Module 4.7 - Cables & Connectors (EMC)
Design review checkpoints for cable EMC design, shield termination, connector selection, and cable routing for interference control
Design review checkpoints for cable EMC design, shield termination, connector selection, and cable routing for interference control
For EMC purposes, cable shields must be terminated at both ends of the cable to provide effective shielding above the cable's first resonant frequency. Single-end grounding only provides electrostatic (E-field) shielding but loses H-field shielding effectiveness above the first resonance. The "ground loop" argument for single-end grounding applies only to audio-frequency analog systems.
| Cable/Termination | Z_t at 1 MHz | Z_t at 100 MHz | Z_t at 1 GHz |
|---|---|---|---|
| Semi-rigid coax | 0.5 mohm/m | 0.5 mohm/m | 0.5 mohm/m |
| Double-braid RG-223 | 2 mohm/m | 5 mohm/m | 20 mohm/m |
| Single braid RG-58 | 5 mohm/m | 50 mohm/m | 200 mohm/m |
| Foil + drain wire | 20 mohm/m | 500 mohm/m | 2 ohm/m |
| Braid + 360-deg connector | 5 mohm/m | 10 mohm/m | 50 mohm/m |
| Braid + 25mm pigtail | 200 mohm | 16 ohm | 157 ohm |
HDMI cable with proper termination: HDMI cable with braided shield terminated 360 degrees to connector metal shell at both ends. Shell provides continuous contact to PCB-mounted receptacle through multiple spring fingers. Receptacle shell grounded to PCB ground plane via 8 grounding tabs. Measured common-mode rejection: 55 dB at 300 MHz. System passes FCC Class B radiated emissions with HDMI cable attached.
Sensor cable with shield grounded at one end only: 3m shielded sensor cable carrying 10 MHz digital signal. Shield grounded at controller end, floating at sensor end. First quarter-wave resonance at 25 MHz. At 30 MHz (3rd harmonic of 10 MHz signal): shield provides approximately 0 dB shielding (near resonance). Cable radiates as if unshielded at 30 MHz. Radiated emission at 30 MHz exceeds FCC Class B by 18 dB. Fix: ground shield at both ends -- emission immediately drops below limit.
Foil shields with drain wire: Foil-shield cables use an aluminum foil with a drain wire for termination. The drain wire is essentially a pigtail -- at high frequencies, it provides poor termination. Foil shields are suitable only for electrostatic shielding; for RF EMC, always specify braided shields or braid-over-foil construction.
Connector-less shield termination: In some harnesses, the shield is twisted and soldered to a ground pad on the PCB. This creates a pigtail that negates shielding above 10 MHz. Always use a connector with shell-to-shell contact for RF shield continuity.
The connector shell (metal body/housing) serves as the transition point between cable shield and enclosure shield. It must provide low-impedance connection to the chassis/enclosure at the point of entry. Any impedance in this path degrades both the cable shielding and the enclosure shielding at that penetration point.
DB-9 connector with proper chassis bonding: DB-9 connector mounted on rear panel via threaded inserts (metal-to-metal contact between connector shell and panel). Connector shell ground lug also bonds to PCB ground plane via 4-via pad. Internal shield fence connects PCB ground to chassis at the connector location. Measured shell-to-chassis impedance: 2 mohm DC, 8 mohm at 500 MHz. Cable shield current has continuous low-impedance path through entire system.
USB connector with plastic housing gap: USB Type-A receptacle mounted on PCB, but the enclosure panel cutout is 2mm larger than the connector body (for tolerance). Air gap between connector shell and panel edge. No spring contacts or gasket to bridge the gap. This 2mm gap is a slot antenna at each connector location. At 2.4 GHz: 2mm gap provides 0 dB shielding. Cable shield current that flows to connector shell has no path to chassis -- it re-radiates from the connector body into the enclosure interior.
Even with a properly shielded cable and grounded shell, noise can propagate on the signal conductors inside the connector. Filtered connectors incorporate capacitors, inductors, or Pi-filters on each pin to attenuate conducted noise at the enclosure boundary. This is particularly important for signals that cannot use shielded pairs internally.
Type 1: Capacitive (C-type) filtered connector
- Capacitor from each pin to shell (ground)
- Typical values: 100 pF to 100 nF per pin
- Provides low-pass filtering with source impedance
- f_c = 1/(2*pi*R_source*C_filter)
- Example: 1 nF with 50 ohm source -> f_c = 3.2 MHz
- Insertion loss at 100 MHz: 20*log10(1 + 50/(1/(2*pi*100e6*1e-9))) = 50 dB
Type 2: Inductive (L-type) filtered connector
- Ferrite or inductor in series with each pin
- Provides high impedance at noise frequencies
- Less common, used for power pins
Type 3: Pi-filter connector (C-L-C)
- Full Pi-filter (capacitor-inductor-capacitor) per pin
- Best filtering performance (60 dB/decade rolloff)
- Most expensive, largest size
- Used in military/aerospace applications
Type 4: Planar array filtered connector
- SMD capacitor array mounted inside connector housing
- Lower inductance than discrete capacitors
- Good for frequencies to 10 GHz
- Examples: Smiths Connectors HyperGrip, Amphenol SIG filters
Standard connector families with filter options:
- D-Sub (DB-9, DB-15, DB-25, DB-37): all types available
- Circular MIL-DTL-38999: C and Pi-filter versions
- Rectangular MIL-DTL-83723: filtered versions available
- Micro-D (MDM series): filtered versions for space/defense
MIL-DTL-38999 connector with per-pin filtering: 26-pin circular connector for industrial controller. Pins 1-4 (power, DC): 100 nF C-type filter. Pins 5-12 (RS-485 data, 500 kbps): 1 nF C-type filter (f_c = 1.3 MHz with 120 ohm source). Pins 13-20 (discrete I/O, DC-100 Hz): 47 nF C-type filter. Pins 21-26 (spare/unused): 100 nF C-type. All filter capacitors rated for 50V DC. Result: 55 dB insertion loss on all pins at 30 MHz, system meets MIL-STD-461G CE102 with 10 dB margin.
Filtered connector with wrong capacitor values: DB-25 connector with 100 nF filters on ALL pins including 10 Mbps Ethernet data pins. The 100 nF capacitor has f_c = 32 kHz with 50 ohm source -- far below the 5 MHz signal bandwidth. Ethernet signal is completely attenuated by the filter. Link fails to establish. Fix: use 47 pF for Ethernet pins (f_c = 68 MHz) or remove filter from data pins and rely on cable shielding instead.
Cables routed in proximity can couple noise through mutual capacitance and mutual inductance. Segregating cables by signal type prevents high-noise cables (power, switching) from coupling into sensitive cables (analog sensors, receivers). Physical separation and cable tray organization are essential EMC practices.
| Category | Cable Types | Shielding | Separation from Cat 3 |
|---|---|---|---|
| 1 (Sensitive) | Low-level analog, sensor, audio, RF receive | Required (braided) | > 100 mm |
| 2 (Medium) | Digital data, control, RS-485, Ethernet | Recommended | > 50 mm |
| 3 (Noisy) | AC power, motor drives, relay, solenoid, PWM | Required | N/A (source) |
| 4 (High-Speed) | USB 3.x, HDMI, PCIe, 10GbE | Required (double) | > 75 mm |
Industrial control cabinet with proper segregation: Three separate cable trays: Top tray (power/motor cables, shielded), Middle tray (control and data cables, 100mm below power), Bottom tray (sensor/analog cables, 100mm below control). Vertical risers have separate conduits for each category. Where cables must cross categories, they cross at 90 degrees (minimizes coupling). Cable tray dividers are grounded metal. Result: No measurable crosstalk between power and sensor cables during motor switching events.
All cables in single bundle: System with VFD motor cable, 4-20 mA sensor cables, and RS-485 data cables all running in the same conduit for 5 meters. During motor acceleration, VFD switching noise (10 kHz - 10 MHz) couples to analog sensors (200 mV noise on 4-20 mA loop = 5% error) and corrupts RS-485 communication (bit errors during motor transitions). Fix: separate into minimum two conduits (power separate from everything else) with 100mm spacing.
Crossing angle matters: Two cables crossing at 90 degrees have minimal coupling (M approaches zero). Two cables running parallel for even 10 cm can have significant coupling at high frequencies. When different cable categories must share a path, minimize parallel run length or use physical barriers (grounded metal separators).
Bundle resonance: A bundle of cables can exhibit transmission-line resonances. At the resonant frequency, coupling between cables in the bundle increases dramatically. This is especially problematic for long parallel runs (> 1m) above 30 MHz.
Snap-on or clamp-on ferrites placed around cables provide common-mode impedance that reduces both conducted and radiated emissions from cables. They are particularly effective as a low-cost retrofit solution and are commonly used on USB cables, HDMI cables, and power cords (the "lumps" visible on many cables).
USB cable ferrite (target: 100-500 MHz):
TDK ZCAT2035-0930 (snap-on, 9.5mm ID)
Impedance: 90 ohm at 100 MHz, 145 ohm at 300 MHz
Single pass: 6 dB attenuation at 300 MHz
2 turns: 580 ohm = 12 dB at 300 MHz
Power cord ferrite (target: 1-30 MHz):
Fair-Rite 0431164281 (snap-on, 13mm ID, mix 31)
Impedance: 200 ohm at 1 MHz, 350 ohm at 10 MHz, 200 ohm at 100 MHz
Mix 31 optimized for 1-300 MHz range
2 turns: 800 ohm at 1 MHz = 14 dB
HDMI cable ferrite (target: 200 MHz - 2 GHz):
Laird 28A2025-0A0 (snap-on, high-frequency mix)
Impedance: 100 ohm at 200 MHz, 200 ohm at 500 MHz, 150 ohm at 1 GHz
Single pass on HDMI cable: 4-6 dB attenuation
Placement rules:
1. Place ferrite as close to the noise source as possible
(near connector at the "dirty" end)
2. For both-end radiation: place at BOTH cable ends
3. Multiple ferrites are better than one large one (distributed)
4. Ferrite should be tight around cable (no air gap)
5. For flat cables: use ferrite designed for ribbon/flat cable geometry
USB cable with optimized ferrite placement: USB cable with two snap-on ferrites: one at the device end (near PCB connector, within 5cm), one at the host end. Device-end ferrite: Fair-Rite 2643102002 (mix 43, 100 ohm at 300 MHz), cable passed through twice (2 turns). Host-end ferrite: same, single pass. Total CM impedance at 480 MHz (USB harmonic): 400 + 100 = 500 ohm. Common-mode current reduced from 15 uA to 4 uA. Radiated emission at 480 MHz drops from 42 dBuV/m to 30 dBuV/m (12 dB improvement, now 7 dB below Class B limit).
Ferrite selected for wrong frequency: Design has conducted emission problem at 150-500 kHz (SMPS switching harmonics). Designer adds Fair-Rite mix 43 snap-on ferrite (optimized for 25-300 MHz) to the power cord. At 300 kHz: mix 43 impedance is only 8 ohm (mostly inductive, low loss). Attenuation: less than 1 dB -- no improvement. Fix: Use mix 31 material (optimized for 1-300 MHz, still provides 150 ohm at 500 kHz) or mix 75/77 for very low frequencies (below 1 MHz).
Cables have resonant frequencies determined by their length and termination conditions. At resonance, the cable acts as an efficient antenna, dramatically increasing both radiated emissions and susceptibility to external fields. Understanding cable resonances is critical for predicting and mitigating EMC problems.
System tested with multiple cable lengths: Pre-compliance testing performed with USB cable lengths of 0.5m, 1.0m, 1.5m, and 2.0m to identify worst-case resonance alignment. Found that 1.5m cable (resonant at 50 MHz) coincides with 5th harmonic of 10 MHz USB frame marker. Added ferrite (mix 31, 200 ohm at 50 MHz) at connector -- provides 6 dB suppression at 50 MHz regardless of cable length. Passes with all cable lengths from 0.3m to 3.0m.
Single cable length tested, fails in field: Product passes FCC with 1m USB cable (lab test configuration). Customer uses 2m cable. Cable resonance at 37.5 MHz aligns with 5th harmonic of 7.5 MHz clock. Radiated emission at 37.5 MHz increases by 18 dB compared to 1m cable test. Product fails at customer site when measured by enforcement authority. Recall required to add ferrite to every unit in field.
Cable Resonance Measurement: Use a current probe (Fischer F-33 or similar) around the cable while sweeping a signal generator at the cable CM source. Plot CM current vs. frequency. Peaks in CM current indicate cable resonances. Alternatively, use a VNA with a current probe to measure the cable's common-mode impedance -- dips indicate resonances.