Inverter Refrigerator: PMSM/BLDC Drive, Sensing & Debug Playbook
← Back to: Smart Home & Appliances
An inverter refrigerator stabilizes temperature and lowers noise/energy by driving the compressor and fans at variable speed, using sensor feedback and event logs to prove root cause when field issues occur. This page focuses on the appliance-level evidence chain—motor drive, sensing, defrost, metering, UI immunity, and power integrity—so failures can be isolated with minimal measurements and fixed reliably.
H2-1 Featured Answer + System Boundary (Definition)
An inverter refrigerator uses a variable-speed compressor and electronically commutated fans driven by a 3-phase inverter. Sensor feedback (temperature, door events, frost/defrost state, and electrical limits) continuously adjusts speed to reduce hard cycling, audible noise, and energy use while improving temperature stability and fault protection.
This page covers
- Motor drive chain: 3-phase inverter, gate driver/IPM choices, current/voltage sensing, protections, and start/ramp behavior.
- Sensing chain: temperature nodes, door events, frost/defrost sensing, wiring/ADC robustness, and sanity checks.
- Defrost power path: heater switching, interlocks, typical failure signatures, and fast proof measurements.
- Device-level metering & logs: kWh/runtime counters for diagnostics (trend + service evidence, not utility-grade standards).
- UI reliability: display/touch noise immunity and power integrity coupling during compressor transients.
Not covered
- Home protocol stacks (Matter/Thread/Zigbee/Wi-Fi) and gateway architecture.
- Whole-home HEMS design and utility meter anti-tamper standards.
- Certification walkthroughs (EMC/safety compliance procedures).
Scope control prevents repeating “shared modules” that belong to other pages. This page stays inside the refrigerator controller coupling.
First 2 measurements (fastest split of root causes)
Proves whether the inverter power domain is stable. Large sag or high ripple commonly explains: start-then-trip, random resets, or harsh acoustic bursts at ramp.
Proves whether the load/control is abnormal: mechanical binding, commutation mismatch, deadtime issues, or false over-current triggers from noisy sensing/reference.
If UI resets/ghost touches are reported, add P3 (logic rail droop) as the third capture only after P1/P2 classify the failure surface.
H2-2 Top-Level Architecture (Power Tree + Motor + Sensors + UI)
The inverter refrigerator is a coupled system: a high-di/dt motor inverter shares ground, rails, and harness paths with sensors and the UI panel. Reliable design and fast debugging start from a single map that shows power domains, switching loads, sensing nodes, and where evidence is captured.
Power & drive chain (functional blocks, appliance-level)
- AC input → internal rails: logic rail(s), driver rail, sensor/reference rail. Rail stability decides whether faults look like “control” or “power”.
- DC bus (bulk cap): ripple/sag during compressor start is the fastest indicator of margin under real load.
- 3-phase bridge + gate drive/IPM: switching timing + protections (OCP/UVLO/OTP) convert electrical stress into clear fault signatures and logs.
Primary loads
- Compressor PMSM/BLDC: the dominant power transient source; start/ramp behavior determines noise and nuisance trips.
- Evaporator & condenser fans: PWM/commutation noise can leak into sensing and touch baselines when reference routing is weak.
- Defrost heater path: high-power switching creates distinct signatures (current step + thermal rise) used to confirm heater/relay/SSR health.
Sensing, UI, and evidence outputs
- Temp nodes (NTC) + frost/defrost sensing: placement and harness integrity determine whether temperature control is stable or oscillatory.
- Door event chain: switch/Hall bounce and moisture/corrosion create false events that amplify energy use and frost formation.
- UI panel (display/touch/backlight): often the first visible failure surface when logic rails dip or ground bounces during inverter transients.
- Device-level metering + logs: kWh/runtime/defrost-energy counters and fault timestamps convert “intermittent” behavior into a reproducible diagnosis trail.
The architecture map below is used as a page index: each later chapter anchors to one chain (drive, sensing, defrost, metering, UI) and returns to a probe point (P1–P4).
H2-3 Motor Drive Deep Dive (PMSM/BLDC + Inverter Stage)
This chapter focuses on the “engine room”: the DC bus, 3-phase bridge, gate drive, sensing, and protection logic that determine start success, acoustic noise, efficiency, and fault behavior. The goal is design-and-debug depth with concrete waveforms and probe points, without drifting into protocol stacks or certification walkthroughs.
3-phase inverter implementation options (what changes in failure behavior)
- Discrete MOSFETs + gate driver: highest flexibility; performance depends heavily on loop inductance, gate tuning, deadtime, and current-sense cleanliness. Typical field failures: false OCP, harsh ramp noise, UI resets from dv/dt coupling.
- IPM (integrated power module): protections are more “self-contained” (OTP/UVLO/OCP often internal). Typical field failures: repeated stop-restart patterns from internal thermal/UVLO events that look like “intermittent compressor.”
- Design decision knob: choose based on required acoustic margin, peak current, thermal path, and service visibility (how clearly logs can distinguish UVLO vs OCP vs OTP).
Control modes (BLDC trapezoidal vs PMSM FOC — what changes in sensing + acoustics)
- BLDC trapezoidal: simpler commutation; more torque ripple at low speed; acoustic “bands” can be more pronounced. Evidence focus: phase current steps, commutation edge timing, and back-EMF/zero-cross stability.
- PMSM FOC (sensorless or with feedback): smoother torque when sampling and estimation are stable; more sensitive to current-sense noise, timing skew, and deadtime compensation. Evidence focus: current ripple shape during ramp, estimator instability signatures, and “limit hits” (Iq clamp) in logs.
- Practical takeaway: the same mechanical load can look like “motor problem” under BLDC but “sensing/timing problem” under FOC; classification depends on P2 current waveform + gate timing + state logs.
Protection primitives (trigger → visible symptom → proof → first fix)
- OCP (over-current): Trigger peak current/di/dt exceeds threshold. Symptom start then immediate stop, or periodic cutouts at certain RPM. Proof P2 shows clipped current + fault flag; P1 bus often stays OK. First fix gate slew/deadtime tuning, shunt layout/Kelvin routing, blanking/filters matched to switching edges.
- DC bus UVLO/OVP: Trigger bus sag at start or surge events. Symptom repeated restart attempts, sometimes UI brownout. Proof P1 sag aligns with trip; logs show UV/OV counters. First fix increase bus hold-up margin (cap ESR/placement), reduce ramp aggressiveness, improve input rail stability.
- Shoot-through prevention: Trigger deadtime too small, Miller turn-on, ringing. Symptom loud click/whine, instant OCP, hot devices. Proof gate timing overlap on scope; abnormal current spikes at each edge. First fix gate resistors, split-gate, deadtime calibration, snubbing where appropriate.
- Desaturation / short detection (IGBT/IPM only): Trigger device fails to saturate under load. Symptom abrupt shutdown under heavy start. Proof desat flag with normal bus; current jumps then cuts. First fix verify device SOA, thermal path, and wiring shorts; validate desat blanking timing.
- Thermal foldback / OTP: Trigger junction/module temperature rises. Symptom works cold, fails after minutes/hours; reduced torque or frequent stops. Proof temperature telemetry aligns with derate; current reduces before stop. First fix heatsink interface, airflow path, switching loss reduction (frequency/gate tuning), derate thresholds validated in hot ambient.
EMI / noise couplings (device-level only)
- dv/dt → ground bounce: fast switching edges inject common impedance noise into sensor references; can distort NTC ADC codes or trip false limits.
- dv/dt → UI resets / ghost touches: rail droop (P3) or touch reference drift (P4) correlates with compressor ramp or specific PWM bands.
- Evidence rule: confirm time alignment: switching edge bursts → P3 dip or P4 drift → UI event (reset/ghost). Avoid “EMI guesswork” without a captured alignment.
Evidence blocks (healthy vs bad signatures)
Healthy: clean rise/fall, stable deadtime, limited ringing. Bad: large ringing, Miller-induced spurs, or overlap hints → correlates with shoot-through/OCP.
Healthy: controlled peak at start, ripple changes smoothly with speed/load. Bad: sharp spikes at edges, clipped/oscillatory ripple, or periodic torque ripple bands → correlates with noise and trips.
Healthy: modest sag at ramp, ripple within expected envelope. Bad: deep sag triggers UVLO/retries; ripple bursts can corrupt sensing and provoke false limits.
H2-4 Control Loop & Operating States (Start, Ramp, Steady, Defrost Transitions)
Many “field-only” issues are timing and state-machine problems: the same hardware behaves differently across cold start, speed ramps, steady regulation, and defrost transitions. This chapter maps each state to its control objective, the most informative measurements, and the log evidence required to prove root cause.
Operating state timeline (bullet format; each stage ties to evidence)
- State A — Cold start: objective is reliable spin-up without OCP/UVLO. Evidence: P1 sag envelope + P2 peak/edge spikes; logs: start-attempt counter, UVLO/OCP reason code.
- State B — Ramp: objective is smooth acceleration with acceptable acoustics. Evidence: P2 ripple shape vs RPM; gate timing stability; logs: speed target vs achieved, limit hits (Iq clamp / current limit).
- State C — Steady regulation: objective is temperature stability with minimal cycling. Evidence: temperature error vs compressor speed commands; logs: duty/runtime counters, door-event correlation.
- State D — Idle / low-speed: objective is avoid short cycling while maintaining setpoint. Evidence: short-cycle count; logs: hysteresis/threshold crossings and minimum-off/on timers.
- State E — Defrost transition: heater ON → compressor OFF/low → drain/settle → resume cooling. Evidence: heater current step + P1 disturbance; restart margin right after defrost. Logs: defrost energy/time, termination condition (temp/time), post-defrost restart outcome.
Fan & damper coupling (how “temperature issues” become drive issues)
- Fan speed ↔ heat exchange: fan PWM changes evaporator/condenser heat transfer, shifting compressor load and current profile; this can move the system into noise or limit bands.
- Damper position ↔ zone temperature lag: airflow redistribution creates delay; poor tuning causes overshoot/oscillation that looks like “unstable compressor,” but the evidence is temperature error leading the speed command.
- Proof pattern: temperature error spikes → speed command step → P2 current change. If P2 spikes occur without a command change, suspect power/drive integrity (return to H2-3 evidence).
Fault logic: retries, lockout, service codes, and event logs (make intermittent faults provable)
- Retry policy: define retry delays and max attempts per fault class; aggressive retries can worsen brownout and UI instability during weak mains.
- Lockout: separate “hard faults” (sustained OCP/OTP) from “recoverable faults” (temporary UVLO). Lockout conditions must be visible in logs to avoid ambiguous service returns.
- Service codes (classification): group by chain: drive (OCP/shoot-through), bus (UV/OV), sensing (open/short/out-of-range), defrost (heater/termination), UI (brownout/watchdog).
- Event log minimum fields: timestamp, operating state, fault reason, bus V snapshot, current snapshot, temperature snapshot, and retry count.
Evidence-first rule: a fault label is not a root cause until it is aligned with a state and a probe capture (P1/P2/P3) or a log snapshot.
State → “first capture” mapping (fastest way to avoid blind swaps)
Capture P1 + P2 at the first 500 ms. If P1 collapses first → bus/rail margin. If P2 spikes with stable P1 → load/control/drive timing.
Capture P2 ripple vs RPM (and gate timing if available). If ripple grows periodically at a fixed RPM band → torque ripple/tuning/estimation. If it coincides with bus ripple bursts → bus impedance/edge coupling.
Capture P3 logic rail during compressor ramp and heater switching. If P3 dips align with events → rail distribution/hold-up; if P3 is clean, check P4 reference drift (touch/ground bounce).
H2-5 Sensing & Calibration (Temp / Door / Defrost / Optional Pressure)
Reliable sensing in an inverter refrigerator is not “just picking NTCs.” The sensor chain must remain trustworthy under inverter dv/dt, fan PWM, heater switching, and condensation. This chapter turns sensing into a debuggable system: node placement, harness noise paths, ADC timing/filtering, and calibration rules that produce clear evidence in logs.
Temperature sensing (NTC placement, harness noise, ADC timing, filtering)
- Placement defines time constant: the same NTC type can read “stable” or “laggy” depending on how it is coupled to air vs coil vs plastic. Treat placement as part of the transfer function, not a mechanical afterthought.
- Harness noise is often dominant: inverter switching and fan PWM inject common-impedance noise into sensor references. Symptoms include “temperature jitter” synchronized with compressor ramp or specific PWM bands.
- ADC timing matters: sampling aligned to switching edges can convert dv/dt into code ripple. Prefer sampling windows away from edge bursts (or use synchronized sampling) when possible.
- Filtering must be state-aware: too much filtering hides door-open thermal events and delays control response, increasing overshoot and energy. Use different filter strengths for Start/Ramp vs Steady states.
- Sanity constraints beat guesswork: enforce plausible range + plausible slope (rate-of-change). A code “jump” without a matching physical event is a strong indicator of wiring/ground/reference issues.
Door sensing (reed/Hall bounce, condensation, false-open) and system impact
- Reed/Hall bounce: mechanical vibration and magnet alignment cause edge chatter that inflates event counters and triggers unnecessary control transitions.
- Condensation/corrosion: moisture on connectors can create leakage paths and intermittent levels, appearing as random “open” events after seasonal humidity changes.
- Why false-open matters: door state affects fan strategy, compressor targets, lighting, and frost formation. A noisy door signal can indirectly increase defrost frequency and energy.
- Evidence rule: compare door events against a physical response (temperature slope change, fan command change). If events occur without any thermal or command correlation, treat them as suspect.
Defrost / frost sensing (coil temperature, time-based vs sensor-based)
- Time-based defrost: simple to implement but sensitive to real-world variation (humidity, door-open frequency, airflow restriction). Failures appear as under-defrost (ice buildup) or over-defrost (energy spikes, warm swings).
- Sensor-based defrost (coil NTC): adapts to conditions but depends on thermal coupling and noise immunity. Misplacement can shift termination and create “looks normal in lab” failures.
- Termination evidence: log the termination reason (temperature threshold, max time, safety cutoff). Without this field, service returns become ambiguous.
- Cross-check principle: coil temperature rise curve must be consistent with heater current evidence (H2-6). If current exists but coil temperature does not rise, suspect coupling or sensor chain.
Calibration & self-check strategy (offset tables, power-up checks, open/short detection)
- Offset tables: use simple per-channel correction tables (or two-point trims) to compensate sensor tolerance and placement bias. Keep corrections traceable (store revision + checksum).
- Power-up self-check: detect open/short by code rails; verify reference health by checking if multiple channels move together abnormally.
- Degrade modes: when a sensor is invalid, switch to a conservative fallback (limit compressor speed, force safer defrost scheduling, constrain fan commands) and record the reason.
- Minimum log snapshot: timestamp + operating state + raw ADC code + filtered value + validity flag + open/short classification.
Evidence blocks (fast checks that separate real thermal change from noise)
Healthy: a door-open event is followed by a plausible temperature slope change (with expected lag). Suspect: frequent door events with no thermal response and no matching control command change.
Healthy: raw codes remain within an expected band for the operating temperature. Suspect: code ripple synchronized with compressor ramp or fan PWM, or abrupt jumps without physical triggers.
Healthy: sensors change independently based on zone physics. Suspect: multiple NTC channels jump in the same direction at the same time → check reference, ground bounce, and sampling timing.
H2-6 Defrost Heater & High-Power Loads (Relays/SSR, Safety Interlocks)
Defrost and auxiliary heaters are common sources of “mystery overheat,” “random trips,” and post-defrost restart failures. This chapter maps the heater control chain, the interlocks that must be logged, and the evidence captures that distinguish stuck relays, triac/SSR leakage, and rail disturbances that cascade into resets.
Heater control chain (relay / triac / SSR drive, switching behavior, inrush)
- Control path: MCU GPIO → driver transistor → relay/triac/SSR → heater load. If current sensing exists, route it into logs as a “heater-on proof.”
- Relay: clear on/off behavior and low leakage; risks include contact wear and sticking under high inrush or arcing.
- Triac/SSR: compact and silent; risks include leakage current and false triggering from dv/dt if layout and snubbing are not robust.
- Zero-cross (if used): can reduce stress and EMI at turn-on; debug still requires proving alignment between commanded ON and measured current.
- Inrush consideration: heater switching can create rail disturbances; correlate heater current step with P1/P3 captures to avoid mislabeling the root cause.
Interlocks (door inhibit, over-temp cutoff, thermal fuse logic — device-level)
- Door-open inhibit: prevents heater activation under unsafe airflow/condition assumptions. A noisy door signal can repeatedly block defrost; log the exact inhibit reason.
- Over-temp cutoff: stop heater if coil or compartment temperature exceeds a threshold. Log both the threshold hit and the sensor used.
- Thermal fuse / thermostat: hardware last-line safety. If it opens, firmware should record the “no-current” evidence rather than guessing heater failure.
- Required log fields: command ON/OFF, interlock status, measured heater current (if available), termination reason, and retry count.
Drain heater / anti-sweat heater (when it runs, power impact, visibility)
- Drain heater: supports water evacuation after defrost; failures appear as refreeze, water pooling, or repeated frost cycles.
- Anti-sweat heater: often a background load; leakage or “always-on” behavior can look like a slow energy creep and localized warming.
- Visibility rule: track runtime and energy contribution per heater class to avoid “mystery energy” complaints and misdiagnosis.
Evidence blocks (current + temperature prove the chain)
Healthy: a clear current step at ON with stable steady-state. Fault A: no current (open load, relay not actuating, interlock blocking). Fault B: abnormal low or noisy current (triac/SSR issues, supply constraints).
Healthy: coil temperature rises with a plausible profile during defrost. If current exists but temperature does not rise, suspect poor thermal coupling, heater detachment, or sensing chain errors (return to H2-5).
Stuck relay: current continues after OFF and temperature keeps rising. Triac/SSR leakage: small persistent current leading to unexpected energy and mild warming over long periods.
H2-7 Energy Metering (Device-Level kWh & Diagnostics)
Device-level energy metering in an inverter refrigerator is for trends, diagnostics, and service evidence—not billing-grade utility standards. This chapter defines what to measure, where to tap, how to sample for event correlation (compressor start, door cycles, defrost), and how to store “energy stories” that explain spikes and abnormal shares.
What to measure (answers “why did kWh rise?” without becoming a utility-meter page)
- Total input energy: kWh/day and rolling averages for user-visible trends and baseline comparison.
- Compressor energy share: identify abnormal load/efficiency behavior (more kWh with no thermal benefit) and short-cycling patterns.
- Defrost energy share: quantify spikes and verify whether defrost duration/triggering is reasonable.
- Optional runtime counters: fan/backlight/heater on-time provide high diagnostic value even when not individually metered as kWh.
Where metering taps go (shunt/CT position and what each can explain)
- AC input tap (trend anchor): best for total kWh and spike correlation. Supports door/defrost correlation directly.
- Heater branch tap (defrost proof): strongest evidence for defrost energy and stuck/leak faults. Reuse the heater current evidence chain from H2-6.
- DC bus / inverter-side tap (if available): improves compressor energy attribution but does not represent total appliance energy by itself.
- Shunt vs CT (device-level): shunt enables detailed transient capture but needs good reference/ground routing; CT is isolation-friendly for AC trend capture but less precise at low power and transient detail.
Sampling for purpose (trend + event correlation, not billing)
- Event-aligned windows: capture higher-rate samples during compressor start/ramp and heater switching; use slower averaging during steady operation.
- Integrate by state: accumulate ΔE by operating state (Start/Ramp/Steady/Idle/Defrost) to make shares explainable.
- Minimum energy event record: timestamp, event type, operating state, ΔE, duration, and snapshots (bus/rail status and key temperatures).
- Spikes must be attributable: an energy spike is only “explained” when it time-aligns with defrost/heater, compressor ramp, or a burst of door events.
Accuracy vs purpose (what “good enough” means for diagnostics)
- Billing-grade is not required: the target is consistent trends and reliable event correlation on the same hardware platform.
- Use anchor events: heater ON/OFF and known operating states are practical anchors to validate that metering responds correctly.
- Share-based diagnostics: large shifts in compressor or defrost share are meaningful even if absolute error exists (e.g., defrost share rising from low single digits to double digits).
- Boundary reminder: no deep anti-tamper, PLC, or utility metering standards in this page.
Evidence blocks (turn kWh into service-ready proof)
Healthy: spikes align with defrost/heater windows or compressor ramp windows. Suspect: repeated spikes with no matching events → investigate rail disturbances, false state triggers, or metering tap noise.
If defrost energy share rises while defrost frequency increases, verify heater current evidence (H2-6) and coil temperature response (H2-5) to separate real ice load from leakage/stuck switching.
If compressor energy rises but temperature stability does not improve, suspect abnormal load behavior or control inefficiency; confirm with state-aligned runtime counters and current profiles (H2-3/H2-4).
H2-8 Display & Touch Integration (Noise Immunity + UX Reliability)
UI failures are common in inverter appliances because display, backlight, and capacitive touch are sensitive to rail dips, ground bounce, and switching noise. This chapter maps UI power domains, touch baselines, and watchdog behavior, then provides evidence captures that explain ghost touches, frozen displays, and random reboots during compressor ramps.
Display types and backlight drivers (reliability-focused)
- Segment LCD: low power and often robust, but can flicker or show missing segments if bias/reference stability is poor.
- TFT / color display: more rails and tighter timing; more vulnerable to brownout and reset loops if power sequencing is marginal.
- Backlight drivers: switching regulators can inject noise into touch/reference paths. Treat backlight as a noise source that must be isolated from touch sensing.
Capacitive touch keys (baseline drift near noisy inverter grounds)
- Baseline drift mechanism: dv/dt and ground bounce shift the touch reference, moving the baseline and creating false triggers.
- Noise timing matters: ghost touches often align with compressor start/ramp bands or heater switching edges.
- Practical robustness: stabilize reference paths, constrain baseline tracking speed, and avoid sampling at the noisiest moments.
- Evidence rule: a touch claim is credible only when baseline drift aligns with a noise or rail event (not just “users reported it”).
UI power domains (keep-alive rail, brownout behavior, watchdog)
- Separate domains: UI logic rail, backlight rail, and touch/reference domain should be treated as distinct noise and brownout risks.
- Brownout behavior: define whether UI should blank safely, restart quickly, or preserve state. Log brownout events with voltage snapshots.
- Watchdog: useful for recovery, but frequent watchdog resets indicate unresolved rail integrity or EMI coupling; log reset cause and state to avoid blind part swaps.
Common symptoms → fastest evidence captures
Capture touch baseline drift (P4) while logging compressor/heater state. If drift time-aligns with inverter PWM bands or load steps, treat it as a reference/noise issue rather than a “bad panel.”
Capture UI/logic rail behavior (P3) during compressor start. If rail dips precede freezes, address sequencing/hold-up. If rails are stable, investigate interface timeouts and reset cause logs.
Capture P3 dips and classify reset reasons (BOR vs watchdog). A BOR signature with ramp correlation is a power integrity issue until proven otherwise.
H2-9 Power Integrity & Thermal Management (Appliance-Level, Not PSU Topologies)
Power integrity in an inverter refrigerator is about keeping control, sensing, and drive rails stable during motor transients and mains disturbances—without drifting into PSU topology design. This chapter maps typical rails, defines DC bus ripple and hold-up expectations, and links thermal coupling to derating, torque limits, and repeatable fault trips.
Typical rails and roles (stability targets by domain)
- Logic rail (3.3V / 5V): MCU, UI, comms, logging. Instability shows as BOR resets, watchdog loops, frozen UI, or corrupted event timelines.
- Gate-driver supply: discrete driver or IPM internal supply. Instability shows as abnormal gate timing, increased switching loss, torque ripple, and protection mis-trips.
- Sensor reference rails (Vref / AVDD): ADC reference and analog front-end supply. Instability shows as multi-channel “move together” drift (temperature, metering, touch baseline).
- Aux rails (backlight / small actuators): can be small loads but strong noise injectors; isolate their switching impact from touch and sensing references.
DC bus ripple & hold-up expectations (during ramp and mains sag)
- Ripple triggers: compressor ramp, abrupt torque demand changes, heater switching, and short mains sag events.
- Hold-up goal (device-level): short disturbances should not destabilize logic rails or cause repeated protection trips. The objective is “controlled behavior,” not uninterrupted operation.
- Correlation rule: if UI resets or state-machine anomalies align with bus ripple bursts, treat DC bus impedance/energy storage as a prime suspect until rail evidence disproves it.
- Domain sensitivity: the weakest rail (logic/Vref/driver) determines field reliability; diagnose per rail rather than assuming “the PSU is bad.”
Transient coupling paths (how motor events become user-visible failures)
- Compressor ramp → bus ripple → logic droop: leads to BOR resets, frozen UI, random reboots, and event log gaps.
- dv/dt → ground bounce → Vref drift: produces sensor code ripple and false state triggers (door/defrost/energy shares become misleading).
- Driver rail instability → abnormal switching: increases current ripple and heat, leading to protective trips and retry loops.
- Backlight/Fan PWM noise: can modulate touch and reference baselines; correlate to UI symptoms before replacing panels.
Thermal coupling (module heat → derating → torque limit → fault patterns)
- Heat source chain: compressor and inverter module heat raises IPM/driver temperature, pushing derating behavior.
- Derating outcomes: torque limit, speed limit, longer ramp times, increased cycle time, and seasonal “only fails in summer” behavior.
- Fault-trip alignment: repeated trips that cluster at high temperature are rarely random; the evidence is IPM temperature vs trip timestamps.
- Recovery behavior: log the derating reason and thermal snapshot so service can distinguish “thermal limit reached” from “electrical anomaly.”
Evidence blocks (scope points and diagnostics that prove root cause)
Capture ripple during compressor start/ramp and during heater switching. A ripple burst that time-aligns with state anomalies is a strong indicator of bus impedance/energy storage limitations.
Capture logic rail droop and compare to brownout thresholds and reset cause logs (BOR vs watchdog). BOR-aligned reboots indicate rail integrity issues before software suspicion.
Trend IPM temperature (or module NTC) against protective trips and torque limiting. Temperature-aligned trips point to thermal coupling and derating rather than random failures.
H2-10 Validation Test Plan (What to Test to Catch Field Failures)
This SOP-style validation plan targets failures that pass the lab but fail in the field: reboot loops during ramps, defrost anomalies, false door events, drifted sensing, and temperature-aligned derating trips. Each test item specifies stimulus, observables, pass/fail rules, and minimum log fields—without relying on wide tables that break mobile layouts.
Functional stress (state-machine coverage)
-
F-01 Start/Stop cycle stress
Stimulus: repeated compressor start/stop with short idle gaps.
Observables: state transitions, retries/lockouts, P1 ripple, P3 droop, reset cause.
Pass/Fail: no state lock, no repeated BOR resets, no escalating retry counter.
Logs: state timeline, start count, retry count, reset cause, snapshots (T sensors, bus/rail flags). -
F-02 Defrost cycle coverage
Stimulus: force defrost entry, run to termination, then resume cooling; include mid-cycle interruptions if supported.
Observables: heater current proof (P7), coil temperature rise, termination reason, restart behavior.
Pass/Fail: termination reason recorded; heater proof aligns with coil response; resume without trip loops.
Logs: defrost count, termination reason, ΔE_defrost, coil temp curve, heater ON duration. -
F-03 Door open/close burst
Stimulus: rapid open/close sequences and long-open holds.
Observables: door event debounce integrity, temperature slope response, fan strategy changes, defrost scheduling impact.
Pass/Fail: no false-open bursts; no unexplained defrost triggers; event counters remain plausible.
Logs: door event counter, door valid flag, temperature snapshot, fan command state, defrost trigger source.
Electrical robustness (device-level disturbances)
-
E-01 Brownout / mains sag observation
Stimulus: controlled supply sag events (short and moderate), aligned to compressor start and steady states.
Observables: DC bus ripple (P1), logic rail droop (P3), reset cause classification (BOR vs watchdog), recovery time.
Pass/Fail: no uncontrolled reboot loops; recovery is deterministic; logs remain coherent.
Logs: reset cause, sag event counter, P1/P3 snapshots, state at event, restart outcome. -
E-02 Load-step disturbance from relay switching
Stimulus: switch high-power loads (defrost heater, auxiliary heaters) while monitoring control rails.
Observables: P7 heater current, P3 droop, UI stability, event alignment to heater commands.
Pass/Fail: no UI freeze; no false sensor spikes; heater proof matches command.
Logs: heater command, heater proof, termination reason, rail dip flags, UI reset flags. -
E-03 Surge observation (device-level)
Stimulus: observe behavior under realistic transient stress conditions without turning into certification procedure.
Observables: protection events, retries/lockouts, rail dip flags, sensor validity flags.
Pass/Fail: protective behavior is logged and bounded; no silent corruption of logs or persistent sensor invalid state.
Logs: trip type, trip count, recovery count, state at trip, rail flags, sensor validity.
Thermal stress (catch temperature-aligned failures)
-
T-01 High ambient endurance
Stimulus: run at high ambient for extended periods, including defrost cycles and door events.
Observables: IPM temperature (P9), derating flags, torque/speed limits, trip clustering.
Pass/Fail: derating is explainable and logged; no uncontrolled trip loops at high temperature.
Logs: T_ipm, derating reason, speed/torque limit flags, trip timestamps. -
T-02 Blocked airflow simulation
Stimulus: restrict airflow path (controlled) to simulate poor ventilation and condenser fouling behavior.
Observables: thermal rise rate, compressor runtime increase, energy share shift, fan resonance.
Pass/Fail: system degrades gracefully with logged reasons; no hidden overheating without evidence.
Logs: runtime counters, kWh/day trend, E_comp share, fan state, thermal snapshots.
Acoustic/noise (torque ripple and resonance bands)
-
A-01 Torque ripple check
Stimulus: sweep compressor speed across operating range (including known ramp patterns).
Observables: phase current ripple signatures, audible bands, vibration peaks, protection sensitivity.
Pass/Fail: no unstable resonance bands in normal range; current ripple remains within expected pattern bands.
Logs: speed command, current ripple metric (or proxy), trip flags, timestamp alignment. -
A-02 Fan resonance scan
Stimulus: sweep fan PWM/speed and monitor UI/touch stability simultaneously.
Observables: resonance peaks, touch baseline drift, display flicker events, rail noise flags.
Pass/Fail: no repeatable ghost touch at specific PWM bands; UI remains stable during sweep.
Logs: fan PWM, touch baseline metric, UI event flags, rail dip flags.
Logging & acceptance thresholds (what proves the test really passed)
- Minimum counters: start count, defrost count, door event count, trip count, reset count.
- Minimum classifications: reset cause (BOR/WDT/other), trip type, termination reason (defrost end / safety cutoff / timeout).
- Minimum snapshots: key temperatures, P1 bus ripple flag, P3 rail droop flag, P7 heater proof, P9 IPM temperature.
- Acceptance framing: define bounded behavior (no uncontrolled loops, no silent corruption, recoveries are deterministic) and require logs that explain any protective action.
H2-11 Field Debug Playbook (Symptom → Evidence → Isolate → Fix)
Fast field diagnosis for inverter refrigerators should converge with minimal tools: two measurements first, a discriminator that proves the root-cause class, a first fix that restores stable operation, and a validation step that prevents repeat returns. The scope is device-level motor drive, sensing, defrost, metering, and UI robustness (no home protocol stacks and no PSU topology deep-dives).
Probe Point Legend (use consistent labels across all symptoms)
Minimal tool kit (practical field set)
- DMM for rails and sensor sanity (open/short, connector integrity).
- 2-ch scope for P1/P3/P6 timing alignment (use differential probe on high-side nodes where required).
- Clamp meter / small power meter for P8 trend and event correlation (ramp/defrost/door cycles).
- Thermal probe (IR thermometer or thermocouple) for P9 clustering vs trip timestamps.
Field Swap Shortlist (MPN examples; verify ratings/agency approvals)
- 3-phase drive (gate driver / motor driver IC): Infineon 6EDL04I06NT, TI DRV8323RS, TI DRV8301.
- IPM / integrated power module (600V class examples): Infineon CIPOS™ Mini IM111-L6E2, ST STGIPN3H60, onsemi SLLIMM™ NBMNS3D3N60 (family example).
- Current sense amplifier (shunt-based): TI INA240A1, TI INA181A1, ADI AD8418A.
- Hall current sensor (if used): Allegro ACS724, Allegro ACS772.
- Brownout / reset supervisor: TI TPS3808, Microchip MCP1316, onsemi NCP302.
- Buck regulator for logic rails (device-level examples): TI TPS62133, MPS MP1584EN, onsemi NCP1587 (family example).
- NTC temperature sensor (10k examples): Murata NCP18XH103F03RB, Vishay NTCLE100E3103JB0, TDK/EPCOS B57861S0103F040.
- Door sensing (Hall / reed examples): Allegro A1104 (Hall switch), Melexis US1881 (Hall latch), Littelfuse 59170 (reed switch family).
- AC load switching (defrost heater chain): Triac ST BTA16-600 + optotriac MOC3063 (zero-cross); Relay Omron G2RL-1A, Omron G5Q-1A (class examples).
- Capacitive touch controller (if external): Microchip AT42QT1070, Microchip CAP1298, Infineon/Cypress CY8CMBR3110.
- ESD protection for UI/touch lines: Nexperia PESD5V0S1UL, Littelfuse SP0503BAHT.
Symptom: Compressor won’t start / starts then stops
- First 2 measurements: P1 (DC bus sag/ripple at start), P6 (phase current sense signature during start).
- Discriminator:
- P1 stable + P6 shows a short current burst then cut-off → protection/drive chain (OCP, gate timing, current-sense offset) is dominant.
- P1 dips strongly + P3 also droops (quick check) → bus energy/impedance or rail hold-up issue is dominant.
- First fix:
- Drive/protection path: verify current-sense offset and OCP blanking/filter window; re-check gate drive integrity (P2) before replacing power stage.
- Bus/hold-up path: inspect DC bus capacitor ESR/connectors/relay contact drop; restore bus integrity before adjusting firmware thresholds.
MPN examples: CSA TI INA240A1 • Gate driver Infineon 6EDL04I06NT • IPM Infineon IM111-L6E2 - Prevent next time: run H2-10 F-01 (start/stop stress) + E-01 (brownout/sag observation) with reset-cause logging enabled.
Symptom: Loud whining at certain speeds (repeatable speed bands)
- First 2 measurements: P6 (current ripple vs speed), P9 (IPM temperature / derating flag alignment).
- Discriminator:
- Noise peak ↔ current ripple peak align tightly → torque ripple / control-to-mechanical resonance coupling.
- Noise peak appears mainly when hot → thermal derating shifts operating point into resonance band.
- First fix:
- Adjust ramp profile and avoid the resonance speed band (limit dwell time / modify target band selection).
- If temperature-aligned, improve module thermal path (contact/airflow/derating thresholds) before aggressive current limiting.
MPN examples: IPM ST STGIPN3H60 • Temp NTC Murata NCP18XH103F03RB - Prevent next time: run H2-10 A-01 torque-ripple check with a controlled speed sweep and log ripple metric vs speed.
Symptom: Random reset when compressor ramps
- First 2 measurements: P3 (logic rail droop during ramp), Reset cause log (BOR vs WDT vs other).
- Discriminator:
- P3 droop crosses BOR region + reset cause=BOR → rail integrity/hold-up is root cause class.
- P3 stable + reset cause=WDT → EMI/noise induced software stalls or peripheral lockups (still device-level).
- First fix:
- BOR path: strengthen local decoupling and rail hold-up on the control domain; isolate noisy auxiliary rails from logic.
- WDT path: tighten timing windows around high-noise events, add event-aligned logging, and ensure deterministic recovery behavior.
MPN examples: Supervisor TI TPS3808 • Supervisor Microchip MCP1316 • Buck TI TPS62133 - Prevent next time: run H2-10 E-01 sag observation aligned to ramp events and require coherent event logs across all resets.
Symptom: Frost builds up / defrost ineffective
- First 2 measurements: P7 (heater current proof during defrost command), coil temperature curve (evaporator sensor response).
- Discriminator:
- P7 present but coil temperature rise is weak → thermal coupling/sensor placement/airflow limitation dominates.
- P7 absent but system reports defrost active → heater drive chain (relay/triac/SSR/thermal fuse/wiring) dominates.
- First fix:
- Restore heater chain integrity (switch device, connector, safety cutout), then re-verify P7 waveform and termination reason.
- If current is real but heat is ineffective, validate sensor mounting and defrost termination logic using measured coil curves.
MPN examples: Triac ST BTA16-600 • Optotriac MOC3063 • Relay Omron G2RL-1A - Prevent next time: run H2-10 F-02 defrost coverage and require “command ↔ P7 proof ↔ coil curve ↔ termination reason” alignment.
Symptom: Temperature swings / overshoot (poor stability)
- First 2 measurements: P5 (Vref/AVDD stability), temperature ADC code vs time aligned with state/door events.
- Discriminator:
- Multiple channels move together → reference/ground-bounce/sampling-coupling dominates (not real temperature physics).
- Single sensor noisy or discontinuous → sensor/harness/connector integrity dominates.
- First fix:
- For Vref instability: restore reference decoupling and sampling timing discipline; isolate sensor return from high dv/dt loops.
- For single-sensor faults: replace the sensor/harness and ensure routing away from inverter edges.
MPN examples: NTC Murata NCP18XH103F03RB • Vishay NTCLE100E3103JB0 - Prevent next time: run H2-10 F-03 door-burst stress with acceptance on “sensor coherence” and stable drift bounds.
Symptom: Energy consumption abnormal spike (kWh/day rises)
- First 2 measurements: P8 (input power trend) and event alignment (defrost/door/ramp timestamps).
- Discriminator:
- Spike aligns with defrost → return to P7 + coil curve evidence chain.
- Spike aligns with door bursts → door sensing integrity and state handling dominate.
- Spike aligns with ramp but no thermal benefit → drive efficiency/control anomalies dominate (confirm with P6 signatures).
- First fix:
- Require every spike to be attributable to an event type (trend without attribution is not serviceable).
- Fix the dominant event chain (defrost heater chain, door false-open, or ramp control) based on alignment proof.
MPN examples: Clamp/power meter for P8 • Energy IC (if used) ADI ADE7953 - Prevent next time: run H2-10 E-02 load-step observation and verify event logging completeness under switching disturbances.
Symptom: UI ghost touch / unresponsive (during motor or backlight activity)
- First 2 measurements: P4 (touch baseline drift vs time), P3 (logic rail micro-drops/glitches).
- Discriminator:
- P4 drift aligns with inverter/backlight PWM bands → touch reference/noise coupling dominates.
- P3 glitches align with UI freeze → UI domain rail integrity and reset behavior dominate.
- First fix:
- For baseline drift: slow baseline tracking, avoid sampling during noisiest windows, and isolate backlight switching from touch reference.
- For rail glitches: add domain decoupling and enforce deterministic brownout behavior (blank safely, recover cleanly, log reset cause).
MPN examples: Touch IC AT42QT1070 • CAP1298 • ESD PESD5V0S1UL • SP0503BAHT - Prevent next time: run H2-10 A-02 fan/backlight sweep with touch stability acceptance and rail-dip monitoring.
Symptom: Fan stalls or hunts (speed oscillation, tach dropout)
- First 2 measurements: P10 (tach/Hall stability), P3 (logic rail stability during fan PWM/speed steps).
- Discriminator:
- P10 dropouts while rails are stable → fan feedback chain or fan mechanics dominates.
- P10 instability aligns with rail noise or UI issues → shared ground/noise coupling dominates (fan PWM pollutes control domain).
- First fix:
- Verify tach wiring, pull-ups, and signal conditioning before fan replacement.
- When coupling is proven, isolate fan PWM return paths and constrain the PWM bands that excite UI/touch noise sensitivity.
MPN examples: Hall switch Allegro A1104 • Melexis US1881 - Prevent next time: run H2-10 A-02 fan resonance scan + T-02 blocked airflow simulation to catch hunting under thermal stress.
Symptom: Trip / retry loop (protection cycling)
- First 2 measurements: P6 (abnormal current spikes vs time), P9 (temperature clustering vs trip timestamps).
- Discriminator:
- Trips cluster at high P9 temperature → thermal derating/protection dominates.
- Trips occur cold with sharp P6 spikes → electrical protection/drive chain dominates (sense offset, gate timing, bus ripple).
- First fix:
- Thermal path: improve module heat flow and airflow; require derating reason logging rather than “mystery stop.”
- Electrical path: validate current sense integrity, OCP blanking, and representative gate waveforms before replacing modules.
MPN examples: CSA TI INA240A1 • IPM Infineon IM111-L6E2 • Supervisor TI TPS3808 - Prevent next time: run H2-10 T-01 high ambient endurance + F-01 start/stop cycles and enforce bounded trip counts with full trip-type logs.
Decision Map (Symptom → 2 measurements → Root-cause class)
Request a Quote
H2-12 FAQs (Evidence-first, mapped back to chapters)
Each answer returns to this page’s evidence chain (motor drive, sensing, defrost, metering, UI, power integrity, validation, field debug). No protocol-stack deep dives and no whole-home energy discussion.
MPN examples below are concrete starting points for proven chains (drive, sensing, heater switching, UI). Always verify voltage/current/temperature, creepage/clearance, isolation class, and agency approvals for appliance production.
1) Why does an inverter fridge still “cycle” sometimes—control loop or sensor noise?
Inverter systems reduce cycling, but they can still enter stop–start when the state machine decides “no cooling needed,” or when noisy temperature readings mimic a real threshold crossing. The fastest proof is time-alignment: sensor codes and state transitions must agree. If multiple sensors move together, treat it as reference/ground noise before tuning control thresholds.
- Temperature ADC code vs time aligned to the state timeline (H2-4).
- P5 (Sensor Vref/AVDD stability) to rule out “all-channels drift together” noise (H2-5).
- Many channels shift together → Vref/ground-bounce / sampling window issue.
- Only one channel toggles at thresholds → sensor/harness or placement issue.
- Stabilize Vref/AVDD and sampling timing before changing hysteresis thresholds.
- Repair/replace the specific sensor/harness when a single channel is proven noisy.
2) Compressor starts then trips in 1–3 seconds: current sense or DC bus sag?
A 1–3 s start-then-trip typically belongs to two proven classes: (A) DC bus energy collapses during ramp, or (B) protection reacts to a false/real over-current from the sensing chain. Do not guess. Capture P1 (bus ripple/sag) and P6 (current-sense output) in the same time window. The alignment identifies the correct chain immediately.
- P1 DC bus sag/ripple during the first seconds of ramp (H2-9).
- P6 phase-current sense signature (burst → cut-off, or sustained rise) (H2-3).
- P1 stable + P6 burst then cut → drive/protection or sense-offset/blanking issue.
- P1 dips hard → bus impedance/hold-up/contact drop dominates.
- Sense/protection chain: verify CSA offset, OCP threshold, and blanking/filter window before swapping power parts.
- Bus chain: inspect bus capacitor ESR, connectors, and contact drop; restore bus integrity first.
3) Why is it noisier at a particular RPM band?
A “noisy RPM band” is usually a resonance where torque ripple excites the mechanical system, or where thermal derating shifts operation into a bad region. Prove it by correlating sound with P6 current ripple during a controlled speed sweep, and checking whether the band appears primarily when hot (P9). Once proven, avoid dwell time in the band or change the ramp profile.
- P6 current ripple metric vs speed during a sweep (H2-3).
- P9 module temperature / derating flag aligned to the noise band (H2-9).
- Noise peak ↔ current ripple peak alignment → torque-ripple resonance coupling.
- Band appears mainly when hot → thermal derating shifts operating point.
- Modify ramp profile and avoid prolonged dwell in the resonance band (H2-4).
- Improve thermal path/airflow if the issue clusters with temperature (H2-9).
4) Door closed but unit thinks it’s open: how to prove it in logs/scope?
False “door open” is either a real input glitch (bounce, moisture leakage, corrosion) or an overly aggressive debounce/threshold strategy. Proof requires aligning the raw door signal with door-event counters. Capture the door input waveform during vibration and temperature transitions, and compare it to log entries (event timestamps and burst counts). Replace hardware only after the raw signal is proven unstable.
- Raw door input waveform during vibration/thermal transitions (H2-11 method).
- Door event log: timestamps + “door_open_burst” counter alignment (H2-4/H2-5).
- Raw signal shows glitches/bounce → sensor/harness/connector chain.
- Raw signal is clean but logs show false opens → debounce/logic thresholds.
- Repair corrosion/connector integrity or replace the proven sensor/harness.
- Adjust debounce window only after raw stability is verified.
5) Frost keeps building though defrost runs: heater drive vs sensor placement?
“Defrost runs” is meaningless without heater proof. First confirm real heater current (P7) during defrost, then verify the evaporator coil temperature rise curve. If current is absent, the drive chain (relay/triac/SSR, wiring, thermal cutoff) is at fault. If current is present but coil temperature rises weakly, placement/thermal coupling/airflow dominates and termination conditions become unreliable.
- P7 heater current proof during defrost command (H2-6).
- Evaporator coil temperature curve (sensor reading aligned to defrost window) (H2-5).
- P7 absent → heater drive chain fault.
- P7 present but weak coil rise → thermal coupling / sensor placement / airflow limitation.
- Restore heater chain integrity (switch device/connector/thermal cutoff), then re-run proof + curve.
- Correct sensor mounting and verify termination logic against measured coil rise.
6) UI resets when compressor ramps: which two rails to capture first?
UI resets on ramp are most often a logic-rail droop problem, not a “software mystery.” Capture P3 (3.3V/5V logic rail) during compressor ramp and read the reset-cause log (BOR vs WDT). If P3 crosses BOR, fix rail hold-up and domain isolation. If P3 is stable but WDT triggers, treat it as noise-induced stalls and enforce deterministic recovery and logging.
- P3 logic rail droop/glitch aligned to ramp event (H2-9).
- Reset-cause log (BOR/WDT/other) aligned to the same timestamp (H2-11).
- P3 droop + BOR → rail integrity/hold-up and domain isolation.
- P3 stable + WDT → EMI/noise induced lockups; strengthen recovery behavior.
- Increase local decoupling/hold-up on UI+MCU domain and reduce shared return with noisy loads.
- Add event-aligned logs and enforce clean brownout behavior (blank safely → recover → record).
7) Ghost touches appear during fan PWM: grounding or sensing thresholds?
Ghost touches during fan PWM almost always come from touch-reference pollution (ground bounce, switching noise) rather than “random thresholds.” Prove it by recording P4 (touch baseline/reference) while sweeping fan PWM bands, and watching whether baseline drift aligns tightly to the PWM timing. If baseline moves, fix return paths and sampling windows. If baseline stays stable, then adjust thresholds and baseline tracking.
- P4 touch baseline/reference drift vs time (H2-8).
- Fan PWM timing (or tach-driven speed steps) aligned to the same capture window (H2-10/H2-11 style).
- P4 drift aligns with PWM bands → grounding/return/sampling coupling dominates.
- P4 stable but false triggers → threshold/baseline tracking strategy dominates.
- Isolate touch reference return from PWM current loops; avoid sampling during the noisiest windows.
- Only after stability is proven, tune thresholds and baseline tracking rates.
8) Energy usage jumped after service: which counters isolate defrost vs compressor?
Separate “where energy went” before changing controls. Use two counters: compressor energy share and defrost energy share (or ΔE per event window), then align spikes to the state timeline. If defrost share grows, return to heater proof and coil curve checks. If compressor share grows without improved temperature stability, suspect ramp efficiency, sensor drift, or abnormal retry behavior proven in logs and current signatures.
- Energy counters: E_comp vs E_defrost (or ΔE per event window) (H2-7).
- State timeline alignment (ramp/steady/defrost/door events) (H2-4).
- E_defrost dominates → heater chain / termination / sensor placement.
- E_comp dominates with no stability gain → drive efficiency/control or sensing drift; confirm with P6/P5 signatures.
- Require every spike to map to an event type; fix the dominating chain proven by alignment.
- Re-validate after service using the same counters as acceptance evidence.
9) Fans hunt / stall at low speed: driver commutation or mechanical drag?
Low-speed hunting is either feedback instability or real mechanical drag. Start by capturing P10 (tach/Hall) to confirm dropout or jitter, and check whether logic rail stability (P3) degrades during fan PWM steps. If tach drops while rails remain stable, treat it as fan/feedback chain or mechanics. If tach instability aligns with rail noise, treat it as shared ground/PWM coupling.
- P10 tach/Hall stability vs time (H2-11).
- P3 rail stability during fan PWM/speed steps (H2-9/H2-8 coupling).
- P10 dropouts with stable rails → mechanical drag or feedback chain weakness.
- P10 instability aligned with rail noise → grounding/return coupling.
- Repair tach wiring/pull-ups and connector integrity before fan replacement.
- When coupling is proven, isolate PWM return paths and avoid sensitive sampling windows.
10) Temperature overshoot after door open: sensor filtering vs damper control?
Post-door overshoot can be caused by (A) sensor filtering/latency that delays true recovery, or (B) state-machine actions (damper/fan/compressor targets) that over-correct. Prove it by aligning door events to temperature ADC slope and to the control state timeline. If temperature codes lag heavily or show shared drift, fix sampling/Vref/harness first. If sensing is clean, adjust state timing and limits.
- Door event timestamp aligned to temperature ADC slope/latency (H2-5).
- State timeline and actuator commands (damper/fan/compressor targets) aligned to the same window (H2-4).
- Sensor lag/drift dominates → filtering/sampling/Vref/harness chain.
- Sensing is clean but overshoot persists → state-machine timing/limits dominate.
- Fix sensing integrity and sampling discipline before changing control limits.
- Then tune door-recovery state timing and actuator constraints using the same evidence window.
11) Random “over-current” alarms in humid season: leakage paths or connector corrosion?
Humid-season OCP alarms are commonly caused by contamination leakage, moisture-driven surface conduction, or intermittent contact resistance from corrosion. Distinguish slow drift from burst faults. Capture P6 current-sense output while recreating humidity/condensation conditions and inspect whether faults are gradual (leakage) or discontinuous (connector). Also check heater-chain switching moments; moisture can create unexpected currents around high-power wiring.
- P6 current-sense output pattern: slow drift vs sharp bursts (H2-11).
- Event alignment: heater switching / door events / ramp windows to see clustering (H2-6/H2-4).
- Gradual drift + humidity correlation → leakage/contamination paths dominate.
- Sharp discontinuous bursts → connector corrosion/contact intermittency dominates.
- Clean/restore creepage surfaces, improve sealing and routing; verify stability under the same humidity condition.
- Replace proven corroded connectors/harness; avoid “threshold-only” fixes without evidence.
12) How to set acceptance thresholds so “good units” don’t get rejected?
Acceptance thresholds must be tied to scenarios, not single static numbers. Build a “good-unit distribution” across key tests (start/stop cycles, defrost cycles, door bursts, sag events, high ambient, blocked airflow) and set limits per scenario: trip counts, reset counts, ripple bounds, and temperature recovery time. Require event-aligned logs so a failure is explainable; unexplained rejects create high false-fail rates.
- Good-unit distributions for each H2-10 test scenario (not a single global threshold).
- Minimum log set: event counters + reset cause + state timeline + key snapshots for correlation.
- Threshold fails only in specific scenarios (hot / sag / blocked airflow) → scenario-specific limit needed.
- Fails without logs → logging coverage is insufficient; fix evidence before limits.
- Define per-scenario pass/fail criteria and cap allowable trip/reset counts with full root-cause logs.
- Re-run the matrix after any service/firmware update to keep thresholds stable.
