Power Path Management for Energy Harvesting IoT Devices: Architecture, Component Selection, and Design Trade-offs

Design Trade-offs in Power Path Management for Energy Harvesting IoT Devices

 

IoT devices that generate their own power from ambient sources — solar cells, thermal gradients, vibration, or RF — operate under a constraint that battery-powered devices do not face: the energy supply is intermittent, variable in magnitude, and sometimes absent entirely. Power path management is the engineering discipline that determines how this unreliable energy flow is captured, stored, regulated, and delivered to the load in a way that keeps the system functional across the full range of harvesting conditions, including zero available harvest.

Getting power path design wrong produces failure modes that are difficult to reproduce and expensive to fix in deployment. A device that boots reliably on a well-lit desk may fail to cold-start in winter indoor conditions. A system that works during active harvesting may brownout during radio transmission bursts because the power path cannot supply peak current from storage. A device with the wrong storage element may lose state overnight due to supercapacitor leakage, even though it collected enough energy during the day to function.

This article covers the specific design decisions in energy harvesting power paths: architecture selection, cold start design, storage element trade-offs, MPPT strategy, PMIC selection criteria, and the firmware coordination patterns that make energy-aware embedded systems reliable in the field.

Energy Harvesting Power Path Architecture

An energy harvesting power path moves energy from the source through four functional stages: harvester interface, storage management, load regulation, and system monitoring. Each stage introduces decisions with tradeoffs between efficiency, complexity, and robustness.

H3: Three Primary Architectures

The three standard power path architectures for energy harvesting systems differ in how they handle the relationship between the harvesting source, storage, and load:

Storage-first architecture routes all harvested energy to storage, and the load draws from storage exclusively. The system wakes when storage voltage reaches a defined threshold and operates until voltage drops below a minimum operating level, then returns to sleep. This architecture is simple, provides the cleanest separation between harvesting and load behavior, and works well for systems with infrequent duty cycles where the energy accumulated over long quiescent periods is sufficient for each active burst. The tradeoff is that any mismatch between harvest profile and duty cycle — either the system wakes before enough energy is stored, or excess energy is wasted because storage is already full — reduces efficiency.

Simultaneous source-and-storage supply connects harvested power to both load and storage simultaneously, with the PMIC managing the split based on available power and storage state. This architecture is more efficient in active harvesting environments where harvested power consistently exceeds immediate load requirements. It requires more careful regulation to prevent harvested power from being directed entirely to load when storage is depleted, leaving the system unable to handle load transients.

Priority power path switching maintains a priority hierarchy between multiple sources — harvested energy, stored energy, and external backup — routing current from the highest-priority available source and falling back as sources become unavailable. This architecture is most robust for systems with multiple harvesting inputs or where an external supply is available intermittently. It requires logic to manage source priority and prevents reverse current flow between source paths.

Architecture comparison

Architecture

Best for

Primary weakness

Storage-first

Low duty cycle, predictable harvest

Energy waste when storage full; wake timing sensitivity

Simultaneous supply

Continuous harvesting, active systems

Regulation complexity; risk of storage depletion under burst loads

Priority switching

Multi-source, hybrid external/harvested

Higher PMIC complexity; requires priority logic and isolation

Cold Start Design

Cold start is the system's ability to power up from a fully discharged storage element. This is one of the most demanding design requirements in energy harvesting, because the conditions that require cold start — extended periods without harvest — are also the conditions under which the least energy is available for the startup sequence.

The minimum energy required for cold start must be met before the system is allowed to enable the main load. If the main load is enabled prematurely — before storage voltage reaches minimum operating level — the resulting current draw collapses the storage voltage and prevents the load from initializing. This produces a stuck state where the system repeatedly attempts to boot, fails, and prevents the remaining storage energy from accumulating to the threshold required for a successful boot.

The e-peas AEM13921 PMIC demonstrates current cold start capability: it initiates cold start from a minimum input of 1.5 µW at 275 mV, and following cold start the boost converter operates down to 120 mV input. The TI BQ25570 cold starts from approximately 330 mV input voltage. These specifications define the minimum illumination or thermal gradient required to initiate system startup — an important parameter for indoor solar applications where light levels may be 10–50 lux rather than direct sunlight at 50,000 lux.

Supercapacitor-based storage creates a specific cold start consideration: the supercapacitor's voltage rises slowly from zero during initial energy accumulation, and the system must wait until sufficient voltage has accumulated before enabling the load. The accumulation time depends on the harvested power level and the supercapacitor capacitance. For a 100 mF supercapacitor charged from 100 µW harvest power, reaching 2V storage takes approximately 200 seconds at perfect efficiency — a waiting period that must be designed into the system behavior, with the PMIC preventing any load enable until the threshold is reached.

Storage Element Selection

The storage element determines how much energy can be buffered between harvest events, how quickly that energy can be delivered to burst loads, and how long the system can operate without any harvesting input. No single storage technology is optimal across all applications.

Storage type

Energy density

Cycle life

Self-discharge

Charge/discharge rate

Key use case

Supercapacitor (EDLC)

Low

>500,000 cycles

High (weeks to months)

Very fast

Burst loads, high duty cycle, no battery permitted

Li-ion / LiPo

High

500–1,000 cycles

Very low

Moderate

Long-term storage, stable voltage, low quiescent current

Thin-film battery

Medium

10,000+ cycles

Very low

Moderate

Ultra-thin form factor, long cycle life, implantable

Hybrid lithium supercapacitor

Medium-high

10,000+ cycles

Low

Fast

Balance of energy density and power delivery

Supercapacitors are preferred when the duty cycle is high (system active frequently, storage needed for seconds to minutes between harvests), when battery chemistry creates safety or regulatory constraints, when cycle life requirements exceed battery capabilities, and when the system must tolerate peak loads that would stress a battery's charge rate. The primary limitation is voltage droop — a supercapacitor's terminal voltage drops linearly with state of charge, which means the load regulator must handle a wide input range, and the usable energy window is determined by the minimum operating voltage of the downstream regulation.

Li-ion batteries are preferred for systems with long quiescent periods where the self-discharge of supercapacitors would drain storage between harvest events, for systems with moderate duty cycles where the higher energy density of Li-ion allows longer inter-harvest operation, and for systems where stable voltage from the storage element simplifies regulation. The cycle life limitation is relevant for high-duty-cycle systems: a system that charges and discharges a Li-ion cell twice daily will consume its cycle budget in approximately one to two years.

For many production IoT deployments, a hybrid approach — a small Li-ion cell for long-term energy storage plus a supercapacitor for peak current buffering — provides the benefits of both: the energy density of Li-ion for overnight operation and the current delivery capability of supercapacitors for radio transmission bursts.

MPPT Design and Photovoltaic Source Matching

Maximum Power Point Tracking (MPPT) is the control technique that maximizes energy extracted from sources with nonlinear power-voltage curves, particularly photovoltaic cells. A PV cell's power output peaks at a specific voltage (the maximum power point) that varies with illumination. Operating the cell at any other voltage delivers less power.

The e-peas AEM13920 PMIC uses open-circuit voltage ratio MPPT: it briefly disconnects the harvest source to measure the open-circuit voltage (Voc), then regulates the harvesting point to a configurable fraction of Voc (default 75%). This Voc ratio is empirically accurate for silicon PV cells across most illumination conditions and adds minimal overhead relative to continuous fractional open-circuit approaches. The AEM13920's average power monitoring (APM) feature allows the MCU to read total energy transferred from each source to storage and from storage to load, enabling state-of-charge estimation and energy-aware scheduling.

For non-photovoltaic sources — thermoelectric generators (TEGs), piezoelectric harvesters, or RF receivers — source impedance matching is more important than voltage-point tracking. TEGs have a relatively constant source impedance; operating the PMIC's input at the source's maximum power point requires the boost converter input impedance to match the TEG's source impedance at the operating temperature. Some PMICs support configuring each boost input independently for MPPT or constant voltage mode, which allows pairing a TEG with fixed MPPT setpoint alongside a PV cell with dynamic VOC-ratio MPPT — as implemented in the e-peas AEM13921 dual-source PMIC.

PMIC Selection Criteria

Energy harvesting PMICs are specialized components and the selection decision has significant consequences for system performance. Generic battery charger ICs are not appropriate for harvesting applications — they lack the low-voltage cold start capability, ultra-low quiescent current, and MPPT logic that harvesting requires.

Primary selection criteria:

  • Minimum startup voltage: Must be lower than the minimum output voltage of the harvest source under the worst operating condition (winter indoor light for solar, minimum vibration for kinetic). Values from 275 mV (e-peas) to 330 mV (TI BQ25570) represent current state of the art.
  • Input power range: Must cover the range from minimum harvesting conditions to maximum. Range of 1.5 µW to 100+ mW covers most ambient source applications.
  • Conversion efficiency: Must remain above 80–85% at the lightest expected load currents (sub-100 µA). Many PMICs designed for battery charging operate at poor efficiency below 1 mA — this is unacceptable for harvesting systems where most energy transfer occurs at µW–mW levels.
  • MPPT strategy: Fractional Voc is standard and adequate for PV applications. Systems combining multiple source types may benefit from independent per-channel MPPT configuration.
  • Storage element compatibility: Verify that the PMIC's charging profile is compatible with the specific storage chemistry (Li-ion, LiPo, Li-ceramic, supercapacitor). Overvoltage on supercapacitors accelerates degradation; overcharge of Li-ion is a safety issue.
  • Quiescent current: The PMIC's own operating current must be negligible relative to harvested power. Sub-100 nA quiescent current is achievable in current devices and is necessary for systems where harvest power is measured in single-digit microwatts.

Reference IC comparison

IC

Manufacturer

Cold start

Input range

MPPT type

Notable feature

AEM10941

e-peas

380 mV

µW to 50 mW

Voc ratio

Solar + RF or TEG

AEM13920

e-peas

275 mV at 5 µW

µW to 100 mW

Voc ratio + APM

Average power monitoring

AEM13921

e-peas

275 mV at 1.5 µW

Dual boost, 120 mV post-startup

Per-channel MPPT/CV

Dual source simultaneous

BQ25570

Texas Instruments

330 mV

µW to 100 mW

Voc fraction

JEITA compliance, wide adoption

LTC3106

Analog Devices

310 mV

µW to 100 mW

Fixed ratio

Priority power path switching

System Reliability Techniques

Brownout detection prevents the system from operating below its minimum functional voltage. Without brownout protection, a system that begins operating with insufficient energy will enter an undefined state as voltage drops — potentially corrupting flash or EEPROM, transmitting corrupted radio packets, or hanging in a state where the watchdog cannot recover it.

The PMIC should be configured to disable the load output when storage voltage drops below the minimum operating voltage of the system. The hysteresis between enable threshold and disable threshold must be sufficient to prevent rapid on/off cycling when storage voltage hovers near the threshold during low-harvest conditions.

Reverse current protection prevents stored energy from flowing back through the harvesting source when the source voltage is lower than storage voltage. This is relevant for PV systems at night and for RF harvesting when the ambient RF field is absent. Most energy harvesting PMICs include integrated reverse current blocking, but this must be verified for the specific power path configuration.

Pre-charge detection for supercapacitors ensures that large supercapacitors are charged slowly on first power-up to prevent in-rush current from collapsing the harvesting source. Some PMICs include soft-start or pre-charge modes for this purpose.

The watchdog timer is the last line of defense against firmware hangs in energy-constrained states. Energy harvesting systems may encounter power levels insufficient to complete normal firmware execution — if the MCU hangs in a partially powered state, only a watchdog reset can recover it. The watchdog must be configured with a timeout appropriate for the slowest legitimate firmware operation, not the fastest.

 

Software Coordination with Energy Availability

Firmware Coordination with Energy Availability

An embedded system that does not know how much energy is available cannot make intelligent decisions about when to perform expensive operations like radio transmission. A firmware policy that transmits unconditionally on a fixed schedule will fail when energy is insufficient for transmission; a firmware policy that checks storage voltage and adapts its behavior can maintain reliable operation across the full range of harvesting conditions.

The practical implementation requires three capabilities: energy state measurement, task scheduling based on energy state, and state persistence across power cycles.

Energy state measurement uses an ADC to monitor the storage voltage. For a supercapacitor, voltage maps directly to stored charge. For a Li-ion cell, voltage maps to state of charge through the cell's discharge curve. PMICs with average power monitoring (APM) provide energy accounting directly without requiring external ADC readings.

Adaptive task scheduling defines energy thresholds for each operational mode. At minimum energy (storage at 20% of operating range): retain only the most critical measurements; do not transmit. At moderate energy (50–80%): transmit at reduced frequency. At full energy: transmit at normal frequency and perform secondary functions. The Silicon Labs xG22E Explorer Kit with e-peas AEM13920 PMIC demonstrates this approach in practice: the MCU reads storage voltage, wakes on RF transmission events when energy permits, and returns to sleep when energy is insufficient.

State persistence before shutdown allows the system to resume from its last state rather than reinitializing from scratch after an undervoltage event. FRAM is preferred over EEPROM for this purpose because it supports unlimited write cycles without wear, executes writes faster (critical when voltage is dropping), and consumes less power per write operation.

Quick Overview

Key Applications: battery-less outdoor sensor nodes (solar or vibration), industrial IoT condition monitoring with TEG harvesting, smart building sensors with indoor light harvesting, remote agricultural monitoring, electronic shelf labels with RF and PV harvesting, wearable biosensors with body heat harvesting

Benefits: eliminates battery replacement maintenance cost (technicians spend 20% of time on battery changes in large IoT deployments); enables deployment in inaccessible locations; dual-source PMICs provide redundancy when one source is unavailable; MPPT extracts 10–30% more energy than fixed-voltage operation; adaptive firmware scheduling maintains reliable operation across the full harvesting range

Challenges: cold start requires minimum input power that may not be available in worst-case field conditions; supercapacitor self-discharge prevents multi-day storage; Li-ion cycle life limits high-duty-cycle systems; brownout without state persistence causes data loss; storage voltage measurement alone is insufficient for precise energy accounting without APM

Outlook: dual-source PMICs (e-peas AEM13921) enabling simultaneous harvesting from independent sources; sub-µW cold start improving reliability in minimal-light indoor applications; hybrid lithium supercapacitors combining energy density with high cycle life; FRAM becoming standard for state persistence in energy-harvesting systems; energy harvesting market driven by elimination of battery replacement cost at large deployment scale

Related Terms: MPPT, cold start, supercapacitor, thin-film battery, PMIC, BQ25570, AEM10941, AEM13920, AEM13921, LTC3106, brownout detection, EDLC, TEG, photovoltaic harvesting, Voc ratio, average power monitoring, FRAM, duty cycle scheduling, quiescent current, storage element sizing, power path switching

 

Contact us

 

 

Our Case Studies in This Industry

 

FAQ

What is the difference between MPPT and fixed voltage operation in energy harvesting systems?

 

Maximum Power Point Tracking, or MPPT, dynamically adjusts the operating voltage of the harvesting source to extract maximum power under varying conditions. For a PV cell, the maximum power point changes with illumination, at low light levels, the Voc drops and the optimal operating voltage shifts accordingly. A fixed-voltage MPPT threshold set for bright-light conditions will underperform at low light. MPPT using fractional Voc, measuring open-circuit voltage periodically and operating at a fixed percentage, typically 70 to 80% of Voc, tracks the maximum power point across changing conditions without requiring a complex hill-climbing algorithm. Fixed voltage operation is simpler and appropriate for sources with stable characteristics, such as TEGs operating across a fixed temperature gradient. For sources with variable output, MPPT delivers 10 to 30% more energy than fixed-voltage operation at typical operating conditions.
 

How do you size a supercapacitor for an energy harvesting IoT device?

 

Supercapacitor sizing balances energy storage capacity against physical size, cost, and self-discharge rate. The minimum required capacitance is determined by the energy needed for one complete duty cycle, in joules, divided by the square of the operating voltage range: C = 2E / (V_max² - V_min²), where V_max is the maximum charge voltage and V_min is the minimum operating voltage before the system must sleep. For a system requiring 1 mJ per duty cycle operating between 3.3V and 2.0V: C = 2 × 0.001 / (3.3² - 2.0²) = 2 × 0.001 / (10.89 - 4.0) = 0.29 mF. This is the theoretical minimum. Practical designs add a safety margin of 2 to 5x to account for capacitance variation with temperature, aging, and ESR losses. Self-discharge must also be checked: a supercapacitor that loses 50% of its charge in 24 hours will leave a system that harvests only during daylight unable to complete its midnight duty cycle regardless of how much energy it collected during the day.
 

What is cold start in energy harvesting and how is it designed for?

 

Cold start is the system's ability to initiate power-up from a fully discharged state using only the energy available from the harvesting source at startup time. It is the most demanding operating condition because the harvest source may provide only microwatts at low illumination or minimal thermal gradient, but the system requires a minimum voltage and current to begin operation. Cold start design requires a PMIC with a dedicated startup circuit that operates at very low input power, below the threshold that triggers the main converter, and accumulates charge in a small bootstrap capacitor until enough energy is available to enable the main power path. Key parameters are minimum startup voltage, how little voltage the PMIC can use to begin accumulating charge, and minimum startup power, how few watts are required. Practical cold start from ambient indoor light requires startup voltage below 400 mV and startup power below 10 µW.