Battery life you can measure, not just estimate
Battery life is decided by the hours a device spends asleep and the milliseconds it spends awake. We measure both, write the budget down and keep later releases from eroding it.
Key characteristics
- Measured
- Sleep current, wake-ups, radio events, updates
- You provide
- Hardware, usage profile and battery choice
- You receive
- Current profile, power budget, firmware changes
- Kept honest by
- A sleep-current test on every release
A battery-powered device spends most of its life doing nothing, and how little current it draws while doing nothing sets the floor for battery life. The rest is decided by short bursts of activity: a sensor reading, a radio event, a retry when the gateway does not answer. Low-power design is the work of making each burst as short and as rare as the product allows, keeping the device in its deepest useful sleep the rest of the time, and proving the result with measurements rather than datasheet numbers.
Where the energy goes
The battery life calculator makes this concrete. Its default example is a 3000 mAh battery with 85% of that capacity usable, a 4 µA sleep current, 12 sensor wake-ups an hour lasting 150 ms each at 3.2 mA, 3,600 radio events an hour lasting 2.5 ms each at 6 mA, and battery self-discharge of 1% a year. That works out to an average current of 24.0 µA and an estimated battery life of 106,194 hours: 4,425 days, or about 12.1 years. The split is the interesting part.
| Part of the average current | Share in the default example |
|---|---|
| Sleep | 16.6% |
| Sensor wake-ups | 6.7% |
| Radio events | 62.5% |
| Battery self-discharge | 14.3% |
In that example the radio is nearly two thirds of the budget, even though each event lasts only 2.5 ms, because it happens once a second. Halving the sleep current would cut the average by less than a tenth; halving the number of radio events would cut it by almost a third. That is why we look at the schedule before we look at the sleep current. The calculator’s own note makes the same point we would: real cells lose capacity in the cold and at high pulse currents, so measure on hardware before you promise a number.
How we measure
Estimates start the budget; measurements finish it. We measure current on your boards in every state the device uses: deep sleep, sensor reading, advertising, connected, transmitting, and downloading an update. We record the measurement setup alongside the numbers, so anyone on your team can repeat them on the next board revision.
Common findings are ordinary and expensive: a peripheral left powered, a floating input pin, debug or log output left enabled, a timer waking the processor more often than anyone intended, a sensor sampled faster than the product needs. Each one is small. Together they can turn years into months.
Sleep states and wake strategy
Microcontrollers offer several sleep depths that trade current against wake-up time and retained memory. On Zephyr, the power management subsystem covers both system power management (the processor and SoC) and device power management (individual peripherals), so the application states what it needs and the kernel picks the deepest state that allows it. On FreeRTOS or bare metal we build the equivalent explicitly and document it.
The wake strategy matters as much as the sleep state. Batch readings and send them together. Let the phone or gateway set connection parameters that match the real data rate. Do not wake the radio just to report that nothing changed. Make sure a failed connection backs off instead of retrying every few seconds until the battery is empty.
Radio duty cycles
Advertising interval, connection interval, readings per message, how long a cellular modem stays attached: each one trades latency against battery, and each belongs to a product decision, not only an engineering one. We put the options in front of you with their cost in battery life. Faster links help where both ends support them. Bluetooth 5’s LE 2M PHY sends data at twice the symbol rate of LE 1M, so the same payload spends less time on air, but support for it is optional in Bluetooth stacks and the other end has to agree. For cellular products we measure the modem’s sleep behavior on the network the device will actually use, because what the network grants can differ from what the firmware asks for. Radio choices are covered in more depth in connectivity.
Updates cost energy too
A firmware download can be the most expensive thing a small device does all year. The update policy checks battery level before a download starts and defers it if the battery is low, and the BLE update time calculator shows how payload size and connection interval change how long the radio stays on. After every release we re-measure sleep current, because one forgotten peripheral can undo a year of battery life. That check runs in the hardware-in-the-loop rig described in test automation, and a release that raises sleep current beyond the agreed limit does not ship until someone has explained why.
Example (fictional). Loamwick’s requirement FW-002 is that the LW-1 probe runs 3 years on two AA lithium cells. It is verified the way we would verify any battery claim: measured average current on 5 boards, fed into the battery life calculator. The probe wakes every 5 minutes to read soil moisture and temperature and reports to its gateway over Bluetooth LE, so the radio schedule is the first place to look. Release 1.1.0 adds battery voltage to every report, so the field trial shows the real discharge curve instead of an estimate. See the example project.
What you receive
- A measured current profile for each device state, with the measurement setup recorded.
- A power budget with its assumptions shown, in a form you can update yourself.
- Firmware changes, each with the measured effect it had.
- A sleep-current regression test that runs on every release candidate.
Where to start
If you have boards and a battery target, start with a firmware review: we read the schematic, the firmware’s sleep and radio behavior and your usage profile, and tell you in writing where the energy is likely going and what to measure first. If you are still choosing a battery, run your own numbers in the battery life calculator. Then request a firmware review, or read more in the connectivity and power section of Field Update.
Send us your board and your plan.
A firmware review reads your schematics, code, flash layout and update plan, and ends with a written plan: what to build first, what could go wrong in the field and how updates will reach your devices.
Or email sales@fumbar.com, or call (720) 675-4202.