Battery life is designed in. It is not discovered later.

Where stated battery life comes from

Often from a spreadsheet built during early design, using datasheet sleep currents and an assumed duty cycle. Real devices then achieve a fraction of it, and the gap is discovered after units are in the field, where replacing batteries is the most expensive thing the product does.

The difference is usually not one mistake. It is an accumulation: a peripheral left enabled, a regulator with higher quiescent draw than assumed, a radio retrying more than expected, a pull-up resistor quietly conducting, and a battery losing capacity in cold weather that nobody modelled.

What the energy budget has to include

• Sleep current. Usually the dominant term over a long life, and the one most often wrong in practice because the whole board matters, not just the processor.
• Active bursts. What the device does when awake, how long it takes and how often it happens.
• Radio energy per transmission. Including retries, reconnection and negotiation, which are frequently omitted from early estimates.
• Peak current. Transmission peaks can exceed what a small or cold cell can deliver without its voltage collapsing.
• Quiescent losses. Regulators, dividers, leakage and anything left powered that nobody accounted for.
• Battery behaviour. Self-discharge, capacity loss with age, and the substantial effect of temperature, especially below freezing.

How we help

• Energy budget from the behaviour. Built from what the product must actually do, with the assumptions written down so they can be challenged.
• Power architecture. Regulator topology and selection, power domains and switching, so parts that are not needed are genuinely off rather than idle.
• Low-power firmware. Sleep states, wake sources, peripheral and clock management, and scheduling that keeps awake time short. See embedded firmware development.
• Battery selection. Chemistry and format chosen against runtime, peak current, temperature range, shelf life and transport constraints.
• Charging and management. Where rechargeable, charge control, protection, state-of-charge estimation and safe behaviour at temperature extremes.
• Energy harvesting. Solar or other sources where the environment genuinely supports it, with realistic assessment of the worst case rather than the average.
• Measurement. Current profiling on the real hardware across the full operating cycle, because measurement is what turns an estimate into a claim you can stand behind.

Measure, then state

We would rather tell you a shorter number we have measured than a longer one derived from a datasheet. A runtime figure is worth having only if it survives the real duty cycle, the real radio behaviour and the coldest temperature the device will see. Where a target is not achievable, it is better to know while the design can still change, whether by reducing reporting frequency, moving a decision onto the device so it transmits less, or accepting a larger cell.

Safety and transport

Lithium chemistries carry handling, shipping and protection requirements. These affect cell choice, protection circuitry, enclosure design and how the product can be shipped. Raise them at scoping, because they constrain the design rather than decorate it, and the applicable rules depend on your product and your markets.

What you receive

A documented energy budget with its assumptions, the power architecture and its design sources, low-power firmware, and measured current profiles with the runtime those measurements support under stated conditions.

Related

Wireless connectivity selection, where most of the energy goes. Edge AI on microcontrollers, where local decisions reduce transmission. Hardware and PCB design.