Skip to content

Power architecture and duty cycling

Designing power architectures for connected devices requires balancing peak pulse requirements, sleep quiescent leakage, supply voltage ranges, and thermal dissipation. Whether an endpoint relies on a decade-long primary lithium cell, ambient solar energy harvesting, or Power-over-Ethernet (PoE), power architecture dictates system reliability and field operating life.

This chapter details power supply topologies, contrasts DC-DC conversion architectures, provides guidelines for firmware duty-cycling, and outlines hardware brownout mitigation mechanisms.

1. Power Supply Topologies

Connected systems utilise four primary power intake topologies:

mermaid diagram

Topology Input Range Primary Advantages Critical Design Challenges Best Use Cases
Primary Lithium 2.0 V to 3.6 V Decade-long life without external wiring Severe peak pulse voltage drop, passivation Remote telemetry, pipeline monitors
Solar Harvesting 0.5 V to 6.0 V (PV) Indefinite operational lifetime Multi-day cloudy periods, winter freeze degradation Outdoor weather stations, agricultural IoT
PoE (802.3af/at) 37 V to 57 V DC Single cable for Gigabit data and 15W-30W power Requires isolated flyback transformer, magnetics Building automation, security cameras, gateways
Mains AC/DC 85 V to 264 V AC Unlimited continuous electrical energy High voltage isolation (4 kV surge), safety certs Industrial gateways, street light controllers

Common power intake architectures across endpoint and gateway designs.

Comparison of power intake architectures.

2. Voltage Regulation: LDO vs DC-DC Switching Converters

Microcontrollers and RF radios require stable voltage rails (typically 3.3 V, 1.8 V, or 1.2 V core). Converting battery or bus voltage down to these rails requires evaluating the trade-offs between Low-Dropout Regulators (LDOs) and Switched-Mode DC-DC Buck Converters:

mermaid diagram

Linear LDO versus switched-mode inductive DC-DC conversion.

2.1 Efficiency Comparison and the Quiescent Current Trap

Linear regulators exhibit electrical efficiency proportional to the voltage ratio:

\[\eta_{LDO} = \frac{V_{out}}{V_{in}} \times 100\%\]

When regulating a \(3.6\,\mathrm{V}\) primary cell down to \(1.8\,\mathrm{V}\) for an ultra-low-power MCU core, an LDO operates at only \(50\%\) efficiency, dissipating the remaining \(50\%\) as waste heat. A synchronous DC-DC buck converter delivers \(90\%\) to \(95\%\) efficiency under load.

Warning

While DC-DC buck converters excel at active currents (\(> 10\,\mathrm{mA}\)), their internal oscillator and switching circuits consume quiescent current (\(I_Q\)) ranging from \(10\,\mu\mathrm{A}\) to \(50\,\mu\mathrm{A}\). In an ultra-low-power sensor that sleeps \(99.9\%\) of the time, this continuous \(I_Q\) draw will drain the battery faster than the radio transmissions!

Architectural Recommendation: Use Dual-Mode Power Management ICs (PMICs) or select ultra-low-quiescent buck converters (such as Texas Instruments TPS62840 with \(I_Q = 60\,\mathrm{nA}\)) that automatically transition to high-efficiency pulse-frequency modulation (PFM) during sleep states.

3. Firmware Duty-Cycling & Low-Power State Execution

Extending battery lifetime requires the firmware to treat every milliampere-second of active compute as a scarce resource.

mermaid diagram

Discrete event timeline for deterministic low-power sensor duty-cycling.

3.1 Firmware Power Optimization Rules

  1. Tickless RTOS Scheduling: Standard RTOS kernels maintain a periodic 1 ms tick interrupt that wakes the processor 1,000 times per second. Firmware must enable tickless idle (for example, configUSE_TICKLESS_IDLE in FreeRTOS), which programs hardware low-power timers (LPTIM) to sleep for seconds or hours until the next actual software timer expires.
  2. Peripheral Power Rail Gating: Idle sensors, level shifters, and external memory chips frequently leak \(5\,\mu\mathrm{A}\) to \(50\,\mu\mathrm{A}\) in standby. Route all peripheral power supplies through a high-side P-channel MOSFET load switch controlled by an MCU GPIO, keeping external circuits completely de-energized during sleep.
  3. Dynamic Voltage and Frequency Scaling (DVFS): When executing sensor math or cryptographic hashing, run the core clock at maximum frequency (for example, 80 MHz) to complete computations as fast as possible (“race-to-sleep”). Once computation finishes, immediately drop the clock down to 32 kHz or enter deep sleep.

4. Hardware Reliability & Brownout Mitigation

When primary batteries degrade or experience cold temperatures, high-current RF transmission bursts induce sharp voltage droops. If the system rail drops below the digital logic threshold, the microcontroller enters a chaotic undefined state.

mermaid diagram

Multi-tier voltage supervision mitigating catastrophic brownouts.

4.1 Independent Watchdog (IWDG) Architecture

Firmware can hang due to electrostatic discharge (ESD), infinite loops, or locked external buses. A dedicated Independent Watchdog Timer (IWDG) is mandatory: - The IWDG must be clocked from an independent dedicated internal low-speed RC oscillator (32 kHz), completely decoupled from the main system PLL and crystal oscillators. - The watchdog timeout must be configured to reset the system if the main loop fails to refresh the watchdog within a bounded window (typically 4 to 8 seconds). - The watchdog service function must only be called from the primary operational task after verifying that all sub-tasks and IPC queues are healthy. Never refresh a watchdog unconditionally inside a hardware timer interrupt service routine (ISR).