A GPS tracker drawing 4 mA average, idling at 40 °C for nine years, will lose more capacity to the calendar than a power tool that runs 800 full cycles in eighteen months. That is the whole problem in one sentence. Cycle life you can measure on a bench in six weeks. Calendar life you find out about after the warranty has expired, in the field, on a product you cannot recall. So the question procurement and design teams actually ask is not "how many cycles does this cell do" but "will it still hold 80% in year eight, and what do I give up to get there".

The answer, stated before I earn it: for idle-dominant products, calendar ageing is the binding constraint, and two levers decide it — storage voltage and cell temperature. Cycle count is mostly noise in this regime. Oversizing the cell and running it between roughly 3.5 V and 3.9 V per cell is the practical fix. It costs you energy density, board area and BOM money. You do not get calendar life for free.

Three variables, and the ones that do not matter

Calendar ageing in lithium cells comes from parasitic reactions at the electrode-electrolyte interface. Those reactions need two things to proceed: a high driving force (voltage) and kinetics (temperature). Add a third variable that really just modifies the first two, and you have the whole model.

  1. State of charge during storage. A cell held at 4.2 V per cell sits at a much higher electrode potential than one held at 3.7 V. The parasitic current at the cathode is roughly an order of magnitude larger at the top of the range. The relationship is not linear. Going from 4.2 V to 4.0 V might buy you a factor of two in calendar life. Going from 4.0 V to 3.8 V might buy you another factor of two. Going from 3.8 V to 3.6 V buys you much less, and you start paying in usable capacity and low-temperature performance.
  2. Cell temperature. Arrhenius behaviour, roughly doubling of reaction rate per 10 °C in the 20-60 °C window. People underestimate this lever because it is invisible. A cell in a sealed ABS enclosure on a pole in Dubai does not see ambient. It sees ambient plus 10 to 15 °C of self-heating and solar gain during the day. That delta is the difference between a six-year and a twelve-year product.
  3. Time itself. Obvious, but worth naming, because it means calendar loss is unavoidable. You cannot schedule it away. You can only slow the rate.

What does not matter much: cycle count, when you are under about 50 full-equivalent cycles per year. C-rate, as long as you are below 0.5C continuous and your thermal design is not a disaster (the reasons are covered in picking a C-rate). Cell brand, within the same chemistry and quality tier. The gap between a good cell and a mediocre one is maybe 30% on calendar life. The gap between 4.2 V and 3.85 V storage is 3x or more. You are optimising the wrong variable if you spend three months qualifying a second supplier before you look at your float voltage.

Why the two budgets are separate

Cycle ageing is mechanical and chemical wear from lithium insertion and extraction: SEI growth, particle cracking, loss of active material. It scales with charge throughput and depth of discharge. Calendar ageing is parasitic side reactions that run whether or not current flows. It scales with time, voltage and temperature.

For a product that cycles hard and dies young, cycle ageing dominates, and you optimise electrode design, particle size, electrolyte additives. For a product that sits at 60% SoC in a warehouse or on a wall, calendar ageing dominates, and none of those levers moves the needle much. The two mechanisms overlap, and the interaction is real (a cell that has cycled is more susceptible to calendar fade). For design purposes, though, you budget them separately: X% loss from cycles, Y% from calendar, and you need the sum under 20% at end of life. If Y is 18% and X is 4%, you have a calendar problem, and a cycle-life datasheet will not tell you.

The reason this trips people up: datasheets report cycle life. A cell rated 500 cycles to 80% at 1C/1C, 25 °C says nothing about what it does at 4.2 V float for six years. Some vendors publish calendar data (typically 80% retention after 12 months at 4.2 V, 25 °C or 60 °C accelerated). Most do not, and the ones that do often bury it. If your supplier cannot produce a calendar-ageing curve, you are buying an unknown, and the honest move is to derate hard and test yourself. The mechanics of reading what is published are covered in how to read a Li-Po datasheet.

The options, honestly costed

Three real strategies exist for a product that needs eight to ten years of service with low cycling. Each has a price.

Option 1: Accept the calendar loss, oversize the cell

You keep the cell at whatever voltage the application wants, and you buy a bigger cell so that 80% retention at year eight still meets the runtime spec. Need 2000 mAh usable at end of life, expect 25% calendar loss? Buy 2700 mAh at beginning of life.

What it costs: volume, mass, money. Roughly linear in capacity. For a wearable or a thin IoT device, that 35% oversizing may not physically fit, which is why capacity selection for constrained devices is genuinely hard (choosing Li-Po capacity for IoT goes through the arithmetic). What it buys you: no changes to the charging architecture, no firmware complexity, no risk that the customer notices the cell is only ever charged to 85%.

Option 2: Partial-SoC operation, charge and float lower

Terminate charge at 4.0 V or 3.95 V per cell instead of 4.2 V. Discharge floor moves up too, from 3.0 V to 3.3 V or 3.4 V. You lose 15-25% of nameplate capacity immediately, but you cut the calendar fade rate by a factor of two to four depending on temperature.

What it costs: you have to run the numbers on usable capacity, and you need a charger IC or BMS that supports a programmable float voltage rather than a fixed 4.2 V target. If the product ships with a fixed-voltage charger, this option is closed until you change the silicon. Using a smart battery with a fuel gauge? You also need to re-bench the SoC model, because the gauge was characterised against a 4.2 V full point. There is also a customer-perception cost: "why does my device only show 90%" is a support ticket you will have to write a script for.

What it buys you: the biggest single reduction in calendar fade available without changing cells. For a device that spends 95% of its life at rest, this is usually the right answer.

Option 3: Thermal management, or just moving the cell

If your cell sits next to a 5 W radio or under a black polycarbonate lid in direct sun, you are paying an Arrhenius penalty every hour of every day. Moving the cell 20 mm away from the heat source, adding a thermal break, or changing the enclosure colour from black to white can drop cell temperature by 8-12 °C in a sealed outdoor enclosure. That is a factor of two on calendar life, for the cost of a mechanical revision.

What it costs: mechanical design time, sometimes a bigger enclosure, sometimes a redesign of the thermal path. What it buys you: the cheapest calendar life you will ever get, if you are currently thermally careless. Most teams are. The thermal problem does not show up in the six-week qualification test.

When the obvious answer is wrong

The "oversize and float low" advice is correct for maybe 80% of idle-dominant products. Here is where it fails.

When the product cycles harder than you think. A security camera with a solar panel may look idle-dominant on paper. If the panel is undersized, though, the battery does a partial cycle every day. Over ten years that is 3650 shallow cycles, and at 20% DoD that is roughly equivalent to 700 full cycles. Suddenly cycle ageing is not noise. Run the actual daily energy balance before you commit to a calendar-life strategy. If the duty cycle is genuinely daily, you are in a hybrid regime and you need a cell that is good at both, which usually means a slightly less aggressive partial-SoC strategy and more attention to the charge protocol.

When the temperature is not controllable and not 25 °C. A cell at 55 °C ambient does not care much about your clever 3.85 V float. The Arrhenius term dominates and the voltage lever is worth less. At 60 °C, calendar life collapses to a couple of years regardless of SoC. The only real options are a different chemistry (LFP tolerates high temperature better than NMC or LCO), active cooling, or accepting a shorter service interval and designing for field replacement. Do not spend six months on a partial-SoC algorithm for a product that will cook in a rooftop enclosure.

When the customer charges it, not you. If the end user plugs the device into a wall and it sits at 100% for months, your carefully designed 4.0 V float is irrelevant. The charger IC in the device will do what it is told, and the customer will leave it plugged in. This applies to products with a removable battery, or to any device where the charging state is not under your firmware's control. The fix is either a charge-management IC that drops the float voltage once the cell is full (many do, but check the register map), or accepting that the customer's behaviour will cost you years of life.

When the qualification test cannot see the problem. IEC 62133-2 does not test calendar ageing. It tests abuse tolerance, overcharge, thermal propagation, mechanical. A cell can pass every safety test and still fade to 70% in four years in the field. The IEC 62133-2 walkthrough covers what the standard does and does not give you. For calendar life you need your own accelerated ageing plan, and you need to be honest about the Arrhenius extrapolation error (typically ±30% on time-to-80% from a 60 °C accelerated test extrapolated to 25 °C).

When the cell is swelling-limited, not capacity-limited. Calendar ageing at high SoC produces gas. Gas produces swelling. Swelling produces mechanical stress, which can crack the pouch and cause a safety event long before capacity falls below 80%. If your form factor is thin and the enclosure has no room for 8% thickness growth, calendar life is not the binding constraint. Mechanical swelling is, and the design response is different (lower SoC, more headroom, or a different cell construction). The root causes are covered in LiPo battery swelling causes.

Where the battery management layer enters

Partial-SoC operation needs a charger that can be told to stop at 4.0 V. Whether that is a simple PCM with a programmable threshold or a full smart battery with a fuel gauge and an SMBus interface is a cost and complexity decision. Protection PCM vs smart battery covers the trade-off. For a ten-year product, fuel gauge accuracy matters more than usual. The customer will notice capacity fade through the gauge before the cell is actually at end of life, and a gauge that is not re-characterised for partial-SoC operation will drift. Budget for a firmware update path, or accept a wider gauge error band.

The other underrated lever is formation. Cells formed and aged properly at the factory will have a more stable SEI and will calendar-age more slowly than cells rushed through formation. This is a manufacturing variable you can specify. Worth asking your supplier what their formation and ageing protocol is. The mechanism is covered in formation cycling.

If/then: mapping your situation to a choice

  1. If the product is genuinely idle-dominant (under 30 full-equivalent cycles per year) and the cell sits at 20-35 °C then partial-SoC float at 3.85-3.95 V plus a 20-30% oversize is the right answer. Accept the BOM cost and move on.
  2. If the product is idle-dominant but the cell sits above 45 °C for more than a few hours a day then thermal design comes first. Move the cell, add a thermal break, change the enclosure. Only after that, tune the float voltage. If you cannot get below 45 °C, consider LFP.
  3. If the product cycles daily (solar, security, duty-cycled radio) then you have a hybrid problem. Size for cycle life first, then apply a moderate partial-SoC (4.05-4.1 V) to slow the calendar component. Do not go to 3.8 V, you will lose too much usable capacity per cycle.
  4. If the customer controls charging then assume 4.2 V float and size for the resulting calendar fade, unless you can verify the charger IC drops float voltage after full charge. Check the register map before you trust the datasheet block diagram.
  5. If the enclosure cannot accommodate swelling then partial-SoC is mandatory, not optional. High-SoC storage is the primary driver of gas generation in idle cells.
  6. If you cannot get calendar-ageing data from your supplier then derate by 25% per year at 4.2 V, 25 °C as a starting assumption, test at 45 °C and 60 °C in-house, and fit an Arrhenius model with a wide error bar. Do not plan a ten-year product on a five-year guess.

The uncomfortable summary: if you are designing a ten-year idle-dominant product and you are still comparing cycle-life numbers between two vendors, you are looking at the wrong column of the datasheet. The cell that wins on cycles often is not the cell that wins on calendar. Ask for the calendar data. If it does not exist, ask why. Then decide whether you are willing to run the test yourself.