Low-power IoT devices treat battery life as an engineering constraint. Total consumption is expressed by the basic equation: Qdaily = Isleep × 86 400 + Σ(Iactive × tactive). In practice, the two main levers are reducing sleep current to the nano/microampere range and shortening the active window.

Duty cycle and sleep mode

A typical telemetry device spends more than 99 % of its time asleep. Modern microcontrollers offer multiple low-power states:

  • Run / Sleep — CPU halted, peripherals running, current in the mA range.
  • Stop / Standby — clock and partial RAM retained, current of a few µA.
  • Shutdown / Backup — only RTC or backup SRAM, current down to nA.

Wake sources include a real-time clock (RTC), GPIO trigger, watchdog or wake-on-radio. The active window is typically held to 2–8 seconds, within which measurement, protocol processing and transmission are completed.

Battery chemistry comparison

Four chemistries are common in field devices:

  • LiSOCl2 (lithium thionyl chloride) — 3.6 V nominal, ~3.6 V plateau; 700 Wh/kg, <1 %/year self-discharge; −55…+85 °C range; passivation can cause a voltage dip on sudden high-current draw, which is solved with a hybrid layer capacitor (HLC) pack.
  • Li-MnO2 (CR/BR series) — 3.0 V; medium energy density; low cost; consumer grade with balanced high/low-current performance.
  • Li-ion (NMC/LCO) — 3.6–3.7 V, 150–250 Wh/kg, rechargeable; ~2 %/month self-discharge; performance falls at temperature extremes (charging is restricted <0 °C).
  • LiFePO4 — 3.2 V, 90–140 Wh/kg; high thermal safety, 2 000–5 000 cycles; better cold-weather behaviour than LCO.
  • Alkaline (Zn-MnO2) — 1.5 V (a 9 V block contains six cells for 9 V); widely available and cheap; capacity drops noticeably in the cold and internal resistance becomes a problem at high pulse currents.

Latching solenoids draw a millisecond-long high-voltage/current pulse; for this profile a pulse capacitor combined with LiSOCl2/HLC or a 9 V block is often preferred.

Radio consumption and reporting period

The radio is typically the hungriest component in an IoT device. A LoRa Class A device draws 30–120 mA on transmit and around 10–15 mA on receive; time-on-air ranges from 50 ms to 2 s depending on spreading factor (SF7–SF12) and packet size. In NB-IoT, PSM (Power Saving Mode) and eDRX modes for long sleep on shared operator infrastructure are defined in 3GPP TS 23.682. Roughly expected battery life by reporting profile:

  • 1 message per day — 3–10 years depending on chemistry.
  • 1 message per hour — 12–24 months.
  • 1 message every 5 minutes — 3–6 months; for intensive monitoring.
Practical note: battery life depends not only on nameplate capacity but also on temperature, discharge profile (continuous vs pulsed), changes in internal resistance and self-discharge. Cold weather (< 0 °C) can cut alkaline capacity by up to 30 %; LiSOCl2 offers a clear advantage in such environments.

Battery health estimation (SoH/SoC)

Field devices measure voltage both under load and at rest during each wake cycle; the difference yields an estimate of internal resistance. Coulomb counting, integrating the charge transferred in active windows, gives a more accurate State of Charge (SoC). More advanced estimates use a Kalman filter or machine-learning models to predict remaining life from voltage, current and temperature; the typical goal is to issue a planned-maintenance alert two to four weeks before the battery is exhausted.

Energy harvesting

At points with continuous solar exposure, a small (0.5–2 W) photovoltaic panel combined with a supercapacitor or rechargeable cell can make the device energy-independent. Relevant standards include IEC 61724 (PV performance monitoring) and IEEE 1725 (portable Li-ion packs). Thermoelectric (TEG) and vibration harvesting appear in industrial niche applications.