A battery cell sorting machine measures every cell — open-circuit voltage, AC internal resistance, capacity and K-value self-discharge — and drops it into a grade bin, so only electrically matched cells enter the same module. NEGUP MACHINE builds sorting machines with 5 to 32 grade bins plus a dedicated NG channel, 0–5 V / 0–200 mΩ channels, barcode binding and MES export.
Send your cell format, daily cell volume and matching tolerances on WhatsApp +86 153 0769 0902.
Updated: Oct 12, 2026 · Author: NEGUP Content Team · Reviewer: pending human fact-check before publish.
What the Sorting Gate Actually Decides
Sorting sits between capacity grading and module assembly and is the last chance to remove a cell before a weld makes removal expensive. A pack is a series-parallel network: the weakest cell sets the usable capacity of its string, and every millivolt of spread becomes a balancing current the BMS carries for the life of the product. Published pack-line practice writes consistency as numbers — capacity spread within 1% of nominal, internal resistance within 0.5 mΩ, open-circuit voltage within 5 mV, and a self-discharge screen at 0.05 mV/h or better. Those are shift-long tolerances, not sample tolerances.
The Market Behind the Gate
Sorting capacity is being bought at roughly the pace of cell capacity, and the equipment is repriced around measurement accuracy rather than mechanical speed. The 2026 numbers below come from three houses that agree on direction and differ on scope: a market of USD 2.6 billion and one of RMB 6.4 billion are counting different things.
| Scope | 2026 number and forecast | Source |
|---|---|---|
| Lithium battery sorting machines, global | USD 2.34B (2025) → 2.60B (2026) → 3.85B (2033), 10.8% CAGR | Stats N Data, 2026 |
| Battery consistency inspection & sorting, equipment plus services | USD 8.73B (2026) → 14.26B (2030), 13.1% CAGR; China 34.2% | ZVZO, 2026 |
| Cell sorting machines, all formats | RMB 6.377B (2025) → 7.494B (2026) → 17.75B (2032), 15.5% CAGR | YHResearch, Jul 2026 |
| Benchmark unit price and volume | USD 36,000 per set; 24,500 sets shipped in 2025; 36% gross margin | YHResearch, Jul 2026 |
| Pouch-cell sorting machines | USD 500.4M (2025) → 523.7M (2026) → 850.5M (2032), 7.87% CAGR | 360iResearch, 2026 |
| Automatic cylindrical sorters | USD 62.4M (2025) → 67.3M (2026) → 108.5M (2034), 6.2% CAGR | 24 Market Reports, 2026 |
| Where the money sits inside the category | Test equipment 52.4%; sorting and matching systems 31.8%; data software 9.3%, growing 21.5% | ZVZO, 2026 |
Two lines are worth reading twice. Inline 100% inspection is the fastest-moving segment at about 19.7% year on year, so buyers are moving the gate upstream into the line rather than auditing batches at the end; and the software slice, still 9.3% of revenue, grows at 21.5% — sorting is turning into a data problem.
The Five Parameters That Set Yield
A sorter is only as good as the parameters it may reject on. Voltage and resistance are the entry ticket; capacity, DCIR and the K-value separate grading from real defect screening.
| Parameter | Group target in published practice | How it is measured |
|---|---|---|
| Open-circuit voltage (OCV) | Spread ≤5 mV inside one group; ≤10 mV at 50% SOC | 0–5 V channel, class down to ±0.1 mV, after 2–4 h rest |
| AC internal resistance (1 kHz) | Spread ≤0.5 mΩ, or ≤10% of the group mean | Four-wire Kelvin probe at the terminals or can |
| DC internal resistance (DCIR) | ±1% to ±5% inside a group at a fixed SOC | Current pulse at 50% SOC, voltage step recorded |
| Capacity | ≤1% of nominal inside one group | Standard full charge and discharge at the grading step |
| K-value (self-discharge) | ≤0.04–0.05 mV/h absolute; ≤0.01 mV/h spread in a group | K = (V1 − V2) / Δt over a 24–48 h rest |
| Outlier screen | K above batch mean + 3σ is quarantined, not sorted | Statistical screen in the host software |
| Appearance and dimensions | 100% rejection of swelling, leakage and pole damage | Vision and laser gauge at the infeed |
The K-value pays for the machine. A cell with an internal micro-short passes capacity and passes a single OCV reading, then drifts apart in the field and drags the whole string down — visible only as a voltage drop over time. Two protocols are used: an accelerated screen at 40–50 °C for 36–48 h, or an ambient rest of 21–36 days. Both are temperature-sensitive to the point that a record taken in a rest room drifting more than about ±2 °C is not evidence of anything.
Bin Ladders, Throughput and the Arithmetic
More bins do not mean more output. Every extra bin divides the same test head and handling system, so throughput per grade falls and the footprint grows. What published machines actually deliver:
| Machine class | Throughput | Measurement class | Source |
|---|---|---|---|
| 9-channel cylindrical sorter, 18650–32700 | ≥5,000 cells/h (≥80 cells/min); 500-cell hopper | Voltage ±0.01% rdg; IR ±0.1% rdg (HIOKI-class meter) | LITH Machine spec, 2026 |
| 11-channel cylindrical sorter, 18650/21700 | 4,000 cells/h | Voltage ±0.1 mV; IR ±0.5% rdg | CowinLink spec, 2026 |
| 10-bin automatic cylindrical sorter, 32700 | Up to 4,500 cells/h; ≥99.5% sorting accuracy | 8 OK + 1 NG bins; under 1 kW; 250 kg | Semco Infratech spec, 2026 |
| 5-bin prismatic sorter (4 OK + 1 NG) | 450–700 cells/h; 99% yield | 0.001 V / 0.1 mΩ; cells 55–270 × 18–60 × 90–225 mm | Semco Infratech spec, 2026 |
| Configurable OCV/IR platform, 8/16/32 bins | Up to 4,000 cells/h | Voltage 0–5 V ±1 mV; IR 0.1–100 mΩ ±0.1 mΩ | Higgs Boson Systems spec, 2026 |
| Parallel test architecture on ESS pack lines | 2.8 s per cell takt | Voltage accuracy ±0.01% | China pack-line practice report, 2026 |
Read those numbers per cell, not per hour, and the case changes shape. A 4,000 cell/h head running 16 hours a day moves about 64,000 cells — roughly 0.3 GWh a year of 21700-class cells, which is why cylindrical sorters must be fast: the cells are small. The same head on prismatic cells runs at 450–700 cells/h, but a 314 Ah cell carries about 1 kWh, so one slow machine still matches a multi-gigawatt-hour prismatic line.
How the Gate Changes by Cell Format
The measurement principle is universal; handling, probe and failure modes are not. A cylindrical cell can be dropped down a chute, a prismatic cell must be clamped at its poles, and a pouch cell may be neither creased at the edge seal nor gripped at the tab.
| Attribute | Tesla 4680 cylindrical | EVE LF314 314 Ah prismatic | 50 Ah NCM power pouch |
|---|---|---|---|
| Handling | Hopper and belt feed; can dents are the handling risk | Servo clamp, copper probes on the poles | Vacuum pad or tray; no edge-seal crease |
| First measurement | OCV and AC IR at the can | OCV and AC IR at the terminals | OCV and AC IR at the tabs |
| Dominant reject | IR drift and feed damage | K-value outliers and pole damage | Seal wrinkles, tab misalignment, swelling |
| Practical bin ladder | 8–32 bins plus NG, on tight IR bands | 4–10 bins plus NG | 5–12 bins plus NG |
| Rest before sorting | ≥24 h at a held temperature | 24–48 h, K screened after grading | 24–48 h, dimension check first |
| Cell identity | Barcode or laser-marked code | Matrix code on the case | Code on the pouch or a tray map |
Buy the recipe library, not the fixture: a plant sorting cylindrical cells today and planning a prismatic line next year should be able to switch in software plus a probe swap.
Traceability Is the Second Product of the Sorter
Every cell should leave with a record — OCV, AC IR, capacity where available, K-value, assigned bin, timestamp, recipe version and operator. Bound to a barcode it becomes the evidence trail behind every pack the plant ships, the SPC input that flags drift before out-of-tolerance cells reach the welder, and increasingly a legal artefact under the EU battery passport. A machine that exports a CSV at the end of the shift is not doing this job; the interface has to write to MES or ERP in real time, with recipe permissions locked, because the sorting rule is a quality decision.
Acceptance Criteria to Write Into the RFQ
"Send your best price for a cell sorting machine" is not a specification — and the reason two quotes for the same line differ by a factor of three.
| Check | Write this into the RFQ |
|---|---|
| Cell envelope | Min/max length, width and height per format; whether prismatic and pouch share the machine |
| Measurement class | Voltage 0–5 V at ±0.1 mV; IR 0–200 mΩ at ±0.1 mΩ or ±0.5% rdg; name the reference meter |
| Bin ladder | Grades needed at your step, a dedicated NG channel, cells per bin |
| Throughput | Cells per hour per format at your takt, not the headline figure |
| K-value protocol | Rest time, temperature hold, accelerated or ambient method |
| Data and integration | Barcode binding, per-cell record, MES/ERP interface, recipe and permission control |
| Changeover | Minutes to switch format, what has to be exchanged, software-only or not |
| Acceptance test | Sorting accuracy, false-reject rate, and a trial run on your own cells |
The Takeaway: Sort on Data, Not on Price
A battery cell sorting machine is the cheapest insurance a pack plant can buy: the only station that sees 100% of incoming cells, and the only one that can reject a cell while rejection is free. Fix the measurement class first, then the bin ladder from your grading rules, then the throughput from your takt. A sorter bought on price alone becomes a bin-filling machine that passes defects downstream. NEGUP MACHINE supplies battery testing equipment including the cell sorting machine and cell grading machine, plus battery assembly equipment and PACK assembly lines configured to your cell formats and shift volume.
WhatsApp +86 15307690902 | rain@negupgroup.com
Frequently Asked Questions
What does a cell sorting machine measure?
OCV, AC internal resistance at 1 kHz, DCIR where required, capacity from the grading step and the K-value self-discharge rate from a rest period, with appearance and dimensions checked at the infeed.
How many bins do I need?
Five to twelve grades plus a dedicated NG channel covers most pack builders. Cylindrical lines on tight IR bands use 9 to 32 bins; prismatic and pouch cells normally need 4 to 10.
Why does the K-value matter more than capacity?
A cell with an internal micro-short passes capacity and a single OCV reading. Only the K-value, measured as a voltage drop over a 24–48 h rest at a held temperature, exposes it — talk to NEGUP MACHINE about a sorting protocol for your cells.