What this page is—and is not

This page defines Smart Plug / Power Strip as a verifiable hardware system with four inseparable functions: (1) energy metering, (2) controlled on/off switching, (3) electrical/thermal protection, and (4) wireless telemetry. Every claim in later chapters should map to at least one measurable artifact: waveform, temperature rise, fault flag, or event log.

Not covered here: whole-home branch monitoring, HEMS panel architecture, UPS/inverter transfer paths, dimmer/phase control, cloud/backend design, app UX tutorials, or certification walkthroughs.

Two form factors—and why they are not “the same device”:

• Single-outlet smart plug: simpler coupling paths, shorter metering loop, but often harsher peak load events (motor/compressor inrush) and tighter thermal enclosure constraints.
• Multi-outlet power strip: stronger cross-coupling between channels (switch transients, RF burst load steps, shared rails), higher total heat stacking, and more complex surge return paths (clamp effectiveness becomes layout-dominated).

Turn “features” into specs, then into evidence

Smart plug requirements are often described as features (metering, protection, wireless control), but robust design starts by translating each use-case into measurable specs and then locking the two most decisive evidence points to verify them. This prevents vague marketing claims and makes validation repeatable across different loads and environments.

How to read the table: Each row defines (1) what must be true, (2) the failure symptom that breaks it, and (3) the first two measurements/logs that settle the debate quickly.

Scope guard reminder: only engineering constraints that change hardware design are referenced (e.g., creepage/clearance implications, clamp loop length, rail budget). Certification procedures and platform/backend architecture remain out of scope.

Pick a sensing topology by system cost: heat, phase, isolation, and evidence

Metering accuracy in smart plugs and power strips is rarely limited by the SoC alone. The sensing topology shapes the entire error chain: heat generation, phase/PF behavior, isolation boundary, and layout sensitivity. A good choice is one that keeps error sources observable and calibratable under real loads (motor inrush, non-linear SMPS currents, and RF burst supply steps).

Shunt (high-side / low-side)

A shunt converts current into a small voltage drop. It is simple and scalable, but the “cost” appears as I²R loss and temperature-driven drift. For power strips, channel-to-channel thermal gradients can dominate long-term accuracy.

High-side vs low-side (why it matters):

• High-side shunt: avoids ground lift errors, but demands higher common-mode tolerance and stricter spacing/layout around the line domain.
• Low-side shunt: simpler common-mode, but more vulnerable to “ground bounce” and shared return coupling from switching transients.

Two decisive evidence points (fast verification):

• (1) ΔT trend at the shunt / nearby copper during steady load (captures drift risk).
• (2) Metering rail/reference ripple time-aligned with Wi-Fi TX burst or relay/eFuse switching (captures jump/offset events).

Current Transformer (CT)

CTs provide a low-loss, naturally isolated current sense path. The trade-off is typically in low-current behavior and phase error, both of which directly impact PF and real-power computation.

Two decisive evidence points (fast verification):

• (1) PF error vs load type (resistive vs SMPS vs motor) after phase compensation.
• (2) Low-load stability (does kWh/W “stick” or jitter) with raw sample distribution before heavy averaging.

Rogowski coil

Rogowski sensing excels at high-frequency current shape and transients, but it measures di/dt and therefore needs integration. That integration chain adds drift and low-frequency sensitivity, which can be unacceptable if the goal is stable low-load energy reporting.

Two decisive evidence points (fast verification):

• (1) Integration stability after warm-up (offset drift mapping into “phantom power”).
• (2) Transient correlation between switching events and captured current shape (does the chain preserve time/shape evidence).

Calibration that must exist (method-only, evidence-oriented)

Regardless of topology, stable metering requires a calibration plan that is repeatable and audit-able via logs. The goal is not “one-time factory trim,” but an evidence chain that can detect drift and prevent silent accuracy collapse.

Switching is energy management: inrush, SOA, recovery policy, and observability

Outlet switching is not merely “on/off.” In smart plugs and power strips, the switching element also defines: how inrush energy is handled, how faults are isolated, how quickly recovery happens, and whether the device can produce a trustworthy fault evidence chain (flags + event logs). This chapter compares three mainstream approaches: eFuse/hot-swap, discrete high-side switches, and relays.

eFuse / Hot-swap

eFuse/hot-swap devices are purpose-built for controlled inrush and fault isolation. They can actively shape current and usually provide fault status outputs, which makes field evidence far cleaner than a pure mechanical switch.

• Inrush control: soft-start and current limiting prevent supply collapse and reduce nuisance resets.
• Limit mode choice: constant-current vs foldback affects both survivability and load start success.
• SOA awareness: long inrush in the linear region can accumulate heat and trigger OTP or lifetime stress.
• Observability: OCP/OTP/UVLO flags + retry counters enable evidence-based debugging.

High-side switch

High-side switches can be compact and efficient, but design quality depends on Rds(on) thermal reality and short-circuit response. Reverse-current behavior and fault reporting completeness are critical in multi-outlet power strips where coupling paths are stronger.

• Loss → heat: Rds(on) turns continuous current into ΔT; board copper and airflow dominate the result.
• Short response: reaction time and limit behavior determine whether rails dip and whether wireless stays stable.
• Reverse current: backfeed can corrupt metering/reference rails and cause “online but wrong readings.”

Relay

Relays provide low on-resistance and a simple separation of control and power, but their failure modes are physical: contact wear, arc-induced EMI, and weld/stick events. Those modes frequently show up as metering jumps, wireless dropouts, or “state mismatch” (reported OFF while power is still ON).

• Arc EMI: switching arcs generate broadband noise that can disturb rails and RF link quality.
• Weld/stick: inrush events can fuse contacts; fault evidence must detect and report state mismatch.
• Contact resistance drift: increases heat and can bias power readings over time.

Power strip strategy (multi-outlet) — per-channel vs inlet protection

For power strips, protection strategy is about fault containment and coupling control:

• Inlet (shared) protection: simpler and cheaper, but one severe event can degrade the entire device’s stability.
• Per-outlet protection: better containment and clearer diagnosis, but higher thermal density and more complex rail transients.
• Hybrid: shared surge/EMI front-end + per-outlet OCP/OTP isolation is often the most practical balance.

Protection as reproducible failure paths: symptoms → two probes → first fix

Ruggedness in smart plugs and power strips is not defined by a parts list. It is defined by where energy flows during surge/EFT/ESD and how that flow maps into rail droop, ground bounce, and state corruption. This chapter expresses each stress type as: typical symptoms + first two measurements + first fix.

Surge (lightning / switching surge)

Surge is an energy problem. The design goal is to force most energy into a short, thick, predictable copper loop before it reaches the PSU isolation boundary or any sensitive metering/reference node. Clamp devices (MOV/GDT/TVS) are a hierarchy: high-energy shunt, fast local clamp, and controlled return.

First fix (path-first):

• Shorten the surge loop: move clamp closer to the entry and tighten the return path (reduce loop area).
• Separate “high-energy return” from sensitive reference/AFE return; prevent surge return from crossing metering ground.
• Apply a clamp hierarchy: high-energy shunt near entry, fast TVS near sensitive nodes (local containment).

EFT (fast transient burst)

EFT behaves like repeated high-speed injection. It primarily attacks power integrity and state-machine robustness: short rail dips, reference disturbance, and timing/glitch sensitivity. The “most common” field result is not destruction but reboots, metering jumps, and link drops.

First fix (domain containment):

• Strengthen decoupling and local energy storage on the always-on rails; protect reference/AFE supplies from burst noise.
• Ensure reset/BOR thresholds and timing avoid “half-alive” states; log reset reason and last good state.
• Reduce coupling: keep noisy switching paths away from metering and radio domains; avoid shared return bottlenecks.

ESD (air / contact discharge)

ESD is a current return path problem. Robust designs force discharge current to return locally through a controlled clamp path, preventing the current from traversing sensitive domains (MCU reset, metering reference, relay control).

First fix (return-path control):

• Place ESD clamps at the entry/interface point with the shortest return path to the intended reference plane.
• Protect control lines that can trigger false switching (relay drive, eFuse enable) with local containment.
• Prevent ESD return from sharing narrow necks with AFE/reference and radio ground.

Brownout (mains dip → PSU dip → state disorder)

Brownout failures are often silent corruption rather than obvious resets. Partial undervoltage can leave the MCU running while storage writes become unreliable and metering accumulators lose coherence. The design goal is “clean fail”: either run within valid rails or reset deterministically with state integrity preserved.

First fix (clean reset + state integrity):

• Use deterministic BOR/UVLO behavior per domain; avoid mid-rail undefined execution.
• Gate non-volatile writes; treat brownout as an event that must set flags and preserve last-known-good markers.
• Prioritize a stable always-on budget (see H2-6) so radio bursts and actuation do not create self-inflicted brownouts.

Always-on credibility: domain rails, peak events, and brownout-proof behavior

Many smart plugs fail in normal usage due to the interaction between always-on metering, radio burst currents, and actuation peaks. A robust design starts from a power-domain plan: which rails must stay stable at all times, which loads can be time-gated, and how peak events are prevented from collapsing the device into brownout or metering jump states.

Aux power chain and domain rails (what must be separated)

A common architecture is: AC-DC → 5V → multiple buck/LDO rails for separate domains. The separation is not cosmetic. It is the primary mechanism to keep noisy events from contaminating the metering/reference chain and to keep RF bursts from triggering resets.

Always-on vs switched domain (metering and radio scheduling)

“Always-on” does not automatically mean “always high power.” It means the domain keeps integrity: metering can remain continuous while reporting is duty-cycled, as long as events are captured and state is coherent. The risk is silent error: metering jumps or log gaps caused by rail ripple during burst events.

Peak current events that trigger self-inflicted brownouts

Three peak events dominate: Wi-Fi TX burst, relay coil pulse, and eFuse/HS switching events. The design task is to prevent these peaks from collapsing rails or injecting noise into the metering/reference chain.

• Wi-Fi TX burst: short, repetitive current peaks that can create ripple and droop on 3V3/1Vx rails.
• Relay actuation: coil pulse can momentarily drag the supply or inject inductive kick noise if not contained.
• Fault/log event: NVM writes and fault processing can spike current and must not occur during unstable rails.

Always-on budget table (state → peak → evidence)

The table below turns “power saving” into a measurable plan. Each state defines the rail risk, the peak event, and the evidence to confirm robustness.

Accuracy is a system error chain: detect, compensate, and flag when invalid

Metering integrity in smart plugs and power strips is determined by the full error chain: sensor → AFE → ADC/REF → digital windowing → phase/PF math → thermal/time drift. A “high-accuracy IC” cannot compensate for ground lift, rail ripple, or temperature gradients that distort the signal before conversion.

Error-source ladder (what it looks like in the field)

Non-linear loads (SMPS / rectifier / PFC): sampling must match waveform reality

Many plug loads are strongly non-linear. Current can be peaky and time-localized, which stresses sampling window selection and phase computation. Errors are amplified when rail ripple or reference movement coincides with those peaks.

• Risk focus: PF and real power can “look stable” while drifting, if phase error and windowing bias are consistent.
• Practical requirement: keep sampling windows away from known high-noise events (radio bursts and switching transients).
• Integrity marker: compare power change vs current change; flag inconsistent deltas as “invalid window / corrupted sample.”

Thermal drift paths: shunt self-heating and contact resistance movement

Drift often follows heat. Shunt self-heating changes resistance; relay contact resistance can move with wear and temperature, creating extra drop and local hotspots. In multi-outlet strips, gradients and heat stacking make single-point compensation less reliable.

Calibration strategy: factory baseline + in-field integrity checks

Calibration is not a one-time action. It is a system policy: establish a factory baseline, then continuously validate integrity in the field. The key output is not only “corrected reading” but also a confidence state backed by logs.

Why metering jumps when Wi-Fi transmits: evidence-based coupling paths

Coexistence failures are rarely “mystical.” They are usually one of three coupling paths: (A) power rail droop/ripple, (B) ground/return-path lift, or (C) switching EMI transient. The fastest diagnosis aligns events in time: radio bursts, switching actions, and metering/reference disturbances.

Primary causal chain: TX burst → rail droop → REF shift → reading jump

A Wi-Fi transmit burst is a short, high peak-current event. If the radio rail shares weak impedance with the ADC reference or metering domain, the burst creates droop and ripple that directly appears as a reading jump.

Return-path coupling: digital ground current lifts analog ground

Even when rails look “okay,” return-path bottlenecks can lift the analog ground used by the AFE/ADC reference. The result is a measurement offset that is synchronized with radio activity.

Switching EMI: relay/eFuse transients disturb RF and reset robustness

Switching edges (relay arcing, eFuse/HS fast transitions) inject wideband noise. This can disturb radio link stability and can also corrupt sampling windows, producing discontinuities and log anomalies.

Four-step debug block: symptom → waveform → isolate → fix

Minimum containment strategy (what changes the outcome fastest)

• Domain rails: isolate RF, MCU, and metering/REF supplies; add local energy storage where peak events occur.
• Return path: avoid shared narrow ground bottlenecks; keep AFE/REF return away from RF burst currents.
• Scheduling: avoid sampling during known TX windows and switching events; protect “log write” from unstable rails.
• Transient containment: contain relay/eFuse edges and return loops to reduce wideband EMI injection.

Bench + pre-compliance validation matrix with evidence points and pass/fail signals

This plan is designed as an executable checklist. Every test defines evidence points (rails/REF/raw/RF/temp/flags), a measurable pass condition, and a concrete artifact (scope capture, histogram, log snippet, or thermal trace).

Validation matrix

Symptom → first 2 measurements → discriminator → first fix (repeatable templates)

This playbook is designed for fast field isolation. Each case uses the same four blocks so diagnosis stays consistent: Symptoms, First 2 measurements, Discriminator, and First fix. Evidence focuses on rails/REF, raw sample behavior, fault flags/reset reasons, RF counters, and thermal points.

Replaceable families per functional block (criteria → MPN examples → drop-in notes)

This chapter is structured for RFQs: each functional block lists selection criteria, representative MPNs grouped by tier, and drop-in notes that prevent rating/package/flag mismatches. Example parts are representative; confirm voltage/current/surge ratings and footprint/thermal constraints before final BOM lock.