Batteryless IoT You Can Actually Deploy: Energy Harvesting, Storage, and Reliable Firmware

Why Batteryless IoT Is Finally Practical

For years, “no battery” sensors sounded like science fair projects: fragile, unpredictable, and too fussy to live in real spaces. That has changed. Energy harvesting parts are mature, microcontrollers sip power in the nanoamp range, and wireless stacks have learned to whisper.

This guide shows how to deploy batteryless devices you can trust. We will cover source selection (light, heat, motion), power managers, storage, communications that don’t drain your budget, and the firmware patterns that make devices reliable when power flickers. You will not need exotic parts or secret factory tuning. You will need good energy math and a few habits that are different from “always-on” design.

What “Batteryless” Really Means

Batteryless does not mean “powerless.” It means your device earns the energy it spends, in small, irregular payments. That has three consequences:

Energy is bursty. You may harvest microwatts for minutes, then spend milliwatts for milliseconds.
Power may disappear anytime. Your device must survive brownouts without corrupting state or bricking itself.
Duty cycle beats peak specs. The average, not the instantaneous number, decides if the design works.

Keep those three in mind. They shape your hardware, radio, and firmware choices.

Harvesting Options: Light, Heat, and Motion

You probably do not need a new kind of physics to power your sensor. Start with sources that are easy to site and model.

Light (Photovoltaic)

Indoor light is your best friend for homes, offices, and stores. It is steady, predictable, and easy to aim. A rule of thumb:

200–300 lux (typical office) → ~10–50 µW/cm² for indoor-optimized PV.
Bright retail lighting → ~100–150 µW/cm².
Sunlit window ledge → orders of magnitude higher, but plan for cloudy days.

Indoor PV modules tuned for LEDs work well. They are small, light, and safe. You mount them where people already expect objects: near signage, on door frames, or the top of shelving. Brands producing indoor PV include Epishine and others focused on low-light performance.

Thermal (TEGs)

Thermoelectric generators (TEGs) produce power from a temperature difference. They shine in places with warm pipes, HVAC vents, or small industrial enclosures where a few degrees exist. Expect microwatts to milliwatts with modest ∆T (2–10°C). Mechanical coupling and heat sinking matter more than the chip here.

Motion and Vibration (Piezo, Electromagnetic)

Vibration harvesters can power event-driven devices like a door-open sensor or a tool-usage counter. If something already moves by design (a latch, a pedal), you can harvest a short, high-power pulse. A small supercapacitor then dribbles it into the PMIC.

RF and Backscatter

Backscatter and RF harvesting exist, but they need dense RF fields or dedicated infrastructure. Systems like ambient IoT “pixels” solve this in retail and logistics with tuned readers. In general buildings, it’s safer to plan around light or motion unless you control the RF environment.

Power Management You Can Trust

The heart of a batteryless design is a harvester PMIC that cold-starts on tiny power, charges a storage element, and gives your MCU a clean rail with brownout handling. Two proven families:

TI BQ25570: Ultra-low-power boost charger + buck converter, excellent for PV and TEG sources with very low input power. Cold-start around 330 mV (typical) and high efficiency at micro-watt levels.
e-peas AEM series (e.g., AEM10941): Optimized for indoor PV with MPPT (maximum power point tracking), multiple storage options, and deep cold-start capability.

Storage: Supercaps, Thin-Film, and “Tiny Batteries”

Your storage picks the burst size you can afford and how your node behaves in the dark.

Supercapacitors (0.1–5 F): Perfect for high-cycle, long-life buffering. Watch leakage current and ESR. Many designs land around 0.47–1 F at 2.7–5.5 V for indoor PV nodes.
Thin-film or solid-state microbatteries: Very low self-discharge, safe chemistry, but smaller capacity and costlier. Good where daily dark periods are certain.
No storage (pure transient): Viable for NFC-scanned tags or backscatter systems. Simpler, but your sensor only works when stimulated.

Whatever you choose, include input clamps and ESD handling, and plan for aging: supercaps will lose capacity over years, and thin-film cells have cycle and temperature limits.

Right-Sizing Your Energy Budget

Energy harvesting succeeds or fails on math. You do not need a simulator—just honest numbers and margins.

Rough Numbers to Guide You

Indoor PV: 10 cm² at 100 µW/cm² → 1 mW under bright LEDs; at 20 µW/cm² → 200 µW.
BLE advertisement burst: 10 mA for 3 ms at 3 V → Q = 30 µC, E ≈ 90 µJ. Three quick adverts might cost ~300 µJ.
Deep sleep MCU: 1–3 µA at 3 V → 3–9 µW continuous.
Sensor read (e.g., temp/humidity): 500 µA for 10 ms at 3 V → 15 µJ.

Put it together: If you harvest 200 µW on average, in 10 seconds you gather 2 mJ. That covers a sensor read + a short BLE advert burst comfortably, with reserves to spare for brownout margin.

Practical Steps

Measure your scene: Use a cheap lux meter. If you see 150–300 lux most days, indoor PV is viable.
Instrument your prototype: Use a USB power meter or source-measure unit to log current during sleep, sensor read, and radio.
Aim for 2× margin: If you need 100 µW average, design for 200 µW harvested in typical lighting.

Communications That Don’t Drain You

Pick radios and modes that keep link time short and state simple.

Bluetooth Low Energy, Advertising-First

BLE is the easiest win for short-range telemetry. Advertising-only payloads avoid connection overhead. Pair it with a phone or a gateway that scans continuously on mains power. You can encode small, self-describing payloads (e.g., CBOR or a compact TLV) in manufacturer data fields.

Pros: Native support in phones, no join process, sub-mJ per event.
Cons: Limited payloads, needs a nearby scanner.

Bluetooth Mesh LPN/Friend

Mesh can work if your node is a Low Power Node (LPN) with a mains-powered Friend. The LPN sleeps and opportunistically polls. Good for buildings with existing mesh lighting or gateways.

802.15.4 (Thread/Zigbee)

Also viable in sleepy end-device modes with a router nearby. Keep polling intervals long and payloads tiny. This is strong when you already plan a Thread border router.

NFC for Configuration, Data, or Power

An NFC interface is gold for provisioning and offline reads. You can wake a dead node by tapping with a phone, fetch the last buffered measurements, or update settings without radio beacons.

What to Avoid

Wi‑Fi for batteryless is rarely a fit. Even brief associations cost too much.
High-rate or long-range bursts unless they happen very rarely with a large storage buffer.

Firmware for Intermittent Power

Even perfect hardware fails if firmware expects stable rails. Intermittent power needs a different mindset.

Make Every Action Idempotent

Assume a brownout can hit at any line of code. Writes must not corrupt, counters must not double-count, and comms must not repeat forever. Patterns that help:

Journaling: Append-only logs in FRAM or flash; commit records with checksums.
Two-phase updates: Write new data to a shadow record, then flip a valid flag.
Monotonic sequence numbers: So gateways can deduplicate transmissions after resets.

Use a Simple Energy State Machine

Track storage voltage and step through Harvest → Ready → Transmit with hysteresis. For example:

Below 2.2 V: Harvest only; MCU in deepest sleep; no sensors.
2.2–2.7 V: Allowed to sample sensors and buffer data; no radio.
Above 2.7 V: One radio burst, then return to Harvest if voltage drops below 2.5 V.

This avoids “thrashing” around the radio threshold and wasting energy on partial transmissions.

Guard Against Brownouts

Enable BOR (brownout reset) at the highest safe threshold.
Boot fast: Do nothing until clocks settle and rails are valid.
Precompute sensor configs; avoid runtime allocations.

Prefer FRAM (or Add External FRAM)

FRAM writes fast at low energy and survives many cycles. If your MCU lacks it, add a small external FRAM (SPI/I²C). If you must use flash, batch writes and wear-level.

OTA Updates (If You Must)

Do not rely on OTA for frequent updates. Batteryless nodes have tight energy windows. If you need updates, use NFC, a pogo pin, or stage the update in many tiny chunks with a rock-solid resume scheme and a dual-bank bootloader that never leaves you bricked.

Hardware Blueprints You Can Build

Indoor PV Room Climate Beacon

Goal: A temperature and humidity beacon every 5–10 minutes in typical office light.

Harvester: Indoor PV (10–20 cm²) into AEM10941 or BQ25570.
Storage: 0.47–1 F 5.5 V supercap.
MCU/Radio: Ultra-low-power BLE SoC.
Sensors: Integrated T/RH like BME280 or similar; sample in burst mode.
Firmware: Sleep at 1–2 µA; wake, sample, encode compact payload; 3 adverts on 3 channels; back to sleep.

Expected: Reliable 5–10 min updates in 200–300 lux and overnight persistence thanks to the supercap.

Door-Open Counter Using Motion Harvesting

Goal: Count door cycles in a storeroom without wires or batteries.

Harvester: Small piezo or electromagnetic harvester coupled to the hinge or latch.
Storage: 0.22 F supercap to catch the pulse and smooth power.
MCU/Radio: BLE advertising-only or a pulse counter with NFC readout.
Sensors: Magnet + reed switch for an unambiguous open/close signal.

Trick: Gate the radio to only report every N events or when storage voltage is healthy. Otherwise, buffer counts in FRAM and send later.

Plant-Soil Moisture Sentinel

Goal: Alert when soil is too dry, with months of indoor operation under diffuse light.

Harvester: Flexible indoor PV on a planter rim.
Storage: Thin-film microbattery (for night) + supercap hybrid.
Sensing: Corrosion-safe AC excitation across probes or a capacitive sensor to avoid electrode wear.
Comms: BLE advert only when threshold crossed, plus a daily heartbeat at midday when light is best.

Detail: Moisture reads are energy cheap; the bottleneck is radio. Batch alerts to peak-light windows.

Design Tactics That Save Orders of Magnitude

Move Work to the Gateway

Keep nodes dumb and cheap. Let the gateway or cloud do parsing, smoothing, and compression. On the node, use a fixed payload layout and simple math. Every CPU cycle you don’t burn is energy you can spend on a better link.

Advertise When It’s Bright

Schedule transmissions when you have harvest headroom: near windows at midday or during open hours. In darkness or low-light, sample-only and defer radio.

Keep Payloads Small

Use compact encodings: CBOR, varints, or tightly packed fields.
Delta encode against the last sample.
Send events not raw streams. “Dry → Wet at 14:32” beats “0.4, 0.41, 0.38…”

Show State Without Power

If users need to see something, consider an e-paper icon that updates only on threshold changes. A tiny tri-color symbol (“OK / Alert”) avoids energy-hungry radios for everyday status checks.

Deployment: Placing, Provisioning, and Monitoring

Site Survey

Check light patterns: A hallway may be dim except during specific hours; aim PV where it sees its best hours.
Mind shadows and dust: Even a thin film of dust can cut harvested power over months. Plan wipe-downs during routine cleaning.
Thermal fit: For TEGs, ensure a safe, stable heat sink and no condensation risk.

Provisioning Without Headaches

Use NFC for keys and config; avoid BLE pairing sessions that might brown out.
Pre-burn identities at assembly and mark devices physically with short codes or QR for mapping to locations.
Gateway-side dedup: Accept repeats; trust sequence numbers.

Security That Fits

Lightweight AEAD (e.g., AES‑CCM with short nonces) is enough for beacons. Rotate keys in batches at the gateway and apply during high-light windows or by NFC tap. Sign payloads if they hop over untrusted relays.

Monitoring and SLAs

Track three simple metrics per node:

Heartbeat rate vs. expectation (e.g., daily at noon).
Storage voltage snapshots (min/max) to spot failing PV or aging caps.
Last-light level or derived harvest health flag.

Alarms should nudge placement (“move closer to light”) before they page someone to “replace a battery” that doesn’t exist.

What Not to Do

Do not use Wi‑Fi. It is a mismatch for batteryless behavior in most cases.
Do not refresh e-paper often. Full refreshes are energy expensive; only update on threshold changes.
Do not run continuous sensors (e.g., mmWave presence) unless harvest is exceptional.
Do not rely on OTA updates as your main config channel.

Choosing Parts: A Practical Shortlist

Harvesters and PMICs

TI BQ25570: For ultra-low-power PV/TEG, with integrated buck for system power.
e-peas AEM10941: MPPT for indoor PV; flexible storage support.

Storage

Supercapacitors: 0.47–1 F, low leakage; pick reputable vendors and check lifetime curves at your max voltage and ambient temp.
Thin-film cells: For overnight resilience where light drops to zero predictably.

MCU/SoC

BLE SoCs with true microamp sleeps and fast wake. Verify sleep current on your board, not just the datasheet.
FRAM MCUs or add external FRAM if you journal often.

Sensors

Temp/humidity with low standby currents and fast conversions.
Magnetic reed or hall sensors with near-zero idle draw.
Capacitive soil probes to avoid corrosion.

Intermittent Compute Patterns That “Just Work”

Checkpoint Small, Often

Keep a tiny status word in FRAM or a redundant flash slot: energy state, last-seq, and last-sensor sample. Write it before any radio call. On boot, assume the previous operation did not finish; resume idempotently.

Bounded Work Units

Make every task finish within a hard energy budget. Example: “Read sensor, average 4 samples, and advertise once” is one unit. If storage V drops mid-unit, abort gracefully and try again later.

Use Event Time, Not Wall Time

Clocks drift when you have intermittent rails. Time stamps can be “days since install” or “gateway time at receipt.” If you need wall time on the node, re-sync often in bright periods and accept low precision.

Real-World Reliability: Failure Modes and Fixes

Dim Winters and Closed Blinds

Expect seasonal dips. Options:

Oversize PV area where aesthetics allow.
Use a thin-film buffer to survive nights plus several dim days.
Reduce cadence in low-light automatically.

Cap Aging

Supercaps lose capacity over years, especially near their max voltage and warm rooms. Derate voltage by 10–20% and place away from heat sources. Monitor min/max Vcap to flag aging early.

Radio Congestion

In dense BLE environments, your 3 adverts may collide. Randomize intervals, use channel maps wisely, and include sequence numbers so gateways can collapse dupes over a wider scan window.

Moisture and Corrosion

Protect exposed pads, conformal coat where needed, and prefer capacitive sensing. For leak detectors, design the pad geometry to encourage wicking only when water is present, not humidity.

Cost and ROI: Where Batteryless Pays Off

Coin cells are cheap until you pay people to replace them—or until you miss an alert because a battery died months ago. Batteryless wins when:

Access is costly: Overhead signage, cold rooms, vaulted ceilings.
Counts are high: Hundreds of nodes where replacing cells is endless.
Data is sparse: Door cycles, threshold alerts, daily climates rather than streams.

The bill of materials may be slightly higher than the absolute cheapest battery node. But the lifecycle cost, reliability, and sustainability story are much better. No stocking batteries. No hazardous waste bins. No surprise outages.

Where to Start This Week

Prototype a single indoor PV node with a harvester PMIC, a 0.47 F cap, a BLE dev kit, and one sensor. Prove an advert every 5–10 minutes under your office lights.
Add NFC for provisioning and off-grid reads. Few things boost confidence like tapping a “dead” node and seeing buffered data appear.
Instrument voltage over a week. If Vcap bottoms out, adjust cadence or move the PV.
Write firmware as if power will vanish—because it will. Journal, bound work, and test BOR paths.

What’s Next: Smarter Harvest, Same Budget

As you scale, you can get fancier without burning more energy:

Adaptive MPPT profiles on the PMIC for different rooms.
Local anomaly flags (simple thresholds) to prioritize urgent adverts.
Gateway-assisted scheduling: Beacons from the gateway that tell nodes “now is a good time to speak.”

The goal is steady, boring reliability that outlasts any coin cell you would have installed.

Summary:

Batteryless IoT is practical today with indoor PV, smart PMICs, and low-power radios.
Design around bursty energy, brownouts, and duty-cycle budgets—not peak specs.
Use harvesters like TI BQ25570 or e-peas AEM10941 and pair with supercaps or thin-film storage.
Favor BLE advertising, NFC provisioning, and tiny payloads; avoid Wi‑Fi.
Write intermittent-safe firmware: idempotent updates, journaling, and energy state machines.
Place and provision thoughtfully; monitor heartbeats and storage voltage to catch issues early.
Start with one PV node, prove the energy math, then scale with confidence.