Field notes

methods ·

Your battery estimate forgot the board

Build an operating-state budget and keep nominal battery arithmetic separate from measured runtime.

On this page

A sensor’s low-power headline can be perfectly accurate while the finished device consumes much more. The regulator, indicator LED, microcontroller, radio, and your decision to sample continuously were not included in the sensor’s publicity department.

Start by identifying the boundary of each number. Bosch lists typical BME280 consumption for specific measurement modes, including 3.6 µA at a 1 Hz humidity/pressure/temperature refresh rate. That is a chip-level operating condition, not the current of every assembled breakout or logger. Bosch Sensortec

Budget operating states

List states such as sleep, sample, process, and transmit. For each, record duration and current at a named rail. A simple weighted average is:

Average current = sum of each state current × its fraction of the cycle.

For an illustrative same-rail load drawing 20 mA for one second and 0.1 mA for nine seconds, the average is:

(20 × 1 + 0.1 × 9) / 10 = 2.09 mA

This is synthetic arithmetic, not a measurement of a particular board. The fractions must sum to one, and concurrent loads must be included. A radio transmitting while the processor works is not necessarily a separate, non-overlapping state.

Keep the rails straight

Adding milliamps from different voltages is not a complete battery budget. Convert to power at each rail, account for conversion efficiency where applicable, and include quiescent loads according to their actual power path.

For an illustrative 3.3 V load averaging 2.09 mA, output power is about 6.90 mW. With a hypothetical 90% efficient converter fed at 3.7 V, the corresponding input current is about 2.07 mA before additional losses and quiescent loads. The efficiency and input voltage here are assumptions chosen to demonstrate the calculation.

A nominal battery capacity divided by that current gives an idealized planning number. Real runtime depends on usable capacity, discharge conditions, cutoff behavior, temperature, and the actual changing load. A calculator should expose these assumptions instead of printing “lasts 20 days” as a fact.

Measure the boundary you care about

If the question is battery runtime, measure the device at its battery input with an appropriate instrument and setup. A measurement at the sensor rail answers a narrower question. Record supply voltage, instrument, sampling behavior, operating sequence, and uncertainty or known limitations.

Averaging slowly can miss brief peaks that matter to the power design. Conversely, sizing everything from a peak alone can misrepresent energy use. Keep peak-current requirements and average-energy estimates as separate outputs.

An INA219 can report electrical quantities through a documented interface, but the chip and breakout must fit the measurement task. TI distinguishes the monitored bus voltage from the device supply and documents calibration behavior. Do not choose a shunt or attach a board based only on a generic module name. Texas Instruments

Show uncertainty without giving up

Use a range for uncertain inputs and display the resulting range. Mark which values are measured, manufacturer-typical, manufacturer-maximum, or assumed. Unknown board current is not zero. Typical sleep current is not a guaranteed worst case.

After a real duty-cycle test, compare measured consumption with the budget and investigate the largest unexplained term. That is usually more productive than adjusting the last decimal place of the smallest one.

The reward is a budget that can improve with the project. The battery does not care how elegant the estimate looked; it participates by going flat.

Next: 3.3 volts is not a personality trait explains the boundaries behind the specifications.

Sources

Put this to work

More field notes