H2-1. Central Thesis + Boundary

This topic covers local leak presence detection near floors, appliances, and plumbing, where the design success metric is simple: trigger on real water while avoiding nuisance alarms from condensation films, dirty humidity, and chemical residue—all under a strict ULP power budget that must survive battery end-of-life and cold conditions.

H2-2. What Counts as “Leak” (and Why False Alarms Happen)

“Water detected” is not a single electrical condition. In real homes, the probe sees a mix of liquid water, thin films, humid dirt, and chemical residue. A robust design first defines event types and then selects discriminators that can be measured repeatedly under humidity and battery aging.

2.1 Leak events worth alarming (practical definitions)

• Sudden wet event: a fast transition caused by dripping or pooling water near the probe; the signal changes quickly and strongly.
• Persistent damp: slow wetting from seepage or wicking; the signal may hover near the decision threshold for a long time.
• Re-wet / repeated wetting: intermittent contact from splashes or vibration; requires confirmation logic to avoid chatter.

2.2 The top nuisance sources (why false alarms happen)

• Condensation film: a thin, weakly conductive layer that can look like “always a little wet,” especially under high humidity or cold surfaces.
• Humid dust / grime: random conduction paths that change with touch, airflow, and drying; often unstable and noisy.
• Electrode corrosion / polarization: DC bias and contamination create drift; the same “dry” state slowly moves toward the threshold.
• Salt/soap water: conductivity can jump by orders of magnitude, instantly defeating a single fixed threshold.

2.3 The 3 discriminators that make leak detection reliable

Event classification becomes robust when decision logic uses discriminators that survive drift and variability:

• Edge speed: how fast the measured impedance changes (fast drops often indicate true wet contact).
• Stability: whether the wet state remains stable (films and humid dirt are often noisy/unstable).
• Dry-back time: how the signal recovers after the wet stimulus is removed (condensation often recovers slowly and repeats).

2.4 Evidence to collect (minimal but decisive)

• Probe impedance vs time across dry → damp → wet → dry-back, capturing both transitions and steady periods.
• One context trace alongside it: humidity/temperature (as a nuisance indicator) or battery voltage (to separate power-related noise).

These two traces create a repeatable “signature” for each nuisance condition, and later enable fast diagnosis in the field (e.g., “humidity-driven film” vs “true water pooling” vs “corrosion drift”).

H2-3. Probe & Physics Model (Electrode, Polarization, Corrosion)

A leak probe is not “two metal pads.” It behaves as a time-evolving interface between electrode surfaces and electrolyte films (water, residue, humid dirt). A practical model explains why the same threshold that worked on day 1 can drift into false alarms or misses after weeks of bias, contamination, and drying cycles.

3.1 Minimal equivalent model (enough to predict drift)

• R_s (water-path resistance): dominated by water type (tap vs salt/soap) and film thickness; sets the “instant wet” range.
• C_dl (double-layer capacitance): the electrode–electrolyte interface; becomes prominent in thin films and during transitions.
• R_p (polarization/contamination path): a slow, bias-dependent drift term created by residue and surface changes; pushes the dry baseline.

3.2 Why DC excitation accelerates drift

Continuous or frequent DC bias drives directional ionic motion and electrochemical surface changes. Over time, this shifts the interface state, effectively changing R_p and C_dl. The observed outcome is a moving dry baseline: “dry” slowly looks more conductive, and recovery after wetting becomes slower or incomplete.

3.3 Geometry & build choices that change the model

• Electrode spacing: smaller spacing increases sensitivity but raises nuisance risk (films and humid dirt form easier bridges).
• Surface finish/plating: affects how quickly R_p drifts under bias and residue exposure.
• Probe cable/connector: adds parasitic capacitance and introduces leakage paths; also becomes an ESD entry and coupling antenna.
• Guard ring (high-Z/impedance sensing): reduces board-surface leakage so measured current tracks the probe, not contamination on the PCB.

3.4 Validation evidence (DC vs AC, same water)

• Dry baseline recovery: after a wet event, the measured baseline should return close to the initial dry value instead of “sticking” near the threshold.
• Repeatability: repeated wet/dry cycles with the same water should not widen the distribution dramatically over days/weeks.

H2-4. Sensing Method A — Pure Conductive (Threshold Comparator)

The conductive threshold approach is the lowest-cost path: it treats the probe as a wet/dry switch and uses a current-limited bias to create a measurable voltage, then a comparator provides an ultra-low-power wake signal. Its success depends on controlling the total DC dose (to reduce drift) and proving margin across water variability (tap vs salt/soap) and nuisance films.

4.1 Reference architecture (why each block exists)

• Probe electrodes create a conduction path when wet.
• R-limit / divider caps electrode current so salt/soap water cannot accelerate corrosion or collapse the battery.
• Comparator with hysteresis converts a noisy analog boundary into a stable wake event (prevents chatter).
• MCU confirmation window rechecks quickly (multiple short samples) before declaring an alarm.

4.2 The “anti-drift kit” for DC sensing (4 knobs)

• Limit current: cap electrode current under the best-conducting water; keeps the electrochemical “dose” bounded.
• Pulse sampling: measure in short bursts; reduce total biased time while preserving detection latency.
• Sampling window: avoid measurement during radio bursts and alarm transients to prevent noise-driven triggers.
• Polarity swap: periodically reverse bias to reduce directional polarization and asymmetric surface aging.

4.3 Evidence to prove before shipping

• Dry vs wet resistance distributions: measure across tap water, salt/soap water, and humid-dust conditions.
• Noise margin: confirm the worst-case dry value remains far from the trip point under battery end-of-life and high humidity.
• Comparator hysteresis effectiveness: demonstrate stable entry/exit behavior without chatter near the boundary.

H2-5. Sensing Method B — Impedance / AC Excitation (More Robust)

Impedance sensing adds a frequency dimension and reduces long-term DC dose at the probe. This improves separation between true water contact and nuisance films (condensation, humid dirt, residue) that can look like weak conduction in a DC threshold design. The practical objective is measurable separability under humidity and contamination without accelerating polarization drift.

5.1 Why AC improves robustness (no long DC bias state)

• Lower polarization risk: short AC bursts reduce directional electrochemical stress that moves the probe baseline over time.
• Film vs droplet behavior: thin films and surface leakage often show higher “capacitive” behavior and noisier stability than true droplets.
• Decision made on response: the design evaluates an impedance response signature, not a continuously biased conduction state.

5.2 Frequency intuition for engineering selection

• Lower frequency: response tends to reflect conductive paths more strongly (closer to “water resistance”).
• Higher frequency: response tends to reflect capacitive paths more strongly (thin films, cable/board parasitics become visible).
• Practical use: one band can detect “likely wet,” and a second band can help reject “film-like” nuisance behavior by stability and ratio.

5.3 Minimal measurement chain (production-friendly)

• AC excite: GPIO square wave + R-limit (or simple RC shaping) into the probe.
• Detect: synchronous sample (aligned window) or rectifier + low-pass to obtain amplitude.
• ADC + decision: short sample window, then threshold/ratio + confirmation bursts.

5.4 Guard & shielding for long probe cables

• Guard ring / driven guard: reduces surface leakage influence on high-impedance nodes so the measurement tracks the probe, not PCB contamination.
• Shield / reference routing: reduces RF coupling and ESD-induced transients that can mimic a wet event.
• Cable parasitics: long leads raise capacitive components; design uses frequency choice and windowing to keep decisions stable.

5.5 Evidence to prove separability (moist dirt vs real droplets)

• Observation 1 — amplitude: record demodulated amplitude under true droplets vs moist dirt/condensation film.
• Observation 2 — stability time: measure time-to-stable and variance inside the window; nuisance conditions are typically less stable.

H2-6. ULP Power Budget (Sleep, Wake, Measurement Duty-Cycle)

“Runs for a year” is determined by a three-state budget: long Sleep, short Measure bursts, and the worst-case Alarm + RF bursts. Many field failures are not caused by average current, but by peak current producing brownout on high-ESR batteries (cold and end-of-life), which can trigger resets and change the false-alarm/miss rate.

6.1 Three-state model (what must be budgeted)

• Sleep: dominant time share; current must be stable across temperature and battery voltage.
• Measure burst: short wake + excite + sample; energy scales with burst length and measurement frequency.
• Alarm + RF burst: highest peak current; dictates minimum rail headroom and reset immunity.

6.2 Duty-cycle strategy (short window + confirmation)

• Short measurement windows: reduce energy and limit probe “dose,” helping long-term stability.
• Confirmation bursts: multiple short checks reject unstable nuisance films without a long biased measurement.
• Wake sources: comparator/RTC wake reduces the need for frequent full MCU run time.

6.3 RF burst brownout (why resets create false alarms/misses)

• Coin-cell ESR + I_peak: voltage droop increases at low temperature and near end-of-life.
• Crossing UVLO/BOR: resets can corrupt state or cause partial sampling, producing inconsistent decisions.
• Design response: keep burst current localized (decoupling near radio), gate sampling windows, and log brownout events if available.

6.4 Battery constraints that must be tested (not assumed)

• Cold behavior: ESR increases; droop worsens under the same transmit power.
• Aging: ESR rises over time; worst-case margin shrinks.
• Impact on accuracy: false-alarm and miss probabilities rise near the rail boundary without robust windowing/confirmation.

6.5 Evidence checklist (minimum two artifacts)

• Artifact #1: battery voltage sag waveform during RF burst (min voltage + recovery time).
• Artifact #2: false-alarm/miss rate shift under low temperature and near end-of-life battery condition.

H2-7. Alarm Chain (Buzzer/LED/Siren Driver) without Killing Battery

The alarm chain must stay audible/visible at low battery while avoiding rail collapse. In leak sensors, the most common failure mode is not “insufficient loudness,” but peak-current transients that pull the battery below UVLO/BOR, causing resets and inconsistent alarm behavior. A robust design treats alarm actuation as a controlled transient event.

7.1 Choose the load type by controllability (not just BOM cost)

• Active buzzer: simple on/off control; predictable current; easiest to keep within brownout margin.
• Passive buzzer (PWM): higher control over loudness and tone, but edges and drive patterns can inject noise and increase peak current.
• LED + optional siren: visibility depends on peak current and flash timing; average power is set by duty-cycle.

7.2 Driver topology (peak current under control)

• Low-side switch (N-MOS/NPN): minimal loss and simplest routing; suitable for most buzzers/LED strings.
• High-side switch (P-MOS/load switch): useful when the alarm return path must not disturb analog reference or when a clean cutoff is required in low-voltage conditions.
• Soft-start by patterning: ramping duty or using short bursts can reduce the effective inrush without extra ICs.

7.3 LED flash strategy (visible, low average power)

• Peak vs duty: choose short, bright pulses instead of long dim on-time to improve visibility per mAh.
• State-based patterns: normal indication, alarm indication, and low-battery indication should use different duty and spacing.
• Noise awareness: avoid sampling the probe during LED/buzzer edges; keep measurement windows clean.

7.4 UVLO/BOR boundary and alarm degradation policy

• OK rail: buzzer + LED pattern + (optional) wireless notification.
• Low rail: LED bursts + short beep, reduced duty, postpone any high-peak bursts.
• Critical rail: minimal LED indication, disable high-peak loads to avoid repeated resets.

7.5 Evidence checklist (must be captured)

• Alarm transient current: peak and repeatability over multiple triggers.
• Battery voltage waveform: minimum droop and recovery time during alarm actions.
• Reset/brownout counters: BOR cause, reset reason, or brownout event flags correlated with alarm activation.

H2-8. Wireless Coexistence (Hardware-Level Only)

If the sensor is stable with radio disabled but becomes noisy when radio transmits, the cause is typically hardware coupling: RF burst peak current creates supply ripple and ground bounce, and the probe cable can behave like an antenna. Robust coexistence is achieved by time-division measurement, supply/return-path isolation, and controlled entry paths for long cables.

8.1 Injection paths (how RF noise reaches the AFE)

• Supply ripple: TX burst current pulls the rail; AFE reference shifts and appears as a false impedance/amplitude change.
• Ground bounce: shared return impedance between radio and AFE injects apparent input steps.
• Probe cable antenna: coupled RF energy is rectified or appears as spikes at high-impedance nodes.
• ESD/common-mode entry: cable or housing paths inject common-mode transients into the front-end.

8.2 Fix strategy (prioritized, hardware-level)

• Time-division: schedule sensing windows away from TX bursts and alarm edges; confirm with repeated short windows.
• RC/LC isolation: separate AFE rail from radio rail (small R + local decoupling or bead + cap) to confine di/dt.
• Return-path control: avoid shared high-current ground paths; keep AFE return local and join at a controlled point.
• Cable entry control: place ESD/CM components at the entry and ensure their return is short and intentional.

8.3 Evidence waveforms (OFF vs ON radio)

• Waveform A: AFE input or demodulated amplitude node (spikes/steps during TX indicate coupling).
• Waveform B: battery voltage and/or analog ground reference (droop and bounce during TX indicate power/return injection).

H2-9. Ruggedization: ESD/Surge, Probe Cable, Coating, IP

Leak sensors are often deployed in kitchens, bathrooms, basements, and utility areas where humidity, detergents, long cables, and nearby motors are common. Ruggedization is therefore not optional: the probe port must survive ESD and induced transients without shifting the decision baseline. The main goal is to force disturbance energy to take a short, intentional return path, while preventing coating or IP choices from creating new leakage paths that increase false alarms or misses.

9.1 Probe-port ESD path (keep energy out of the high-Z front-end)

• Define the return: ESD energy should flow into a short, controlled return (chassis/return node), not through the AFE input.
• Place protection at the entry: protect the port before the signal reaches high-impedance nodes and sensitive references.
• Short loop matters: a long return loop converts ESD into ground bounce and threshold drift even if nothing “burns.”

9.2 TVS selection principles (brief, measurement-aware)

• Working voltage + clamping: protect against contact discharge and cable entry events within the allowed node voltage.
• Leakage vs high-Z: leakage current can look like “wet” on conductive or impedance inputs if not managed.
• Capacitance impact: port capacitance can distort impedance signatures, especially with long probe cables.

9.3 Long-cable induction (near pumps/valves/motors)

• Cable as a loop: long probe leads pick up induced transients that appear as spikes or step changes at the AFE node.
• Common symptom: stable sensing in quiet conditions but false triggers when a nearby motor starts or switches.
• Mitigation (hardware-first): limit energy at the entry (series impedance, controlled filtering) and avoid sampling during known transient windows.

9.4 Coating and IP tradeoffs (protection vs decision integrity)

• Coating side-effects: some materials support surface conduction when contaminated, shifting the dry baseline.
• Keep-out around high-Z: reserve clean, controlled areas around the probe input network and guard structures.
• IP design reality: sealing can trap moisture and promote thin films; mechanical placement should avoid persistent bridging around electrodes.

9.5 Evidence points (required)

• Post-ESD threshold drift: compare the decision baseline/threshold distribution before vs after ESD stress.
• Damp-heat miss-rate shift: quantify detection probability and response time after high-humidity exposure.

H2-10. Validation Matrix (Test Plan You Can Actually Run)

Reliability is only meaningful when it is measurable. This chapter defines a practical validation matrix for leak sensors that can be executed with typical lab resources. The matrix intentionally stresses the decision boundary using representative water samples, environment conditions, battery states, and lightweight EMC checks. Each test cell must report a consistent KPI set: False Alarm Rate, Miss Rate, Response Time, and Recovery Time.

10.1 KPIs and pass criteria (define “good” upfront)

• FAR (false alarms): triggers per hour/day under dry and nuisance-film conditions.
• MISS: probability of not triggering under true leak events within the allowed time window.
• Tresp: time from wet application to confirmed decision.
• Trec: time from removal/drying to stable non-wet state without re-triggering.

10.2 Four test axes (minimum set)

• Water samples: tap, salt, soap, condensation film.
• Environment: cold, high humidity, dry dust.
• Battery: fresh, end-of-life, cold-soak.
• EMC (lightweight): contact discharge and probe-port ESD/EFT as available.

10.3 Repeatable per-cell procedure (keeps results comparable)

• Pre-condition: define dry baseline (probe dryness, cable length, placement) before each run.
• Apply stimulus: consistent wet method (droplet, film, or controlled wetting) and duration.
• Record: decision node amplitude/stability time + event outcome + response/recovery timing.
• Reset/clean: remove residues between samples (soap/salt) to prevent cross-contamination of the baseline.

10.4 Lightweight EMC checks (focus on KPI impact)

• Contact discharge: near enclosure/cable entry; observe any baseline drift or spurious triggers.
• Probe-port ESD/EFT: minimal available stress; quantify drift, FAR rise, or MISS rise.

10.5 Reporting template (what each cell must output)

• Outputs: FAR, MISS, Tresp, Trec, plus baseline/threshold drift indicators after ESD or damp-heat exposure.
• Worst-case summary: identify the combination that produces the maximum FAR or MISS and the slowest recovery.

H2-11. Field Debug Playbook (Symptom → Evidence → Fix)

This chapter is a fixed-format troubleshooting SOP for field failures: each symptom lists the first two measurements, a discriminator that separates root causes, and the first fix with concrete component options. The scope stays inside the leak-sensor chain (probe/AFE/power/alarm/RF coexistence/port protection), avoiding protocol-stack or cloud workflows.