3 Energy Sources for IoT Devices

Batteries, Harvesting, Storage, Derating, and Source-Selection Evidence

energy-power

aware

sources

3.1 Start With the Source as a Constraint

An IoT power source is not a perfect fuel tank. A coin cell sags under pulses, a rechargeable pack ages, and a harvester delivers power only when the environment cooperates. The source sets the shape of the design as much as the firmware does.

The practical move is to match the load ledger to usable source behavior: capacity, peak current, derating, recharge pattern, buffer storage, and field validation.

Phoebe’s Field Notes: What Is Actually Inside That 65% Usable Fraction

Phoebe the physics guide

Phoebe’s Why

This chapter’s own ledger treats “usable capacity” as a single derating fraction applied to a coulomb count, and that is the right move for a review record – but it quietly bundles together two physically different effects. Self-discharge is a slow chemical leak: charge the cell never gets to deliver, whether or not the device ever wakes up. Voltage sag under load is a different animal entirely – a real, undamaged battery has an internal resistance, and every amp you pull develops a voltage drop across it, so the terminal voltage a regulator sees is always lower than the cell’s open-circuit voltage while current flows. A cell can be nearly full in coulombs and still trip a low-voltage cutoff during a radio burst, because charge remaining and voltage available are not the same question. Reviewers who read “35% derating” as “the battery leaks 35% on the shelf” are misreading the ledger.

The Derivation

Charge remaining after \(t\) years of shelf self-discharge at annual rate \(r\):

\[Q(t) = Q_0(1-r)^t\]

Terminal voltage under load, from a Thevenin battery model with open-circuit voltage \(V_{oc}\) and internal resistance \(R_{int}\):

\[V_{terminal} = V_{oc} - I\,R_{int}\]

The device is only usable while \(V_{terminal}\) stays above the regulator’s cutoff voltage, independent of how much coulomb capacity remains:

\[V_{terminal} \ge V_{cutoff}\]

Worked Numbers: This Chapter’s Own 1,489 mAh / 65% Example

Self-discharge share: a catalog-typical low-leakage primary lithium cell self-discharges around \(1\%\)/year. Over this chapter’s own 2-year interval: \(Q(2)/Q_0=(0.99)^2=0.980\), a \(2.00\%\) charge loss.
Against this chapter’s own 35-percentage-point derating gap (\(1-0.65\)), self-discharge accounts for only \(2.00/35=5.71\%\) of it – leaving \(33.0\) percentage points, or about \(94\%\) of the derating, to come from cutoff-voltage, temperature, aging, and pulse effects the ledger’s “usable_fraction” line does not break out.
Pulse-sag check with a catalog-typical coin-cell Thevenin model, fresh cell (\(V_{oc}=3.0\) V, \(R_{int}=25\ \Omega\)) under a catalog-typical 15 mA radio-burst pulse: \(V_{terminal}=3.0-0.015\times25=2.625\) V – a comfortable \(0.625\) V above a 2.0 V cutoff.
Same pulse, catalog-typical near-end-of-life cell (\(V_{oc}=2.6\) V, \(R_{int}=200\ \Omega\), an order-of-magnitude rise that is well documented for coin cells near depletion): the linear model gives \(V_{terminal}=2.6-0.015\times200=-0.400\) V. A real cell current-limits rather than going negative, but the sign flip is the diagnostic: this chapter’s own pulse-droop and cutoff-mismatch failure modes can end a battery’s service life while its coulomb counter still shows charge remaining – exactly why the review ledger asks for pulse-load testing at end-of-life voltage, not only a nominal mAh check.

3.2 The Source Is A System

An IoT energy source is more than a battery label or a panel rating. It is the complete path that stores, harvests, converts, protects, and delivers energy to the device while the service promise is being met. The design has to support average load, short peak events, sleep leakage, startup, maintenance delay, and the field environment.

The strongest source decision starts from evidence. A measured load budget tells the team how much energy the device needs. The deployment context tells the team which losses, reserves, and failure modes must be included. Only then does it make sense to choose between a primary cell, rechargeable cell, supercapacitor, harvester plus buffer, wired supply, or hybrid design.

Source selection is a system decision: load evidence, field conditions, and service access shape the source path, then validation proves it can survive startup, peak current, autonomy gaps, and recovery.

For example, a sensor that averages 85 uA for a two-year service interval needs about 0.085 mA x 24 h x 365 x 2 = 1489 mAh before derating. If cutoff voltage, cold operation, self-discharge, pulse support, and reserve leave only 65% usable capacity, the selected source path needs roughly 1489 / 0.65 = 2291 mAh of rated capacity or an equivalent harvested-and-buffered energy plan. The printed label is not enough without the usability assumptions.

Wired sources still need the same review. A mains-powered gateway may not need a multi-year cell, but it may need isolation, surge tolerance, safe shutdown, backup energy for log flushing, and proof that a brownout does not corrupt storage or leave an actuator in an unsafe state.

If you only need the intuition, use this rule: choose the source from measured load and field conditions, not from the largest printed capacity that fits the enclosure.

The Four Source Questions

Load

What average current, peak current, pulse duration, sleep baseline, retry behavior, and maintenance activity must the source support?

Environment

What temperature range, enclosure, vibration, moisture, light access, service access, and storage time change usable energy?

Source Family

Which battery, storage, harvester, regulator, and protection path matches the workload and maintenance model?

Evidence

Which tests prove startup, pulse load, cold behavior, low-energy behavior, autonomy, recharge, and recovery?

Common Source Families

Primary batteries fit long maintenance intervals when the load is low and replacement is planned.
Rechargeable batteries fit devices with a reliable charging path, but they require charge control and end-of-life planning.
Supercapacitors fit frequent charge and discharge cycles or short hold-up intervals, but leakage matters during long idle periods.
Harvesters plus buffers fit sites where ambient energy is measured, variable, and supported by storage for source gaps.
Wired or mains sources fit powered installations, but they may still need backup, isolation, surge handling, and safe shutdown behavior.

Overview Knowledge Check

3.3 Build The Source-Selection Ledger

A source-selection ledger makes the decision reviewable. It separates measured demand from source assumptions, then records the validation gates that must pass before the design can claim a maintenance interval or lifetime target.

For battery-powered nodes, the planning estimate often starts with charge: raw capacity need equals average current multiplied by deployment time. That result is not the selected battery size. The selected source must include usable-capacity derating, reserve, pulse-current support, regulator behavior, and low-energy policy.

Source-Selection Workflow

Measure the load. Capture whole-device current for sleep, wake, sensing, compute, storage, radio transmit, receive, retries, and maintenance states.
Record the deployment. State temperature range, enclosure, installation date, shelf time, replacement access, source availability, and safety constraints.
Estimate raw need. Convert average current and deployment time into raw capacity, or convert state power and time into energy.
Apply source derating. Account for cutoff voltage, temperature, pulse load, aging, self-discharge, converter losses, leakage, and reserve.
Choose the source path. Select the source family, storage buffer, regulator, protection, connector, and low-energy behavior.
Validate the path. Test startup, peak pulses, sleep leakage, cold or hot behavior, autonomy, recharge, recovery, and shutdown behavior.

Review Ledger

Ledger Item

What To Record

Common Failure

Validation Gate

Measured load

Average current, peak current, pulse duration, sleep baseline, retry behavior, startup, maintenance events, and measurement setup.

The estimate uses datasheet currents or omits receive windows, regulator quiescent current, or whole-board leakage.

Repeat the trace with production firmware, final sensors, selected regulator, and realistic radio conditions.

Usable capacity

Rated capacity or stored energy after cutoff voltage, temperature, aging, self-discharge, peak load, storage time, and reserve are considered.

The budget treats a label value as fully usable in every field condition.

Run end-of-life voltage and pulse-load tests under the expected environmental range.

Harvesting and buffer

Measured site energy, conversion path, storage size, autonomy window, load-shedding policy, and recovery after low-harvest intervals.

The harvester is treated as a continuous supply even though the source is intermittent.

Prove the buffer can bridge source gaps and recover without violating the service promise.

Protection and behavior

Charge limits, reverse-current protection, regulator limits, brownout handling, safe shutdown, degraded mode, and maintenance signal.

The device works at nominal voltage but behaves unpredictably near depletion or during recharge faults.

Test cold start, brownout, recharge limits, low-energy mode, and recovery with logs enabled.

Practitioner Knowledge Check

3.4 Why Nominal Energy Disappears

The arithmetic behind source sizing looks simple, but the path from stored or harvested energy to useful device work has many losses and limits. A cell may have enough rated capacity while still failing a radio burst. A harvester may provide enough energy on a good day while failing after a low-harvest interval. A regulator may look efficient during transmit but drain the system during long sleep through quiescent current.

Use charge units when the voltage path is stable enough for current budgeting. Use energy units when voltages, converters, harvesters, and storage elements differ. In both cases, the result is a model that must be tied to measurement and retested when hardware, firmware, radio policy, source chemistry, or the environment changes.

Derating should be explicit rather than hidden in a large safety factor. If a node needs 1577 mAh raw charge for a one-year interval, a 70% usable-capacity assumption raises the source requirement to 1577 / 0.70 = 2253 mAh. Adding a 20% service reserve changes the target again to about 2253 / 0.80 = 2816 mAh. Reviewers should be able to see which step came from chemistry, environment, reserve policy, or measurement uncertainty.

Harvested systems need an energy balance, not only a peak harvester rating. If a load needs 20 mW for a six-hour source gap, the buffer must deliver 20 mW x 6 h = 120 mWh. With 75% storage and converter efficiency, the harvested-side buffer must store about 120 / 0.75 = 160 mWh, plus reserve. At a 3.7 V nominal storage voltage, that is roughly 160 mWh / 3.7 V = 43 mAh before aging and low-temperature effects. The same check should be repeated for startup, radio retries, and recovery after depletion.

Equations To Use Carefully

Raw Capacity

Qraw = Iavg x Tdeployment. Use measured average current and the required service duration.

Rated Capacity

Qrated = Qraw / usable_fraction. State what derating and reserve are included.

Energy Path

Eload = Pload x t. Use energy when converter losses, mixed voltages, or harvesting paths matter.

Harvest Balance

Eharvest usable >= Eload + reserve. Check this across poor-source intervals, not only typical days.

Failure Modes To Surface

Pulse droop: short radio or actuator peaks can collapse voltage even when average current looks low.
Cutoff mismatch: stored energy below the regulator or device cutoff voltage is not usable by the product.
Temperature loss: cold or hot conditions can reduce usable capacity, charging acceptance, and pulse performance.
Idle drains: regulators, sensors, pull-ups, indicators, protection circuits, and supercapacitor leakage can dominate long sleep periods.
Harvest gaps: solar, vibration, thermal, and RF sources can disappear or weaken, so storage and recovery behavior matter.
Unsafe recharge: rechargeable systems need chemistry-appropriate limits, fault handling, and temperature-aware charge policy.

Under-the-Hood Knowledge Check

3.5 Summary

Select IoT energy sources from measured load, deployment conditions, maintenance access, and safety constraints.
Rated capacity is not the same as usable field capacity; cutoff voltage, temperature, pulse load, aging, self-discharge, converter behavior, leakage, and reserve all matter.
Harvesting designs need measured site energy, storage autonomy, load-shedding policy, and recovery tests after source gaps.
Rechargeable and hybrid systems require charge control, protection, low-energy behavior, and end-of-life planning.
A defensible source decision leaves evidence: load trace, assumptions, source path, validation results, and retest triggers.

Key Takeaway

Choose the power source from the whole source system: measured demand, usable capacity, storage, conversion, protection, environment, and validation evidence.