combining Multiple Filter Stages
For maximum CM suppression, combine multiple filter types. A typical high-performance CM filter chain for a USB 3.0 port might be:
- Stage 1 (at connector): Common mode choke (90 ohm @ 250 MHz) - blocks CM at fundamental
- Stage 2 (near choke): Y-capacitors (47 pF to chassis ground) - shunts remaining CM above 500 MHz
- Stage 3 (on cable): Ferrite clamp (snap-on ferrite) - adds 50-100 ohm CM impedance for compliance margin
The combined suppression of all three stages can exceed 40 dB at the problem frequency, compared to 15-25 dB for any single stage alone.
7. Practical Case Studies
USB 3.0 Hub Failing FCC Class B at 500 MHz
A USB 3.0 hub design was failing FCC Class B radiated emissions at 500 MHz (second harmonic of 250 MHz LFPS) by 8 dB. The emission correlated with USB 3.0 cable attachment.
Root Cause
The USB 3.0 SuperSpeed differential pair had 15 mil trace length mismatch (skew ~2.5 ps per mil = 37.5 ps total). At 500 MHz, this converted ~12% of the DM signal energy to CM. The 1.8m USB cable then radiated efficiently at 500 MHz (cable length = 0.3 lambda).
Solution
1) Corrected trace length matching to < 2 mil (< 5 ps skew). 2) Added Murata DLW21SN common mode choke (90 ohm @ 500 MHz) at the USB connector. 3) Added 100 pF Y-caps to chassis ground at connector. Result: 14 dB margin at 500 MHz.
4K HDMI Output Failing EN 55032 at 742.5 MHz
A 4K video processor board was failing EN 55032 radiated emissions at 742.5 MHz (HDMI 2.0 TMDS clock frequency 5x harmonic) with an HDMI cable attached. Emission disappeared when cable was removed.
Root Cause
The HDMI connector PCB footprint had unequal via stub lengths on the differential pairs (one via had a 12 mil stub, the other 8 mil). This created mode conversion at the stub resonance frequency. Additionally, the CM choke had SRF at 300 MHz - too low for the 742.5 MHz problem.
Solution
1) Back-drilled vias to equalize stub lengths to < 2 mil difference. 2) Replaced CM choke with higher SRF model (TDK ACM series, SRF 800 MHz, 20 dB CM attenuation at 742 MHz). 3) Added EMI gasket around HDMI connector shell for improved shielding. Result: 10 dB margin.
Gigabit Ethernet Port Causing 125 MHz Emission Spike
A network switch failed CISPR 32 at exactly 125 MHz and harmonics (250, 375 MHz). The emission was strongest when all Ethernet ports were active with traffic.
Root Cause
The Ethernet PHY magnetics (pulse transformers) had poor CM rejection (CMR) at 125 MHz. The transformers were spec'd for 30 dB CMR, but PCB layout had asymmetric trace routing to the transformer pins, degrading effective CMR to only 15 dB. The 125 MHz clock component of the Gigabit Ethernet PAM5 signal passed through as CM current onto the cable.
Solution
1) Redesigned PCB layout for symmetric routing to transformer pins (matched within 5 mil). 2) Added CM choke between PHY and magnetics (Murata BLM18KG, 90 ohm @ 125 MHz). 3) Grounded transformer center tap through 1000 pF capacitor to chassis. Result: 8 dB margin at 125 MHz.
CAN Bus CM Emission on Automotive ECU
An automotive ECU failed CISPR 25 radiated emissions on the CAN bus cable harness at 66 MHz and 132 MHz (harmonics of the 33 MHz microcontroller clock). The emissions were present even when CAN communication was idle.
Root Cause
The 33 MHz microcontroller clock was routed on an outer PCB layer, crossing a ground plane gap between the digital and CAN transceiver power domains. The clock current's return path detoured around the gap, and the resulting CM noise coupled to the CAN bus connector pins through parasitic capacitance in the CAN transceiver IC package.
Solution
1) Eliminated the ground plane split (unified ground plane). 2) Rerouted the clock to an inner layer. 3) Added a common mode choke on the CAN bus lines at the connector (Wurth WE-CNSW, 90 ohm @ 100 MHz). 4) Added 22 pF Y-caps from each CAN line to chassis ground. Result: 12 dB margin at 66 MHz.
Industrial Ethernet Switch Failing at 62.5 MHz on Cable
An industrial 8-port Ethernet switch failed EN 55032 at 62.5 MHz and harmonics when Ethernet cables were connected. The emission tracked the 125 MHz RGMII clock at its sub-harmonic.
Root Cause
The RGMII interface between the switch IC and PHY used a 125 MHz clock with significant duty cycle distortion (42% duty cycle instead of 50%). This created strong even harmonics, including the 62.5 MHz component (half the fundamental). The even harmonics coupled as CM current through the PHY magnetics, which had degraded CM rejection due to a layout asymmetry in the center-tap grounding.
Solution
1) Added a series resistor (33 ohm) on the RGMII clock to slow the edges and reduce harmonic amplitude. 2) Fixed the magnetics center-tap grounding layout for symmetric routing. 3) Added a CM choke between the PHY and magnetics (TDK ACT45B, 45 ohm @ 62.5 MHz). Result: 9 dB margin.
Key Takeaways from Case Studies
1) The root cause is almost always an asymmetry or discontinuity that converts DM to CM. 2) The solution always involves a combination of fixing the root cause (reducing CM generation) AND adding filtering (CM choke, Y-caps). 3) Fixing only the symptom with filtering is a bandaid -- always identify and address the root cause for robust compliance margin.
Systematic EMI Debug Methodology for CM/DM Problems
When a product fails radiated emissions testing, a structured debug process helps identify whether the root cause is CM or DM, and guides the corrective action:
Record all frequencies where the emission exceeds the limit. Look for patterns:
- Harmonically related (N * f_clk): Points to a specific clock source. Identify which oscillator or clock generator has that fundamental frequency.
- Broadband elevation: Suggests data-dependent noise (e.g., SerDes traffic, DDR bus activity). The emission level changes with data activity.
- Single frequency, non-harmonic: Could be a spurious oscillation, PLL spur, or intermodulation product.
Apply the ferrite clamp test: Place a snap-on ferrite clamp on the offending cable.
- Emission drops 6+ dB: The noise is CM on the cable. The cable is acting as a monopole antenna driven by CM currents from the PCB.
- Emission unchanged: The noise is either DM radiation from the cable loop, or direct radiation from the PCB/enclosure. Try absorber material on the PCB to isolate.
- Emission changes with cable routing/length: Confirms cable radiation. CM is likely if emission peaks at cable resonance (lambda/4).
Use a near-field probe (H-field loop or E-field monopole) to scan the PCB surface:
- H-field probe (magnetic loop): Detects current flow. High signal over traces carrying the clock or data at the problem frequency.
- E-field probe (monopole tip): Detects voltage hotspots. High signal over plane gaps, connector pins, or IC packages with noise.
- Correlate with layout: Overlay the near-field scan results on the PCB layout to identify the exact traces, vias, or plane gaps responsible.
Based on the mechanism identified in steps 2-3, apply targeted fixes:
| Root Cause | Quick Fix (Board Rework) | Proper Fix (Next Spin) |
|---|---|---|
| CM on cable from ground bounce | Add ferrite clamp + CM choke bodge | Improve PDN, add CM choke in schematic |
| CM from plane gap crossing | Bridge the gap with copper tape | Eliminate split or reroute trace |
| Mode conversion (skew) | Add delay on shorter trace (series cap) | Match trace lengths in layout |
| DM radiation from large loop | Add ground wire to reduce loop | Route on inner layer, reduce loop area |
| Clock harmonic on I/O cable | Add ferrite bead on clock trace | Reroute clock away from I/O, use SSC |
After applying fixes, re-measure to confirm:
- Re-scan with near-field probe to verify the noise source is suppressed
- Re-measure radiated emissions in the EMC chamber to confirm the limit is met with margin
- Document the root cause, fix applied, and margin achieved for future designs
- Target a minimum 6 dB margin above the fix to account for production variation
CM/DM Debug Equipment Summary
| Equipment | Purpose | Frequency Range | Cost Range |
|---|---|---|---|
| Near-field probe set (H+E) | Localize noise sources on PCB | 30 MHz - 6 GHz | $500 - $5,000 |
| Spectrum analyzer | Measure emission frequencies and levels | 9 kHz - 26 GHz | $5,000 - $50,000 |
| Current probe (RF) | Measure CM/DM current on cables | 10 kHz - 1 GHz | $500 - $3,000 |
| Ferrite clamp kit | Quick CM identification and suppression | 1 MHz - 1 GHz | $50 - $200 |
| TDR (Time Domain Reflectometer) | Identify impedance discontinuities | DC - 20 GHz | $10,000 - $80,000 |
| 4-port VNA | Measure mixed-mode S-parameters (Scd21) | 10 MHz - 40 GHz | $30,000 - $200,000 |
Industry Standards for CM/DM Emission Limits
| Standard | Application | Frequency Range | Class B Limit (3m/10m) |
|---|---|---|---|
| CISPR 32 / EN 55032 | Information Technology Equipment | 30 MHz - 6 GHz | 30-37 dBuV/m (10m) |
| FCC Part 15 Subpart B | Unintentional radiators (USA) | 30 MHz - 40 GHz | 29-46 dBuV/m (3m) |
| CISPR 25 | Automotive components | 150 kHz - 2.5 GHz | 24-52 dBuV/m (1m) |
| MIL-STD-461G RE102 | Military equipment | 10 kHz - 18 GHz | Varies by platform |
| IEC 61000-6-3 | Residential/light industrial | 30 MHz - 1 GHz | 30-37 dBuV/m (10m) |
Before visiting an accredited EMC test lab ($2,000-$10,000 per day), perform pre-compliance testing in your own lab using a near-field probe and spectrum analyzer. While this does not replace formal testing, it identifies 80% of problems at 10% of the cost. Focus on frequencies where the pre-compliance scan shows less than 10 dB margin -- these are your highest-risk areas.
8. Knowledge Check - CM / DM Noise
CM/DM Noise Engineering Reference Summary
| Concept | Key Formula | Critical Threshold |
|---|---|---|
| CM Voltage | V_cm = (V1 + V2) / 2 | < 10 mV for low-EMI systems |
| DM Voltage | V_dm = V1 - V2 | The intended signal |
| CM Radiation | E = 1.26e-6 * f * I_cm * L / r | I_cm < 5 uA at 200 MHz for FCC Class B |
| Mode Conversion (skew) | V_cm/V_dm = sin(pi*f*delta_t) | delta_t < 5 ps for USB 3.0, < 1 ps for Gen5 |
| CMC Impedance | Z_cm = j*2*pi*f*L_cm | > 100 ohm at problem frequency |
| CMC Rejection | Z_cm/Z_dm = L_cm/L_leakage | > 100:1 ratio |
| Scd21 Compliance | Measured per IEEE 370 | < -25 dB across bandwidth |
EMI Debugging Flowchart for CM/DM Problems
When facing a radiated emission failure that correlates with cable attachment, follow this systematic debugging approach:
- Confirm CM source: Use a CM current probe on the cable. If reading is significant (> 10 uA), CM is the dominant source.
- Identify frequency: Note the exact emission frequency. Is it a harmonic of a clock? A data rate component? This identifies the noise source.
- Trace the coupling path: Use a near-field probe on the PCB near the connector to find where CM noise is strongest. Look for plane gaps, trace crossings, and asymmetric routing.
- Fix the root cause: Correct layout asymmetries, fill plane gaps, match trace lengths, fix connector grounding.
- Add filtering: Place CM choke at connector. Add Y-caps to chassis ground. Consider ferrite clamp on cable.
- Verify: Re-measure with CM probe and far-field antenna to confirm fix effectiveness.
In practice, 80% of radiated emission failures in commercial electronics are caused by CM currents on cables. Of those, 80% can be traced to one of three root causes: (1) ground plane discontinuity under an I/O signal path, (2) trace length mismatch on a differential pair, or (3) poor connector grounding. Focusing on these three areas during design review eliminates the vast majority of CM-related EMI problems before they reach the compliance test lab.
Industry Standards for CM/DM Testing
| Standard | Scope | CM/DM Relevance |
|---|---|---|
| CISPR 32 / EN 55032 | Radiated emissions from ITE | Cable CM currents are the dominant failure mechanism |
| CISPR 25 (Automotive) | Radiated emissions from vehicles | CM on wire harnesses is the primary concern |
| IEC 61000-4-6 | Conducted immunity (CM injection) | Tests device immunity to externally induced CM noise |
| IEC 61000-4-3 | Radiated immunity | External fields couple as CM onto cables |
| DO-160 Section 21 | Aircraft EMI emissions | Very strict CM limits on all cable bundles |