Cell Grading, Matching and Sourcing
Why a series string is only as good as its worst cell, what spreads are acceptable, and why Grade B cells cost more than they save.
Battery Fundamentals · Part 4 — Specification and Sourcing · Chapter 11 · 15 min read
A pack made from mismatched cells has usable capacity equal to its worst cell, and the mismatch widens over time rather than settling. That single fact governs everything about how cells should be bought, tested and binned — and it is why cell selection is a commercial decision as much as a technical one.
≤2 %
Capacity spread
≤10 mV
OCV at assembly
≤5 %
ACIR spread
25–40 %
Grade B discount
11.1 — What the grade letters buy you
The grades and where they are decided are covered in Chapter 4. The point to carry into a purchase decision is that there is no international standard behind A, B and C. They are trade terms, and the only way to know what a seller means by them is to ask which parameter bands they refer to.
Important
The distinction that actually matters is not strong versus weak. It is dispersion versus defect. A cell 5 per cent under capacity is a performance limitation you can design around by matching. A cell that failed the self-discharge screen has an internal micro-short, and no amount of matching helps — it will fail.
11.2 — Why a series string is only its worst cell
In a series string, every cell carries the same current. There is no way for a strong cell to compensate for a weak one, and the BMS — correctly — protects the weakest in both directions.
- •On charge: the highest-voltage cell reaches the cut-off first, the BMS stops charging, and the pack never fills. Every other cell is left short.
- •On discharge: the lowest-voltage cell reaches the cut-off first, the BMS disconnects, and the pack never empties. Energy remains in every other cell, unusable.
Worked example 11.1 — What a 10 per cent outlier costs
A 20S1P pack of 100 Ah cells, nineteen of them healthy and one at 90 Ah.
Nominal pack energy: 20 × 3.2 V × 100 Ah = 6,400 Wh
Actual usable, limited by the worst cell: 20 × 3.2 V × 90 Ah = 5,760 Wh
You paid for 6.4 kWh and you have 5.76 kWh — a 10 per cent loss from one cell out of twenty. Worse, that cell now takes a deeper discharge on every cycle than its neighbours, so it ages faster, and the gap grows. A mismatch is not a fixed penalty; it is a diverging one.
Technical framing
Note that balancing does not fix this. Balancing equalises charge so that cells reach the same voltage together. It cannot equalise capacity, because capacity is a physical property of the cell. A well-balanced pack of mismatched cells is still limited by its worst cell — balancing simply stops the mismatch getting worse than it needs to be. The distinction is developed in Chapter 2.
11.3 — Matching criteria
Acceptable spread across a batch destined for one pack
| Parameter | Acceptable | Tight | Why |
|---|---|---|---|
| Capacity | ≤2 % | ≤1 % | Sets usable pack capacity directly, and the spread widens with age. |
| OCV at assembly | ≤10 mV | ≤5 mV | Cells joined in parallel at different voltages equalise instantly, at very high current. |
| Internal resistance (ACIR) | ≤5 % | ≤3 % | Drives differential heating, which becomes differential ageing. |
| Manufacturer, model and production batch | Identical | Identical | Not a tolerance — a requirement. Different batches have different capacity distributions. |
Use ACIR for this rather than DCIR: it is fast, repeatable and non-destructive, which is what a sorting measurement needs to be. The distinction is in Chapter 9.
11.4 — What you must never mix
- •Different batches. Even from the same manufacturer and model, batches differ in coating weight and therefore capacity distribution.
- •Different vendors. Even at identical specifications, ageing behaviour differs, so a pack that starts matched will not stay matched.
- •New with used. This is the one that costs money in service. Adding one new cell to an aged pack destroys the new cell: it has more capacity than its neighbours, so it is worked hardest, and it ages down to meet them. The correct repair for a failed cell in an aged pack is rarely a single new cell.
- •Different ages. Cells more than six to twelve months apart in date code have measurably different calendar ageing behind them.
Important
One more, from the pack-design side: never parallel cells at different states of charge. The instantaneous equalisation current between two cells 200 mV apart can be hundreds of amps through a millohm of connection, with nothing in the circuit to limit it. Bring cells within 50 mV before joining them.
11.5 — Sourcing due diligence
Six things to request or verify on every purchase. None of them are unusual requests, and all of them are cheaper than a field failure.
- •The factory test report for your batch — capacity, OCV, IR and K-value per cell, or at minimum as a distribution. A generic model report is not this.
- •The UN 38.3 test summary — legally required for transport, and it must name the exact model rather than a family.
- •The date code on the physical cell — verified on arrival, not taken from the invoice. Cells more than six to twelve months old have already spent calendar life you are paying for.
- •Storage state of charge on arrival — should be around 30 to 50 per cent. Arriving at 0 per cent means over-discharge in transit; arriving at 100 per cent means hard calendar ageing on a hot ship.
- •The MSDS and a dimensional drawing with tolerances — you need the first to ship and the second to design the enclosure and compression fixture.
- •Your own incoming inspection on a sample. Trust, then verify. The procedure is in Chapter 15.
11.6 — The commercial trap
Grade B cells cost 25 to 40 per cent less and pass a bench test perfectly on day one. That is the whole difficulty: the saving is immediate and measurable, and the cost arrives later and lands somewhere else on the balance sheet.
In plain English
They fail at month six to eighteen — inside warranty, in the customer’s hands, at your cost. And on a financed pack that is not merely a warranty claim. It is a vehicle off the road, an operator who cannot earn, a defaulted instalment, and a lender revising its view of your portfolio.
Important
Work the arithmetic before deciding. A 30 per cent saving on cells is perhaps 15 per cent of pack cost. One replacement inside warranty costs a full pack plus labour plus logistics plus the customer relationship — several times the saving, on a single unit. The break-even failure rate is low enough that it is not a close call.
Second-life and refurbished cells are a separate question and a legitimate one. Harvested, re-tested and honestly re-graded cells have a real place in stationary storage where the duty cycle is gentle, the installation is monitored and the labelling is truthful. They do not belong in a vehicle pack, and they certainly do not belong in a financed one.
Quick check: test yourself
1.One cell in a 20S1P pack is 10 % below the rest. How much capacity have you lost?
Show answer
2.A service technician replaces one dead cell in a two-year-old pack with a new one. What happens?
Show answer
3.You have two batches of identical cells from the same manufacturer, both Grade A. Can you build one pack from both?
Show answer
4.Grade B cells save you 30 %. What is the actual break-even?
Show answer
Chapter summary
- ✓Grade letters are trade terms with no standard behind them. Ask which parameter bands a seller means, and distinguish dispersion (a design constraint) from defect (a failure).
- ✓A series string is limited by its worst cell in both directions, and the mismatch widens with age rather than settling.
- ✓One cell 10 per cent low in a 20S1P pack costs about 10 per cent of pack capacity — and then costs more each month.
- ✓Balancing equalises charge, not capacity. It cannot rescue a mismatched pack; it only stops things getting worse than necessary.
- ✓Match to ≤2 % capacity, ≤10 mV OCV and ≤5 % ACIR, from the same manufacturer, model and batch. Use ACIR for sorting.
- ✓Never mix batches, vendors, ages, or new with used — and never parallel cells more than about 50 mV apart, because the equalisation current is unlimited.
- ✓Request the batch test report, UN 38.3 summary for the exact model, date code, arrival SOC, MSDS and dimensional drawing. Then inspect a sample yourself.
- ✓Grade B saves 25 to 40 per cent on cells and pays it back as in-warranty failures. On a financed pack a cell failure is also a defaulted instalment.
Frequently asked questions
How closely do cells need to be matched?+
Capacity within 2 per cent — 1 per cent for a tight build — open circuit voltage within 10mV at assembly, and internal resistance measured as ACIR within 5 per cent. Same manufacturer, same model and same production batch is not a tolerance but a requirement, because batches differ in coating weight and therefore in capacity distribution.
What happens if one cell in a pack is weaker than the rest?+
The whole pack is limited by it in both directions. On charge, that cell reaches the cut-off first and the BMS stops, so the pack never fills. On discharge, it reaches the low cut-off first and the BMS disconnects, so the pack never empties. One cell 10 per cent low in a 20S1P pack costs about 10 per cent of pack capacity — and it worsens, because that cell now takes a deeper swing every cycle and ages faster.
Can I replace a single dead cell in an old pack with a new one?+
You can, and it destroys the new cell. It has more capacity than its aged neighbours, so in a series string it is worked hardest and ages rapidly down to meet them. Either replace the whole string with matched cells, or source a used cell matched in capacity and resistance to the existing pack. This should be a written policy for a service network, because the intuitive repair is the wrong one.
Does balancing fix mismatched cells?+
No. Balancing equalises charge so cells reach the same voltage together; it cannot equalise capacity, which is a physical property of the cell. A perfectly balanced pack of mismatched cells is still limited by its worst cell — balancing only stops the mismatch getting worse than it needs to be.
Are Grade B cells worth the saving?+
Rarely. Cells are roughly half of pack cost, so a 30 per cent cell saving is about 15 per cent on the pack. One in-warranty failure costs a replacement pack plus labour, logistics and the customer relationship — several times that saving on a single unit. On a financed pack add a vehicle off the road, an operator unable to earn, a defaulted instalment and a lender re-rating your book.
Why must cells be at similar voltage before paralleling them?+
Because the equalisation current is unlimited. Two cells 200mV apart joined through a millohm of busbar will exchange hundreds of amps instantaneously, with nothing in the circuit to restrict it. Bring cells within 50mV of each other before connecting them in parallel.
Reviewed by
Battery Fundamentals is an original educational series on lithium battery technology. Threshold tables, cycle-life curves and worked examples use representative values drawn from published product specifications and widely-observed industry patterns, not measured data from a specific product. Always consult the current manufacturer datasheet before making design, purchasing, warranty or certification decisions.