123 Main Street, New York, NY 10001

← Back to: Supervisors & Reset

What It Solves

Map hard-to-see yet fatal imaging artifacts to controllable power-supervision strategies and measurable acceptance: UV/OV windows, programmable delay/debounce, soft clamp / controlled discharge, multi-rail PG aggregation, and safe OD/PP semantics.

Cold/Resume Black-Level Jump

VREF/VCM stabilize first + tPG_filter ≥ 0.5–1 ms + keep-bias soft clamp.

Row Banding / FPN on Power-Up

AVDD controlled ramp; DVDD after analog; avoid half-powered sampling.

Thermal Black-Level Drift

Narrow VREF/VCM windows + ppm/°C margin; validate across −40~+85 °C.

False PG / Back-Power

OD outputs, proper pull-up rail, fanout buffering; PG aggregation with first-fault tagging.

Acceptance (examples):

  • ΔBlack-Level@resume ≤ 2 LSB (25 °C); ≤ 4 LSB at 85 °C
  • FPN RMS ≤ baseline + 10%
  • Sleep→Stable resume time ≤ 5 ms (with VREF/VCM kept)
  • Re-reset events ≤ 10 ppm (ramp + brownout scripts)
Imager rails and dependencies mapped to black-level stability Center chain VREF/VCM→AVDD→DVDD; right-side symptom-to-capability cards; bottom OD/PP PG semantics. What It Solves VREF / VCM Reference & common-mode AVDD Analog front-end rail DVDD Digital core / IO rail then then Symptom → Capability • Cold/resume level jump → VREF/VCM first + tPG_filter + keep-bias clamp • Row banding/FPN → AVDD controlled ramp; DVDD last • Thermal drift → Narrow VREF/VCM windows with ppm/°C margin • False PG/back-power → OD outputs; correct pull-up; PG aggregation PG / RESET semantics Prefer OD across mixed domains; PP only in same-level fanout; define low/high-active consistently. Debounce 0.5–2 ms; record first-fault/last-good when aggregating rails.

Rails & Bias Map for Imager/AFE

Recommended order: VREF/VCM first → AVDD next → DVDD last. Set narrow windows for references, controlled discharge for analog, and allow faster discharge for digital with proper debounce.

Rail Window % Thermal Margin tPG / Debounce Discharge Policy Output (OD/PP) Pull-up Rail Notes
VREF ±1.5–2% ≥ 30–50 ppm/°C tPG ≥ 2 ms keep (no fast) OD always-on ref Bias conservation across sleep
VCM ±2% ≥ 50 ppm/°C tPG ≥ 1.5 ms keep (no fast) OD reference domain Common-mode for sensor black level
AVDD ±5% ≥ 100 ppm/°C tPG ≥ 1 ms slow-discharge OD/PP (same domain) analog rail Control dV/dt to avoid sampling artifacts
DVDD ±8–10% ≥ 150 ppm/°C tPG ≥ 0.5–1 ms fast-allowed OD (mixed domains) digital or IO rail Debounce stronger to mask bus chatter
VANA* ±3–4% ≥ 80 ppm/°C tPG ≥ 1–1.5 ms slow-discharge OD/PP per domain analog/ISP domain Project-specific analog sensitive rail
Threshold windows and power-up dependencies across VREF/VCM/AVDD/DVDD Five-rail cards with low/high window bars, recommended delays, and power-up order arrows; bottom legend shows keep/slow/fast discharge. Rails & Bias Map VREF ±1.5–2% tPG ≥ 2 ms keep / OD VCM ±2% tPG ≥ 1.5 ms keep / OD AVDD ±5% tPG ≥ 1 ms slow / OD/PP DVDD ±8–10% tPG ≥ 0.5–1 ms fast / OD VANA* ±3–4% tPG ≥ 1–1.5 ms slow / per domain Discharge / Clamp Policies • keep: preserve VREF/VCM across sleep • slow: AVDD controlled decay • fast: DVDD allowed with stronger debounce

Soft Clamp & Controlled Discharge

Use keep-bias + slow discharge to avoid black-level re-calibration; allow faster discharge on digital rails when paired with stronger debounce. Default policy: VREF/VCM = keep, AVDD = slow, DVDD = fast-allowed.

Strategy Matrix

{VREF, VCM, AVDD, DVDD} × {keep / slow / fast / cut}. Default highlights emphasize bias conservation.

Discharge Path

External FET with controlled Rds(on) and time window; avoid rapid collapse of references.

Sleep/Resume

Keeping VREF/VCM reduces resume calibration from ms to µs; capture Δblacklevel_resume as KPI.

Engineering: t_discharge, Rds(on), E_hold, I_leak_total, Δblacklevel_resume; abnormal power-loss → “limit current + keep bias” hook.

Discharge/soft-clamp policies per rail with default safe choices Matrix with rails (VREF, VCM, AVDD, DVDD) versus actions (keep, slow, fast, cut). Default choices highlighted; includes iconography for FET, timer, and reference caps. Soft Clamp & Controlled Discharge keep slow fast cut VREF VCM AVDD DVDD FET t VREF/VCM t_discharge ≈ C_load·ΔV / I_path · E_hold = ½·C_bias·V_bias²

Power-Up Order & PG/RESET Semantics

Turn the slogan “reference → analog → digital” into a repeatable recipe: OD-first across mixed domains, PP only within same-level fanout; aggregate PG with first-fault/last-good tagging; apply debounce to tolerate slow ramps.

Recipe: tPG(VREF)=2 ms → tEN(AVDD)=+1 ms → tEN(DVDD)=+2 ms; tPG_filter ≥ 0.5–1 ms; tPW_reset_min per MCU (50–200 µs typical).

Safe power-up order and PG delay windows to avoid black-level drift Three-rail time diagram (VREF, AVDD, DVDD) with enable points, PG windows, and reset pulse width; OD/PP semantics and a slow-ramp tolerance zone. Power-Up Order & PG Semantics time → VREF/VCM tPG=2ms PG window AVDD tEN=+1ms PG window DVDD tEN=+2ms PG window RESET tPW_reset_min Slow-ramp tolerance OD across mixed domains (preferred) PP only in same-level fanout

Parameters: tPW_reset_min, tPG_valid_min, I_fanout, R_pull, C_filter, dV/dt. Unify active-low/high via small adapter board if mixed; ensure pull-up rail is always present to avoid false PG.

Back-Power & Level Domains

Identify back-power paths created by long harness/FPC and shared pull-ups, then design fan-out buffering and level compatibility to eliminate false stability and half-powered states.

Back-Power Paths

Pull-ups tied to always-on rails, input clamp diodes/ESD arrays, oversized series resistors causing slow charge “fake-PG”.

Fanout & Level

Budget PG fanout current; OD across mixed domains; buffer or translate levels; pick a pull-up rail that is always present.

Test Hooks

Pluggable/shortable pull-ups; I–V back-power sweeps to locate knee points; DVDD-off with VREF/VCM-kept A/B.

Parameter Guideline Notes
I_backpower_max ≤ 10% standby budget Measure with DVDD off, references kept
R_series (IO/PG) 100–470 Ω (typ.) Coordinate with pull-up and ESD model
OD pull-up rail Always-on, level-compatible Avoid DVDD-only in sleep paths
PG fanout budget I_fanout ≤ Σ I_sink_node (≥30% margin) Buffer if wiring is long or multi-drop
Common back-power paths and PG fan-out with level compatibility Left: three typical back-power routes with blocking components; Right: OD pull-up to always-on rail driving multiple nodes with level adapter. Bottom: test hooks icons (jumper, I–V curve, FPC plug). Back-Power Paths & Fan-Out Route A: Pull-up → Clamp Diode Always-on pull-up Clamp D → Off-domain back-power Route B: ESD Array Reverse ESD reverse energy path Route C: Oversized R_series R slow charging → fake PG PG Fan-Out & Level Domains OD driver Pull-up (always-on) Node A Node B Node C Level adapter / buffer if domains differ; PP only within same-level fanout. Budget I_fanout; avoid multi-point pull-ups. Test Hooks Pluggable/shortable pull-ups I–V back-power curve FPC insert/remove cycles

Validation & Tuning

Turn “feels stable” into a data-closed loop: ramp spectra, temperature cycling, sleep-resume A/B, and EMI/glitch immunity, with reproducible KPIs for acceptance.

Ramp Spectrum

0.1/0.5/1/2/5 V/ms; log false-PG, re-reset, start-up jitter.

Temperature Cycling

−40/−10/25/60/85 °C; track black-level curve, FPN RMS, t_resume_s2s.

Sleep A/B

Keep VREF/VCM vs full-off; compare re-calibration cost and noise floor.

EMI/Glitches

Reuse ramp/instant-brownout scripts; validate tPG_filter adequacy.

Metric Target (example) Notes
Δblacklevel_85C (LSB) ≤ X Curve vs temperature
FPN_RMS (LSB) ≤ Y Row/column fixed pattern
t_resume_s2s (ms) ≤ Z Sleep→stable black level
Re-reset / False-PG (ppm) ≤ 10 / ≤ 5 Ramp + transient scripts
Ramp/temperature/resume validation plan and KPIs Four-quadrant experiment grid (ramps, temperature, sleep A/B, back-power & EMI), right-hand metrics column, and bottom KPI pass bands. Validation & Tuning Plan Ramp Spectrum 0.1/0.5/1/2/5 V/ms false-PG, re-reset, jitter Temperature Cycling −40/−10/25/60/85 °C ΔBL curve, FPN_RMS, t_resume Sleep A/B Keep VREF/VCM Full off Compare re-cal cost & noise floor Back-Power & EMI Pull-up removal/swap; transient scripts Metrics Δblacklevel_85C (LSB) FPN_RMS (LSB) t_resume_s2s (ms) Re-reset (ppm) False-PG (ppm) I_standby / I_backpower (µA) KPI PASS band MARGINAL FAIL

Tuning: If false-PG is high → increase tPG_filter, move OD pull-up, add buffer/RC. If ΔBL/FPN exceed targets → tighten VREF/VCM windows, delay DVDD enable, limit dV/dt(AVDD). If resume time is long → keep references or adjust post-reset wait. If back-power is noticeable → raise R_series, add one-way elements/isolation, choose an always-on pull-up rail.

Cross-Brand Shortlist (Procurement-Oriented)

Directional candidates for Imager/AFE rails (VREF/VCM/AVDD/DVDD). Final PNs must match your inventory/BOM to avoid guesswork. Focus on window/threshold accuracy, delay/debounce, OD/PP output strategy, and cooperation with controlled discharge.

Brand Part / Family Rails (n) Window / Delay / Debounce Output (OD/PP) Discharge Control Why It Fits Imager/AFE Docs
TI TPS3851-Q1 (high-accuracy + watchdog)
TPS386596 (quad-rail supervisor)
1 / 4 Tight UV thresholds, selectable reset/WDT delays; quad version supports fixed or divider-set windows OD preferred across mixed domains; PP available in family Use external FET + timing for controlled discharge/soft clamp Narrow windows for VREF/VCM, watchdog-friendly RESET tree, and multi-rail aggregation to gate AVDD/DVDD cleanly. TPS3851-Q1 · TPS386596
ST STM6717/18/19/20 (multi-rail + manual reset)
STWD100 (stand-alone watchdog)
2–3 / — Fixed thresholds, robust reset timing; WDT timeouts selectable OD or PP by variant Pair with external FET for slow/fast discharge policies Practical for splitting VREF/AVDD monitoring and keeping a human-reset path; WDT decouples from PG aggregation to avoid false stability. STM6717 · STWD100
NXP PF5020 (multi-rail PMIC, integrated PGOOD/RESET)
VR5510 (automotive multi-output PMIC)
≥5 (PMIC) Built-in power-up sequence, OV/UV monitors, synchronized PGOOD/RESET masking PGOOD/RESET pins; translate to OD/PP if needed Coordinate with external soft-clamp/discharge switch If rails are PMIC-sourced, reuse its native supervisors to reduce discretes and lock a consistent RESET release point for black-level stability. PF series · VR5510 DS
Renesas ISL88022 (configurable 3-rail)
ISL88013 (dual: one fixed + one adjustable)
3 / 2 Adjustable thresholds, selectable reset delays; valid at low V PP/OD by option (see DS) External discharge path Bind VREF/AVDD/DVDD with different window widths and release delays in one device; reduce re-calibration across temperature. ISL88022 · ISL88013
onsemi NCV809 (ultra-low IQ single-rail)
NCV303 (OD, programmable delay)
1 / 1 Fixed thresholds; NCV303 adds delay to reject slow-slope glitches OD (preferred across domains); PP on some variants External controlled discharge Cost-effective guards for DVDD/IO with long harness/FPC; easy to aggregate into a unified RESET tree. NCV809 · NCV303
Microchip MCP1316 (single-rail, wide temp)
MIC2775 (single-rail, dual PP outputs, MR pin)
1 / 1 Fixed thresholds; typical 280 ms reset (device-dependent); manual reset support PP; OD available in family options External soft-discharge with FET Good for narrow-window, long-debounce guarding near VREF or AVDD with minimal disturbance; MIC2775’s dual PP eases RESET fan-out. MCP1316 · MIC2775
Melexis — (no dedicated supervisor family) In imaging programs typically paired with the supervisor/PMIC options above; keep system-level compatibility only.
  • Automotive: prefer -Q1 / AEC-Q100 variants where available; keep same package/codes for drop-in alternates.
  • Outputs: OD across mixed/long domains; PP allowed only within same-level short fan-out.
  • Discharge: most supervisors lack integrated controlled discharge—pair with eFuse/load switch or external FET for soft-clamp policies.

BOM & Procurement Notes

Required Fields

V_rail (VREF/VCM/AVDD/DVDD…), n_rails, threshold tolerance/window, t_delay/debounce, discharge policy (off/slow/fast), output (OD/PP), AEC-Q100, package height, second-source (Y/N).

Optional Fields

I²C/PMBus, PG/FAULT semantics, back-power notes, dV/dt requirements, temperature coefficients, production test hooks (ramp/thermal/sleep A/B).

Risks & Mitigations

  • Pin/semantics mismatch → adapter board (OD↔PP or polarity), RESET-tree decoding.
  • EOL/NRND → dual-source plan and transition BOM; keep package/thresholds aligned.
  • Sample lead time / MOQ → lock samples early (same lot if possible) and keep drop-in alternates.
Submit your BOM (48h cross-brand)
Reference Docs (quick check)
TI: TPS3851-Q1 · TPS386596
ST: STM6717 · STWD100
NXP: VR5510 DS · PF series
Renesas: ISL88022 · ISL88013
onsemi: NCV809 · NCV303
Microchip: MCP1316 · MIC2775

FAQs

Why is black level highly sensitive to power-up order?
VREF/VCM must settle before AVDD/DVDD so the AFE’s bias network and offset memory never see half-powered domains. If AVDD or DVDD rises first, pre-bias and leakage shift the dark reference and create persistent banding. Use VREF→AVDD→DVDD enables with held RESET until all-good. Acceptance: Δblacklevel_cold ≤ X LSB and re-reset ≤ 10 ppm.
Should VREF/VCM be retained in standby or powered down?
Retaining VREF/VCM gives µs-class resume and stable black level; powering down saves leakage but forces re-calibration and risks drift during warm starts. Default to retain VREF/VCM and gate AVDD/DVDD. For deep sleep, use controlled discharge with guard timers. Acceptance: t_resume_s2s ≤ Z ms and Δblacklevel_resume ≤ X LSB across −40~+85 °C.
Soft clamp vs fast discharge FET—how do I choose?
Use soft/controlled discharge on VREF/VCM to preserve bias and avoid offset collapse; allow fast discharge on DVDD and some AVDD loads when data retention is not required. Implement discharge Rds(on) and windowed timing per rail. Default matrix: VREF/VCM=retain/slow, AVDD=slow, DVDD=fast. Acceptance: no offset jump after power-down; E_hold within budget.
On mixed-voltage boards, is OD or PP safer for PG?
Prefer open-drain PG across mixed domains, long harnesses, and FPCs; pull up to a known always-on rail. Use push-pull only within a single level and short fan-out. Add RC or digital filters at the aggregator and enforce minimum pulse width. Acceptance: false-PG ≤ 10 ppm across 0.1–5 V/ms ramps and EMI tests.
How much temperature margin do window thresholds need?
Budget the sum of sensor/AFE tempcos, supervisor accuracy, divider drift, and wiring drops, then add guard margin. VREF/VCM windows should be narrow yet cover worst-case ppm/°C across −40~+85 °C. Start with ≥ ±(aggregate ppm/°C × ΔT) + safety. Acceptance: no nuisance trips; window tracking error ≤ X ppm/°C equivalent.
How do I prove slow ramps don’t false-trigger?
Sweep a ramp spectrum (0.1, 0.5, 1, 2, 5 V/ms) and log PG/RESET chatter, re-arm latency, and start-up jitter. Combine debounce time with slope immunity—debounce alone misses slow-edge metastability. Add hold-off on “all-good.” Acceptance: zero false resets across spectrum; tPG_filter ≥ 0.5–1 ms; jitter within Y ms.
How do I tag first-fault/last-good in multi-rail PG without chatter?
Latch the first-fault source and release only when all rails are back in window plus a hold-off. Use OR-tree with RC or digital filters and minimum pulse enforcement. Provide timestamped “first-fault” metadata if available. Acceptance: no oscillation at window edges; max chatter count = 0 over test ramps and temperature.
How do I quantify black-level step after standby and set a pass line?
Capture N dark frames immediately after resume with VREF/VCM retained vs fully powered off. Compute mean black-level step and FPN_RMS change. Define a per-SKU acceptance envelope and gate firmware recalibration accordingly. Acceptance: Δblacklevel_resume ≤ X LSB, FPN_RMS change ≤ Y LSB, with resume time ≤ Z ms.
How do I locate and isolate back-power from long cables or FPC?
Perform DVDD-off A/B tests, swap or short the pull-ups, and measure I–V to find the diode knee through protection paths. Insert series resistors, ideal-diode elements, or level isolators on IO nets. Pull OD lines to an always-on rail. Acceptance: I_backpower_max ≤ budget and no false “all-good” with DVDD removed.
How do I migrate from single-rail to multi-rail windowing without a respin?
Add a small adapter board that provides adjustable thresholds, OD pull-ups to an always-on rail, and a PG aggregator. Reuse the existing reset net and preserve pulse width and polarity. Validate across slope and temperature before locking. Acceptance: no baseboard ECOs; RESET timing preserved; zero regressions in start-up jitter.
When is I²C programmability mandatory instead of fixed hooks?
Use I²C when thresholds or delays must be trimmed per sensor, temperature, or SKU, when logs/telemetry are required, or when a single PCB must cover multiple variants. Prefer OTP-locked settings for production after characterization. Acceptance: #SKUs covered with one PCB ≥ target and calibration time within factory takt.
What AEC-Q100 and PPAP pitfalls affect supervisors in automotive projects?
Match device grade to ambient and verify ESD/latch-up margins early. Confirm PPAP package completeness and lot traceability, and monitor PCNs that shift thresholds or timings. Keep a validated alternate. Acceptance: grade matrix approved for locations, PPAP on file, and second-source build passes correlation without window retuning.

Need help mapping rails, policies, and parts for your camera board? Submit your BOM (48h).