Key takeaways
A battery pack is a group of cells asked to behave as one. They never quite do. Cells leaving the same production line differ measurably in capacity, in internal resistance and in how quickly they lose charge at rest, and those differences determine how the finished pack performs for the rest of its life.
Sorting is what limits the damage. By grouping cells that are electrically similar, a manufacturer ensures the differences within any one group are small enough that the group behaves predictably. Skip it, or do it loosely, and the pack works perfectly on day one and disappoints steadily afterward.
This article covers what mismatch actually does inside a pack, why battery management cannot correct most of it, and how grading strategy is decided. The measurements themselves are covered in the guide to OCV, IR and ACIR testing.
Three Independent Dimensions of Mismatch
Cell matching is often discussed as though it were one property. It is at least three, and a cell can match well on one and poorly on another.
| Mismatch type | What it does in a pack | How it is detected | Can balancing fix it? |
|---|---|---|---|
| Capacity | The smallest cell limits how much of the pack can be used, in both directions | Formation and grading data, or capacity testing | No; balancing shifts charge but cannot create capacity |
| Internal resistance | Higher-resistance cells dissipate more heat and run warmer under load | ACIR at incoming inspection, DCIR in characterization | No; balancing does not change how current distributes under load |
| Self-discharge | Cells drift apart in state of charge over time at rest | Voltage decay tracked over days of aging | Partly; balancing corrects the drift but consumes energy doing it |
| Temperature position | Cells in warmer locations age faster regardless of their initial match | Thermal modeling and in-service data | No; this is a pack design problem rather than a cell one |
The right-hand column is the one most often assumed away. Battery management systems are frequently treated as the answer to cell mismatch, and for one of these dimensions they help. For the others they do not, and understanding why is the foundation of any sensible grading policy.
What Mismatch Does in Series
Cells in series all carry the same current, so differences show up as differences in voltage and in state of charge. Capacity mismatch is the clearest case. A battery management system must stop charging when any cell reaches its upper voltage limit, and stop discharging when any cell reaches its lower limit. The cell with the smallest capacity therefore reaches both limits first, and the usable capacity of the whole string is set by that one cell. Every other cell in the string is prevented from using capacity it has.
The practical consequence is direct: a string of cells with a two percent capacity spread does not deliver the average capacity of those cells. It delivers roughly the capacity of the smallest, and the difference is paid for and never used.
Resistance mismatch behaves differently. Under load, a higher-resistance cell drops more voltage and dissipates more power internally, because power dissipated rises with the square of current through that resistance. It therefore runs warmer than its neighbors.
What Mismatch Does in Parallel
Cells in parallel share a common voltage, so they self-equalize in state of charge. That sounds like parallel connection solves matching, and it partly does.
What it does not equalize is current. Under load, current distributes between parallel cells inversely to their resistance, so the lowest-resistance cell carries the largest share. It therefore works hardest and heats most, which is the opposite of the series case but produces the same outcome: one cell in the group experiences harsher conditions than the others.
There is a second effect worth knowing. Because parallel cells equalize their voltage, a cell with high self-discharge quietly draws charge from its neighbors rather than simply drifting down itself. That is invisible from outside the group and represents a continuous small energy loss.
The Divergence Problem
The most important consequence of resistance mismatch is not the immediate heating. It is that the effect compounds. A cell running warmer than its neighbors ages faster, because degradation mechanisms in lithium cells accelerate with temperature. As it ages, its internal resistance rises further. Higher resistance means more heat under the same load, which means faster aging still.
That is a positive feedback loop, and it explains a pattern familiar to anyone analyzing returned packs: cells do not drift apart linearly. Differences that were small and stable for the first years widen increasingly quickly toward the end of life. A pack built with tight matching stays coherent for far longer than the initial difference alone would suggest, and a pack built loosely diverges faster than expected.
This is why matching is a service life decision rather than a day-one performance decision. The initial difference is barely measurable in a new pack. What it determines is the shape of the degradation curve.
Why Balancing Is Not a Substitute
Battery management systems balance cells, and this is frequently taken to mean mismatch is a solved problem. It is worth being precise about what balancing does.
Passive balancing bleeds charge off cells that are higher in state of charge, dissipating it as heat through resistors. Active balancing moves charge from higher cells to lower ones, which wastes less energy but costs considerably more in hardware.
Both correct differences in state of charge. Neither changes capacity, and neither changes how current distributes under load. A cell with less capacity than its neighbors still has less capacity after balancing; a cell with higher resistance still runs warmer.
Balancing also works within limits. Balancing current is small relative to pack current, so a system can correct slow drift but cannot keep up with large differences. Where cells are badly matched, the balancing system runs constantly, consumes energy doing so, and may still fail to hold the pack together.
The useful way to see it: good matching reduces the work balancing has to do, and reduces the size of the balancing system the pack needs. Poor matching cannot be balanced away, only managed.
Grading Strategy
Grading turns measurements into bins, and bin width is the central decision. narrow bins produce better-matched groups and longer pack life. They also produce more cells falling outside the target band, and a subtler problem: bin fragmentation. A module requires a specific number of cells from the same bin, so with narrow bins a plant accumulates partial bins waiting to be completed. That is work in progress, storage space and cells sitting idle, and in extreme cases production waits for a bin to fill.
Wider bins avoid fragmentation and improve cell utilization, at the cost of matching quality. Somewhere between the two is a width that suits the application, and it is chosen rather than inherited.
Three factors set it. Application demand, since a high-power automotive pack is less tolerant of mismatch than a low-rate stationary system. Warranty exposure, because a longer warranty makes the divergence problem more expensive. And cell cost, since tighter bins mean more cells diverted to lower-grade uses.
Downgrading rather than rejecting
Cells outside the target band are rarely scrap. They are cells for a less demanding application: a lower-power product, a stationary system with a shorter warranty, or a secondary market. A grading policy that treats out-of-band cells as waste is discarding value, and one that has an established downgrade path can afford tighter primary bins.
Matching Within Modules and Across Packs
Matching operates at two levels and they are not equally important. within a module, cells are directly connected in series and parallel, so mismatch produces the effects described above and matching matters most. This is where tight grading pays.
Between modules in a pack, the requirement depends on architecture. Modules in series see the same current, so a weaker module limits the string in the same way a weaker cell limits a module. Matching modules by measured capacity after assembly, rather than assuming they match because their cells did, is a step that catches accumulated variation.
Where an architecture removes the module level entirely, the whole burden falls on cell grading, since there is no intermediate assembly to test or match. This is one of the manufacturing consequences discussed in cell-to-pack versus module-based design.
“The right grading strategy balances matching quality, cell utilization, warranty requirements, and application demands rather than simply choosing the narrowest possible sorting bands.”
See it in action
Implementation: Sorting Must Be Physical
A point that sounds obvious and is regularly missed in practice: measuring cells is not sorting them.
Measurement produces data. Sorting requires that cells be physically diverted according to that data and kept separated afterward, which means mechanical handling, defined destinations and traceability that survives to the point of assembly. Where measurement happens but diversion is manual or absent, the grading decision is effectively being made by whoever picks up the next cell.
Cell Sorting in Cybernetik Battery Lines
Cybernetik builds battery pack assembly automation with testing and physical sorting integrated at the head of the line, so grading is a machine decision executed mechanically rather than a measurement recorded and acted on later.
| Cybernetik cell sorting capability | Specification |
|---|---|
| Measurements | OCV, IR and ACIR testing on every incoming cell |
| Sorting mechanism | SCARA robots diverting cells onto parallel conveyors by test result |
| Throughput | Up to 6,000 cells per hour through testing and sorting |
| Cell formats | Cylindrical 18650, 21700 and 32140, plus prismatic, pouch and blade |
| Cell diameters | 32, 33, 35, 40, 42 and 46 mm |
| Identity capture | Barcode scanning at entry with results written against the cell identifier |
| Downstream verification | Plasma cleaning, vision-confirmed insulation, polarity checking and inline weld integrity testing |
| Data | MES connectivity with process logging and predictive maintenance analytics |
| Line flexibility | Assembly lines adaptable to multiple module configurations |
Two elements of that specification address the failure modes in this article directly. Robotic diversion onto parallel conveyors means the sorting decision is executed rather than recorded. And barcode capture at entry, with results written against the cell identifier, means a cell can be traced to its bin and its module position years later, which is what makes a field failure analyzable rather than merely regrettable.
Cybernetik has operated for more than three decades, is headquartered in Pune with additional facilities in Gujarat and Raigad and international offices in the United States and UAE, and has installed over 6,000 systems across 30 plus countries, including more than 400 custom robotic automation solutions. In battery manufacturing that work has been delivered for manufacturers including Hero MotoCorp, TVS Motor, Livguard and Matter. Further background is on the Cybernetik about page.