A few connected pieces of a Bluetooth LE battery budget: an activity card (one measured activity, reduced to the numbers that transfer), a separate sleep baseline for the chip, a coin-cell model, and a calculator that composes them into a product-specific battery-life band, plus a brown-out check the energy-only view cannot give you.
Honesty note: only non-connectable advertising is bench-measured here (nRF54L15 on a Joulescope JS320, 19.15 µJ per event). Scannable and connectable advertising are shown as "not yet characterized" rather than guessed, and any composed number is labeled measured or estimate.
Think of an activity as the software side ("I advertise"), and the hardware as what grounds it (the chip, its RAM, its power source). An activity card describes one such activity, at one stated operating point, stripped to essentials. It declares the activity's defining spec up front (connectable or not, scannable or not, channels, PHY, TX power), and its face is the number that transfers to any battery budget: the energy the chip spends each time the activity happens. Underneath it carries the current shape a battery has to deliver: the transmit-burst current and the turn-on peak.
What is not on the card is as important as what is: no sleep floor, no battery, no interval. Those moved out, so the same card drops onto any chip's sleep baseline, any cell, and any run-rate without carrying stale assumptions. In the library below, the sleep baseline and the battery each get the same kind of label, chosen from a dropdown, so the whole system reads in one consistent format.
A real product is a chip, a battery, and a handful of activities each running at its own rate, and the calculator composes exactly that: pick a chip (it brings its own intrinsics, sleep baselines, and measured activity cards), set the sleep baseline, pick a battery, and build the activity mix. It returns a battery-life band and a brown-out / deliverability flag that asks whether the worst burst sags the cell past the chip's brown-out threshold near end-of-life. That deliverability check is the honesty layer an energy-only score cannot express, and it is where a constrained cell (a small coin cell, or a printed flexible cell) fails in the field even when the average current looks fine. Have a PowerScope capture of your own firmware? Import it below to get your own numbers.
| Activity | Rate | Per | µA |
|---|
None of this replaces BlueJoule. It layers on top of it:
For "which chip is most efficient," the EM•erald leaderboard remains the right tool. This adds the layer that answers "how long will my product last, and will it still deliver at end-of-life."