This page turns “EMI good/bad” into measurable targets and a repeatable closure loop: identify whether peaks come from reflection or differential-to-common-mode conversion, then use one-change-at-a-time knobs (slew/drive/symmetry/SIC + harness/termination) to pass emission and RF immunity with evidence and numeric pass criteria.
Outcome: stable links (no bus-off/false-wake), reproducible measurements, and a production-ready checklist that locks comparability, validates worst-case conditions, and prevents “fake improvements” from settings or harness differences.
H2-1 · Scope & Boundary (What this page is / is NOT)
This page exists to keep EMI emission and RF immunity decisions measurable and comparable:
define target metrics, map dominant mechanisms, choose the right knobs (slew / SIC / harness / termination),
and close the loop with validation and production-ready checks—without turning the site into a mixed encyclopedia.
This page covers
Metric contract: what “good” means for emission and immunity, and what must be logged to compare results.
Mechanism map: how edge-rate, symmetry, reflection, and common-mode conversion create peaks or wideband noise.
Actionable knobs: programmable slew/drive/symmetry and SIC waveform integrity as levers for emission vs robustness.
System terms: harness, stub, termination, and return-path contributions that can dominate chip-level improvements.
Test & validation: measurement setups, comparability traps, and evidence-first verification loops.
Selection logic: how to choose a transceiver family by EMI/immunity needs without mixing protocol/controller scope.
Engineering closure: design → bring-up → production checklist items and pass/fail criteria placeholders.
Rule of thumb: if a section cannot be tied back to a measurable emission/immunity metric or a validation step, it is likely out of scope for this page.
Diagram · “Solar-system” boundary map (this page vs siblings)
Use this map as a scope guard: stay on measurable emission/immunity outcomes and the knobs that move them.
H2-2 · EMI & Immunity Targets (Metrics that actually matter)
Treat emission and immunity as a metric contract. Without a stable measurement definition, “improvement” often comes from
setup differences (bandwidth, averaging, harness, termination, grounding) rather than real robustness.
The four buckets to track (keep results comparable)
1) Emission outcomes
Narrow-band peaks: often driven by reflection/ringing + imbalance.
Wideband noise floor rise: often driven by excessive edge energy and common-mode current.
Common-mode current: a direct proxy for harness-as-antenna radiation risk.
2) Immunity outcomes
No bus drop: link remains functional under injected disturbance.
Error counters: error rate stays below threshold (placeholder X/Y) during and after injection.
No false wake (if selective wake): wake source attribution remains stable (bus/local/timed).
3) Signal-quality proxies
Edge rate (slew): moves harmonic energy but can steal timing margin.
One or a few narrow peaks; sensitive to stub/termination changes.
Average / quasi-average level
dBµV
Broad energy content; reflects persistent noise, not single spikes.
Detector (avg), dwell time (X), sweep method, harness routing locked.
Noise floor rises across a band; often edge-energy and CM current driven.
Common-mode current
mA
Direct proxy for harness radiation potential and imbalance.
Clamp position (X), frequency point list, harness return path fixed.
High CM current aligns with high peaks; changes with symmetry/termination.
Error counter rate
errors/min
Immunity success must be tied to quantified communication stability.
Injection method (BCI/ALSE), level (X), load (A/B), time window (T).
Counters spike at certain bands; may precede bus-off or dropouts.
Wake-source attribution (if applicable)
events
Separates real bus wake from RF/noise-induced false wakes.
Wake table locked, logging fields fixed, harness + environment recorded.
False wakes correlate with specific injection bands or harness routing.
Comparability rule: if any “How measured” field changes, treat the result as a different dataset—do not claim improvement without re-baselining.
Pass criteria (three tiers; use placeholders)
Tier A · Emission pass
Peak/average levels meet the limit under locked setup (band X–Y, RBW/VBW X, distance X, harness config A).
Tier B · Immunity pass
Under injection method M at level X, the network remains functional, and error counters stay ≤ Y per time window T.
Tier C · System pass
Tier A and B remain true across harness variants/batches and temperature range, with the same measurement contract and logging fields.
Diagram · Metrics mapping (System → Signal → Outcomes)
This mapping prevents “random tuning”: each knob must be tied to a proxy change, then verified against emission and immunity outcomes under a locked setup.
H2-3 · Emission Mechanisms (Where radiation really comes from)
A differential bus radiates when differential energy is converted into common-mode current. Once common-mode current flows on the harness,
the harness behaves like an antenna and produces narrow peaks (resonance / ringing) or a raised noise floor (excess edge energy).
Emission causal chain (use this to avoid random tuning)
1) Source
Edge energy (dv/dt, di/dt): faster edges push more harmonic energy into the spectrum.
Asymmetric transitions: unequal rise/fall or dominant/recessive behavior increases conversion risk.
Ringing seeds peaks: reflections create narrow-band spikes at specific frequencies.
2) Conversion (Diff → CM)
Imbalance: connector asymmetry, unequal return paths, or output impedance mismatch.
Termination bias: midpoint/return differences turn differential transitions into common-mode current.
Stub & T-branch: asymmetric reflection and discontinuities increase conversion and ringing.
Reference loop: ground offsets and return-path detours expand the common-mode loop area.
3) Antenna (Harness)
Effective length: harness length and routing set resonance bands and coupling efficiency.
Loop area: return-path choices determine how strongly CM current radiates.
Topology: branching and shielding breaks create “unexpected antennas.”
4) Radiation outcome
Narrow-band peaks: usually reflection/ringing or harness resonance dominated.
Wideband floor rise: usually excessive edge energy + persistent CM current.
High CM current: strong indicator of imbalance and antenna-like harness behavior.
Cross-page boundary: layout placement recipes and termination value optimization belong to sibling pages; this section focuses on mechanism → checkpoints → measurable evidence.
Evidence-first checks (before changing hardware)
Peak vs floor classifier
If the failure is a few sharp peaks, suspect reflection / resonance / discontinuity first. If the noise floor rises across a band, suspect edge-energy and CM-current first.
CM current correlation
Track common-mode current at fixed clamp locations. If CM current tracks the emission failure band, the dominant problem is almost always Diff → CM conversion + harness antenna behavior.
Single-change discipline
Change one lever at a time (slew/drive/symmetry OR a single harness/termination variable), then re-measure under the same metric contract. Otherwise, conclusions become non-comparable.
Diagram · Energy path (Driver → Imbalance → CM loop → Harness antenna → Emission)
Reading guide: emission is rarely “just differential.” The dominating step is often the Diff → CM conversion, which then drives harness radiation.
Immunity failures often happen because injected disturbance becomes common-mode stress on the harness, then shifts the receiver reference
or modulates the sampling window. The visible waveform can look acceptable while the decision threshold is moving.
Injection path map (outside → inside)
Layer 1 · Harness pickup
RF/EFT energy couples into the harness and becomes common-mode disturbance. Harness length, routing, and return-path geometry strongly shape which bands are worst.
Layer 2 · CM → decision point
Threshold movement: the receiver reference shifts, turning a clean-looking waveform into wrong decisions.
Ground/reference drift: return-path and reference bounce couples into the comparator input window.
Layer 3 · System response
Error frames / counters: bursts at specific bands often precede bus-off and dropouts.
Dominant/recessive flips: threshold stress can cause wrong state decisions under injection.
False wake: in selective-wake designs, injected disturbance can mimic a wake pattern without a real bus frame.
Boundary note: immunity here means RF/EFT susceptibility impact on communication decisions. ESD/surge device details and placement belong to the protection pages.
Immunity failure signatures (symptom → path → first evidence)
Symptom
Most likely path
Priority evidence
Error counter spikes at specific bands
Harness CM injection → threshold/window modulation
Band sensitivity repeats; CM current correlates; counters rise before link drop
Reading guide: immunity failures are best debugged by mapping symptom → injection band → CM correlation → decision-point movement, then validating with counters and pass criteria placeholders.
H2-5 · Programmable Slew / Drive Strength (The knobs and trade-offs)
Programmable slew and drive settings reduce high-frequency spectral energy, but they can also reduce timing and decision margin.
The goal is to choose the smallest emission footprint that still preserves stable threshold crossing under worst-case load and temperature.
Knob panel (what can be tuned)
1) Tx slew rate (rise / fall)
Primary proxy change: dv/dt and harmonic energy (wideband floor tendency).
Typical EMI effect: slower slew lowers broadband energy and reduces over-excited ringing peaks.
Typical robustness effect: slower slew can shrink sampling margin, especially on long lines and heavy loads.
2) Drive strength (current / output impedance)
Primary proxy change: source impedance and reflection excitation/damping.
Typical EMI effect: stronger drive can increase peaks by exciting discontinuities; weaker drive may lower peaks but risks edge flattening.
Typical robustness effect: stronger drive improves threshold crossing certainty under heavy load; too strong can increase ringing sensitivity.
3) Symmetry trim (rise/fall, dominant/recessive)
Primary proxy change: diff→CM conversion risk (imbalance-driven CM current).
Typical EMI effect: improved symmetry often reduces CM current and suppresses narrow-band peaks caused by conversion.
Typical robustness effect: more stable decision threshold crossing under injected CM stress; beware of margin shifts across temperature.
Boundary note: register maps and vendor-specific tuning sequences belong to each transceiver page; this section defines knob classes, effects, and validation logic.
Costs (expressed as measurable risks)
Timing margin shrink
Slower edges increase threshold-crossing uncertainty and reduce usable sampling window under worst-case load.
Evidence: increased crossing spread (X), higher error counters at the same bus utilization.
Edge flattening under heavy load
Excessively weak drive or too-slow slew can cause edge “plateau” near the decision threshold on long harness/heavy loads.
Evidence: slow recovery, dominant/recessive levels approach threshold (X), BER/error spikes increase under load.
Temperature drift & repeatability
A “good” setting at room temperature can lose symmetry or margin at temperature extremes.
Evidence: symmetry metrics drift (X), emission peaks shift bands, and immunity counters rise at hot/cold.
Decision table (Knob → effect → side effects → when to use)
Knob
Primary proxy
Likely EMI impact
Side effects
Validation
Tx slew
dv/dt, harmonic energy
Floor ↓, peaks may ↓
Timing margin ↓
Same metric contract; counters ≤ X over T
Drive
source impedance / damping
Peaks can ↑ if over-driven
Ringing sensitivity ↑
Peak band shift? CM current correlation?
Symmetry trim
diff→CM conversion
CM current ↓, peaks ↓
Hot/cold drift risk
Symmetry metric ≤ X; immunity counters stable
Discipline: change one knob at a time, re-measure with identical setup and metric contract, then decide with counter-based pass criteria placeholders (X/T).
Diagram · Edge vs spectrum trade-off (Fast slew vs Slow slew)
Use case: choose slower slew only after confirming counters and threshold-crossing stability across load and temperature.
H2-6 · SIC Waveform Integrity (Symmetry, shaping, and why it helps EMI & robustness)
SIC waveform integrity can be assessed as measurable signal quality: symmetry, controlled ringing, and stable dominant/recessive levels.
These reduce diff→CM conversion, suppress peak excitation, and stabilize threshold crossing under injected disturbance.
Waveform integrity dimensions (measurable)
1) Symmetry
Rise/fall and dominant/recessive transitions should be balanced to reduce diff→CM conversion.
Evidence: symmetry metric ≤ X under fixed load and harness setup.
2) Overshoot & ringing
Overshoot and ringing should be controlled to avoid narrow-band peaks and threshold-crossing jitter.
Evidence: overshoot ≤ X and ring decay within Y cycles.
3) Level stability
Dominant/recessive levels should remain stable with minimal droop/drag under heavy load.
Evidence: level droop ≤ X and recovery time ≤ Y under worst-case conditions.
Boundary note: SIC architecture evolution and standard history belong to the SIC page; this section focuses on EMI/immunity outcomes and acceptance checks.
SIC waveform acceptance checklist (setup locked)
Setup lock: harness length/routing, termination, node count, and bus utilization must remain identical across comparisons.
Symmetry: rise/fall and dominant/recessive transition mismatch ≤ X; CM current trend must improve or remain stable.
Ringing: overshoot ≤ X; ring decay within Y cycles; peak bands should not shift into more sensitive ranges.
Level stability: dominant/recessive droop ≤ X under heavy load; no prolonged plateau near decision threshold.
Immunity linkage: error counters ≤ X over T during injection; false wake count = 0 over T (if applicable).
Temperature: repeat checks at hot/cold; waveform metrics and counters remain within pass criteria placeholders.
Result format: record waveform metrics + CM current + counters in the same run sheet to preserve comparability.
Diagram · SIC vs non-SIC waveform (same load condition)
The comparison must be made under identical load and harness conditions; otherwise, waveform “improvement” can be a setup artifact.
When emissions or immunity failures persist after swapping transceivers, the dominant term is often the system: harness geometry, termination behavior, and stubs/T-branches.
The correct sequence is evidence first (CM current / peak-band / waveform signatures), then parameter changes.
Harness (length, routing, return path)
Why it dominates: harness geometry sets antenna efficiency and common-mode loop area.
Mechanism chain: imbalance → diff→CM conversion → CM loop area → radiation / injection.
First evidence: CM current trend at a fixed clamp position and peak-band drift when harness routing changes.
First action: lock routing/bundling and re-check CM current + worst-band peaks before changing transceiver knobs.
Termination (reflection, ringing, CM contribution path)
Why it dominates: mismatch creates reflections that excite ringing and narrow-band peaks.
Signature: emission failures appear as stable spectral lines/peaks aligned with ringing behavior.
CM path (high-level): termination imbalance and midpoint reference behavior can increase diff→CM conversion.
First evidence: ringing amplitude/decay + peak-band correlation; CM current changes under the same electrical setup.
Boundary note: detailed termination network sizing belongs to the CMC & Split Termination page; this section focuses on contribution paths and signatures.
Stubs & T-branches (FD/XL sensitivity)
Why it dominates: stubs create extra discontinuities and asymmetric branches that amplify diff→CM conversion.
FD/XL sensitivity: higher edge content and tighter margins make peak excitation and threshold modulation more visible.
Signature: narrow-band peak bands move when stub length changes; CM current rises when branches are asymmetrical.
First action: confirm with a controlled A/B change (one stub variable) before transceiver tuning.
System-term priority table (what often eats chip advantage)
System term
Dominance trigger
Typical signature
First evidence
First action
Harness
Long runs, large loop area, routing changes
Peak bands drift with routing / bundling
CM current @ fixed clamp position
Lock geometry, re-check worst band
Termination
Discontinuities, heavy loads, peak-like failures
Stable narrow-band peaks, ringing
Waveform ringing + peak correlation
Evidence first, then knob changes
Stubs/T-branch
Multi-drop, asymmetry, FD/XL edges
Peak bands move with stub length
A/B stub change + CM trend
Single-variable experiment
Reference / ground offset
Large return ambiguity, injection sensitivity
Immunity counters spike without EMI shift
Counters vs injection band correlation
Lock return path, re-test
Rule: evidence first (CM current + peak-band + waveform signature), then change one variable and re-test with the same measurement contract.
Interpretation: the system sets conversion and antenna efficiency; chip knobs help only after dominant system terms are evidenced and constrained.
H2-8 · Measurement & Validation (How to measure without fooling yourself)
The measurement chain is part of the system. Consistent setup, fixed analyzer settings, and structured logging are required to avoid false improvements.
Use a minimal closed-loop workflow: lock setup, baseline, sweep, correlate, change one variable, then re-test.
Emission measurement (comparability first)
Probe selection: near-field (local hot spots) vs current clamp (CM current trend) vs antenna (system-level signature).
Peak capture discipline: lock probe position/orientation and harness condition; use the same sweep bands and dwell.
False improvement traps: RBW/VBW, detector mode, averaging, and peak-hold settings can hide worst-case peaks.
Minimum record fields: probe type + position, distance, RBW/VBW, detector/average, worst band + peak level (X).
Immunity measurement (define functional pass)
Injection setup: define injection point and sweep plan; keep clamp location and harness state fixed.
Monitors that matter: error counters, dropouts, false-wake count, and retry storms (with a fixed counting window).
Trigger conditions: log band/level (X), dwell time (T), bus utilization, temperature/humidity, and grounding state.
Pass criteria placeholders: counters ≤ X per window, dropouts ≤ X per T, false wakes = 0 per T (if applicable).
Correlation (bench vs vehicle / harness differences)
Harness delta: routing, bundling, shielding continuity, and node count can shift worst bands and CM current.
Environment delta: temperature/humidity and return path conditions can change both peaks and immunity counters.
System delta: gateway/wake strategy and traffic profile change the observable failure signature.
Correlation rule: only compare results that share the same measurement contract and logging fields; otherwise, conclusions are not portable.
Guardrail: never change analyzer settings and declare an improvement; lock the contract, log counters, and prove repeatability.
H2-9 · EMI/Immunity Debug Playbook (From symptom to root cause)
Debugging becomes repeatable when each symptom maps to a minimal evidence pack, a short list of likely causes, single-variable fixes, and a measurable pass criterion (X/T placeholders).
Playbook rules (non-negotiables)
Lock the contract: harness state, probe/clamp position, analyzer settings, traffic profile, and counting window.
Collect evidence first: worst band, peak vs floor classification, CM current trend, waveform signature, counters.
Change one thing: one knob or one system term per iteration; log every change.
Define pass: peak ≤ X dB, counters ≤ X per window, dropouts ≤ X per T, false wake = 0 per T (where applicable).
Emission triage (first split: narrow-band peak vs wideband floor)
Narrow-band peak: stable lines/peaks often tied to ringing, reflections, discontinuities, or diff→CM conversion.
Wideband floor: elevated noise floor often tied to dv/dt/di/dt energy and CM loop efficiency.
Minimal evidence pack: worst band (X–Y), peak level (X), CM current @ fixed clamp, waveform ringing/overshoot flag.
Next step: select the matching branch below, then enforce single-variable changes.
Evidence to collect: ringing amplitude/decay vs peak-band, peak-band shift vs stub length change, CM current correlation.
Fix options (choose one bucket per iteration): reduce excitation (slew/drive), improve symmetry (SIC/symmetry trim), constrain system term (stub/termination topology).
Pass criteria: peak band ≤ X with repeatability across ≥ N runs; CM current reduced by ≥ X% (optional).
Boundary note: termination network sizing and CMC details belong to the dedicated termination/CMC pages; this branch is signature-driven.
Emission branch B · Wideband floor (edge energy / CM loop efficiency)
Likely causes: overly fast edges, strong drive, large CM loop area, harness routing that increases antenna efficiency.
Evidence to collect: floor level over band, CM current trend, sensitivity to harness routing/bundling, edge-rate change response.
Fix options (single variable): tune slew/drive within timing margin, reduce conversion by symmetry improvements, constrain CM loop (system term check).
Pass criteria: floor ≤ X across the defined band; functional margin preserved (timing criteria X).
Immunity triage (first split: PHY decision vs policy vs power/reset)
PHY decision: threshold modulation and CM injection → bit errors and counter spikes.
Controller policy: retry storms, bus-off behavior, wake filtering, and logging gaps can mimic PHY weakness.
Power/reset chain: injection disturbs rails or reset/monitor lines → dropouts or false wakes.
Minimal evidence pack: counters vs band/level, state logs (policy), rail/reset correlation (power) with timestamps.
Worst band + peak (X), ringing signature, CM current trend
Reflection / discontinuity / conversion
Tune slew/drive OR improve symmetry OR constrain stub/topology
Peak ≤ X across ≥ N repeats
Emission over-limit, raised floor
Floor level vs band, CM trend, edge-rate sensitivity
Edge energy / CM loop efficiency
Tune slew/drive OR constrain loop via system term check
Floor ≤ X, timing margin ≥ X
Immunity: counters spike
Counters vs band/level, CM clamp response, state logs
PHY threshold modulation or conversion
Improve symmetry/SIC OR reduce excitation OR tighten system term
Counters ≤ X/window at worst band
Immunity: dropouts / resets
Reset/PG timestamps, rail ripple, dropout count vs level
Power/reset chain sensitivity
Stabilize chain (checkpoints) OR reduce injected CM conversion
Dropouts ≤ X per T
Immunity: false wake
Wake source attribution, injection band, logging completeness
Policy/logging gaps or threshold modulation
Fix attribution/logging OR improve immunity margin
False wake = 0 per T
Diagram · Debug flow (Emission vs Immunity, tree-style)
Discipline: classify first, collect the minimal evidence pack, then apply a single-variable fix and re-test under the same contract.
H2-10 · Selection Logic (Turn EMI results into a decision tree)
Selection becomes objective when test results and system constraints are translated into gates: system-first vs chip-first, need for waveform symmetry (SIC), and required knob strength (slew/drive/symmetry).
Quick table is for fast screening; strict table enforces comparability and evidence requirements before selection conclusions.
Quick selection table
Constraint
Failure signature
Class tendency
Must-have
First validation
Long harness / many branches
Peak + CM trend
SIC / stronger symmetry
Symmetry/SIC + repeatable logging
A/B under one contract
Floor-dominant emission
Raised floor over band
FD/XL with strong knobs
Programmable slew/drive
Timing + EMI re-test
Immunity counters at band
Counters spike vs band/level
SIC / stronger symmetry + stability
Stable threshold behavior
Worst-point retest
Strict selection table (comparability enforced)
Inputs
System-first gate
Class choice
Required knobs
Evidence pack
Pass criteria
Harness + topology + goals
CM & band stable? yes/no
Classic/FD/SIC/XL (lens)
slew/drive/symmetry
worst band + CM + counters
X/T/N repeatability
Non-comparable data
Fail (re-run)
No decision
No decision
Rebuild the contract
—
Diagram · Selection tree (Inputs → Gates → Class output)
Outcome rule: selection conclusions are valid only when inputs and results are gathered under one comparable contract (same harness/termination/settings/logging).
Purpose: convert EMI emission/immunity targets into a repeatable engineering closure loop with evidence, single-variable changes, and pass criteria placeholders (X/Y).
Design Gate (lock comparability before touching knobs)
Freeze the network “identity”: harness version, node count, worst-load definition, termination topology (end/split/stubs) and any CMC/TVS population options. Evidence: photo + BOM snapshot. Pass: identity record completeness ≥ X%.
Define a single “default knob policy”: slew/drive/symmetry default states and allowed adjustment window (only one knob per experiment). Evidence: knob table with allowed states. Pass: policy approved + versioned.
Pick a reference PHY/transceiver class for correlation: at least one “SIC-class” and one “non-SIC-class” device for A/B (e.g., TJA1462 vs baseline FD transceiver). :contentReference[oaicite:15]{index=15}
Pre-plan split termination footprint (if used): include two RT/2 resistors + midpoint capacitor footprint for quick A/B. Example parts: RC0603FR-0760R4L + GCM188R71H472KA37D. :contentReference[oaicite:16]{index=16}
Pre-plan CMC footprint options: allow “DNP / CMC / alt-CMC” in the same line location to isolate differential→common-mode conversion effects. Example WE-CNSW parts: 744232090 / 744235900. :contentReference[oaicite:17]{index=17}
Pre-plan bus protection footprints (low-C): keep placement close to connector, with a “DNP option” for SI comparison; example PESD2CANFD24VQB-Q, SM24CANB-02HTG, ESDCAN24-2BLY, ESD2CANFD24. :contentReference[oaicite:18]{index=18}
Define the logging contract: counters (error frames, bus-off, retries), link state transitions, wake-source attribution (bus/local/timed), and timebase alignment. Evidence: field dictionary. Pass: log completeness ≥ X% for all test runs.
Define the “peak capture rule”: narrowband peak vs broadband noise classification + how the worst point is stored (frequency, amplitude, detector mode). Evidence: screenshot + exported trace. Pass: worst point reproducible within ±X dBµV and ±Y kHz.
Define the immunity pass rule: which observable defines failure (BER/counters/dropouts/false wake) and the time window. Evidence: pass/fail script. Pass: failure is unambiguous (no subjective calls).
Document “only-one-change” discipline: each run must include a diff summary (“what changed vs baseline”). Evidence: change log. Pass: 100% runs have a diff record.
Run baseline emission capture first: store worst band + worst point, and label peak type (narrowband/broadband). Evidence: trace export + peak table. Pass: baseline repeatability within X dB.
Capture baseline common-mode current trend: fixed clamp location; store current vs frequency or key band points. Evidence: saved trace. Pass: repeatability within X%.
Near-field scan for localization: use a consistent probe set (e.g., RF1/RF2/TBPS01) and store “top-3 hotspots” for correlation. :contentReference[oaicite:19]{index=19}
Slew sweep (one knob only): slow→fast (or fixed order), record (a) peak shift, (b) CM current shift, (c) counters impact. Pass: selected setting improves emission ≥ X dB without violating error criteria.
Drive strength sweep (if available): keep slew fixed; sweep drive states and record ringing/overshoot signature vs peak behavior. Pass: no new narrowband peaks beyond X dB over baseline.
Symmetry/shape sweep (SIC-class or symmetry trim): quantify dominant/recessive symmetry (qualitative bins) and relate to CM conversion evidence. Pass: symmetry improves while counters remain ≤ X/window.
System A/B before “changing chips”: DNP/insert CMC or change termination option (one change) and re-run baseline + one sweep point. Example CMC footprints: WE-CNSW 744232090/744235900. :contentReference[oaicite:20]{index=20}
Immunity validation with fixed injection setup: keep traffic fixed; sweep frequency/level; record first failure point + signature. If BCI is used, document clamp model (e.g., FCC F-140). :contentReference[oaicite:21]{index=21}
Separate failure layers: classify as (A) PHY decision errors, (B) controller strategy side-effects (timeouts/retries), (C) power/reset coupling. Evidence: counters + timestamps + reset flags. Pass: layer is identified, not guessed.
Verify protection A/B effect on SI/EMI: populate vs DNP TVS arrays (one change), re-check peak + error signature. Example parts listed above (PESD2CANFD24VQB-Q / SM24CANB-02HTG / ESDCAN24-2BLY / ESD2CANFD24). :contentReference[oaicite:22]{index=22}
Freeze the “worst-case regression”: harness state + termination option + knob state + traffic profile + environment fields. Evidence: a single frozen test recipe. Pass: recipe reproduced on 2 different days within X margin.
Lock analyzer settings to avoid “fake improvements”: RBW/VBW/detector/averaging must remain unchanged across A/B runs. Evidence: settings checksum. Pass: 0 unauthorized changes.
Production Gate (consistency control → drift watch → sampling rules)
BOM identity lock: harness/termination/CMC/TVS part numbers and alternates are controlled; any alternate triggers a re-run of the frozen worst-case recipe. Pass: no uncontrolled alternates.
Parameter lock: analyzer settings, clamp/probe positions, traffic profiles, and pass/fail windows are frozen per station. Pass: station-to-station delta ≤ X.
Environment fields are mandatory: temperature, humidity, ground reference conditions, and fixture ID must be logged for each run. Pass: field completeness ≥ X%.
Golden unit strategy: maintain a golden reference build and re-run worst-case weekly (or per X lots) to detect drift. Pass: drift ≤ X dB and counters within X/window.
Incoming harness/termination inspection: verify correct population (split termination cap, CMC DNP state, TVS presence) and connector integrity. Pass: 0 critical escapes per X units.
AOI-friendly packages when required: if production requires AOI wettable flank packages, keep ordering code documented (e.g., Nexperia PESD2CANFD24VQB-Q). :contentReference[oaicite:23]{index=23}
Immunity spot-check: test at the pre-identified “first-fail band” + margin, not a full sweep; log counters and link stability. Pass: no failure at level X for Y minutes.
Emission spot-check: measure the previously worst narrowband peak and the broadband floor at fixed bands. Pass: peak margin ≥ X.
Change-control triggers: any PCB stack, connector vendor, harness routing, or shield bond change triggers a Bring-up Gate partial rerun. Pass: change tickets include evidence pack.
Field-return correlation pack: returns must include firmware version, logging fields, and a photo of harness/connector state. Pass: pack completeness ≥ X%.
False-wake watch (if applicable): track wake-source attribution in fleet logs; define acceptable false-wake rate. Pass: false-wake ≤ X/day per node.
Sampling rules: define per-lot sample size and escalation steps if drift is detected. Pass: sampling executed as written (no ad-hoc).
One-page checklist template (copy into a project bible)
Use a single-row-per-check format. Keep “Evidence” mandatory and “Pass criteria” numeric (X/Y placeholders until program targets are known).
Gate
Check item
Why it matters
How to check
Evidence
Pass criteria
Owner
Design
Network identity frozen
Prevents non-comparable EMI data
Record harness + termination + population options
Photo + BOM + version tag
X
—
Bring-up
Single-knob sweep
Avoids multi-variable confusion
Change only slew/drive/symmetry per run
Trace export + counters + diff note
X/Y
—
Production
Spot-check worst band
Detects drift quickly
Measure fixed peak + floor bands
Station log + trace ID
X
—
Mobile note: the table is intentionally scroll-wrapped to prevent horizontal page drift.
Diagram: three gates with fail-back loops (evidence-driven closure)
Keep the process linear in success cases, but force a controlled rollback when evidence is missing or drift appears.
Fixed format per question: Likely cause / Quick check / Fix / Pass criteria (numeric placeholders X/Y). Scope stays strictly within emission + RF immunity for the bus PHY.
Emission fails only in one narrow band—reflection peak or common-mode conversion first?
Likely cause: a resonance-driven narrowband peak (reflection/ringing) or diff→CM conversion that produces a CM-current spike at f0.
Quick check: lock the analyzer settings and log (1) peak @ f0 (dBµV) and (2) CM current @ f0 (mA_rms); if CM rises with the peak, prioritize CM conversion.
Fix: change one variable only: (A) terminate/stub option A/B for reflection, or (B) symmetry/slew/CMC A/B for CM conversion; re-run only the f0 band first.
Pass criteria: peak @ f0 margin ≥ X dB and CM current @ f0 ≤ X mA_rms with counters stable (errorFrames ≤ X/min).
Lowering slew passed EMI but random errors increased—what margin did you steal first?
Likely cause: slower edges reduced high-frequency energy but consumed timing/threshold margin under heavy load (long harness, high Cbus, low Vdiff at sampling).
Quick check: keep EMI setup unchanged and correlate errors with load/temperature; record edge-rate proxy (rise/fall time bins), sample-point window (X% UI), and errorFrames (/min).
Fix: back off one step from the slowest slew, then use termination/CM suppression (split/CMC option) instead of further edge slowing; keep “one-change” discipline.
Pass criteria: emission margin ≥ X dB while link integrity remains: errorFrames ≤ X/min and bus-off = 0 over Y minutes at worst load.
SIC “enabled” but emissions got worse—what waveform integrity check is fastest?
Likely cause: “SIC on” changed edge symmetry/shape in a way that increased diff→CM conversion or created ringing at the actual harness/termination condition.
Quick check: under the same load, capture only three fingerprints: rise/fall symmetry (bin), overshoot/ringing magnitude (bin), and dominant/recessive plateau stability; compare vs non-SIC baseline.
Fix: freeze the load condition, then adjust one control path (symmetry trim or drive vs slew) to restore symmetry first; re-check the worst band only after symmetry is improved.
Pass criteria: worst-band peak reduces by ≥ X dB and symmetry/plateau bins stay within target (e.g., “balanced” bin), with errorFrames ≤ X/min.
BCI test causes bus-off but scope looks OK—what counter/log correlation is first?
Likely cause: RF injection modulates receiver threshold/decision timing intermittently; the time-domain view may look acceptable while the controller accumulates error counters → bus-off.
Quick check: time-align injection frequency/level steps with (1) error counters (TEC/REC or errorFrames), (2) state transitions, and (3) dropouts; identify the first-fail band and first-fail level.
Fix: hold the first-fail band constant and change one variable (slew/drive/symmetry or CM suppression option) while keeping traffic constant; re-run only the first-fail point before sweeping again.
Pass criteria: at first-fail band + margin, bus-off = 0 and errorFrames ≤ X/min over Y minutes, with unchanged test setup parameters.
Different harness vendor changes results dramatically—what parameter should be locked first?
Likely cause: harness geometry changes CM loop area, impedance profile, stub distribution, and shield/return behavior—making emission/immunity data non-comparable.
Quick check: lock “identity fields” before comparing: total length, branch count, stub max length, termination topology, and clamp/probe placement; record them as a versioned harness ID.
Fix: compare only after freezing the same topology/termination; then adjust one knob (slew or symmetry) to re-optimize for the new harness—do not cross-compare un-locked data.
Pass criteria: harness identity completeness = 100% and repeated peak measurements match within ±X dB; immunity first-fail shifts ≤ X dB at fixed settings.
Split termination improves emission but hurts immunity—what midpoint return-path check?
Likely cause: the midpoint network reduces CM emission, but an unfavorable midpoint return path (or parasitic coupling) increases susceptibility during RF injection.
Quick check: keep injection conditions constant and A/B the midpoint capacitor return path (to ground reference) while logging first-fail level and CM current trend around the fail band.
Fix: keep split termination but correct the midpoint return (short, controlled reference) or revert midpoint cap value in one step; retest only the first-fail band before full sweep.
Pass criteria: emission peak improves by ≥ X dB and immunity first-fail level improves by ≥ X dB, with bus-off = 0 over Y minutes.
CMC reduces a peak but creates a new band—what placement/parasitic signature to confirm?
Likely cause: the CMC altered the impedance/CM conversion; parasitics introduced a new resonance (new narrowband peak) tied to placement or the local return path.
Quick check: compare “before/after CMC” with identical analyzer settings and log top-3 peaks (frequency/amplitude); if the new band is narrow and repeatable, treat it as a resonance signature.
Fix: keep one variable: move between “CMC / DNP / alternative CMC value” options (one at a time) and recheck only the new-band peak; avoid simultaneous slew changes.
Pass criteria: original peak reduced ≥ X dB and no new peak exceeds baseline by > X dB in the monitored bands.
Cold start immunity fails, hot passes—threshold drift or edge-rate shift first?
Likely cause: temperature-dependent receiver threshold/offset or edge-rate/drive variation changes decision margin under RF injection.
Quick check: run the first-fail immunity point at two temperatures and log (1) errorFrames (/min) and (2) whether the failure is gradual (counter buildup) or abrupt (dropout/bus-off).
Fix: keep injection constant; adjust one knob that improves decision robustness (drive/symmetry before slowing slew further), then retest cold only; do not re-tune across temps simultaneously.
Pass criteria: at cold condition, immunity passes at level X for Y minutes with bus-off = 0 and errorFrames ≤ X/min.
Only heavy-load nodes fail immunity—drive strength or symmetry trim first?
Likely cause: under heavy load the edge/plateau becomes less predictable, and RF injection pushes the decision over threshold; weak drive or asymmetry amplifies this effect.
Quick check: compare light vs heavy load at the same immunity point; log Vdiff/edge bin (qualitative), CM current trend, and errorFrames rate—look for a margin collapse only at heavy load.
Fix: adjust drive strength first (restore amplitude/edge certainty) while keeping slew constant; if emission worsens, apply CM suppression options next rather than re-slowing edges.
Pass criteria: heavy-load immunity passes at level X for Y minutes and emission peak increase ≤ X dB vs baseline.
Pass on bench, fail in vehicle—what grounding/return-path evidence is most diagnostic?
Likely cause: the vehicle environment changes the CM return path and ground offsets, increasing CM loop area and injection sensitivity compared to the bench setup.
Quick check: capture CM current and worst-band peaks with the same clamp/probe placement rule; compare “bench vs vehicle” deltas and log ground-reference conditions as identity fields.
Fix: do not re-tune knobs blindly; first reproduce the failure with a controlled vehicle-like return condition, then apply one change (termination/CMC/symmetry) and re-check only worst bands.
Pass criteria: vehicle delta vs bench ≤ X dB at worst bands and immunity passes at level X for Y minutes without dropouts.
Tester settings change “improvement”—how to detect an RBW/VBW/averaging artifact?
Likely cause: spectrum analyzer settings (RBW/VBW/detector/averaging) changed the displayed peak without changing the actual emission.
Quick check: re-run the same DUT condition twice: once with the “old settings” and once with the “new settings” (one setting difference); if peak moves with settings, it is an artifact.
Fix: freeze a single settings profile (RBW/VBW/detector/avg) for all A/B decisions; store it as a checksum and block comparisons across profiles.
Pass criteria: A/B comparisons use identical settings 100% of the time and repeatability is within ±X dB at the tracked peaks.
Selective wake false-wake spikes during RF—how to attribute wake source quickly?
Likely cause: RF injection perturbs the wake decision path (bus pattern detect, local pin threshold, or timed wake logic), causing source misclassification or genuine false wake.
Quick check: enforce wake attribution logging (bus/local/timed) with timestamps; run a controlled RF step at the first-fail band and count wake events per source (events/min).
Fix: isolate by one change: tighten the wake filter (bus pattern table) or harden the local threshold path; re-run the same RF step and compare per-source counts.
Pass criteria: false-wake rate ≤ X/hour per node and attribution completeness = 100% across Y RF steps.
Note: Each answer is intentionally evidence-first. If a run is not comparable, it is not a valid improvement—lock identity fields and measurement settings before deciding.
