Electronics Guide

Single-Inductor Multiple-Input Architectures

Single-inductor multiple-input converters, described in the literature as SIMI or multiple-input single-inductor topologies, share one magnetic element among several harvesting sources. Because the inductor is usually the largest and most expensive passive component in a microwatt-scale converter, sharing it markedly reduces board area and cost. These converters use time-division multiplexing to allocate the inductor to different sources during successive switching cycles, and advanced control sequences can approximate simultaneous extraction from multiple sources through interleaved operation. The same idea extends to single-inductor multiple-input multiple-output arrangements that also serve several regulated rails from the shared inductor.

The primary challenge in this design is managing cross-regulation, where changes in one source's operating point affect the power extracted from the others. Careful control loop design, appropriate switching sequences, and adequate output capacitance help maintain stable operation. The switching frequency must be high enough to visit every source often relative to the fastest source dynamics, yet low enough that gate-drive and core losses do not consume the harvest. Designers also budget for the dead time between phases, during which the inductor serves no source at all; with many inputs, that idle fraction becomes a first-order efficiency term.

Multiple-Input Buck Converters

When multiple sources produce voltages higher than the desired output, multiple-input buck converters can efficiently step down and combine the sources. Each input has its own high-side switch, with a shared inductor and output stage. Time-multiplexed control activates one input at a time, or more sophisticated control enables simultaneous conduction from multiple sources with appropriate current sharing.

Multiple-Input Boost Converters

Low-voltage harvesting sources such as thermoelectric generators and single photovoltaic cells typically require boost conversion. Multiple-input boost topologies share the boost inductor and output diode among sources, with separate low-side switches for each input. The topology naturally supports maximum power point tracking through duty cycle adjustment for each source.

Buck-Boost and SEPIC Topologies

When source voltages vary above and below the desired output voltage, buck-boost or SEPIC (Single-Ended Primary-Inductor Converter) topologies provide the flexibility to handle the full range. Multiple-input versions of these converters can accommodate sources with widely varying voltage levels, though the additional complexity and component count may reduce efficiency compared to simpler topologies.

Resonant and Soft-Switching Converters

High-frequency operation benefits from resonant or soft-switching techniques that reduce switching losses. Zero-voltage switching (ZVS) and zero-current switching (ZCS) topologies minimize the power dissipated during transistor transitions. These techniques become increasingly important as switching frequencies increase to reduce magnetic component sizes in space-constrained hybrid harvesting systems.

Independent MPPT for Each Source

The most straightforward approach to maximizing power from multiple sources implements independent MPPT controllers for each harvester. Each controller tracks its source's optimal operating point without regard to the others. This approach extracts the most power from every source but requires separate control loops and, in most cases, a separate power stage per source.

Commercial dual-source harvesting power management ICs follow this pattern. The e-peas AEM13920, for example, conditions two independent inputs at the same time, and each input can be configured for maximum power point tracking or for a fixed constant-voltage operating point, so a photovoltaic cell and a thermoelectric generator may be served with settings appropriate to each. The device claims source-to-storage and storage-to-load conversion efficiencies above 90 percent and includes the storage charger and a regulated output rail on the same die.

Perturb-and-Observe Algorithms

Perturb-and-observe (P&O) algorithms incrementally adjust the operating point and measure the resulting power change. If power increases, the algorithm continues in the same direction; if power decreases, it reverses direction. P&O is simple to implement and works across source types, but suffers from oscillation around the maximum power point and may track incorrectly under rapidly changing conditions.

Incremental Conductance Methods

Incremental conductance methods compare the instantaneous conductance to the incremental conductance to determine the position relative to the maximum power point. This approach can theoretically find the exact MPP without oscillation, though practical implementations still exhibit some hunting behavior due to measurement noise and discrete adjustments.

Fractional Open-Circuit Voltage Tracking

For certain source types, the maximum power point voltage is approximately a fixed fraction of the open-circuit voltage. For silicon photovoltaic cells the factor commonly falls between roughly 0.7 and 0.8, while for thermoelectric generators the maximum power point sits near half the open-circuit voltage because of their near-resistive source impedance. Periodically sampling the open-circuit voltage and setting the operating point to the appropriate fraction provides simple, low-overhead tracking. The momentary disconnection from the load and the approximation error are acceptable trade-offs for many applications, and energy-harvesting power management ICs frequently implement this method to keep control overhead within a microwatt-scale budget.

The Texas Instruments BQ25570 illustrates the practical form. Its tracking network briefly disconnects the input, samples and holds the open-circuit voltage, and then regulates the input to a selected fraction of that stored value. Pin strapping selects a ratio of 80 percent, suited to photovoltaic inputs, or 50 percent, suited to thermoelectric inputs, and an external resistor divider sets any intermediate ratio for sources that match neither default. In a hybrid design this pin-level choice is the point at which the converter is tuned to the transducer actually connected.

Model-Based MPPT

If the source characteristics are well-known, model-based algorithms can calculate the optimal operating point from measured parameters such as temperature, illumination, or vibration amplitude. This approach avoids the oscillation and convergence time of iterative methods but requires accurate models and appropriate sensors. Hybrid approaches combine model-based initialization with iterative refinement.

Multi-Source MPPT Coordination

When sources share power conditioning resources, MPPT must be coordinated to avoid conflicts. Time-multiplexed converters may track each source during its connection interval, maintaining operating point estimates between intervals. Coordinated control can also exploit correlations between sources, using conditions measured at one source to improve tracking at another.

Resistive Source Matching

Sources with predominantly resistive internal impedance, such as thermoelectric generators, achieve maximum power transfer when the load impedance equals the source impedance. DC-DC converters present an adjustable effective input resistance through duty cycle control, enabling continuous matching as source conditions change. For an ideal boost converter, the effective input resistance equals the load resistance multiplied by the square of one minus the duty cycle, so increasing the duty cycle lowers the emulated input resistance; a buck converter instead presents the load resistance divided by the square of the duty cycle. This relationship lets a single converter both track the maximum power point and synthesize the matched source impedance.

Capacitive Source Matching

Piezoelectric harvesters present capacitive sources that require different matching strategies. A plain full-wave bridge feeding a smoothing capacitor wastes much of the available energy, because the element's own capacitance must first be charged through the rectifier before any current reaches the load. Optimal extraction instead transfers charge from the piezoelectric capacitance at the correct phase of the mechanical cycle, near the displacement extremes.

Synchronized switch harvesting on inductor (SSHI) circuits close a switch through a small inductor at each vibration peak, resonantly inverting the voltage across the element so that the rectifier conducts sooner and for longer. Published implementations report several times the power of a full-bridge rectifier under comparable excitation, with reported factors spanning roughly three to seven depending on the topology variant, the quality factor of the flipping inductor, and the electromechanical coupling of the harvester. Synchronous electric charge extraction takes a related approach, emptying the element's charge into an inductor at each peak, which makes extracted power largely independent of the load or storage voltage. Both techniques must be self-powered and synchronized without external sensing if they are to remain net contributors at the microwatt level.

Nonlinear Source Matching

Photovoltaic cells and RF rectifiers exhibit strongly nonlinear current-voltage characteristics. The optimal operating point varies with input conditions, requiring adaptive matching that tracks the changing maximum power point. The steep current roll-off beyond the maximum power point demands precise control to avoid operating in the current-limited region.

Reactive Power Considerations

Some harvesting sources exhibit significant reactive impedance components at their operating frequencies. Conjugate matching that cancels the reactive component maximizes real power transfer. Passive matching networks or active impedance synthesis can implement the required reactive compensation, with active approaches enabling adaptive matching as conditions change.

Voltage Bus Combining

Individual source converters can output to a common voltage bus, with currents from each source combining naturally. This approach requires each converter to regulate its output to the bus voltage while maximizing input power extraction. Output current limiting and reverse-current blocking prevent sources from loading each other.

Current Summing

Current-output converters can sum their outputs into a common node that feeds a final voltage regulation stage. This approach naturally accommodates sources with different power levels and avoids circulating currents between sources. The final regulator determines the system output voltage.

Direct Charge Combining

For battery or supercapacitor charging applications, sources can directly contribute charge to the storage element with minimal intermediate regulation. Each source converter implements current limiting and end-of-charge detection. This approach maximizes efficiency by avoiding unnecessary conversion stages but provides less output voltage regulation.

Time-Multiplexed Combining

Where board area or cost forbids one converter per source, a single shared power stage serves the sources in sequence, and combining happens in the output capacitor or storage element rather than in a summing node. Each source is granted the converter for a defined interval, during which its own operating point is tracked, and the controller retains that operating point estimate for the next visit. Allocation may be fixed round-robin or weighted toward whichever source is currently delivering the most power. The trade-off is bandwidth: a source visited only intermittently cannot be tracked through fast changes, so time multiplexing suits slowly varying sources such as thermoelectric generators better than impulsive kinetic sources.

Source Prioritization

Power routing algorithms assign priorities to available sources based on instantaneous power levels, efficiency, or other criteria. Higher-priority sources receive first allocation of conversion resources, with lower-priority sources engaged as capacity permits or as primary sources become unavailable. Priorities may be fixed or adjusted dynamically according to operating conditions.

Prioritization is often implemented in hardware. The Analog Devices LTC3331, for instance, combines a harvesting input path with a battery-powered buck-boost path behind an input prioritizer: whenever harvested energy is sufficient, the harvesting path supplies the load and the battery path stands down, drawing only about 200 nanoamperes from the cell. Fixed hardware arbitration of this kind costs almost nothing to run and cannot be confused by a firmware fault, which is why safety-relevant designs frequently prefer it to a software policy.

Load Matching

Intelligent routing can match source characteristics to load requirements. Steady loads may be served by stable sources such as thermoelectric generators, while intermittent high-power demands draw from storage that has been charged by variable sources. This approach optimizes both energy capture and delivery efficiency.

Storage Management Integration

Power routing decisions integrate with storage management to maintain appropriate state of charge while meeting load demands. Excess harvested energy charges storage, while storage supplements harvesting during high-demand periods. Coordinated control prevents storage overcharge or deep discharge while maximizing energy throughput.

Predictive Energy Management

Learning algorithms can predict future source availability based on time-of-day, historical patterns, or correlated environmental measurements. Predictive knowledge enables proactive routing decisions, such as charging storage before an anticipated low-energy period or deferring non-critical loads until energy availability improves.

Minimum Operating Voltage Challenges

Power management circuits require minimum supply voltage to begin operation, typically several hundred millivolts for CMOS control logic. Starting from a completely discharged state with only microwatts of available power presents a bootstrapping challenge. The system must accumulate sufficient energy to start the power management electronics before it can efficiently harvest additional energy.

Ultra-Low-Voltage Startup Circuits

Specialized startup circuits can operate from sub-100-millivolt sources using techniques such as mechanical switches, depletion-mode transistors, or transformer-coupled oscillators. A common commercial approach pairs a step-up transformer with a self-oscillating depletion-mode switch to form a resonant boost stage; the Analog Devices LTC3108 is the familiar example, starting from inputs as low as about 20 millivolts and suiting small thermoelectric generators whose output falls to a few tens of millivolts at low temperature differentials. Once started, such circuits raise the source voltage enough to power conventional CMOS circuitry, then hand off to the main power management system for efficient operation.

Fully integrated harvester ICs avoid the transformer at the cost of a higher starting threshold. The BQ25570 requires roughly 330 millivolts to bring its cold-start oscillator up and charge the storage node, after which the main boost converter continues to harvest from inputs near 100 millivolts. The e-peas AEM13920 specifies a cold-start condition of about 275 millivolts and 5 microwatts. The distinction that matters in practice is between the one-time starting threshold and the far lower running threshold, and a design that ignores it can commission successfully on the bench yet fail to restart in the field after a deep discharge.

Multi-Source Cold Start Strategy

In hybrid systems, the source with the best cold-start capability may serve as the primary startup source. Once the power management system is running, it can engage additional sources with higher power but more demanding startup requirements. Sequencing logic ensures orderly initialization without overloading the nascent power supply.

Auxiliary Power Elements

Small primary cells or charged supercapacitors can provide guaranteed cold-start capability when harvested energy may be insufficient. These auxiliary elements provide the initial energy to start the power management system, which then maintains itself from harvested energy under normal conditions. The auxiliary element requires replacement or recharge only after extended periods without sufficient harvested energy.

Digital versus Analog Control

Simple hybrid systems may use analog control for low overhead, while complex systems benefit from digital control's flexibility. Mixed-signal approaches use analog circuits for fast inner loops and digital processing for slower optimization functions. The control approach must balance performance requirements against the power budget allocated to control functions.

Control Processor Selection

Ultra-low-power microcontrollers designed for energy harvesting applications provide the processing capability for sophisticated power management algorithms while consuming only microwatts during active operation and nanowatts in sleep modes. Appropriate processor selection matches control requirements to available power budget.

Real-Time Constraints

Power management control loops must respond fast enough to track changing source conditions and maintain output regulation. Inner current and voltage loops typically require microsecond response, while MPPT and source selection can operate on millisecond timescales. Hierarchical control structures separate fast and slow functions appropriately.

Stability Analysis

Multi-input converters present complex stability challenges due to interactions between sources and control loops. Small-signal modeling and analysis techniques adapted for multi-source systems help ensure stable operation across the range of operating conditions. Worst-case analysis considers stability with various combinations of active and inactive sources.

Input Overvoltage Protection

Harvesting sources that appear feeble on average can produce damaging peaks. An unloaded piezoelectric element struck by a mechanical shock may swing to tens or hundreds of volts, and a photovoltaic string reaches its highest voltage precisely when it is open-circuited in full illumination. Input clamps, shunt regulators, and series blocking devices bound these excursions before they reach the converter's switches. In a hybrid system the clamp must be sized for the harshest source on the board, not for the nominal operating range of the aggregate.

Reverse Current and Source Isolation

When several sources share a bus or a storage element, one source can sink current from another, converting harvested energy into heat in a transducer. Ideal-diode controllers, back-to-back MOSFET switches, and controlled disconnect logic isolate each path. Simple Schottky blocking diodes accomplish the same task but cost several hundred millivolts of forward drop, which is unacceptable on a low-voltage thermoelectric input and merely wasteful on a higher-voltage one.

Storage Protection and Recovery

Storage elements require their own limits: overvoltage cutoff during abundant harvest, undervoltage lockout to prevent deep discharge, and temperature-dependent charge restriction for lithium chemistries. Undervoltage lockout interacts directly with cold start, because a system that disconnects its load at too low a threshold may lack the energy to restart the power management circuitry. Setting the disconnect threshold above the cold-start requirement, with margin for leakage during the dormant period, is what allows an unattended node to recover on its own after an energy drought.

Light-Load Efficiency

Energy harvesting systems often operate at light loads relative to converter ratings, where switching and quiescent losses dominate. Pulse-frequency modulation, burst-mode operation, and dynamic voltage scaling help maintain efficiency at light loads. Some systems include separate high-efficiency paths for light-load operation.

Conversion Loss Minimization

Each power conversion stage introduces losses that reduce net harvested energy. Minimizing the number of conversion stages, using high-efficiency topologies, and selecting appropriate components for the power levels involved all contribute to maximizing net output. Direct paths from sources to loads or storage, bypassing conversion stages when voltage matching permits, can significantly improve efficiency.

Quiescent Current Reduction

The power consumed by control circuits, reference generators, and bias currents must be minimized relative to harvested power. Specialized regulators and power management ICs designed for energy harvesting achieve sub-microampere quiescent consumption: the LTC3331, for example, draws roughly 950 nanoamperes at no load when running from its battery path, which at a few volts amounts to single-digit microwatts. Aggressive power gating of unused circuit blocks reduces overhead further, and in hybrid systems the interface for a source that is currently unavailable should be gated off entirely rather than left biased.

Quiescent current also sets a floor on which sources are worth interfacing at all. If a harvester's average output does not comfortably exceed the overhead of the stage that conditions it, adding that source lowers net system energy. Evaluating each candidate source against the marginal overhead of its interface is a routine and often decisive step in hybrid system design.