HDMI / DisplayPort Tx/Rx Redriver & Retimer (HDCP/EDID, Compliance)
← Back to:Interfaces, PHY & SerDes
This page helps HDMI/DisplayPort links stay stable at high data rates by turning “black screen / flicker / snow / training & HDCP/EDID failures” into actionable, measurable steps—budget the channel, pick redriver vs retimer, and verify with pass/fail criteria.
It focuses on extension/repair/retiming scenarios across Main Link/TMDS/FRL and control-plane (HPD, DDC/AUX), so teams can debug faster and lock production using logged fields and threshold-based acceptance.
Definition & Page Boundary (HDMI/DP Tx/Rx Redriver/Retimer)
This page helps you
- Extend HDMI/DP links with redriver/retimer placement and practical EQ/retiming rules.
- Debug HDCP/EDID/HPD (and DP AUX) failures that masquerade as “signal integrity”.
- Build an evidence chain for eye/jitter compliance and bring-up readiness.
This page does NOT cover
- USB-C Type-C/PD/Alt-Mode policy and port controller flows (link out only).
- General CDR/equalization theory or math derivations (reference to the CDR/EQ hub page).
- ESD/TVS fundamentals; only placement and parasitic impact relevant to HDMI/DP links.
Definitions (HDMI/DP-specific, engineering-facing)
Redriver (linear re-drive)
- Role: reshape channel loss with gain/CTLE/FFE (no clock recovery).
- Best when: loss is moderate and jitter is mostly ISI-driven.
- Trap: “bigger eye” can still worsen BER if noise is amplified or EQ is over-peaked.
Retimer (CDR + re-transmit)
- Role: recover clock/data and re-launch, breaking jitter/noise accumulation.
- Best when: loss is high, channel is harsh (dock/backplane), or jitter must be re-timed.
- Trap: lock/re-lock behavior and latency can trigger visible “blink” and control-plane re-auth cycles.
Repeater / Mux / Switch (topology shapers)
- Repeater is ambiguous in vendor docs; in this page it must resolve to redriver or retimer.
- Mux/Switch changes loss/return-loss and may impact HDCP/EDID/HPD/AUX transparency.
- Trap: control-plane faults can look like SI faults; triage starts with HPD/EDID/AUX before eye tuning.
In-scope (must be solved here)
- HDMI TMDS/FRL + DP Main Link extension using redriver/retimer.
- Control-plane robustness: HDCP/EDID/HPD and DP AUX pass-through/debug.
- Practical EQ/retiming knobs tied to eye/jitter and margin evidence.
- Bring-up/compliance proof: test points, de-embed plan, logs and pass criteria (threshold placeholders).
Out-of-scope (link only, no deep dive)
- USB-C port policy, PD negotiation, and Alt-Mode discovery state machines.
- Generic CDR math, equalizer derivations, or SERDES training theory beyond actionable knobs.
- Full protocol spec commentary; only engineering checks that affect link stability and compliance.
- Standalone ESD/TVS selection theory (only parasitic impact + placement here).
Cross-links (avoid duplication)
- USB-C / Type-C & Bridges (dock-focused system pages).
- CDR / Retimer & Equalization (generic knobs and measurement artifacts).
- High-speed ESD/TVS arrays + CM chokes (parasitics, placement, matching).
- Protocol bridges & format conversion (if HDMI↔DP or HDMI↔MIPI conversion exists).
Minimum verification loop (do not skip)
Functional closure
Identify/handshake succeeds: EDID readable, training completes (DP), HDCP stable (if enabled), no repeated hotplug cycles.
Margin closure
Target mode holds across cable variance and temperature. Record negotiated rate/lane, error counters, and drop events (thresholds: TBD).
Compliance closure
Measurement setup is reproducible (reference plane + de-embed plan). Save eye/jitter artifacts and logs as evidence for reviews.
Scope map: where redrivers/retimers sit, and which paths are “main link” vs “control-plane”.
System Topologies & Where the Problem Appears
Topology determines whether failures are driven by main-link margin, link training (DP), or control-plane (HDCP/EDID/HPD/AUX). Fast triage prevents “eye tuning” from being applied to EDID/AUX problems.
Looks like physical-layer margin
- Only fails at higher modes; lower modes are stable.
- Sensitive to cable length, temperature, or connector batch.
- First check: negotiated mode + error counters (thresholds: TBD).
Looks like DP link training
- Training repeats; final state drops to a lower rate/lane count.
- AUX/DPCD readbacks show retries or timeouts.
- First check: training log + DPCD status snapshots (fields: TBD).
Looks like control-plane instability
- EDID read fails or HDCP re-auth loops even at low modes.
- Hot-plug changes the outcome; “replug fixes it”.
- First check: HPD bounce + EDID success rate + HDCP retries (thresholds: TBD).
First 5 probes / logs (fastest to differentiate root cause)
- HPD event log: count + timestamps; flag any bounce or repeated toggles (pass: TBD).
- EDID/AUX success rate: retry count, timeout rate (pass: TBD).
- Negotiated mode snapshot: HDMI mode or DP rate/lane count; record any downshift.
- Error counters: CRC/BER-like indicators where available; correlate with temperature and cable ID.
- Training/handshake trace: DP training steps or HDCP auth phases; store for regression comparisons.
Topology gallery (trigger → failure mode → first check → fix hint)
Direct attach
- Trigger: high modes compress margin; small layout defects become dominant.
- Failure: only top mode fails; lower mode is clean.
- First check: negotiated mode + repeatability across temperature.
- Fix hint: prefer layout/return-loss cleanup; redriver only if loss is the limiter.
Cable extension
- Trigger: insertion loss + cable variance; EMI/ESD exposure increases.
- Failure: snow/sparkles or intermittent dropouts; “one cable works”.
- First check: cable ID vs error trend; mode downshift behavior.
- Fix hint: redriver for moderate loss; retimer if loss is high or jitter accumulates.
Dock / Panel / Backplane
- Trigger: stacked connectors/vias create return-loss and reflection hotspots.
- Failure: DP training loops; rare drops after warm-up.
- First check: AUX/DPCD stability and reference-plane correctness.
- Fix hint: retimer is common; verify control-plane transparency before tuning EQ.
KVM / Matrix / Multi-display
- Trigger: mux/switch adds loss and can destabilize HPD/EDID/HDCP timing.
- Failure: hot-plug dependent; HDCP/EDID failures dominate.
- First check: HPD bounce + EDID read success + HDCP re-auth count.
- Fix hint: stabilize control-plane first; then address main-link margin.
Topology gallery: small structural differences that change the first debug step.
Link Basics You Must Not Get Wrong (HDMI vs DP)
Before selecting a redriver or retimer, identify the main high-speed bundle and the control-plane. Control-plane instability (EDID/HDCP/HPD/AUX) can look like “signal integrity” but requires a different first fix.
HDMI (TMDS vs FRL)
- TMDS: clock/data relationship is explicit; mode downshift often stabilizes quickly.
- FRL: higher bandwidth compresses margin; connector/via/cable variance becomes dominant.
- Control-plane: DDC + HPD (+ HDCP phase). If these flap, do not start with EQ.
DisplayPort (Main Link + AUX)
- Main Link: rate and lane count define high-speed stress and EQ demand.
- Training: repeats/downshift are direct margin signals; logs are first-class evidence.
- AUX/HPD: sideband failures frequently present as training failures.
Three non-negotiable rules
- If EDID/HDCP/HPD/AUX is unstable, treat it as control-plane first.
- DP training downshift is a margin signal; capture the negotiated state.
- Higher modes compress eye/jitter budget; topology variance becomes visible.
Rate / lanes / control-plane cheat sheet (verify per spec)
| Mode family | Main-link bundle | Device-choice drivers | Control-plane to not forget | Typical failure signature |
|---|---|---|---|---|
| HDMI · TMDS |
TMDS lanes + clock (verify per spec) |
Loss window, EQ range, connector/via RL sensitivity | DDC (I²C), HPD, HDCP phase stability | High-mode fails, low-mode stable; cable length sensitive |
| HDMI · FRL |
Multi-lane high-speed bundle (verify per spec) |
Higher stress → tighter eye/jitter margin; redriver window narrows; retimer may be needed | HPD + HDCP re-auth sensitivity; DDC stability across hot-plug | Top mode fails; intermittent blink; “eye looks bigger but BER worsens” if EQ is over-peaked |
| DP · Main Link |
Lane count + link rate (verify per spec) |
Training outcome, EQ/retiming strategy, channel loss and RL hotspots | AUX + HPD must be reliable; DPCD reads are first debug evidence | Training loops or downshift; AUX timeouts masquerade as SI failures |
| DP · AUX (sideband) |
Not a high-speed lane (verify per spec) |
Pass-through integrity, ESD/EMI robustness, timing margin for handshakes | AUX retries/timeouts drive training instability | “Only replug works”; training never converges until AUX stabilizes |
Note: values and exact lane definitions vary by revision; treat this table as a selection and debugging structure, and verify details per the applicable HDMI/DP specifications.
Signal bundle diagram: thick lines are main lanes; thin lines are control-plane paths that often fail first.
Channel Loss & Eye Closure: Budgeting the Link
Link extension succeeds when the channel budget is quantified. The goal is to separate loss-driven eye closure from control-plane instability and to prevent measurement artifacts from producing false “improvements”.
Insertion loss (IL)
- Effect: high-frequency attenuation shrinks vertical opening.
- Signature: top modes fail first; cable length sensitive.
- Leverage: EQ window and placement; retimer if loss is beyond linear repair.
Return loss (RL) / reflections
- Effect: ringing and pattern-dependent errors.
- Signature: certain cables/connectors are always “bad”.
- Leverage: connector/via cleanup; avoid stubs; verify reference plane.
Crosstalk (XT) / noise coupling
- Effect: BER degrades even if the eye “looks fine”.
- Signature: worsens with nearby aggressors or power noise.
- Leverage: spacing/return paths; filtering; avoid over-peaked EQ that amplifies noise.
Discontinuities (vias/stubs)
- Effect: localized reflection + common-mode conversion.
- Signature: EMI increases; sporadic failures after integration.
- Leverage: backdrill/stackup choices; placement discipline around connectors.
Link budget template (copy/paste structure)
| Item | Value / placeholder | Evidence / measurement plan | Pass criteria |
|---|---|---|---|
| Target mode (rate / lanes) | TBD (verify per spec) | Negotiated state snapshot; training logs if DP | No downshift; no drops (TBD) |
| Allowable loss window (IL/RL) | IL@fN: TBD, RL: TBD | Reference plane + de-embed plan (TBD) | Meets mask/jitter limits (TBD) |
| Segment contributions (waterfall) | Conn A: TBD · Cable: TBD · Conn B: TBD · PCB/vias: TBD | VNA/S-params or fixture plan (TBD) | Segment limits met (TBD) |
| Margin reserve (variance) | Cable batch: TBD · Temperature: TBD · Process: TBD | Multi-cable + temp sweep logs | No failures across sweep (TBD) |
| Control-plane stability (must be clean) | EDID/AUX success: TBD · HPD bounce: TBD · HDCP retries: TBD | Event counts + timestamps | Stable handshake (TBD) |
Measurement traps (false “improvements” to eliminate)
Symptom
Eye/jitter looks “better” after moving the reference plane or changing fixture settings.
Quick check
Re-run with the same plane + same de-embed file; compare to a known-good cable/fixture baseline.
Fix
Lock measurement settings; document the reference plane; treat the de-embed plan as part of the spec.
Pass criteria
Same DUT + different fixtures converge within agreed correlation limits (TBD), and results are reproducible.
Loss budget waterfall: segment-by-segment loss accumulation and shrinking eye opening.
Redriver vs Retimer: Decision Rules
A redriver is a linear channel conditioner (gain/CTLE/FFE) that improves eye shape but does not recover the clock. A retimer uses CDR + re-transmit to re-time the stream and reduce channel-related jitter transfer, at the cost of added latency and re-lock behavior.
When redriver is usually enough
- Channel loss is within a linear EQ window (budget margin remains).
- Failures track with loss slope, not with re-lock events.
- System latency budget is tight; re-lock must be avoided.
When retimer is usually justified
- Loss/ISI is beyond what linear EQ can recover reliably across variance.
- Jitter/BER strongly correlates with channel segments and connectors.
- A clean re-timed hop is needed for long cables or multi-hop topologies.
Selection evidence (collect first)
- DP: negotiated rate/lane, training attempts, downshift counts (TBD).
- HDMI: HPD bounce, EDID read success, HDCP retry counts (TBD).
- Across cables: batch-to-batch variance and temperature sweep stability.
Side-by-side comparison (selection impact)
| Dimension | Redriver (linear) | Retimer (CDR + re-TX) | What it means for selection |
|---|---|---|---|
| Clock recovery | No (passes jitter) | Yes (re-times) | If channel-driven jitter dominates, retimer can cut transfer (TBD limits). |
| EQ range | Limited by linear stability / noise gain | Wider tolerance across loss/ISI variance | If the budget waterfall shows “dominant loss segment”, retimer is more robust. |
| Latency | Minimal | Added (TBD) | If UI/UX switching is sensitive, budget latency + re-lock time explicitly. |
| Re-lock behavior | None (no CDR) | Present (lock time TBD) | Hot-plug / power events may look like “blink”; lock robustness must be verified. |
| Protocol transparency | Often simpler pass-through | Must verify handshake/training behavior | Confirm EDID/HDCP (HDMI) and AUX/training (DP) are stable through the hop. |
Keep the decision evidence close to the selection: negotiated states, retry counts, and variance sweeps are often more predictive than a single eye capture.
Decision flow (evidence-driven)
- Control-plane stable? (EDID/HDCP/HPD for HDMI; AUX/HPD/DPCD for DP) → If not, fix control-plane first.
- Is channel loss beyond the linear EQ window? → If yes, retimer is favored.
- Is BER/jitter strongly correlated with channel segments? → If yes, retimer can cut transfer.
- Must pass training/HDCP transparently? → Verify hop behavior before committing.
- Can latency + re-lock time be tolerated? → If not, redriver is preferred.
Two-lane pipeline compare: same channel, different hop strategy.
Equalization & Retiming Knobs (CTLE/FFE/DFE/Peaking/Loop BW)
Treat knobs as a diagnostic tool, not as a tuning ritual. The same eye can hide very different BER outcomes. The table below maps each knob to typical symptoms, safe fix direction, and placeholder pass criteria.
Linear EQ knobs (redriver)
- CTLE peaking: compensates HF loss, but can amplify noise.
- Gain/slope: improves vertical opening; watch noise gain.
- FFE taps: shapes precursor/postcursor ISI (device-specific).
Retiming knobs (retimer)
- CDR loop BW: trades jitter filtering vs tracking ability.
- Lock time: impacts blink behavior on events (TBD).
- Jitter transfer: verify with the target channel variance.
Two archetypal mistakes
- Over-EQ: eye grows but BER worsens (noise gain / overshoot).
- Under-EQ: high modes fail first; training downshifts.
- Wrong loop BW: stable on bench, drops in system events.
Knob → symptom → fix direction → pass criteria (TBD)
| Knob | Symptom (field) | Fix direction (safe) | Pass criteria (placeholder) |
|---|---|---|---|
| CTLE peaking | High mode fails; DP downshift; “snow” at top resolution | Increase gradually until margin improves; stop if noise artifacts rise | No downshift; stable image; error counters flat (TBD) |
| Gain / slope | Eye height low; link marginal across cables | Raise gain cautiously; validate BER (do not trust eye alone) | BER meets target; no intermittent blinks (TBD) |
| FFE taps | Pattern-dependent errors; sensitivity to cable type | Adjust taps to reduce ISI; re-check across variance set | Stable across cable batch + temperature sweep (TBD) |
| Over-EQ indicator | Eye looks larger but BER worsens; overshoot/EMI grows | Reduce peaking/gain; prioritize BER and stability over “pretty” eyes | Error counters improve; EMI artifacts do not regress (TBD) |
| CDR loop BW | Stable on bench, drops in system events; periodic unlocks | If unlocks track events, widen tracking; if noise dominates, tighten filtering (device-specific) | No unlocks; negotiated state stays fixed (TBD) |
Knob-to-symptom matrix: minimal text, iconized symptoms.
HDCP / EDID / HPD / AUX/DDC: Control-Plane That Breaks the Link
Many “main-link” failures are actually control-plane instability. Before touching EQ or retiming, verify the presence and stability of 5V/HPD, the integrity of DDC/EDID (HDMI), and the reliability of AUX/DPCD transactions (DisplayPort).
Troubleshooting priority (control-plane first)
- Power presence: HDMI 5V present and stable (no drops, no bounce).
- HPD stability: HPD asserted cleanly; count bounce events (TBD) across hot-plug and motion.
- DDC/EDID (HDMI): EDID read succeeds consistently; track ACK/retry/timeout (TBD).
- AUX/DPCD (DP): no AUX timeout; DPCD reads/writes are reliable during training.
- Main link last: only after the above are stable, tune EQ/retiming and evaluate eye/jitter.
EDID / DDC symptoms (HDMI)
- EDID read fails: no ACK / timeout; often looks like “no signal”.
- EDID read wrong: intermittent capability mismatch; unstable modes.
- Hot-plug sensitive: depends on HPD/5V timing window.
Practical DDC checks (no I²C theory)
- Log: EDID success rate, retry count, and timeout count (TBD fields).
- Verify pull-ups and ESD capacitance do not slow edges excessively (TBD).
- If long DDC wiring exists, consider a dedicated buffer/switch (HDMI-only scope).
HDCP failures (transparency traps)
- HPD bounce → re-auth: short dips can trigger repeated auth.
- Sideband handling: verify pass-through behavior through switch/retimer.
- Symptom fingerprint: stable at low mode, fails at high mode (TBD).
DisplayPort AUX focus
- AUX timeout: training loops and downshifts mimic main-link issues.
- DPCD reads/writes: errors often correlate with ESD/cap load (TBD).
- Priority: if training repeats, confirm AUX/HPD before EQ tuning.
Handshake timeline (event blocks): hot-plug → control-plane → link stable.
Compliance: Eye / Jitter / Mask & Test Setups (CTS/ATC Thinking)
“The system works” is not the same as “the link is compliant.” Compliance requires repeatable measurements, consistent reference planes, and evidence packages that survive cross-lab correlation.
Turn specs into steps (setup-first)
- Pick the reference plane: define where “compliance” is evaluated (connector / fixture / coupon) (TBD).
- Describe fixture & interconnect: fixture type, cable, adapters, and exact stack-up.
- De-embed responsibly: specify model source and confirm before/after sanity checks.
- Lock down key settings: trigger/clock recovery, RBW/VBW, and CTLE preset consistency.
- Build an evidence pack: screenshots + report fields + training logs for repeatability.
Setup traps (common)
- Reference plane drift: results become incomparable across runs.
- De-embed mismatch: “improvement” can be a modeling artifact.
- Preset inconsistency: CTLE/receiver presets change mask outcome.
Settings to document (only pitfalls)
- RBW/VBW: can reshape apparent jitter/noise (settings artifact).
- Trigger/CR: wrong recovery hides real failures.
- Preset/CTLE: must match the compliance target profile.
Evidence pack (fields)
- Eye: mask hit status, sweep conditions, preset ID (TBD).
- Jitter: RJ/DJ summary, RBW/VBW, CR mode (TBD).
- Logs: training attempts, downshifts, AUX/DDC counters (TBD).
Compliance evidence flow: DUT → fixture → scope/BERT → report (with de-embed and reference plane).
Board Design & Layout: Return Current, Vias, Connectors, ESD/EMI
For HDMI/DP, link margin is often lost at the connector launch, via transitions, and protection placement. The most common “bench OK, system fails” pattern is caused by return-path discontinuity, stubs/reflections, or parasitics from ESD/CM components.
Differential pairs (HDMI TMDS/FRL, DP Main Link)
- Impedance: match the stackup target (TBD) and keep geometry stable through launch.
- Length symmetry: control skew, but never sacrifice return-path continuity to chase perfect length.
- Reference continuity: avoid crossing split planes; provide a nearby return bridge at layer changes.
Vias & connector launch
- Stubs: long via stubs create notches and reflections; mitigate with backdrill (TBD).
- Via fence: add ground via stitching around the launch to keep return currents local.
- Fanout: keep the first centimeters clean—no sudden width/spacing jumps, no sharp bends.
ESD arrays & CM choke (placement + parasitics)
- Place near the connector: keep branch length minimal to avoid loading the main link.
- Parasitic C: excessive capacitance closes the eye and can break training at high modes.
- Matching: imbalance converts differential to common-mode → EMI up, margin down.
Power noise → jitter (what to measure first)
- Correlate events: flicker/downshift aligned with load/fan/USB changes suggests supply coupling.
- Probe near the IC: check Tx/retimer rails at the pins (TBD bandwidth/probe method).
- Avoid artifacts: poor probing makes ground-bounce look like rail noise.
Layout acceptance checklist (HDMI/DP-focused)
| Item | Why it breaks links | Quick check | Fix action | Pass criteria |
|---|---|---|---|---|
| Return path continuity | Forces current detours → common-mode + EMI + jitter | No plane split crossings under main link | Add return bridges / re-route to continuous reference | TBD (layout review sign-off) |
| Via stub control | Creates reflection notches at high modes | Identify long stubs at connector/retimer transitions | Backdrill / shorten / avoid unused via barrels | TBD (eye/margin stable) |
| Connector launch symmetry | Mode conversion and impedance steps collapse eye | Launch region is short, symmetric, and fenced | Tight geometry control + ground stitching | TBD (training stable across cables) |
| ESD / CM placement | Parasitic C loads the link; mismatch drives EMI | Placed near connector; branch length minimal | Move closer, reduce branch, choose lower C (TBD) | TBD (mask/jitter margin) |
| Rail noise correlation | Supply ripple modulates analog front-end/PLL → jitter | Event time aligns with rail noise bursts (TBD) | Decoupling/filters/layout return fixes before EQ rework | TBD (repeatable across load/thermal) |
Layout placement blueprint: connector → protection/magnetics → redriver/retimer → SoC/PHY, with return-path rules.
Bring-up & Debug Playbook (Symptom → Probe → Fix → Pass)
Debug faster by enforcing a strict layer order: Power/HPD → EDID/AUX → Training → Main link eye/jitter. Each symptom below is mapped to a minimal probe set, a fix action, and a pass criterion (TBD thresholds).
Debug ladder (always in this order)
- Power / HPD: HPD bounce count (TBD), 5V present stability.
- EDID (HDMI) / AUX+DPCD (DP): success rate, timeout counters.
- Training / Negotiation: attempts, downshift events, relock loops.
- Main link: eye/mask margin and jitter evidence, then EQ/retiming knobs.
Only fails at high resolution / refresh
Likely cause
Channel loss exceeds the effective EQ range, or jitter margin collapses at higher modes.
Quick check
Check training downshift count (DP) or mode fallback (HDMI) and capture eye at a consistent plane (TBD).
Fix
Reduce launch/stub loss, adjust EQ presets conservatively, or insert a retimer when channel-driven jitter dominates.
Pass criteria
No repeated training loops; stable mode hold across cables and temperature (TBD margin + log evidence).
Likely cause
Via-stub notch or connector launch asymmetry becomes critical at the top rate.
Quick check
Compare behavior across two cable lengths and one alternative connector path; look for rate-specific failure.
Fix
Shorten stubs (backdrill/TBD), improve launch fencing, and avoid abrupt geometry changes near the connector.
Pass criteria
High mode runs continuously with zero visual artifacts and stable error counters (TBD).
Only fails on hot-plug / switching
Likely cause
HPD bounce or a timing race during EDID/AUX transactions triggers re-auth or retraining loops.
Quick check
Record HPD edges and count bounces during plug-in; log EDID/AUX timeout counters (TBD).
Fix
Improve HPD filtering/termination and ensure the control-plane path is not overloaded by ESD parasitics.
Pass criteria
Plug/unplug cycles succeed repeatedly with zero re-auth loops and stable mode negotiation (TBD).
Likely cause
Retimer re-lock window or switch path settling causes transient control-plane failures.
Quick check
Measure time-to-stable after switching; correlate failures with the first EDID/AUX attempt (TBD).
Fix
Enforce a settle delay in system sequencing and validate sideband pass-through behavior end-to-end.
Pass criteria
Switching and hot-plug succeed across repeated trials without intermittent artifacts (TBD).
Only fails on long cable / certain cable batches
Likely cause
Return loss and reflections accumulate; common-mode noise increases with cable/connector variance.
Quick check
Compare training stability and error counters across two cable models and lengths; note downshift patterns.
Fix
Improve connector launch + return path, reduce discontinuities, and apply compatible EQ presets (avoid over-EQ).
Pass criteria
Stable operation across a defined cable set with margin evidence and no recurring training loops (TBD).
Likely cause
Protection parasitics (ESD/CM choke) differ across revisions and reduce high-frequency margin.
Quick check
A/B test a board variant or bypass path (TBD) and watch for training success rate changes.
Fix
Re-select lower-capacitance protection and minimize branch stubs at the connector.
Pass criteria
Cable-to-cable variability stays within margin budget (TBD) with repeatable evidence logs.
Intermittent blink / snow / drop
Likely cause
Control-plane timeouts or power-noise coupling causes periodic retraining or re-auth events.
Quick check
Correlate symptoms with HPD edges, EDID/AUX counters, and rail noise bursts near the IC (TBD).
Fix
Stabilize HPD/control paths, improve supply decoupling and return paths, then retune EQ/retiming only if needed.
Pass criteria
Long-run stress test with zero events and preserved evidence pack (eye/jitter/logs) (TBD duration).
Likely cause
Over-EQ improves eye height but increases BER due to noise amplification and overshoot.
Quick check
Reduce peaking/DFE strength one step and compare error counters and stability (TBD).
Fix
Tune EQ to the minimum that meets training and stability; prefer channel fixes over aggressive peaking.
Pass criteria
Stable operation with consistent preset ID and improved BER margin (TBD) across cables.
Debug decision tree: symptom → probe point → fix action (box-diagram, minimal text).
theme: bg #0b1120, card-border #1f2a44, text #e5e7eb, mutetext #9ca3af, accent #38bdf8
H2-11 · Engineering Checklist (design → bring-up → production)
This checklist is optimized for HDMI/DisplayPort link extension and repair using redrivers/retimers/switch-with-EQ. Every line is written as a doable action with acceptance placeholders and record-field placeholders to enable repeatable engineering and production gating.
Design
Budget → Placement → Protection → Config
-
Freeze the target mode set (HDMI TMDS vs FRL; DP RBR/HBR/HBR2/HBR3) and map them to a “worst-case” channel condition.
Accept: [Target modes pass] · Log:
MODE_SET, MAX_RATE, LANE_COUNT -
Choose the “first-fix component class” by the link-budget boundary:
If loss/margin is moderate → Redriver; if jitter accumulation / clock recovery needed → Retimer; if multi-source routing is required → Switch-with-EQ.
Accept: [Class locked] · Log:
FIX_CLASS, LOSS_EST_DB, JITTER_SUSPECT -
Lock placement order at the connector (recommended): Connector → ESD/TVS → (optional CM choke) → Redriver/Retimer → SoC/PHY.
Accept: [No stubs, return path continuous] · Log:
PLACEMENT_SEQ, RETURN_PATH_OK -
Control-plane integrity rule: HPD + DDC/AUX must be treated as “first-class” links (ESD choice + pull-ups + routing) before touching Main Link/TMDS/FRL EQ.
Accept: [No HPD bounce / no DDC/AUX timeouts] · Log:
HPD_EVENTS, DDC_ERR, AUX_ERR -
Pre-select “known-good reference parts” for schematic bring-up (examples; verify package/speed/availability):
Accept: [Ref-BOM frozen] · Log:HDMI redriver / retimer examples• TMDS1204 / TDP1204 (HDMI 2.1 FRL-capable redriver family)
• TMDS181 (HDMI 2.0 TMDS retimer), TMDS171 (HDMI 1.4b TMDS retimer)
• IT66313 / IT66319 (HDMI 2.1 retimer family; retiming buffer)DP redriver / retimer examples• TDP142 (DP 1.4 HBR3 linear redriver; AUX/HPD-aware)
• ANX7496 / ANX7497 (DP 1.4 retimer family)
• PS8461 / PS8461E (DP mux + retimer class device)
• PI3DPX1207C / PI3DPX1207B1 (DP1.4 linear redriver class)Control-plane + protection examples• TPD12S016 (HDMI companion: DDC level shifting + load switch + ESD)
• TPD4E05U06 / TPD4E05U06-Q1 (ultra-low-C ESD arrays)
• RClamp0524P (low-capacitance TVS array class)
• DLW21SN261XQ2L (Murata DLW21 series common-mode choke example)REF_BOM_REV, IC_PN_LIST, ESD_PN_LIST, CMC_PN_LIST
Bring-up
HPD → DDC/AUX → Training → Main Link
-
Control-plane first gate: verify HPD stability + EDID read + DP AUX transactions before tuning any EQ knob.
Accept: [0 unexpected HPD toggles] · Log:
HPD_TOGGLE_CNT, EDID_RD_OK, AUX_TXN_OK -
Preset/EQ sweep with evidence: sweep CTLE/FFE (redriver) or loop/EQ presets (retimer) and record “pass window”, not just “one lucky setting”.
Accept: [Pass window exists] · Log:
PRESET_ID, CTLE_DB, FFE_TAPS, PASS_WINDOW -
Corner capture plan: temperature + supply ± corners must be included early (video links are “almost pass” by default).
Accept: [All corners meet criteria] · Log:
TEMP_C, VDD_V, LINK_RATE, TRAIN_RESULT -
Measurement hygiene: record fixture + reference plane + de-embed settings as immutable metadata (avoid “fake improvement”).
Accept: [Settings reproducible] · Log:
FIXTURE_ID, REF_PLANE, DEEMBED_FILE, RBW_VBW
Production
Cable/connector variation + fixture correlation
-
Golden cable set: qualify multiple cables/lengths/vendors and freeze a golden set for line correlation.
Accept: [Golden set defined] · Log:
CABLE_VENDOR, CABLE_LEN, GOLDEN_SET_ID -
Line quick-screen: use a short test script that exercises (1) HPD/EDID or AUX/DPCD, then (2) a worst-case link rate.
Accept: [0 retries beyond threshold] · Log:
EDID_RETRY_CNT, AUX_RETRY_CNT, LINK_UP_TIME_MS -
Fixture-to-system mismatch guard: if fixture passes but system fails, log the control-plane waveforms (HPD/DDC/AUX) before touching the main-link tuning.
Accept: [Correlation documented] · Log:
STATION_ID, FIXTURE_REV, SYSTEM_FAIL_SIG -
Incoming control-plane components (ESD/CMC) must be version-locked; small C changes can destroy eye margin.
Accept: [AVL locked] · Log:
AVL_REV, ESD_PN, CMC_PN
Diagram · Lifecycle Checklist Strip (Design → EVT → DVT → PVT → MP)
Note: acceptance thresholds are intentionally left as placeholders to be filled with the project’s compliance target, fixture method, and golden-cable correlation.
H2-12 · Applications & IC Selection Notes (HDMI/DP)
Applications are presented as scenario cards. Selection notes are presented as requirement → check → part examples. Part numbers are concrete starting points; always verify the exact data-rate mode, package, and compliance constraints.
Applications (scenario cards)
Long cable / meeting room / classroom
loss + reflections + cable variance
- Link trait: loss grows with length; worst-case cable dominates margin.
- Common pitfall: main link “almost OK” but EDID/HPD/AUX becomes intermittent due to ESD and weak pull-ups.
- Recommended architecture: connector-side protection + (sink-side) redriver/retimer close to the receiver.
-
Example parts (pick by mode):
• HDMI FRL redriver: TMDS1204 / TDP1204
• HDMI TMDS retimer (legacy TMDS): TMDS181 (6G) / TMDS171 (3.4G)
• DP 1.4 redriver: TDP142 (AUX/HPD-aware)
• Protection: TPD4E05U06 or RClamp0524P (choose by capacitance/standard)
Dock / multi-monitor / complex routing
mux + tuning + repeatability
- Link trait: routing introduces discontinuities; “good at one config” is not sufficient.
- Common pitfall: a switch/mux changes insertion loss and return loss, breaking training stability.
- Recommended architecture: mux-with-equalization or mux + redriver/retimer, with preset sweep evidence.
-
Example parts:
• DP mux + retimer class: PS8461 / PS8461E
• DP linear redriver class: PI3DPX1207C / PI3DPX1207B1
• HDMI 2-port switch (1080p-class): TMDS261B
• HDMI port companion (DDC/5V/ESD): TPD12S016
KVM / matrix switch / frequent hot-plug
hot-plug + control-plane stress
- Link trait: HPD toggles are common; EDID/HDCP can repeatedly re-run.
- Common pitfall: HPD bounce or DDC/AUX corruption triggers re-authentication loops that look like “random black screen”.
- Recommended architecture: harden HPD/DDC/AUX, then stabilize main link with a right-class EQ/retiming stage.
-
Example parts:
• HDMI control-plane + ESD: TPD12S016
• HDMI FRL redriver: TMDS1204 / TDP1204
• ESD arrays for control lines: TPD4E05U06 (or automotive TPD4E05U06-Q1)
Industrial display / harsh EMI / field ESD
immunity + robustness
- Link trait: ground noise and common-mode coupling can translate into jitter at the analog front-end.
- Common pitfall: “works on bench” but fails in system due to ESD/CM filtering choices and return current breaks.
- Recommended architecture: connector-first protection, controlled CM strategy, and an EQ/retiming stage placed to minimize stubs.
-
Example parts:
• ESD arrays: TPD4E05U06 / TPD4E05U06-Q1, RClamp0524P
• CM choke example: DLW21SN261XQ2L (select by insertion loss / skew impact)
• DP redriver: TDP142
IC Selection Notes (requirements → checks → part examples)
1) Redriver class (linear EQ / gain shaping)
- Check: supported signaling mode (TMDS/FRL or DP HBRx) + EQ range vs expected insertion loss.
- Check: additive jitter/noise behavior and whether “bigger eye” can still worsen BER (over-EQ risk).
- Check: whether AUX/HPD/DDC are passed/managed correctly (don’t assume).
- Examples: TMDS1204 / TDP1204 (HDMI 2.1 redriver), TDP142 (DP 1.4 redriver), PI3DPX1207C / PI3DPX1207B1 (DP redriver class)
2) Retimer class (CDR + re-transmit / jitter cut)
- Check: re-lock time and behavior during hot-plug or training retries.
- Check: latency budget impact (system-level tolerance differs by product).
- Check: control-plane transparency (HDCP/EDID timing, AUX/DPCD access).
- Examples: ANX7496 / ANX7497 (DP retimer family), TMDS181 / TMDS171 (HDMI TMDS retimer family), IT66313 / IT66319 (HDMI 2.1 retimer family)
3) Switch-with-EQ / Mux + conditioning
- Check: routing-dependent discontinuities (each path is a different channel).
- Check: whether the device changes return loss enough to destabilize training.
- Check: AUX/DDC/HPD handling across the switching event.
- Examples: PS8461 / PS8461E (DP mux + retimer class), TMDS261B (HDMI/DVI 2:1 mux with adaptive EQ; use within its speed class)
4) “Do-not-get-burned” list (practical traps)
- Trap: selecting by “max data rate” but ignoring loss compensation range.
- Trap: fixing main link first while HPD/DDC/AUX is unstable (it will re-break the link).
- Trap: over-EQ makes eye look larger yet increases BER (noise amplification / ringing).
- Trap: production yield collapses due to cable vendor/connector variation (no golden set / no correlation).
Diagram · Selection Scorecard (Redriver vs Retimer vs Switch-with-EQ)
Suggested “first buys” (examples): HDMI FRL redriver TMDS1204/TDP1204, DP redriver TDP142, DP retimer ANX7496/ANX7497, DP mux+retimer PS8461/PS8461E, control-plane companion TPD12S016.
Recommended topics you might also need
Request a Quote
H2-13 · FAQs (HDMI/DP redriver/retimer) — troubleshooting only
Each answer is intentionally short and executable. Every item follows the same 4-line structure and uses measurable fields with threshold placeholders (X/Y/Z) so it can be gated in bring-up and production.
Common fields used below:
MODE_SET,
LINK_RATE,
LANE_COUNT,
TRAIN_RETRY_CNT,
DOWNGRADE_EVENTS,
HPD_TOGGLE_CNT,
EDID_OK_RATE,
AUX_TIMEOUT_CNT,
HDCP_REAUTH_CNT,
RELOCK_TIME_MS,
CABLE_SET_ID,
TEMP_C,
VDD_MV.
Low resolution is stable, but high resolution goes black — budget first or HDCP/training first?
Likely cause: Worst-case rate collapses margin (loss/return-loss/EQ window), or control-plane retriggers (training/HDCP) under stress.
Quick check: Check
DOWNGRADE_EVENTS, TRAIN_RETRY_CNT, HDCP_REAUTH_CNT at the failing mode (MODE_SET).Fix: Stabilize HPD/DDC/AUX first; then retune EQ/retimer preset and validate a pass window (not one lucky setting).
Pass criteria:
DOWNGRADE_EVENTS ≤ X, HDCP_REAUTH_CNT ≤ X, LINK_UP_TIME_MS ≤ Y (TBD).After adding a redriver the eye looks larger, but snow/flicker happens more often — noise or over-EQ?
Likely cause: Over-peaking amplifies noise/ringing; apparent eye opening does not guarantee BER margin.
Quick check: Run a controlled sweep and record stability vs setting (
CTLE_DB, FFE_TAPS, PRESET_ID).Fix: Back off peaking, prioritize return-loss/overshoot control, then retest worst-case cable and temperature corner.
Pass criteria:
PASS_WINDOW ≥ X presets, FLICKER_EVENTS = 0 over T ≥ Y min (TBD).DP keeps repeating link training and finally downshifts to a lower rate — AUX first or Main Link first?
Likely cause: Main Link margin is insufficient at target
LINK_RATE, or AUX/DPCD transactions are unreliable (timeouts/ESD damage).Quick check: Separate symptoms: high
AUX_TIMEOUT_CNT indicates AUX; repeated TRAIN_RETRY_CNT with low AUX timeouts indicates Main Link.Fix: If AUX is failing, harden AUX ESD/termination and routing; if Main Link is failing, tune EQ/retimer presets and reduce discontinuities.
Pass criteria:
AUX_TIMEOUT_CNT ≤ X, TRAIN_RETRY_CNT ≤ Y, DOWNGRADE_EVENTS = 0 (TBD).Hot-plug sometimes fails — how to capture HPD bounce, and what “pass” looks like?
Likely cause: HPD chatter (bounce) triggers re-detect/re-auth/retrain loops; sometimes looks like random black screen.
Quick check: Count and timestamp HPD edges (
HPD_TOGGLE_CNT, HPD_BOUNCE_MAX_MS) and correlate with HDCP_REAUTH_CNT/TRAIN_RETRY_CNT.Fix: Debounce/shape HPD (hardware + firmware policy), harden 5V/HPD routing and ESD, and ensure sideband paths remain stable through plug events.
Pass criteria:
HPD_TOGGLE_CNT ≤ X per plug, HPD_BOUNCE_MAX_MS ≤ Y, LINK_UP_TIME_MS ≤ Z (TBD).EDID is intermittently unreadable — how to quickly validate DDC pull-ups / cable length / buffering?
Likely cause: DDC rise-time too slow (pull-up too weak / too much C), cable variance, or DDC path disturbed by ESD/level shifting.
Quick check: Log EDID retries and success rate (
EDID_RETRY_CNT, EDID_OK_RATE) and measure DDC rise time at the connector (DDC_RISE_TIME_NS).Fix: Adjust pull-ups / reduce parasitic C / add an HDMI DDC companion or buffer; keep DDC/HPD ESD capacitance low and close to the connector.
Pass criteria:
EDID_OK_RATE ≥ 99.9%, EDID_RETRY_CNT ≤ X, DDC_RISE_TIME_NS ≤ Y (TBD).HDCP authentication intermittently fails — how to localize transparency issues vs re-lock triggers?
Likely cause: Sideband instability (HPD/DDC) triggers re-auth, or a retimer/switch path breaks “transparent behavior” under certain events.
Quick check: Correlate failures with
HPD_TOGGLE_CNT and RELOCK_EVENTS; track HDCP_REAUTH_CNT per hour.Fix: Stabilize HPD/DDC first; then validate the chosen device path for hot-plug and mode-switch events; avoid retimer configurations that re-lock frequently during steady video.
Pass criteria:
HDCP_REAUTH_CNT ≤ X per hour, RELOCK_EVENTS ≤ Y, HPD_TOGGLE_CNT ≤ Z (TBD).A cable swap makes it pass/fail — how to build a cable-consistency screen (golden set)?
Likely cause: Cable vendor/length variance shifts insertion/return loss and pushes the system outside its EQ/retime pass window.
Quick check: A/B test with a frozen set and record IDs (
CABLE_SET_ID, CABLE_VENDOR_ID, CABLE_LEN_M) while logging DOWNGRADE_EVENTS/TRAIN_RETRY_CNT.Fix: Define a golden cable set (multiple vendors/lengths) and gate tuning/compliance on worst-case; avoid “tuned to one cable”.
Pass criteria: For all cables in
GOLDEN_SET_ID: DOWNGRADE_EVENTS = 0, TRAIN_RETRY_CNT ≤ X (TBD).The system works, but compliance fails — how to debug reference plane / de-embed / setup traps?
Likely cause: Measurement setup mismatch (reference plane moved, wrong de-embed model, RBW/VBW/trigger artifacts) hides true margin.
Quick check: Make the setup reproducible: record
REF_PLANE_ID, DEEMBED_MODEL_ID, FIXTURE_ID, RBW_VBW.Fix: Re-run with the correct reference plane and validated de-embed; then retune presets using the same compliance setup (do not mix setups).
Pass criteria:
COMPLIANCE_MASK_PASS = true and metadata complete: REF_PLANE_ID/DEEMBED_MODEL_ID present (TBD).EMI fails but functionality is OK — where do CM choke / ESD capacitance trade-offs usually break the link?
Likely cause: ESD/CMC parasitics (capacitance, imbalance) reduce eye/jitter margin while improving emissions; placement can add stubs.
Quick check: Compare before/after part changes with the same preset and cable set; record
ESD_PN, CMC_PN, PLACEMENT_DISTANCE_MM.Fix: Keep ESD close to connector with lowest feasible capacitance; use CM choke only when justified and verify it does not shrink the pass window.
Pass criteria: EMI pass and link margin stable:
PASS_WINDOW ≥ X presets, DOWNGRADE_EVENTS = 0 (TBD).After temperature rises the link starts flickering — where is the first probe point for “power noise → jitter”?
Likely cause: Supply ripple/ground noise couples into the analog front-end (redriver/retimer/PHY), shrinking jitter margin at high rate.
Quick check: Log
TEMP_C/VDD_MV and measure ripple at the device supply pins (not at a distant rail point): RAIL_RIPPLE_MVPP.Fix: Improve local decoupling and return path; reduce coupling loops; re-validate presets across temperature corners.
Pass criteria:
FLICKER_EVENTS = 0 over T ≥ X min at TEMP_C corners; RAIL_RIPPLE_MVPP ≤ Y (TBD).Retimer occasionally re-locks and the picture “jumps” — how to evaluate loop behavior vs configuration?
Likely cause: CDR loop behavior or reference dependency causes occasional loss-of-lock under stress (temperature, EMI, marginal channel).
Quick check: Count lock events and time-to-recover (
RELOCK_EVENTS, RELOCK_TIME_MS) per mode (MODE_SET) and per cable (CABLE_SET_ID).Fix: Tune retimer preset/loop behavior to maximize stability (wider pass window) and reduce re-lock triggers; harden control-plane so re-lock is not cascaded into full retrain.
Pass criteria:
RELOCK_EVENTS ≤ X per hour, RELOCK_TIME_MS ≤ Y, no visible jump events over T ≥ Z min (TBD).In a multi-output distribution, one output is worse — how to separate reflection/return-loss vs connector batch issues?
Likely cause: One branch has worse launch/connector discontinuity or batch-to-batch variation; reflections cut the pass window on that path.
Quick check: Swap outputs while keeping source/cable constant; record per-path IDs (
PATH_ID, CONNECTOR_BATCH_ID) and compare TRAIN_RETRY_CNT/DOWNGRADE_EVENTS.Fix: Rework the worst path’s launch/return path; lock connector AVL/batch screening; avoid stub-inducing placements near the split.
Pass criteria: Across all outputs:
DOWNGRADE_EVENTS = 0, TRAIN_RETRY_CNT ≤ X, and no path-specific fail (TBD).