Every pass/fail must bind to: layer + reference plane + case ID + instrument setup snapshot. Missing any one item makes results non-comparable.

Scope guard: this section defines gates and evidence packages only. Protocol-specific mechanisms belong to the USB/PCIe/HDMI/MIPI sub-pages.

• Reference plane (A/B/C) + fixture/cable identity
• Instrument bandwidth, filters, trigger and clock recovery mode
• Pattern/traffic definition + observation duration (time window)
• Calibration status/date + de-embedding package version (if used)
• Reporting rule: confidence + sample size policy (X / Y / N placeholders)

The eye is a statistical view of amplitude and timing at a defined reference plane. “Looks open” is not a definition; a valid definition names the mask/template, sampling method, and plane.

• Fix plane + fixture identity; lock bandwidth/filter and clock recovery.
• Capture enough UI samples to satisfy the confidence rule (X/Y placeholders).
• If de-embedding is used, attach the package and sanity checks.

• Comparing eyes measured at different planes (connector vs after fixture).
• Silent bandwidth/filter changes when switching probes/instruments.
• Mask fails only under corners, while nominal captures look clean.
• Over-aggressive processing that creates non-physical improvement.

Mask pass at Plane=__, duration T=__, confidence C=__.

Jitter is meaningful only with a stated method and BER point. TJ without “@BER” is incomplete; RJ/DJ splits depend on model assumptions and instrument configuration.

• Lock timebase/reference and recovery settings; record them in the evidence set.
• Specify BER point for TJ and method for RJ/DJ extraction (placeholders).
• Keep filtering/bandwidth consistent across runs; label differences explicitly.

• Comparing TJ values derived from different extrapolation assumptions.
• Changing trigger/recovery modes between runs (procedural drift).
• Ignoring instrument floor (measurement system becomes the bottleneck).

TJ@BER=__ under limit X, method M=__.

BER is a tuple: pattern + duration + detector rule + confidence. “No errors observed” must be reported as a confidence bound, not as BER=0.

• Declare pattern/stimulus, observation time, and detector thresholds.
• Record counters and reset points; make windows comparable across runs.
• Attach confidence policy: X runs / Y minutes / N corners (placeholders).

• Too short a window (rare errors never appear).
• Pattern mismatch across labs (“same BER test” but different stimulus).
• Detector saturation or mis-thresholding under stress.

BER ≤ X with pattern=__, T=__, C=__.

Electrical margin measures headroom in eye/jitter/BER under controlled stimulus. Protocol margin measures headroom in real-case execution under corners. Mixing them hides root cause.

• Sweep one axis at a time (temp/voltage/cable loss/noise); log the cliff point and dispersion.
• Bind margin to a guardband policy for production (placeholders).

• Equating “compliance pass” with “manufacturing safe”.
• Attributing a margin cliff to DUT while the setup floor is limiting.
• Changing multiple axes together (root cause becomes unidentifiable).

Maintain ≥ X electrical margin and ≥ Y protocol margin across N corners.

Scope guard: this section defines metric computation and evidence rules only. Protocol-specific tuning belongs to the dedicated sub-pages.

Scope guard: methodology only (setup control + evidence fields). Protocol-specific details belong to the dedicated USB/PCIe/HDMI/MIPI pages.

• Sampling chain fixed: probe/fixture/cable identities recorded; plane defined for each run.
• Instrument snapshot: bandwidth/filter, trigger mode, recovery mode, and acquisition settings saved.
• Calibration trace: calibration status/date and any deskew/offset steps recorded.
• Repeatability guard: temperature, power state, and cable routing constraints documented.
• Evidence bundle: raw capture pointers + photos + IDs so results remain auditable.

• External reference: improves timebase stability and comparability (when the reference is known-good).
• Golden kit: golden cables + golden fixture + golden peer set for fast A/B isolation.
• Automation: scripted runs reduce operator variance; version control scripts and configs.
• Environmental control: controlled airflow/temperature improves repeatability for thermal-sensitive links.
• Health checks: quick pre/post sanity captures to detect drift within a session.

• Untracked adapters: unknown dongles/headers quietly change bandwidth and reflections.
• Hidden filter changes: switching probes/instruments changes bandwidth/filters without being recorded.
• Trigger drift: different trigger/recovery modes produce non-comparable jitter/eye statistics.
• Post-processing drift: scripts/packages change without versioning (results “move” with software).
• Connector wear: frequent re-mates without a wear log introduces time-varying contact behavior.

• Bandwidth & loading: probe/fixture capacitance and front-end bandwidth reshape edges and overshoot.
• Discontinuities: connectors/adapters/vias introduce reflection timing that can masquerade as jitter or eye closure.
• Reference plane: moving the plane changes the reported metric even when the DUT is unchanged.

Probe ID · Fixture revision · Cable/adapter IDs · Plane label · Mate count · Photo of routing · Any inline attenuator or adapter.

• Pattern trigger: define the trigger source and thresholds; drifting triggers corrupt statistics.
• Multi-channel alignment: record deskew/offset settings so timing comparisons remain valid.

• Recovery mode matters: different modes can shift how jitter is classified and reported.
• Lock & record: treat recovery settings as part of the metric definition and store snapshots.

If results change materially when switching internal vs external reference, the measurement system is likely limiting. Require a reference disclosure and a setup snapshot in the evidence bundle.

• Temperature: allow stabilization time; record ambient and any airflow changes.
• Power: fix power mode and load state; log supply settings and any current limiting behavior.
• Cable routing: define bend radius and keep away from noisy bundles; take a routing photo.
• Grounding: keep return paths consistent; changing ground leads changes noise pickup.
• Fixture pressure/mate count: log mates and fixture pressure/torque when applicable.

Fix plane + snapshot settings → run pre-capture sanity → capture → run post-capture sanity → log mates + photo + IDs → archive raw + snapshots.

Scope guard: methods and trust checks only (no protocol-specific masks, limits, or certification clause details).

• Plane mismatch: pass/fail plane ≠ measurement plane → de-embedding required.
• Fixture dominates: fixture/probe clearly shapes edges or reflections → de-embedding strongly recommended.
• Cross-lab comparison: multiple fixtures/instruments → plane normalization required for comparability.

Comparing results from different fixtures without a declared plane and without a versioned de-embedding package.

• Plane A: instrument measurement plane
• Plane B: fixture end / connector plane
• Plane C: DUT reference plane (the plane where pass/fail is reported)

• Store model source, revision, and conditions (temperature / mate state).
• Bind model to a fixture ID and keep a wear/mate-count log.

• Record tool name/version and package version for every run.
• Archive raw input pointers and output pointers with plane labels.

• Non-physical gain: de-embedded result shows unrealistic improvement → model or plane definition is suspect.
• Phase continuity: abrupt phase/group-delay behavior → plane mismatch or bad fixture model.
• Causality: non-causal signatures → reject the package and re-check the model chain.

Frequent re-mates change contact behavior and invalidate previously “trusted” models. Track mate count and require a quick sanity capture before/after the measurement session.

• Plane label (A/B/C) appears in every chart, metric, and report.
• Fixture/probe/cable IDs recorded; mate count included.
• De-embedding package has a version, tool name, and timestamp.
• Sanity checks completed (no non-physical gain; phase/causality acceptable).
• Results are reproducible across a golden kit (if available), within placeholders (X/Y/N).

Rule: pass/fail must bind to a declared plane. Plane-free results should be treated as non-comparable.

Scope guard: stimulus selection motivations and failure signatures only. Protocol-specific named patterns and clause details belong to the dedicated standard pages.

• Role: stable, repeatable stimulus for channel + receiver tolerance comparisons.
• Best observables: eye opening, bathtub trends, BER confidence vs time window.
• Common blind spots: worst-case density corners and system-side behaviors driven by real traffic.
• Failure signatures: good PRBS margin but failures appear only under corner transitions or real workload.

• Role: force worst-case conditions (dense transitions, long runs, slow variation) to reveal hidden margins.
• Best observables: worst-eye corner, jitter decomposition stability, sensitivity to slow variations.
• Common blind spots: full system behavior (buffers, thermal/power coupling) unless paired with traffic.
• Failure signatures: only specific stress classes fail; errors cluster around corner densities or slow drift windows.

• Role: reproduce real workload interactions (firmware timing, buffering, power/thermal coupling).
• Best observables: error bursts vs workload, correlation to temperature/power states, repeatability across runs.
• Common blind spots: not guaranteed to hit the strict worst-case electrical corners unless designed to.
• Failure signatures: PRBS passes but real workload fails; errors track with load steps or long-duration drift.

• Transition density: include baseline + corner density classes.
• Low-frequency content: include at least one class that stresses long-run behavior.
• Slow variation: include a class that probes long-window sensitivity (timebase/reference stability).
• System realism: include a traffic window (with controlled overlays if needed).

Every report must declare the tested categories and explicitly list any missing categories as “coverage gaps.” Untested categories should not be implied as passing.

Scope guard: workflow and artifacts only. Protocol-specific masks/limits are not included here.

• Plane label · stimulus class · objective
• Instrument setup snapshot · trigger/recovery/timebase settings
• Temperature · power state · firmware/build ID
• Probe/fixture/cable IDs · adapter IDs · mate count
• Raw capture pointer/hash · processed pointer/hash · package/script version

• Within-setup repeat: 3 consecutive runs with the same chain (short-term stability).
• Sensitivity checks: swap cable / swap port / swap fixture to isolate chain sensitivity.
• Operator variance: a second operator follows the same SOP using the same snapshot.
• Regression trigger: if results differ beyond placeholders (X/Y/N), return to Step 2/3/6 before concluding.

Scope guard: universal margin knobs and measurement logic only. Protocol-specific tool names, register controls, or clause limits are not listed here.

• Most sensitive axis: which knob causes the fastest loss of BER/eye/jitter margin.
• Knee point: the transition boundary between safe and unstable behavior.
• Outliers: rare samples that fail early (often the true production risk).
• Guardband gate: a documented buffer to absorb spread, drift, and aging (X/Y/N placeholders).

• Stresses: headroom, noise coupling, threshold margins.
• Primary observables: BER bursts, eye height collapse, jitter floor shifts.
• Failure signature: “sudden” error onset near a knee point when supply droops or noise increases.
• Pass placeholder: stable BER/eye/jitter margin within X over Y windows.

• Stresses: drift, gain/offset shifts, timing drift, impedance changes.
• Primary observables: slow margin drift, knee point movement, repeatability degradation.
• Failure signature: “works cold / fails hot” (or reverse) with a moving margin boundary.
• Pass placeholder: margin stays above X across Y temperature points.

• Stresses: equalization boundaries, sensitivity to tuning and training outcomes.
• Primary observables: eye width vs setting, error-rate sensitivity, jitter decomposition stability.
• Failure signature: sharp boundary between “one click works” and “one click fails.”
• Pass placeholder: margin maintained for X setting steps around nominal.

• Stresses: insertion/return loss, reflections, crosstalk, connector mating variance.
• Primary observables: eye closure with length, BER vs channel class, outlier detection by cable batch.
• Failure signature: passes on short bench cable; fails on longer/poorer channel classes.
• Pass placeholder: maintain margin above X through Y channel classes.

• Stresses: sensitivity to long-window behavior and reference/timebase stability.
• Primary observables: long-window BER trend, jitter floor movement, extrapolation sensitivity.
• Failure signature: margin “moves” when capture window/timebase choices change.
• Pass placeholder: stable margin within X over Y long windows.

• Stresses: robustness to external perturbations (electrical noise paths).
• Primary observables: BER bursts, outlier amplification, jitter floor deterioration.
• Failure signature: narrow knee point + outliers expand under injected disturbance.
• Pass placeholder: no out-of-family behavior beyond X under Y injection levels.

• Ranked sensitivity axes (which knob dominates).
• Knee points and “safe zone” boundaries.
• Outlier list with evidence bundle references.

• Guardband declaration (X/Y/N placeholders) tied to evidence, not single-point pass.
• Retest triggers when drift/outliers appear in regression.
• Minimal “must-hold” conditions for stable operation across spread and aging.

Scope guard: no EDID field explanations and no HDCP mechanism walkthrough. Only case organization and acceptance criteria placeholders.

• Case: short name that uniquely identifies the scenario.
• Preconditions: topology, firmware/build, cable class, power/temperature state.
• Trigger: the action that starts the case (event, hot-plug, injection, topology change).
• Expected: what “correct” looks like (observable behavior, not internal implementation).
• Evidence / Logs: timestamps, captures, dumps, snapshots (evidence ladder).
• Pass criteria: X/Y/N placeholders tied to the expected outcome.
• Failure signature: how the failure presents (useful for fast triage).

• Read path: read succeeds and results remain consistent across repeated reads.
• Abnormal content: missing/corrupted/boundary content handling (concept-level).
• Hot-plug re-read: re-read behavior remains stable under repeated connect/disconnect cycles.
• Evidence: capture + timestamp + raw blob reference (no field interpretation).
• Pass criteria: X/Y/N placeholders (consistency rate, retry limit, stability window).

• Authenticate / re-authenticate: stable behavior across repetitions and topology states.
• Failure injection: controlled disturbances to confirm expected fallback/behavior (concept-level).
• Topology variations: different peer/switch configurations with consistent acceptance outcomes.
• Evidence: event logs + timestamps + capture references + peer identification (concept-level).
• Pass criteria: X/Y/N placeholders (stability window, retry behavior, no persistent lock-up).

• Level 1: pass/fail result + timestamps.
• Level 2: key event logs + peer identification + execution snapshot.
• Level 3: captures + dumps + configuration IDs (auditable reproduction).
• Level 4: evidence bundle manifest aligned with the workflow SOP (settings/raw/processed/report).

Scope guard: universal interop strategy only. No protocol mechanism walkthroughs and no compliance clause references.

• Peer variability: different firmware, defaults, and tolerance boundaries.
• Environment coupling: cable class, adapter behavior, powering method, and layout/grounding variance.
• Coverage gap: tests focused on “nice-looking” combinations rather than “most demanding” ones.

• Bucket A · Mass-market peers: represents the most common field counterparts.
• Bucket B · Strict peers: tends to expose tight tolerance boundaries and fragile edges.
• Bucket C · Edge peers: different generations/topologies/powering methods to probe corner behavior.

Peer ID · firmware/build ID · mode · cable class · adapter class · powering method · environment note · result + evidence pointer.

1. Define factors: device · firmware · cable · adapter · mode · powering · temperature.
2. Define levels: short/typical/worst cable classes, normal/stress modes, bus/self powering.
3. Pairwise coverage: ensure every pair of factors is exercised at least once (concept-level).
4. Risk-weighted must-run: add cases that target known sensitive axes and prior failure signatures.

• Minimal coverage set: smallest set that still meaningfully spans the space.
• Regression set: frozen, repeated set for continuous stability tracking.

Scope guard: evidence packaging and report structure only. Measurement SOP details are referenced conceptually, not repeated.

• Environment: temperature point(s), powering method, physical setup notes.
• Versions: DUT firmware/build ID, peer firmware/build ID, host/driver ID, config ID.
• Fixture: fixture revision, reference plane statement (concept-level), connection notes.
• Calibration: instrument calibration state and timestamps (concept-level).
• Raw pointers: raw data IDs/paths (where the original measurements live).
• Processing: script version + parameters + output identifiers (concept-level checks).
• Figures: eye/mask results, jitter statistics, BER + confidence view.
• Decision: pass/fail + margin + explicit “not covered” risks.

Every pass/fail statement must point to raw evidence and processing identifiers. If a risk was not covered, it must be listed explicitly.