EC-20.6 Low Power Design, and Measuring It

What this is and why it exists

Battery life is arithmetic, and the arithmetic is usually dominated by one thing nobody measured.

A radio transmission of two hundred milliamperes for eight milliseconds costs more energy than an hour of sleeping at two microamperes. So the design question is almost never how fast the processor runs. It is how often the expensive thing happens, and how deeply the device sleeps between.

This topic builds the budget on paper first, then makes it real with an instrument that can see both ends of the range at once. The two halves are equally necessary: a budget nobody measured is a guess, and a measurement without a budget has nothing to compare against.

The vocabulary

• Dynamic power — energy spent charging and discharging capacitance when signals change.
• Leakage — current flowing whenever the part is powered, whether or not anything changes.
• Sleep mode — a state in which parts of the device are powered down.
• Wake source — the event permitted to bring the device out of sleep.
• Duty cycle — the fraction of time spent in each state.
• Energy budget — the average current computed from those states and fractions.
• Race to sleep — running at full speed to finish sooner and sleep longer.
• Self-discharge — the current a cell loses internally with nothing connected.

The mental model

Energy goes to three places. Dynamic power depends on how often signals change and on the square of the supply voltage. Leakage flows whenever the part is powered at all, and grows sharply with temperature. The radio dwarfs both while it is transmitting. Knowing which of the three dominates decides which optimisation is worth doing at all.

Sleep modes are a ladder. Deeper states save more and take longer to leave, and the wake-up itself costs energy and time. That leads to a result worth deriving once. A device that wakes very often can use more energy than one staying awake, because it pays the transition cost repeatedly.

The duty cycle is where the arithmetic happens. Multiply each state's current by the fraction of time spent in it, add them up, and divide the cell's capacity by the result. That gives a battery life directly, and doing it before the board exists is what makes a cell choice defensible.

Voltage and frequency scaling is often misunderstood. Running slower at a lower voltage reduces energy per operation, but takes longer, so leakage is paid for longer. When the amount of work is fixed, racing to sleep frequently wins. Which strategy is better depends on the leakage of the specific part.

Then measurement. The interesting behaviour spans five orders of magnitude and lasts milliseconds, which defeats an ordinary meter. A shunt with a wide dynamic range instrument, or a dedicated current analyser, shows both. The sleep current and the radio burst appear in one trace. Without that, the budget cannot be checked.

Finally, the cell. A rated capacity assumes a temperature and a discharge rate that your product may not have. For a device that sleeps for years, self-discharge alone can dominate everything else in the budget. The cell's own curves, not its headline number, are what the budget has to be checked against.

What you should now be able to explain or do

• Say where the energy goes in a battery-powered device, and which term dominates.
• Build an energy budget from a duty cycle and convert it into a battery life.
• Choose a sleep mode and a wake source for a stated behaviour.
• Explain when racing to sleep beats slowing down, and why.
• Measure current from microamperes to hundreds of milliamperes in one capture.
• Check a budget against a cell's temperature and self-discharge behaviour.

Check yourself