Microwatts to Watts

Budgeting a Battery-Free Sensor Node in Microwatts

A harvested node lives on whatever the room gives it. An indoor cell, a thermoelectric module clamped to a warm pipe, a piezo beam bolted to a pump housing — all of them produce power in the microwatt range. Everything you have to reconcile that against is written in watts: the power-management chip's quiescent draw, the sensor's active figure, the radio's burst. Turning μW into W is what puts the harvest and the load on the same line of the same budget sheet.

Conversion factor: 1 μW = 0.000001 W, so the decimal point moves six places left. A 50 cm² indoor cell producing 100 μW under office lighting is 0.0001 W; set beside a node averaging 0.00002 W, that says the design has roughly five times the margin it needs.

Where the Microwatts Come From

Indoor light is the dependable source

Office lighting at 200–500 lux gives a cell tuned for indoor spectra only a few μW per cm², but it arrives on a schedule and does not depend on the machinery running.

Heat only pays while a gradient lasts

A thermoelectric module produces nothing without a temperature difference across it. The cold side is the hard part: once the heat sink warms up, the ΔT collapses and so does the output.

Vibration has to match the mount

A piezo harvester is tuned to a frequency. Bolted to a pump turning at its design speed it gives hundreds of μW; a metre away on the same frame it can give almost nothing.

Ambient fields are a trickle

Power scavenged from fields already in the air lands in single μW at any realistic distance — enough to keep a buffer topped up between events, never enough to run a node outright.

From Harvester Datasheet to Node Power Budget

The numbers arrive in mixed units — μW from the harvester, microamps and volts from the microcontroller, milliwatts from the radio. Normalise everything to watts first, then compare.

Enter the harvester's rated output

Type the μW figure on the left — 5, 250, 100, whatever the module is rated at under its stated conditions. The watt value appears as you type, and a comma works as a decimal separator if that is how your notes are written.

Set it against the node's average, not its peak

A node that sleeps at 3 μW and wakes for 50 ms every quarter hour might average 20 μW — that is 0.00002 W, and that is the figure the harvest has to beat, with margin for the worst week of the year.

Work back from the load with the swap button

If you already know the node needs 0.00015 W, press ↔ to run W → μW and read 150 μW: the minimum a harvester has to deliver on its worst day. By hand, multiply the watt figure by 1,000,000.

Paste the bare number into the budget sheet

The copy button hands over the plain value with no unit and no spaces, which is what a spreadsheet cell or a simulation parameter expects. Ctrl + C inside a field does the same thing.

Rated output is a best case: harvester datasheets quote one specific illuminance, ΔT, acceleration or distance. Derate for the darkest, coldest, quietest hour the installation will ever see — that is the hour the energy buffer has to carry alone.

What Ambient Sources Deliver Against a Node's Sleep Draw

Representative harvest figures under the conditions each source is normally quoted at, with a microcontroller's deep-sleep draw and a whole node's average at the bottom for scale.

Source	Stated condition	Output (μW)	Output (W)
Indoor PV cell	50 cm², 500 lux office lighting	100 μW	0.0001 W
Outdoor mini PV cell	10 cm², direct daylight	10 000 μW	0.01 W
Thermoelectric generator	4 cm² module, ΔT of 10 °C on a warm pipe	500 μW	0.0005 W
Piezo vibration harvester	Machine housing near 0.5 g, on resonance	250 μW	0.00025 W
Ambient field harvesting	A few metres from a strong source	5 μW	0.000005 W
Microcontroller in deep sleep	1 μA with the real-time clock running, at 3 V	3 μW	0.000003 W
Whole node average	One measurement and report every 15 minutes	20 μW	0.00002 W

Read the last two rows first. The node's average sits above the sleeping microcontroller but three orders of magnitude below a small cell in daylight, which is why almost every workable battery-free design is either outdoors, or indoors with a generous cell area and a very long reporting interval.

How the Converter Fits an Energy-Harvesting Workflow

Sweep a harvest budget without clearing fields

Both boxes stay editable and track each other, so you can walk 5, 20, 100 and 500 μW through in one pass while the watt column keeps up.

Size the harvester from the load, or the load from the harvester

The swap button turns the page around so a watt-denominated load target becomes the μW specification you take to a supplier.

Reach mW when the radio joins the sum

Searchable dropdowns on both sides carry every power unit, so a burst figure quoted in milliwatts drops into the same comparison without a second tool.

Sub-microwatt leakage terms stay readable

Results carry up to eight decimals and flip to exponent form below 0.000001 W, so a 0.5 μW leakage figure shows as 5.000000e-7 instead of a row of zeros.

Energy-Harvesting Budget Questions

My node pulls 15 mA while it transmits — how can 100 μW of harvest ever run it?

The harvester never has to supply the peak. It fills a buffer slowly and the buffer covers the burst. A 50 ms wake at 60 mW costs 0.003 J; harvesting 0.0001 W refills that in 30 seconds, so at one report per quarter hour the source is comfortably ahead. Size the harvester against the average and size the buffer against the peak.

Why does deep-sleep current end up deciding whether a harvested design works?

Sleeping is what the node does almost all of the time. At one report an hour, a 3 μW sleep draw costs 0.0108 J over that hour while a 0.003 J burst costs a third as much — the idle state is nearly four fifths of the budget. Halving quiescent current usually buys more than halving transmit time, and that quiescent figure has to include the harvesting front end and any always-on sensor, not just the microcontroller.

How large a supercapacitor do I need between harvests?

Work from usable energy, which is ½C(V_high² − V_low²), not the full ½CV². A 1 F cell taken from 3.3 V down to a 2.2 V cut-off yields about 3 J — a few hundred bursts, or several days of a 3 μW sleep draw. Filling it from 0.0001 W takes roughly eight and a half hours, so the buffer has to cover the whole dark or still period, and its own leakage counts against the node's average.

Is a harvester actually better than simply fitting a coin cell?

Compare average powers. A 225 mAh CR2032 at 3 V holds about 0.675 Wh; spread over ten years that is only 7.7 μW, under 0.000008 W, and that ignores self-discharge. Any source reliably giving tens of μW therefore beats a coin cell over a long deployment. Below that, or where light and vibration come and go unpredictably, a primary cell is simpler and the honest choice.

The harvester is producing power but the circuit never wakes up. What is missing?

You are under the cold-start threshold. A harvesting front end needs a minimum input voltage — commonly a few hundred millivolts from a single thermoelectric module — and a minimum input power, often around 10 μW, before its boost converter can bootstrap itself from a flat buffer. Below that it sits there consuming the little it makes. Series-connect cells for more open-circuit voltage, or pre-charge the buffer once at commissioning so the chip starts in its efficient mode.