Battery Life Calculator

Average current is the only number that matters

A battery holds a certain amount of charge and the device takes it out at some average rate; life is the ratio. Everything difficult about the estimate is in the word average. A sensor node that sleeps at 5 µA, wakes for 10 ms at 5 mA and transmits for 5 ms at 20 mA does not draw "about 5 mA" or "about 20 mA" — it draws the charge of each phase summed over a wake-up, divided by the interval. For a 60 s interval that is 0.3 mC asleep, 0.05 mC active and 0.1 mC transmitting: 0.45 mC per minute, or 7.5 µA. The 20 mA burst that looks like the whole story is a fifth of the drain.

Sleep current wins or loses it

That arithmetic is why sleep current dominates almost every low-power design. It is small, but it is on for 99.97 % of the time. Cutting the transmit burst in half saves 0.05 mC per wake-up; cutting sleep current from 5 µA to 2 µA saves 0.18 mC. The breakdown table shows the share each phase takes, and the phase with the biggest share is the one to work on — which is usually the one with the smallest current.

The sleep figure has to be the whole board's, not the microcontroller's. A 100k/100k divider to sense the battery voltage draws 15 µA from a 3 V cell continuously, three times the MCU's sleep current, and it is the kind of thing that appears on a schematic without anyone thinking of it as a load. LDO quiescent current, pull-ups, the brown-out detector, an unused peripheral clock, a debug pin left driving a pull-down — measure the board asleep with a meter that resolves microamps, then put that number in.

Rated capacity is not usable capacity

The mAh on the label was measured at a low, steady current down to a cutoff voltage, at room temperature, on a fresh cell. Your device runs pulses, may reset above that cutoff, sits in the cold, and the cell has been on a shelf. Each takes a slice. Eighty percent is a fair working figure for a gentle load; a coin cell delivering radio bursts may give half its rating, because the pulses pull the voltage below the device's minimum while plenty of charge is still inside. The usable field is where that judgement goes, and the tool reports the usable figure it actually divided by.

Self-discharge is a load too

Every cell leaks. A lithium coin cell loses about a percent a year, an alkaline two or three, a modern low-self-discharge NiMH a percent or two a month, an older NiMH twenty. The tool converts the rate into an equivalent current and adds it to the load, and for a device that wakes rarely the comparison can be startling: a 225 mAh coin cell at 1 % a year is leaking 0.3 µA, which is a third of a well-designed node's sleep current before the device does anything at all. When self-discharge takes the biggest share, the chemistry is the lever, not the firmware.

Coin cells and the pulse problem

Capacity is not the only thing a cell has to supply. A CR2032 has an internal resistance of ten to twenty ohms fresh and thirty or more as it empties. A 20 mA radio burst through 15 Ω drops 0.3 V; through 40 Ω, 0.8 V. The cell that reads 3.0 V at rest is at 2.2 V during the burst, and if the radio's minimum is 2.3 V the device resets mid-packet with most of the charge still in the cell. The tool computes the voltage under the largest pulse and compares it with the device minimum you give it. The standard fix is a capacitor across the cell — a few hundred microfarads supply the burst and recharge slowly between them — or a cell that can deliver the current. Lithium thionyl chloride cells have superb shelf life and the same weakness; the hybrid-layer-capacitor versions exist for exactly this reason.

Regulators: what the efficiency field means

An LDO passes the load current through: what the load draws, the battery supplies, plus the regulator's own quiescent current, which belongs in the sleep figure. Set efficiency to 100 for an LDO. A buck converter draws less current from a higher-voltage battery — power in equals power out over efficiency — and the field lets you account for it. But most bucks are efficient at milliamps and poor at microamps, and their quiescent current can exceed the sleep load; check the datasheet's light-load curve rather than the headline figure before assuming a buck helps a device that mostly sleeps.

Measure, then calculate

The estimate is exactly as good as the currents you type in, and the currents are the part that surprises people. The active and peak figures come from the datasheets and are usually close; the sleep figure comes from the board and is usually not. Measure it. Then the wake interval is a firmware decision, and the timer prescaler calculator gives the register values for the RTC or timer that sets it. If the device senses its own battery, the voltage divider calculator shows what that divider draws — and it is a reason to switch it through a transistor rather than leave it connected.