Watt-hours to Joules for IoT Battery Energy Budgets

Budgeting a Coin Cell in Microjoules per Operation

A battery-powered sensor node lives or dies on a single sum: how much energy the cell contains, divided by how much each cycle of firmware spends. The cell arrives described in milliampere-hours, which is a charge count and useless for adding things up across different supply voltages. Multiply it by the nominal voltage and it becomes watt-hours; multiply again by 3 600 and it becomes joules, the unit in which a radio transmission, an ADC reading and a year of sleep can finally be summed on one line. That is why an energy budget spreadsheet has a joule column and never a milliampere-hour one.

Conversion factor: 1 Wh = 3 600 J exactly, since a joule is a watt-second and an hour holds 3 600 of them. A CR2032 rated 225 mAh at 3 V therefore stores 0.225 × 3 = 0.675 Wh, or 2 430 J. At 45 µJ per advertising event that is 54 million adverts on paper — before any of the derating below.

What Actually Drains the Cell

Sleep Sets the Floor, Radio Sets the Slope

A microcontroller idling at 2 µA on 3 V burns 6 µW, which is 0.518 J a day and 17.5 mAh over a year. That floor is fixed. Everything above it is the radio, and how often you choose to use it is the one number the product manager can still negotiate.

A Coin Cell Cannot Deliver a Pulse

Internal resistance on a fresh CR2032 is around 10 Ω and climbs steeply as it ages. Draw 15 mA for a transmit burst and the terminal voltage sags far enough to trip a brown-out detector, so the practical answer is a bulk capacitor across the cell rather than a bigger battery.

Self-Discharge Is a Fixed Tax

Lithium manganese dioxide coin cells lose roughly 1 % of capacity a year, lithium thionyl chloride cells rather less. Over a ten-year design that quietly removes about a tenth of the joules before the firmware has spent a single one, and no amount of duty-cycling recovers it.

Duty Cycle Is the Only Real Lever

Halving the transmit interval doubles radio energy; dropping a LoRa spreading factor from SF12 to SF7 cuts the same message by a factor of twenty-eight. Firmware timing changes energy by orders of magnitude, while a change of cell chemistry rarely buys more than a factor of ten.

Working a Datasheet Current Down to Joules per Event

Datasheets talk in microamps and milliseconds, marketing talks in years, and the bridge between them is the joule. The workflow below is the one an energy budget is normally built with, and the converter handles the step where the cell rating has to meet the per-event figures.

Turn each current figure into an energy per event

Multiply the current by the supply voltage to get watts, then by the duration in seconds. A 5 mA wake lasting 10 ms on a 3 V rail is 0.005 × 3 × 0.01 = 150 µJ. Do it once per state the firmware enters.

Put the cell's watt-hour rating into the left box

Enter 0.675 for a CR2032, or whatever mAh × V gives for the cell you have chosen, and the joule total appears beside it as you type. Decimal commas are accepted, so a value pasted from a European datasheet needs no editing first.

Reverse it to price a budget back in watt-hours

Once the yearly total is known in joules, the swap arrows put joules on the left and give the watt-hours a cell selection table is indexed by. Both boxes stay live either way, so typing into the right-hand one produces the same result without touching the arrows.

Take the plain figure into the budget sheet

The copy control above a field yields the number with no unit attached, and Ctrl+C inside the field behaves identically — which is what a spreadsheet cell or a firmware constant wants, with nothing to strip before it will parse.

Nameplate joules are not deliverable joules: the rated capacity is measured at a gentle continuous drain down to a cutoff a node would never reach. Under pulsed loads, at low temperature, and with a decade of self-discharge in the way, planning on 60 to 70 % of the printed figure is normal engineering practice. Measure the finished board with a current profiler before anybody prints a battery-life claim.

Energy per Operation for Common IoT Radio and Sensor Tasks

Figures below assume a 3 V rail for the sleep and sensor rows and a 3.3 V rail at 40 mA for the LoRa transmit rows, giving 132 mW of radio draw. LoRa airtimes are computed for a 12-byte payload on 125 kHz with coding rate 4/5. The last column divides the 2 430 J of a nominal CR2032 by the event cost, before derating.

Operation	Energy	In joules	In watt-hours	Count on one CR2032
One second of 2 µA deep sleep	6 µJ	0.000006	1.67 × 10⁻⁹	405 000 000
BLE connection event, empty	20 µJ	0.00002	5.56 × 10⁻⁹	121 500 000
BLE advertisement, three channels	45 µJ	0.000045	1.25 × 10⁻⁸	54 000 000
MCU wake plus sensor read	150 µJ	0.00015	4.17 × 10⁻⁸	16 200 000
A full day of 2 µA sleep	518 mJ	0.5184	1.44 × 10⁻⁴	4 687 days
LoRa uplink, SF7 (41.2 ms)	5.44 mJ	0.00544	1.51 × 10⁻⁶	446 600
LoRa uplink, SF12 (1.155 s)	152 mJ	0.1525	4.24 × 10⁻⁵	15 900
NB-IoT message with connection setup	1.5 J	1.5	4.17 × 10⁻⁴	1 620

The spread across that table is nearly six orders of magnitude, and it explains most of the architecture decisions in a low-power product. A node that advertises over Bluetooth every second spends 45 µJ a time and could in theory run for over a year on a coin cell; the same node sending one NB-IoT message an hour spends 1.5 J a time and empties the identical cell in about ten weeks. The SF7 and SF12 rows are the same message on the same radio at the same output power — the twenty-eight-fold difference is nothing but airtime.

What the Converter Contributes to the Budget

Nothing Rounds to Zero at the Microjoule Scale

Results carry up to eight decimals and drop into exponent form below a millionth, so a 1.25 × 10⁻⁸ Wh advertising event stays a real number instead of collapsing to a row of zeros.

A Cell Budget and an Event Cost, Same Two Boxes

Cell ratings arrive in watt-hours and firmware costs arrive in joules, and both sides accept typing, so a 0.675 Wh coin cell and a 0.0054 J uplink can be compared without a spreadsheet in between.

Plain Digits for the Energy-Budget Spreadsheet

Copying returns the value on its own, so a converted joule total drops straight into a budget column or a firmware constant without a unit label that would have to be deleted afterwards.

The Unit Menu Reaches kWh for Rechargeable Nodes

Type into either dropdown's search box to reach the kilojoule, the kilowatt-hour or the electronvolt, which covers the energy-harvesting and rechargeable end of a fleet as well as the primary-cell end.

Battery-Life Questions From the Firmware Bench

How do I get from microjoules per event to years on one coin cell?

Build a daily total, then divide. Take every state the firmware enters in twenty-four hours, multiply each event cost by how many times it happens, and add the sleep floor. A node that advertises every two seconds spends 43 200 × 45 µJ = 1.944 J a day on radio plus 0.518 J on sleep, so 2.46 J daily. Against 2 430 J that is 988 days, or about 2.7 years — and against a realistic 65 % deliverable fraction, closer to 1.8 years. Stretching the interval to ten seconds drops radio to 0.389 J and pushes the same cell past four years.

Why can I not use all 225 mAh of a CR2032?

Because that number was measured under conditions your node does not reproduce. The rating comes from a continuous drain of a few hundred microamps at room temperature, run down to about 2.0 V. A real node pulls milliamp pulses through an internal resistance that starts near 10 Ω and can reach several hundred ohms late in life, and each pulse pulls the terminal voltage down by I × R. Once the sag crosses the brown-out threshold the node resets while plenty of chemical capacity remains. Add a 100 µF or larger bulk capacitor to supply the burst, and design the cutoff generously.

The datasheet gives microamps — how do I turn that into joules per day?

Multiply amps by volts by seconds. A 2 µA sleep current on a 3 V rail is 6 µW, and 86 400 seconds later that is 0.5184 J — put 0.000144 in the watt-hour box and the same figure appears. Over a year it becomes 189 J, which is 7.8 % of a CR2032 spent doing nothing at all, or 17.5 mAh if you prefer to work in charge. Take care with mixed rails: a sensor on 1.8 V and a radio on 3.3 V give different joules for the same current, which is precisely why energy rather than current is the quantity worth tracking.

Why budget energy per message instead of current draw?

Because current alone hides the two variables that matter — how long the draw lasts and what voltage it happens at. A radio pulling 40 mA sounds alarming until you notice the SF7 burst lasts 41.2 ms and costs 5.44 mJ, while a 12 mA sensor left powered for two seconds quietly costs 72 mJ, thirteen times more. Energy per message also composes: uplink, downlink windows, sensor warm-up and processing simply add, and the sum divides straight into the cell's joules. Current figures cannot be added like that at all.

What does self-discharge remove from a ten-year budget?

Roughly a tenth, and more if the device runs warm. At about 1 % a year a lithium manganese dioxide coin cell has surrendered near 10 % of its 2 430 J before the firmware touches it, and self-discharge accelerates markedly with temperature, so a node on a sunlit roof does considerably worse than one in a corridor. Lithium thionyl chloride primaries are the usual answer for genuinely long deployments — well under 1 % a year, and a C-size cell holds 30.6 Wh or 110 160 J, some forty-five times a coin cell — at the cost of a passivation layer that has to be depassivated before the first real load.

Watt-hours to Joules