Air Purifier Hardware: PM2.5/VOC Sensing & Fan Drive
← Back to: Smart Home & Appliances
An air purifier is best engineered as a closed-loop electromechanical system: PM/VOC sensing AFEs create measurable signals, the MCU validates and filters them, then controls the BLDC fan and indicators while keeping power/EMC noise from corrupting readings. This page shows how to design, verify, and debug that evidence chain so sensor accuracy stays stable across fan speed, Wi-Fi bursts, lighting PWM, and ESD/surge stress.
H2-1 — Scope & system boundary (what this page covers)
An air purifier is best treated as a closed-loop mechatronic system: sensors observe particles and gas trends, firmware stabilizes the measurement and decides airflow setpoints, the fan actuator delivers CADR, and the platform survives power/EMC stress without silently degrading accuracy.
What this page delivers (engineering takeaways)
A repeatable hardware path that enables: (1) stable PM/VOC signal chains, (2) fan-drive selection with low noise/EMI, (3) validation + field debug evidence, and (4) IC selection by function buckets.
Out of scope (kept off this page)
Whole-home IAQ platforms, HVAC/ERV/HRV system design, cloud/backend architecture, router/network tuning, and deep protocol-stack walkthroughs (Matter/Thread/Zigbee).
Evidence chain keywords used throughout this page: rail ripple / droop, AFE reference noise, AFE output noise, TX burst current pulses, PWM edge coupling, tach jitter, brownout counters, self-test baselines, calibration drift fields.
H2-2 — Architecture overview: signal chain + power tree + control loop
A robust purifier architecture becomes easier to design and debug when it is separated into three engineering lanes with explicit measurement points: a PM lane (optical signal chain), a VOC/Gas lane (bias + low-noise path), and a Fan lane (BLDC actuation + feedback). This separation prevents “mixed-root-cause” confusion (e.g., PWM noise misdiagnosed as sensor drift).
Three lanes (each must be testable with evidence)
- PM lane: optical chamber → photodiode → TIA/ADC → digital filters → PM2.5 trend fields.
- VOC/Gas lane: sensor element → stable bias → low-noise amp/ADC → drift/humidity compensation fields.
- Fan lane: BLDC driver → motor → FG/tach → RPM stability → airflow estimate and acoustic control.
Power tree (practical partitioning)
- Actuation rail: AC-DC → 12V/24V fan domain (PWM edges, high di/dt).
- Logic rail: step-down → 5V / 3.3V for MCU + RF (TX burst current pulses).
- Sensitive rails: post-regulation → clean references for AFE/ADC (noise floor & drift).
- Design intent: isolate “dirty” and “clean” return paths so sensing is not dominated by motor/RF coupling.
Minimum measurements to keep the lanes honest: (1) ripple/droop on 12/24V and 3.3V during fan start + RF TX, (2) AFE reference noise and AFE output noise, (3) tach/RPM variance under fixed setpoints, (4) log fields aligned to timestamps (brownout counter, TX bursts, self-test baselines).
H2-3 — PM sensing deep dive (PM1/2.5/10): physics → AFE → digital
PM readings become trustworthy only when the optical path, analog front-end (AFE), and digital processing are treated as one measurable chain. The dominant failure modes are not “mysterious sensor variance”, but repeatable couplings: humidity-driven bias, optics contamination / aging, and electrical injection from fan PWM, switching rails, or RF bursts.
Measurement mechanism and dominant error sources
- Scattering intensity & particle distribution: different particle-size mixes can shift response even at similar mass concentration.
- Humidity coupling: typically appears as baseline lift + higher variance, often correlated with RH and with a slower time constant.
- Optics contamination / aging: chamber window dust, LED/laser aging, and photodiode drift can reduce sensitivity or raise baseline over time.
- Electrical injection: fan PWM edges, SMPS ripple, and RF burst droop can masquerade as “real PM changes” unless measured and gated.
AFE focus (photodiode → TIA → ADC): what must be engineered
- TIA stability & noise floor: input current noise + feedback resistor thermal noise set the minimum detectable change.
- Bandwidth vs settling: bandwidth too low hides real dynamics; too high imports PWM/RF energy and increases false spikes.
- Reference integrity: ADC reference / analog rails must not share bursty return paths with motor/RF domains.
- Flicker rejection: the chain must suppress 50/60 Hz and lighting-induced modulation via sampling and filtering strategy.
Digital processing essentials (keep it evidence-driven)
- De-spike: remove impulsive events that align with PWM edges or TX bursts before computing trend fields.
- Sliding window vs latency: choose window sizes that stabilize display/alarms without masking true step changes.
- Baseline + drift fields: track baseline, baseline slope, noise RMS, and flicker metrics; run a power-on self-test baseline.
- Sampling window discipline: align sampling to “quiet time” slots away from motor commutation and RF transmit bursts.
Six must-lock metrics (do not ship without these): noise floor (RMS, dark/quiet), dynamic range (no saturation), bandwidth/settling, sampling synchronization (quiet window), flicker immunity (50/60 Hz metric), drift monitoring (baseline + slope + self-test baseline).
H2-4 — Gas/VOC sensing deep dive: biasing, drift, humidity coupling
Gas/VOC sensing is dominated by low-frequency behavior: bias stability, leakage, and 1/f noise. The design goal is not “a sensor that outputs a number”, but a chain where drift becomes measurable fields and where humidity coupling is separated from true gas changes by evidence.
Sensor types (circuit view only)
- MOX: heater + sensing element; watch heater power stability, baseline drift, and humidity sensitivity.
- Electrochemical: small signals and high source impedance; bias/leakage/guard and ESD micro-leak dominate drift.
- NDIR (if present): optical drive + receiver chain; contamination/aging and drive timing can mimic drift.
AFE focus: bias, leakage, 1/f noise, and humidity coupling
- Bias source stability: drift in bias or reference directly shifts baseline and can masquerade as “gas trend”.
- Input leakage & bias current: leakage paths on humid/contaminated PCBs create slow baseline lift; guard zones reduce it.
- Low-frequency noise: 1/f noise limits low-level detection and increases false alarms if not filtered with evidence-aware windows.
- Humidity coupling: track RH correlation and time constant; compensation must be validated with logged fields.
Drift = evidence chain: baseline + baseline slope + RH correlation + warm-up fields (and heater power, if used). Lifetime edge is typically signaled by slower response, irreversible baseline lift, or self-test deviations that persist across power cycles.
H2-5 — Fan drive & acoustics: BLDC driver selection + speed feedback + EMI
Fan performance is perceived through airflow (CADR), acoustic comfort, and stability. Hardware choices that maximize efficiency can still fail the user experience if they introduce audible tones at low speed or create EMI that corrupts sensor readings and RF links.
Drive architecture essentials
- 3-phase BLDC (sensorless or Hall): Hall improves low-speed start and robustness; sensorless simplifies wiring but needs stronger start/recover logic.
- PWM frequency strategy: avoid audible bands and resonance regions; keep switching transitions consistent with EMI goals.
- Soft-start + protections: limit inrush/di/dt, handle stall/lock, and define retry/backoff behavior with logged events.
- Feedback options: FG/tach, Hall edges, or estimated speed; each must be validated for jitter and dropout under EMI stress.
Speed → airflow trend (engineering approximation)
- Goal: use RPM as a repeatable trend index for airflow and filter loading, not as an absolute CFM claim.
- Baseline mapping: store a clean-filter reference curve: RPM ↔ airflow_index (device-internal, relative).
- Trend detection: rising drive current + falling RPM under same setpoint indicates higher flow resistance (filter clog trend).
- Evidence trio: RPM variance, bus/phase current trend, and acoustic proxy (tach jitter / torque ripple signatures).
| Selection choice | Best for | Cost / risk symptom | Evidence to confirm |
|---|---|---|---|
| Hall vs sensorless | Hall: stable start and low-speed control | Hall adds sensors/wiring; sensorless may lose lock at low speed or heavy load | Lock-fail count, restart attempts, tach dropout rate at low RPM |
| Lower vs higher PWM freq | Higher can push tones above audible range | Higher switching can increase losses/EMI; lower can create audible modulation | Noise peaks vs PWM frequency, sensor noise aligned to PWM harmonics |
| Stronger vs softer edges | Strong edges improve efficiency and commutation margin | Strong edges increase common-mode noise; soft edges reduce EMI but may reduce margin | Phase node dv/dt at TP, CM noise correlation with AFE/ADC ref noise |
| Bus vs phase current sense | Phase improves control fidelity (FOC-ready) | Phase sense adds analog complexity; bus sense is simpler but less informative | Current ripple vs speed stability, stall signature clarity |
| Closed-loop vs open-loop speed | Closed-loop stabilizes low-speed and reduces audible hunting | Poor feedback quality can create oscillation and new tones | RPM variance, loop-induced oscillation in tach and bus current |
Minimum logged events: stall_count, lock_fail, restart_attempt, ocp_event, pwm_mode, rpm_mean, rpm_var. These fields enable fast correlation between acoustics, stability, and EMI.
H2-6 — Power integrity & isolation inside a noisy appliance
Sensor instability is often caused by internal noise sources rather than true air changes: switching ripple and ground bounce, motor PWM common-mode injection, and RF transmit burst current. The fix is a layered approach: domain partitioning, post-regulation cleaning, and measurement timing discipline.
Three couplings that corrupt sensor readings
- SMPS ripple / ground bounce: reference noise and local ground shift raise AFE noise floor and bias baselines.
- Motor PWM common-mode: dv/dt on phase nodes couples through return paths and parasitics into AFE/ADC reference.
- RF TX burst droop: Wi-Fi/BT peak current creates 3.3V droop and resets or measurement jumps aligned to TX timestamps.
Executable isolation tactics
- Partition domains: Dirty (motor/driver), Burst (RF), Clean (AFE/ref) with controlled return paths.
- Clean critical rails: post-regulate AFE/ref with LDO and targeted RC/LC filtering near the load.
- Control return paths: keep high-di/dt loops small; prevent dirty return from flowing through clean reference regions.
- Schedule measurements: sample in quiet windows away from PWM edges, commutation events, and TX bursts.
Three must-capture waveforms: (1) main bus ripple/droop (12/24V), (2) AFE reference + AFE ground, (3) RF TX burst current + 3.3V droop aligned to timestamps. These traces separate “true drift” from “injected noise”.
H2-7 — Backlight & indicators (UX hardware): LED drivers, flicker, light sensor coupling
Backlight and indicators are frequent hidden interference sources in air purifiers. Dimming edges can inject noise electrically through rails/ground, and light leakage can modulate the optical PM sensing path, creating measurement errors that look like real air events.
Two coupling paths to control
- Optical coupling: LED light leaks (direct/reflective) into the PM optical chamber and modulates photodiode current.
- Electrical coupling: PWM edge current creates rail ripple and ground bounce that corrupts ADC reference and AFE baselines.
Driver choices and dimming strategy
- Constant-current (CC) control: stabilizes LED current and reduces brightness drift; still requires return-path discipline.
- PWM dimming: efficient and common, but edges can cause EMI and optical modulation; frequency selection must be validated.
- Frequency selection principle: avoid alignment with PM sampling cadence and avoid low-frequency components that appear as flicker.
- ALS / light sensor (if used): schedule ALS sampling and feedback updates away from PM sampling windows to prevent closed-loop oscillation.
Interference-avoidance rules (must-pass checklist)
- Keep LED PWM edges out of PM sampling windows: sample only during a quiet region between edges.
- Partition LED power: do not share AFE reference rails with LED supply; isolate returns from clean reference regions.
- Control light leakage: use mechanical shielding and black walls near the PM chamber; avoid direct line-of-sight paths.
- Log LED states: store backlight PWM mode and brightness steps with timestamps for correlation analysis.
H2-8 — EMC/ESD/surge robustness for home appliances
Robustness comes from controlling entry points and return paths. The most effective designs treat protection as a map: where surges enter, where they are clamped, and where current returns without crossing clean sensing references.
Four-segment checklist (entry → harness → panel → antenna)
H2-9 — Validation plan: measurable specs → test matrix → pass/fail evidence
A validation plan must convert product claims into measurable evidence. Each item below is structured as: test condition → test points → logs → pass/fail. The goal is not only “still alive” after stress, but no measurable degradation in sensing stability, fan control, and coexistence.
Minimal instrumentation (practical baseline)
- Waveforms: basic scope channels for bus ripple, AFE reference, and key digital timing markers.
- Current/droop proxy: 3.3V rail min capture during Wi-Fi bursts (scope or rail monitor).
- Logs: baseline/slope/noise metrics, tach statistics, brownout/reset counters, event timestamps.
Pass/Fail should be written as measurable bounds (variance, slope, correlation, and event counts) rather than vague “looks OK”.
H2-10 — Field debug playbook: symptom → first 2 measurements → isolate → fix
Field debugging should follow a repeatable SOP. Each symptom below begins with the first two measurements to avoid guessing, then provides a discriminator and a first fix. The focus is on evidence from rails / references / timing alignment / tach stability.
PM readings jump randomly (no obvious trigger)
TP-REF (ADC ref noise) and TP-AFE (PM AFE output noise) in the same time window.
If TP-REF noise rises with jumps, the root is rail/reference injection; if TP-REF is stable but TP-AFE is noisy, the issue is optical/AFE path.
Strengthen clean-domain filtering (REF/AFE), then apply sampling window discipline to avoid switching edges.
PM spikes when the fan turns on or speed increases
TP-AFE (PM output) vs TP-PI (bus ripple) during fan speed steps.
If PM spike aligns with bus ripple/ground bounce, it is electrical injection; if PM spike aligns with airflow/optical leakage changes without rail movement, it is optical coupling.
Electrical: reduce loop area, slow edges, isolate returns. Optical: add shielding/black walls near chamber and lock sampling to quiet windows.
PM jumps during Wi-Fi upload / reconnect
TP-RF (3.3V droop) and log alignment (tx_ts vs pm_update_ts overlap).
Droop aligned to jumps indicates RF current transient; stable 3.3V with aligned jumps indicates sampling schedule conflict.
Add RF-domain decoupling/partition and reschedule PM sampling to avoid TX bursts; freeze updates during TX peaks.
PM changes when backlight/indicator brightness changes
Log backlight_step_ts and measure TP-AFE while stepping brightness (fixed fan mode).
A synchronous step without rail movement suggests optical leakage; a synchronous modulation with REF noise suggests PWM electrical injection.
Optical: add barriers and remove line-of-sight paths. Electrical: isolate LED supply/return and avoid LED edges inside sampling windows.
Fan is unstable at low speed (shake, hunt, audible whine)
TP-FAN (tach jitter / missing pulses) and rpm_var across the low-speed band.
High rpm_var with tach dropouts points to sensing/commutation instability; high rpm_var without dropout points to control tuning or mechanical resonance.
Adjust PWM frequency band and deadtime policy; improve low-speed commutation strategy and confirm return-path and edge control for EMI.
Fan occasionally fails to start or stops unexpectedly
TP-PI (bus droop at start) and stall_count / ocp_event logs.
Droop + brownout indicates power path weakness; repeated ocp/stall without droop indicates protection threshold or mechanical blockage.
Power: increase hold-up and reduce inrush; Protection: confirm stall policy and restart timing; verify airflow path and rotor friction.
Filter replacement indicator behaves erratically
rpm_mean vs motor current trend (or bus current proxy) at a fixed speed, plus baseline_slope trend.
If current rises while rpm is held, airflow resistance is increasing; if indicator changes without resistance evidence, thresholds or mapping drift is likely.
Re-anchor mapping using a clean filter reference; add hysteresis and require multi-window confirmation before changing indicator state.
Random reboot / reset in the field
reset_reason + brownout_counter, and TP-PI droop during suspect events (fan step, TX burst).
Brownout-aligned resets point to rail stability; watchdog-only resets point to firmware timing stress (often triggered by power noise or RF bursts).
Strengthen rail margins (partition, decoupling, sequencing) and enforce scheduling windows for high-noise actions.
VOC baseline slowly rises (no obvious air quality change)
voc_baseline / voc_slope trends and humidity_corr trend during steady air periods.
Strong RH correlation suggests humidity coupling; weak RH correlation with monotonic drift suggests sensor aging or bias/leakage issues.
Confirm bias stability and leakage control (guard/clean zone), then adjust drift monitoring thresholds and recalibration policy.
After ESD, the unit still works but sensing becomes noisier
Compare pm_noise_rms and selftest_time before vs after ESD; capture TP-REF noise under the same operating mode.
If REF noise increases, the damage is likely in rail/reference integrity; if only AFE noise increases, input leakage/ESD path damage is suspected.
Re-check clamp return paths and sensitive-node protection; add post-ESD self-test gates to flag measurable degradation.
H2-11 — IC selection map (MPN examples by function bucket)
This section maps an air purifier into function buckets and ties each bucket to selection parameters, common failure signatures, and concrete MPN examples. Focus stays on hardware evidence: power integrity, sensor fidelity, motor EMI, and RF peak-current behavior.
Selection order (prevents “good sensor, bad system” failures)
- Define system targets: CADR/airflow range, acoustic target, cost tier, enclosure constraints.
- Partition power domains: Dirty (motor/LED switching), Clean (AFE/ADC reference), RF (Wi-Fi/BT burst).
- Lock evidence hooks: TP points + required log fields before choosing parts.
- Choose sensor chain (PM, gas/VOC) with drift + interference behavior defined.
- Choose motor chain (3-phase BLDC + current sense) based on low-speed stability & EMI control.
- Choose RF module based on peak current and supply droop tolerance.
- Place protection (ESD/TVS/CM choke) driven by entry points and return paths.
Minimum “carry-forward logs” for selection and debug: pm_noise_rms, pm_baseline, pm_slope, voc_baseline, rpm_mean, rpm_var, rail_min_3v3, brownout_counter, stall_count, tx_ts.
A) PM sensor / optical front-end AFE (photodiode → TIA/ADC)
Key specs to lock
- Input bias/leakage sensitivity (contamination & humidity coupling).
- Low-frequency noise (1/f) and baseline stability for trend metrics.
- TIA bandwidth vs sampling window; avoid modulation from PWM edges.
- ADC reference sensitivity (TP-REF noise → concentration noise floor).
- Optical isolation constraints: light leakage paths and chamber reflections.
Common failure signatures (symptom → evidence)
- PM jumps with backlight changes → correlation between backlight_step_ts and pm_update_ts.
- PM rises when fan PWM increases → TP-REF ripple aligns with PWM edges; pm_noise_rms increases with RPM.
- Slow drift after weeks → pm_baseline and pm_slope monotonic shift; self-test baseline time increases.
Concrete MPN examples
| Function | MPN examples (2–6) |
|---|---|
| Module-type PM sensor (integrated optics + AFE) |
Sensirion SPS30,
Plantower PMSA003I,
Honeywell HPMA115S0
Fast integration; selection driven by stability vs dust contamination behavior and interface noise immunity.
|
| Photodiode TIA (discrete optical AFE) |
TI OPA380,
TI OPA381,
ADI ADA4530-1
Used when optics/chamber is custom and the PM signal chain must be tuned for bandwidth + baseline stability.
|
| Particle/photometric AFE (integrated optical front end) |
ADI ADPD4100,
TI AFE4404
Useful when the design needs integrated timing/photodiode front-end management and controlled sampling windows.
|
| Low-noise ADC (if not integrated) |
TI ADS1220,
ADI AD7685
Tie ADC reference and sampling schedule to LED/fan/RF quiet windows.
|
B) Gas/VOC sensing AFE (bias + low-drift amplify + ADC)
Key specs to lock
- Bias stability and temperature drift (baseline credibility).
- 1/f noise floor for slow-drift tracking (baseline/slope estimation).
- Input leakage management (board contamination + humidity coupling).
- ADC/reference isolation from motor/RF domains.
- Built-in diagnostics support (open/short detect, bias monitor) if available.
Common failure signatures (symptom → evidence)
- VOC baseline rises slowly → voc_baseline and voc_slope drift correlates with humidity/temperature trends.
- VOC jumps during Wi-Fi upload → rail_min_3v3 droop aligns with tx_ts; AFE output has synchronous steps.
Concrete MPN examples
| Function | MPN examples (2–6) |
|---|---|
| Electrochemical gas sensor AFE (potentiostat) |
TI LMP91000,
ADI ADuCM355
Best fit for electrochemical sensors when bias/control and low-power operation must be programmable.
|
| MOX VOC sensor (sensor+processing) |
Sensirion SGP40,
Sensirion SGP41,
Bosch BME688
Integrated digital sensors reduce analog pitfalls; still require clean supply and careful placement away from switching noise.
|
| Low-drift IA / current sense for sensor interface |
TI INA333,
ADI AD8420
Useful when discrete sensor interface is required and drift/CMRR must be controlled.
|
C) 3-phase BLDC fan driver + current sense (acoustics + EMI)
Key specs to lock
- Low-speed stability (RPM variance, tach dropout behavior).
- PWM strategy vs audible band; edge control vs EMI budget.
- Protection observability: OCP/OTP/UVLO flags and stall counters.
- Current-sense bandwidth/CMRR under high dv/dt common-mode noise.
- Control interface (3x/6x PWM, cycle-by-cycle limit) aligned to MCU timers.
Concrete MPN examples
| Function | MPN examples (2–6) |
|---|---|
| Integrated 3-phase driver (FETs included) |
TI DRV8316,
MPS MP6540
Fewer externals; selection driven by supply range, peak current, and EMI behavior at target PWM frequency.
|
| 3-phase gate driver (external FETs) |
TI DRV8323,
Infineon 6EDL7141
Used when higher current or lower RDS(on) is needed; layout and return paths become critical for EMI.
|
| Current-sense amplifier (motor phase/bus) |
TI INA240,
TI INA181,
ADI AD8418
Pick by CMRR vs switching edges; confirm with TP-FAN current waveform and noise injection into TP-REF.
|
D) Power (AC-DC / buck / LDO): dirty/clean/RF domains
Key specs to lock
- Ripple spectrum placement (avoid AFE/ADC sensitive bands).
- Transient response (Wi-Fi/BT burst droop: rail_min_3v3).
- Noise-cleaning chain: buck → LDO for AFE/REF.
- Start-up sequencing and brownout behavior (reset reason + counter).
Concrete MPN examples
| Function | MPN examples (2–6) |
|---|---|
| Offline flyback switcher (compact AC-DC) |
Power Integrations INN3167C,
Power Integrations INN3168C,
ST VIPer26K
Good fit for appliance auxiliaries; confirm conducted noise does not elevate TP-REF noise floor.
|
| Step-down buck (system rails) |
TI TPS62135,
TI TPS62130,
MPS MP1584EN
Choose by ripple, load-step response, and EMI; place “dirty” buck away from AFE/REF.
|
| Low-noise LDO (AFE/ADC reference cleanup) |
TI TPS7A20,
TI TPS7A02,
ADI ADM7150
Use to create a clean reference island; validate with TP-REF noise and sensor noise floor tests.
|
E) LED driver (indicator/backlight): flicker + sensor interference control
Key specs to lock
- Dimming method (PWM/analog) and controllable frequency range.
- Edge behavior and supply ripple injected into analog domains.
- Ability to align brightness steps with “measurement freeze” windows.
Concrete MPN examples
| Function | MPN examples (2–6) |
|---|---|
| White LED backlight driver |
TI TPS61169,
TI TPS61161,
Diodes AL8860
Pick PWM range to avoid PM sampling windows; confirm no synchronous PM modulation at TP-AFE.
|
| I²C LED driver (status indicators) |
NXP PCA9633,
TI TLC59711
Useful for multi-channel indicators with controlled dimming steps and low firmware overhead.
|
F) Wi-Fi/BT module / SoC (hardware power & burst behavior only)
Key specs to lock
- Peak TX current and allowable supply droop (budget for rail_min_3v3).
- Decoupling strategy (local bulk + HF caps) and return path separation from AFE/REF.
- Wake/sleep transitions and current-step edges (align with sampling schedule).
Concrete MPN examples
| Function | MPN examples (2–6) |
|---|---|
| Wi-Fi + BLE module/SoC (common appliance choice) |
Espressif ESP32-S3-WROOM-1,
Espressif ESP32-C3-MINI-1,
u-blox NINA-W106
Selection driven by burst current profile and antenna placement constraints; avoid RF return currents through analog reference grounds.
|
| Wi-Fi module (Murata ecosystem option) |
Murata Type 1DX
Useful when a certified module footprint is preferred; still requires careful supply droop control.
|
G) Protection (TVS/ESD arrays + common-mode choke): entry-point driven
Key specs to lock
- Low capacitance on high-speed/clock lines; robust clamp at panel connectors.
- Return path discipline: clamp current must not traverse AFE/REF islands.
- Common-mode impedance at relevant frequencies for harness emissions.
Concrete MPN examples
| Function | MPN examples (2–6) |
|---|---|
| ESD diode arrays (panel/buttons/high-speed pins) |
TI TPD4E05U06,
TI TPD2E2U06,
Littelfuse SP0502BAHT
Place at the entry point; verify ESD does not increase self-test time or sensor noise floor.
|
| Common-mode choke (signal/harness EMI suppression) |
TDK ACM2012-102-2P,
Murata DLW21SN series
Pick by common-mode impedance at target band; confirm reduced RF dropouts and lower PM noise coupling.
|
Figure F11 — IC selection flow (power/Noise/Motor/Sensor/RF → evidence)
Flow enforces domain partition first, then selects sensor/motor/RF chains, and finally anchors every choice to measurable evidence (TP points + logs).
Request a Quote
H2-12 — FAQs ×12 (evidence-based, within scope)
Each FAQ is constrained to this page’s evidence chain (PM/Gas AFE, fan drive, power/EMC, indicator interference, validation and field evidence). Every answer starts with the first two checks (TP points + log fields), then a discriminator, then a first fix.
Evidence shorthand used below: TP-PI (rails droop/ripple), TP-REF (AFE/ADC reference noise), TP-AFE (PM/VOC output), TP-FAN (phase current/tach/PWM edges), TP-RF (TX burst alignment). Common logs: pm_baseline, pm_noise_rms, pm_slope, voc_baseline, voc_slope, humidity_corr, rpm_mean, rpm_var, rail_min_3v3, brownout_counter, stall_count, tx_ts.
Q1Why is PM2.5 always high even when the air “looks clean”?
High readings can be real fine particles, optical contamination/leakage, or electrical injection into the measurement chain. First check TP-AFE for baseline offset and noise floor while logging pm_baseline/pm_noise_rms. Then check TP-REF for ripple correlated to the PM update. If TP-REF correlation is strong, treat it as power/return-path noise; otherwise prioritize chamber cleanliness and light sealing.
Maps to: PM AFE + PI (H2-3/H2-6/H2-9/H2-10)
Q2PM drifts when backlight/LED effects turn on: optical crosstalk or power coupling?
Separate “light leakage” from “electrical coupling” with alignment evidence. First check correlation between backlight_step_ts and pm_update_ts. Then probe TP-REF during LED PWM edges. If PM changes track brightness steps but TP-REF stays clean, optical crosstalk dominates (shielding/geometry). If TP-REF shows synchronous ripple, isolate LED power and move PM sampling into a quiet window away from PWM transitions.
Maps to: Backlight interference + sampling rules (H2-7/H2-6/H2-10)
Q3PM jumps twice during Wi-Fi upload: TX current droop or sampling-window conflict?
Start by time-aligning events. First check TP-PI 3.3V droop while logging rail_min_3v3 aligned to tx_ts. Then check whether tx_ts overlaps pm_update_ts. Droop-aligned jumps indicate insufficient RF decoupling or shared return paths. Jumps without droop but with overlap indicate scheduling conflict; fix by creating quiet windows, freezing updates during TX, and adding local bulk/HF caps at the RF module.
Maps to: RF coexistence + PI (H2-6/H2-2/H2-9/H2-10)
Q4Same unit reads very differently in two rooms: humidity coupling or airflow path?
Use correlation to avoid guesswork. First check humidity_corr/temp_corr against PM/VOC trend. Then check airflow proxy (rpm_mean, rpm_var) and whether readings change with fan speed steps. Strong humidity correlation suggests sensor/compensation coupling; weak humidity correlation but strong dependence on RPM or placement suggests sampling representativeness and internal flow path effects. Fix by validating compensation inputs and stabilizing intake/optics shielding before changing algorithms.
Maps to: PM+VOC sensing + fan loop (H2-3/H2-4/H2-5/H2-9)
Q5Filter replaced but “life low” persists: wrong estimate or threshold drift? How to reset reliably?
Treat filter reset as a verifiable event, not a UI action. First compare before/after fan proxy metrics (rpm_mean and motor current proxy if available) and log the reset event. Then verify persistent state by reading back cal_version/crc (or equivalent) after power-cycle. If state does not survive reboot, fix atomic write + readback + CRC. If state is correct but indication is wrong, recalibrate the mapping thresholds using a controlled airflow step test and re-baseline counters.
Maps to: Fan evidence + validation discipline (H2-5/H2-9/H2-10)
Q6Low-speed fan whine: change PWM frequency or driver dead-time first? How to measure audible components?
Use electrical signatures to infer acoustic risk quickly. First probe TP-FAN phase current ripple while sweeping PWM frequency; log rpm_var. If ripple and RPM variance peak in the audible band, prioritize PWM frequency move or spread-spectrum. If current ripple grows at commutation transitions, adjust dead-time and gate-slew to reduce torque ripple. A practical first fix is a frequency sweep + edge-rate limit, then re-check rpm_var and current ripple under identical airflow.
Maps to: Fan drive & acoustics (H2-5/H2-9/H2-10)
Q7Fan occasionally stops and recovers: stall protection or UVLO? What evidence proves it?
Split the root cause by event ordering. First check uvlo_event and rail_min_3v3 (or motor-bus minimum) aligned to the stop time at TP-PI. Then check stall_count and ocp_event aligned to the same moment. Voltage dip before stop indicates UVLO/PI; protection flags before dip indicate stall/overcurrent. Fix UVLO with hold-up/partitioning; fix stall with soft-start, current limit tuning, and tach dropout handling.
Maps to: Motor chain + PI (H2-5/H2-6/H2-10)
Q8After ESD the unit runs, but sensor readings become noisier: input leakage or damaged reference?
Compare “reference integrity” versus “input leakage” effects. First probe TP-REF noise and ripple before/after ESD stress; log pm_noise_rms and self-test settle time. Then probe TP-AFE for baseline shift and noise increase at constant conditions. If TP-REF degrades, strengthen the clean reference island (buck→LDO, return path control). If TP-REF is stable but baseline/noise worsens, suspect leakage paths (contamination, damaged clamps); fix with cleaning, guard/keepout, and low-leak ESD placement at the entry point.
Maps to: EMC/ESD + AFE integrity (H2-8/H2-6/H2-3/H2-9)
Q9VOC readings slowly rise over time: sensor aging or bias drift? Which three fields decide?
Use a three-field drift fingerprint. Track voc_baseline and voc_slope over days, and compare against humidity_corr (or temperature correlation). Strong correlation implies environmental coupling/compensation issues; weak correlation with monotonic slope suggests bias/reference drift or sensor aging. Confirm by checking TP-REF stability and bias monitor (if available). First fix: isolate bias/REF power, reduce leakage sensitivity, and schedule periodic re-zero under known clean-air conditions (when feasible) with versioned calibration storage.
Maps to: Gas/VOC AFE drift strategy (H2-4/H2-6/H2-9/H2-10)
Q10Only a few units in the same batch are “very inaccurate”: fastest two self-checks to screen hardware?
Two quick checks catch most hardware-related outliers. First measure baseline settle time on boot (self-test duration or the time to stable pm_baseline). Second measure static noise floor (pm_noise_rms, and voc_baseline stability if VOC exists) with fan and LEDs held constant. Long settle time plus high noise suggests reference/PI/leakage issues; normal settle time with large offset points to optics alignment or calibration constants. First fix: add screening limits to production logs and rework paths tied to TP-REF noise and leakage cleanliness controls.
Maps to: Validation matrix + debug SOP (H2-9/H2-10)
Q11Readout latency becomes much larger: filter window too long or sampling drops? Which counters first?
Separate algorithmic delay from data-path loss. First check configuration/version fields such as filter_window_len (or equivalent) and confirm it did not change. Then check sampling health counters: sample_drop_count, queue_overrun, and timestamp gaps between pm_update_ts and upload events. Stable counters with longer delay indicate filtering/config; rising drop/overrun indicates scheduling/CPU contention (often RF bursts). First fix: reduce window, prioritize acquisition, use DMA/interrupt discipline, and align uploads to non-sampling periods.
Maps to: Architecture + validation hooks (H2-2/H2-9/H2-10)
Q12Baseline changes after reboot: what calibration must be stored, and how to avoid wear/write failures?
Store only what is needed, and store it safely. Persist pm_baseline_cal, voc_baseline_cal, and compensation coefficients with cal_version + crc. First check version/CRC consistency after reboot, then check whether baseline rebuild occurred under a quiet window (no RF/LED/fan edge conflicts). CRC mismatch indicates storage/atomic-write issues; stable CRC with large shifts indicates unstable boot-time sampling conditions. First fix: atomic updates with readback, wear-leveling or FRAM, and “write on event” instead of frequent continuous writes.
Maps to: Drift + validation discipline (H2-3/H2-4/H2-9/H2-10)
Figure F12 — FAQ evidence map (Q1–Q12 → first checks)
A single-page map linking each FAQ to the first two evidence hooks (TP points + logs). Minimal text, more structure.
ICNavigator helps engineers find verified ICs, alternatives, and fast quotes.
Categories
- Power Management
- MCU
- Automotive ICs
- Sensors
- Interface & Transceivers
- Analog Front-End
Get in Touch
- Email: hello@icnavigator.com
- WeChat/WhatsApp: +86-xxxx