Skip to main content

3.3 EMI/EMC Management

Grounding, shielding, and cable segregation.

EMI (electromagnetic interference) and EMC (electromagnetic compatibility) are less about mystery and more about geometry: where currents travel, how they return, and what the metalwork lets them do. A quiet box starts by choosing a ground scheme that makes sense, then bonding lids, trays, and studs so the chassis becomes the obvious return path—not the wiring harness. Shields only earn their keep when they’re terminated properly (ideally 360° at the entry), while cable segregation and tight pairing keep loops small so wires stop acting like antennas. These choices belong in assembly flow as much as in design, because clamps, gaskets, and routing are what turn the intent into physics on the floor. With a couple of fast checks—a low-ohm bond read and a quick sniff for hot spots—the build leaves the bench already calm.

3.3.1 The goal (in one line)

Keep noise in the noisy places and out of the quiet ones—by giving currents short, obvious return paths and making cables poor antennas.




3.3.2 Know your players: sources vs victims

Noisy stuff (emitters): switching PSUs, motor/solenoid drivers, DC–DCs, relays, PWM lines, long unshielded harnesses, poor shield bonds.

Sensitive stuff (susceptors): radios/GNSS, analog sensors, high-impedance nodes, USB/Ethernet PHYs, high-speed clocks, long signal runs.

First rule: shorten loop area (route out and back together) and give noisy currents a nearby chassis to hug.




3.3.3 Grounding strategy (decide once, build to it)

  • Chassis vs 0V: treat chassis/earth as the EMC sink; treat 0V as the circuit reference. Bond them deliberately, where designed.
  • Single-point (low-freq analog): tie 0V↔chassis at one slot (star) to avoid hum loops.
  • Multi-point (RF/fast edges): use short, wide bonds at several points to lower HF impedance.
  • Equipotential bonding: all metalwork (lids, frames, trays) must read < 0.1 Ω to chassis at their bond pads (log at first article).
  • Earth joints: serrated/star washer on bare metal (23.1/23.2), torque per map, measure < 0.1 Ω.

Rule of thumb: if the problem is low-frequency ground loop (mains ripple), think single-point. If it’s RF spray, think multi-point/short bonds.



3.3.4 Cable shields & terminations (where the magic happens)

  • 360° termination at entry: clamp braid/foil to chassis as the cable enters the box (backshell, gland, or clamp). Avoid pigtails; if unavoidable, keep ≤ 10 mm.
  • Drain wires: bond alongside the 360° clamp, not to a distant PCB pad.
  • Which end(s) to bond?
    • HF noise / radiated control: both ends (360°) to kill common-mode.
    • Low-freq instrumentation (e.g., 4–20 mA): single-end at the quiet end to avoid DC loops—but still add a high-frequency path (capacitor/RC or 360° with DC-isolation hardware) if the spec allows.
  • Coax: treat the shield as signal return; bond shells properly (U.FL click, SMA torque).
  • EMI gaskets: ensure clean bond lands; target 20–30% compression; verify seam < 0.1 Ω (23.4/23.5).



3.3.5 Segregation & routing (turn geometry into a filter)

  • Zoning inside the box:
    1. Noisy zone (PSUs, motors, relays).
    2. Digital zone (processors, high-speed).
    3. Quiet/analog/RF zone (sensors, radios, GNSS).
      Keep short paths to their exits; do not meander.
  • Separation: keep noisy power/switching harnesses ≥ 100 mm from low-level signals; when crossing, do it at 90°.
  • Pairing: run supply with its return (twisted if possible) to shrink loop area.
  • Chassis-hugging: clamp long runs along metal rails (24.1) to reduce radiation and pickup.
  • Ferrites: last-resort band-aids; pick cores sized for the cable and place right at the entry/exit to choke common-mode.
  • Antenna keepouts: keep digital and power wiring away from RF modules/antennas; avoid routing behind antenna ground cuts.



3.3.6 Interface panels & filters (stop noise at the wall)

  • Feedthroughs / filtered connectors: use at the bulkhead for nasty lines (PWM, long I/O).
  • Common-mode chokes on I/O just inside the entry; tie the choke’s shield/can to chassis nearby.
  • RC/TVS right at the connector where ESD enters; shortest leads to chassis.



3.3.7 Inside the enclosure (bonds that actually work)

  • Paint scrape pads under every intended bond (23.1).
  • Board-to-chassis bonds: short, wide straps or mounting posts near the I/O shield.
  • Lid seams: conductive gasket or fingers around the perimeter, even compression (23.5).


3.3.8 Quick verification & pre-compliance sanity

  • Bond resistance: seam/strap/chassis points < 0.1 Ω (low-ohm meter).
  • Shield continuity: shell ↔ shell and braid ↔ chassis low ohms.
  • Near-field sniff: handheld probe (or small loop) over hot zones—compare “clamp on/off”, “ferrite in/out” deltas.
  • Common-mode current clamp: on external cables during function—look for drops after 360° bonds/ferrites.
  • Functional abuse test: toggle relays/PWM while watching radios/GNSS/analog noise; verify no resets or data errors.



3.3.9 Acceptance cues (fast eyes)

Feature

Accept

Reject

Shield terminations

360° clamp at entry; pigtails ≤ 10 mm only if spec’d

Foil braid twisted to long pigtail; clamp far from entry

Bond pads

Bare, clean; < 0.1 Ω to chassis

Painted pad; star washer on paint; high Ω

Segregation

Power/relay harnesses away from analog/RF; 90° crosses

Paralleling noisy and quiet runs for long spans

Pairing/returns

Twisted or taped pairs; returns co-routed

Return separated → big loop area

Gaskets

Continuous contact, even squeeze

Gaps/crushed foam; inconsistent seams

Ferrites

At entry/exit; sized to cable

Random mid-span beads “just because”



3.3.10 Common traps → smallest reliable fix

Trap

Symptom

First move

Shield bonded to PCB 0V only

Radiates on cable

Move to 360° chassis bond at entry; add HF decoupling if single-end needed

Long drain pigtails

Fails radiated emissions

Replace with 360° clamp; shorten to ≤ 10 mm if constrained

Ground loop fear everywhere

Hum fixed, RF worse

Use single-point for LF, multi-point for HF; add DC-isolated HF bond

Paint under earth lugs

High leakage, ESD upsets

Scrape pad; serrated washer; verify < 0.1 Ω

Big harness loops

Analog noise, resets

Pair with return; route along chassis; add clamp spacing ≤ 200–300 mm

Filters far from wall

Internal re-radiation

Move chokes/feedthroughs to the bulkhead

Ferrite scattergun

Costs rise, little effect

Measure current; place few, correct cores at entries



3.3.11 Pocket checklists

Before build

  • Ground strategy chosen (single- vs multi-point) and drawn
  • Bond pads exposed; Ω meter at station; star washers in kit
  • Shield hardware (360° clamps/backshells) present; ferrites (if any) sized

During routing

  • Noisy runs segregated; crosses at 90°; power paired with return
  • Shields bonded at entry; pigtails ≤ 10 mm (only if spec’d)
  • Gaskets seated; seam compression even; lid bonds in place

Quick verify

  • Chassis bonds < 0.1 Ω logged (tray, lid, straps)
  • Sniff/common-mode check shows improvement vs open case
  • Radios/GNSS/analog behave while PWM/relays chatter




Bottom line: decide the ground scheme, bond metalwork properly, 360°-terminate shields at the wall, and route cables in tight, paired, chassis-hugging paths with sensible distance between noisy and quiet worlds. Prove it with a couple of quick measurements, and most EMC headaches never get a chance to start.