Battery life estimator

The physics

Average current from a duty cycle

Q_cycle = I_sleep·t_sleep + I_wake·t_wake + I_tx·t_tx

I_avg = Q_cycle / T_period + I_quiescent

Charge is current times time, so a duty cycle is just a weighted average. The useful insight is which term dominates: at a 15-minute reporting interval with a few microamps of sleep current, sleep is usually more than half the budget even though the radio is what everyone worries about.

Self-discharge is a parallel load

I_self = C_rated · (rate / 100) / 8766 h

A cell loses charge whether or not you use it. At 3% per year a 2500 mAh cell leaks about 8.6 µA continuously — which can be twice your entire sleep budget. Below a certain average current, improving the circuit changes nothing at all, because the chemistry sets the floor. The first chart shows exactly where you sit relative to that floor.

Efficiency is not one number

I_battery = I_load · (V_load / V_batt) / η

A switching converter rated 92% at 100 mA may be 50% or worse at 10 µA, because its own control circuitry consumes a fixed amount regardless of load. In a design that sleeps 99.9% of the time, the converter's light-load behavior matters far more than its headline figure. Low-quiescent parts and pulse-skipping modes exist precisely for this.

Temperature and pulse loads

Capacity falls with temperature — an alkaline cell at −10 °C may deliver half its rated capacity. Internal resistance rises at the same time, so pulse loads sag the terminal voltage further. The derating applied here is a rough approximation of both effects; a real design measures the cell it is going to use, at the temperature it will see, under the load profile it will run.

What to do with the answer

• If sleep dominates, chase leakage: floating inputs, pull-ups on idle-high lines, sensors that never enter standby, a regulator with high quiescent current.
• If self-discharge dominates, the circuit is done. Change chemistry or accept the number.
• If transmit dominates, shorten airtime rather than reducing power — see the link budget calculator, where a lower spreading factor cuts airtime by a factor of two per step.
• If wake dominates, the processor is awake too long. Usually a blocking delay, a slow sensor settle, or a boot sequence that runs every cycle instead of once.

Where this stops being accurate

• Cell capacity is not a constant. It depends on discharge rate, temperature, cut-off voltage and age.
• Internal resistance rise with depletion is not modeled, and it is what actually ends a coin-cell device's life.
• Self-discharge accelerates at high temperature — often doubling for every 10 °C.
• Wake and transmit currents are treated as flat. Real profiles have peaks several times the average.
• No allowance for retries. A radio that retransmits on failure will consume more than the profile says, and failure rates rise as the battery sags.

Related: power and battery · IoT and telemetry · connected products

Design guidance, not a manufacturing instruction. Verify every result against your fabricator's stack-up and the applicable standard before release. Closed-form models are approximations; the fab's field solver and process window are the authority.