Marketplace titles that say “rechargeable CR2032” usually describe a LIR2032. The can is 20 × 3.2 mm, so the part drops into a CR2032 tray. The electrical system is not the same part. Nominal voltage, charge voltage, capacity, and whether the cell may be charged all change when the chemistry letters change from CR to LIR. A firmware team that treats those letters as a packaging detail will find out later, at brownout or at the charger.

PlainBattery is a free battery cross-reference, not a cell vendor. Its method is to match marking, voltage, chemistry, and physical size before calling anything a replacement, and it says a chart match is not a device guarantee. The three pages below are useful when you read them that way: as first checks on voltage class, on what a milliamp-hour rating actually includes, and on whether a coin cell can survive a radio pulse.

Same outline, different rail

PlainBattery’s CR2032 vs LIR2032 comparison states the mismatch without softening it. CR2032 is a 3.0 V primary lithium-manganese dioxide cell, typically listed at about 220–240 mAh (about 0.66–0.72 Wh at the 3.0 V nominal voltage). LIR2032 is a rechargeable lithium-ion cell in the same 20 × 3.2 mm outline, with a nominal voltage of about 3.6–3.7 V and a typical capacity of only about 35–45 mAh (about 0.13–0.16 Wh at 3.6 V). Fresh open-circuit voltage is not a minor detail either. The comparison lists about 3.2–3.4 V for a fresh CR2032, which is normal for a Li-MnO₂ primary, and about 4.1–4.2 V for a LIR2032 just off a charger. That last figure is a charge-voltage neighborhood, not the 3.6–3.7 V nominal value, and it is well above what a 3 V-only input was designed to see.

Those energy figures are just nominal voltage times typical capacity. They are consistent arithmetic, not a claim that every brand, lot, and cutoff will land in the same place. Capacity is always tied to a load, a temperature, and a cutoff. The gap is still large enough that “rechargeable” is not an upgrade in a holder that specified a primary CR2032. You give up most of the stored energy, and you may apply an over-voltage to the board.

The comparison’s safety notes are the ones that matter in a lab as well as in a consumer gadget. Installing a LIR2032 in a device that specifies a disposable CR2032 can over-voltage the circuit. Putting a disposable CR2032 into a charger meant for LIR2032 can overheat or rupture the primary cell. Charging is required for LIR2032 and unsafe for CR2032. Use a LIR2032 only when the product is designed to charge it, with the charge control that design assumes. This is not a reason to improvise a charger, to bypass a protection circuit, or to rebuild a pack.

ML2032 is a different rechargeable chemistry again: it can share the 20 × 3.2 mm outline and still need a different charge voltage. Follow the schematic, not the photograph of the can.

mAh is a test condition, not a runtime

PlainBattery’s article on measuring current and using mAh correctly starts from a formula everyone already knows and then restricts it:

Average current = Σ(current × duration) ÷ cycle duration

Runtime hours = usable capacity in mAh ÷ average current in mA

The formula holds only when both numbers describe the same conditions. A printed milliamp-hour rating was taken at a specified load, temperature, and cutoff. A product sleeps, wakes, transmits, and shuts down at its own minimum voltage. The article’s example is straightforward and arithmetically consistent. A sensor sleeps at 12 µA for 59 seconds and transmits at 18 mA for one second:

Sleep: 0.012 mA × 59 s = 0.708 mA·s

Transmit: 18 mA × 1 s = 18 mA·s

Average: 18.708 ÷ 60 = 0.312 mA

A nominal 235 mAh cell then gives an ideal 235 ÷ 0.312, or about 753 hours. The 235 mAh figure is taken from the Energizer CR2032 sheet as cited on that page: a 15 kΩ load at 21 °C, down to 2.0 V. A board that resets at 2.7 V cannot use all of the capacity measured to 2.0 V. Cold, age, and rising internal resistance reduce usable energy further.

The same article treats derating as a scenario, not a law. It shows 90%, 70%, and 50% usable fractions as optimistic, planning, and conservative cases. At 70%, the example becomes 235 × 0.70 ÷ 0.312, or about 527 hours. Replace the percentage with a measured discharge when you can. Do not advertise 753 hours, or 527 hours, as a property of every CR2032 in every enclosure.

If the load sits behind a regulator, capacity in mAh is the wrong first unit. Energy is the one that crosses a conversion boundary:

Battery Wh = nominal voltage × capacity in Ah

Load W = output voltage × average output current

Runtime = battery Wh × efficiency ÷ load W

Read efficiency at the expected input voltage and load, and add regulator quiescent current separately. At sleep, that quiescent current can exceed the current of the circuit you thought you were budgeting. Distinguish energy (watt-hours) from capacity (milliamp-hours), and distinguish nominal voltage from the open-circuit voltage of a fresh or freshly charged cell.

A resistor discharge characterizes a cell. A complete-device test, in the real enclosure and firmware, shows whether the product can still do its job at the cutoff you actually use. The article is also clear that a slow multimeter can miss short pulses; a profiler that can see microamps and milliamps in one trace is the better instrument for a duty-cycled radio.

Select the can after you have seen the pulse

PlainBattery’s guide to coin-cell power for MCUs is a selection article, not a shopping list. CR2032 and CR2450 are both 3 V primary lithium coins, with different dimensions and capacity. A larger cell generally stores more energy, but neither should be chosen from the catalog mAh alone. The questions the page wants measured on the assembled board are more specific: charge consumed over a complete duty cycle; minimum voltage during the largest pulse; behavior near end of life; cold-start and cold-transmit behavior; and whether the regulator is helping or quietly spending the budget.

An MCU data sheet may advertise nanoamp sleep while the finished board consumes tens of microamps. Radio bursts, sensor warm-up, flash writes, pull-ups, regulators, and indicator LEDs decide whether a coin cell works. Measure release firmware with the debugger and USB interface disconnected. Capture startup, sleep, sampling, processing, transmit and receive, retries, and non-volatile writes. Measure voltage at the MCU pins as well as current at the holder. A dip on a long trace, a weak spring, or a contaminated contact can reset the chip while the cell terminals still look fine.

For a short pulse, the page’s first-order capacitor estimate is the usual one:

C = current × pulse duration ÷ allowed voltage drop

A 15 mA, 8 ms pulse with 0.2 V of allowed droop gives 0.015 × 0.008 / 0.2 = 600 µF in the ideal calculation. Real capacitors have tolerance, leakage, ESR, and temperature dependence. The cell still has to supply the average energy and recharge the capacitor between events. More capacitance will not fix a continuous over-budget current or a retry loop.

End-of-life behavior is not a fresh-cell transmission that happened to work. Internal resistance rises as the cell discharges, which deepens pulse sag. The guide suggests emulating that with a documented series resistance or a programmable source, then repeating the test at the lowest intended temperature. Do not treat a homemade “aged cell” recipe, a charger experiment, or a holder bypass as part of the method.

Holder depth is an electrical problem in disguise. CR2032, CR2025, and CR2016 share a 20 mm diameter and not a thickness. CR2450 is wider. A mechanically loose cell can look like a firmware reset. If the production holder was cut for 3.2 mm, a 2.5 mm can is the wrong part, however attractive its price.

Read the letters, then the current

A 2032 outline tells you the can will probably fit a 20 × 3.2 mm tray. It does not tell you the rail, the charge chemistry, or how long the board will run. CR means a 3.0 V primary Li-MnO₂ coin in this family. LIR means a rechargeable lithium-ion cell at a higher nominal and charge voltage, with far less capacity. mAh means a rating taken under stated conditions. Average current means a measured duty cycle, including the pulses a meter display will miss.

Keep coin cells in child-resistant packaging and out of reach of children, and do not charge a primary CR2032. A cross-reference can keep you from treating a LIR2032 as a drop-in. A current profile can keep you from believing a catalog capacity. Neither replaces the manufacturer’s limits, and neither is a certificate for a medical, automotive, or mains-connected design.