123 Main Street, New York, NY 10001

Submit Your BOM

EMI Mitigation & Spread Spectrum — Sources, Paths, and Fix Loops

← PMIC Home Layout & Snubbers Synchronous Buck Controller Boost & Buck-Boost Controller PMBus / Power System Manager
PMIC ↑

This page shows fast, low-cost ways to bring power-converter EMI under control: identify where noise is generated, how it couples, and what to change first. We focus on practical levers—loop geometry, damping, input/output filtering, and spread-spectrum trade-offs—plus a repeatable measure → hypothesis → action → verify loop. Deep loop-compensation, minimum on-time, valley/peak current details are covered on sibling pages (Synchronous Buck / Boost & Buck-Boost / Layout & Snubbers / PMBus-PSM).

Spread-spectrum switching reduces narrowband peaks in power converter EMI spectra.
Spread spectrum smears energy to lower peak amplitude.

EMI Basics — Sources · Paths · Victims

Sources. Fast edges (high dV/dt and dI/dt) excite parasitic L/C and create narrowband peaks and ringing. The hot-loop area and impedance largely set the raw amplitude and radiation efficiency—shrink the loop first, then consider filters and spread-spectrum. Expect a line spectrum around the switching frequency with sidebands from ringing and load modulation.

Paths & victims. Three dominant couplings: DM (conducted differential) through the rail/return loop and LISN path, CM (common-mode) via stray capacitances, shields and harness (the whole cable “rides” the noise), and radiated from large loop areas, slots, and cable antennas. Always classify the path first; different tools solve different paths (DM→LC/π+damping; CM→choke/shield termination; radiated→loop geometry & seams).

  • Shrink the hot loop (HS FET → diode/SR → Cin) and keep the return tight.
  • Keep SW copper compact; avoid routing under sensitive analog nodes.
  • Fingerprint the dominant path (near-field probe / current clamp) before adding parts.
  • Fix geometry & damping first; use spread-spectrum as a finishing lever.
  • Log before/after spectra with identical setups to avoid “peak whack-a-mole”.
Path Symptoms First Levers Verify
DM (Conducted) Peaks at fsw and harmonics on LISN; ripple interaction LC/π + damping (ESR/series-R/RC); snubber Before/after LISN overlays; Bode after heavy filtering
CM (Common-mode) Cable/housing currents; radiated fails despite DM fixes CM choke; 360° shield termination; edge-rate control Clamp current on harness; chamber/antenna near-field scan
Radiated Peaks vs loop geometry; enclosure seams and slot resonances Shrink hot loop; stitch returns; fix seam impedance; shorten cables Near-field probe maps; A/B on cable routing and seam bonding
Hot and cold current loops around switch, diode or synchronous rectifier, input capacitor, inductor and load in a buck converter.
Shrink the hot loop first; filters and spread spectrum work better on a quiet geometry.

With the path classified, move to Measurement: keep cables repeatable, run a quick LISN pre-scan and near-field sweep to pick the right lever before investing in parts.

Measurement & Limits — Quick, Repeatable, Honest

Conducted (150 kHz–30 MHz): LISN · Cable Discipline · Termination Consistency

For conducted pre-scans, verify LISN bonding, fix cable length and routing across runs, and keep a consistent ground/termination topology. Start with a wide sweep to reveal peaks, then re-scan around hot spots. Log both Peak and Average detectors under identical operating points.

Radiated: Near-Field Probes · Current Clamp · Reproducible Posture

Use near-field H/E probes around the SW node, inductor, and harness to fingerprint sources. Place a current clamp on the cable to expose common-mode components. Photograph the setup (distances, cable paths, antenna/probe posture) so every fix is re-measured in the same geometry.

Domain Focus Tools Reproducibility
Conducted (150 kHz–30 MHz) Peaks at fsw & harmonics on LISN; ripple interaction LISN, spectrum analyzer (Peak/Avg), stable mains impedance Lock cable length/route; repeat operating points & terminations
Radiated Hot spots near SW node, inductor, seams; harness as antenna Near-field H/E probes, current clamp on harness, chamber/antenna checks Photograph layout & posture; annotate distances for exact A/B overlays

Pre-scan steps (fast loop)

  1. Fix cable length & route; confirm LISN ground/bonding.
  2. Record operating points (VIN, load, modes) and keep constant across runs.
  3. Wide sweep → identify top-N peaks; then zoom around each bucket.
  4. Capture Peak & Average traces; export raw files with timestamps.
  5. Near-field & clamp fingerprint (DM vs CM) to choose the right lever.
  6. Photograph the setup; save into the same run folder as spectra.
Guardrail mindset: treat limits as guardrails, not a tightrope. Keep a 3–6 dBµV margin to absorb build variance, cable routing drift, and temperature spread.

Helpful resources: EMI Pre-Compliance Checklist (PDF), Spread-Spectrum Playbook (PDF).

Spread Spectrum — Profiles, Trade-offs, and Verification

Profiles & selection logic

Choose between Triangular sweep and Random jitter, and decide center-spread vs down-spread. For tight timing/sync domains, down-spread is often safer; for RF-sensitive products, random jitter tends to reduce discrete tones, while triangular is easier to model and reproduce.

Side-effects checklist

  • Audible noise & beat tones with other clocks or load cycles.
  • Loop-gain variation near crossover; ripple redistribution & efficiency shifts.
  • Energy relocation into sensitive bands (AM/FM/ISM/automotive); conducted vs radiated can diverge.
Practical start: set depth to 3–6%. Above 10–15%, reassess loop stability, ripple, and efficiency. Keep modulation rate out of the audible band or randomize sufficiently.
Type / Bias Strengths Risks Best-fit scenarios Notes
Triangular sweep Predictable skirts; easy to tune & reproduce Discrete skirts may land on RF masks Systems that favor modeling/validation repeatability Sweep rate: avoid audible; confirm with LISN overlay
Random jitter Lower discrete tones; often quieter for radios Harder to analyze; repeatability depends on RNG RF-sensitive products; coexistence with wireless Seed & bounds must be documented in production
Center-spread Average fsw unchanged; minimal timing drift Risk of pushing energy to both sides of masks Tight regulation loops; synced PWM domains Verify loop gain around crossover
Down-spread Safer for timing budgets; avoids exceeding max fsw Lower average fsw may affect ripple & efficiency Timing-critical / multi-domain sync systems Check ripple change & audible interaction
Fixed-frequency vs triangular vs random-jitter spectra with reduced peak levels.
Spread spectrum lowers peak amplitude by distributing energy—verify side-effects before committing.

Verification flow

  1. Capture baseline conducted & radiated spectra at fixed operating points.
  2. Enable profile A/B; sweep depth and modulation rate.
  3. Select the minimum effective depth that meets limits with margin.
  4. Re-check loop stability, ripple, efficiency, and cross-domain timing/audio.

See also: Spread-Spectrum Playbook (PDF), RC Snubber Quick Rules (PDF).

Input/Output Filtering — LC/π Selection & Damping

LC/π selection: target corner & controlled damping

Pick L and C from the desired corner frequency and ripple goals, then add loss on purpose to avoid peaking: via capacitor ESR, a series resistor, or an RC damper across L/C. π filters provide steeper DM attenuation but may introduce a mid-node resonance— verify interactions with the control loop.

Damping strategies (from light to heavy)

Method Mechanism Strengths Trade-offs Use when… Verify
Cap ESR Intrinsic loss flattens the LC peak No extra parts; simple ESR varies with temp/age; limited range Mild peaking; cost/space constrained Ripple, temperature drift, LISN overlay
Series-R Adds controlled DM loss in series path Predictable damping; scalable Efficiency drop; power rating needed Moderate peaking; tunable attenuation Thermals, step load transient, Bode
RC across L Bypasses HF energy and damps the L branch Strong peak control; localizes loss at HF Power in R; layout proximity matters Sharp peaking near LC corner; ringing Scope overshoot; resistor heating
RC across C Adds HF shunt loss at the cap legs Simple routing; broad HF damping May raise ripple; part tolerance sensitive Distributed HF hash; broad skirts Ripple spectrum; LISN & scope decay

When to use a CM choke (cable egress & shields)

If common-mode dominates (harness currents, radiated fails despite DM fixes), place a CM choke near the cable egress and implement 360° shield termination. CM chokes do little for DM peaking; pair them with geometry fixes and damping.

Control-loop interaction: after adding significant filtering, always recheck Bode (crossover, phase margin) and transient response; adjust compensation if needed.
LC and pi filters with damping options—ESR, series R, and RC placement to avoid peaking.
Damping turns unstable peaking into a well-behaved corner; verify loop stability after changes.

If peaking is driven by SW-node ringing, move to Snubbers. For CM-dominated fixes, see “Shielding & Cables”.

Snubbers — Fast Sizing, Placement, and Scope Verification

Types & placement

Use RC to ground at the SW node, RC across the diode/synchronous MOSFET, or an RC-R variant for stronger damping. Keep loops short, place parts physically close to the node, and return to a quiet ground.

Quick sizing & scope method

  1. Measure ringing frequency fr and estimate node capacitance Cnode.
  2. Estimate L ≈ 1 / ( (2π·fr)² · Cnode ).
  3. Start Csnub ≈ 0.5–1.0 × Cnode.
  4. Pick R ≈ √( L / Csnub ); begin slightly overdamped, then trim.
  5. On the scope, minimize overshoot and ringing without excessive power loss.
Symptom Likely cause Quick action Verify
Resistor overheating Over-damping or too small package/power rating Reduce C or increase R; choose higher-power package Thermal rise at worst-case duty/load
Overshoot persists Under-damping; parasitic L in placement Increase C slightly; move parts closer to SW node Scope overshoot/settling vs baseline
New peak appears Alternate resonance excited by values/placement Sweep R in small steps; fine-trim C LISN overlay / near-field map after change
CM noise increases Snubber return creates CM path; geometry too large Shorten return; review ground reference; consider CM choke Clamp current on harness; radiated pre-scan
Power & placement: estimate P on the resistor and ensure voltage and pulse ratings are met. Place the network as close as possible to the SW node with the shortest return.
RC snubber placements around the switching node and diode/MOSFET for ringing control.
Keep the snubber loop compact; trim R/C on the scope for minimal overshoot with acceptable loss.

Detailed sizing steps: RC Snubber Quick Rules (PDF).

Layout Levers — Minimal Checklist That Moves the Needle

Core principles

  • Shrink the hot loop: HS FET → diode/SR → Cin → return. Keep the physical triangle compact.
  • SW copper compact: do not route under sensitive analog nodes; define a keepout if needed.
  • Rsense / analog star-ground: separate high-dI/dt returns; Kelvin sense only to the sense element.
  • Place DM/CM filters at the cable egress: control the path at the boundary first.

Return paths & reference planes

Maintain a continuous reference plane. Put HF bypass caps close to the switch and return via short, low-inductance vias. Avoid slots under fast return paths. Prefer multilayer stacks: parts on top, solid inner-layer ground, short and direct traces.

Minimal actionable checklist

  • Draw the hot-loop outline; make it visibly shorter/tighter after each placement tweak.
  • Keep SW node local; fence it from analog/Kelvin sense routes.
  • Place Cin HF cap so HS/LS/source pins and cap pads form a tight triangle.
  • Star the analog ground; single-point return to power ground after Rsense/Kelvin.
  • Put CM/DM filters next to the connector; route cables consistently.
Geometry first: when the geometry is quiet, filters can be smaller and spread spectrum becomes a finishing lever—not the primary fix.

Deep layout patterns, scope shots, and snubber placement live on the sibling page: Layout & Snubbers.

Shielding & Cables — Control the Common-Mode Path

360° shield termination beats long pigtails

Use a 360° clamp or full-wrap termination for the cable shield at the chassis. Keep the bond short, wide, and circumferential to lower HF impedance. Long pigtails add inductance and undo shielding at high frequency—avoid them wherever possible.

Seams and conductive finishes

Lower seam impedance with frequent fasteners and wide, clean contact. Break anodize/paint where electrical bonding is required: use conductive gaskets, finger-stock, plating, or localized mask-offs. Slot geometry and poor contact increase radiated peaks.

Cables and CM choke placement

If common-mode dominates, put a CM choke near the cable egress on the shield side. Keep cables short and routed over a reference plane or the chassis. Where appropriate, evaluate single-ended shield termination vs multi-point schemes with pre-scans.

Common-mode choke and 360-degree shield termination at cable egress to reduce radiated EMI.
Place the CM choke close to the egress and use a 360° shield clamp; avoid long pigtails.
Item Do Don’t Verify
Shield termination 360° clamp; short, wide bond to chassis Long pigtail ground leads Radiated pre-scan; clamp current on harness
Seams & coatings Break paint/anodize where bonding is needed; use gaskets Rely on cosmetic contact through coatings A/B with near-field probe along seam
CM choke placement Place near egress; pair with 360° shield clamp Place far from connector or before long cable runs Clamp current drop; chamber scan delta
Cable length & routing Keep short and repeatable; route over ground/chassis planes Floating in air; inconsistent lengths and loops Photographic record of posture; repeatability across runs
CM-first: confirm the common-mode path with a current clamp or near-field map before adding chokes or shield rework. Change one variable at a time and re-measure in the same posture.

Next: close the loop with a repeatable Fix-Loop Playbook, or revisit Measurement to ensure posture consistency.

Fix-Loop Playbook — Fingerprint → Classify → Change One → Verify

  1. Fingerprint: run LISN pre-scan (150 kHz–30 MHz), near-field map around SW node/inductor, and a current-clamp on the harness. Log top-N peaks.
  2. Classify: assign each peak to DM, CM, or radiated. Map levers: DM→LC/π + damping; CM→choke & 360° shield; radiated→geometry/seams.
  3. Change one variable: add one lever at a time (e.g., RC snubber, series-R, CM choke, down-spread). Keep operating points and posture identical.
  4. Verify & record: overlay before/after spectra (Peak/Avg), photograph posture, document settings and side-effects. Aim for a 3–6 dBµV margin.
Evidence, not whack-a-mole: always save paired before/after plots with the same posture, operating points, and filenames (timestamp + build + change note). One change per run.

Remediation log template

Change Path (DM/CM/Rad) Domain Before After Δ Notes / Side-effects
Add RC snubber @SW→GND DM Conducted 123 kHz @ 66 dBµV (Peak) 123 kHz @ 58 dBµV (Peak) −8 dBµV Ripple +2 mVpp; R hot @50°C
Add CM choke @egress CM Radiated 40–60 MHz peak over limit Within limit +4 dB margin pass Clamp current −60%

Templates & procedures: EMI Pre-Compliance Checklist (PDF), RC Snubber Quick Rules (PDF), Spread-Spectrum Playbook (PDF).

Vendor Options — Spread-Spectrum Capabilities (TI · ST · NXP · Renesas · onsemi · Microchip · Melexis)

Spread spectrum reduces peak amplitude by redistributing energy; it does not remove energy. Profiles (Triangular / Random) and bias (Center / Down) have different system impacts. The table below is a series-level overview—always verify the latest datasheet for exact capabilities.

Vendor Product type Spread spectrum Depth / Rate (e.g.) Notes Docs
TI Buck controllers / PMIC Triangular / Random; Center/Down (model-dependent) e.g., 3–10% depth; kHz-range sweep Sync options, phase interleaving on some families Datasheet
ST Buck / Module Triangular; often Down e.g., 3–8%; audio-safe rates Automotive-grade options available Datasheet
NXP PMIC / Automotive Random / Triangular (series-specific) e.g., 3–6%; randomized rates Focus on car-grade coexistence Datasheet
Renesas Controllers / PMIC Triangular / Random; Center/Down e.g., 3–10%; vendor-defined ranges Sync & frequency foldback notes Datasheet
onsemi Controllers / Modules Triangular; often Down for timing safety e.g., 4–8%; kHz sweep Automotive & industrial lines Datasheet
Microchip Controllers / PMIC Random / Triangular (device-specific) e.g., 3–6%; audio-aware dithering Good docs for spread configuration Datasheet
Melexis Automotive PMIC / drivers Random / Triangular (series-specific) e.g., 3–6%; EMI-friendly modes Automotive sensors & drivers ecosystem Datasheet
Datasheet-first: capabilities differ by device and revision. We can prepare a pin-to-pin comparison matrix and run a quick bench validation on your shortlist.
Submit your BOM (48h) Contact us

Need deeper loop/layout guidance? See Layout & Snubbers. For side-effects and verification steps, revisit Spread Spectrum.

Resources & Downloads

Submit your BOM (48h) Contact

FAQ — Practical Answers

Will enabling spread spectrum guarantee compliance?
No. Spread spectrum reduces peak amplitude by redistributing energy, but total energy remains. If common-mode paths or geometry dominate, the impact is limited. Classify DM/CM first, fix geometry and damping, then use spread spectrum as a finishing lever. See #spread.
Center-spread vs down-spread: how do I choose?
Down-spread preserves timing budgets by avoiding excursions above nominal fsw; good for synchronized domains. Center-spread keeps the average fsw unchanged. Choose based on clock coexistence and regulation margins, then verify on LISN and near-field. See #spread.
What depth should I start with?
Start at 3–6%. Beyond 10–15%, reassess loop stability, ripple and efficiency. Set modulation rate outside the audible band—or randomize sufficiently—and recheck both conducted and radiated spectra. See #spread.
Triangular sweep or random jitter—what’s safer?
Triangular is predictable and easy to tune; random tends to lower discrete tones near radios but is harder to analyze. Match to your coexistence needs, then verify with identical posture and overlays. See #spread.
Why did output ripple increase after enabling spread?
Frequency dithering redistributes ripple energy and may interact with filter Q. Re-check damping (ESR/series-R/RC), verify loop gain near crossover, and confirm LISN overlays. Adjust depth/rate to avoid moving energy into sensitive bands. See #spread and #filters.
Why can a CM choke fix cases where DM filters fail?
DM filters target differential current in the rail/return loop. Common-mode currents ride the entire harness and bypass DM elements. A CM choke impedes this path—especially when paired with 360° shield termination at the egress. See #filters / #shielding.
How do I size and verify an RC snubber quickly?
Measure ringing frequency and estimate node capacitance; pick Csnub ≈ 0.5–1.0×Cnode and R ≈ √(L/Csnub). Start slightly overdamped, then trim on the scope. Check resistor heating and CM side-effects. See #snubbers.
Should I change the PCB layout or enable spread first?
Fix geometry first: shrink the hot loop, keep SW copper local, and separate sensitive returns. When geometry is quiet, filters shrink and spread becomes a finishing lever. Verify with identical posture between revisions. See #layout / #spread.
Conducted passes but radiated fails—why?
Likely a common-mode or antenna geometry issue: long cables, high-Z seams, poor shield bonds. Add 360° shield termination, move CM choke to the egress, and retest with clamp current/near-field maps. See #shielding / #filters.
How do I avoid fix-one, break-another ping-pong?
Enforce the single-variable rule, keep posture identical, and overlay before/after spectra. Photograph setups, log operating points, and hold a 3–6 dBµV margin for build drift. Track changes in a remediation log. See #playbook.
Cable rules: length, routing, and LISN consistency?
Fix cable length and route across runs, keep routing over a reference plane or chassis, and bond LISN consistently. Photograph posture and store images with spectra files. Consistency is the cheapest way to get trustworthy overlays. See #measurement / #shielding.
How to assess audible noise and beat tones conflicts?
Sweep depth and modulation rate; keep rates outside the audible band or apply randomized jitter. Check for beats with other clocks and load cycles. Validate both conducted and radiated spectra under audio-sensitive operating modes. See #spread.
What changes when rails are paralleled/interleaved?
Interactions include phase alignment, loop coupling, and CM current stacking. Prefer synchronized or phase-shifted operation, constrain return geometry, and keep cable rules consistent. Validate interactions with overlays and clamp-current checks. See #spread.
How should I manage low-frequency harmonics?
Watch for load-modulation and burst/hiccup modes that create LF envelopes. Limit LF excursions by mode selection or current limits. Use Average detector traces, then align damping/filtering without hurting stability. See #measurement.
Automotive notes: what’s different?
Tighter margins across temperature and aging, larger harness variability, and stricter coexistence with radios. Favor 360° shield terminations, dense seam bonding, and conservative spread depth. Validate multiple postures and cable sets before freezing. See #shielding.

icnavigator

ICNavigator helps engineers find verified ICs, alternatives, and fast quotes.

Explore
Categories
  • Power Management
  • MCU
  • Automotive ICs
  • Sensors
  • Interface & Transceivers
  • Analog Front-End
Get in Touch
  • Email: hello@icnavigator.com
  • WeChat/WhatsApp: +86-xxxx

© 2025 ICNavigator. All rights reserved.