123 Main Street, New York, NY 10001

Video Format Bridges for HDMI↔MIPI, DP↔LVDS & CSI↔Parallel

← Back to:Interfaces, PHY & SerDes

A Video Format Bridge is an engineering tool that makes two video interfaces interoperable by owning the control plane (EDID/HPD/AUX/DDC), the pixel/transport plane (format/packing/timing), and the clock/CDC plane (PLL/buffers/latency) so bring-up and mode-switching are repeatable.

Most “no image” failures are not raw link errors—success comes from validating these three planes with logs, counters, and pass/fail thresholds, then selecting the right bridge/buffer architecture for the required modes.

H2-1 · What is a Video Format Bridge

Page boundary (scope control)

This page covers format/protocol bridging (control handshakes, pixel transport/packing, and clock-domain decoupling). It does not deep-dive PHY eye/equalization/long-cable SI, ESD/TVS sizing, or retimer internals (handled in sibling pages).

Engineering definition

A video format bridge terminates the source-side control link, translates pixel transport/packing, and decouples clock/reset domains to present a different downstream interface with predictable latency and repeatable bring-up behavior.

What it solves (typical failure pattern)

  • “Black screen” caused by control-plane mismatch (EDID/HPD/AUX/DDC), not raw signal integrity.
  • Color/timing artifacts caused by packing/colorspace/bit-depth mismatch, not random bit errors.
  • Periodic tear/drop caused by clock-domain drift and buffer under/overflow, not cable loss.

What it does NOT replace (boundary guard)

  • Retimer/redriver tasks: cleaning an eye, recovering margin on long links, or compensating insertion loss.
  • ESD/EMI compliance building blocks: TVS selection, CM choke tuning, surge/IEC robustness.
  • Full PHY training theory: link equalization presets and deep electrical compliance (refer to interface-specific pages).

The 3-plane responsibility model (debug starts here)

1) Link / Control plane

Owns EDID/HPD, AUX/DDC transactions, sideband I²C proxying, and lane/mode announcements. Failures look like: EDID reads but no image; different sources behave differently; hot-plug causes endless retrain loops.

2) Pixel / Transport plane

Owns packing/unpacking, colorspace/bit-depth constraints, stream↔packet conversion, and lane mapping. Failures look like: color tint, snow/tiles, stable still image but unstable motion, “works at 720p but fails at 1080p”.

3) Clock / Reset plane

Owns PLL/gearbox, CDC isolation, elastic or frame buffering, and mode-set/reset sequencing. Failures look like: periodic tear, one-frame corruption, every-N-seconds blink, underflow/overflow counters incrementing.

Bridge vs retimer/redriver vs simple converter (quick discriminator)

  • Bridge: input and output protocols differ (HDMI↔MIPI, DP↔LVDS, CSI↔Parallel) and control-plane proxying is required (EDID/HPD/AUX/DDC).
  • Retimer/Redriver: protocol stays the same; the goal is margin recovery (training/BER/CRC that worsens with insertion loss or cable length).
  • Simple converter: electrical/media conversion without terminating or emulating upstream control and capability negotiation (limited interoperability).
Three-plane bridge model (Control / Pixel / Clock-Reset) Inputs HDMI DisplayPort MIPI CSI-2 Bridge Core Control plane EDID / HPD · AUX / DDC · sideband I²C Pixel / Transport plane Packing · Colorspace · Packetize / Depacketize Clock / Reset plane PLL · CDC FIFO · Buffer depth · Mode-set sequencing Outputs MIPI CSI/DSI LVDS Parallel RGB Recommended probe points A EDID/HPD/AUX/DDC transactions B Packing/colorspace/bit-depth status C PLL lock + FIFO fill level + under/over counters
Figure: A bridge is best understood as three planes—control negotiation, pixel transport/packing, and clock/reset decoupling. Most “mystery failures” become deterministic once classified by plane.

H2-2 · Use Cases & Mapping Matrix (within this page boundary)

The purpose of this section is strict scope control and fast routing: identify the bridge class (In→Out), attach the dominant risk labels, then jump to the first chapter that yields actionable probes and pass/fail criteria.

Mapping matrix (practical routing fields)

In → Out Control proxy Transport / packing CDC / buffer Mode-set risk Audio Protection Bring-up safe mode Start here
HDMI → CSI-2 EDID/HPD + DDC TMDS/FRL → packetize (CSI) Elastic FIFO (typ.) / optional frame buffer Medium Drop / extract Mention-only (HDCP cases) 1280×720@60, RGB 8b, no audio H2-3 / H2-4 / H2-5
HDMI → DSI EDID/HPD + DDC HDMI pixels → DSI video mode Tight CDC + panel timing constraints High Usually drop Mention-only 800×480 or 1024×600 (panel-safe), RGB 8b H2-6 / H2-5
DP → LVDS AUX training + EDID-like sink data Main Link → panel stream timing Often needs buffering for training/mode-set Medium Drop Mention-only 1280×720@60 or native panel timing, RGB 8b H2-3 / H2-4
CSI-2 → Parallel Sensor sideband (I²C) proxy (optional) Depacketize + DE/HS/VS synthesis CDC to pixel clock domain Medium N/A N/A Match sensor native format first (8b), then expand H2-4 / H2-5
Parallel → CSI-2 (optional) Upstream camera expectations (strict) Packetize + timing compliance Buffering required for deterministic framing High N/A N/A Lower rate first; lock frame markers before scaling up H2-6 / H2-5

Tip: Start with the “Bring-up safe mode” row values to eliminate format and negotiation variables before increasing resolution, rate, or optional features.

Quick routing by symptom (first chapter to read)

  • EDID reads OK but screen stays black → Control-plane (EDID/HPD/AUX/DDC) next.
  • Image appears but shows tint/snow/tiles → Pixel/packing next.
  • Periodic tear/blink every N seconds → Clock/CDC + buffer next.
  • Fails only on mode switch or hot-plug → Mode-set sequencing next.

Quick routing by system requirement (buffer/scaler fork)

  • Needs resolution/frame-rate conversion (or strong tolerance to input timing drift) → prefer frame buffer architecture (higher latency, higher interoperability).
  • Needs low latency and fixed timing (format shift only) → prefer elastic/line buffer architecture (tight CDC and sequencing discipline).
  • Interoperability must survive many sources/sinks → prioritize control-plane robustness (EDID fallbacks, HPD debounce, AUX retry policies).
Bridge routing map (tags indicate dominant risk plane) Inputs HDMI DisplayPort MIPI CSI-2 Parallel RGB Bridge core Transport + packing (colorspace / bit-depth / packetize) Control proxy (EDID/HPD/AUX/DDC) Clock / CDC / buffer (PLL + FIFO + sequencing) optional scaler Outputs MIPI CSI/DSI LVDS Parallel RGB Control Pixel Clock/CDC
Figure: Use the map and the matrix to lock scope and route debugging. Each bridge path is dominated by one plane (control, pixel, or clock/CDC), which determines the first probe and the first chapter to read.

H2-3 · Control-Plane: EDID / HPD / AUX / DDC Ownership

In video bridging, “no picture” is most often a control-plane mismatch, not corrupted pixel data. This section locks the ownership model (who generates, caches, and advertises capabilities) and the sequencing rules that make bring-up repeatable.

Scope guard (no cross-page expansion)

This section focuses on control ownership and transaction behavior (EDID/HPD/AUX/DDC). Physical signaling margins, eye diagrams, equalization tuning, and long-cable SI are treated as external prerequisites and belong to the interface-specific PHY pages.

Ownership matrix (who owns what, and when)

Bridge role What it must provide Cache / update policy “No picture” signature if wrong First evidence to collect
HDMI input (bridge acts as sink) EDID template or synthesized EDID, stable HPD behavior, DDC transaction handling EDID must be ready before HPD assertion; updates only on explicit state transitions (hot-plug / mode profile change) EDID read succeeds, but source selects an unsupported mode; black screen after mode-set or on hot-plug DDC log: EDID blocks read, checksum, chosen mode, HPD transitions (timestamps)
DP input (bridge terminates AUX as sink) AUX responses, capability mapping, training/parameter visibility (as implemented by the bridge) Cache down-stream constraints and serve stable AUX reads; define retry/timeout behavior with counters Training loops, periodic black screen on mode-set, different hosts show different lock behavior AUX transcript: request/response, timeouts, retries, chosen link parameters (plus bridge counters)
Downstream panel (DSI/LVDS/Parallel) A deterministic “capability contract” equivalent: supported timing set, pixel format constraints, and mode-change rules Panel constraints must dominate the upstream advertisement; any EDID/capability merge must be reproducible across resets Source selects a mode “allowed” upstream but not actually realizable downstream; black screen or unstable image after switching Mode-set record: selected mode, downstream applied timing, and transition timing (reset/enable)

Practical rule: upstream advertisement must never exceed what the downstream can actually render. If the merge policy is not deterministic, interoperability becomes source-dependent.

Control-plane state machine (events and triggers)

S0 · Power rails not stable

HPD forced low; DDC/AUX blocked. Trigger to exit: rails OK + ref clock stable (threshold X placeholder).

S1 · Capability ready (EDID/AUX data prepared)

EDID cached or synthesized; AUX mapping ready. Trigger to exit: HPD assert allowed (debounce X placeholder).

S2 · Negotiation (source reads EDID / performs AUX reads)

DDC/AUX transactions must be consistent and repeatable. Trigger to exit: mode selected + stable transactions (retry policy visible).

S3 · Mode-set transition

Bridge applies pixel/clock settings and transitions enable. Trigger to exit: pixel enable + stable counters (no under/over events).

S4 · Steady state + hot-plug handling

HPD pulses (if used) must be deliberate and debounced; EDID updates must precede any re-negotation. Trigger to re-enter: hot-plug / source reset / explicit mode change.

Packet capture / logging checklist (evidence-first)

  • HPD transitions with timestamps (assert, deassert, pulse width; debounce status).
  • EDID blocks read (including checksum status) and the selected mode derived by the source.
  • DDC error counters: NACK/timeout/retry count; arbitration conflicts (if any).
  • AUX transcript (DP): request/response codes, timeout/retry counts, and final parameter selection.
  • Mode-set record inside the bridge: applied pixel format + timing profile ID + transition timestamp.

Routing rule (avoid dead ends)

If EDID/AUX transactions are stable and HPD behavior is clean, but black screen persists after mode-set, the dominant cause is likely in pixel packing/timing (H2-4) or clock/CDC buffering (H2-5), not the control plane.

Control sequence: HPD → EDID/AUX read → mode-set → pixel enable time → HPD DDC / EDID AUX (DP) Mode-set / Pixel EN HPD low (rails/refclk not stable) HPD high (capability ready) EDID read blocks + checksum AUX reads map + retry policy Mode set apply profile Pixel EN steady state HPD jitter EDID mismatch AUX timeout
Figure: The control plane must be deterministic. HPD should only assert after EDID/AUX data is ready; any jitter, mismatch, or AUX timeout creates source-dependent failures that look like “black screen”.

H2-4 · Pixel/Transport Plane: Packing, Color, Timing, Audio

In bridging, artifacts are more often caused by format and timing mismatches than random bit errors. This section turns pixel transport into a deterministic check path: lock a minimum-viable link (MVP), validate packing and timing, then add complexity (colorspace, bit depth, audio).

Pixel-plane model (4 dimensions that must agree)

1) Colorspace

RGB vs YUV; 4:4:4 / 4:2:2 / 4:2:0. Mismatches typically show tint (purple/green) or incorrect luma/chroma separation.

2) Bit depth

8/10/12 bpc support must match both ends. Mismatches show contouring, banding, or “sparkle-like” artifacts that correlate with edges.

3) Timing model

Active/blanking, porch, sync polarity, and DE/HS/VS behavior (especially for CSI↔Parallel). Timing disagreement often shows unstable lock, tearing, or intermittent black frames.

4) Packing / transport

Stream↔packet conversion, lane mapping, and gearbox packing. Errors here show snow/tiles, line slips, or “still image OK but motion fails”.

MVP bring-up recipe (freeze variables first)

Step 1 · Lock a conservative mode

Start with 1280×720@60, RGB 8b (or a known panel-safe timing) to avoid bandwidth edges. Pass criteria: stable image for X minutes, no line slips, no periodic tear.

Step 2 · Disable optional features

Disable audio, high bit depth, complex colorspace conversions, and any non-essential overlays. Goal: validate packing and timing with minimal degrees of freedom.

Step 3 · Use golden patterns

Validate with color bars + moving checkerboard + a frame counter overlay (if supported). These separate “format errors” (tint/tiles) from “CDC errors” (periodic tear/skip).

Step 4 · Add one variable at a time

Change only one dimension per step: colorspace → bit depth → resolution/rate → audio. Pass criteria: no new artifacts and counters stay within threshold X placeholders.

Symptom → likely cause → first probe (evidence-driven)

Symptom Likely cause (dimension) First probe / check
Purple/green tint or inverted colors Colorspace / range / byte order Read detected input format + configured output packing; verify RGB↔YUV mode matches the sink expectation
Snow/tiles or periodic “line slip” Packing / gearbox / lane mapping Check packet/line counters, mapping configuration, and “payload size per line” consistency (avoid dynamic changes during motion)
Works at 720p but fails at 1080p Bandwidth edge or bit-depth/colorspace combination not supported Reduce one axis: keep resolution, drop bit depth; or keep bit depth, drop colorspace complexity; verify sink constraints
Periodic tear or blink every N seconds Timing drift + CDC buffer under/overflow Read FIFO fill level trend + under/over counters; correlate with blink period (route to clock/CDC section)
Audio disappears after resolution switch Mode-set sequencing resets the audio path Confirm mode-change event order and audio enable gating; verify whether audio is pass/drop/extract per profile

Note: if the control plane is clean (stable EDID/AUX and HPD), persistent pixel issues almost always reduce to a mismatched combination of the four dimensions above.

Audio handling (bridge boundary)

  • Audio modes are typically drop, pass-through, or extract (e.g., to an external audio path).
  • If audio fails only after mode switching, treat it as a sequencing issue (order of resets/enables), not a pixel packing issue.
  • Audio protocol details and electrical transport belong to dedicated audio/interface pages; this page only covers bridging control interactions.
Pixel / transport pipeline (optional blocks shown as dashed) Input unpack Colorspace (optional) Bit depth (optional) Packing gearbox / map Output packetize / format Probe points P1 Detected input format (colorspace/bit depth) P2 Packing configuration + line/frame counters P3 Output payload sizing consistency (per line/frame)
Figure: Most pixel-plane failures are deterministic configuration mismatches. Validate the MVP first, then add one variable per step (colorspace, bit depth, resolution/rate, audio).

H2-5 · Clocking & Domain Crossing: PLL, Gearbox, Buffer Depth, Latency

Clock-domain crossing is the root of long-run bridge stability. If write and read clocks drift, FIFO fill will move, and failures appear as underflow/overflow, periodic tearing, or a black flash every N seconds. This section makes CDC behavior observable and controllable.

Scope guard (no cross-page expansion)

Focus: PLL/gearbox/elastic buffering, fill-level observability, and resync policy. Deep protocol/PHY training math and SI/eye margins are out of scope for this page and should be handled by interface-specific pages.

CDC architectures (trade-offs that matter in bridging)

Architecture Latency (typ / worst) Robustness to drift/jitter Failure signature Best-fit scenarios Must-have observability
Line buffer Low / Low (lines) Weak against ppm drift; limited jitter absorption Periodic tear/flash every N seconds when clocks mismatch Lowest-latency pipelines with controlled clocks or lock-to-reference systems Line counters, occasional slip flags, timing drift indicator (placeholder)
Frame buffer High / High (frames) Strong compatibility; can hide drift and many transient behaviors Longer blackout on mode switch; memory/bandwidth edge failures Multi-source interoperability, timing mismatch tolerance, systems that allow end-to-end delay Frame boundary markers, buffer occupancy, memory under/over events
Elastic buffer + PLL tracking Medium / Medium (lines to sub-frames) Balanced; absorbs short-term jitter and manages long-term drift via tracking Fill-level hunting, rare resync flashes if thresholds/rates are poorly tuned Low-latency systems needing robust multi-source support without full frame buffering FIFO fill level + slope, high/low watermarks, resync counters, tracking status

Engineering note: periodic artifacts strongly suggest deterministic drift. Measure fill-level slope and watermark hit frequency before changing pixel-format settings.

Buffer sizing checklist (from evidence → thresholds)

  1. Identify write clock and read clock sources (common reference vs free-run). Record expected drift drivers (temperature, supply, mode changes).
  2. Define whether the pipeline requires frame boundary alignment (packet/frame markers, DE/HS/VS) or only line continuity.
  3. Log FIFO fill level over time and compute a drift indicator (fill slope; placeholder).
  4. Choose headroom: set low watermark and high watermark with margin for worst-case drift (threshold X placeholders).
  5. Define resync policy: soft resync (adjust/read pacing in blanking) vs hard resync (clear/re-align boundaries).
  6. Pass criteria: zero under/over events for X minutes; watermark hits below threshold X; no periodic tear/flash across temperature sweep (placeholders).

Latency budget (placeholders for system constraints)

Static latency

Line-buffer style: latency ≈ (N_lines × line_time). Frame-buffer style: latency ≈ (N_frames × frame_time). Use placeholders: line_time = X µs, frame_time = X ms, N_lines / N_frames = X.

Dynamic latency (resync & recovery)

Any resync makes latency non-constant. Define worst-case: max blackout = X ms, max alignment time = X lines, and allowable frequency = X per hour (placeholders).

Symptom cue

If artifacts are periodic (every N seconds), prioritize fill-level slope and watermark hit logs before modifying format settings or assuming random errors.

CDC / elastic buffering: drift → fill movement → watermark → resync Input HDMI / DP / CSI pixel stream Write domain WCLK write pointer Elastic FIFO depth = N (placeholder) HIGH WM LOW WM fill level monitor under/over counters slope (drift) indicator Read domain RCLK read pointer Output MIPI / LVDS / Parallel Resync controller trigger: watermark hit soft: pace adjust (blanking) hard: clear + boundary align PLL tracking (optional) rate adjust avoid hunting
Figure: With free-run clocks, fill level drifts. Watermarks plus a defined resync policy are mandatory to prevent periodic tearing/black flashes and to make recovery deterministic.

Bring-up must be executable, not “try-and-see”. This section defines a deterministic sequence from power-on to steady video, with a first probe point and pass criteria placeholders at every step. The same structure applies to hot-plug and re-training behavior.

Bring-up 12-step flow (each step: action → first probe → pass criteria)

Step 1 · Power rails

Probe: rail-good flags / brownout logs. Pass: stable for X ms (placeholder). If fail: power/PMIC domain.

Step 2 · Ref clock stable

Probe: refclk stable indicator / PLL pre-lock. Pass: stable within X ms (placeholder). If fail: clock tree.

Step 3 · Reset asserted

Probe: reset pin state + internal reset cause. Pass: asserted until rails/clock ok. If fail: reset wiring/polarity.

Step 4 · Reset deassert + settle

Probe: device ready flag / firmware boot log. Pass: ready within X ms (placeholder). If fail: boot/straps.

Step 5 · Prepare capability data

Probe: EDID cache valid / AUX mapping ready. Pass: valid before HPD assert. If fail: route to control-plane (H2-3).

Step 6 · HPD assert (gated)

Probe: HPD transition timestamp + debounce status. Pass: no jitter; debounce = X ms (placeholder). If fail: control-plane (H2-3).

Step 7 · Negotiation

Probe: DDC/AUX transactions (timeouts/retries). Pass: retry count < X (placeholder). If fail: control-plane (H2-3).

Step 8 · Apply mode profile

Probe: selected mode ID + output timing registers. Pass: exact match to profile. If fail: pixel/timing (H2-4).

Step 9 · Arm CDC/PLL

Probe: PLL lock + FIFO fill within window. Pass: fill stable; no under/over. If fail: CDC (H2-5).

Step 10 · Enable pixel output

Probe: pixel enable gate + line/frame counters. Pass: counters monotonic; no slips for X frames (placeholder). If fail: H2-4/H2-5.

Step 11 · Steady-state monitoring

Probe: FIFO watermark hits + error counters. Pass: < X per minute (placeholder). If fail: CDC thresholds/resync policy (H2-5).

Step 12 · Hot-plug / re-train handler

Probe: HPD pulse width + post-pulse settle. Pass: pulse width = X ms; settle = X ms (placeholders). If fail: H2-3 sequencing.

Failure localization rule

If negotiation is clean (EDID/AUX stable) but the image tears or flashes periodically, prioritize CDC logs: FIFO fill, watermark hits, and resync counters (H2-5).

Deterministic bring-up: gate each step with logs/measurements Power rails OK measure Refclk stable log Reset deassert + settle log EDID/AUX ready log HPD assert + negotiate log Mode-set → CDC armed measure Hot-plug / re-train HPD pulse width debounce + settle log
Figure: Gate each step with a measurable signal. If a step fails, collect the first probe evidence, then route to control-plane (EDID/HPD/AUX), pixel-plane (format/timing), or CDC (fill/watermarks/resync).

H2-7 · Signal Integrity Boundary: What to Assume vs What to Push to PHY Pages

This page treats signal integrity as an input condition, not a tuning guide. The bridge scope keeps only capability constraints (lane count/rate/encoding class) and uses a routing rule to separate control-plane, pixel-plane, clock/CDC, and PHY-layer issues without expanding into HDMI/DP/MIPI/retimer SI theory.

Scope guard (SI boundary for the bridge page)

  • Keep here: lane count, lane-rate range, encoding family class, and required “link health” flags/counters.
  • Push out: eye/jitter decomposition, CTLE/DFE/FFE tuning, channel loss/reflection/crosstalk modeling, and compliance measurement procedures.
  • Use the router below to decide whether the symptom is control-plane, pixel-plane, clock/CDC, or PHY-layer.

Boundary matrix: keep vs route out (no cross-page expansion)

Category Keep on bridge page Route to PHY pages First probe point (bridge scope)
Link constraints Lane count (X), lane-rate range (X–Y), encoding family class, allowable pixel-clock range (X–Y placeholders) Eye/jitter margin methodology, channel budget modeling, compliance test setup Training/lock status + error counters + “link stable” flag
Equalization EQ as requirements (e.g., “must train to a stable preset”, placeholder) and “known-good preset” artifacts CTLE/DFE/FFE scan strategy and tuning loops Retry/timeout counters; stability window X; preset ID captured
Symptom mapping “Which plane?” routing: Control vs Pixel vs Clock/CDC vs PHY Deep SI root-cause (cables, return paths, reflections, crosstalk) Control logs (EDID/AUX), mode record, FIFO fill/watermarks, training result
Compliance mindset Save artifacts and presets so issues are reproducible (see H2-8) PHY compliance templates and measurement details Artifacts: training logs + preset IDs + firmware/config snapshot

Practical rule: if negotiation is clean but artifacts are periodic (every N seconds), treat it as CDC until proven otherwise (FIFO fill, watermarks, resync counters).

Troubleshooting router (symptom → first probe → primary chapter)

Black screen but EDID reads OK

First probe: mode-set record + output timing; then FIFO fill window. Primary: Pixel plane (H2-4) → Clock/CDC (H2-5).

Periodic tear/flash every N seconds

First probe: FIFO fill slope + watermark hits + under/over counters. Primary: Clock/CDC (H2-5).

Hot-plug works once, then fails to recover

First probe: HPD pulse width + debounce + EDID-ready gating order. Primary: Sequencing (H2-6) → Control plane (H2-3).

Color is wrong / image is stable but “looks off”

First probe: colorspace + packing + bit depth settings. Primary: Pixel plane (H2-4).

Training fails or link never locks

First probe: training/lock status + retry/timeout counters. Primary: PHY pages (HDMI / DP / MIPI / Retimer). This page keeps only “input condition” requirements.

Symptom router: decide the plane before changing knobs Observed symptom black / tear / color / no lock Control plane EDID / HPD / AUX probe: timeouts/retries route: H2-3 / H2-6 Pixel plane Color / Packing / Timing probe: mode record route: H2-4 Clock / CDC FIFO / Watermarks / Resync probe: fill + slope route: H2-5 PHY layer Eye / Jitter / EQ go to pages HDMI / DP / MIPI / Retimer
Figure: Use plane routing to localize first. PHY/SI is a separate page scope; this bridge page uses link constraints and observability to decide when escalation is justified.

H2-8 · Compliance Hooks & Artifacts (EDID/HDCP/PCI-SIG-like mindset)

Bridge projects fail late when interoperability artifacts are missing. This section defines the minimum reproducibility package: EDID governance, content-protection presence boundaries, and logs/presets that must be saved from bring-up through validation and production.

Artifacts checklist (copy-ready for project tracking)

Bring-up (field reproducibility)

  • EDID dump (raw) + checksum OK + hash ID (placeholders)
  • Fallback EDID profile + activation condition (placeholders)
  • AUX/DDC transaction log (timeouts, retries, error codes)
  • Mode-set record (resolution, color, bit depth, packing, output timing)
  • CDC snapshot (FIFO fill, watermark hits, under/over counters)
  • Firmware revision + config snapshot (must bind to artifact bundle ID)

Validation (EVT/DVT regression)

  • Pass matrix: modes × sources × sinks (placeholder structure)
  • Known-good presets: training preset IDs / EQ preset IDs (saved, not explained)
  • Failure bundle: symptom + logs + dumps + exact FW/config ID
  • Mode-switch black screen budget: max blackout X ms (placeholder)

Production (traceability)

  • Release bundle ID → firmware revision → config checksum mapping (placeholders)
  • Locked EDID set + policy (when/why fallback triggers)
  • Manufacturing log schema: required fields and retention window (placeholders)
  • Compatibility change note: what changed, why, and how verified

EDID governance rule

Any EDID update must happen before HPD pulse that triggers re-read. Store both raw and fallback dumps and bind them to the same artifact bundle ID (placeholders).

Content protection boundary (no crypto detail)

Treat HDCP/DRM as a presence constraint: supported/transparent/not supported (placeholders). Do not expand into key management or encryption flows on this page.

Artifacts pipeline: reproducibility bundle across lifecycle Bring-up Validation Production EDID dump + hash AUX/DDC log Mode-set record CDC snapshot FW + config ID Artifact bundle ID Pass matrix Known-good presets Failure bundle Change notes Bundle ID persists FW+config lock Locked EDID set Traceability pack Retention policy Release bundle ID Binding rule: bundle ID ↔ firmware revision ↔ config checksum (placeholders)
Figure: Save dumps/logs/presets as a single artifact bundle that persists from bring-up to production, so any interoperability issue is reproducible and traceable.

H2-9 · Debug & Instrumentation: Probes, Logs, and Golden Patterns

Debug must start with first probe points and consistent observability. Without a fixed probe order, parameter changes become infinite. This section defines a minimal instrumentation loop: where to probe (control/pixel/clock), which counters to expose, and golden patterns to separate tearing, color mismatch, and frame loss.

First 5 probes (fixed order)

Probe-1 (Control): DDC/AUX + HPD edges

  • Observe: timeouts, retries, NACK/error codes, HPD rise/fall/glitch counts
  • Pass criteria: no timeout in X consecutive reads; HPD glitches < X (placeholders)
  • Route next: Control-plane ownership (H2-3) / Sequencing (H2-6)

Probe-2 (Config): Mode-set record snapshot

  • Observe: resolution/refresh, colorspace, bit depth, packing, output timing ID
  • Pass criteria: record hash remains stable across X attempts (placeholder)
  • Route next: Pixel/transport packing & timing (H2-4)

Probe-3 (Pixel): After input unpack

  • Observe: in-frame counter, in-line counter, DE/HS/VS presence, format tag
  • Pass criteria: monotonic frame count; expected lines per frame (placeholder)
  • Route next: Pixel plane mismatch (H2-4)

Probe-4 (Clock): PLL lock stability window

  • Observe: lock flag, lock-loss count, relock timestamps
  • Pass criteria: lock-loss = 0 over X seconds; relock count < X (placeholders)
  • Route next: Clocking & CDC fundamentals (H2-5) / Sequencing (H2-6)

Probe-5 (CDC): FIFO fill + watermarks + under/over

  • Observe: min/max fill, watermark hits, underflow/overflow, resync events
  • Pass criteria: fill stays within safe band; under/over = 0 (placeholders)
  • Route next: Buffer depth & resync strategy (H2-5)

Counters to expose in firmware (minimal schema)

Control-plane

edid_read_ok / edid_read_fail · aux_timeout / aux_retry · ddc_nack / ddc_timeout · hpd_rise_cnt / hpd_fall_cnt / hpd_glitch_cnt · last_error_code(enum)

Pixel/transport

in_frame_cnt / out_frame_cnt · in_line_cnt / out_line_cnt · format_tag_in / format_tag_out(enum) · drop_frame_cnt / repeat_frame_cnt (if buffering) · crc_or_packet_err_cnt (if available)

Clock/CDC

pll_lock_loss_cnt · fifo_underflow_cnt / fifo_overflow_cnt · fifo_wm_hi_cnt / fifo_wm_lo_cnt · resync_event_cnt · fill_min / fill_max (windowed) · fill_slope_est (optional)

Logging rule: every field above must be tied to an artifact bundle ID (H2-8) and a mode-set record hash (Probe-2).

Golden patterns (fast discrimination)

Color bars

Use to detect colorspace/range/bit-depth mismatch. Typical signatures: stable image but wrong hues, crushed blacks, banding. Next: Pixel plane (H2-4) + confirm format_tag_in/out.

Moving checkerboard

Use to separate tearing, packing disorder, lane mapping issues. Motion amplifies periodic artifacts; edges reveal packing mistakes. Next: Pixel plane (H2-4) then CDC if periodic.

Frame counter overlay

Use to detect drop/repeat/freeze. Counter discontinuity correlates with FIFO under/over or resync events. Next: Clock/CDC (H2-5) + check fifo_underflow/overflow and resync_event_cnt.

Probe map: control, pixel, and clock observability points Input IF HDMI / DP / CSI Unpack format tag CSC (opt.) RGB/YUV Pack gearbox Output IF MIPI/LVDS Control plane DDC / AUX / HPD Clock / CDC PLL + FIFO / elastic buffer PLL lock loss cnt FIFO fill wm / under / over P1 DDC/AUX/HPD P2 mode record P3 after unpack P4 before out P5 FIFO fill Use probes with golden patterns Color bars → colorspace/bit depth · Checkerboard → tearing/packing · Frame counter → drop/repeat Bind logs to artifact bundle ID + FW/config ID (H2-8)
Figure: Probe points P1–P5 anchor debugging to observability before changing parameters. Use golden patterns to classify tearing vs color mismatch vs frame anomalies quickly.

H2-10 · Design Checklist (Schematic → Layout → Firmware → Bring-up → Production)

A bridge is a system: electrical ownership, control-plane behavior, pixel formatting, and clock/CDC must be verified across the full lifecycle. The checklist below stays within bridge scope: layout focuses on partitioning and clock loops, while deep SI tuning is routed to PHY pages.

Schematic checklist

  • Power rails sequencing: rails stable before releasing reset (X ms placeholder)
  • Refclk ownership: source, gating, and “refclk-good” definition (placeholder)
  • I²C pull-up ownership: which side provides pull-ups and when they are valid
  • HPD level compatibility: default level, tolerance, and clamp policy (placeholders)
  • Straps/config defaults: safe-mode fallback must be deterministic
  • Pass criteria: EDID readable X times after HPD assert; no DDC timeout (placeholders)

Layout checklist (bridge-only)

  • Clock loop partitioning: keep PLL/refclk loop compact and isolated
  • Control-plane routing hygiene: DDC/AUX routing consistent with pull-up ownership
  • I/O zoning: separate high-activity pixel logic from control rails/lines
  • Return continuity rule: avoid split references around bridge I/O regions
  • ESD return awareness: define return path policy for connector protection parts
  • Pass criteria: no HPD glitches induced by local switching events (X placeholder)

Note: SI details (eye/jitter, equalization tuning, channel modeling) are routed to PHY pages to avoid cross-page overlap.

Firmware checklist

  • EDID fallback policy: trigger condition, versioning, and checksum validation
  • Hot-plug debounce: HPD pulse width X ms + glitch filter (placeholders)
  • Safe defaults: minimal-known-good mode to start (e.g., 720p/60 RGB 8b placeholder)
  • Expose counters: schema from H2-9 + bind to bundle ID (H2-8)
  • Resync strategy: FIFO watermark thresholds and max resync rate (placeholders)
  • Pass criteria: mode record hash stable; no lock-loss in X seconds (placeholders)

Bring-up checklist

  • Verify Probe-1 to Probe-5 in order; do not tune before observability is clean
  • Golden patterns: color bars + moving checkerboard + frame counter overlay
  • Clock/CDC: fifo_under/over = 0 over X minutes; fill within safe band (placeholders)
  • Mode switch: blackout < X ms; no permanent black after switch (placeholders)
  • Artifacts: save dumps/logs as a bundle with FW/config ID (H2-8)
  • Pass criteria: stable output for X minutes across hot-plug cycle X times (placeholders)

Production checklist

  • Self-test/BIST: deterministic pattern path + counters snapshot
  • Traceability fields: bundle ID, FW revision, config checksum, mode hash (placeholders)
  • Locked EDID policy: approved EDID set + fallback mapping
  • Regression gate: known-good presets cannot change without pass matrix update
  • Retention: define artifact retention window X days (placeholder)
  • Pass criteria: station-to-station mismatch < X; failure bundle reproducible (placeholders)
Bridge lifecycle checklist: schematic → layout → firmware → bring-up → production Schematic Rails HPD Pull-up Refclk Layout Clock loop Partition Return Firmware Fallback Debounce Counters Bring-up Probes Golden FIFO Prod BIST Trace Lock Pass-criteria band (placeholders) HPD pulse width X ms · EDID stable X reads · lock-loss=0 over X s · fifo_under/over=0 over X min · blackout < X ms · bundle ID bound to FW/config Deep SI methods are routed to PHY pages (HDMI/DP/MIPI/Retimer) to keep this bridge page non-overlapping.
Figure: Checklist pipeline enforces stage-gated verification. Each stage should produce artifacts and pass criteria to prevent late interoperability regressions.

IC Selection Logic: Bridge vs Scaler vs Retimer (Boundaries + Key Metrics)

Selection is treated as an engineering decision tree plus verification gates. This section lists representative material numbers for bridge-class devices (and a few boundary examples for retimers) while avoiding product-page sprawl.

Page boundary (no cross-page expansion)
  • Retimer/Redriver appears only as a routing result; SI/eye/equalization methods belong to PHY/retimer pages.
  • Scaler is included only as a need-or-not branch (resolution/frame-rate change), not algorithm details.
  • Material numbers are representative; verify package, suffix, licensing (HDCP), and availability.

A) Decision Tree (classify first, then compare)

Gate-1: Must resolution / frame-rate change?
Yes → Bridge + Scaler (often frame-buffer based).
No → continue to Gate-2.
Gate-2: Must protocol / packing / handshake ownership change?
Yes → Video Format Bridge (control-plane + pixel-plane + clock-plane).
No → continue to Gate-3.
Gate-3: Is it only reach-extension (same protocol)?
Yes → Retimer/Redriver category (route to PHY/retimer pages).
No → continue to Gate-4.
Gate-4: Is control-plane proxying required (EDID/HPD/AUX/DDC, content protection)?
Yes → Prefer bridge-class devices with configurable firmware + logs/counters.
No → minimal-control bridge may still be sufficient (MVP-first).
Diagram: Selection Decision Tree (class boundaries)
Decision gate Result category Boundary route-out Input interface HDMI / DP / CSI-2 / Parallel Output interface MIPI (CSI/DSI) / LVDS / Parallel Gate-1: change resolution / frame rate? If YES → scaler path (buffer/latency dominant) YES Bridge + Scaler Frame buffer Latency ↑ NO Gate-2: change protocol / packing / handshake? If YES → format bridge (control + pixel + clock) YES Video Format Bridge EDID/HPD/AUX/DDC Packing / CDC / PLL NO Gate-3: same protocol, only reach-extension? If YES → retimer/redriver (route out) YES Retimer / Redriver Route to PHY pages Eye/jitter/EQ there NO Gate-4: need strong control-plane proxy + logs? If YES → firmware-rich bridge preferred Firmware-rich Bridge EDID fallback • counters • traces
Labels are intentionally short; the objective is fast classification before deep verification.

B) Key Metrics to Compare (bridge-class, within page boundary)

Prefer metrics that map directly to bring-up and failure modes (black screen, tearing, color shifts, periodic dropouts).
Metric Why it matters Quick check Pass criteria (placeholder)
Supported input/output mode set Prevents late surprises (unsupported bit-depth/colorspace or blanking limits). Enumerate modes; record any conditional limitations (e.g., 4:2:0 only at high rates). Target modes covered ≥ X%; no silent fallback.
Buffer architecture (line / elastic / frame) Determines latency and tolerance to clock mismatch / mode switching. Confirm FIFO/frame store presence; verify counters: under/over, watermark events. Under/over counters = 0 over X hours; no periodic tearing.
Control-plane ownership (EDID/HPD/AUX/DDC) Black-screen is often a control-plane mismatch rather than data corruption. Verify: EDID cache/fallback; HPD debounce/pulse; AUX/DDC proxy timeouts and retry. EDID stable across reset; HPD pulse width within X ms; AUX/DDC error rate < X.
Mode-set behavior (blackout / recovery) Prevents “works once, fails on switch” field issues. Execute repeated mode switches; log transitions and any PLL/FIFO events. Blackout < X ms; no latch-up; recovery success ≥ X%.
Debug counters / logs / dumps Enables deterministic troubleshooting (first probe points). Confirm exposure of: PLL lock, FIFO fill, under/over, EDID/AUX/DDC transactions. Required counters present; trace export in < X minutes.

C) Representative Material Numbers (examples, not exhaustive)

Use these as starting points for category-fit. Always verify: package/suffix, HDCP licensing requirements, lane/rate limits, and firmware tooling.

Category Material No. Typical mapping Notes (engineering)
HDMI → CSI-2 bridge TC358743XBG HDMI 1.4 → MIPI CSI-2 Control-plane ownership + CSI packetization; validate EDID/HPD bring-up first.
HDMI → DSI bridge TC358870XBG HDMI 1.4b → MIPI DSI Mode-set sequencing and clocking are frequent failure points; log PLL/FIFO events.
HDMI → DSI (with scaling) TC358779XBG HDMI → MIPI DSI (scaler/de-interlace) Choose only when resolution/frame-rate conversion is required; latency budget must be explicit.
HDMI 2.0 → MIPI/LVDS converter LT6911UXC HDMI 2.0 → CSI/DSI/LVDS Useful when higher HDMI bandwidth is needed; confirm firmware tooling and debug visibility.
DSI → LVDS bridge (single-link) SN65DSI83 MIPI DSI → LVDS Confirm bpp support and pixel clock range; verify colorspace/packing expectations.
DSI → LVDS bridge (dual-link) SN65DSI85 MIPI DSI → LVDS (higher bandwidth) Prefer when resolution/refresh pushes beyond single-link margin; validate mode-set repeatability.
(e)DP → LVDS bridge PTN3460 / PTN3460I (e)DP main link → LVDS panel DP control-plane (AUX) ownership matters; treat training/mode-set as first-class logs.
CSI-2 ↔ Parallel bridge TC358746AXBG CSI-2 → Parallel (and reverse) Useful for CSI↔Parallel bring-up; validate PCLK limits and timing polarity expectations.
CSI-2 ↔ Parallel bridge TC358748XBG CSI-2 ↔ Parallel (variant) Same class as TC358746; confirm register map, interface options, and tool support.
Boundary examples (retimer/redriver) TDP142 / TMDS181 DP redriver / HDMI TMDS retimer Listed only to clarify “same protocol reach-extension”; SI methodology belongs elsewhere.
BOM placeholders (keep the page non-commercial)
  • U? Video format bridge IC (selected class + firmware/counters requirement)
  • U? Optional scaler/frame-buffer IC (only if Gate-1 = YES)
  • U? Optional retimer/redriver (only if Gate-3 = YES)
  • R/C? DDC/AUX pull-ups / level considerations (ownership defined in H2-3)

D) Verification Use-Cases (turn selection into deliverables)

Tier-1: MVP bring-up (fixed safe mode)
  • Lock a conservative mode (e.g., 720p60 RGB 8b) and keep it stable.
  • Pass criteria: EDID stable, no lock-loss, FIFO under/over = 0 for X minutes.
Tier-2: Mode stress (coverage and switch repeatability)
  • Run target modes across bit-depth/colorspace boundaries used in the system.
  • Pass criteria: blackout < X ms, recovery success ≥ X%, no periodic tearing.
Tier-3: Long-run + recovery (field realism)
  • Run for X hours and perform hot-plug cycles (X times).
  • Pass criteria: artifacts captured (EDID dump, AUX/DDC logs, mode-set record), and counters show no accumulating errors.
Recommended “selection record” to store with the design
  • Chosen class: Bridge / Bridge+Scaler / Retimer boundary
  • Target modes list (input/output) + any conditional restrictions
  • Latency budget and buffer architecture assumption
  • Required counters/logs + how to retrieve them in X minutes

Request a Quote

Accepted Formats

pdf, csv, xls, xlsx, zip

Attachment

Drag & drop files here or use the button below.

FAQs (Troubleshooting, within this page boundary)

Each answer is intentionally short and executable, using a fixed four-line structure: Likely cause / Quick check / Fix / Pass criteria (threshold placeholders X).

EDID read OK but black screen after mode-set — control plane or pixel plane first?
Likely cause: Mode-set succeeds on control-plane, but pixel/clock-plane does not start cleanly (packing mismatch, PLL not locked, FIFO not primed).
Quick check: Capture the mode-set record (resolution, refresh, colorspace, bpc, blanking) and verify PLL lock + FIFO fill-level + output packetizer enable at the first active frame.
Fix: Force a safe MVP mode (e.g., 720p60 RGB 8b), pin explicit packing/colorspace, then re-enable the target mode with the same sequence order (EDID ready → HPD stable → mode-set → pixel enable).
Pass criteria: Blackout < X ms after mode-set; PLL unlock events = 0; FIFO under/over counters Δ = 0 over X minutes.
HPD toggling makes the source constantly re-train — what to debounce and how?
Likely cause: HPD is glitching (debounce too short, ground reference noise, rail/reset timing causing pulses).
Quick check: Measure HPD at the source pin (with system ground reference) and log HPD transitions with timestamps; record pulse widths and count retrain triggers.
Fix: Add firmware HPD debounce X ms + minimum stable-high time X ms; keep HPD low until EDID cache is valid; apply level/RC conditioning if HPD high level is marginal.
Pass criteria: Retrain count = 0 over X minutes; HPD pulse width stays within X–Y ms (no micro-glitches < X ms).
Image shows but color is wrong (purple/green) — first packing/colorspace check?
Likely cause: Colorspace/packing mismatch (RGB↔YCbCr, 4:4:4 vs 4:2:2, channel swap, wrong bpc alignment).
Quick check: Display a color-bar test pattern and read bridge format registers (input colorspace, output colorspace, swap bits, bpc); verify the expected mapping is explicitly set (not auto).
Fix: Pin the full chain: force RGB 8b or a known-good YUV mode, disable “auto colorspace,” apply the required swap/CSC setting, then re-test with the same pattern.
Pass criteria: Color-bar sanity: channel order correct; per-channel average error < X% (or ΔE < X); no mode-dependent color flips across X switches.
Periodic tear every few seconds — FIFO underflow/overflow or source timing drift?
Likely cause: CDC/buffering issue: FIFO fill-level drifts until a watermark triggers (under/over), often periodic when clocks are slightly mismatched.
Quick check: Trend FIFO fill-level vs time and read underflow/overflow + watermark counters; correlate the tear period with the fill-level ramp.
Fix: Increase effective elasticity (deeper buffer, better watermark policy), adjust PLL tracking/gearbox settings, or enable “free-run + resync on threshold” if supported.
Pass criteria: FIFO under/over counters Δ = 0 over X minutes; no visible tear over X minutes; fill-level stays within X–Y% band.
Works at 720p, fails at 1080p — bandwidth limit or clocking/gearbox setting?
Likely cause: Throughput/clocking boundary (lane rate limit, pixel clock limit, bpc/packing pushes beyond gearbox range) rather than random corruption.
Quick check: Compute required throughput for 1080p mode (pixel clock × bpp) and compare to configured lanes/rates; verify PLL lock and that the intended gearbox divisor/packing is active.
Fix: Reduce bpc/colorspace (e.g., 8b RGB), adjust lane count/rate, tighten blanking if allowed, and re-apply the correct divisor/gearbox preset for the target mode.
Pass criteria: Target mode runs > X hours; counters Δ = 0; computed throughput margin ≥ X%.
DP→LVDS: backlight on but no image — AUX training vs panel timing mismatch?
Likely cause: DP link training incomplete (AUX side) or LVDS timing generator mismatch (DE/HS/VS polarity, porch/blanking, mapping).
Quick check: Read DP link status (training success, lane count/rate, link quality flags) and compare LVDS timing parameters against the panel timing table (active/blanking/polarity).
Fix: Lock DP to a known-good lane/rate preset first, then set LVDS timing exactly to panel requirements (polarity/porches), and validate mapping (JEIDA/VESA, bit order) if applicable.
Pass criteria: Training success = 100%; stable image over X minutes; no retrain events; timing hash matches the panel spec across X reboots.
HDMI→MIPI: stable still image, fails on motion — buffer depth or frame pacing?
Likely cause: Motion increases burst stress: buffer/DDR margin or pacing issue reveals FIFO watermark events and occasional frame drops/corruption.
Quick check: Run a moving checkerboard and monitor: frame-drop counter, FIFO near-empty/full events, and any frame-buffer error/ECC flags (if present).
Fix: Increase effective buffering (deeper FIFO or safer watermark), stabilize clock/PLL policy, and disable aggressive low-power transitions during active video if they disturb pacing.
Pass criteria: Frame drops < X/hour; FIFO under/over Δ = 0 over X minutes on motion pattern; no visible corruption.
Random one-frame corruption under load — DDR/frame buffer margin or CDC?
Likely cause: Frame-buffer/DDR margin (burst collisions, timing margin) or a CDC threshold hit causing a single bad frame, not a persistent format mismatch.
Quick check: Enable a frame counter overlay / internal CRC (if available) and correlate bad frames with DDR error flags, FIFO watermark, or PLL unlock events.
Fix: Reduce DDR speed or tighten DDR timing margins, increase buffer watermarks, and add retry/scrub policy if supported; keep mode-set parameters pinned (no auto changes).
Pass criteria: Corruption rate < X frames/hour over X hours; DDR error flags = 0; watermark events Δ < X.
Hot-plug works on bench, fails in system — HPD level/ground reference or firmware timing?
Likely cause: System integration effects: HPD high level marginal, ground offset/return path noise, or power/reset timing shifts HPD/EDID readiness.
Quick check: Measure HPD high level and rise/fall in-system at the source pin; log the exact sequence timestamps (rails good → refclk stable → reset deassert → EDID ready → HPD assert).
Fix: Add HPD level translation or conditioning, enforce “EDID-ready before HPD” with delays X ms, and apply hot-plug debounce/pulse shaping consistent with source expectations.
Pass criteria: Hot-plug success ≥ X% over N cycles; retrain count ≤ X; blackout < X ms per plug.
Two sources behave differently with same bridge — what EDID field is usually the culprit?
Likely cause: EDID advertises capabilities that trigger different source policies (preferred timing list, colorspace support, deep color, limited/full range, audio blocks).
Quick check: Diff EDID dumps from both scenarios; identify changes in CTA extension blocks and the “preferred mode” list; test with a conservative fallback EDID that advertises only MVP modes.
Fix: Provide a conservative EDID fallback, remove unsupported modes, and pin the operational mode (colorspace + bpc + timing) to prevent source auto-selection surprises.
Pass criteria: Both sources select the same mode (mode hash matches) across X power cycles; no unexpected mode switches; stability ≥ X minutes per cycle.
“Eye/jitter looks fine” but link unstable — which non-SI counters to check first?
Likely cause: Non-SI instability: control-plane timeouts, PLL unlocks, FIFO watermark hits, packetizer errors, or mode-set sequencing glitches.
Quick check: Read counters/logs in this order: AUX/DDC timeoutPLL unlockFIFO under/overpacket/CRC/frame drop; identify which counter increments at failure.
Fix: Fix the first incrementing counter’s root (timeouts/debounce, PLL preset, buffer policy, sequencing). If all counters stay clean, route-out to PHY/SI pages for eye/jitter work.
Pass criteria: All listed counters Δ = 0 over X minutes; no drops in X retrain attempts; route-out only if counters remain clean.
Audio disappears after resolution switch — mode-set sequencing or audio path reset?
Likely cause: Audio path not re-initialized after mode-set (mute latched, audio clock gating not restarted, infoframes/packets not regenerated).
Quick check: Log audio enable/status bits across the switch; confirm audio packets/infoframes resume and the audio clock domain is ungated after video re-lock.
Fix: Re-run audio init on each mode-set (clear mute, restart audio clock, re-announce capability), and enforce a safe order: video lock → audio reset release → audio enable.
Pass criteria: Audio resumes within X ms after each switch; no mute events over X consecutive mode switches; audio-drop rate < X.