How to choose a lithium-ion cell for a pack

Cell selection goes wrong in a predictable order. First someone picks a part number they have used before. Then the pack is designed around that cell's dimensions. Then, at the end, someone checks whether the cell can deliver the current the load needs, and discovers it cannot — or that it can, but only while the cell stays below a temperature the housing never achieves. The sequence below runs the same checks in the opposite order, which is the order that survives contact with a prototype.

Step 1 — Write down the load, not the cell

Four numbers decide almost everything downstream, and none of them appear on a distributor listing:

• Peak current and how long it lasts (a 60 A burst for two seconds is a different requirement from 60 A for two minutes).
• Continuous current — the figure the pack will sit at for most of its life, not the figure on the marketing page.
• Duty cycle — how often the load returns, and how much of that time the pack is actually working.
• Required runtime or energy at the end of life, not at the start.

If the four numbers are not written down before the shortlist, the shortlist is a guess. In practice the continuous figure and the runtime figure pull in opposite directions, and the whole job of selection is deciding how much of one to trade for the other.

Step 2 — Convert the duty into current per cell

A cell is only ever asked for its share of the pack current. That share is set by the parallel count, so the arithmetic has to happen before comparing cells:

Series count sets voltage, parallel count sets the current each cell sees. Both are decided by the mechanical envelope as much as by the electrical requirement.

The first row is the one worth staring at. A pack that looks like a 30 A design is a 7.5 A per-cell design, and a mid-grade 21700 with a 9.8 A continuous rating covers it with margin. Choosing a 60 A cell for that pack buys nothing except cost and a lower capacity in the same can.

Step 3 — Apply the temperature cut-off before you trust the rating

A continuous discharge figure is valid only while the cell stays inside the temperature the manufacturer specified when measuring it — commonly 80 °C or 83 °C surface temperature for high-power cylindrical cells. The number is not a property of the cell alone; it is a property of the cell plus a thermal path.

This is where most first prototypes fail. A cell rated 35 A in free air at 25 °C ambient will not deliver 35 A inside a sealed enclosure with no airflow, because the cell reaches its cut-off earlier and the pack either derates itself or runs hotter than the design assumed. If the housing cannot guarantee the condition, derate the cell until it can, and write the derating into the specification rather than leaving it as an assumption.

Step 4 — Check impedance, then check it again at end of life

Impedance decides how much voltage the pack loses under load, and it rises with age. Two cells with identical capacity can differ by a factor of two in DCIR, and the higher-impedance one will hit the low-voltage cut-off earlier under the same current — which reads to the user as a capacity problem.

The arithmetic is worth doing by hand once. A cell with 25 mΩ DCIR carrying 7.5 A drops 188 mV. Thirteen in series drop 2.4 V across the pack. If the pack cuts off at 3.0 V per cell, that sag consumes a meaningful slice of the usable window before the cells are anywhere near empty. By mid-life, with impedance up by half, the same current consumes half again as much.

AC impedance at 1 kHz (ACIR) is the figure most data sheets publish, because it is quick to measure. DCIR under a current pulse is closer to what the pack actually experiences. Where both exist, design against DCIR; where only ACIR exists, treat it as a lower bound rather than the design value.

Step 5 — Match charge acceptance to the duty cycle

Charge current is the constraint people forget until the first field complaint. A pack that must be ready in an hour needs a cell whose maximum charge rate supports it: at 5,000 mAh, a 1 C charge is 5 A and a 5 C charge is 25 A, and the difference decides whether the product is usable in a shift-based operation.

Standard charge and maximum charge are separate columns for a reason. The standard figure is the one the cycle-life rating was measured at; running at the maximum figure continuously will shorten life, and the data sheet usually says so in a footnote rather than a heading. Charge below 0 °C is a separate prohibition: most cells require a reduced rate or no charge at all, and a pack without a temperature-aware charger will violate that quietly.

Step 6 — Confirm the envelope, the terminal and the weld method

Mechanical constraints eliminate more candidates than electrical ones. Diameter and height maxima, terminal style (flat, raised, tabbed), and the interconnect method all have to be settled before the shortlist is final:

• Cylindrical cells need resistance welding or laser welding of the interconnect. Soldering directly to a cell can is not an acceptable production method — the heat damages the seal and the internal connections.
• Holder geometry differs between nominally identical formats; a holder designed for a 65.0 mm cell may not accept a 65.2 mm one.
• If the pack uses a plastic carrier, check the cell's maximum diameter including the sleeve, not the bare can.

Step 7 — Ask for the documentation before the quotation

The documentation question is a selection criterion, not paperwork. Ask for the UN38.3 test summary, the safety data sheet matched to the shipping configuration, and the certificate numbers with the model list they cover. A supplier who cannot produce them is telling you something about the rest of the answer too.

On this site every specification carries its provenance: values read from a manufacturer data sheet are marked verified and the PDF is linked; models without a data sheet on file are marked pending. That distinction matters at this step — a pending model may still be the right cell, but you are designing against an unverified figure, and you should know that before the first prototype rather than after.

Where this sequence stops working

The order above assumes a duty that is roughly continuous and a thermal environment you control. Three cases break the assumption, and each needs a different approach:

• Very short, very high pulses (hundreds of amps for milliseconds): the relevant figure is pulse capability and impedance, not continuous rating, and the thermal mass of the interconnect matters as much as the cell.
• Extreme cold: charge acceptance collapses below 0 °C and usable capacity falls well before that. Selection has to start from the cold-end requirement, not the room-temperature data sheet.
• Safety-critical or regulated products: the standard the product must meet (IEC 62133-2 for portable, IEC 62619 for industrial) constrains the cell choice before any of the above arithmetic matters.

If the application is in one of those three, run the sequence but start at the boundary condition rather than at the load profile — the rest of the method is unchanged.

Questions this guide does not answer in passing