Remote monitoring battery life: what decides ten years

Start with the money, because the money is the point. Once the engineer's time, the vehicle and the overhead behind both are counted, a fully loaded UK site visit costs £150 to £200. Across a 1,000-site estate, one battery-change round is £175,000. It does not matter how efficiently the round is scheduled or how cheap the replacement cells are; the visits are the cost, and the estate pays it every time the batteries run out.

Whether that round happens once a decade or three times was settled before the first unit shipped, by three choices: the sleep behaviour of the firmware, the cell that was specified, and the reporting interval the deployment runs at. None of the three can be changed cheaply once tags are fitted across an estate. That is why a battery change is not maintenance. It is the invoice for a design decision, arriving years after the decision was taken.

Take a concrete design: a D-size bobbin lithium thionyl chloride cell holding 19 Ah, a tag reporting every three minutes, and a ten-year target. Ten years is 87,600 hours, and at three-minute reporting it is 1.75 million transmissions.

Each transmission is a 20 ms burst at 25 mA, which costs 0.000139 mAh. Multiply by 1.75 million and the entire decade of transmissions costs 243 mAh: 1.3% of the cell. The radio, the component everyone instinctively worries about, is nearly free.

Sleep current is the opposite case. The device is asleep for 99.98% of its life, so whatever it draws while asleep is multiplied by an enormous number of hours. Transmission is not.

The failure modes that push a design down this table are mundane. A floating GPIO. A pull-up resistor left enabled. A voltage regulator sitting in normal mode when a low-power mode exists. An ungated peripheral clock. Each costs tens of microamps, and the table prices what tens of microamps means: it is the difference between ten years and three. On the bench these leaks are invisible, because the current probe is pointed at the transmit pulse and nobody is watching the floor. In the field, across an estate, they are decisive.

Sensing has its own budget line, and it is the one most designs overlook. A sensor that needs 10 mA for 200 ms to stabilise before each reading costs 2 mA·s per sample. Sampled once a minute, that is 5.26 million samples over ten years: around 2,900 mAh, 15% of the cell, and twelve times the radio. The measurement, not the transmission, is the larger energy event, and it is hidden inside a line of firmware that looks free.

The pattern that resolves it is to sample locally and report on change. The device reads its sensor on its own schedule, as cheaply as the sensor allows, and spends transmission energy only when the value has done something worth reporting. A temperature that has not moved does not need announcing every minute; a temperature that has crossed its band needs announcing immediately. Getting the sensor's stabilisation time and duty cycle into the power budget, next to sleep current, is the difference between a calculated life and a hopeful one.

A primary cell consumes itself while it sits, and the rate is a property of how the cell was built. Tadiran's bobbin lithium thionyl chloride cells are specified at under 1% annual self-discharge, which is what enables up to 40-year operating life on low-current applications, per Tadiran's published technical data on long-life lithium batteries. The same comparison notes that a lower-quality bobbin cell can run up to 3% per year, losing 30% of capacity every ten years.

Put those two figures against the 19 Ah cell and the gap is stark: roughly 1,800 mAh lost over ten years at the specified rate, against roughly 4,900 mAh from the lower-quality cell. The 3,100 mAh between them is more than twelve times the entire ten-year radio budget, and it is lost before the firmware draws a microamp. In estate terms that gap is roughly three years of life, and on a 1,000-site estate it is one extra £175,000 battery round, paid for a component substitution that no spreadsheet line ever showed.

There is an architectural decision underneath all of this arithmetic. A receiving device must listen, and listening cannot be compressed to milliseconds the way transmitting can, because the device does not know when the other end will speak. Every listening window is paid for in full whether anything arrives or not, and the windows recur for the whole life of the device.

Our tags transmit and do not receive, and the reasoning starts with security rather than power: a device with no receive path has no inbound attack surface to defend for a decade in the field. The battery consequence follows as a second benefit. With no receive window to fund, the radio budget really is the 243 mAh of transmission arithmetic above, and nothing more. The full engineering case is set out in our article on one-way telemetry as an architectural decision.

Bobbin construction is what buys the energy density and the low self-discharge, and it carries a cost: a low-rate design with high internal impedance that struggles to deliver the radio's pulse. Tadiran call the resulting voltage sag transient minimum voltage, and a cell that cannot hold voltage through the transmit pulse browns the device out regardless of how much capacity remains on paper.

There are two engineering answers. One is a hybrid layer capacitor in parallel with the cell, which delivers the pulse and recharges quietly between transmissions: that is the construction behind Tadiran's PulsesPlus series. The other is a cell designed for the duty from the start, which is the TRR series. Tadiran's families split by pulse capability set against operating life:

Passivation is the last trap. A film forms on the anode during storage, and it is part of what makes the low self-discharge possible, but it raises the cell's impedance until the first significant load burns it through. A tag fitted after a long spell in a warehouse can stumble on its first transmissions with a cell that is essentially full. It is a known property of the chemistry with known handling, and a supplier who has fielded the chemistry at scale will have an answer ready when asked about it.

Cold is the environment where all of the above compounds. Low temperature raises the cell's internal impedance exactly where the design is least able to absorb it, on the transmit pulse, and reduces usable capacity at the same time. That is not a reason to avoid battery-powered monitoring in a freezer. It is a reason to specify the cell for the temperature rather than for the brochure.

Tadiran specify a typical bobbin LiSOCl2 operating range of −55°C to +125°C, with cells that can be modified for cold chain use operating continuously and reliably down to −80°C. They have also reported cryogenic chamber testing in which cells were taken progressively down to −100°C and continued to operate, which is a test result rather than a rated operating range. A −55°C requirement, covering commercial freezers at −18 to −25°C, deep freeze and blast freeze, therefore sits inside the standard range with margin rather than at its edge. This is the specification thinking behind our cold storage monitoring work, where the tag lives inside the cold space with the stock.

The test of a supplier is simple here. A battery life figure quoted without asking the operating temperature is a figure calculated at room temperature, and a supplier quoting one battery life figure without asking the operating temperature has not done the calculation.

Where nobody visits anyway, the battery sets the visit schedule. On an unmanned site the tag's cell is often the only consumable in the building, and its life decides whether the site sees an engineer once a decade or every three years. A retrofit deployment across a multi-site estate multiplies that decision by every site in it, which is why battery life is a first-order specification for remote monitoring and not a datasheet footnote.

Everything in this article is visible in the design of the ZARC family of tags: transmit-only radio, sample-locally-report-on-change firmware, and cells specified for the application and its temperature. The same arithmetic sits underneath our IoT remote monitoring work across estates of exactly this kind: unmanned, distributed, and expensive to visit.

Battery life claims are cheap to make and expensive to be wrong about. These six questions separate a calculated figure from a hopeful one.

One. What reporting interval is the quoted life calculated at, and what does halving the interval do to it? If the answer is that halving barely moves the figure, the radio was never the constraint, which is correct, and the supplier should be able to say what the constraint actually is.

Two. What is the measured sleep current of the finished device, not the microcontroller's datasheet figure? The datasheet describes the chip. The estate pays for the board, with every peripheral, pull-up and regulator it carries.

Three. What is the specified annual self-discharge of the cell, and which cell is it? Under 1% a year and up to 3% a year are both sold as bobbin lithium thionyl chloride, and the difference is roughly three years of estate life.

Four. Is the life calculated at our operating temperature, or at 20°C? A freezer, a plant room and a south-facing cabinet are three different batteries.

Five. How long does the sensor take to stabilise before a reading, and does it stay powered between samples? Sensor duty is routinely the largest line in the budget and the least often quoted.

Six. What pulse current does the design draw on transmit, and how is passivation handled after storage? Both questions have good answers; the tell is a supplier who has never been asked.